{ "row014": { "id": "row014", "row": 14, "doi": "10.3389/fnut.2023.1016809", "measure": "OR", "contrast": "serum 25(OH)D of 100 nmol/L or more vs under 50 nmol/L", "published": { "measure": "OR", "estimate": 1.91, "low": 1.59, "high": 2.3, "n": 21334, "events": null, "contrast": "serum 25(OH)D of 100 nmol/L or more vs under 50 nmol/L" }, "published_b": [0.647103242058538, 0.46373401623214, 0.832909122935104], "constants": null, "original": { "estimate": 1.89805079068682, "ci": [1.56694494033386, 2.29912150152466], "b": 0.640827459938658, "se": 0.0969561976141016, "df": 139, "p": 7.62479488270029e-10, "n": 21365, "events": 3536, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, "harmonized": { "estimate": 1.87165580859843, "ci": [1.5444828205579, 2.26813495056865], "b": 0.626823498222349, "se": 0.0971760836203569, "df": 139, "p": 1.71820121896052e-09, "n": 22826, "events": 3712, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.990301729751939, "published_z": 6.87098900053576, "interval_ratio_original": 0.962899872745971, "ratio_harmonized": 0.978147063614206, "variants": [ { "label": "BMI in the text's groups (<25, 25-30, 30+)", "estimate": 1.79938072516084, "ci": [1.48797471145096, 2.17595834738556], "b": 0.587442564128924, "se": 0.0961100362912798, "df": 139, "p": 9.34928394223408e-09, "n": 21365, "events": 3536, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "Model 1, crude (published 2.56, 2.16-3.04)", "estimate": 2.5810910874462, "ci": [2.16309803356841, 3.07985634414544], "b": 0.948212211646562, "se": 0.0893550137408583, "df": 139, "p": 1.22902349866235e-19, "n": 21365, "events": 3536, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "Model 2 (published 1.93, 1.61-2.32)", "estimate": 1.93364025230662, "ci": [1.59693188560238, 2.34134258264249], "b": 0.659404367477707, "se": 0.0967642656776179, "df": 139, "p": 2.63783334780269e-10, "n": 21365, "events": 3536, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "Model 3 (published 1.86, 1.55-2.25)", "estimate": 1.85030605336582, "ci": [1.52401238794347, 2.24645975203792], "b": 0.615351059659292, "se": 0.0981219196701476, "df": 139, "p": 4.23846545866973e-09, "n": 21365, "events": 3536, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "per 10 nmol/L (published 1.07, 1.05-1.09)", "estimate": 1.07329313969711, "ci": [1.05160739697033, 1.09542607539625], "b": 0.0707316226998683, "se": 0.0103236924951339, "df": 139, "p": 2.1764102379998e-10, "n": 21365, "events": 3536, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "50, 75, 100 closing their groups (>100 vs <=50)", "estimate": 1.88107330717448, "ci": [1.55502273710174, 2.27548877745628], "b": 0.631842522007378, "se": 0.0962747820769485, "df": 139, "p": 9.68080040813915e-10, "n": 21365, "events": 3536, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "2017-2018 alcohol as ever drinking (ALQ111)", "estimate": 1.85948172990555, "ci": [1.5378886727154, 2.24832418964855], "b": 0.620297809062091, "se": 0.0960395834409845, "df": 139, "p": 1.64657747024304e-09, "n": 21366, "events": 3537, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "any supplement (DSD010) for vitamin D supplements", "estimate": 1.8765628429167, "ci": [1.55065872062119, 2.27096269255489], "b": 0.629441828462878, "se": 0.0964819724609408, "df": 139, "p": 1.18156745205405e-09, "n": 21579, "events": 3572, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "without vitamin D supplements", "estimate": 2.04577902639933, "ci": [1.69511870736603, 2.46897860702548], "b": 0.715778658959346, "se": 0.0950981211278786, "df": 139, "p": 5.92831968867153e-12, "n": 21365, "events": 3536, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "unweighted", "estimate": 1.65390517897855, "ci": [1.4295224443094, 1.91350779551667], "b": 0.503139266650158, "se": 0.0743885266934382, "df": null, "p": 1.34541583168963e-11, "n": 21365, "events": 3536, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": "alcohol as at least 12 drinks in any one year (ALD100/ALQ101, asked through 2015-2016; 2017-2018 and 2021-2023 ask only about ever drinking and the past year)", "choices": [ { "choice": "BMI in the models", "decision": "continuous", "reason": "the covariate text groups BMI (<25, 25-30, 30+), as Table 1 does, but Table 3's numbers show BMI entered continuously: so entered, every adjusted estimate is reproduced (Model 2 1.25, 1.63, 1.93 against 1.23, 1.62, 1.93; Model 3 1.85 against 1.86; Model 4 1.26, 1.62, 1.90 against 1.25, 1.62, 1.91; per 10 nmol/L 1.07), while the groups give 1.82, 1.73, and 1.80 for the top group (a variant; both reproduce the headline)" }, { "choice": "osteoarthritis", "decision": "told of arthritis (MCQ160A) and its type osteoarthritis, with the type codes of each cycle (library arthritis()); other types are not osteoarthritis, an unknown type is missing", "reason": "the paper's definition; this gives its flow exactly (91,351 participants, 58,284 under 40, 3,702 without 25(OH)D, 3,641 without osteoarthritis information) and its 14.3% prevalence in 2001-2006, which reading MCQ190 with the 2011 codes would make 6.5%" }, { "choice": "25(OH)D", "decision": "LBDVIDMS (2001-2006) and LBXVIDMS (2007 on), groups <50, 50 to <75, 75 to <100, 100+ nmol/L", "reason": "the LC-MS/MS-equivalent values the paper describes; it prints <50 and >=100 and leaves 50 and 75 unstated, so all groups are closed below (a variant closes them above)" }, { "choice": "alcohol in 2017-2018", "decision": "12 or more drinks in the past year from ALQ121 and ALQ130", "reason": "the 12-drinks question was dropped; this matches Table 1's 72.1% (72.0%) and Table 3, and keeps the cycle (requiring the old question would drop it, leaving 18,858)" }, { "choice": "vitamin D supplements", "decision": "any reported product containing vitamin D in the Dietary Supplement Database, missing when a product has no information", "reason": "the paper's 'supplement containing vitamin D' (its 'more than one' read as one or more: Table 1 has 46.6% users, this 47.4%); NCHS recommends the database over nutrient totals; with products of unknown content missing, the analytic sample is 21,365 against the paper's 21,334 (counting them as non-users gives 21,579)" }, { "choice": "vitamin D supplements in 2021-2023", "decision": "kept, built the same way from DSQTOT_L and DSQIDS_L, under the dietary day-one weight (WTDRD1) as the plan's weight rule directs", "reason": "they can be built, and leaving them out moves the paper-cycle estimate from 1.90 to 2.05 (a variant); 2021-2023 asked them by telephone after the first dietary recall (6,754 participants have the data, by the codebook), so its complete-case sample is smaller, and NCHS directs WTDRD1 for any analysis using them, with examination data too; the sensitivity analysis uses the examination weight" }, { "choice": "recreational activity", "decision": "any moderate or vigorous recreational activity (PAD200/PAD320 in 2001-2006, 'unable' as no; PAQ650/PAQ665 from 2007)", "reason": "the paper's definition; gives Table 1's 54.8% active" }, { "choice": "covariate coding", "decision": "race: Mexican American, non-Hispanic White (reference), non-Hispanic Black, other (other Hispanic included); education under high school, high school or GED, more; PIR <1.3, 1.3-3.5, 3.5+; season of examination; smoking never, former, current; self-rated health excellent or very good, good, fair or poor", "reason": "the paper's definitions; Table 1's weighted shares are reproduced to within 0.4 points" }, { "choice": "weights", "decision": "examination weight over nine cycles (WTMEC2YR/9; the phlebotomy weight in 2021-2023)", "reason": "the paper weighted 'under the NHANES analytical guidelines' without naming the weight; 25(OH)D is measured at the examination" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The covariate text groups BMI (under 25, 25 to 30, 30 or more), but Table 3's adjusted estimates are reproduced with BMI continuous (Model 4's three upper groups 1.26, 1.62, and 1.90 here against 1.25, 1.62, and 1.91; Model 3's top group 1.85 against 1.86), while the text's groups give 1.82, 1.73, and 1.80 for the top group in Models 2, 3, and 4 (published 1.93, 1.86, 1.91)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The covariate text defines vitamin D supplement use as taking more than one supplement containing vitamin D in the past 30 days, but Table 1's 46.6% users is the weighted share taking one or more (47.4% here); two or more gives 12.5%." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The abstract gives the subgroup estimate for BMI under 25 as 1.01 (1.04 to 1.08), outside its own interval; the Results give 1.04 (1.00 to 1.08)." } ], "projected_power": 0.572426870167927 }, "row028": { "id": "row028", "row": 28, "doi": "10.31083/j.rcm2504130", "measure": "OR", "contrast": "highest SII tertile (597 or more) against the lowest (under 384); SII = platelets x neutrophils / lymphocytes, counts in 10^3 cells/uL", "published": { "measure": "OR", "estimate": 1.18, "low": 1.01, "high": 1.42, "n": 57600, "events": 2368, "contrast": "highest SII tertile (597 or more) against the lowest (under 384); SII = platelets x neutrophils / lymphocytes, counts in 10^3 cells/uL" }, "published_b": [0.165514438477573, 0.00995033085316809, 0.350656871613169], "constants": { "bmi": 27.95, "met": 720, "met_leisure": 28, "diabetes": 0, "cancer": 0 }, "original": { "estimate": 1.16399263944222, "ci": [0.960674536394484, 1.41034118564302], "b": 0.151856025786002, "se": 0.0972213264879133, "df": 163, "p": 0.120236347988451, "n": 52590, "events": 2063, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 320, "strata": 157 }, "harmonized": { "estimate": 1.16717647388575, "ci": [0.963193436405333, 1.4143586009852], "b": 0.154587561996702, "se": 0.09727853019311, "df": 163, "p": 0.113970364201238, "n": 52590, "events": 2063, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 320, "strata": 157 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.917479025895242, "published_z": 1.90429181437951, "interval_ratio_original": 0.887369271021432, "ratio_harmonized": 1.01798767086496, "variants": [ { "label": "unweighted", "estimate": 1.28670633891078, "ci": [1.14624020280547, 1.44438591365143], "b": 0.25208572768364, "se": 0.058979925761011, "df": null, "p": 1.91916341883138e-05, "n": 52590, "events": 2063, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude (published 1.46, 1.23-1.74)", "estimate": 1.3801067808291, "ci": [1.15090138854719, 1.65495910027079], "b": 0.322160873588028, "se": 0.0919746569448914, "df": 163, "p": 0.000594479696577829, "n": 52590, "events": 2063, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 320, "strata": 157 }, { "label": "Model 1 (published 1.32, 1.11-1.58)", "estimate": 1.33566942625639, "ci": [1.10422330147622, 1.615626851789], "b": 0.289432609060424, "se": 0.0963678938306048, "df": 163, "p": 0.00309099031557658, "n": 52590, "events": 2063, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 320, "strata": 157 }, { "label": "Model 2 (published 1.22, 1.02-1.47)", "estimate": 1.22166831288046, "ci": [1.00602664776587, 1.4835327374385], "b": 0.200217394184298, "se": 0.0983522802068843, "df": 163, "p": 0.0433980000199946, "n": 52590, "events": 2063, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 320, "strata": 157 }, { "label": "2017-2018 stacked with 2017-March 2020, as the paper did", "estimate": 1.18328730542293, "ci": [0.98753474028806, 1.41784262371025], "b": 0.168296417237017, "se": 0.0916396936209935, "df": 178, "p": 0.067951462901607, "n": 57606, "events": 2305, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 350, "strata": 172 }, { "label": "2017-2018 files in place of 2017-March 2020", "estimate": 1.1579139254101, "ci": [0.968485260114385, 1.38439345839944], "b": 0.146620045999651, "se": 0.090424649375462, "df": 153, "p": 0.106977795953495, "n": 49503, "events": 1885, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 301, "strata": 148 }, { "label": "Model 3 without cancer (the Methods' list)", "estimate": 1.1693487815771, "ci": [0.964856834977779, 1.41718079139414], "b": 0.156446996909436, "se": 0.0973463720920101, "df": 163, "p": 0.109964121948145, "n": 52590, "events": 2063, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 320, "strata": 157 }, { "label": "complete cases, no imputation", "estimate": 1.11483243397052, "ci": [0.916093988303085, 1.35668541841957], "b": 0.108704110183882, "se": 0.0994317706530946, "df": 163, "p": 0.275895669000518, "n": 51378, "events": 1909, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 320, "strata": 157 }, { "label": "per unit of log10 SII (published 1.30, 0.99-1.70)", "estimate": 1.28512640566472, "ci": [0.971368610354171, 1.70022982103009], "b": 0.250857083676252, "se": 0.141751674783118, "df": 163, "p": 0.0786478888528984, "n": 52590, "events": 2063, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 320, "strata": 157 } ], "left_out": "physical activity over work, transport, and leisure: 2021-2023 asks only about leisure time, so the harmonized version groups leisure-time MET-minutes (4 METs moderate, 8 vigorous) into the paper's four groups in every cycle", "choices": [ { "choice": "cycles", "decision": "1999-2000 to 2015-2016 and 2017-March 2020 (each participant once); variants stack 2017-2018 with 2017-March 2020 as the paper did, or use 2017-2018 alone", "reason": "the paper says only 1999 to 2020; its 66,568 participants over 18 and 59,732 with SII are exactly the counts over the ten 2-year files to 2017-2018 plus the pre-pandemic files, which hold the 2017-2018 participants again under new SEQNs. Stacked, our sample is 57,606 (published 57,600) and Model 3 gives 1.18" }, { "choice": "weights", "decision": "examination weight (WTMEC2YR, the 4-year WTMEC4YR for 1999-2002, WTMECPRP for 2017-March 2020; WTPH2YR in 2021-2023), pooled by years as NCHS directs, with strata and PSUs", "reason": "the paper says it used appropriate weights with the R survey package; the weighted crude OR implied by Table 1's weighted stroke percentages (1.46) is the published crude OR, and weighted fits come closest (unweighted gives 1.29)" }, { "choice": "age", "decision": "18 and older, as stated", "reason": "the flow chart's 66,568 counts those over 18; the stroke question is asked from age 20, so the analytic sample is 20 and older either way" }, { "choice": "stroke", "decision": "MCQ160F yes or no; refused and don't know are missing and excluded", "reason": "this excludes 2,062 for missing stroke (flow chart 2,059)" }, { "choice": "SII tertiles", "decision": "the published cutpoints: under 384, 384 to under 597, 597 or more", "reason": "stated; they are the unweighted tertiles of the stacked sample (384.0 and 596.7)" }, { "choice": "income to poverty ratio", "decision": "1 or less, over 1 to 3, over 3, and missing as unknown", "reason": "the Methods' groups; on the stacked sample they reproduce Table 1's counts (10,721, 22,022, 19,444, 5,419 against 10,721, 22,019, 19,443, 5,417), which Table 1 labels <1, 1 to <3, 3 or more" }, { "choice": "education", "decision": "under high school (DMDEDUC2 1-2), high school (3), college or higher (4-5); those with no answer (61) go under high school", "reason": "on the stacked sample this reproduces Table 1's counts (14,963, 13,379, 29,264 against 14,960, 13,379, 29,261); the median would put them in college or higher" }, { "choice": "missing covariates", "decision": "BMI, MET-minutes, diabetes, and cancer imputed with the study population's median (BMI 27.9, so the 25-29.9 group), income missing as its own group", "reason": "the paper imputes each covariate's median and keeps unknown income as a category; Table 1's BMI counts match" }, { "choice": "physical activity", "decision": "MET-minutes a week in the paper's four groups: 2007-March 2020 GPAQ work, transport, and leisure (8 METs vigorous, 4 moderate, 4 walking or cycling); 1999-2006 leisure activities from PAQIAF at NCHS's MET values plus walking or cycling to get places at 4 METs", "reason": "the paper gives the groups but not the items; this uses the GPAQ's domains that 1999-2006 asked in minutes. No construction tried matches Table 1's groups (published 27.7%, 20.6%, 11.4%, 40.2%; ours 29.7%, 15.0%, 11.2%, 44.1%); dropping activity moves the estimate from 1.16 to 1.19" }, { "choice": "diabetes", "decision": "told by a doctor, insulin or diabetes pills, fasting glucose 126 mg/dL or more, or HbA1c 6.5% or more", "reason": "stated; on the stacked sample Table 1 has 10,335 against our 9,939 (counting borderline gives 10,716); either changes the estimate by under 0.01" }, { "choice": "hyperlipidemia", "decision": "triglycerides 150 mg/dL or more, total cholesterol 200 or more, LDL 130 or more, HDL 40 or less (men) or 50 or less (women), or now taking cholesterol medicine (BPQ100D; BPQ101D in 2021-2023); not asked counts as not taking", "reason": "stated; the flow chart excludes only 2 for missing hyperlipidemia, which counting unasked as not taking reproduces (1), while treating it as missing would exclude 687. Triglycerides and LDL only from those who fasted, as NCHS directs: on the stacked sample this gives 41,012 with hyperlipidemia (Table 1: 40,996; 41,101 with every value in the files)" }, { "choice": "hypertension", "decision": "told by a doctor, taking medicine for it, or mean measured pressure 140/90 or more", "reason": "unstated; the flow chart's 24 missing fits a combined definition (13 missing) rather than self-report alone (205); the estimate changes by under 0.01 with self-report alone" }, { "choice": "cancer", "decision": "in Model 3 (MCQ220)", "reason": "Table 2's note lists it; the Methods' Model 3 sentence omits it (variant)" }, { "choice": "log base for the continuous SII", "decision": "log10 (variant only)", "reason": "unstated; per unit of log10 SII, Model 3 gives 1.29 (1.29 stacked, 1.30 published, 0.99-1.70), per unit of the natural log 1.12" } ], "departures": [ { "kind": "sample", "affects_headline": true, "followed": false, "detail": "The text describes adults in NHANES 1999 to 2020, but its 66,568 adults and 59,732 with SII are exactly the counts over the ten 2-year files to 2017-2018 plus the 2017-March 2020 files, which hold the 2017-2018 participants a second time under new SEQNs. Stacked that way, the sample here is 57,606 (published 57,600) and Model 3 gives 1.18 (0.99 to 1.42)." }, { "kind": "reporting", "affects_headline": true, "followed": true, "detail": "Model 3's interval for the highest tertile, 1.18 (1.01 to 1.42), is asymmetric on the log scale beyond rounding (0.16 below the estimate, 0.19 above), unlike Table 2's other intervals: its estimate and upper bound imply a lower bound of 0.97 to 0.99, and the stacked sample here gives 1.18 (0.99 to 1.42)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text imputes each covariate's missing values with its median, which for education is college or higher, but Table 1's education counts (14,960, 13,379, 29,261) put those with no answer under high school, as coding them so does in the stacked sample here (14,963, 13,379, 29,264)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The text and Table 1 give 2,368 strokes among the 57,600 analyzed, more than the 2,310 adults with SII who answered yes to the stroke question in the stacked files; the stacked sample here has 2,305." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1 labels the sexes the other way round: its 29,870 males (51.86%, repeated in the Results) are the stacked sample's 29,873 women." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's note calls its percentages weighted, but the Total column's are unweighted shares: its 4.11% with stroke is 2,368 of 57,600, above every tertile's weighted share (2.58% to 3.72%), and the stacked sample's weighted share is 2.94%." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's race and ethnicity rows count 2,986 participants in the Total column (1,707, 668, 456, and 155), not 57,600." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 3 gives 1,757 strokes among the 37,767 under 60 and 611 among the 19,833 aged 60 or more, where the stacked sample has 587 and 1,718, and 4,871 strokes among the 52,200 without cancer, more than the 2,368 in all (1,832 in the stacked sample)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The text's 66,568 adults 'aged 18 or above' are exactly those aged 19 or older in the stacked files; the analytic sample is 20 and older either way, as the stroke question is asked from age 20." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The abstract gives the sample as 53,600 people and the Discussion as 53,111, against the 57,600 of the Methods and Table 1." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The abstract says the risk of stroke 'decreased by 34%' for every unit of log SII and cites Model 3's OR of 1.30 (0.99 to 1.70), which is an increase; the 34% is Model 2's (1.34)." } ], "projected_power": 0.0850933548260145 }, "row034": { "id": "row034", "row": 34, "doi": "10.1186/s12889-023-17556-w", "measure": "OR", "contrast": "DII quartile 4 (2.95 or more) vs quartile 1 (under 0.23)", "published": { "measure": "OR", "estimate": 1.87, "low": 1.53, "high": 2.29, "n": 44019, "events": 1486, "contrast": "DII quartile 4 (2.95 or more) vs quartile 1 (under 0.23)" }, "published_b": [0.625938430866495, 0.425267735404344, 0.828551817566148], "constants": { "sample_cuts": [0.0468253212718225, 1.49075093151774, 2.66582595264511] }, "original": { "estimate": 1.78189857281343, "ci": [1.4317875698437, 2.21762123842236], "b": 0.577679409815632, "se": 0.110729282129546, "df": 153, "p": 5.82420961058141e-07, "n": 41097, "events": 1288, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 301, "strata": 148 }, "harmonized": { "estimate": 1.78515164600176, "ci": [1.43408143875881, 2.22216557100197], "b": 0.57950336736635, "se": 0.110842229694153, "df": 153, "p": 5.53433392339483e-07, "n": 41097, "events": 1288, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 301, "strata": 148 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.922901329154598, "published_z": 6.08413193231676, "interval_ratio_original": 0.921768180266668, "ratio_harmonized": 1.00315738715926, "variants": [ { "label": "unweighted (published 1.89, 1.60-2.23)", "estimate": 1.91956353760562, "ci": [1.61029741842003, 2.28822584744645], "b": 0.652097836033895, "se": 0.0896337603885553, "df": null, "p": 3.46070508452471e-13, "n": 41097, "events": 1288, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude (published 2.47, 2.08-2.94)", "estimate": 2.46847816285923, "ci": [2.03630976104868, 2.99236616995582], "b": 0.903601832357713, "se": 0.097420301067047, "df": 153, "p": 1.59614580183298e-16, "n": 41097, "events": 1288, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 301, "strata": 148 }, { "label": "Model I (published 2.43, 2.02-2.93)", "estimate": 2.38979850445713, "ci": [1.93596578062999, 2.95001954530781], "b": 0.871209054630799, "se": 0.106602436561597, "df": 153, "p": 1.06326890157077e-13, "n": 41097, "events": 1288, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 301, "strata": 148 }, { "label": "crude, unweighted (published 2.37, 2.04-2.77)", "estimate": 2.46417793742938, "ci": [2.08578658457105, 2.91121486360623], "b": 0.901858257929726, "se": 0.0850588155451298, "df": null, "p": 2.89311807029614e-26, "n": 41097, "events": 1288, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model I, unweighted (published 2.23, 1.91-2.60)", "estimate": 2.36438776369146, "ci": [1.99032579639771, 2.8087509628885], "b": 0.860519114726914, "se": 0.0878693557464987, "df": null, "p": 1.20464606767943e-22, "n": 41097, "events": 1288, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "MEC exam weight", "estimate": 1.84501374518717, "ci": [1.5025782746024, 2.26548977678408], "b": 0.612486727430685, "se": 0.103920455349408, "df": 153, "p": 2.32290098196651e-08, "n": 41097, "events": 1288, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 301, "strata": 148 }, { "label": "ever smoker and past-year drinker, as the text says", "estimate": 1.80624250786742, "ci": [1.45483260554161, 2.24253428525036], "b": 0.59124872497653, "se": 0.109515559509475, "df": 153, "p": 2.51089143538252e-07, "n": 41036, "events": 1287, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 301, "strata": 148 }, { "label": "DII with total folate, as the parameter list says", "estimate": 1.79469328957457, "ci": [1.43778175771663, 2.24020369319418], "b": 0.58483413802263, "se": 0.112236153663837, "df": 153, "p": 5.99454267549519e-07, "n": 41097, "events": 1288, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 301, "strata": 148 }, { "label": "missing covariates as their own level", "estimate": 1.74158083753953, "ci": [1.4096096896537, 2.15173309033512], "b": 0.55479322808686, "se": 0.107046676762658, "df": 153, "p": 6.81331703369488e-07, "n": 43477, "events": 1347, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 301, "strata": 148 }, { "label": "sample quartiles of this DII", "estimate": 1.67692011504178, "ci": [1.33859301811955, 2.10075880732004], "b": 0.51695884607246, "se": 0.114061993773723, "df": 153, "p": 1.16948218061583e-05, "n": 41097, "events": 1288, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 301, "strata": 148 } ], "left_out": "the two-hour OGTT glucose part of the diabetes definition: 2021-2023 has no OGTT (NHANES ran it in 2005-2016 only)", "choices": [ { "choice": "weights", "decision": "day-one dietary weight (WTDRD1; the four-year WTDR4YR for 1999-2002), pooled over the ten cycles as NCHS directs", "reason": "the paper says only that it used 'the sample weights corresponding to different research periods'; NCHS directs the day-one dietary weight for day-one intakes; the examination weight is a variant" }, { "choice": "dietary recall", "decision": "day one (the MEC recall), reliable recalls only (status 1; 1999-2000 names it DRDDRSTS)", "reason": "the Methods: intake 'recorded in the mobile examination center'" }, { "choice": "DII parameters", "decision": "the 26 the paper lists, with folic acid as DR1TFA (dii() as the library builds it)", "reason": "the parameter list names total folate, but the weighted medians the paper reports (1.39 overall, 1.99 with stroke, 1.37 without) fit folic acid (1.36, 1.99, 1.34) and not total folate (1.19, 1.81, 1.16); total folate is a variant" }, { "choice": "nutrients a cycle lacks", "decision": "left out of those participants' DII: vitamin D in 1999-2006; beta-carotene, vitamin A as RAE, alpha-tocopherol, and folic acid in 1999-2000", "reason": "NHANES files lack them; the paper kept those cycles and does not say how it scored them; dii() leaves a missing parameter out" }, { "choice": "quartile cutpoints", "decision": "the published ones: Q1 under 0.23, Q4 2.95 or more", "reason": "Statistical analysis; this DII's own quartiles are lower (0.05, 1.49, 2.67), so the published cutpoints put 28%, 28%, 25%, and 19% of the sample in Q1 to Q4, yet they reproduce the crude and Model I estimates (2.47 and 2.39 against 2.47 and 2.43), while the sample's own quartiles give lower ones (variant)" }, { "choice": "population", "decision": "ages 18-79, not pregnant (RIDEXPRG = 1), a reliable day-one recall, and a yes or no to the stroke question (MCQ160F, asked from age 20)", "reason": "the paper's exclusions; 44,013 here against 44,019 published, whose age step prints 46,469 where its other counts imply 46,369 (the count here); 1,421 strokes against 1,486" }, { "choice": "smoking", "decision": "current smoker (100 cigarettes in life and smoking every day or some days) against not", "reason": "the text calls everyone who smoked 100 cigarettes a smoker, but Table 1's smoking shares (22.3% overall, 29.6% with stroke, 22.1% without) are those of current smoking (22.6%, 30.3%, 22.4% weighted here), not of ever smoking (46.8%, 59.9%, 46.5%); the text's version is a variant" }, { "choice": "drinking", "decision": "12 or more drinks in any one year or in life (ALQ100, ALD100, or ALQ101, then ALQ110); from 2017, which asks neither, ever having had a drink (ALQ111)", "reason": "the text says 12 drinks in the year before the survey, but Table 1's drinker shares (89.3% without stroke, 84.3% with) are matched by this (89.4%, 83.3%) and not by 12 drinks in the past year (59.5%, 33.2%); the text's version is a variant" }, { "choice": "education", "decision": "under 9th grade, 9th grade to high school graduate or GED, more than high school", "reason": "Table 1's shares (5.4%, 35.2%, 59.3%) are those of these groups (5.3%, 35.3%, 59.4%), not of under high school (16.5%)" }, { "choice": "race", "decision": "RIDRETH1's five groups, non-Hispanic White as reference", "reason": "Covariates; the reference is unstated" }, { "choice": "age and BMI", "decision": "continuous", "reason": "unstated; Table 1 reports age as a mean" }, { "choice": "diabetes", "decision": "told by a doctor, HbA1c 6.5% or more, fasting glucose 126 mg/dL or more, or two-hour OGTT glucose 200 mg/dL or more (2005-2016 only); a test a person lacks counts as not met", "reason": "the paper's definition; 12.0% weighted here against Table 1's 12.3%" }, { "choice": "hypertension", "decision": "mean of the first three readings 140/90 mmHg or more, told (BPQ020), or taking prescribed medicine (BPQ050A, BPQ150 in 2021-2023)", "reason": "the paper's definition; 2021-2023 measured blood pressure with an oscillometric device" }, { "choice": "missing covariates", "decision": "complete case (41,097 of 44,013; drinking is missing for 2,423, BMI for 536)", "reason": "unstated; keeping missing values as their own level changes the estimate little (variant)" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text counts as smokers everyone who smoked 100 cigarettes, quitters included, but Table 1's smoking shares (22.3% overall, 29.6% with stroke, 22.1% without) are those of current smoking (22.6%, 30.3%, 22.4% weighted here), not of ever smoking (46.8%, 59.9%, 46.5%)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text counts as drinkers those with at least 12 drinks in the year before the survey, but Table 1's drinking shares (89.31% without stroke, 84.3% with) are matched by 12 or more drinks in any one year or in life (89.4% and 83.3% here), not by 12 or more in the past year (59.5% and 33.2%)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text and Table 1 group education as below high school, high school, and above high school, but Table 1's shares (5.4%, 35.2%, 59.3%) are those of under 9th grade, 9th grade to high school graduate, and more (5.3%, 35.3%, 59.4% here); below high school is 16.5%." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The DII's parameter list names total folate, but the DII values in Table 2 and the Results (1.39 overall, 1.99 with stroke, 1.37 without) fit a DII with folic acid (1.36, 1.99, 1.34 here) and not one with total folate (1.19, 1.81, 1.16)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The exclusion counts (46,469 aged under 18 or 80 and over, 1,516 pregnant, 5,747 without dietary data, 3,665 without stroke status) leave 43,919 of the 101,316, not the 44,019 analyzed; 46,369 are under 18 or 80 and over in the data, which leaves 44,019." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's laboratory rows don't fit their labels: the WBC, neutrophil, lymphocyte, platelet, and hemoglobin rows repeat the age, male sex, non-Hispanic White, non-Hispanic Black, and Mexican American rows, the TC row repeats the monocyte row, and the FBG, HbA1c, eGFR, TG, HDL-C, and RBC values don't fit their labels and units." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's note calls its categorical shares unweighted, but they are weighted (68.3% non-Hispanic White; its smoking shares match weighted ones here), and its overall drinking share (83.4%) lies outside both groups' (89.31% and 84.3%)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 2 gives a mean vitamin E intake of 66.6 mg (66.8 without stroke, 56.0 with); the same recalls give 8.4 mg here." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 3 gives Model I's estimate for quartile 2 as 1.26 (1.01 to 1.58) with P 0.002, which its interval rules out (a lower bound of 1.01 puts P near 0.04)." } ], "projected_power": 0.43680390681585 }, "row040": { "id": "row040", "row": 40, "doi": "10.1186/s12944-024-02262-2", "measure": "OR", "contrast": "per unit of ln NHHR ((total cholesterol - HDL cholesterol) / HDL cholesterol)", "published": { "measure": "OR", "estimate": 1.77, "low": 1.11, "high": 2.83, "n": 3772, "events": null, "contrast": "per unit of ln NHHR ((total cholesterol - HDL cholesterol) / HDL cholesterol)" }, "published_b": [0.570979546585738, 0.104360015324243, 1.04027671165515], "constants": null, "original": { "estimate": 1.49079912673828, "ci": [0.25605115329355, 8.6798360706294], "b": 0.399312302855608, "se": 0.898838118282842, "df": null, "p": 0.656859056485693, "n": 3040, "events": 214, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 1.50073202509554, "ci": [0.258565888804907, 8.71033925455921], "b": 0.40595300579709, "se": 0.897239820826743, "df": null, "p": 0.650947363265534, "n": 3040, "events": 214, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.699346071577097, "published_z": 2.39145075967613, "interval_ratio_original": 0.265630308347821, "ratio_harmonized": 1.01663034896243, "variants": [ { "label": "weighted (WTMECPRP)", "estimate": 5.94660159921424, "ci": [0.524339966180368, 67.4411123710015], "b": 1.78281989658772, "se": 1.17911602477911, "df": 25, "p": 0.143073713733442, "n": 3040, "events": 214, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, { "label": "crude, unweighted (published 1.42, 1.00-2.00)", "estimate": 1.43601797632077, "ci": [1.06636091886377, 1.93381770827991], "b": 0.361873988876906, "se": 0.151850824678841, "df": null, "p": 0.017168055752498, "n": 3040, "events": 214, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, weighted", "estimate": 1.50396587104833, "ci": [0.895014826206812, 2.52723561112885], "b": 0.408105533148055, "se": 0.252008163529878, "df": 25, "p": 0.117906089848221, "n": 3040, "events": 214, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, { "label": "Model 2, unweighted (published 1.80, 1.22-2.66)", "estimate": 1.89106734625954, "ci": [1.35450482926252, 2.64017937096332], "b": 0.637141403140273, "se": 0.170261013921089, "df": null, "p": 0.000182456175374675, "n": 3040, "events": 214, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2, weighted", "estimate": 2.21847754738361, "ci": [1.2552388957361, 3.92088123222077], "b": 0.796821171156184, "se": 0.276515950992101, "df": 25, "p": 0.00800909703226665, "n": 3040, "events": 214, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, { "label": "Model 3 without total and HDL cholesterol", "estimate": 1.08898450956663, "ci": [0.746101147856871, 1.58944570113912], "b": 0.08524561939293, "se": 0.192931973957479, "df": null, "p": 0.658602905406775, "n": 3040, "events": 214, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 3 without BMI, total and HDL cholesterol", "estimate": 1.65944551311394, "ci": [1.17578415513979, 2.34206201789356], "b": 0.506483518323975, "se": 0.175793142097967, "df": null, "p": 0.00396247713509797, "n": 3040, "events": 214, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 3, BMI and PIR continuous, race as one 1-5 code", "estimate": 2.02052240451342, "ci": [0.356226838490057, 11.4604244993029], "b": 0.703356094073486, "se": 0.885497729570959, "df": null, "p": 0.427017354982489, "n": 3040, "events": 214, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 3, never-drinkers missing for alcohol", "estimate": 1.47839329838721, "ci": [0.246980919845199, 8.84945584495395], "b": 0.390955888877696, "se": 0.912976001559443, "df": null, "p": 0.668489915615351, "n": 2813, "events": 200, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 3, NHHR Q4 vs Q1 (published 1.86, 1.04-3.32)", "estimate": 0.747441478272809, "ci": [0.277945852847437, 2.0099913624158], "b": -0.291099266718154, "se": 0.504718300579629, "df": null, "p": 0.564104352237422, "n": 3040, "events": 214, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": "physical activity at work (PAQ620, whether work involves moderate-intensity activity): 2021-2023 asks only about leisure-time activity", "choices": [ { "choice": "weights", "decision": "unweighted", "reason": "the paper does not say the regressions were weighted; Table 2's crude quartile ORs equal the unweighted ORs from Table 1's counts, and its crude and Model 2 intervals are as wide as unweighted fits on 2,117 people with 157 cases give (our unweighted standard errors on all 3,772, scaled to that size: 0.177 and 0.199; published 0.177 and 0.199)" }, { "choice": "analytic sample", "decision": "Fig. 1's 3,772 (aged 50 or younger, total and HDL cholesterol, a yes or no to MCQ550), complete cases for Model 3's covariates", "reason": "Fig. 1 is reproduced exactly (15,560, 10,719, 6,683, 3,772); the paper says it multiply imputed missing covariates, but its models were fitted on Table 1's 2,117 participants, a subset the paper does not define (not the fasting subsample, people who fasted 8 or 9 hours, complete cases, or those under 50), so complete cases of the stated population are used" }, { "choice": "total and HDL cholesterol in Model 3", "decision": "continuous, mmol/L, as Table 2's note and the covariate text say", "reason": "stated; with both in the model ln NHHR is close to a function of them and its interval is about four times as wide as the published one, so the published Model 3 cannot have held both as continuous terms (variant without them)" }, { "choice": "BMI", "decision": "Fig. 2's categories: 25 or less, 25 to 30, 30 or more", "reason": "Fig. 2 defines the covariate by these categories; Table 1 uses them" }, { "choice": "income to poverty ratio", "decision": "Table 1's categories: 1.3 or less, 1.3 to 3.5, 3.5 or more", "reason": "Table 1" }, { "choice": "race and education", "decision": "RIDRETH1's five groups; education as less than high school (DMDEDUC2 1-2), high school (3), college or above (4-5)", "reason": "Table 1's categories; its 61% 'college graduate or above' matches DMDEDUC2 4 and 5 together (61.1%)" }, { "choice": "smoking, alcohol, physical activity", "decision": "SMQ020 yes; ALQ151 yes, with never-drinkers (ALQ111 no) as no; PAQ620 yes", "reason": "Fig. 2's definitions; never-drinkers have not had 4/5 drinks every day" }, { "choice": "hypertension and diabetes", "decision": "self-reported diagnosis: BPQ020 yes; DIQ010 yes, borderline as no", "reason": "unstated; Table 1's shares (19.0% and 5.6%) match BPQ020 (18.9%) and DIQ010 yes (5.6%) in the population, not borderline counted as diabetes (7.4%)" }, { "choice": "CHD, heart attack, COPD", "decision": "MCQ160C, MCQ160E, MCQ160P yes", "reason": "unstated; the 2017-March 2020 questions for these conditions" }, { "choice": "sedentary time", "decision": "PAD680 minutes, continuous", "reason": "the covariate text says minutes" } ], "departures": [ { "kind": "sample", "affects_headline": true, "followed": false, "detail": "The text analyzes 3,772 participants with missing covariates multiply imputed, but Table 1 covers 2,117, Table 2's crude quartile estimates are those of Table 1's counts (quartile 4: 1.51, 0.95 to 2.41), and its crude and Model 2 intervals for ln NHHR are as wide as unweighted fits on 2,117 people with 157 cases give (standard errors 0.177 and 0.199 here, scaled to that size, against the published 0.177 and 0.199)." }, { "kind": "model", "affects_headline": true, "followed": false, "detail": "Table 2's note puts BMI and continuous total and HDL cholesterol in Model 3, but its interval (1.11 to 2.83) is about a quarter as wide on the log scale as a model with both cholesterol terms gives (0.26 to 8.68). Its estimate is also close to Model 2's (1.77 against 1.80), while adjusting for BMI lowers it here (1.09 with BMI, 1.66 without, both without the cholesterol terms) and the paper's own BMI subgroups give 0.94, 1.35, and 1.37." }, { "kind": "coding", "affects_headline": true, "followed": false, "detail": "Fig. 2 defines alcohol consumption as ever having 4/5 or more drinks every day (ALQ151), but Table 1's 18.7% drinkers (396 of 2,117) matches ALQ151 yes together with those who never drank (704 of the 3,772 here, 18.7%), not ALQ151 yes alone (422, 11.2%)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's 'Male' row (70.89%, 61.81%, 50.85%, 32.26% by NHHR quartile) gives the share of women (70.70%, 60.31%, 50.79%, 30.93% of the study population here)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's PIR rows hold more people than quartiles 2 and 3 have (629 and 644 of 529, their shares summing to 118.85% and 121.65%)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's note and the Methods give continuous variables as mean and standard error, but the values are standard deviations (age 32.67 +- 8.96 in a quartile of 529)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "In Fig. 4 the small COPD and total cholesterol groups carry the narrow intervals and the large ones the wide intervals, the reverse of what their sizes give: COPD yes 1.58 (1.04 to 2.41) and no 1.72 (0.42 to 7.16), with 157 and 3,610 people here; total cholesterol 6.2 mmol/L or more 1.74 (1.11 to 2.73) and under 6.2 1.43 (0.49 to 4.18), with 287 and 3,485." } ], "projected_power": 0.424579292106431 }, "row050": { "id": "row050", "row": 50, "doi": "10.3389/fpsyt.2024.1433990", "measure": "OR", "contrast": "per 10 mmHg of diastolic blood pressure", "published": { "measure": "OR", "estimate": 1.06, "low": 1, "high": 1.12, "n": 26581, "events": 2261, "contrast": "per 10 mmHg of diastolic blood pressure" }, "published_b": [0.0582689081239758, 0, 0.113328685307003], "constants": { "kcal_median": 1951, "kcal2_median": 1920.5, "met_median": 2280, "met_known_median": 1200, "leisure_median": 960 }, "original": { "estimate": 1.05871419382572, "ci": [0.999812169457931, 1.12108631845894], "b": 0.0570551471213013, "se": 0.0288819042704613, "df": 109, "p": 0.0507433236235202, "n": 28029, "events": 2381, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, "harmonized": { "estimate": 1.06539923022319, "ci": [1.00863346256831, 1.12535976832445], "b": 0.0633495929767407, "se": 0.02762567292659, "df": 109, "p": 0.0237583979665313, "n": 27997, "events": 2375, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.979169662831299, "published_z": 2.01546432894882, "interval_ratio_original": 0.989891293264033, "ratio_harmonized": 1.11032213872058, "variants": [ { "label": "unweighted (published 1.05, 1.01-1.09)", "estimate": 1.04692932548963, "ci": [1.00780829499374, 1.08756895335632], "b": 0.0458614276899162, "se": 0.0194306936077768, "df": null, "p": 0.0182622858075624, "n": 28029, "events": 2381, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude (published 1.05, 1.00-1.10)", "estimate": 1.04919672573926, "ci": [0.999384293263198, 1.10149196532556], "b": 0.0480248482844552, "se": 0.0245416460655252, "df": 109, "p": 0.0529200842177016, "n": 28029, "events": 2381, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "Model 1 (published 1.07, 1.01-1.13)", "estimate": 1.07484204597234, "ci": [1.02151090722851, 1.13095750188751], "b": 0.0721737168045511, "se": 0.0256769620040222, "df": 109, "p": 0.00585994010981378, "n": 28029, "events": 2381, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "Model 2 (published 1.08, 1.02-1.14)", "estimate": 1.08466112775234, "ci": [1.02928256367451, 1.14301922871106], "b": 0.0812676135567026, "se": 0.0264411976508927, "df": 109, "p": 0.00267299394277719, "n": 28029, "events": 2381, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "Model 3 (published 1.07, 1.01-1.13)", "estimate": 1.0678341387333, "ci": [1.01142315141977, 1.12739138534021], "b": 0.0656324276660663, "se": 0.0273839142553711, "df": 109, "p": 0.0182430946088583, "n": 28029, "events": 2381, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "age continuous", "estimate": 1.09875790627361, "ci": [1.04045950688862, 1.16032284640176], "b": 0.0941803656897575, "se": 0.0275069681272513, "df": 109, "p": 0.000870838364462336, "n": 28029, "events": 2381, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "age in three groups (<40, 40-59, 60+)", "estimate": 1.07159857464725, "ci": [1.01319380600331, 1.13337004073854], "b": 0.0691515285710147, "se": 0.0282769536007299, "df": 109, "p": 0.016065576299832, "n": 28029, "events": 2381, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "BMI continuous", "estimate": 1.05114854191491, "ci": [0.992677870380435, 1.11306325056525], "b": 0.0498834157948076, "se": 0.028876600690914, "df": 109, "p": 0.0869159776385551, "n": 28029, "events": 2381, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "activity unknown only if missing (no activity is low)", "estimate": 1.05755717076459, "ci": [0.99885835001183, 1.11970548118494], "b": 0.0559616926657222, "se": 0.0288117717007826, "df": 109, "p": 0.0546784580690622, "n": 28029, "events": 2381, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "energy intake as the two-day mean", "estimate": 1.06191157439416, "ci": [1.00015355229621, 1.12748306421884], "b": 0.0600706560708363, "se": 0.0302311293564851, "df": 109, "p": 0.0494260030689888, "n": 24923, "events": 2115, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "CKD only where eGFR and ACR are both known", "estimate": 1.05705389784898, "ci": [0.997453000866479, 1.12021613247649], "b": 0.0554856969305456, "se": 0.0292819864528922, "df": 109, "p": 0.0607585763846012, "n": 27494, "events": 2323, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "pregnant women left out", "estimate": 1.05553130353922, "ci": [0.996736423606815, 1.11779433997158], "b": 0.0540442454119126, "se": 0.028917305244648, "df": 109, "p": 0.0643179502371607, "n": 27512, "events": 2346, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 } ], "left_out": ["antidepressant use (Multum category 249 in the prescription file): 2021-2023 releases no drug names, and no question asks about antidepressants", "antihypertensive drug use from the prescription file (Multum blood-pressure-lowering classes): 2021-2023 releases no drug names; replaced by self-reported current treatment for high blood pressure (BPQ050A; BPQ150 in 2021-2023)", "physical activity over work, transport, and leisure (GPAQ MET-minutes): 2021-2023 asks only about leisure time; replaced by leisure-time moderate (4 METs) and vigorous (8 METs) activity in the paper's groups, split at the paper cycles' median"], "choices": [ { "choice": "weights", "decision": "examination weight WTMEC2YR over 7 (NCHS's pooled weight for seven 2-year cycles)", "reason": "the paper names 1/7 * WTMEC2YR; weighted fits reproduce its crude to Model 4 sequence, and the unweighted fit its sensitivity analysis 1" }, { "choice": "blood pressure", "decision": "auscultatory readings (BPX, BPX_J in 2017-2018) averaged by the protocol the paper quotes: the only reading, or the mean of those after the first; a diastolic reading of 0 is no reading", "reason": "the paper quotes the protocol and names mercury sphygmomanometers; with all-zero diastolic readings as missing, 948 with a PHQ-9 lack a pressure against the paper's 954 (825 if zeros counted)" }, { "choice": "depression", "decision": "PHQ-9 total of 10 or more, missing if any item is missing, refused, or don't know", "reason": "the paper's cutoff; the convention for incomplete answers (36,259 with a PHQ-9 against the paper's 36,397)" }, { "choice": "study population", "decision": "everyone with a PHQ-9 and a diastolic pressure and complete covariates; no age bound, which leaves adults 20 and older because education and the medical conditions are asked from 20", "reason": "the paper's flow has no age or pregnancy step" }, { "choice": "age", "decision": "two groups, under 60 and 60 or older", "reason": "Table 1's groups; the Methods list 'age (<40 years, >=60 years)', a typo for these groups or a list of three; two groups reproduce Table 2's sequence (1.049, 1.075, 1.085, 1.068, 1.059 against 1.05, 1.07, 1.08, 1.07, 1.06), three groups less closely, and continuous age puts Models 1-4 0.02-0.04 too high (variants)" }, { "choice": "BMI", "decision": "two groups, under 25 and 25 or more", "reason": "Table 1's groups; the Methods give none, but with BMI continuous Model 2 (1.065) falls below Model 1 (1.075) where the paper's rises (1.07 to 1.08), and Model 4 is 1.051 (variant)" }, { "choice": "education", "decision": "less than college (DMDEDUC2 1-3) or college or higher (4-5)", "reason": "the paper's groups; 62.3% college or higher (weighted) against Table 1's 62.56%" }, { "choice": "income to poverty ratio", "decision": "under 1.3, 1.3 to under 3.5, 3.5 or more", "reason": "Table 1" }, { "choice": "race", "decision": "RIDRETH1's five groups", "reason": "the paper's five groups" }, { "choice": "drinking", "decision": "never (fewer than 12 drinks in life; never a drink from 2017), former (none in the past 12 months), current", "reason": "unstated; the library's definition, built in this file because alcohol() stops on 2005-2010; Table 1 has fewer former drinkers (13.1% against our 15.3%)" }, { "choice": "smoking", "decision": "never (fewer than 100 cigarettes), former, current (SMQ020, SMQ040)", "reason": "unstated; convention" }, { "choice": "energy intake", "decision": "first-day recall split at the analytic sample's unweighted median, high at the median or above", "reason": "unstated; Table 1's groups are near equal in size (13,284 and 13,297), so the median is the sample's; the two-day mean would drop another 3,106 people (variant)" }, { "choice": "physical activity", "decision": "GPAQ MET-minutes a week (8 METs vigorous work and leisure, 4 moderate and walking or cycling), low or high at the median of those who reported any, unknown for those who reported none or were not asked (2005-2006 had no GPAQ)", "reason": "unstated; Table 1 keeps an 'Unknown' group that is far more common among the depressed (32.4% against 19.1%), as no activity is and missingness is not; counting no activity as low changes the estimate little (variant)" }, { "choice": "medical conditions", "decision": "ever told by a doctor: arthritis (MCQ160A), thyroid problem (MCQ160M), cancer (MCQ220), liver condition (MCQ160L), CHD (MCQ160C), CHF (MCQ160B), stroke (MCQ160F)", "reason": "unstated; ever-told prevalences match Table 1 (thyroid 11.0% against 10.8%, liver 3.5% against 3.4%)" }, { "choice": "diabetes", "decision": "told by a doctor, insulin or pills, HbA1c 6.5% or more, or fasting glucose 126 mg/dL or more", "reason": "unstated; common definition (12.9% against Table 1's 13.6%)" }, { "choice": "CKD", "decision": "eGFR (CKD-EPI 2009) under 60 or urine ACR 30 mg/g or more, from whichever was measured", "reason": "the paper's A2 or G3a and above; requiring both measures drops 535 people and changes the estimate little (variant)" }, { "choice": "hyperlipidemia", "decision": "NCEP ATP III thresholds or cholesterol medication", "reason": "unstated; 69.8% against Table 1's 70.3%" }, { "choice": "antihypertensive and antidepressant drugs", "decision": "prescription medicines taken in the past 30 days (RXQ_RX) in Multum's blood-pressure-lowering classes (42, 43, 44, 47, 48, 49, 53, 55, 56, 340, 342) and antidepressants (249)", "reason": "the Results speak of use in the past month, the prescription file's window; 27.9% and 13.1% against Table 1's 26.8% and 12.9%" }, { "choice": "missing covariates", "decision": "complete case", "reason": "the paper's flow excludes those missing covariates" }, { "choice": "harmonized physical activity", "decision": "leisure-time moderate (4 METs) and vigorous (8 METs) activity in the paper's groups: unknown if none or missing, low or high at the paper cycles' median of those with any (960 MET-minutes a week)", "reason": "2021-2023 asks only about leisure time; the plan's rule for activity" }, { "choice": "harmonized antihypertensive drug", "decision": "self-reported current treatment for high blood pressure (BPQ050A; BPQ150 in 2021-2023); untreated if never told of high blood pressure or not told to take medicine", "reason": "the closest 2021-2023 measure of the same covariate, since its prescription file has no drug names" }, { "choice": "2021-2023 weight", "decision": "phlebotomy weight (WTPH2YR, from BIOPRO_L)", "reason": "CKD (serum creatinine), diabetes (HbA1c, glucose), and hyperlipidemia (lipids) use blood analytes, for which NCHS directs the phlebotomy weight" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The covariate text gives age as '<40 years, >=60 years', but Table 1 groups it as under 60 and 60 or more, and those two groups reproduce Table 2's estimates from the crude model to Model 4 (1.049, 1.075, 1.085, 1.068, 1.059 against 1.05, 1.07, 1.08, 1.07, 1.06), where three groups (under 40, 40 to 59, 60 or more) give 1.072 for Model 4." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text splits physical activity into high and low at the median MET and excludes participants missing covariates, but Table 1 has a third activity group, 'Unknown' (6,534 of the 26,581 analyzed; 32.38% of the depressed against 19.12% of the others)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 2's row for diastolic pressure of 90 or more gives Model 2 the same estimate, interval, and P as Model 1 (1.34, 1.02 to 1.76, 0.04), while every other row changes between the two models, as this row does in Table 3's unweighted and imputed analyses (1.28 to 1.25, 1.46 to 1.40)." } ], "projected_power": 0.110139114610244 }, "row058": { "id": "row058", "row": 58, "doi": "10.3389/fpsyt.2023.1274648", "measure": "OR", "contrast": "highest vs lowest quartile of non-HDL cholesterol (total minus HDL cholesterol, mmol/L)", "published": { "measure": "OR", "estimate": 1.22, "low": 1.03, "high": 1.45, "n": 28041, "events": 2419, "contrast": "highest vs lowest quartile of non-HDL cholesterol (total minus HDL cholesterol, mmol/L)" }, "published_b": [0.198850858745165, 0.0295588022415444, 0.371563556432483], "constants": { "cutpoints": [2.8446, 3.51696, 4.2669], "cutpoints_weighted": [2.87046, 3.54282, 4.29276] }, "original": { "estimate": 1.26037038254781, "ci": [1.06428102424848, 1.49258839067006], "b": 0.23140563217028, "se": 0.0853223660155583, "df": 109, "p": 0.00777012062176199, "n": 29467, "events": 2564, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, "harmonized": { "estimate": 1.27009882486562, "ci": [1.07322956155736, 1.50308105805826], "b": 0.239094712297902, "se": 0.0849773413210373, "df": 109, "p": 0.00581273562502079, "n": 29467, "events": 2564, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.16371452268574, "published_z": 2.27915265305228, "interval_ratio_original": 1.01121320213695, "ratio_harmonized": 1.0332277138439, "variants": [ { "label": "crude (published 1.30, 1.14-1.50)", "estimate": 1.29501908797474, "ci": [1.11769261099904, 1.5004791314849], "b": 0.258525434791707, "se": 0.07429942421606, "df": 109, "p": 0.000723262756198726, "n": 29467, "events": 2564, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "Model 1 (published 1.40, 1.22-1.62)", "estimate": 1.4049227166416, "ci": [1.20671729214176, 1.63568372856609], "b": 0.339982295323786, "se": 0.0767311295509202, "df": 109, "p": 2.24418958173561e-05, "n": 29467, "events": 2564, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "Model 2 (published 1.13, 0.96-1.33)", "estimate": 1.1889081751767, "ci": [1.01432185214953, 1.39354451055797], "b": 0.173035386111303, "se": 0.0801300342698023, "df": 109, "p": 0.03300860488152, "n": 29467, "events": 2564, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "unweighted", "estimate": 1.23190121896741, "ci": [1.08989045264656, 1.39241573279982], "b": 0.208558682487218, "se": 0.0624917061854605, "df": null, "p": 0.000845716871311867, "n": 29467, "events": 2564, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, unweighted", "estimate": 1.25153543488093, "ci": [1.11749572770818, 1.40165273649406], "b": 0.224371145416889, "se": 0.0577974503062712, "df": null, "p": 0.000103590236409632, "n": 29467, "events": 2564, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1, unweighted", "estimate": 1.31587018160324, "ci": [1.17310349842405, 1.47601156859447], "b": 0.274498181839124, "se": 0.0585956594179584, "df": null, "p": 2.8048485678752e-06, "n": 29467, "events": 2564, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 3 without HDL (the Methods' list)", "estimate": 1.27683195964633, "ci": [1.08517153140936, 1.5023429992279], "b": 0.244381978451457, "se": 0.0820618457829818, "df": 109, "p": 0.00357593542745568, "n": 29467, "events": 2564, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "fasting subsample and weight (Methods' 'fasting')", "estimate": 1.16038698803712, "ci": [0.900746555214304, 1.49486884430605], "b": 0.148753559859533, "se": 0.127794686282749, "df": 109, "p": 0.246963820344225, "n": 13393, "events": 1128, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "weighted quartile cutpoints", "estimate": 1.23717169885023, "ci": [1.04541668155588, 1.46409928159746], "b": 0.212827886405666, "se": 0.0849723030297048, "df": 109, "p": 0.0137363365789666, "n": 29467, "events": 2564, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 } ], "left_out": ["marital status in four groups (widowed apart from divorced or separated): 2021-2023 releases only DMDMARTZ's three groups; the harmonized version uses married or living with a partner; widowed, divorced, or separated; never married", "two-hour OGTT glucose in the diabetes definition: 2021-2023 (like 2017-2018) has no OGTT; the harmonized version counts diabetes as told by a doctor, insulin or pills, HbA1c 6.5% or more, or fasting glucose 126 mg/dL or more"], "choices": [ { "choice": "weights", "decision": "examination weight (WTMEC2YR) over 7, with strata and PSUs; in 2021-2023 the phlebotomy weight (WTPH2YR from TCHOL_L)", "reason": "the paper built 14-year weights as NHANES recommends, and the PHQ-9 and lipids come from the examination; weighted fits reproduce the crude and Model 1 estimates (1.30 and 1.40), unweighted ones do not (1.25 and 1.32). NCHS directs the phlebotomy weight for blood analytes in 2021-2023" }, { "choice": "fasting", "decision": "no fasting restriction: everyone with total and HDL cholesterol", "reason": "the Methods say non-HDL cholesterol was computed from fasting subjects' lipids, but Figure 1 keeps 34,585 of 42,143 adults with a PHQ-9 and non-HDL cholesterol (ours 34,459), about twice the morning fasting subsample; the text's version (fasting subsample, fasting weight) is a variant" }, { "choice": "quartiles", "decision": "unweighted quartiles of non-HDL cholesterol (mmol/L) in the analytic sample, intervals closed on the right (2.84, 3.52, 4.27)", "reason": "unreported; they reproduce Table 2's crude and Model 1 Q4 estimates to the second decimal (Q2 and Q3 within 0.03); weighted quartiles (2.87, 3.54, 4.29) put the crude Q4 at 1.27 (variant)" }, { "choice": "HDL in Model 3", "decision": "included (continuous, mmol/L)", "reason": "Table 2's note lists HDL in Model 3 and the Methods do not; the published estimate is closer with it (variant without it)" }, { "choice": "study population", "decision": "adults 18 and older with a PHQ-9 and non-HDL cholesterol, complete case for Model 3's covariates; 18- and 19-year-olds drop out because education and the medical conditions are asked from 20", "reason": "the paper's Figure 1 (42,143 adults, as ours). We keep 29,467 (2,564 depressed) against the paper's 28,041 (2,419): the paper excluded 6,544 for missing covariates against our 4,992, and no coding we tried accounts for the gap (requiring a measured blood pressure drops 636, counting borderline diabetes as missing 411)" }, { "choice": "PHQ-9", "decision": "sum of the nine items, missing if any is missing, refused, or don't know; depression at 10 or more", "reason": "the paper's cutoff and its exclusion of incomplete responses; 36,259 adults have a PHQ-9 against the paper's 36,397 (allowing one missing item would give 36,401)" }, { "choice": "age, income to poverty ratio, BMI", "decision": "continuous", "reason": "Table 1 reports them as means (weighted 47.3, 3.04, 29.1 against 47.26, 3.06, 29.08); BMI or income in groups changes Model 3 by under 0.02" }, { "choice": "race, marital status, education", "decision": "non-Hispanic White, Mexican American, non-Hispanic Black, other; married or living with a partner, never married, widowed, divorced or separated (DMDMARTL); less than high school, high school or GED, some college, college graduate", "reason": "the paper's groups; weighted shares within 0.3 points of Table 1's" }, { "choice": "smoking and drinking", "decision": "never (fewer than 100 cigarettes), former, current (SMQ020, SMQ040); never (fewer than 12 drinks in life; never a drink from 2017-2018), former (none in the past 12 months), current", "reason": "unstated beyond never, former, current; smoking matches Table 1 (54.7, 25.2, 20.2% against 54.8, 25.0, 20.2%), drinking less closely (10.2, 15.4, 74.4% against 10.4, 13.2, 76.5%)" }, { "choice": "CVD", "decision": "ever told of heart failure, angina, heart attack, or stroke (coronary heart disease not included)", "reason": "the paper's list; 7.36% weighted against Table 1's 7.12%" }, { "choice": "diabetes", "decision": "told by a doctor, insulin or diabetes pills, HbA1c 6.5% or more, fasting glucose 126 mg/dL or more, or two-hour OGTT glucose 200 mg/dL or more (2005-2016)", "reason": "unstated; this matches Table 1 (13.97% weighted against 13.89%), while told only gives 9.42% and the same without the OGTT 13.07%" }, { "choice": "hypertension", "decision": "told by a doctor, taking medicine for it, or mean blood pressure 140/90 mmHg or more", "reason": "unstated; this matches Table 1 (37.62% weighted against 37.53%), while told only gives 31.80%" }, { "choice": "harmonized marital status and diabetes", "decision": "DMDMARTZ's three marital groups; diabetes without the OGTT", "reason": "2021-2023 releases only three marital groups and gave no OGTT" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text computes non-HDL cholesterol from fasting subjects' lipids, but Figure 1 keeps 34,585 adults with a PHQ-9 and non-HDL cholesterol (34,459 here with no fasting restriction), about twice the fasting subsample (13,393 of the 29,467 analyzed here have a fasting weight)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Figure 1 excludes 5,575 under 18, though it goes from 70,190 participants to 42,143 adults, 28,047 fewer; the abstract gives the 42,143 adults as the study's participants, of whom 28,041 were analyzed." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The abstract reports a positive association in participants of normal BMI with OR 0.93 (0.66 to 1.32), where Table 3 gives 1.39 (0.92 to 2.11), and 1.29 (1.01 to 1.66) in those without hypertension, where Table 3 gives 1.30 (1.01 to 1.68)." } ], "projected_power": 0.127359076700501 }, "row062": { "id": "row062", "row": 62, "doi": "10.3389/fnut.2023.1218166", "measure": "OR", "contrast": "DII tertile 3 (2.60 or more) vs tertile 1 (under 0.79)", "published": { "measure": "OR", "estimate": 1.31, "low": 1.19, "high": 1.44, "n": 31781, "events": 5491, "contrast": "DII tertile 3 (2.60 or more) vs tertile 1 (under 0.79)" }, "published_b": [0.27002713721306, 0.173953307123438, 0.364643113587909], "constants": null, "original": { "estimate": 1.23882047057061, "ci": [1.13457747162274, 1.35264113442145], "b": 0.214159693496538, "se": 0.0448474483535676, "df": null, "p": 1.79446043183894e-06, "n": 25142, "events": 4489, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 1.23631188646089, "ci": [1.13687251244511, 1.34444897195831], "b": 0.212132662528642, "se": 0.0427822046557347, "df": null, "p": 7.10643185153781e-07, "n": 27949, "events": 4961, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.793104336500665, "published_z": 5.55083120171565, "interval_ratio_original": 1.08470502702292, "ratio_harmonized": 0.990534955785555, "variants": [ { "label": "weighted (WTDRD1)", "estimate": 1.21029598586869, "ci": [1.05965973812361, 1.38234597457074], "b": 0.190864946115335, "se": 0.0669430984751698, "df": 94, "p": 0.0053548312243026, "n": 25142, "events": 4489, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 184, "strata": 90 }, { "label": "T2 vs T1 (published 1.17, 1.07-1.29)", "estimate": 1.12733406247384, "ci": [1.03431409730644, 1.22871968169379], "b": 0.119855608591487, "se": 0.0439381081274301, "df": null, "p": 0.00637528099322945, "n": 25142, "events": 4489, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "per unit of DII (published 1.06, 1.04-1.09)", "estimate": 1.0479563602209, "ci": [1.02782581200191, 1.06848117658033], "b": 0.0468419440212442, "se": 0.00989622007157287, "df": null, "p": 2.20880405137075e-06, "n": 25142, "events": 4489, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1, crude (published 1.21, 1.12-1.30)", "estimate": 1.20544116482621, "ci": [1.12200815736538, 1.29507828648021], "b": 0.186845611828984, "se": 0.0365953328436378, "df": null, "p": 3.29533102861837e-07, "n": 31784, "events": 5491, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2, age sex BMI (published 1.26, 1.17-1.36)", "estimate": 1.26163634192946, "ci": [1.16862575711951, 1.36204961218768], "b": 0.232409562473631, "se": 0.0390726927517457, "df": null, "p": 2.71218637461846e-09, "n": 31494, "events": 5435, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1, crude, weighted", "estimate": 1.12381501494564, "ci": [1.00687997384458, 1.25433042728509], "b": 0.116729160476648, "se": 0.0554361997489003, "df": 109, "p": 0.0375314683441927, "n": 31784, "events": 5491, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 214, "strata": 105 }, { "label": "Model 2, weighted", "estimate": 1.1739608537241, "ci": [1.04881351981871, 1.31404111411039], "b": 0.160383376492303, "se": 0.0568747131028451, "df": 109, "p": 0.00570734043218709, "n": 31494, "events": 5435, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 214, "strata": 105 }, { "label": "missing covariates as their own level", "estimate": 1.2405030120566, "ci": [1.14484414529705, 1.3441547735934], "b": 0.215516952245601, "se": 0.040943835081527, "df": null, "p": 1.41168109278325e-07, "n": 31485, "events": 5435, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "library DII grouping (alpha-linolenic acid in n-3)", "estimate": 1.25601453986375, "ci": [1.14846468113236, 1.3736360815151], "b": 0.227943644303639, "se": 0.0456731116009947, "df": null, "p": 6.01414649517563e-07, "n": 25142, "events": 4489, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Cockcroft-Gault as printed (exclusion and covariate)", "estimate": 1.21226549776272, "ci": [1.11134489442094, 1.32235064419999], "b": 0.192490921217522, "se": 0.0443477652643032, "df": null, "p": 1.42166949953366e-05, "n": 24891, "events": 4737, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": ["physical activity (MET-minutes a week over work, transport, and leisure): 2021-2023 asks about leisure-time activity only", "the two-hour OGTT glucose part of the diabetes definition: 2021-2023 has no OGTT", "education of 18- and 19-year-olds (DMDEDUC3): 2021-2023 records education from age 20 only, so the harmonized version, like a complete-case analysis there, leaves 18- and 19-year-olds out"], "choices": [ { "choice": "weights", "decision": "unweighted", "reason": "the paper mentions no weights and used EmpowerStats; Table 1 is unweighted, and only unweighted fits reproduce Models 1 and 2 (1.21 and 1.26, with their intervals)" }, { "choice": "dietary recall", "decision": "day one, reliable recalls (DR1DRSTZ = 1)", "reason": "unstated; the 9,549 excluded for lacking dietary data are exactly those without a reliable day-one recall" }, { "choice": "DII parameters", "decision": "the library's 28 (folic acid as DR1TFA) with n-3 as EPA + DPA + DHA and n-6 as 18:2 + 18:3 + 18:4 + 20:4", "reason": "the paper names the parameters but not their nutrients; so built, the DII matches Table 2's component medians and IQRs, Table 1's mean and SD by sex (1.15, SD 1.88; 1.87, SD 1.80), the range (-5.28 to 5.79), and the tertile cutpoints (0.786 and 2.599 here); folic acid fits Table 2's folate scores and total folate does not, and with alpha-linolenic acid in n-3 the means are 0.74 and 1.52" }, { "choice": "vitamin D in 2005-2006", "decision": "left out of those participants' DII", "reason": "NHANES dietary files have vitamin D only from 2007" }, { "choice": "tertile cutpoints", "decision": "the published 0.79 and 2.60, T2 from 0.79 and T3 from 2.60", "reason": "Methods and Table 3; the sample's own tertiles of this DII are 0.786 and 2.599 (10,617, 10,579, and 10,588 people at the published cutpoints against 10,594, 10,593, and 10,594)" }, { "choice": "eGFR", "decision": "CKD-EPI 2009, with 2005-2006 creatinine recalibrated as NCHS directs, for both the exclusion (under 60) and the covariate", "reason": "the Methods name CKD-EPI but print a Cockcroft-Gault formula (in mmol/L); Table 1's eGFR means and SDs by sex (98.2, SD 19.3; 101.2, SD 20.5) and the 2,956 excluded match CKD-EPI 2009 with the recalibration (2,954 here; 3,031 without it); Cockcroft-Gault gives means near 123 with SDs near 44" }, { "choice": "pregnancy", "decision": "RIDEXPRG = 1", "reason": "unstated; excludes 641 (published 642)" }, { "choice": "drinking", "decision": "Rattan's five groups as alcohol5() builds them (never, former, mild, moderate, heavy)", "reason": "the paper's definition; Table 1's shares are close (men: heavy 25.4% here against 25.9%, moderate 12.6% against 12.4%, mild 36.4% against 38.1%, former 17.4% against 15.2%); counting ever drinking 4 or 5 drinks almost every day (ALQ150 or ALQ151) as heavy would make 33.0% heavy" }, { "choice": "physical activity", "decision": "GPAQ MET-minutes a week (8 METs vigorous, 4 moderate and transport) over work, transport, and leisure, in the paper's three groups; missing in 2005-2006, which asked no GPAQ", "reason": "the paper's groups; no definition tried reproduces Table 1's shares, and leaving MET out moves the estimate by under 1%" }, { "choice": "education", "decision": "under high school (DMDEDUC2 1-2), high school or GED (3), more (4-5); 18- and 19-year-olds by DMDEDUC3", "reason": "Table 1's three groups; under high school matches (26.1% of men)" }, { "choice": "race", "decision": "non-Hispanic White, non-Hispanic Black, Mexican American, other (other Hispanic with other races)", "reason": "Table 1; shares match" }, { "choice": "diabetes", "decision": "told, HbA1c 6.5% or more, fasting glucose 126 mg/dL or more, random glucose 200 or more, OGTT 200 or more, or insulin or diabetes pills", "reason": "the Methods list the first five (HbA1c 'greater than' 6.5%); Table 1's share in men (16.53%) is matched only with 6.5% or more and the medicines (16.53%; 16.06% as printed)" }, { "choice": "hypertension", "decision": "mean of the available readings 140/90 or more, told, or taking prescribed medicine (BPQ050A)", "reason": "unstated readings; Table 1's shares match (38.0% of men, 36.4% of women)" }, { "choice": "hyperlipidemia", "decision": "NCEP ATP III thresholds with fasting triglycerides and LDL, or cholesterol medicine (hyperlipidemia_status)", "reason": "unstated sources; Table 1's shares are close (65.8% vs 67.0% of men); serum triglycerides overshoot (70.3%)" }, { "choice": "missing covariates", "decision": "complete case (25,142 of 31,784)", "reason": "unstated; Table 3 heads every model n = 31,781, but its Model 3 interval is wider than Model 2's by more than the covariates alone explain, as a smaller complete-case sample would make it; keeping missing values as their own level changes the estimate little (variant)" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods name the CKD-EPI creatinine equation for eGFR, an exclusion step and a Model 3 covariate, but print a Cockcroft-Gault formula with creatinine in mmol/L; Table 1's eGFR means and SDs (98.23, SD 19.26 in men; 101.22, SD 20.45 in women) and the 2,956 excluded are those of CKD-EPI 2009 (2,954 excluded here), while the printed formula gives means near 123 with SDs near 44." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The DII's n-3 and n-6 fatty acids were summed as EPA, DPA, and DHA (n-3) and 18:2, 18:3, 18:4, and 20:4 (n-6), counting alpha-linolenic acid (18:3, an n-3 fat) as n-6: so grouped, the DII reproduces Table 2's component medians, Table 1's means by sex (1.15 and 1.87), and the tertile cutpoints (0.79 and 2.60; 0.786 and 2.599 here), while 18:3 counted as n-3 gives means of 0.74 and 1.52." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text defines diabetes by a doctor's diagnosis, HbA1c greater than 6.5%, or fasting, random, or two-hour OGTT glucose, which gives 16.06% of men here, but Table 1's 16.53% is matched only with HbA1c of 6.5% or more and insulin or diabetes pills added (16.53% here)." }, { "kind": "sample", "affects_headline": true, "followed": true, "detail": "Table 3 gives n = 31,781 for every model, but Model 3's interval for tertile 3 is too wide for that many: it is 27% wider on the log scale than Model 2's (1.19 to 1.44 against 1.17 to 1.36), where keeping everyone (missing covariates as their own level) widens it by 5% here (1.14 to 1.34 against 1.17 to 1.36)." } ], "projected_power": 0.501291733663888 }, "row066": { "id": "row066", "row": 66, "doi": "10.1371/journal.pone.0300566", "measure": "OR", "contrast": "per mg of caffeine a day among those in the highest quartile of caffeine intake (the cell Table 2 labels quartile 4 against quartile 1)", "published": { "measure": "OR", "estimate": 1.002, "low": 1.0005, "high": 1.0036, "n": 10001, "events": null, "contrast": "per mg of caffeine a day among those in the highest quartile of caffeine intake (the cell Table 2 labels quartile 4 against quartile 1)" }, "published_b": [0.00199800266267306, 0.000499875041650993, 0.00359353551013022], "constants": { "intercept": -6.58277149246062, "slope": 0.0145572094552852, "mean": 19.9817974452555, "adjusted_cuts": [2.41810146681459, 9.40445014224782, 21.6307973040179], "raw_cuts": [0.5, 4.5, 21.5] }, "original": { "estimate": 1.00147502447179, "ci": [1.00002399105847, 1.00292816333257], "b": 0.00147393769175045, "se": 0.00073978243095004, "df": null, "p": 0.04632790220726, "n": 2543, "events": 1079, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 1.00147502447179, "ci": [1.00002399105847, 1.00292816333257], "b": 0.00147393769175045, "se": 0.00073978243095004, "df": null, "p": 0.04632790220726, "n": 2543, "events": 1079, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.737705569310162, "published_z": 2.53163739185596, "interval_ratio_original": 1.06681852402665, "ratio_harmonized": 1, "variants": [ { "label": "crude (published 1.0008, 0.9996-1.0019)", "estimate": 1.0011161672972, "ci": [0.99984308242196, 1.00239087317191], "b": 0.00111554484561385, "se": 0.000649233734161568, "df": null, "p": 0.0857513109861768, "n": 2543, "events": 1079, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "sex, age, race (published 1.0009, 0.9997-1.0022)", "estimate": 1.00127954945131, "ci": [0.999917660181122, 1.00264329361661], "b": 0.00127873152555445, "se": 0.000694438645443352, "df": null, "p": 0.0655646068059409, "n": 2543, "events": 1079, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "weighted", "estimate": 1.00166163014994, "ci": [0.999266773016087, 1.00406222683287], "b": 0.00166025116991639, "se": 0.00120079801619561, "df": 72, "p": 0.171053990854558, "n": 2543, "events": 1079, "weighted": true, "weights": ["WTDR2D", "WTDR2DPP"], "dropped": [], "psu": 140, "strata": 68 }, { "label": "within quartile 3, crude (published 1.0171, 1.0005-1.0340)", "estimate": 1.01853876629511, "ci": [1.00150195593553, 1.03586539426863], "b": 0.0183690181057938, "se": 0.00860637701049619, "df": null, "p": 0.0328141665625376, "n": 2555, "events": 861, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "within quartile 3, sex, age, race (published 1.0088, 0.9918-1.0262)", "estimate": 1.00923165809763, "ci": [0.991884097780693, 1.02688261863], "b": 0.00918930679099956, "se": 0.00884624532165629, "df": null, "p": 0.29890685061983, "n": 2555, "events": 861, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "as the text describes: quartile 4 vs 1 of energy-adjusted caffeine, weighted", "estimate": 1.48286938423971, "ci": [1.19536531039342, 1.83952268950464], "b": 0.393978983915983, "se": 0.10811692261618, "df": 72, "p": 0.000503256251890596, "n": 10106, "events": 3600, "weighted": true, "weights": ["WTDR2D", "WTDR2DPP"], "dropped": [], "psu": 140, "strata": 68 }, { "label": "as the text describes, crude", "estimate": 1.66270553646726, "ci": [1.39090302287229, 1.98762218180377], "b": 0.508446116863945, "se": 0.089539095120345, "df": 72, "p": 2.68077108979657e-07, "n": 10106, "events": 3600, "weighted": true, "weights": ["WTDR2D", "WTDR2DPP"], "dropped": [], "psu": 140, "strata": 68 }, { "label": "as the text describes, quartiles of unadjusted caffeine (Table 1's grouping)", "estimate": 1.56089426625347, "ci": [1.25459758098205, 1.94197003673149], "b": 0.445258904616184, "se": 0.109580149700056, "df": 72, "p": 0.000121858184287012, "n": 10106, "events": 3600, "weighted": true, "weights": ["WTDR2D", "WTDR2DPP"], "dropped": [], "psu": 140, "strata": 68 }, { "label": "per mg of caffeine, all participants", "estimate": 1.00252080885282, "ci": [1.00137136468414, 1.003671572434], "b": 0.00251763694358419, "se": 0.000585322858094431, "df": null, "p": 1.69815051162261e-05, "n": 10106, "events": 3600, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "obese only (BMDBMIC 4) as the outcome", "estimate": 1.00182827856731, "ci": [1.00025351413977, 1.00340552224933], "b": 0.00182660930033192, "se": 0.000802630712731433, "df": null, "p": 0.022859301459942, "n": 2543, "events": 621, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Methods' Model 3 list (+K, Mg, vitamin D, -selenium)", "estimate": 1.00121649821302, "ci": [0.999734195386084, 1.00270099884581], "b": 0.00121575887860307, "se": 0.000755931658190388, "df": null, "p": 0.107771230304432, "n": 2543, "events": 1079, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "income-to-poverty ratio imputed by regression, for the paper's multiple imputation", "estimate": 1.00146720436332, "ci": [1.00006594521398, 1.00287042691038], "b": 0.00146612907064919, "se": 0.000714393755191991, "df": null, "p": 0.0401434028366883, "n": 2760, "events": 1167, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": [], "choices": [ { "choice": "outcome", "decision": "overweight or obese (BMDBMIC 3 or 4) against underweight or normal weight (1 or 2)", "reason": "the Methods compute ORs 'for overweight and obese individuals', and Table 1's weighted 'rate of obesity' of 34.87% fits BMDBMIC 3 or 4 (35.4% in this sample) and not 4 alone (19.3%)" }, { "choice": "study population", "decision": "aged 2-19 with two reliable 24-hour recalls and a BMI category (n 10,960), and of them, for the headline, those in the highest quartile of caffeine intake", "reason": "the paper's flow (16,225 aged 2-19, less those without caffeine and without a BMI category); the 16,225 match, the 192 without a BMI category are close to the 198 here, but the 10,193 with caffeine (10,001 in the end) match no definition tried (two recalls 11,158, the first recall 13,144)" }, { "choice": "caffeine", "decision": "mean of the two recalls' totals (DR1TCAFF, DR2TCAFF)", "reason": "the Methods: 'the average caffeine intake from two total nutrient intake recalls'" }, { "choice": "energy adjustment", "decision": "for the variants that follow the text's description: residual method, caffeine regressed on energy (two-day means) by least squares in the analytic population, the residual plus the mean caffeine; the paper's cycles fix the coefficients", "reason": "the Methods say caffeine was adjusted for total energy 'with a residual model'; this is the usual (Willett) form" }, { "choice": "quartiles", "decision": "unweighted quartiles of unadjusted caffeine in the analytic population of the paper's cycles (quartile 1: no caffeine), reused in 2021-2023; the text's quartiles of energy-adjusted caffeine are used in its variants", "reason": "Table 1's quartiles are of unadjusted caffeine (quartile 1 has no caffeine; unadjusted quartiles here have mean energy 1,594, 1,727, 1,878, 2,077 kcal against Table 1's 1,602, 1,741, 1,890, 2,098, adjusted ones 2,351, 1,647, 1,418, 1,884), and Table 2's cells are reproduced within them; the printed cutpoints (236.25, 472.5, 708.05: steps of 236.25) have no units and fit no caffeine scale" }, { "choice": "weights", "decision": "unweighted, as the paper's computation was; the two-day dietary weight (WTDR2D, WTDR2DPP for 2017-March 2020, pooled 2/9.2 and 3.2/9.2) in a variant and in the 'as the text describes' variants", "reason": "the paper names 'combined dietary sample weights', but Table 2's cells and their interval widths are reproduced only by unweighted models" }, { "choice": "Model 3 covariates", "decision": "the Table 2 note's list: sex, age (continuous), race, income-to-poverty ratio (continuous), and two-day mean energy, protein, total sugar, total fat, cholesterol, vitamins B6 and C, calcium, phosphorus, zinc, copper, sodium, selenium", "reason": "the note of the headline's table; the Methods' list (with potassium, magnesium and vitamin D, without selenium) is a variant" }, { "choice": "'total intake (from dietary and supplements)'", "decision": "not a covariate of its own; the dietary covariates come from the recalls' food and beverage totals", "reason": "it names no variable, and the Methods say the dietary factors are 'the average of two total nutrient intakes' (DR1TOT, DR2TOT); adding the 24-hour supplement intakes to each dietary covariate is a variant (2021-2023 has no 24-hour supplement recall)" }, { "choice": "energy adjustment of the dietary covariates", "decision": "raw two-day means, with energy in the model", "reason": "the notes say dietary confounders were also residual-adjusted; with energy in the model a linear residual is a reparametrization that leaves the caffeine estimate unchanged" }, { "choice": "race", "decision": "RIDRETH1's five groups as indicators", "reason": "Table 1's groups; 'characteristics with three or more categories as indicator variables'" }, { "choice": "missing income-to-poverty ratio", "decision": "complete cases", "reason": "the paper reports 8 multiple imputations by fully conditional specification, which kept those missing the ratio (Table 1's 2,028 Mexican American participants exceed the 1,889 with a ratio in this larger sample); multiple imputation's random draws can't be reproduced, so, as for every paper that used it, the re-implementation uses complete cases, and a single regression imputation of the ratio is a variant" }, { "choice": "what Table 2 reports", "decision": "the per-milligram odds ratio of unadjusted caffeine among those in its highest quartile, unweighted, as the paper computed the cell it labels quartile 4 against quartile 1; the quartile contrast the Methods and Results describe is a variant", "reason": "Table 1's weighted prevalences (33.80% in Q1, 40.77% in Q4) give a crude Q4 vs Q1 OR near 1.35, not Table 2's 1.0008. Table 2's cells match per-milligram ORs of unadjusted caffeine within quartiles of unadjusted caffeine from unweighted models (crude Q3 1.0181, 1.0017-1.0348 here against 1.0171, 1.0005-1.0340; crude Q4 1.0010 against 1.0008; Model 2 Q3 1.0080 against 1.0088; Model 3 Q4 1.0015, 1.0001-1.0029 against 1.0020, 1.0005-1.0036), with interval widths that scale with each quartile's spread of caffeine; weighted fits and energy-adjusted caffeine match less well" } ], "departures": [ { "kind": "reporting", "affects_headline": true, "followed": true, "detail": "Table 2 and the Results give 1.0020 (1.0005 to 1.0036) as Model 3's odds ratio for quartile 4 against quartile 1, but Table 1's weighted obesity rates (33.80% in quartile 1, 40.77% in quartile 4) give a crude contrast near 1.35. Table 2's cells are per-milligram odds ratios of caffeine among those in each quartile: so computed, Model 3 gives 1.0015 (1.0000 to 1.0029) for quartile 4 here, and the crude cells for quartiles 3 and 4 give 1.0185 and 1.0011 (published 1.0171 and 1.0008)." }, { "kind": "weighting", "affects_headline": true, "followed": true, "detail": "The abstract says all data were survey-weighted and the Methods name combined dietary sample weights, but Table 2's cells and their interval widths are reproduced only unweighted (Model 3, quartile 4: 1.0015, 1.0000 to 1.0029 unweighted, and 1.0017, 0.9993 to 1.0041 weighted, against the published 1.0020, 1.0005 to 1.0036)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods adjust caffeine for energy by the residual method, and Tables 1 and 2 print quartile cutpoints of 236.25, 472.5, and 708.05, but Table 1's quartiles are of unadjusted caffeine, with mean intakes of 0.00, 1.79, 9.86, and 70.09 mg a day and mean energy (1,602 to 2,098 kcal) matching unadjusted quartiles here (1,594 to 2,077), not energy-adjusted ones (2,351, 1,647, 1,418, 1,884); Table 2's cells are reproduced within unadjusted quartiles." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's Overall column is the mean of its four quartile columns (caffeine 20.43, SD 18.86; 507 Mexican American participants, the mean of 434, 466, 598, and 530), not the whole sample's, and its 214 other-race participants in quartile 2 should be 428 for that quartile's counts to sum to its 2,497." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's weighted 'rate of obesity' (34.87%) is the share overweight or obese (BMI-for-age category 3 or 4: 35.4% here), not the share obese (19.3%)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Fig 1 starts from 45,932 participants and excludes 29,707 outside ages 2 to 19, but the four cycles' files hold 45,462, among them exactly the paper's 16,225 aged 2 to 19." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The Results give the inflection point of caffeine as 479.15 where Table 3 prints 26.99, and Table 3's interval below that point (1.0009 to 1.0013) excludes its own estimate (1.0022)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The abstract and Results give per-quartile effects that match no printed estimate (a 0.05% increase in Model 3 and a 0.03% decrease in Model 1, where Table 2's per-quarter odds ratios are 1.0022 and 1.0009), and the abstract gives Table 4's quartile 4 against quartile 1 estimates (1.5961 for boys, 1.4418 for girls) as per-quartile increments." } ], "projected_power": 0.197602333334283 }, "row068": { "id": "row068", "row": 68, "doi": "10.1038/s41598-022-05124-y", "measure": "beta", "contrast": "workday sleep under 6 hours vs 6 to under 8 hours (systolic blood pressure, mmHg)", "published": { "measure": "beta", "estimate": 3.58, "low": 1.6, "high": 5.56, "n": 2887, "events": null, "contrast": "workday sleep under 6 hours vs 6 to under 8 hours (systolic blood pressure, mmHg)" }, "published_b": [3.58, 1.6, 5.56], "constants": null, "original": { "estimate": 3.5811169085125, "ci": [1.08371723103702, 6.07851658598799], "b": 3.5811169085125, "se": 1.22285333172369, "df": 30, "p": 0.00644662896711037, "n": 2887, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, "harmonized": { "estimate": 3.46591170321774, "ci": [0.859989316356447, 6.07183409007902], "b": 3.46591170321774, "se": 1.27599154501691, "df": 30, "p": 0.0108509956607612, "n": 2887, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.00031198561802, "published_z": 3.54377326497646, "interval_ratio_original": 0.792824639907658, "ratio_harmonized": 0.967829811693408, "variants": [ { "label": "told hypertension and snorting recorded in both cycles, as the text defines them", "estimate": 3.67797597355338, "ci": [1.12318098043623, 6.23277096667053], "b": 3.67797597355338, "se": 1.25095698433114, "df": 30, "p": 0.00626226075208735, "n": 2887, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "unweighted", "estimate": 2.49044584983188, "ci": [0.558459021954714, 4.42243267770905], "b": 2.49044584983188, "se": 0.985725678184107, "df": null, "p": 0.0115202086499333, "n": 2887, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, weighted (published 6.15, 3.88-8.42)", "estimate": 6.15097357859336, "ci": [3.2852266390379, 9.01672051814881], "b": 6.15097357859336, "se": 1.40321480158706, "df": 30, "p": 0.000131876510873201, "n": 2887, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "crude, unweighted", "estimate": 4.9714162770975, "ci": [2.62965103184983, 7.31318152234518], "b": 4.9714162770975, "se": 1.19480014108383, "df": null, "p": 3.17027980227764e-05, "n": 2887, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model I, weighted (published 4.17, 2.19-6.15)", "estimate": 4.16791905286195, "ci": [1.45212552320172, 6.88371258252219], "b": 4.16791905286195, "se": 1.32979001958378, "df": 30, "p": 0.00383400763687638, "n": 2887, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "Model I, unweighted", "estimate": 2.88227355040947, "ci": [0.932161650268138, 4.83238545055079], "b": 2.88227355040947, "se": 0.994973333961012, "df": null, "p": 0.00376947985047432, "n": 2887, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": "snort, gasp, or stop breathing while asleep (SLQ040): not asked in 2021-2023", "choices": [ { "choice": "population", "decision": "everyone with workday sleep hours (SLD012, asked from age 16), a mean systolic reading, not taking prescribed blood pressure medicine (BPQ050A), and BMI under 25; no age limit", "reason": "the paper states no age limit; these steps reproduce its flow exactly (6,818, 1,055, 2,944, and 5,521 excluded; 2,887 left; 1,378 men, 1,509 women)" }, { "choice": "weights", "decision": "examination weight, WTMEC2YR, each cycle's halved; in 2021-2023 the phlebotomy weight (WTPH2YR from BIOPRO_L), because the model holds blood analytes", "reason": "the paper names no weight; Table 1's weighted means and percentages (age 38.54, SBP 115.33, BMI 21.97, sleep groups 40.53/6.26/53.21) are reproduced exactly with it, and so are the crude and Model I estimates and CIs (as weighted least squares, which the paper's CIs are); unweighted fits do not reproduce them. NCHS directs the phlebotomy weight for blood analytes in 2021-2023; it leaves out those with no blood sample (about 6% there), whom the paper's missing indicators would keep" }, { "choice": "sleep groups", "decision": "under 6 hours, 6 to under 8 (reference), 8 or more", "reason": "the paper's labels (< 6 h, 6-8 h, >= 8 h); they reproduce Table 1's weighted shares exactly" }, { "choice": "systolic blood pressure", "decision": "mean of all available auscultatory readings (BPXSY1-4); 2021-2023 measured only with the oscillometric device (BPXO, three readings)", "reason": "the paper's definition; the library reads BPX in 2015-2018 and BPXO in 2021-2023" }, { "choice": "race", "decision": "Mexican American, non-Hispanic White, non-Hispanic Black, and other (other Hispanic and other or multiracial, RIDRETH1 2 and 5)", "reason": "reproduces Table 1's weighted shares and Table 4's subgroup counts (302, 968, 550, 1,067)" }, { "choice": "alcohol, smoking, diabetes", "decision": "drinking in the past 12 months (ALQ120Q above 0 in 2015-2016, ALQ121 1-10 after) or not (0), else not recorded; smoking now (SMQ040 every day or some days), not at all, else not recorded (never smokers and under 18s skip it); DIQ010 yes, no, borderline, else not recorded", "reason": "the paper's definitions; each reproduces Table 1's weighted percentages exactly" }, { "choice": "snort or stop breathing", "decision": "SLQ040 rarely, occasionally, or frequently as yes, never as no, don't know as not recorded", "reason": "unstated; this split reproduces Table 1's percentages exactly (for 2017-2018, the only cycle the paper's data recorded)" }, { "choice": "told hypertension and snorting in 2015-2016", "decision": "unrecorded in 2015-2016 in the paper version, as the paper's data had them; recorded from BPQ020 in every cycle in the harmonized version (2021-2023 has no snorting item)", "reason": "Table 1's 'not recorded' shares (50.04% and 51.37%) equal 2015-2016's weighted share plus a few nonresponses: the paper's data lost both items for 2015-2016, a data-handling failure of its own files, which the plan follows in the paper version only. That version reproduces Model II to the second decimal (3.58, 1.60-5.56; 1.70 for 8 hours or more); recorded in both cycles (variant) it gives 3.68" }, { "choice": "missing laboratory covariates", "decision": "albumin, creatinine, AST, hemoglobin, total cholesterol, and HDL each with a missing indicator and the missing value set to 0", "reason": "the paper's missing-indicator method; the fill value does not change the fit" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "Model II adjusts for told hypertension and snort or stop breathing, but Table 1's 'not recorded' shares for them (50.04% and 51.37%) equal 2015-2016's weighted share plus a few nonresponses, so the analysis had neither item for 2015-2016. With both unrecorded there, Model II gives 3.581 (1.70 for 8 hours or more), as published, and the paper version codes them so; recorded in both cycles, as the text defines them and the harmonized version records them, it gives 3.678." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's snort or stop breathing rows swap yes and no in the columns by sex: 6.72% of all participants are yes (as the Results say) and 41.91% no, which are the averages of the men's and women's 'No' shares (8.56%, 5.22%) and 'Yes' shares (39.08%, 44.22%)." } ], "projected_power": 0.649293051581877 }, "row072": { "id": "row072", "row": 72, "doi": "10.1097/MD.0000000000037315", "measure": "OR", "contrast": "per 1% of red cell distribution width", "published": { "measure": "OR", "estimate": 1.145, "low": 1.036, "high": 1.266, "n": 1236, "events": 102, "contrast": "per 1% of red cell distribution width" }, "published_b": [0.135404637006203, 0.0353671438372913, 0.235862323721984], "constants": null, "original": { "estimate": 1.15151232538928, "ci": [1.0453029512157, 1.26851324200443], "b": 0.141076143989688, "se": 0.0485433477371543, "df": 72, "p": 0.00485833698091838, "n": 1235, "events": 102, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 140, "strata": 68 }, "harmonized": { "estimate": 1.15151232538928, "ci": [1.0453029512157, 1.26851324200443], "b": 0.141076143989688, "se": 0.0485433477371543, "df": 72, "p": 0.00485833698091838, "n": 1235, "events": 102, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 140, "strata": 68 }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.04188561860865, "published_z": 2.64732760183566, "interval_ratio_original": 1.03594312871644, "ratio_harmonized": 1, "variants": [ { "label": "unweighted", "estimate": 1.11015758269273, "ci": [0.995599274020999, 1.2378975061247], "b": 0.104501971637813, "se": 0.0555685768581932, "df": null, "p": 0.0600270720074395, "n": 1235, "events": 102, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "each cycle's weight over 4", "estimate": 1.15154002210325, "ci": [1.04009869202385, 1.2749217287499], "b": 0.141100196168964, "se": 0.0510591680949082, "df": 72, "p": 0.00725603363111486, "n": 1235, "events": 102, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 140, "strata": 68 }, { "label": "interview weight (WTINT2YR, WTINTPRP)", "estimate": 1.14891541758654, "ci": [1.04108573377735, 1.26791348103358], "b": 0.138818382220109, "se": 0.0494386957914902, "df": 72, "p": 0.00641319872451973, "n": 1235, "events": 102, "weighted": true, "weights": ["WTINT2YR", "WTINTPRP"], "dropped": [], "psu": 140, "strata": 68 }, { "label": "interview weight, each cycle's over 4", "estimate": 1.14781067871337, "ci": [1.03497725819485, 1.27294522052234], "b": 0.13785637028482, "se": 0.0519081052184803, "df": 72, "p": 0.00973633150752748, "n": 1235, "events": 102, "weighted": true, "weights": ["WTINT2YR", "WTINTPRP"], "dropped": [], "psu": 140, "strata": 68 }, { "label": "Model 2, weighted (published 1.168, 1.060-1.287)", "estimate": 1.15824740208278, "ci": [1.02567022001784, 1.30796138783105], "b": 0.146908002344505, "se": 0.0609801615042767, "df": 72, "p": 0.0185507283190232, "n": 1235, "events": 102, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 140, "strata": 68 }, { "label": "Model 2, unweighted", "estimate": 1.13439543575058, "ci": [1.00551332991712, 1.27979706122625], "b": 0.126099853295549, "se": 0.0615325929968814, "df": null, "p": 0.0404310250903317, "n": 1235, "events": 102, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 3, weighted (published 1.187, 1.065-1.322)", "estimate": 1.12662544357589, "ci": [0.962137718712936, 1.31923410279612], "b": 0.11922683161533, "se": 0.0791710037104027, "df": 72, "p": 0.136457252741912, "n": 1082, "events": 92, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 140, "strata": 68 }, { "label": "Model 3, unweighted", "estimate": 1.16994657814454, "ci": [1.02279808871297, 1.33826510903487], "b": 0.15695808806171, "se": 0.0685808481022684, "df": null, "p": 0.0220992468814111, "n": 1082, "events": 92, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 3 with moderate recreational activity, weighted", "estimate": 1.12295020154387, "ci": [0.957166021688626, 1.31744872527208], "b": 0.115959330644736, "se": 0.0801307606033842, "df": 72, "p": 0.152201580287305, "n": 1082, "events": 92, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 140, "strata": 68 } ], "left_out": [], "choices": [ { "choice": "headline model", "decision": "Model 1: RDW alone", "reason": "the abstract's estimate is Table 3's Model 1 and Table 2's univariate estimate, which the table note says adjusts for nothing, although the abstract says it holds after adjustment" }, { "choice": "cycles", "decision": "2011-2012, 2013-2014, 2015-2016, and the 2017-March 2020 files, without the 2017-2018 files", "reason": "the paper's 45,462 participants are exactly these four cycles' DEMO files; with the 2017-2018 files instead of the 2017-March 2020 ones the total would be 39,156" }, { "choice": "population", "decision": "RDW measured (LBXRDW), coronary heart disease answered yes or no (MCQ160C 1 or 2), and told of arthritis (MCQ160A = 1) of the rheumatoid type (MCQ195 = 2); no age step, since the arthritis and heart disease items are asked from age 20", "reason": "the paper's steps; they reproduce its flow exactly (36,770 with RDW, 23,659 with heart disease answered) and give 1,235 with rheumatoid arthritis and 102 cases, Table 1's totals (its tertiles sum to 1,235, its cases to 102; the text says 1,236). Counting refused or don't know as answered would give 23,739 and 1,243" }, { "choice": "weights", "decision": "examination weight (WTMEC2YR; WTMECPRP for 2017-March 2020), each cycle's times its share of the 9.2 pooled years as NCHS directs, with strata and PSUs; WTPH2YR from CBC_L in 2021-2023", "reason": "the paper says only 'a weighted approach'; RDW is measured at the examination, so NCHS's guidelines call for the examination weight. Weighted fits come closer than unweighted ones to Model 1 (1.152 against 1.145; unweighted 1.110) and Model 2 (1.158 against 1.168; unweighted 1.134). No weighting reproduces every model: the interview weight with each cycle's weight over 4 (weights stacked unscaled) matches Model 1 most closely (1.148, 1.037-1.271, P .0098 as published) but gives 1.149 (P .04) for Model 2, against the published 1.168 (P .0026). The interview weight and dividing each cycle's weight by 4 are variants" }, { "choice": "RDW scale", "decision": "continuous, per percentage point (LBXRDW, %)", "reason": "the paper reports the OR per unit of RDW and describes RDW as a percentage" }, { "choice": "coronary heart disease", "decision": "ever told of coronary heart disease (MCQ160C = 1) against not (2)", "reason": "the paper's 'CHD values'; refused or don't know answers are missing, which reproduces the flow's 13,111 excluded" }, { "choice": "Model 2 and Model 3 covariates (variants only)", "decision": "age, income-to-poverty ratio, and BMI continuous; gender; race in four groups (other Hispanic with other); education less than high school (DMDEDUC2 1, or unanswered), high school (2-3), university or above (4-5); hypertension (BPQ020), hypercholesterolemia (BPQ080), smoking (SMQ020), each yes against anything else; diabetes (DIQ010) yes or borderline against anything else; drinking as ever 4 or 5 or more drinks almost every day (ALQ151) yes, no, or other; complete cases for income and BMI", "reason": "unstated definitions; each grouping reproduces Table 1's counts exactly (race 161/434/398/242, education 169/502/564, hypertension 773/462, hypercholesterolemia 632/603, smoking 641/594, diabetes 389/846, drinking 242/750/243). Income is missing for 133 and BMI for 24, so Model 3 has 1,082 participants and 92 cases. Model 3 is not reproduced closely (1.127 weighted, 1.170 unweighted, against 1.187), and neither are Table 2's univariate estimates for the covariates under any weighting (women 0.66 weighted and 0.37 unweighted against 0.235; poverty 0.97 and 1.07 against 1.205), so those rows come from some other computation" }, { "choice": "physical activity in Model 3", "decision": "left out", "reason": "the Methods list moderate recreational activity, but Table 1, which defines Model 3, does not; adding it (PAQ665) is a variant" } ], "departures": [ { "kind": "reporting", "affects_headline": true, "followed": true, "detail": "The abstract presents OR 1.145 (1.036-1.266) as holding after adjusting for age, gender, race, education level, smoking, and drinking, but it is Table 3's Model 1, which the table note says adjusts for no covariates, and Table 2's univariate estimate; RDW alone reproduces it (1.152, 1.045-1.269)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The abstract, the text, and Tables 1, 2, and 4 name two race groups 'Hispanic White' and 'Hispanic Black', but Table 1's counts for them (434 and 398) are those of non-Hispanic White and non-Hispanic Black participants." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The text and Figure 1 give 1,236 participants, but Table 1's tertile columns sum to 1,235 (402, 408, 425), the number the paper's exclusion steps give when reproduced (with its 36,770 with RDW and 23,659 with heart disease answered exactly)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 2's univariate odds ratios for the covariates are not reproduced under any weighting by data that reproduce Table 1's counts exactly (women: 0.235 published; 0.66 with the examination weight, 0.37 unweighted)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The Results give the participants' mean age as 61.107 years, citing Table 1, but Table 1's tertile means (57.858, 61.027, and 62.915 years for 402, 408, and 425 participants) average to 60.645." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The abstract names the age subgroup 55-66 years, which the Results and Table 4 give as 56 to 66, and the Results give the male subgroup's P as .084, which its interval (1.034-1.442) rules out and Table 4 gives as .0184." } ], "projected_power": 0.211708509099763 }, "row087": { "id": "row087", "row": 87, "doi": "10.3389/fpubh.2024.1280163", "measure": "OR", "contrast": "per unit of log10 whole-blood manganese (ug/L), among participants below log10 Mn = 1.10", "published": { "measure": "OR", "estimate": 3.936, "low": 2.631, "high": 5.887, "n": 8542, "events": 4370, "contrast": "per unit of log10 whole-blood manganese (ug/L), among participants below log10 Mn = 1.10" }, "published_b": [1.37016497919001, 0.967364002058156, 1.77274653004612], "constants": null, "original": { "estimate": 4.15987509548748, "ci": [2.48904738755468, 6.95228258673595], "b": 1.42548504869921, "se": 0.262037971490049, "df": null, "p": 5.32823017234197e-08, "n": 4431, "events": 2269, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 4.25274873822194, "ci": [2.54712549920257, 7.10050283588714], "b": 1.44756553581054, "se": 0.261535452553331, "df": null, "p": 3.11453804595475e-08, "n": 4439, "events": 2271, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.04037475074126, "published_z": 6.6688161681367, "interval_ratio_original": 0.784079119172739, "ratio_harmonized": 1.01548980617613, "variants": [ { "label": "2017-2018 and 2017-2020 files stacked, as the authors did", "estimate": 4.05542925364087, "ci": [2.71127496527047, 6.06596772439325], "b": 1.40005653982941, "se": 0.20543109560131, "df": null, "p": 9.41246285821122e-12, "n": 7131, "events": 3687, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "stacked, at or above the knot (published 0.458, 0.121-1.726)", "estimate": 0.447991290236055, "ci": [0.118703337547277, 1.69073760076403], "b": -0.802981488193516, "se": 0.677638151794784, "df": null, "p": 0.236028954119072, "n": 1411, "events": 756, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "at or above the knot", "estimate": 0.388868449274188, "ci": [0.0701173586084102, 2.15665098974184], "b": -0.944514169231355, "se": 0.874031734005767, "df": null, "p": 0.279857019378182, "n": 842, "events": 452, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "continuous hinge at the knot, both segments", "estimate": 3.81902999846392, "ci": [2.3988685986489, 6.07994541150854], "b": 1.33999646326985, "se": 0.237248877002571, "df": null, "p": 1.62266276874407e-08, "n": 5273, "events": 2721, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "weighted (MEC weight)", "estimate": 5.38459241753507, "ci": [2.23873133962613, 12.9510115795396], "b": 1.68354161926381, "se": 0.426130542158521, "df": 25, "p": 0.00056216709458278, "n": 4431, "events": 2269, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 } ], "left_out": "the requirement that chromium and cobalt were measured (2021-2023 does not measure them): the harmonized version requires an age of 40 or more instead, since NHANES measured them only at those ages and requiring them limited the paper's sample to them", "choices": [ { "choice": "cycles", "decision": "2017-March 2020 files only", "reason": "the authors stacked the 2017-2018 and 2017-March 2020 files, which counts about 3,100 people twice (variant); 2015-2016 has no elastography, so it adds no one" }, { "choice": "model", "decision": "logistic, MAFLD on log10 Mn with no covariates, fitted on those below log10 Mn = 1.10", "reason": "Table 3 names no covariates; a refit of the authors' data this way reproduces its four estimates exactly, while a hinge model with a shared intercept does not (variant)" }, { "choice": "weights", "decision": "unweighted", "reason": "the paper mentions no weights; the authors' data reproduce Table 3 unweighted" }, { "choice": "steatosis", "decision": "CAP over 268 dB/m, partial exams kept, no reliability filter", "reason": "unstated; in the authors' data every MAFLD case has CAP 269 or more and MAFLD is known for everyone with a CAP value" }, { "choice": "metabolic criteria", "decision": "the paper's thresholds in mmol/L; drugs from the questionnaire (DIQ050, DIQ070, BPQ050A, BPQ100D); BP the mean of the readings; a missing measurement counts as not met", "reason": "unstated; this agrees with the authors' MAFLD for 9,612 of the 9,698 participants of 2017-March 2020 with a CAP value (and 5,901 of 5,948 in their 2017-2018 rows); no single change to a criterion explains the rest" }, { "choice": "metals required", "decision": "all ten (lead, cadmium, total, inorganic, methyl and ethyl mercury, manganese, selenium, chromium, cobalt)", "reason": "the paper excludes those missing heavy metal data; chromium and cobalt limit the sample to ages 40 and over" } ], "departures": [ { "kind": "sample", "affects_headline": true, "followed": false, "detail": "The 2017-2018 files are stacked with the 2017-March 2020 files, which already hold their participants: in the authors' supplementary dataset 3,141 of the 3,269 analytic rows labeled 2017-2018 have an exact twin among those labeled 2019-2020, so the 8,542 rows hold about 5,400 people. Stacked, the model gives 4.055 on 7,131 rows below the knot (the authors' count); counted once, 4.160 on 4,431." }, { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The paper says it analyzed adults aged 20 and over, but it required blood chromium and cobalt, which NHANES measured only at ages 40 and over, so its sample is aged 40 to 80 (Table 1's medians are 60 and 60; the rows with chromium in the authors' dataset are aged 40 to 80)." }, { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The paper says it used the 2015-2020 files, but 2015-2016 has no elastography, so no 2015-2016 participant has MAFLD status and none is in the analysis, in the authors' dataset as here." }, { "kind": "model", "affects_headline": true, "followed": true, "detail": "The Methods describe a piecewise linear regression, but Table 3's estimates are reproduced exactly from the authors' dataset by separate unadjusted, unweighted logistic models below and above log10 Mn = 1.10 (3.936, 2.631-5.887; 0.458, 0.121-1.726); a continuous hinge model with a shared intercept gives 3.546." }, { "kind": "reporting", "affects_headline": true, "followed": true, "detail": "The abstract gives the turning point as log10 Mn = 1.10 and as Mn = 12.61 ug/L, but 10^1.10 is 12.59 ug/L; Table 3 is reproduced exactly from the authors' dataset with the split at log10 Mn = 1.10, which the file uses." } ], "projected_power": 0.99841999821502 }, "row096": { "id": "row096", "row": 96, "doi": "10.3389/fmed.2022.925344", "measure": "beta", "contrast": "ng/mL of serum ferritin per unit of METS-IR", "published": { "measure": "beta", "estimate": 0.29, "low": 0.14, "high": 0.44, "n": 4182, "events": null, "contrast": "ng/mL of serum ferritin per unit of METS-IR" }, "published_b": [0.29, 0.14, 0.44], "constants": null, "original": { "estimate": 0.297893152999844, "ci": [0.145694709294407, 0.450091596705282], "b": 0.297893152999844, "se": 0.0776536941015036, "df": null, "p": 0.000124965423468804, "n": 4182, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 0.326212223445354, "ci": [0.173392013131117, 0.479032433759591], "b": 0.326212223445354, "se": 0.0779709277923796, "df": null, "p": 2.8671751619655e-05, "n": 4182, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.02721776896498, "published_z": 3.7892637034441, "interval_ratio_original": 0.985555412710446, "ratio_harmonized": 1.09506452283421, "variants": [ { "label": "Model 1 (published 0.54, 0.41-0.68)", "estimate": 0.544258450949698, "ci": [0.411573016690363, 0.676943885209033], "b": 0.544258450949698, "se": 0.0676978940970041, "df": null, "p": 9.01923984999784e-16, "n": 4182, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2 (published 0.49, 0.36-0.63)", "estimate": 0.492685657001226, "ci": [0.360074643187074, 0.625296670815379], "b": 0.492685657001226, "se": 0.0676599237844019, "df": null, "p": 3.29408745815233e-13, "n": 4182, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1, weighted (MEC weight)", "estimate": 0.524680525872982, "ci": [0.365230644446668, 0.684130407299297], "b": 0.524680525872982, "se": 0.079259666114043, "df": 47, "p": 3.08791557658727e-08, "n": 4182, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 93, "strata": 46 }, { "label": "Model 2, weighted (MEC weight)", "estimate": 0.491039781851723, "ci": [0.340943868382824, 0.641135695320622], "b": 0.491039781851723, "se": 0.0746099770047488, "df": 47, "p": 3.5311461839576e-08, "n": 4182, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 93, "strata": 46 }, { "label": "Model 3 with METS-IR's components, as the text says", "estimate": 2.4045732376389, "ci": [1.32309280782844, 3.48605366744935], "b": 2.4045732376389, "se": 0.551785868689953, "df": null, "p": 1.31375504094461e-05, "n": 4182, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "weighted (MEC weight)", "estimate": 0.284454944900144, "ci": [0.13819936366829, 0.430710526131999], "b": 0.284454944900144, "se": 0.0727010169719637, "df": 47, "p": 0.000292924467713829, "n": 4182, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 93, "strata": 46 }, { "label": "education in DMDEDUC2's five levels", "estimate": 0.291099395155135, "ci": [0.138269773053786, 0.443929017256485], "b": 0.291099395155135, "se": 0.0779757298128182, "df": null, "p": 0.000189058492164056, "n": 4182, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "as the text describes it: fasting values, all five cycles, ferritin converted", "estimate": 0.14220539633546, "ci": [-0.0447776410430356, 0.329188433713955], "b": 0.14220539633546, "se": 0.0954012618871539, "df": null, "p": 0.136065786063993, "n": 2956, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "as the text describes it, weighted (fasting subsample weight)", "estimate": 0.143776828269668, "ci": [-0.0783804137534501, 0.365934070292787], "b": 0.143776828269668, "se": 0.11156647916892, "df": 77, "p": 0.20135750739667, "n": 2956, "events": null, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 153, "strata": 76 }, { "label": "as the text describes it, with METS-IR's components", "estimate": 2.95244074041884, "ci": [1.55114848226784, 4.35373299856984], "b": 2.95244074041884, "se": 0.714958167192979, "df": null, "p": 3.63506587305573e-05, "n": 2956, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "fasting values, 2005-2010 only, no ferritin conversion", "estimate": 0.172407064709173, "ci": [-0.0679748216440922, 0.412788951062438], "b": 0.172407064709173, "se": 0.122646073218369, "df": null, "p": 0.159804798980322, "n": 1866, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "serum values, all five cycles, no ferritin conversion", "estimate": 0.227934207758171, "ci": [0.0928791013571574, 0.362989314159184], "b": 0.227934207758171, "se": 0.068906932712188, "df": null, "p": 0.000940131518714602, "n": 6699, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": ["activity intensity (2021-2023 asks no work activity, and the paper's intensity combines work and recreation)", "blood transfusion (MCQ092 is not asked in 2021-2023)", "drinking as Table 1 codes it, from ALQ101 (at least 12 drinks in any one year), which 2021-2023 does not ask; the harmonized version uses the text's definition, drinking at least 12 times in the past year, from the past-year frequency questions every cycle asks"], "choices": [ { "choice": "glucose and triglycerides in METS-IR", "decision": "the biochemistry profile's non-fasting serum glucose and triglycerides (LBXSGL, LBXSTR), as the paper's computation used them", "reason": "the paper defines METS-IR with fasting values, but its flow chart and Table 1 (FPG 89.92 and 95.00, TG 123.47 and 136.72 mg/dL) match the serum values, and only they reproduce its Models 1-3; the published estimate comes from them, so the replication follows them (the text's version is a variant)" }, { "choice": "cycles", "decision": "2005-2010", "reason": "the paper names 2005-2010 and 2015-2018, but no 2015-2018 participant got a METS-IR in its data (its 6,193 with METS-IR and 1,990 under 20 are exactly 2005-2010's)" }, { "choice": "Model 3 covariates", "decision": "every Table 1 covariate but METS-IR's components (BMI, HDL, glucose, triglycerides)", "reason": "the text says all Table 1 covariates, but only leaving the components out reproduces the published 0.29 (0.14-0.44): 0.298 (0.146-0.450) on the paper's sample; with them it is 2.40" }, { "choice": "weights", "decision": "unweighted", "reason": "Models 1 and 2 are reproduced to the second decimal only unweighted; the text names 2-year weights" }, { "choice": "ferritin across analyzers", "decision": "2005-2008 values converted to the Elecsys scale with NCHS's Deming equation", "reason": "the paper says it converted them; its mean ferritin (54.04) and group sizes match unconverted values, and the conversion barely moves the estimate (variant)" }, { "choice": "pregnancy", "decision": "not excluded", "reason": "no exclusion is listed, and the paper's 4,182 include 420 pregnant women" }, { "choice": "age range", "decision": "20-49", "reason": "the paper reports ages 20-49 (Discussion; Table 1 means); ferritin was measured at every age in 2017-2018 and only to 49 otherwise" }, { "choice": "education", "decision": "less than high school, high school, more than high school", "reason": "the text and Table 1; DMDEDUC2's five levels give the printed Model 3 to the digit on the paper's sample, 0.291 (0.138-0.444) against 0.298 (0.146-0.450), so its model may have used them (variant)" }, { "choice": "race", "decision": "Hispanic (RIDRETH1 1-2), White, Black, other", "reason": "Table 1's 'Mexican American' count (1,322) is Mexican American and other Hispanic together" }, { "choice": "drinking", "decision": "ALQ101 (at least 12 drinks in any one year) with missing as its own level; in 2017-2018, which no longer asks it, the text's definition (at least 12 times in the past year)", "reason": "Table 1's counts are ALQ101's" }, { "choice": "activity intensity", "decision": "vigorous if any vigorous work or leisure activity, else moderate if any moderate, else mild (PAD200 and PAD320 in 2005-2006)", "reason": "reproduces Table 1's counts (1,462, 1,369, 1,351) exactly" }, { "choice": "diet covariates", "decision": "two-day means split at the sample mean (Table 1's cuts; fat at 68.69 g and moisture at 2,607.39 g) with missing as its own level", "reason": "Table 1 shows means as cuts and 621 unclear, the count missing either day; its fat row prints 73.01 but its counts are the 68.69 split, and its moisture row repeats the fat row; its sugar row has 907 unclear where we find 621" }, { "choice": "missing PIR and CRP", "decision": "the sample mean when 10% or less is missing, else split at the mean with missing as its own level", "reason": "the paper's stated rule" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The paper computes METS-IR from fasting glucose and triglycerides, but its flow chart (6,193 with METS-IR) and Table 1's FPG and fasting TG means (89.92 and 95.00, 123.47 and 136.72 mg/dL) are those of the biochemistry profile's non-fasting serum values (LBXSGL, LBXSTR), which alone reproduce Models 1, 2, and 3 (0.544, 0.493, 0.298 against 0.54, 0.49, 0.29). Fasting values give 0.172 on the same cycles." }, { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The paper says it used NHANES 2005-2010 and 2015-2018, but no 2015-2018 participant has a METS-IR in its analysis: the flow chart's 6,193 with METS-IR and 1,990 under 20 are exactly 2005-2010's. Serum values on all five cycles give 6,699 participants and 0.228." }, { "kind": "model", "affects_headline": true, "followed": true, "detail": "Table 2's note says Model 3 adjusts for all the covariates in Table 1, which include BMI, HDL, FPG, and fasting TG, the components of METS-IR, but the published 0.29 (0.14-0.44) is reproduced only without them (0.298, 0.146-0.450); with them Model 3 gives 2.40." }, { "kind": "weighting", "affects_headline": true, "followed": true, "detail": "The Methods say 2-year sample weights were used, but Models 1 and 2 are reproduced to every printed digit, intervals included, only unweighted (0.544, 0.412-0.677 and 0.493, 0.360-0.625 against 0.54, 0.41-0.68 and 0.49, 0.36-0.63)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The paper says 2005-2008 ferritin was converted to the Elecsys 170 scale with the Deming equation, but Table 1's mean ferritin (54.04 ng/ml) and the groups split at it (2,766 and 1,416) match the unconverted values." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods give the race groups as Mexican, White, Black, and other races, but Table 1's 'Mexican American' count (1,322) is Mexican American and other Hispanic participants together, and Model 3 adjusts for all the covariates in Table 1." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's fat row prints its cut as 73.01 g, but its counts (1,961 below) are the split at 68.69 g, the sample's mean fat intake; the moisture row prints 68.69 g with the fat row's counts, so the moisture cut is not printed." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The flow chart starts from 84,724 participants and excludes 67,020 without ferritin, but the five cycles it names hold 50,259 participants; the 17,704 with ferritin it leaves are reproduced exactly." } ], "projected_power": 0.534623717403654 }, "row100": { "id": "row100", "row": 100, "doi": "10.1186/s40001-023-01205-4", "measure": "OR", "contrast": "per unit of WWI (cm per square root of kg)", "published": { "measure": "OR", "estimate": 3.44, "low": 3.09, "high": 3.82, "n": 6587, "events": 2874, "contrast": "per unit of WWI (cm per square root of kg)" }, "published_b": [1.23547147138531, 1.12817109090965, 1.34025042261848], "constants": null, "original": { "estimate": 3.49700501943481, "ci": [3.16061726413789, 3.86919487047981], "b": 1.25190689343474, "se": 0.0516027584980858, "df": null, "p": 5.12771745153655e-130, "n": 5756, "events": 2544, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 3.45195815462157, "ci": [3.12260907871649, 3.81604446822281], "b": 1.2389416510798, "se": 0.0511605092376722, "df": null, "p": 1.48181607220974e-129, "n": 5773, "events": 2548, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.01330295553568, "published_z": 22.8356018319261, "interval_ratio_original": 1.0484490763943, "ratio_harmonized": 0.989643604949428, "variants": [ { "label": "weighted, as the paper says", "estimate": 4.31062005762337, "ci": [3.53911948015145, 5.25030177291154], "b": 1.46108175864591, "se": 0.0957514471368482, "df": 25, "p": 3.55980307471211e-14, "n": 5756, "events": 2544, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, { "label": "crude, unweighted (published 2.53, 2.37-2.71)", "estimate": 2.51302215566755, "ci": [2.33556070540218, 2.70396754846433], "b": 0.921486074944217, "se": 0.0373650140689193, "df": null, "p": 2.75422021121655e-134, "n": 5756, "events": 2544, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, weighted", "estimate": 2.95306700764605, "ci": [2.56447986359258, 3.40053547522533], "b": 1.08284429388465, "se": 0.0685049657660021, "df": 25, "p": 1.59660005822493e-14, "n": 5756, "events": 2544, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, { "label": "age, sex, race, unweighted (published 3.28, 3.01-3.56)", "estimate": 3.16902908144248, "ci": [2.89788385364572, 3.46554445458322], "b": 1.1534252575182, "se": 0.0456357837365654, "df": null, "p": 6.08092236917463e-141, "n": 5756, "events": 2544, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "age, sex, race, weighted", "estimate": 4.02562058748128, "ci": [3.36401222201148, 4.81734905964851], "b": 1.39267908221068, "se": 0.0871771527113911, "df": 25, "p": 1.25335471285062e-14, "n": 5756, "events": 2544, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 } ], "left_out": ["serum ferritin (measured in 2021-2023 only in children 1-5 and females 12-49)", "blood transfusion (not asked in 2021-2023)"], "choices": [ { "choice": "weights", "decision": "unweighted", "reason": "only unweighted fits reproduce the published crude, Model 2, and Model 3 estimates; the text names WTMECPRP" }, { "choice": "alcohol intake for the heavy-drinking exclusion", "decision": "mean of the two 24-hour recalls, or the first day's where there is one", "reason": "unstated; the dietary covariates use two-day means" }, { "choice": "race groups", "decision": "RIDRETH1: non-Hispanic White, non-Hispanic Black, all others", "reason": "Table 1 shows White, Black, Other" }, { "choice": "education", "decision": "high school or less versus more, with missing as its own level", "reason": "Table 1 categories, with 'Unclear'" }, { "choice": "income to poverty ratio", "decision": "Table 1 categories (<1.35, 1.35-3.45, >3.45) with missing as its own level", "reason": "Table 1" }, { "choice": "hypertension and diabetes", "decision": "self-reported diagnosis (BPQ020, DIQ010), with missing or borderline as their own level", "reason": "unstated; Table 1 shows yes, no, unclear" }, { "choice": "dietary covariates", "decision": "two-day means split at Table 1's cutpoints, with missing as its own level", "reason": "Table 1" }, { "choice": "activity intensity", "decision": "sedentary minutes (PAD680), continuous", "reason": "the covariate text says sedentary; Table 1's means fit PAD680" }, { "choice": "missing CRP, ferritin, or sedentary minutes", "decision": "complete case", "reason": "unstated" } ], "departures": [ { "kind": "weighting", "affects_headline": true, "followed": true, "detail": "The Methods say the examination weight WTMECPRP was used, but Table 2's Model 1, Model 2, and Model 3 estimates are reproduced only unweighted (Model 3: 3.50, 3.16-3.87 unweighted and 4.31, 3.54-5.25 weighted, against 3.44, 3.09-3.82; Model 1: 2.51 and 2.95 against 2.53, 2.37-2.71)." } ], "projected_power": 1 }, "row101": { "id": "row101", "row": 101, "doi": "10.1186/s40001-022-00977-5", "measure": "beta", "contrast": "per 1 ug/L of blood manganese (liver stiffness in kPa)", "published": { "measure": "beta", "estimate": 0.08, "low": 0.03, "high": 0.12, "n": 4690, "events": null, "contrast": "per 1 ug/L of blood manganese (liver stiffness in kPa)" }, "published_b": [0.08, 0.03, 0.12], "constants": null, "original": { "estimate": 0.0881365805186167, "ci": [-0.0241943232933169, 0.20046748433055], "b": 0.0881365805186167, "se": 0.0527016480619683, "df": 15, "p": 0.115174727640297, "n": 4690, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 30, "strata": 15 }, "harmonized": { "estimate": 0.0902331390712927, "ci": [-0.0240086063616989, 0.204474884504284], "b": 0.0902331390712927, "se": 0.053598146703017, "df": 15, "p": 0.112966266007682, "n": 4690, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 30, "strata": 15 }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.10170725648271, "published_z": 3.4843804169601, "interval_ratio_original": 0.400602135947733, "ratio_harmonized": 1.02378760941642, "variants": [ { "label": "unweighted", "estimate": 0.0726152184037017, "ci": [0.0312270263620146, 0.114003410445389], "b": 0.0726152184037017, "se": 0.0211168125374506, "df": null, "p": 0.000584429215699711, "n": 4690, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "interview weight (WTINT2YR)", "estimate": 0.0899801971915871, "ci": [-0.02116135503306, 0.201121749416234], "b": 0.0899801971915871, "se": 0.0521436467760527, "df": 15, "p": 0.104950072899968, "n": 4690, "events": null, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 30, "strata": 15 }, { "label": "crude, weighted (published 0.11, 0.07-0.15)", "estimate": 0.104529184890508, "ci": [-0.00926997358894244, 0.218328343369958], "b": 0.104529184890508, "se": 0.0533905007118353, "df": 15, "p": 0.0691202308806124, "n": 4690, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 30, "strata": 15 }, { "label": "crude, unweighted", "estimate": 0.0427456606268414, "ci": [0.00236716374093045, 0.0831241575127524], "b": 0.0427456606268414, "se": 0.020601652481582, "df": null, "p": 0.0379989900312444, "n": 4690, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, interview weight", "estimate": 0.107844760312803, "ci": [-0.00492753651131601, 0.220617057136922], "b": 0.107844760312803, "se": 0.0529087338985085, "df": 15, "p": 0.0595515242516667, "n": 4690, "events": null, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 30, "strata": 15 }, { "label": "age, sex, race, weighted (published 0.15, 0.11-0.19)", "estimate": 0.145416832441427, "ci": [0.020445240463912, 0.270388424418942], "b": 0.145416832441427, "se": 0.0586322074748872, "df": 15, "p": 0.0254839278883155, "n": 4690, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 30, "strata": 15 }, { "label": "age, sex, race, unweighted", "estimate": 0.103710796431686, "ci": [0.0610109758359562, 0.146410617027416], "b": 0.103710796431686, "se": 0.02178602307621, "df": null, "p": 1.93182748321584e-06, "n": 4690, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "age, sex, race, interview weight", "estimate": 0.148671120317395, "ci": [0.0251764236537618, 0.272165816981028], "b": 0.148671120317395, "se": 0.0579393009423547, "df": 15, "p": 0.0215052701270976, "n": 4690, "events": null, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 30, "strata": 15 }, { "label": "without COPD", "estimate": 0.0881911210485256, "ci": [-0.0232332099086907, 0.199615452005742], "b": 0.0881911210485256, "se": 0.052276316457659, "df": 15, "p": 0.112279020887382, "n": 4690, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 30, "strata": 15 }, { "label": "age and BMI in Table 1's groups", "estimate": 0.114156059490504, "ci": [-0.00040065138206713, 0.228712770363075], "b": 0.114156059490504, "se": 0.053745917237973, "df": 15, "p": 0.0507091544675362, "n": 4690, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 30, "strata": 15 } ], "left_out": ["diet status ('How healthy is the diet?', DBQ700): not asked in 2021-2023, so the harmonized model drops it and no longer needs it answered", "COPD alone (MCQ160O): 2021-2023 asks one question about COPD, emphysema, or chronic bronchitis together (MCQ160P), so the harmonized version adjusts for that combined condition (in 2017-2018, a yes to MCQ160O, MCQ160G, or MCQ160K), and in 2021-2023 its population is those who answered MCQ160P yes or no"], "choices": [ { "choice": "population", "decision": "adults 20 and older who answered the COPD question yes or no, with liver stiffness (LUXSMED) and blood manganese", "reason": "Table 1's groups (COPD 223, non-COPD 4,467) are exactly MCQ160O yes and no, and these steps give the published 4,690 with Table 1's and Table 2's distributions; the flow chart's middle steps differ (697 without elastography where 699 lack stiffness here; 7 without diet status and 2 without smoking, where no one here lacks either), so its 7 'diet status' exclusions are probably the 7 who answered the COPD question 'don't know'" }, { "choice": "weights", "decision": "examination weight (WTMEC2YR; the phlebotomy weight WTPH2YR from PBCD_L in 2021-2023)", "reason": "the paper says only '2-year sampling weights'; NCHS directs the examination weight for exam and laboratory data. The interview weight (variant) matches the crude and age-sex-race estimates slightly better (0.108 and 0.149 against 0.105 and 0.145), so the paper may have used it; both reproduce the headline, and the unweighted crude and Model 2 estimates (0.04, 0.10) do not" }, { "choice": "race", "decision": "Table 1's three groups: RIDRETH1 1-3 as 'White', 4 'Black', 5 'Other race'", "reason": "Table 1's percentages are reproduced only with Mexican American and other Hispanic counted in 'White'" }, { "choice": "alcohol", "decision": "drinking more than twice a week (ALQ121 every day, nearly every day, or 3-4 times a week) versus less, with no answer (including lifetime abstainers, who skip ALQ121) as 'unclear'", "reason": "the paper's definition; these codes reproduce Table 1's yes, no, and unclear counts exactly" }, { "choice": "income to poverty ratio", "decision": "Table 1's groups, cut at 1.39 and 3.49 (each cut point in the higher group), missing as 'unclear'", "reason": "Table 1; those cuts reproduce its percentages exactly (PIR is missing for 12.7%, over the paper's 12% limit for mean imputation)" }, { "choice": "age and BMI", "decision": "continuous; BMI's missing values (40, under 1%) take the sample mean", "reason": "the model's form is unstated and Table 1 shows both; continuous is the convention; the paper mean-imputes continuous covariates missing for under 12%, and Table 1's BMI SDs fit imputed values" }, { "choice": "categorical covariates", "decision": "smoking (SMQ020) yes or no; diet status (DBQ700) in five levels; prescription medicine (RXDUSE), diabetes (DIQ010, with borderline), heart failure (MCQ160B), and liver condition (MCQ160L) with refused or don't know as 'unclear'; COPD (MCQ160O) yes or no", "reason": "the paper's definitions and Table 1's levels, which these reproduce exactly" }, { "choice": "COPD in the 'Total' model", "decision": "included", "reason": "the Methods list COPD among the covariates; leaving it out (variant) changes the estimate by under 0.001" }, { "choice": "manganese outliers", "decision": "kept", "reason": "the paper limits manganese to 0-35 ug/L only for its curve fitting, not for Table 3" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "Table 1 labels its race groups White, Black, and Other race, and Model 3 adjusts for all the covariates in Table 1, but Table 1's race percentages (White 70.85% with COPD, 57.09% without) are reproduced only with Mexican American and other Hispanic participants counted as White." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's smoker 'Yes' and 'No' labels are swapped: the shares who smoked at least 100 cigarettes (SMQ020) are those printed as 'No' (85.20% with COPD, 40.0% without)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The flow chart excludes 697 participants without elastography, 7 without diet status, and 2 without smoking, but in the data 699 lack liver stiffness and none lacks diet status or smoking; the published 4,690 (223 with COPD, 4,467 without) is reached by keeping those who answered the COPD question yes or no." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 2 prints the share of participants without COPD whose LDL is under 120 mg/dL as 209.46%; the column's other two shares (17.62% and 52.92%) leave 29.46%." } ], "projected_power": 0.902184446966579 }, "row104": { "id": "row104", "row": 104, "doi": "10.1186/s12876-024-03394-6", "measure": "OR", "contrast": "per unit of NHR (neutrophils, 10^3 cells/uL, over HDL cholesterol, mmol/L)", "published": { "measure": "OR", "estimate": 1.2, "low": 1.09, "high": 1.31, "n": 4761, "events": 2123, "contrast": "per unit of NHR (neutrophils, 10^3 cells/uL, over HDL cholesterol, mmol/L)" }, "published_b": [0.182321556793955, 0.0861776962410524, 0.27002713721306], "constants": null, "original": { "estimate": 1.10077493725388, "ci": [1.01675617623123, 1.19173651541282], "b": 0.0960144201814634, "se": 0.0385509073403571, "df": 25, "p": 0.0197559025027104, "n": 5162, "events": 2292, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, "harmonized": { "estimate": 1.10077493725388, "ci": [1.01675617623123, 1.19173651541282], "b": 0.0960144201814634, "se": 0.0385509073403571, "df": 25, "p": 0.0197559025027104, "n": 5162, "events": 2292, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.526621327010559, "published_z": 3.8873513352247, "interval_ratio_original": 1.15778465138702, "ratio_harmonized": 1, "variants": [ { "label": "unweighted", "estimate": 1.10731139923782, "ci": [1.05951822131166, 1.15726045123045], "b": 0.101934914293535, "se": 0.022510932871361, "df": null, "p": 5.94769104006327e-06, "n": 5162, "events": 2292, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, weighted (published 1.48, 1.38-1.56)", "estimate": 1.3777865243993, "ci": [1.3003423496648, 1.45984302311283], "b": 0.320478243608954, "se": 0.0280891405686812, "df": 25, "p": 2.10276794425222e-11, "n": 5162, "events": 2292, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, { "label": "crude, unweighted", "estimate": 1.353114703359, "ci": [1.30600303905392, 1.4019258345468], "b": 0.30240912265293, "se": 0.0180808244881609, "df": null, "p": 8.55915529624879e-63, "n": 5162, "events": 2292, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2, weighted (published 1.49, 1.37-1.62)", "estimate": 1.36850381176061, "ci": [1.27967754956668, 1.4634957716008], "b": 0.313718034851393, "se": 0.0325849219801892, "df": 25, "p": 6.8567412961596e-10, "n": 5162, "events": 2292, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, { "label": "Model 2, unweighted", "estimate": 1.34661934618465, "ci": [1.29683847767168, 1.39831111949616], "b": 0.29759726371577, "se": 0.019218670224476, "df": null, "p": 4.39427320396337e-54, "n": 5162, "events": 2292, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": [], "choices": [ { "choice": "weights", "decision": "examination weight (WTMECPRP; WTPH2YR in 2021-2023)", "reason": "the paper says it weighted as NCHS directs; weighted and unweighted fits both reproduce the headline" }, { "choice": "age", "decision": "20 and older", "reason": "Results describe participants aged over 20; no age step is listed" }, { "choice": "alcohol grams a day", "decision": "questionnaire: drinks per drinking day times drinking days a year over 365, 14 g a drink", "reason": "unstated" }, { "choice": "elastography completeness", "decision": "exam status complete with CAP and stiffness recorded", "reason": "unstated" }, { "choice": "MASLD cardiometabolic criteria", "decision": "2023 multisociety criteria with missing fasting values counted as not met", "reason": "fasting glucose and triglycerides exist only for the fasting subsample" }, { "choice": "hypertension", "decision": "told, taking medication, or mean blood pressure 140/90 or more", "reason": "unstated" }, { "choice": "diabetes grade", "decision": "diabetes (told, medication, HbA1c 6.5% or more, or fasting glucose 126 or more); prediabetes (told, HbA1c 5.7% or more, or fasting glucose 100 or more); otherwise normal", "reason": "unstated" }, { "choice": "covariate coding", "decision": "age, BMI, sedentary hours, cholesterol, ALT, HbA1c, albumin, creatinine, uric acid continuous; income in the paper's three groups", "reason": "unstated; the paper lists them as continuous measures" } ], "departures": [ { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The Results describe Model 3's estimate for the fourth NHR quartile against the first (OR 3.00, 2.10-4.30) as 'each unit increase in the NHR ratio was associated with a 3-fold increase in the odds of MASLD prevalence', but Table 2's Model 3 estimate per unit of NHR is 1.20." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The Results give the NHR quartiles as 0.191-2.065, 2.067-2.993, 3.000-4.207, and 4.211-19.028, but Tables 2 and 3 label them 0.231-2.022, 2.023-2.948, 2.951-4.172, and 4.173-16.769." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1 has cells that contradict their own columns: the first quartile's prediabetes count is 1,110 where the column needs 110, the female total is 2,497 where its columns sum to 2,494, and the third quartile's MASLD shares (50.75% and 59.25%) and BMI shares (19.66%, 20.11%, 50.23%) sum to 110% and 90%." } ], "projected_power": 0.81904881885987 }, "row106": { "id": "row106", "row": 106, "doi": "10.1265/ehpm.24-00050", "measure": "OR", "contrast": "per unit of ln blood cadmium (ug/L)", "published": { "measure": "OR", "estimate": 1.33, "low": 1.21, "high": 1.45, "n": 10868, "events": 1173, "contrast": "per unit of ln blood cadmium (ug/L)" }, "published_b": [0.285178942233662, 0.19062035960865, 0.371563556432483], "constants": null, "original": { "estimate": 1.29519575102517, "ci": [1.18159831519014, 1.41971430722944], "b": 0.258661842820911, "se": 0.0468344404195855, "df": null, "p": 3.33453112783549e-08, "n": 10868, "events": 1173, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 1.29731090994619, "ci": [1.18368250246972, 1.42184715373746], "b": 0.260293591292132, "se": 0.0467678223741132, "df": null, "p": 2.61168652631561e-08, "n": 10868, "events": 1173, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.907015927595997, "published_z": 6.17807649846479, "interval_ratio_original": 0.985595783054123, "ratio_harmonized": 1.00630842359053, "variants": [ { "label": "weighted (WTMEC2YR in every cycle)", "estimate": 1.31566967892897, "ci": [1.18527705098659, 1.46040683282632], "b": 0.274345797543018, "se": 0.0525651069037222, "df": 94, "p": 1.07083049688854e-06, "n": 10868, "events": 1173, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "weighted as NCHS directs (WTSH2YR in 2013-2016)", "estimate": 1.29939886481679, "ci": [1.16530640913319, 1.4489214138478], "b": 0.26190174584285, "se": 0.054855902971825, "df": 94, "p": 6.60939465988987e-06, "n": 10868, "events": 1173, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "crude (published 1.56, 1.45-1.69)", "estimate": 1.56478688497373, "ci": [1.45102004524083, 1.68747358343997], "b": 0.447749638983721, "se": 0.0385123660413671, "df": null, "p": 3.03562379428431e-31, "n": 10868, "events": 1173, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, weighted (WTMEC2YR)", "estimate": 1.6178408737309, "ci": [1.46470860510809, 1.78698280571732], "b": 0.481092466291153, "se": 0.0500805711343376, "df": 94, "p": 1.26239109100522e-15, "n": 10868, "events": 1173, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "age, race (published 1.67, 1.55-1.81)", "estimate": 1.67225139813865, "ci": [1.54864349483447, 1.80572529953099], "b": 0.514170861098348, "se": 0.0391800454375569, "df": null, "p": 2.42207348701568e-39, "n": 10868, "events": 1173, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "age, race, weighted (WTMEC2YR)", "estimate": 1.68795941242348, "ci": [1.52533317204938, 1.86792435265859], "b": 0.523520351111864, "se": 0.0510230649022711, "df": 94, "p": 5.14445622146781e-17, "n": 10868, "events": 1173, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "BMI continuous, as Table 1 summarizes it", "estimate": 1.31810997881275, "ci": [1.20180855036167, 1.4456661302026], "b": 0.276198876311335, "se": 0.0471290949824194, "df": null, "p": 4.61545412423571e-09, "n": 10868, "events": 1173, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "diabetes yes or no (borderline as no)", "estimate": 1.29383773468261, "ci": [1.18044745033816, 1.41811995375912], "b": 0.257612789979686, "se": 0.0467963844744521, "df": null, "p": 3.69226779981518e-08, "n": 10868, "events": 1173, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": "marital status as married against everything else: 2021-2023's DMDMARTZ puts living with a partner together with married, so the harmonized version counts married or living with a partner as yes", "choices": [ { "choice": "weights", "decision": "unweighted", "reason": "the paper names no weight and says it accounted for the design, but only unweighted fits reproduce its crude (1.56, 1.45-1.69) and age-race (1.67, 1.55-1.81) estimates to the second decimal, CIs included; Table 1's shares are unweighted; weighted fits are kept as variants, with the examination weight and with the blood metal subsample weight NCHS directs for 2013-2016" }, { "choice": "missing depression data", "decision": "PHQ-9 total missing if any of the nine items is missing, refused, or don't know", "reason": "reproduces the flow chart exactly: 29,745 excluded for depression, then 6,265 for metals, 12,309 men, and 723 under 20" }, { "choice": "blood metals", "decision": "cadmium, lead, and mercury all measured (LBXBCD, LBXBPB, LBXTHG); values below detection as NHANES releases them (detection limit over the square root of 2)", "reason": "the flow chart excludes those missing any of the three; detection limits are not discussed" }, { "choice": "exposure scale", "decision": "natural log of blood cadmium (ug/L), per unit", "reason": "Table 2's title and the Results; the abstract's 'each incremental unit of blood cadmium' is per unit of ln cadmium" }, { "choice": "pregnancy", "decision": "pregnant women kept", "reason": "no pregnancy exclusion is listed, and the flow chart and n = 10,868 are reproduced without one" }, { "choice": "missing covariates", "decision": "complete case", "reason": "1,026 excluded for missing covariates, reproduced exactly (most for income to poverty ratio)" }, { "choice": "BMI", "decision": "three groups (under 25, 25 to 29.9, 30 or more), as the covariate text says it was stratified", "reason": "stated; Table 1 summarizes BMI by its mean, which says nothing of how the model held it, and nothing in the paper's numbers shows the model held it otherwise (continuous BMI is a variant)" }, { "choice": "income to poverty ratio", "decision": "continuous", "reason": "Table 1 reports it as a mean; no categories are given" }, { "choice": "education", "decision": "DMDEDUC2 in three levels: less than high school (1-2), high school or GED (3), above high school (4-5)", "reason": "Table 1's categories; its shares (23.77, 21.83, 54.41%) are reproduced exactly" }, { "choice": "smoking", "decision": "smoked 100 cigarettes in life (SMQ020) yes or no", "reason": "the covariate text; Table 1's 37.15% yes is reproduced exactly" }, { "choice": "marital status", "decision": "married (DMDMARTL 1) against widowed, divorced, separated, never married, or living with a partner", "reason": "unstated mapping; Table 1's 47.52% yes is reproduced exactly by married alone (married or living with a partner gives 54.91%)" }, { "choice": "hypertension", "decision": "told by a doctor (BPQ020) yes or no", "reason": "the covariate text: self-reported physician diagnosis; Table 1's 35.48% is reproduced exactly" }, { "choice": "diabetes", "decision": "DIQ010 in three levels: yes, no, borderline", "reason": "Table 1 and Table 3 show three levels (text says yes or no); the two-level coding is a variant" }, { "choice": "race", "decision": "RIDRETH1 in five groups", "reason": "the covariate text; Table 1's shares are reproduced exactly" } ], "departures": [ { "kind": "weighting", "affects_headline": true, "followed": true, "detail": "The Methods say the analyses adjusted for the complex multistage cluster survey design, but Table 2's Model 1 (1.56, 1.45-1.69) and Model 2 (1.67, 1.55-1.81) are reproduced to every printed digit only unweighted (Model 1: 1.565, 1.451-1.687; with the examination weight 1.618, 1.465-1.787)." }, { "kind": "reporting", "affects_headline": true, "followed": true, "detail": "The abstract reports OR 1.33 (1.21-1.45) for each incremental unit of blood cadmium, but Table 2 and the Results give it per unit of ln-transformed blood cadmium, the scale on which it is reproduced (1.318, 1.202-1.446)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text defines diabetes as a physician's diagnosis, yes or no, but Tables 1 and 3 show it in three levels (yes, no, borderline); the two codings give nearly the same estimate here (1.318 and 1.317)." } ], "projected_power": 0.654928092706238 }, "row111": { "id": "row111", "row": 111, "doi": "10.1093/ajh/hpw010", "measure": "OR", "contrast": "a usual source of care (any place other than the emergency room) vs none or the emergency room", "published": { "measure": "OR", "estimate": 3.89, "low": 2.15, "high": 6.98, "n": null, "events": null, "contrast": "a usual source of care (any place other than the emergency room) vs none or the emergency room" }, "published_b": [1.35840915763035, 0.765467842139571, 1.94304891677428], "constants": null, "original": { "estimate": 3.40568890278204, "ci": [1.98349776012563, 5.84760776427502], "b": 1.22544724007187, "se": 0.269004811746209, "df": 49, "p": 3.48514106429398e-05, "n": 4368, "events": 2315, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 94, "strata": 45 }, "harmonized": { "estimate": 3.64362115668325, "ci": [2.23205583368908, 5.9478687464048], "b": 1.29297801040455, "se": 0.243859967046943, "df": 49, "p": 2.72384276359068e-06, "n": 4733, "events": 2397, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 94, "strata": 45 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.902119389573004, "published_z": 4.52186789100833, "interval_ratio_original": 1.08917208681299, "ratio_harmonized": 1.05510704020902, "variants": [ { "label": "missing pressure counted hypertensive, uncontrolled (Stata)", "estimate": 3.05245056119698, "ci": [1.81492815407051, 5.13378692575495], "b": 1.1159447307035, "se": 0.25871081544925, "df": 49, "p": 7.75851632192072e-05, "n": 4944, "events": 2315, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 94, "strata": 45 }, { "label": "Table 2's covariates only", "estimate": 3.5589233536534, "ci": [2.28937094967306, 5.53249591945263], "b": 1.2694580703741, "se": 0.219539421494636, "df": 49, "p": 5.05621671136818e-07, "n": 5163, "events": 2711, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 94, "strata": 45 }, { "label": "Table 2's covariates only, Stata missing pressure", "estimate": 3.19575376093829, "ci": [2.0559816353113, 4.96738002185746], "b": 1.16182297892153, "se": 0.219483976245733, "df": 49, "p": 2.80753750304688e-06, "n": 5961, "events": 2711, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 94, "strata": 45 }, { "label": "unweighted", "estimate": 2.93685823278891, "ci": [2.18838760440293, 3.94132020403817], "b": 1.07734038165717, "se": 0.150092228742123, "df": null, "p": 7.08131646281712e-13, "n": 4368, "events": 2315, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "interview weight", "estimate": 3.39897161854446, "ci": [1.98051544532116, 5.83333398937331], "b": 1.22347292073633, "se": 0.268771118398432, "df": 49, "p": 3.52473009946193e-05, "n": 4368, "events": 2315, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 94, "strata": 45 }, { "label": "COPD added (Table 1)", "estimate": 3.40573435272271, "ci": [1.98342265300425, 5.84798528127464], "b": 1.22546058528299, "se": 0.269030295701563, "df": 49, "p": 3.48958613249776e-05, "n": 4368, "events": 2315, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 94, "strata": 45 }, { "label": "without physical activity, paper's sample", "estimate": 3.38606291227369, "ci": [1.98239755355836, 5.7836138998937], "b": 1.21966786365461, "se": 0.266404987458017, "df": 49, "p": 3.22998087079365e-05, "n": 4368, "events": 2315, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 94, "strata": 45 } ], "left_out": ["meeting the physical activity guideline over work, transport, and leisure (GPAQ, 2007-2012): 2021-2023 asks about leisure time only", "the paper's complete-case drop of adults who never had their cholesterol checked (BPQ060 = 2, who were not asked BPQ080, so hyperlipidemia was missing): 2021-2023 asks everyone whether they were told of high cholesterol and has no BPQ060, so the harmonized version counts the never checked as not told and keeps them"], "choices": [ { "choice": "model covariates", "decision": "every variable the Measures section defines: age group, race and ethnicity, sex, married, insured, education, income, self-rated health, physical activity, smoking, diabetes, heart failure, prior heart attack, prior stroke, high cholesterol, creatinine, BMI group", "reason": "Table 2 shows nine covariates and says not all are shown; with all the Measures variables the shown odds ratios match Table 2's (age 18-34 0.39 vs 0.38, over 74 0.60 vs 0.58, 35-44 0.85 vs 0.84, Black 0.69 vs 0.70, male 0.73 vs 0.76, heart failure 1.28 vs 1.30, diabetes 1.04 vs 1.02, high cholesterol 1.53 vs 1.55, obese 2.05 vs 2.03), better than with the shown ones alone (over 74 0.66, 35-44 0.95, Black 0.79, heart failure 1.40, diabetes 1.22)" }, { "choice": "population and missing blood pressure", "decision": "adults 18 and older taking prescribed blood pressure medicine or with mean pressure 140/90 or more; control is undefined without a reading, so those without one leave the model", "reason": "the paper's Table 1 counts (n = 7,653; by age group, usual source, and treatment within 1%) are reproduced only if adults without a pressure reading (mostly those interviewed but not examined, hence its 10% missing BMI) are counted hypertensive, the way Stata compares missing values; its Table 1 control shares (54.8% and 17%) and Table 2 odds ratios are those without them, so control was missing for them (a variant counts them as uncontrolled)" }, { "choice": "blood pressure readings", "decision": "mean of the readings after the first (2-4; diastolic zeros left out, as the library does)", "reason": "the paper averages after discarding the first reading" }, { "choice": "usual source of care", "decision": "no place (HUQ030 = 2) or the emergency room (HUQ040 = 3) is none; any other place is one; a place of unknown kind is missing", "reason": "the paper's definition; 'some other place' and more than one place count as a usual source, since only none and the emergency room are named as none" }, { "choice": "2021-2023 place types", "decision": "urgent care or retail clinic, VA, some other place, and no one place most often count as a usual source; the emergency room as none", "reason": "only the emergency room and no place are none in the paper; urgent care and retail clinics would have been reported as a clinic or health center in 2007-2012, and 'no one place most often' is said by those with more than one place" }, { "choice": "weights", "decision": "examination weight over the three cycles (WTMEC2YR/3; the phlebotomy weight in 2021-2023)", "reason": "pressure and BMI come from the examination; Table 1's weighted shares (90.7% with a usual source, 54.8% and 17% controlled, 70.7% and 20.1% treated) match the examination weight" }, { "choice": "categorical coding", "decision": "married or living with a partner vs not; insured vs not (HIQ011); diabetes told (borderline as no); income <150%, 150-249%, 250-349%, 350% or more of poverty; smoking never, former, current; self-rated health in its five answers; creatinine continuous", "reason": "the paper's definitions where it gives them; smoking in Table 1's three groups; Table 1's diabetes counts match DIQ010 = 1" }, { "choice": "physical activity", "decision": "30 minutes on 5 days of moderate or 20 minutes on 3 days of vigorous activity in any GPAQ domain", "reason": "the ACSM guideline the paper cites, which gives Table 1's counts" }, { "choice": "missing covariates", "decision": "complete case", "reason": "Model 1 is the model without imputation" } ], "departures": [ { "kind": "sample", "affects_headline": false, "followed": null, "detail": "The Methods restrict the sample to adults taking blood pressure medication or with a systolic pressure of 140 or diastolic of 90 at the examination, but Table 1's population (n = 7,653, with its counts by age group, usual source, and treatment within 1%) is reproduced only by also counting as hypertensive about 1,270 adults with no pressure reading. Table 1's control shares (54.8% and 17%) and Table 2's odds ratios leave them out (counted as uncontrolled, they give 3.05 for the headline, against 3.41 without them)." } ], "projected_power": 0.684786724040109 }, "row121": { "id": "row121", "row": 121, "doi": "10.1016/j.pmedr.2023.102306", "measure": "OR", "contrast": "per unit of ln(VAI); VAI from waist, BMI, and triglycerides and HDL cholesterol in mmol/L", "published": { "measure": "OR", "estimate": 1.23, "low": 1.02, "high": 1.48, "n": 6085, "events": null, "contrast": "per unit of ln(VAI); VAI from waist, BMI, and triglycerides and HDL cholesterol in mmol/L" }, "published_b": [0.207014169384326, 0.0198026272961797, 0.392042087776024], "constants": null, "original": { "estimate": 1.26777689759507, "ci": [0.989429400928989, 1.62442945455905], "b": 0.237264892265446, "se": 0.124009649394933, "df": 62, "p": 0.0603328907115609, "n": 2906, "events": 965, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 121, "strata": 59 }, "harmonized": { "estimate": 1.30802526200516, "ci": [1.01762862036886, 1.68129124102614], "b": 0.26851856630811, "se": 0.125586345577342, "df": 62, "p": 0.0364559587868236, "n": 2906, "events": 965, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 121, "strata": 59 }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.14612875520109, "published_z": 2.17999626240444, "interval_ratio_original": 0.750810484822273, "ratio_harmonized": 1.13172481501266, "variants": [ { "label": "unweighted", "estimate": 1.31717037663636, "ci": [1.14506337908312, 1.51514565287869], "b": 0.275485781619714, "se": 0.0714430440247419, "df": null, "p": 0.000115248227832529, "n": 2906, "events": 965, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "fasting subsample weight (WTSAF2YR), as NCHS directs for LDL-C", "estimate": 1.26116576812608, "ci": [0.978587637484264, 1.62534149601764], "b": 0.232036506016318, "se": 0.126905979860397, "df": 62, "p": 0.0722989623447054, "n": 2761, "events": 906, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 121, "strata": 59 }, { "label": "Model 3 without physical activity, as Table 2's note lists it", "estimate": 1.30802526200516, "ci": [1.01762862036886, 1.68129124102614], "b": 0.26851856630811, "se": 0.125586345577342, "df": 62, "p": 0.0364559587868236, "n": 2906, "events": 965, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 121, "strata": 59 }, { "label": "diabetes and hypertension as the Methods text defines them", "estimate": 1.27309960938306, "ci": [0.992502823422027, 1.63302569742123], "b": 0.24145456426196, "se": 0.124554043820791, "df": 62, "p": 0.057111580188387, "n": 2906, "events": 965, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 121, "strata": 59 }, { "label": "LDL-C from non-fasting lipids (Friedewald), keeping the full sample", "estimate": 1.22041891365384, "ci": [1.08143765867866, 1.37726138243036], "b": 0.199194171653955, "se": 0.0604825277697187, "df": 62, "p": 0.0016380128192585, "n": 5984, "events": 2014, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 121, "strata": 59 }, { "label": "Model 1, full sample (published 1.35, 1.20-1.52)", "estimate": 1.43147648507517, "ci": [1.29495416769513, 1.58239185481783], "b": 0.358706418648617, "se": 0.0501413426907565, "df": 62, "p": 1.16394766823101e-09, "n": 6159, "events": 2077, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 121, "strata": 59 }, { "label": "Model 1, full sample, unweighted", "estimate": 1.31794578882128, "ci": [1.22914550674591, 1.41316149531417], "b": 0.276074303839312, "se": 0.0355899834847318, "df": null, "p": 8.6907983134301e-15, "n": 6159, "events": 2077, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2, full sample (published 1.42, 1.26-1.61)", "estimate": 1.52766872045376, "ci": [1.37157100783669, 1.70153182454167], "b": 0.423742861252754, "se": 0.0539207569632524, "df": 62, "p": 6.96206396795434e-11, "n": 6159, "events": 2077, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 121, "strata": 59 }, { "label": "Model 2, full sample, unweighted", "estimate": 1.47876471518466, "ci": [1.36807257051027, 1.59841307399325], "b": 0.391207087357927, "se": 0.0396967605321788, "df": null, "p": 6.52901271138941e-23, "n": 6159, "events": 2077, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": "physical activity (high or low at 600 MET-minutes a week, from the GPAQ's work, transport, and leisure items; 2021-2023 asks only about leisure-time activity, so the score can't be built)", "choices": [ { "choice": "weights", "decision": "examination weight (WTMEC2YR pooled over four cycles; WTPH2YR in 2021-2023)", "reason": "Table 1 is survey-weighted over the examined sample and the text doesn't say the models were unweighted; the published Model 1 and 2 intervals are as wide as design-based ones (unweighted ones are about 40% narrower), and only the weighted Model 3 shows the published Q3 above Q4" }, { "choice": "LDL cholesterol and Model 3's sample", "decision": "NHANES's LDL-C (fasting subsample) with complete-case analysis, so Model 3 is fitted on the fasting subsample (2,906) while Models 1 and 2 (variants) use the full sample", "reason": "Table 1's LDL-C mean (2.8 mmol/L) is the fasting value's (2.83; 2.73 by Friedewald from non-fasting lipids); Model 3's published intervals are about 1.5 times as wide as Models 1 and 2's, as halving the sample would make them; Table 2's caption repeats the cohort's 6,085" }, { "choice": "triglycerides for VAI", "decision": "standard biochemistry profile (LBXSTR, non-fasting), in mmol/L", "reason": "only 696 of 6,781 lost for VAI (fasting values exist for about half); Table 1's triglyceride mean (1.7 mmol/L) is the serum value's (1.72, fasting 1.35); VAI's sample quartiles (1.07, 1.82, 3.06) and mean (2.49) match the paper's (1.1, 1.8, 3.0; 2.5)" }, { "choice": "renal function exclusion", "decision": "both serum creatinine and urine ACR required", "reason": "the paper excludes 'missing data on renal function, including serum creatinine and urine ACR'; this gives 6,159 against the paper's 6,085 (its intermediate 6,781 matches no definition tried)" }, { "choice": "CKD", "decision": "CKD-EPI 2009 creatinine equation with its race coefficient; ACR over 30 mg/g", "reason": "as stated (Levey et al., 2009, and 'ACR >30 mg/g')" }, { "choice": "diabetes", "decision": "HbA1c 6.5% or more, fasting glucose 126 mg/dL or more, told by a doctor, or taking insulin or pills", "reason": "Table 1's 26.5% matches this exactly; the Methods text's definition, without fasting glucose, gives 24.8% (variant)" }, { "choice": "hypertension", "decision": "taking blood pressure medicine or mean pressure 140/90 mmHg or more", "reason": "Table 1's 63.0% matches this (63.1%); the Methods text adds a self-reported diagnosis, which gives 66.8% (variant)" }, { "choice": "physical activity in Model 3", "decision": "included", "reason": "the Methods, the Results sentence reporting the headline, and the figure captions include it; only the Table 2 and 3 notes omit it (variant)" }, { "choice": "CVD history", "decision": "told of coronary heart disease, angina, heart attack, or stroke", "reason": "as stated; gives 19.7% against Table 1's 22.5% (21.3% with heart failure added), unresolved" }, { "choice": "cholesterol-lowering medicine", "decision": "taking prescribed cholesterol medicine (BPQ100D; BPQ101D in 2021-2023); those never asked count as not taking", "reason": "unstated item; gives Table 1's share (43.6% against 43.7%)" }, { "choice": "blood pressure", "decision": "mean of up to four auscultatory readings (oscillometric in 2021-2023), a diastolic 0 as missing", "reason": "number of readings unstated; SBP and DBP means match Table 1 (132.0 and 68.8 against 131.9 and 68.2)" }, { "choice": "covariate coding", "decision": "age, SBP, DBP, HbA1c, total and LDL cholesterol continuous (complete case); sex, race (five groups), education (three), smoking (never, former, current), activity, and the yes/no conditions as factors with missing as its own level", "reason": "the paper groups missing categorical values; Table 1's categories" } ], "departures": [ { "kind": "sample", "affects_headline": true, "followed": true, "detail": "Table 2's caption gives every model n = 6,085, but Model 3 adjusts for LDL-C, which NHANES measures only in the fasting subsample, and the paper's numbers fit that subsample: Table 1's LDL-C mean (2.8 mmol/L) is the fasting value's (2.83; 2.73 by Friedewald from non-fasting lipids), and Model 3's intervals are about 1.5 times as wide as Models 1 and 2's. Our Model 3, on that subsample, has n = 2,906." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods define diabetes as HbA1c of 6.5% or more, a self-reported diagnosis, or diabetes medication, but Table 1's 26.5% is reproduced exactly only with fasting glucose of 126 mg/dL or more added (24.8% as defined)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods define hypertension as mean pressure of 140/90 mmHg or more, antihypertensive medication, or a self-reported diagnosis, but Table 1's 63.0% is reproduced without the self-reported diagnosis (63.1%; 66.8% as defined)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 2 prints Model 2's Q3 odds ratio as 1.04 (0.81, 1.33), the same as its Q2, with P = 0.017, which that interval rules out." } ], "projected_power": 0.175380674333234 }, "row131": { "id": "row131", "row": 131, "doi": "10.3389/fnut.2022.936926", "measure": "OR", "contrast": "highest versus lowest tertile of fiber intake from foods and supplements, first 24-hour recall (tertile means 7.51 and 28.79 g/day)", "published": { "measure": "OR", "estimate": 0.64, "low": 0.46, "high": 0.91, "n": 8872, "events": null, "contrast": "highest versus lowest tertile of fiber intake from foods and supplements, first 24-hour recall (tertile means 7.51 and 28.79 g/day)" }, "published_b": [-0.446287102628419, -0.776528789498996, -0.0943106794712413], "constants": { "fiber": [11.3, 19.3], "fiber_food": [11.3, 19.2], "fiber_dsq": [11.3, 19.3] }, "original": { "estimate": 0.694567024143134, "ci": [0.543536510329319, 0.887563837679932], "b": -0.364466614398695, "se": 0.122658968309561, "df": 62, "p": 0.00421485624698168, "n": 15065, "events": 1048, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 121, "strata": 59 }, "harmonized": { "estimate": 0.773580636470219, "ci": [0.599020175212344, 0.999009759411753], "b": -0.256725365572271, "se": 0.127933105695989, "df": 62, "p": 0.0491458974039761, "n": 15071, "events": 1049, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 121, "strata": 59 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.816664009002635, "published_z": -2.56430204668957, "interval_ratio_original": 1.39119284302856, "ratio_harmonized": 0.704386507378245, "variants": [ { "label": "crude (published 0.64, 0.48-0.84)", "estimate": 0.678949478828278, "ci": [0.541829602998884, 0.85077004329374], "b": -0.387208559451918, "se": 0.112855611602613, "df": 62, "p": 0.00107546286638379, "n": 15065, "events": 1048, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 121, "strata": 59 }, { "label": "age, sex, race (published 0.56, 0.42-0.76)", "estimate": 0.622627271009326, "ci": [0.49129740771172, 0.789063228340892], "b": -0.473807220120013, "se": 0.118510140874187, "df": 62, "p": 0.000172510237798637, "n": 15065, "events": 1048, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 121, "strata": 59 }, { "label": "per g/day, Model 3 (published 0.98, 0.96-1.00)", "estimate": 0.985767350769326, "ci": [0.97248296839106, 0.999233201431261], "b": -0.0143349047872949, "se": 0.00678739639987248, "df": 62, "p": 0.038723576288783, "n": 15065, "events": 1048, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 121, "strata": 59 }, { "label": "examination weight", "estimate": 0.726782278448199, "ci": [0.572802652092975, 0.922154390061409], "b": -0.319128325740056, "se": 0.119104102047995, "df": 62, "p": 0.00943253375806688, "n": 15065, "events": 1048, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 121, "strata": 59 }, { "label": "unweighted", "estimate": 0.617672844165261, "ci": [0.502381191544159, 0.759422822431974], "b": -0.481796340068012, "se": 0.105409978603494, "df": null, "p": 4.86120837376069e-06, "n": 15065, "events": 1048, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "current smoking coded in every cycle", "estimate": 0.778486962749216, "ci": [0.605170494955086, 1.00144001768539], "b": -0.250403034477448, "se": 0.125985795288078, "df": 62, "p": 0.0512829593425513, "n": 15065, "events": 1048, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 121, "strata": 59 }, { "label": "current smoker yes or no, never smokers no", "estimate": 0.780504268362782, "ci": [0.60826943181406, 1.00150834658208], "b": -0.247815070287239, "se": 0.124725279141155, "df": 62, "p": 0.0513582291990093, "n": 15065, "events": 1048, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 121, "strata": 59 }, { "label": "fiber from foods only", "estimate": 0.751956711310577, "ci": [0.58766826153858, 0.962173615101549], "b": -0.285076521435784, "se": 0.123321491984252, "df": 62, "p": 0.0241336692885794, "n": 15065, "events": 1048, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 121, "strata": 59 }, { "label": "supplement fiber from the 30-day questionnaire", "estimate": 0.756800968995776, "ci": [0.590170348925266, 0.970478621496237], "b": -0.278654980859749, "se": 0.124408513970892, "df": 62, "p": 0.028696465600824, "n": 15065, "events": 1048, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 121, "strata": 59 } ], "left_out": ["vigorous work activity (PAQ605; 2021-2023 asks only leisure-time activity), as a covariate and as Figure 1's exclusion of those missing it", "sleep disorder (told a doctor of trouble sleeping, SLQ050, not asked in 2021-2023), as a covariate and as an exclusion", "fiber from supplements on the recall day (DS1TFIBE): 2021-2023 dropped the 24-hour supplement recall, so there the supplement part is the 30-day supplement questionnaire's average daily fiber (DSQTFIBE)", "antihypertensive and lipid-lowering medicine coded as the paper's cycles ask them: 2021-2023 asks everyone told of high blood pressure (BPQ150) and everyone (BPQ101D) whether they take one, so the harmonized model codes both cycles' answers that way"], "choices": [ { "choice": "headline", "decision": "tertile 3 versus tertile 1, Model 3 (Table 3)", "reason": "the study's rule: the main table's most-adjusted highest-versus-reference estimate, since the per-gram estimate's CI (0.96-1.00) does not exclude 1 and no p-value is given" }, { "choice": "weights", "decision": "dietary day-one weight (WTDRD1) over the four cycles", "reason": "the paper names only 'an appropriate NHANES sample weight'; NCHS directs the weight of the smallest subsample, the day-one recall; weighted univariate odds ratios match Table 2 (race, BMI, HbA1c) where unweighted ones do not" }, { "choice": "population", "decision": "18 and older with every key variable known, in Figure 1's steps; education, asked from 20, makes it 20 and older", "reason": "reproduces Figure 1's 20220 and 16223 exactly" }, { "choice": "vigorous activity exclusion", "decision": "drop only those missing PAQ605 (fewer than 10)", "reason": "Figure 1 drops 6590 for missing vigorous activity, which PAQ605 cannot produce; no NHANES variable's missingness fits the per-cycle losses Supplementary Figure S1 implies (keeping 71%, 53%, 58%, 54%), so the sample is 15065 against 8872; weighted descriptives still match Table 1 (tertile means, prevalence by tertile 6.34/5.50/4.39 against 6.50/5.45/4.25, sex, age, vigorous 23%, sleep 30%)" }, { "choice": "blood pressure", "decision": "mean of the first three readings, anyone missing one excluded, diastolic zeros kept as recorded", "reason": "the paper averages three readings; requiring all three reproduces Figure 1's 2008" }, { "choice": "glucose, triglycerides, cholesterol", "decision": "serum glucose and triglycerides from the biochemistry panel (LBXSGL, LBXSTR), total cholesterol LBXTC, HDL LBDHDD, HbA1c LBXGH", "reason": "unstated; the fasting values exist for half the sample, while these reproduce Figure 1's 996 missing and Table 1's means" }, { "choice": "fiber", "decision": "day-one food fiber (DR1TFIBE, reliable recalls) plus that day's supplement fiber (DS1TFIBE, none counted as 0)", "reason": "the paper sums food and supplements from the first 24-hour recall without naming the supplement file; the day-one supplement recall is that recall's" }, { "choice": "tertiles", "decision": "cutpoints at the unweighted thirds of the analytic sample (11.3 and 19.3 g/day), groups closed on the left", "reason": "cutpoints are unpublished; Table 1's unequal weighted shares (30.88/34.82/34.29) show unweighted tertiles, Table 2's sizes (2929/2959/2984) fall as left-closed groups split values tied at a cutpoint, and the weighted tertile means come out 7.51/15.07/29.23 against 7.51/14.90/28.79" }, { "choice": "outcome", "decision": "yes to heart attack or stroke (MCQ160E, MCQ160F); no to both; missing otherwise", "reason": "the paper's composite" }, { "choice": "current smoker", "decision": "yes if SMQ040 every day or some days, no if not at all, unknown where not asked (never smokers) or not answered; in the paper version for 2011-2012 only (unknown in the other cycles), as the paper's computation merged it, and in the harmonized version for every cycle", "reason": "Table 2's counts (546/583/7743) and odds ratios (no 1.32, unknown 0.86 against our 1.36, 0.89) are SMQ040 coded this way in 2011-2012 alone, so the paper merged smoking for one cycle only; a data-handling failure of the paper's own files, followed in the paper version and not in the harmonized one, as the plan sets; the paper version with smoking coded in every cycle is a variant" }, { "choice": "medicines and aspirin", "decision": "DIQ070, BPQ050A, BPQ100D, and RXQ520 (taking low-dose aspirin on one's own) as yes, no, or unknown where not asked", "reason": "Table 2's counts fit these items, aspirin's 190/3437 fitting RXQ520 and not RXQ515" }, { "choice": "conditions", "decision": "diabetes DIQ010 in three levels; hypertension BPQ020; hypercholesterolemia BPQ080; sleep disorder SLQ050 (told a doctor of trouble sleeping); vigorous activity PAQ605 (work)", "reason": "the paper's categories; SLQ060 was asked only through 2014, and Table 1's 30% fits SLQ050; the paper's definition of vigorous activity is PAQ605's wording" }, { "choice": "other covariates", "decision": "race from RIDRETH1 with other Hispanic among other races; marital status in three groups; education in three; PIR under 1.2 or not; BMI under 25, 25-30, 30 or more; age, blood pressure, labs, and day-one energy continuous", "reason": "the paper's categories" }, { "choice": "2021-2023 blood pressure", "decision": "the oscillometric readings (BPXO)", "reason": "2021-2023 measures blood pressure only by the oscillometric device; the library maps it" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods adjust for current smoking (yes, no, or unknown), but Table 2's counts (546 yes, 583 no, 7,743 unknown) and odds ratios (no 1.32, unknown 0.86; ours 1.36, 0.89) are those of SMQ040 in 2011-2012 alone, with everyone in 2013-2018 unknown. Coded so, Model 3 gives 0.69 for the headline, and coded in every cycle 0.78." }, { "kind": "sample", "affects_headline": true, "followed": false, "detail": "Figure 1 drops 6,590 participants for missing vigorous activity, but the vigorous activity item (PAQ605) is missing for fewer than 10 of the 16,223 at that step, and no variable tried has missing values that fit the per-cycle losses Supplementary Figure S1 implies (71%, 53%, 58%, and 54% kept). The file keeps them (n = 15,065 against 8,872), and its weighted descriptives still match Table 1." }, { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The Methods exclude only those younger than 18, but the education question is asked from age 20, so Figure 1's 1,071 excluded for missing education include the 18- and 19-year-olds (its 20,220 is reproduced exactly so), and the sample is 20 and older." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The Discussion says higher dietary fiber intake was independently associated with an increased prevalence of stroke and MI, but every odds ratio for higher intake in Table 3 is below 1." } ], "projected_power": 0.224729389024504 }, "row133": { "id": "row133", "row": 133, "doi": "10.1136/bmjopen-2023-082601", "measure": "beta", "contrast": "weekday sleep under 7 hours vs 7 to 9 hours (difference in mean VAI)", "published": { "measure": "beta", "estimate": 0.15, "low": 0.01, "high": 0.28, "n": 11252, "events": null, "contrast": "weekday sleep under 7 hours vs 7 to 9 hours (difference in mean VAI)" }, "published_b": [0.15, 0.01, 0.28], "constants": null, "original": { "estimate": 0.140860458213945, "ci": [0.00701165518149233, 0.274709261246398], "b": 0.140860458213945, "se": 0.0674123509247621, "df": 94, "p": 0.0393606128506471, "n": 11318, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, "harmonized": { "estimate": 0.141650590134942, "ci": [0.00854783475530813, 0.274753345514577], "b": 0.141650590134942, "se": 0.0670366073615849, "df": 94, "p": 0.0372474324727567, "n": 11351, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.939069721426302, "published_z": 2.17773776060006, "interval_ratio_original": 1.00860072665176, "ratio_harmonized": 1.00560932380184, "variants": [ { "label": "seven conditions as separate terms", "estimate": 0.119805083293741, "ci": [-0.00764282749011959, 0.247252994077602], "b": 0.119805083293741, "se": 0.0641885701757548, "df": 94, "p": 0.0650952058013746, "n": 11318, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "unweighted", "estimate": 0.137823621767992, "ci": [0.0245845701282251, 0.251062673407759], "b": 0.137823621767992, "se": 0.0577760879959948, "df": null, "p": 0.0170569145195302, "n": 11318, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "only those NCHS counts as fasted", "estimate": 0.135034802124706, "ci": [-0.00156353374039936, 0.271633137989812], "b": 0.135034802124706, "se": 0.0687971408369214, "df": 94, "p": 0.0526260941947947, "n": 10829, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "fasting subsample weight (WTSAF2YR)", "estimate": 0.128950046735082, "ci": [-0.00592463013568692, 0.263824723605851], "b": 0.128950046735082, "se": 0.0679290276945672, "df": 94, "p": 0.0607237317706782, "n": 10829, "events": null, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "Model I, crude (published 0.20, 0.06-0.33)", "estimate": 0.195190659689669, "ci": [0.0650245418471051, 0.325356777532234], "b": 0.195190659689669, "se": 0.0655575830019887, "df": 94, "p": 0.00369729075731448, "n": 11318, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "Model I, crude, unweighted", "estimate": 0.127721752241659, "ci": [0.0143234901556203, 0.241120014327697], "b": 0.127721752241659, "se": 0.0578573193081656, "df": null, "p": 0.0272770725986025, "n": 11318, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model II, age (published 0.21, 0.07-0.34)", "estimate": 0.202908548231314, "ci": [0.0728334718780518, 0.332983624584576], "b": 0.202908548231314, "se": 0.0655117303631417, "df": 94, "p": 0.00257496825458895, "n": 11318, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "Model II, age, unweighted", "estimate": 0.13112397374921, "ci": [0.0177396831159329, 0.244508264382487], "b": 0.13112397374921, "se": 0.057850190884954, "df": null, "p": 0.0234138993970989, "n": 11318, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "borderline diabetes counted as no diabetes", "estimate": 0.140400421576577, "ci": [0.00560378741618803, 0.275197055736965], "b": 0.140400421576577, "se": 0.067889721832579, "df": 94, "p": 0.0413808852318618, "n": 11602, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "middle 7 to under 9 h, long 9 h or more", "estimate": 0.144514085458557, "ci": [0.0107093254007707, 0.278318845516344], "b": 0.144514085458557, "se": 0.0673901688775811, "df": 94, "p": 0.0345768872905965, "n": 11318, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 } ], "left_out": ["physical activity (inactive, insufficient, sufficient, from minutes a week over the GPAQ's work, transport, and leisure domains): 2021-2023 asks only about leisure-time activity, so the harmonized model leaves it out", "married vs unmarried: 2021-2023 records marital status only as married or living with a partner, widowed, divorced, or separated, and never married (DMDMARTZ), so the harmonized model codes married or living with a partner against the rest", "Healthy Eating Index-2015 (a completeness requirement only, not in the model): computing it needs USDA food pattern equivalents from outside NHANES, so neither version computes it; both require the reliable day-1 recall it is computed from, which energy intake requires anyway"], "choices": [ { "choice": "weights", "decision": "examination weight, WTMEC2YR divided by 6 (WTPH2YR from HDL_L in 2021-2023)", "reason": "as the paper states; the weighted crude and age-adjusted fits reproduce Models I and II (0.195 and 0.203 against 0.20 and 0.21, with matching intervals) and unweighted ones do not (0.13). NCHS would weight fasting triglycerides with the fasting subsample weight (variant)" }, { "choice": "age", "decision": "16 and older (the text says 18)", "reason": "the flow chart's 38,562 'aged 18 or older' and 15,786 with VAI and sleep are exactly the counts at 16 and older (the sleep questionnaire's age) in these cycles (15,012 at 18 and older); 16 and 18 give the same analytic sample, because marital status and education are asked from 20 (Results: ages 20 to 80)" }, { "choice": "triglycerides for VAI", "decision": "the fasting subsample's values (TRIGLY files), in mmol/L, including the 5% NCHS doesn't count as fasted (fasting weight 0)", "reason": "unstated; every value in the files gives the flow chart's 15,786 with VAI and sleep exactly (14,928 counted as fasted); those counted as fasted only is a variant" }, { "choice": "sleep hours across cycles", "decision": "SLD010H (whole hours, 2007-2014) and SLD012 (half hours, asked in 2015-2016 and derived from sleep and wake times in 2017-2018, as in 2021-2023), used as recorded", "reason": "the same question (usual sleep at night on weekdays or workdays) in the same unit throughout" }, { "choice": "sleep groups", "decision": "under 7 hours short, 7 to 9 inclusive middle (reference), over 9 long, so 6.5 is short and 9.5 long", "reason": "the paper's labels (<7, 7-9, >9); Table 1's group sizes (3,885, 6,823, 544) fit them (3,865, 6,868, 585), and so do the group VAI means (2.11, 1.92, 2.12 against 2.13, 1.92, 2.12). Table 1's group sleep means (7.54, 9.51) fit 9 hours in the long group instead, which would make it 2.3 times larger (variant)" }, { "choice": "self-reported chronic diseases", "decision": "one covariate: told of any of hypertension, diabetes, stroke, heart attack, congestive heart failure, coronary heart disease, or cancer", "reason": "the Methods and Table 2's note adjust for 'self-reported chronic diseases, including' the seven, which reads as one covariate or seven; one term reproduces Model III's short-sleep estimate (0.140 against 0.15) and its long-sleep estimate (0.015, -0.25 to 0.28, against 0.01, -0.26 to 0.29), and seven terms give 0.119 and -0.040 (variant)" }, { "choice": "diabetes", "decision": "told by a doctor (DIQ010 1) or not (2); borderline missing", "reason": "the paper codes yes or no with no borderline group and excludes missing self-reported diseases; this gives 11,318 participants (published 11,252) against 11,602 with borderline counted as no (variant, 0.140 either way)" }, { "choice": "marital status", "decision": "married (DMDMARTL 1) against all others, living with a partner included", "reason": "Table 1's married shares (53.1, 58.8, 41.0) match married alone (52.8, 58.6, 40.5), not married or living with a partner (62.0, 66.2, 54.5)" }, { "choice": "education, race, smoking", "decision": "DMDEDUC2 1-2, 3, 4, 5; RIDRETH1 white, Mexican American, black, other (other Hispanic and other or multiracial); never, former, current from SMQ020 and SMQ040", "reason": "the paper's groups; Table 1's weighted shares match each within half a point in the short and middle groups and within three points in the small long group" }, { "choice": "energy intake", "decision": "day-1 recall (DR1TKCAL, reliable recall), kcal, continuous", "reason": "Table 1's means (2,248, 2,176, 1,995) match day 1 (2,252, 2,181, 1,942), not the two-day mean (about 2,140, 2,090)" }, { "choice": "physical activity", "decision": "moderate-equivalent minutes a week (vigorous counted twice) over the GPAQ's five domains: none inactive, 1-149 insufficient, 150 or more sufficient; missing if any domain is", "reason": "Table 1's minutes a week (1,144, 895, 752) match these (1,154, 904, 790); the 150-minute cutpoint is the physical activity guidelines', the convention for these labels" }, { "choice": "drinking status", "decision": "never, former, mild, moderate, heavy by Rattan and colleagues' definitions (library alcohol and alcohol5, which read 2007-2010's binge items for 5 or more drinks, ALQ140Q/U and ALQ150, in place of the later ones)", "reason": "unstated; the five labels are this convention's. Table 1 can't check it (its 'Never' row repeats 'Heavy'), and its former (24.0, 19.6, 23.0) and heavy (15.3, 11.1, 15.2) shares match no definition tried (these give about 17, 13, 24 and 28, 24, 30); alcohol coding moves the estimate by under 0.01" }, { "choice": "covariate form and missing data", "decision": "age, poverty income ratio (INDFMPIR), and energy continuous; complete-case analysis", "reason": "Table 1 reports them as means; the paper excludes participants with missing covariates" } ], "departures": [ { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Figure 1 labels its 38,562 starting participants as aged 18 or older, but that count and its 15,786 with VAI and sleep data are exactly those aged 16 or older (15,012 at 18 or older). The analytic sample is 20 and older either way (the Results give ages 20 to 80), since marital status and education are asked from 20." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's mean sleep in the middle and long groups (7.54 and 9.51 hours) fits 9 hours counted as long, but its group sizes (3,885, 6,823, 544) and mean VAI fit the stated groups, with 9 hours in the middle (3,865, 6,868, 585 here); with 9 hours counted as long, the long group would be 2.3 times as large." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 2 prints Model II's long-sleep estimate as 0.20 (0.09 to 0.49) with P = 0.17, which needs a negative lower limit (Model I's is -0.08)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's 'Never' drinking row repeats its 'Heavy' row (15.27, 11.09, 15.17%)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The abstract and Results give the participants' mean sleep and VAI as 7.05 hours and 2.03, which match the analytic sample's unweighted means (7.07 and 2.02 here; weighted 7.15 and 1.98), though the Methods say continuous variables are presented as weighted means." } ], "projected_power": 0.132531289318008 }, "row189": { "id": "row189", "row": 189, "doi": "10.1186/s12991-020-00315-1", "measure": "beta", "contrast": "PHQ-9 points per 1 pg/mL of serum estradiol", "published": { "measure": "beta", "estimate": 0.014, "low": 0.001, "high": 0.028, "n": 548, "events": null, "contrast": "PHQ-9 points per 1 pg/mL of serum estradiol" }, "published_b": [0.014, 0.001, 0.028], "constants": null, "original": { "estimate": 0.0143046563755793, "ci": [0.000842423626459985, 0.0277668891246986], "b": 0.0143046563755793, "se": 0.00656108418360618, "df": 27, "p": 0.038130977896102, "n": 261, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 57, "strata": 30 }, "harmonized": { "estimate": 0.00557672034547094, "ci": [-0.00622444840145316, 0.017377889092395], "b": 0.00557672034547094, "se": 0.00577844974137154, "df": 30, "p": 0.342215830302089, "n": 548, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.02176116968423, "published_z": 2.03255524322672, "interval_ratio_original": 1.00280542251679, "ratio_harmonized": 0.389853499381603, "variants": [ { "label": "duration with months converted to years", "estimate": 0.0150148729762888, "ci": [0.00133525339762581, 0.0286944925549519], "b": 0.0150148729762888, "se": 0.00666703193503969, "df": 27, "p": 0.0326523030501019, "n": 261, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 57, "strata": 30 }, { "label": "testosterone, duration as coded (published -0.033, -0.048 to -0.018)", "estimate": -0.0331482863050075, "ci": [-0.0476907950311439, -0.0186057775788711], "b": -0.0331482863050075, "se": 0.00708757795019183, "df": 27, "p": 7.25681559739598e-05, "n": 261, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 57, "strata": 30 }, { "label": "unweighted", "estimate": 0.00754590038503455, "ci": [-0.0109674922744016, 0.0260592930444707], "b": 0.00754590038503455, "se": 0.00944578206817442, "df": null, "p": 0.424368959903829, "n": 261, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "univariate, all 548 (published -0.003, -0.011 to 0.004)", "estimate": -0.00335574141873237, "ci": [-0.0110018170350543, 0.00429033419758954], "b": -0.00335574141873237, "se": 0.00374390576180503, "df": 30, "p": 0.377217722195965, "n": 548, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "users of estrogen-only pills, no duration (Table 3: 0.014, 0.001-0.027)", "estimate": 0.0139188379425201, "ci": [0.00111738960398365, 0.0267202862810566], "b": 0.0139188379425201, "se": 0.00623903789114907, "df": 27, "p": 0.0341877773330577, "n": 267, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 57, "strata": 30 }, { "label": "all women, 0 years for never taking estrogen-only pills", "estimate": 0.00586353336643484, "ci": [-0.00620288237496634, 0.017929949107836], "b": 0.00586353336643484, "se": 0.00590832809999048, "df": 30, "p": 0.328931050626433, "n": 528, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "PHQ-9 only when all nine items are answered", "estimate": 0.0150148729762888, "ci": [0.00133525339762581, 0.0286944925549519], "b": 0.0150148729762888, "se": 0.00666703193503969, "df": 27, "p": 0.0326523030501019, "n": 261, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 57, "strata": 30 } ], "left_out": ["duration of estrogen-only pill use (RHQ560Q/U): 2021-2023's reproductive health file asks no hormone-use questions; leaving it out also lifts the restriction it put on the paper's sample, which complete-case analysis limited to the 261 women who took estrogen-only pills and said for how long, so the harmonized version takes every woman with both ovaries removed (548 in 2013-2016)", "chronic respiratory disease as the paper counted it (still having asthma, emphysema, or still having chronic bronchitis, MCQ035, MCQ160G, MCQ170K): 2021-2023 asks emphysema and chronic bronchitis only within one question on ever having COPD, emphysema, or chronic bronchitis (MCQ160P), so the harmonized version counts still having asthma or ever any of those, in every cycle"], "choices": [ { "choice": "the multivariate model's sample", "decision": "complete cases: the women with a duration of estrogen-only pill use (261 of 548)", "reason": "the model adjusts for that duration, which only women who took estrogen-only pills have; Table 1 gives its mean over them alone (10.697 years), and Table 3's models by estrogen use leave it out. The paper reports 548 and states the result for ovariectomized women in general, but only this sample reproduces Table 2 (to the digit with the duration as coded); 0 years for never-users keeps 528 women and gives 0.006 (variant)" }, { "choice": "duration of estrogen use", "decision": "RHQ560Q read as years whatever its unit, as the paper's computation read it", "reason": "the paper's numbers (Table 1's 10.697 +- 1.108 years, the univariate -0.104 and the multivariate -0.110) are RHQ560Q read as years whatever its unit, so 43 answers in months count as years; the published estimate comes from that coding, and months divided by 12 is a variant" }, { "choice": "history of estrogen use", "decision": "ever took estrogen-only pills (RHQ554)", "reason": "Table 1's 267 yes and 18 no (55.42% and 4.92%), reproduced exactly; it enters only a variant" }, { "choice": "PHQ-9 score", "decision": "sum of the items answered 0-3, missing only if none was answered", "reason": "unstated; this gives the paper's n of 548 and Table 1's 4.522 +- 0.316 exactly (all nine items required: 544 women, 4.491)" }, { "choice": "weights", "decision": "examination weight (WTMEC2YR over 2, with strata and PSUs; WTPH2YR from TST_L in 2021-2023)", "reason": "'discharge weights' (sic); the MEC weight reproduces Table 1's shares and means and Table 2's estimates and CIs; unweighted fits do not (variant)" }, { "choice": "age", "decision": "older than 20, as written", "reason": "RHQ305 is asked from 20, and every woman in 2013-2016 with both ovaries removed was 31 or older, so it changes nothing there" }, { "choice": "CVD, chronic respiratory disease, arthritis", "decision": "ever told of heart failure, coronary heart disease, angina, heart attack or stroke; still having asthma, emphysema, or still having chronic bronchitis; ever told of arthritis; everyone else no", "reason": "unstated; these reproduce Table 1's counts and weighted shares exactly (109, 117, 342 yes; COPD (MCQ160O) is not part of the respiratory count)" }, { "choice": "education", "decision": "below high school (DMDEDUC2 1-2) against high school or above", "reason": "Table 1's 110 and 438" }, { "choice": "estradiol below the detection limit", "decision": "NCHS's fill value (the limit over the square root of 2), as released", "reason": "unstated; 135 of the 548 are below the limit. 2021-2023 measures estradiol by the same ID-LC-MS/MS method family in a wider steroid panel with a lower limit (1.72 pg/mL against 2.994 in 2013-2016), so fill values drop from 2.12 to 1.22 pg/mL: a measurement change recorded, not corrected" } ], "departures": [ { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The paper reports n = 548 and states the result for ovariectomized women, but its multivariate model adjusts for the duration of estrogen-only pill use, which only the 261 women who took them and gave a duration have, and only that sample reproduces Table 2's multivariate estimates and intervals to the digit (estradiol 0.0143, 0.0008 to 0.0278). With never-users given 0 years (528 women), the estimate is 0.006 (-0.006 to 0.018)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "Table 2 gives the duration of estrogen use in years, but Table 1's mean (10.697 years) and Table 2's estimates for it (univariate -0.104, multivariate -0.110) are reproduced only with RHQ560Q read as years whatever its unit, so the 43 durations given in months count as years. Converted to years, they give 0.0150 for the headline, against 0.0143." } ], "projected_power": 0.270097968738529 }, "row191": { "id": "row191", "row": 191, "doi": "10.1186/s12888-023-04935-1", "measure": "OR", "contrast": "highest vs lowest quartile of serum albumin", "published": { "measure": "OR", "estimate": 0.77, "low": 0.6, "high": 0.99, "n": 13681, "events": 1551, "contrast": "highest vs lowest quartile of serum albumin" }, "published_b": [-0.261364764134408, -0.510825623765991, -0.0100503358535015], "constants": { "cutpoints": [40, 42, 44], "cutpoints_published_sample": [40, 42, 44] }, "original": { "estimate": 0.821460126172217, "ci": [0.635334312627373, 1.06211285220895], "b": -0.196671880513347, "se": 0.129634851665448, "df": 109, "p": 0.132130868257232, "n": 13681, "events": 1551, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, "harmonized": { "estimate": 0.821460126172217, "ci": [0.635334312627373, 1.06211285220895], "b": -0.196671880513347, "se": 0.129634851665448, "df": 109, "p": 0.132130868257232, "n": 13681, "events": 1551, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.752480469831838, "published_z": -2.04588979087468, "interval_ratio_original": 0.974528606736282, "ratio_harmonized": 1, "variants": [ { "label": "Model I (published 0.52, 0.42-0.65)", "estimate": 0.522608078056707, "ci": [0.418512933398837, 0.652594415737795], "b": -0.64892346857784, "se": 0.112072482482042, "df": 109, "p": 6.87055903585317e-08, "n": 13681, "events": 1551, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "Model I, unweighted", "estimate": 0.549253982075431, "ci": [0.473026991396629, 0.637764741362848], "b": -0.599194317720157, "se": 0.0762302324497692, "df": null, "p": 3.83141433734498e-15, "n": 13681, "events": 1551, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model II (published 0.60, 0.48-0.76)", "estimate": 0.603471802459822, "ci": [0.477565361789042, 0.762572509446313], "b": -0.505055962883558, "se": 0.118063632945229, "df": 109, "p": 4.06654186946271e-05, "n": 13681, "events": 1551, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "per g/L (published 0.98, 0.95-1.00)", "estimate": 0.976219167882055, "ci": [0.954528337073705, 0.998402904058304], "b": -0.0240681605220478, "se": 0.0113371122278474, "df": 109, "p": 0.0360211500437543, "n": 13681, "events": 1551, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "PHQ-9 score, linear (Fig. 3: -0.39, -0.67 to -0.11)", "estimate": -0.389064577852454, "ci": [-0.670214321003364, -0.107914834701543], "b": -0.389064577852454, "se": 0.141853862188371, "df": 109, "p": 0.00712511364601879, "n": 13681, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "all adults, as the text describes the sample", "estimate": 0.889873485104011, "ci": [0.745626237590776, 1.06202649473524], "b": -0.116675977928971, "se": 0.0892319623379175, "df": 109, "p": 0.193774495424311, "n": 31716, "events": 2743, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "all adults, Model I", "estimate": 0.534559639976507, "ci": [0.461149425203012, 0.619656001015323], "b": -0.626311973868928, "se": 0.0745325958371105, "df": 109, "p": 1.80864750172225e-13, "n": 31716, "events": 2743, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "all adults, Model II", "estimate": 0.631194788614709, "ci": [0.538246278198, 0.740194363272139], "b": -0.460140765798388, "se": 0.0803738166355218, "df": 109, "p": 9.23691337689521e-08, "n": 31716, "events": 2743, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 }, { "label": "all adults, unweighted", "estimate": 0.850453264848307, "ci": [0.750435657236203, 0.963801158323014], "b": -0.16198581886379, "se": 0.0638356351970194, "df": null, "p": 0.0111632905651889, "n": 31716, "events": 2743, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "all adults, drinking in the past 12 months (the text's question)", "estimate": 0.880766689790158, "ci": [0.738656414296417, 1.05021759349755], "b": -0.126962512419856, "se": 0.0887804103414157, "df": 109, "p": 0.155556158754447, "n": 31693, "events": 2741, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 214, "strata": 105 } ], "left_out": [], "choices": [ { "choice": "study population", "decision": "the sample the paper's computation analyzed: adults aged 20 and older with a complete PHQ-9 under 27, serum albumin, and Model III covariates, which complete-case coding limited to ever smokers (SMQ040 is asked only of them) who had drunk (the daily-drinking item is not asked of lifetime abstainers)", "reason": "that sample reproduces n = 13,681 and 1,551 cases, every count in Table 1 (8,226 men, 6,263 current smokers, 3,588 drinkers, quartiles of 3,971, 3,159, 3,167, 3,384), its weighted shares (56.74% men, 44.52% current smokers), and Models I and II; the published estimate comes from it. The text describes adults aged 20 and older; that version is a variant (n = 31,716)" }, { "choice": "confidence interval of the headline", "decision": "the abstract's 0.60-0.99 kept as published", "reason": "the abstract prints 0.77 (0.60-0.99); the Results text and Fig. 2 print 0.77 (0.60-0.97), P = 0.044" }, { "choice": "albumin quartiles", "decision": "unweighted quartiles of albumin in g/L (LBDSALSI) over the analytic sample, intervals closed on the right: 40 or less, 41-42, 43-44, 45 or more", "reason": "cutpoints and unit are not reported. Unweighted quartiles of the published sample (40, 42, 44 g/L) reproduce Table 1's quartile sizes and Fig. 2's cases per quartile (577, 369, 316, 289) exactly; weighted ones (41, 43, 45) do not. All adults give the same unweighted cutpoints (weighted: 40, 43, 45). Table 1's header row (2,745, 2,744, 3,292, 4,900) is the same cutpoints with intervals closed on the left, which its body and Fig. 2 do not use. The continuous estimate is per g/L (variant)" }, { "choice": "albumin across analyzers", "decision": "values as released in every cycle, no crosswalk", "reason": "the paper mentions none and Table 1 is reproduced without one. NCHS's bridging study says 2017-2018's Roche Cobas 6000 reads 4.4% below the earlier Beckman analyzer and recommends its equation when pooling; 2021-2023's Cobas 8000 differs from the Cobas 6000 by 1.5%, with no adjustment recommended. So the fixed cutpoints, set mostly on Beckman values, will put more of 2021-2023 in the lowest quartile (about half of 2017-2018's adults were in it)" }, { "choice": "weights", "decision": "examination weight (WTMEC2YR) over 7, with strata and PSUs (WTPH2YR from BIOPRO_L in 2021-2023)", "reason": "the paper combined MEC examination weights over seven cycles as NCHS directs; on the published sample, weighted fits reproduce Model I (Q2-Q4 0.70, 0.62, 0.52) and Model II (0.74, 0.70, 0.60), CIs included, and unweighted ones do not" }, { "choice": "smoking status", "decision": "current smoker (SMQ040 every day or some days) against former or never smoker", "reason": "the text's 'Do you smoke now?'; Table 1's 6,263 smokers are SMQ040 1-2" }, { "choice": "drinking status", "decision": "ever drank 4 (women) or 5 (men) drinks almost every day (ALQ151; ALQ150, 5 drinks for everyone, in 2005-2010), lifetime abstainers as no", "reason": "no NHANES item reads 'Do you drink alcohol now?'; this item alone reproduces Table 1's 3,588 drinkers (25.23% weighted); drinking in the past 12 months is a variant" }, { "choice": "PHQ-9", "decision": "sum of the nine items, missing if any is missing, refused, or don't know; 10 or more", "reason": "the paper's definition; with the published sample's steps it reproduces n and cases exactly" }, { "choice": "BMI", "decision": "continuous", "reason": "Table 1 reports 'BMI status' as a mean, and with continuous BMI the same covariates reproduce Fig. 3's linear Model III exactly (variant); BMI in three groups does not" }, { "choice": "education", "decision": "DMDEDUC2 in its five levels", "reason": "Table 1's five categories, reproduced exactly" }, { "choice": "diabetes", "decision": "DIQ010 yes, no, borderline", "reason": "Table 1's three levels" }, { "choice": "chronic conditions", "decision": "ever told of congestive heart failure (MCQ160B), coronary heart disease (MCQ160C), a liver condition (MCQ160L), cancer (MCQ220), a thyroid problem (MCQ160M); yes or no", "reason": "the Methods' self-reported history; Table 1's counts are reproduced exactly" }, { "choice": "pregnancy", "decision": "pregnant women kept", "reason": "no pregnancy exclusion is described, and the published sample is reproduced without one" }, { "choice": "Model III quartile estimates", "decision": "the paper's covariates and codings, unchanged", "reason": "on the published sample, Fig. 3's linear Model III (Q2-Q4 -0.31, -0.52, -0.39, CIs included) and Fig. 2's continuous Model III (0.98, 0.95-1.00) are reproduced exactly, but Fig. 2's logistic Model III quartiles are not: we get 0.85, 0.86, 0.82 (0.64-1.06) against 0.81, 0.80, 0.77. Codings tried without a match: education, race, or diabetes as numbers; three or four BMI groups; SMQ040 in three levels" } ], "departures": [ { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The text describes adults aged 20 and older with complete data, but n = 13,681, its 1,551 cases, every count in Table 1, and Models I and II are reproduced only in ever smokers who had drunk alcohol (the smoking question used, SMQ040, skips never smokers, and the drinking item skips lifetime abstainers), less 8 with a PHQ-9 score of 27. All adults give n = 31,716 and an OR of 0.89 (0.75-1.06), against 0.82 (0.64-1.06) in that sample." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods code drinking status from the question 'Do you drink alcohol now?', which NHANES doesn't ask, and Table 1's 3,588 drinkers (25.23% weighted) are reproduced only by ever having had 4 or 5 or more drinks almost every day (ALQ151; ALQ150, 5 drinks, through 2009-2010)." }, { "kind": "reporting", "affects_headline": true, "followed": true, "detail": "The abstract prints the headline's interval as 0.60-0.99, and the Results and Fig. 2 print it as 0.60-0.97 (P = 0.044); the file keeps the abstract's." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's header gives quartile sizes of 2,745, 2,744, 3,292, and 4,900, those of the same cutpoints with groups closed on the left, while its body (3,971, 3,159, 3,167, 3,384) and Fig. 2's cases per quartile (577, 369, 316, 289) are those of groups closed on the right." } ], "projected_power": 0.112010868929842 }, "row194": { "id": "row194", "row": 194, "doi": "10.1186/s12872-023-03333-5", "measure": "beta", "contrast": "IU/L of serum creatine phosphokinase per 1 umol/L of serum uric acid", "published": { "measure": "beta", "estimate": 0.12, "low": 0.09, "high": 0.15, "n": 8431, "events": null, "contrast": "IU/L of serum creatine phosphokinase per 1 umol/L of serum uric acid" }, "published_b": [0.12, 0.09, 0.15], "constants": null, "original": { "estimate": 0.110140247410076, "ci": [0.0671972235110576, 0.153083271309094], "b": 0.110140247410076, "se": 0.021027078814348, "df": 30, "p": 1.18943841166122e-05, "n": 8431, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, "harmonized": { "estimate": 0.110140247410076, "ci": [0.0671972235110576, 0.153083271309094], "b": 0.110140247410076, "se": 0.021027078814348, "df": 30, "p": 1.18943841166122e-05, "n": 8431, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.917835395083964, "published_z": 7.83985593816021, "interval_ratio_original": 0.698600081599888, "ratio_harmonized": 1, "variants": [ { "label": "unweighted", "estimate": 0.132295661180231, "ci": [0.102613692340831, 0.161977630019631], "b": 0.132295661180231, "se": 0.0151441399298803, "df": null, "p": 2.42010711187496e-18, "n": 8431, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1, weighted (published 0.30, 0.27-0.33)", "estimate": 0.281966686586017, "ci": [0.239292214626471, 0.324641158545562], "b": 0.281966686586017, "se": 0.0208955821873216, "df": 30, "p": 2.81628743905637e-14, "n": 8431, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "Model 1, unweighted", "estimate": 0.293020814033278, "ci": [0.265297708574357, 0.3207439194922], "b": 0.293020814033278, "se": 0.014144701472883, "df": null, "p": 2.48792139879498e-95, "n": 8431, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2, weighted (published 0.14, 0.11-0.17)", "estimate": 0.13337969357877, "ci": [0.087978054523495, 0.178781332634045], "b": 0.13337969357877, "se": 0.0222309412807252, "df": 30, "p": 1.39532109826404e-06, "n": 8431, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "Model 2, unweighted", "estimate": 0.154082437825494, "ci": [0.12607019509425, 0.182094680556739], "b": 0.154082437825494, "se": 0.0142922232001208, "df": null, "p": 4.2391015496601e-27, "n": 8431, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Black adults, Model 1, weighted (published 0.42, 0.34-0.51)", "estimate": 0.420808952889198, "ci": [0.333031486801328, 0.508586418977067], "b": 0.420808952889198, "se": 0.0428515719833608, "df": 28, "p": 1.43457391737182e-10, "n": 1789, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 58, "strata": 30 }, { "label": "Black adults, Model 1, unweighted", "estimate": 0.356284909298468, "ci": [0.286007247123282, 0.426562571473654], "b": 0.356284909298468, "se": 0.0358566089629848, "df": null, "p": 2.89142802392891e-23, "n": 1789, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "complete case (LDL and TG only in the fasting subsample)", "estimate": 0.123401318752623, "ci": [0.0588632392135786, 0.187939398291667], "b": 0.123401318752623, "se": 0.031601111467727, "df": 30, "p": 0.000495476030482065, "n": 2966, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "text's definitions: BMI at 25 and 30, BP 130/80 or treated, diabetes treated", "estimate": 0.108452582078866, "ci": [0.065513554797262, 0.151391609360471], "b": 0.108452582078866, "se": 0.0210251218681034, "df": 30, "p": 1.49006917665265e-05, "n": 8431, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "drinking as the text defines it", "estimate": 0.110231534992324, "ci": [0.0674468137469343, 0.153016256237713], "b": 0.110231534992324, "se": 0.0209495658198695, "df": 30, "p": 1.11231254311132e-05, "n": 8431, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 } ], "left_out": [], "choices": [ { "choice": "weights", "decision": "examination weight (WTMEC2YR over 2, with strata and PSUs; WTPH2YR from BIOPRO_L in 2021-2023)", "reason": "the paper says it used NHANES sample weights without naming one; weighted fits come closest to Table 2's subgroup estimates (Black adults' Model 1: 0.42, 0.34-0.51, exactly, against 0.36 unweighted; Mexican Americans 0.36 against the published 0.41, 0.26 unweighted; Whites 0.21 against 0.22, 0.18 unweighted)" }, { "choice": "missing covariates", "decision": "continuous covariates filled with the analytic sample's mean, categorical ones with missing as its own level; all 8,431 kept", "reason": "unstated; the paper analyzed 8,431 although LDL and fasting triglycerides exist for only 3,864 and 3,907 of them, Table 1 shows drinking with a 'Missing data' group, and Table 1's means and SDs are the mean-filled values' (LDL 2.9 +- 0.6 mmol/L, 0.92 unfilled; triglycerides 1.3 +- 0.7, 0.97 unfilled; income-poverty ratio 3.1, 3.17 unfilled; waist SD 16.0, 16.37 unfilled). Complete-case analysis keeps 2,966 (variant)" }, { "choice": "triglycerides and total cholesterol", "decision": "fasting triglycerides and LDL (TRIGLY, LBDTRSI and LBDLDLSI) and total cholesterol from TCHOL (LBDTCSI)", "reason": "unstated; Table 1's means fit these (TC 5.0 against 5.03 from TCHOL and 5.08 from the biochemistry profile; TG 1.3 against 1.33 fasting and 1.77 non-fasting)" }, { "choice": "BMI", "decision": "three groups cut at 24 and 28 kg/m2", "reason": "the text and Table 1's labels cut at 25 and 30, but Table 1's shares are the cuts at 24 and 28 (17.9, 25.5, 55.6% against 17.9, 25.4, 55.7), so Model 3 used them; 25 and 30 is part of a variant" }, { "choice": "hypertension and diabetes", "decision": "self-reported diagnosis (BPQ020; DIQ010 yes or no, borderline with missing as its own level)", "reason": "the text adds measured pressure of 130/80 or more, treatment, and diabetes treatment, but Table 1's shares are self-report alone (38.1% hypertension against 38.0; 13.7% and 83.6% for diabetes against 13.6 and 83.7); the text's definitions are a variant" }, { "choice": "drinking", "decision": "the text's groups (often once a week or more, sometimes 1 to 3 times a month, none or rarely less than monthly or never) as the paper's computation coded them: ALQ121's less-than-monthly answers (2017-2018) and lifetime abstainers missing, 2015-2016 answers grouped by their unit, missing as its own level", "reason": "Table 1's shares show this coding; the published estimate comes from it, and the text's coding (lifetime abstainers none) is a variant" }, { "choice": "race", "decision": "non-Hispanic White, non-Hispanic Black, Mexican American, other (other Hispanic included)", "reason": "the text's four groups; Table 1's 16.0% other is RIDRETH1 2 and 5 (16.1%)" }, { "choice": "education", "decision": "less than high school, high school or GED, college graduate or above with some college included", "reason": "Table 1's shares (13.1, 22.8, 64.1%) are DMDEDUC2 1-2, 3, 4-5 (13.3, 22.7, 63.9%)" }, { "choice": "covariate coding", "decision": "age, income-poverty ratio, waist, ALT, AST, HDL, total cholesterol, triglycerides, LDL (mmol/L) and total protein (g/L) continuous; activity, smoking, kidney disease yes or no", "reason": "Table 1 gives the continuous ones as means and the others as yes or no" }, { "choice": "moderate recreational activity in 2021-2023", "decision": "any moderate leisure-time activity (PAD790Q above 0)", "reason": "2021-2023 does not ask PAQ665 (in a typical week, at least 10 minutes); it asks how often" }, { "choice": "CPK and uric acid across analyzers", "decision": "values as released in every cycle, no crosswalk", "reason": "the paper pooled the DxC800 (2015-2016) and Cobas 6000 (2017-2018) values as released. 2021-2023 measured on a Cobas 8000; NCHS's bridging study recommends no adjustment for uric acid but gives Cobas 8000 = -0.5307 + 1.017 x Cobas 6000 for CPK (and Cobas 8000 = 1.477 + 0.9663 x Cobas 6000 for ALT, a covariate). These are recorded as measurement changes, not applied, as the plan reports measurement changes without correcting them" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods and Table 1's labels cut BMI at 25 and 30 kg/m2, but Table 1's shares (17.9, 25.4, 55.7%) are those of cuts at 24 and 28 (17.9, 25.5, 55.6%; 25 and 30 give 23.4, 32.7, 42.8%), and Model 3 adjusts for the covariates as shown in Table 1." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods define hypertension as mean pressure of 130/80 mmHg or more, antihypertensive drugs, or a diagnosis, but Table 1's 38.0% is self-reported diagnosis alone (38.1%; the stated definition gives 57.9%)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods define diabetes as a diagnosis or use of insulin or diabetes medication, but Table 1's shares (13.6% yes, 83.7% no) are self-reported diagnosis alone (13.7% and 83.6%; adding treatment gives 14.8% yes)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods group drinking as none or rarely, sometimes (1 to 3 times a month), and often (once a week or more), but Table 1's shares (25.3, 18.6, 29.3%, and 26.7% missing) are those of 2015-2016's answers grouped by the unit they were given in, with 2017-2018's less-than-monthly answers and all lifetime abstainers left missing (25.3, 18.5, 29.3, 27.0%)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods and Table 1 name the top education group college graduate or above, but Table 1's shares (13.1, 22.8, 64.1%) are those with some college counted in it (13.3, 22.7, 63.9%)." } ], "projected_power": 0.99931936023765 }, "row218": { "id": "row218", "row": 218, "doi": "10.1038/s41598-024-51216-2", "measure": "OR", "contrast": "per unit of WWI (waist cm over the square root of weight kg)", "published": { "measure": "OR", "estimate": 1.2, "low": 1.13, "high": 1.28, "n": 14118, "events": 3338, "contrast": "per unit of WWI (waist cm over the square root of weight kg)" }, "published_b": [0.182321556793955, 0.122217632724249, 0.246860077931526], "constants": null, "original": { "estimate": 1.2162888613364, "ci": [1.13809267263525, 1.29985776183371], "b": 0.195804305774458, "se": 0.0339039590813785, "df": null, "p": 7.68326366311569e-09, "n": 12016, "events": 2816, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 1.21292680061388, "ci": [1.13477782949119, 1.29645767251821], "b": 0.193036282398148, "se": 0.0339799078413338, "df": null, "p": 1.3399186274593e-08, "n": 12014, "events": 2813, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.07395038314499, "published_z": 5.73390042737324, "interval_ratio_original": 0.937858801329121, "ratio_harmonized": 0.98586331712491, "variants": [ { "label": "weighted, as the paper says (WTMEC2YR)", "estimate": 1.19673447248161, "ci": [1.08007674758886, 1.32599225084976], "b": 0.179596574467736, "se": 0.0509231421983465, "df": 45, "p": 0.00098081246116096, "n": 12016, "events": 2816, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "covariates imputed (mode, median), as the paper says", "estimate": 1.2299473815872, "ci": [1.15722424609723, 1.3072406377373], "b": 0.206971389271873, "se": 0.0310960525140243, "df": null, "p": 2.81622562785799e-11, "n": 14118, "events": 3338, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "missing categories as their own level", "estimate": 1.20145148878965, "ci": [1.12674281017079, 1.28111372611826], "b": 0.183530399840071, "se": 0.0327553969417037, "df": null, "p": 2.1060230636745e-08, "n": 12777, "events": 3016, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "2017-2018 never drinkers counted as not drinking monthly", "estimate": 1.21551859816417, "ci": [1.13856548463969, 1.2976728017981], "b": 0.195170815491352, "se": 0.0333688235045454, "df": null, "p": 4.94843036275826e-09, "n": 12376, "events": 2906, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "alcohol 2017-2018 as 12+ drinks in the past year", "estimate": 1.21884768973605, "ci": [1.14169025498231, 1.3012195595888], "b": 0.197905895798487, "se": 0.0333659455904861, "df": null, "p": 3.00410400311645e-09, "n": 12372, "events": 2906, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "alcohol left out", "estimate": 1.21322848417087, "ci": [1.13541999166808, 1.29636906660514], "b": 0.193284975101737, "se": 0.0338181499384358, "df": null, "p": 1.0943254747952e-08, "n": 12016, "events": 2816, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude (published 1.96, 1.87-2.06)", "estimate": 1.96145040716757, "ci": [1.86718439261785, 2.06047550257413], "b": 0.673684203231647, "se": 0.0251293288112944, "df": null, "p": 2.55948527723652e-158, "n": 14118, "events": 3338, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, weighted", "estimate": 2.03301191889632, "ci": [1.88201701343702, 2.19612120021508], "b": 0.709518397342415, "se": 0.0383169585845041, "df": 45, "p": 1.12207648259142e-22, "n": 14118, "events": 3338, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "age, sex, race (published 1.30, 1.22-1.38)", "estimate": 1.29714553185806, "ci": [1.22316637990632, 1.37559906686463], "b": 0.260166105560872, "se": 0.0299613748671779, "df": null, "p": 3.84166674761375e-18, "n": 14118, "events": 3338, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "age, sex, race, weighted", "estimate": 1.2908094498475, "ci": [1.16830533338613, 1.42615888860701], "b": 0.25526950209194, "se": 0.0495084993298653, "df": 45, "p": 5.4614144690045e-06, "n": 14118, "events": 3338, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 } ], "left_out": "alcohol use as ever having had 12 drinks in one year (ALQ101, asked through 2015-2016 only): the harmonized version uses drinking at least once a month in the past year in every cycle instead (ALQ120Q/U through 2015-2016, ALQ121 after), which is how the paper itself coded 2017-2018", "choices": [ { "choice": "weights", "decision": "unweighted", "reason": "the crude and age-sex-race estimates for overall, weekly, and daily UUI (Table 2) are reproduced exactly only unweighted; the weighted crude OR (2.03) lies outside the published interval; the text says it weighted" }, { "choice": "missing covariates", "decision": "complete-case analysis", "reason": "the text says missing values were imputed with the mode or median, but Table 3's intervals are those of a model on about 12,000 people, not 14,118, and a complete-case fit with the paper's alcohol coding (n = 12,016) gives 12 of its 13 covariate odds ratios and their intervals to two decimals (income-to-poverty ratio 0.95, 0.92-0.98, against 0.94, 0.91-0.98); imputation as stated matches few of them and gives 1.23 (variant)" }, { "choice": "pregnancy exclusion", "decision": "pregnant at the exam (RIDEXPRG 1)", "reason": "'missing data related to pregnancy (n = 190)' is exactly the number pregnant among those 18 and over" }, { "choice": "missing UUI", "decision": "KIQ044 other than yes or no (missing, refused, don't know)", "reason": "gives the paper's n = 3040 exactly" }, { "choice": "WWI", "decision": "waist (cm) over the square root of weight (kg)", "reason": "the Methods' wording inverts it, but the published quartile cutpoints (10.54, 11.12, 11.70) and range (8.46-14.79) are this index's" }, { "choice": "alcohol use in 2017-2018", "decision": "drinking at least once a month in the past year (ALQ121 1-7) against less often, never drinkers missing, as the paper's computation coded it", "reason": "unstated (ALQ101 was not asked then); this coding gives Table 1's 69.92% and Table 3's alcohol OR (0.95, 0.86-1.05) exactly, and the complete-case model drops the 360 never drinkers; the published estimate comes from it. Counting never drinkers as no is a variant. The harmonized version uses drinking at least monthly in every cycle, with never drinkers missing where ALQ121 skips them (2017-2018 and 2021-2023), as the paper's coding left them" }, { "choice": "diabetes", "decision": "told by a doctor (DIQ010) yes against no, borderline and don't know missing", "reason": "unstated; Table 1's shares (11.04% overall; 1.88, 7.64, 13.01, 25.44% by quartile) are exactly these" }, { "choice": "education, smoking, insurance, race", "decision": "DMDEDUC2 1-2, 3, 4-5; SMQ020 yes or SMQ040 now; HIQ011; RIDRETH1 five groups", "reason": "Table 1's weighted shares match each to the hundredth" }, { "choice": "income-to-poverty ratio", "decision": "continuous", "reason": "Table 3 gives an OR per unit" } ], "departures": [ { "kind": "weighting", "affects_headline": true, "followed": true, "detail": "The Methods say the models were weighted, but Table 2's crude and age-sex-race estimates for overall, weekly, and daily UUI are reproduced exactly only unweighted (crude 1.96, 1.87-2.06; weighted 2.03, 1.88-2.20), and Model 3's interval is as narrow as the unweighted fit's (1.14-1.30; weighted 1.08-1.33)." }, { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The Methods say missing covariates were imputed with the mode or median, and the paper reports n = 14,118, but Table 3's intervals are those of a model on about 12,000, and a complete-case fit (n = 12,016) gives 12 of its 13 covariate odds ratios and their intervals to two decimals; imputing as stated gives 1.23 (1.16-1.31) for the headline." }, { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The Methods exclude only those under 18, but the 3,040 excluded for missing UUI, a count reproduced exactly, include all 904 participants aged 18 or 19, whom the urge-leakage question (asked from 20) skips, so the sample is 20 and older." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods compute WWI as the square root of waist circumference (cm) divided by weight (kg), but the paper's quartile cutpoints (10.54, 11.12, 11.70) and range (8.46-14.79) are those of waist circumference divided by the square root of weight." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods define alcohol use as at least 12 drinks within a year, which 2017-2018 doesn't ask; Table 1's 69.92% and Table 3's odds ratio (0.95, 0.86-1.05) are reproduced exactly with drinking at least once a month in the past year there and never drinkers, whom that question skips, left missing, so the complete-case model drops 360 of them." }, { "kind": "reporting", "affects_headline": true, "followed": false, "detail": "The abstract prints the headline's upper limit as 12.8; Tables 2, 3, and 4 and the Results give 1.28, which the file uses." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 3 labels its smoking odds ratio of 1.25 (1.13, 1.38) as no against yes, and the Results say non-smokers had 25% higher odds, but our Model 3, which gives Table 3's sex, age, race, diabetes, and alcohol odds ratios to two decimals, gives 0.80 (0.72-0.88) for no against yes, so 1.25 is smokers against non-smokers." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The Results give non-Hispanic White and other Hispanic participants 25% and 10% lower odds of UUI than Mexican Americans, but Table 3, like our Model 3, gives 0.90 for non-Hispanic White and 0.75 for other Hispanic." } ], "projected_power": 0.871271935668844 }, "row220": { "id": "row220", "row": 220, "doi": "10.1007/s11255-023-03917-2", "measure": "OR", "contrast": "ABSI (x 1000) quartile 4 (86.42 or more) vs quartile 1 (below 80.94)", "published": { "measure": "OR", "estimate": 1.91, "low": 1.12, "high": 3.27, "n": 11013, "events": 492, "contrast": "ABSI (x 1000) quartile 4 (86.42 or more) vs quartile 1 (below 80.94)" }, "published_b": [0.647103242058538, 0.113328685307003, 1.18478998490916], "constants": null, "original": { "estimate": 1.92473977804938, "ci": [1.12470928156522, 3.29384959644859], "b": 0.654790778367989, "se": 0.27173388908771, "df": 139, "p": 0.0172753677330607, "n": 11022, "events": 493, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, "harmonized": { "estimate": 1.91554371712656, "ci": [1.12216031929926, 3.26986052626986], "b": 0.650001507725882, "se": 0.270459158594161, "df": 139, "p": 0.0175658050052142, "n": 11022, "events": 493, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.01187992241392, "published_z": 2.3674192417118, "interval_ratio_original": 0.997141933153956, "ratio_harmonized": 0.992685800105426, "variants": [ { "label": "unweighted", "estimate": 1.88337667545276, "ci": [1.32309352556569, 2.68091985419014], "b": 0.633066269740171, "se": 0.180153154779128, "df": null, "p": 0.000441338252963362, "n": 11022, "events": 493, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model I (published 6.43, 3.96-10.42)", "estimate": 6.46017848517484, "ci": [3.98294826129882, 10.4781441591478], "b": 1.86565694669686, "se": 0.244608578416333, "df": 139, "p": 3.43001794679784e-12, "n": 11022, "events": 493, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "Model II (published 1.92, 1.15-3.23)", "estimate": 1.94652125189209, "ci": [1.15991173590285, 3.26658043606938], "b": 0.666043806011386, "se": 0.261837814129964, "df": 139, "p": 0.0120613076577534, "n": 11022, "events": 493, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "2017-2018 former drinkers kept, as mild drinkers", "estimate": 1.90446719389536, "ci": [1.10998461105909, 3.26760862852231], "b": 0.644202281214707, "se": 0.273043805425464, "df": 139, "p": 0.0196990790308127, "n": 11355, "events": 509, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "heavy drinking at 3 drinks a day, as printed", "estimate": 1.91827222795132, "ci": [1.12058753814003, 3.28378481402448], "b": 0.651424899628114, "se": 0.271888438548255, "df": 139, "p": 0.0179096271135186, "n": 11022, "events": 493, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "hypertension at diastolic 80 mmHg, as printed", "estimate": 1.94352423723993, "ci": [1.14143264669571, 3.30925041584693], "b": 0.664502942155933, "se": 0.269181042516037, "df": 139, "p": 0.0147767813115965, "n": 11022, "events": 493, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "education as the Methods split it", "estimate": 1.92648756848728, "ci": [1.13066886070438, 3.28244146497846], "b": 0.655698432162802, "se": 0.269520061763307, "df": 139, "p": 0.0162518974351298, "n": 11022, "events": 493, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "age continuous", "estimate": 1.33963464551092, "ci": [0.742205565541671, 2.4179567855213], "b": 0.292396924182546, "se": 0.298671155629031, "df": 139, "p": 0.329284344947298, "n": 11022, "events": 493, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "ABSI continuous, per unit (published 1.05, 1.02-1.08)", "estimate": 1.04701474182367, "ci": [1.01622386493457, 1.07873856088459], "b": 0.0459430118500901, "se": 0.015096947073966, "df": 139, "p": 0.00279975271693715, "n": 11022, "events": 493, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 } ], "left_out": ["the two-hour oral glucose tolerance test from the diabetes definition: 2021-2023 gave none, so diabetes there is told, HbA1c, fasting glucose, or medication (the harmonized model uses that definition in every cycle)", "the fewer-than-12-drinks-in-life item behind never drinkers (asked 2001-2016): 2021-2023, like 2017-2018, asks only whether one ever had a drink (ALQ111), so never drinkers there are those who never had one, as in the paper's own 2017-2018 data"], "choices": [ { "choice": "weights", "decision": "examination weight (WTMEC2YR) over nine cycles, each divided by nine, with strata and PSUs; WTPH2YR in 2021-2023", "reason": "the paper weights as CDC directs without naming the weight; ABSI comes from the examination; NCHS uses 2-year weights for 2001-2002 when 1999-2000 is not pooled; weighted confidence intervals match the published widths, unweighted ones are far narrower" }, { "choice": "quartile cutpoints", "decision": "the published 80.94, 83.66, and 86.42, each opening the higher quartile", "reason": "Table 2's row labels; boundary side unstated" }, { "choice": "prostate cancer and other cancers", "decision": "cases told of cancer with prostate among the first three kinds; men with any other kind, more than three kinds, or a kind not named are left out; controls never told of cancer", "reason": "the paper keeps men with no tumor history but prostate cancer; Fig. 1's 91,351, 45,010, and 16,127 are reproduced exactly" }, { "choice": "drinking status", "decision": "Rattan et al.'s groups for men: heavy at 4 or more drinks on a drinking day or 5 or more drinks on 5 or more days a month; never fewer than 12 drinks in life (never a drink from 2017 on); mild otherwise, former drinkers included", "reason": "the text prints 3 drinks, Rattan's threshold for women, but cites Rattan (ref. 20), and Table 1's 2,048 heavy drinkers (31 with prostate cancer) fit 4 (2,068 and 33 here) rather than 3 (3,164); its 801 never drinkers (47 with prostate cancer) match exactly; 3 drinks is a variant" }, { "choice": "2017-2018 former drinkers", "decision": "dropped, as the paper's computation dropped them (and in 2021-2023, whose drinking items skip them the same way)", "reason": "its n (11,013, 492 with prostate cancer) and Table 1's margins are matched within 20 men (most within 5) only if the 333 men of 2017-2018 who drank before but not in the past year, whose ALQ130 and ALQ142 are skipped, are dropped, as complete-case coding of those items does; keeping them as mild drinkers, as the paper's definitions classify them, is a variant" }, { "choice": "hypertension", "decision": "told, taking prescribed medicine, or a mean systolic pressure of 140 mmHg or more or diastolic of 90 or more over the second and third readings", "reason": "the text prints diastolic 80 but cites the 2018 ESC/ESH guideline (ref. 21), whose threshold is 90; on the sample the paper ran (2017-2018 former drinkers dropped), Table 1's 5,909 hypertensive men (358 with prostate cancer) are matched exactly with 90 over readings 2 and 3 (5,949 over all readings; 6,855 with 80); 80 is a variant" }, { "choice": "diabetes", "decision": "told, HbA1c 6.5% or more, fasting glucose (LBXGLU) 126 mg/dL or more, two-hour OGTT glucose 200 mg/dL or more where the cycle gave the test (2005-2016), or insulin or diabetes pills; no random glucose", "reason": "the paper's list; NHANES has no random glucose measure (refrigerated serum glucose would add 3 men); Table 1's 2,752 with diabetes (149 with prostate cancer) compare with 2,747 (148) on the sample the paper ran; fasting glucose as recorded, as is usual in this literature" }, { "choice": "income to poverty ratio", "decision": "low 1 or less, middle over 1 to 3, high over 3", "reason": "the text puts 3 itself in high, but Table 1's counts fit R's default intervals (10 men have exactly 3)" }, { "choice": "education", "decision": "Table 1's high school or less versus above high school", "reason": "the Methods say less than high school versus high school or above, but Table 1's counts (5,690 and 5,323) fit its own split; the Methods' split is a variant" }, { "choice": "age", "decision": "three groups: under 60, 60 to 79, 80 or older", "reason": "the Methods divide age into these groups; continuous age is a variant" }, { "choice": "living status", "decision": "married or living with a partner versus widowed, divorced, separated, or never married (DMDMARTL; DMDMARTZ in 2021-2023)", "reason": "Methods" }, { "choice": "race, BMI, smoking", "decision": "RIDRETH1's five groups; BMI under 25, 25 to under 30, 30 or more; never, former, current smoking from SMQ020 and SMQ040", "reason": "Methods and Table 1" }, { "choice": "2021-2023 items", "decision": "never drinker from ALQ111 and heavy drinking from ALQ130 and ALQ142, as in 2017-2018; oscillometric readings 2 and 3 (BPXO); medication from BPQ150", "reason": "2021-2023 asks the 2017-2018 alcohol items and measures blood pressure only by the oscillometric device" } ], "departures": [ { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The paper's definitions class men who drank before but not in the past year as mild drinkers, but its n of 11,013 (492 with prostate cancer) and Table 1's margins are matched (11,022 and 493 here, every margin within 20 men) only if the 333 such men of 2017-2018, whose drinks-a-day and binge items (ALQ130, ALQ142) are skipped, are dropped. Kept as mild drinkers, they give 11,355 men (509 with prostate cancer) and Model III 1.90 (1.11-3.27), against 1.92 (1.12-3.29) without them." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods define heavy drinking as 3 or more drinks a day, but Table 1's 2,048 heavy drinkers (31 with prostate cancer) are those of 4 or more for men (2,068 and 33 here), not of 3 (3,164 and 65); Model III is 1.92 either way." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods define hypertension by a diastolic pressure of 80 mmHg or more, but Table 1's 5,909 hypertensive men (358 with prostate cancer) are matched exactly with 90 or more over the second and third readings, while 80 gives 6,855 (363); Model III is 1.92 with 90 and 1.94 with 80." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods split education into less than high school and high school or above, but Table 1's counts (5,690 and 5,323) are those of its own split, high school or less and above high school (5,693 and 5,329 here), not the Methods' (3,141 and 7,881); Model III is 1.92 with Table 1's split and 1.93 with the Methods'." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods put an income-to-poverty ratio of exactly 3 in the high group, but Table 1's counts (1,898, 4,525, and 4,590) are matched within 3 men in each group with it in the middle group, and are off by 7 and 13 in the middle and high groups with it in the high group (10 men have exactly 3)." } ], "projected_power": 0.114642528351159 }, "row221": { "id": "row221", "row": 221, "doi": "10.1016/j.heliyon.2024.e27764", "measure": "OR", "contrast": "sitting 7 hours a day or more against less than 7 hours", "published": { "measure": "OR", "estimate": 1.2, "low": 1.1, "high": 1.3, "n": 22916, "events": 5222, "contrast": "sitting 7 hours a day or more against less than 7 hours" }, "published_b": [0.182321556793955, 0.0953101798043249, 0.262364264467491], "constants": null, "original": { "estimate": 1.18033346592798, "ci": [1.07029816140698, 1.30168128940642], "b": 0.165796996798265, "se": 0.0492866160921, "df": 94, "p": 0.00111331288732618, "n": 22916, "events": 5222, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, "harmonized": { "estimate": 1.18067922241956, "ci": [1.07076615231936, 1.30187475877318], "b": 0.166089885096372, "se": 0.0492139556675322, "df": 94, "p": 0.00107469191869871, "n": 22916, "events": 5222, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.9093658463307, "published_z": 4.27817955654234, "interval_ratio_original": 0.85353843657523, "ratio_harmonized": 1.00176654766831, "variants": [ { "label": "unweighted", "estimate": 1.20279679692255, "ci": [1.12372986831237, 1.28742696575277], "b": 0.18464950911119, "se": 0.0346925340174768, "df": null, "p": 1.02375173877141e-07, "n": 22916, "events": 5222, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude (published 1.1, 1.0-1.2)", "estimate": 1.11008681437023, "ci": [1.01665933371531, 1.21209995774624], "b": 0.104438223410264, "se": 0.0442785677420404, "df": 94, "p": 0.020411559065016, "n": 22916, "events": 5222, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "Model 2 (published 1.1, 1.0-1.2)", "estimate": 1.11455609303231, "ci": [1.00939068030798, 1.23067837731217], "b": 0.108456202799167, "se": 0.0499159761756973, "df": 94, "p": 0.0323115073026899, "n": 22916, "events": 5222, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "lifetime abstainers kept, as drinking on 0 days a week", "estimate": 1.1438441603984, "ci": [1.04526203591653, 1.251723891541], "b": 0.134394660250377, "se": 0.045392087694861, "df": 94, "p": 0.0038847170903652, "n": 26478, "events": 6085, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "lifetime abstainers kept, crude", "estimate": 1.08920774753678, "ci": [1.00519630441982, 1.18024062770395], "b": 0.0854505948446627, "se": 0.0404264905335179, "df": 94, "p": 0.0371865124281263, "n": 26478, "events": 6085, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "lifetime abstainers kept, Model 2", "estimate": 1.08294487182034, "ci": [0.988669379190831, 1.1862100921562], "b": 0.079684063510257, "se": 0.0458717146961802, "df": 94, "p": 0.0856439038929959, "n": 26478, "events": 6085, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "pregnant women kept", "estimate": 1.18434443493711, "ci": [1.07559706867752, 1.3040866151589], "b": 0.169189402042392, "se": 0.0485078543487285, "df": 94, "p": 0.000742410952798499, "n": 23144, "events": 5264, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "urge without stress leakage (Table 1's UUI group)", "estimate": 1.13131535202535, "ci": [1.02357551562817, 1.25039570230711], "b": 0.123380984109556, "se": 0.0504043804494819, "df": 94, "p": 0.0162283841816068, "n": 22916, "events": 2951, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "marital status left out", "estimate": 1.19159144878253, "ci": [1.08036818998202, 1.31426507553437], "b": 0.175289765582941, "se": 0.0493511503463071, "df": 94, "p": 0.000599983547025325, "n": 22916, "events": 5222, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 184, "strata": 90 } ], "left_out": "marital status in six groups (2021-2023's DMDMARTZ has three: married or living with a partner; widowed, divorced, or separated; never married); the harmonized version uses those three, which leaves the estimate on the paper's cycles unchanged (1.144 either way)", "choices": [ { "choice": "outcome", "decision": "any urge leakage (KIQ044 yes against no), mixed UI included", "reason": "unstated (Table 1's groups are exclusive); on the paper's population this coding reproduces Table 2's urgency UI estimates (crude 1.11, p 0.02; Model 1 1.18, 1.07-1.30, p 0.001; Model 2 1.12, p 0.03), and coding stress UI as any stress leakage and mixed UI as both reproduces theirs; urge without stress leakage gives 1.13 (p 0.02)" }, { "choice": "weights", "decision": "examination weight (WTMEC2YR), each cycle's over 6", "reason": "the leakage questions are asked in the examination center; weighted fits reproduce Table 2, unweighted ones do not (crude 1.14, p < 0.001, against the published p 0.02)" }, { "choice": "lifetime abstainers", "decision": "dropped, as the paper's computation dropped them", "reason": "the paper's drinking variable left them missing (the alcohol items skip them), so its 'unknown covariates' step dropped 3562 people whose other covariates were known; with them dropped the population is the paper's 22,916 (12,148 men, 10,768 women) exactly and every Table 2 estimate is reproduced (Model 1 1.18, 1.07-1.30). Keeping them, as drinking on 0 days a week, is a variant" }, { "choice": "pregnant women", "decision": "excluded (RIDEXPRG 1)", "reason": "unstated, but the paper's counts leave them out: without them its 10,768 women and Table 1's counts by race and UI group are matched exactly; with them there would be 228 more" }, { "choice": "missing UI or sitting time", "decision": "KIQ042 and KIQ044 each yes or no, and PAD680 0-1380 minutes", "reason": "gives the paper's n = 4863 and 29,907 exactly" }, { "choice": "unknown covariates", "decision": "complete data on every Model 2 covariate", "reason": "with the paper's alcohol coding this gives its 22,916 (12,148 men, 10,768 women) exactly" }, { "choice": "covariate coding", "decision": "age continuous; RIDRETH1 five groups; DMDEDUC2 1-2, 3, 4-5; income-to-poverty ratio under 1.3, 1.3 to 3.5, 3.5 or more; DMDMARTL six groups", "reason": "Table 1's categories, whose counts these codings match" }, { "choice": "drinking", "decision": "drinking days a week (under 1, 1 to 3, 4 or more)", "reason": "the paper says drinks a week, but Table 1's counts are those of drinking days (men exactly, women within one), not of drinks; only Model 2 and the population step use it" }, { "choice": "recreational activity in 2021-2023", "decision": "any moderate or vigorous leisure-time activity (PAD790Q, PAD810Q above 0)", "reason": "2021-2023 asks how often, not yes or no for a typical week of 10-minute bouts; it enters only the complete-data step" }, { "choice": "Model 2's diabetes and hypertension", "decision": "diabetes: told, medication, HbA1c 6.5% or more, or fasting glucose 126 mg/dL or more; hypertension: told, medication, or a mean pressure of 140/90 mmHg or more", "reason": "unstated; Table 1's counts (diabetes 2427 men, 1806 women; hypertension 5370, 4497) are far above those told alone and near these definitions' (2309, 1672; 5348, 4482)" } ], "departures": [ { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The text excludes people with unknown covariates, but its population also lacks the 3,562 lifetime abstainers, whom the alcohol items skip and whose other covariates are known. Without them the population is the paper's 22,916 (12,148 men, 10,768 women) exactly and every Table 2 estimate is reproduced (Model 1 1.18, 1.07-1.30); kept, as drinking on 0 days a week, they give 26,478 people and 1.14 (1.05-1.25)." }, { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The text lists three exclusions (under 20, no answer on UI or sitting time, unknown covariates), but its population also leaves out pregnant women: Table 1's 10,768 women and its counts by race and UI group are matched exactly without them, and keeping them adds 228 (Model 1 1.18 either way)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text groups alcohol use by drinks a week (under 1, 1 to 3, 4 or more), but Table 1's counts in those groups are those of drinking days a week (men exactly, women within one). Alcohol is in Model 2 and reaches Model 1 only through the complete-case step, where drinking days keep 11 drinkers who gave no number of drinks a day, whom drinks a week would leave out." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The text and Fig. 1 exclude 25,372 people under 20, but the 59,842 participants less the 34,770 adults Fig. 1 next shows, which the data reproduce, leave 25,072." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The Results give the headline estimate (OR 1.2, 1.1 to 1.3, p = 0.001) as that of men, but Table 2 prints it in the whole-population column, where it is reproduced here (1.18, 1.07-1.30); Table 2's Model 1 estimate for men is 1.2 (0.99 to 1.4), p = 0.1." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's title calls it weighted, but its percentages are unweighted shares (1,798 of 12,148 men, 14.8%, are Mexican American), as are the weighted proportions with UI that the Results give (2,202 of 12,148 men, 18.1%; 5,860 of 10,768 women, 54.4%)." } ], "projected_power": 0.372776395140947 }, "row236": { "id": "row236", "row": 236, "doi": "10.1186/s12944-024-02306-7", "measure": "OR", "contrast": "TyG-BMI quartile 4 (287.23-679.46) vs quartile 1 (112.58-204.87)", "published": { "measure": "OR", "estimate": 2.36, "low": 2.03, "high": 2.78, "n": 18751, "events": 4352, "contrast": "TyG-BMI quartile 4 (287.23-679.46) vs quartile 1 (112.58-204.87)" }, "published_b": [0.858661619037519, 0.708035793053696, 1.02245092770255], "constants": null, "original": { "estimate": 2.420536431289, "ci": [2.04908029692893, 2.8593299266888], "b": 0.883989181426257, "se": 0.0842605712798975, "df": 139, "p": 2.50668411479929e-19, "n": 16314, "events": 3766, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 274, "strata": 135 }, "harmonized": { "estimate": 2.41394886754055, "ci": [2.04359413790002, 2.85142192719746], "b": 0.881263940819088, "se": 0.0842381788961158, "df": 139, "p": 2.98485348765765e-19, "n": 16332, "events": 3776, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 274, "strata": 135 }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.02949655816354, "published_z": 10.7052470619772, "interval_ratio_original": 0.943633475549187, "ratio_harmonized": 0.996917110905394, "variants": [ { "label": "unweighted", "estimate": 2.34636025854165, "ci": [2.07009869748623, 2.65948984439679], "b": 0.852865301332441, "se": 0.063913937314604, "df": null, "p": 1.2844319302346e-40, "n": 17186, "events": 3969, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1, crude (published 1.86, 1.63-2.11)", "estimate": 1.89492224402024, "ci": [1.64681525460926, 2.18040870148162], "b": 0.639177805513254, "se": 0.0709771962942922, "df": 139, "p": 1.47621582778433e-15, "n": 16314, "events": 3766, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "Model 1, crude, unweighted", "estimate": 2.14204229923761, "ci": [1.93545171324271, 2.37068441456269], "b": 0.761759719373814, "se": 0.0517453266485843, "df": null, "p": 4.70360229177766e-49, "n": 17186, "events": 3969, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2, age, sex, race (published 2.46, 2.14-2.84)", "estimate": 2.5572952167492, "ci": [2.18582037281538, 2.99190130485654], "b": 0.938950143985423, "se": 0.0793853416887279, "df": 139, "p": 9.07363346073431e-23, "n": 16314, "events": 3766, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "Model 2, unweighted", "estimate": 2.51535793624482, "ci": [2.23683443737345, 2.82856229397969], "b": 0.922415113999571, "se": 0.0598753069775456, "df": null, "p": 1.50086242089846e-53, "n": 17186, "events": 3969, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "examination weight (WTMEC2YR)", "estimate": 2.34437725825496, "ci": [1.99053729148863, 2.76111618331589], "b": 0.852019805118796, "se": 0.0827518968452512, "df": 139, "p": 7.92457408187695e-19, "n": 17186, "events": 3969, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "Q3-Q4 cutpoint as printed (287.23)", "estimate": 2.41704802355692, "ci": [2.04609150219108, 2.85525898618137], "b": 0.882546970600052, "se": 0.0842694005340642, "df": 139, "p": 2.79136425524861e-19, "n": 16314, "events": 3766, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "diabetes counts HbA1c 6.5% itself", "estimate": 2.42171135382413, "ci": [2.04888950260577, 2.8623729458236], "b": 0.884474461242712, "se": 0.0845531076156472, "df": 139, "p": 3.00244196179715e-19, "n": 16314, "events": 3766, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "hypertension over 140/90 mmHg, as printed", "estimate": 2.42474512257664, "ci": [2.05341447760295, 2.86322560476075], "b": 0.885726414773516, "se": 0.0840705447492069, "df": 139, "p": 1.92859074326253e-19, "n": 16314, "events": 3766, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 274, "strata": 135 }, { "label": "missing covariates as their own level, for imputation", "estimate": 2.36166009127311, "ci": [2.01602049699281, 2.76655837330605], "b": 0.859364800253201, "se": 0.0800329080064978, "df": 139, "p": 5.83566149765784e-20, "n": 17759, "events": 4124, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 274, "strata": 135 } ], "left_out": ["the requirement that the question on leakage during nonphysical activities (KIQ046) be answered: 2021-2023 does not ask it (on the paper's cycles it removes 20 people)", "being told to take cholesterol medicine (BPQ090D) as a criterion of high cholesterol: 2021-2023 asks only whether one takes it (BPQ101D), which the harmonized version uses in every cycle (BPQ100D before 2021)"], "choices": [ { "choice": "cycles", "decision": "the nine cycles 2001-2002 to 2017-2018, without the 2017-March 2020 files", "reason": "the paper's 91,351 participants are the nine cycles' DEMO files exactly, and its 41,150 under 20 leave 50,201 adults, as here" }, { "choice": "population", "decision": "adults answering yes or no to all three leakage items (KIQ042, KIQ044, KIQ046), with BMI, triglycerides, fasting glucose, and waist circumference, not pregnant (RIDEXPRG)", "reason": "the paper names only the UI items and BMI, triglycerides, and glucose, but only with KIQ046 and waist circumference also required are its steps (6,365, 24,593, 492) and its 18,751 matched, with Table 1's counts (sex, race, education, marital status, income, smoking, alcohol, activity, insurance, UI types, missing counts) exact and its TyG-BMI range (112.58-679.46) too; without them there are 19,138" }, { "choice": "weights", "decision": "fasting subsample weight (WTSAF2YR) over 9 cycles, strata and PSUs", "reason": "the paper says it weighted but not with which weight; the exposure needs fasting labs, so NCHS's guidelines call for the fasting weight; with it the abstract's UUI and MUI prevalences (19.42%, 9.32%) are matched exactly (SUI 23.44% against 23.59%), and the crude (1.89 against 1.86) and Model 2 (2.56 against 2.46) estimates are reproduced, while the unweighted crude estimate (2.14) is not" }, { "choice": "missing covariates", "decision": "complete cases", "reason": "the paper imputed them (multiple imputation, no details); keeping everyone, with missing values as their own level, gives 2.36 (2.02-2.77), the published estimate (variant)" }, { "choice": "stress UI", "decision": "KIQ042 yes against no, with or without urge leakage", "reason": "Table 1's SUI (4352), UUI (4264), and MUI (1970) counts are reproduced exactly with overlapping groups" }, { "choice": "quartiles", "decision": "the paper's cutpoints 204.87 and 242.54, and 287.30 between Q3 and Q4 where it prints 287.23", "reason": "the paper's quartiles are equal groups of its sample (4688, 4687, 4688, 4688), whose third quartile is 287.30 on the reproduced sample; with 287.23 six people move to Q4 (4682 and 4693), while with 287.30 Table 1's SUI, UUI, and MUI counts by quartile are matched exactly (variant: as printed)" }, { "choice": "alcohol use", "decision": "at least 12 drinks in any one year (ALD100 in 2001-2002, ALQ101 in 2003-2016); ever a drink (ALQ111) in 2017-2018 and 2021-2023", "reason": "the paper's item was not asked from 2017 on; ever a drink gives Table 1's 13,793 yes, 4,907 no, and 51 missing exactly" }, { "choice": "vigorous and moderate activity", "decision": "leisure-time only: PAD200 and PAD320 (2001-2006), PAQ650 and PAQ665 (2007-2018), any PAD810Q or PAD790Q (2021-2023)", "reason": "unstated; leisure alone gives Table 1's counts exactly (4,530 and 8,131 yes; 2 and 7 missing)" }, { "choice": "diabetes", "decision": "told, insulin or pills (DID070 in 2005-2008), or HbA1c over 6.5% as printed", "reason": "gives Table 1's 2,900 within 2; counting 6.5% itself gives 2,978 (variant)" }, { "choice": "hypertension", "decision": "told, taking medication, or a mean of all readings of 140/90 mmHg or more", "reason": "the paper prints over 140/90; at or over gives 8,092 and over 8,015 against Table 1's 8,182 (neither exact; which readings is unstated), so the conventional threshold, nearer Table 1, is used (variant: as printed)" }, { "choice": "high cholesterol", "decision": "told (BPQ080), told to take medicine (BPQ090D), or total cholesterol 240 mg/dL or more", "reason": "the paper's sentence is garbled; this reading gives Table 1's 7,659 and 11,092 exactly, while taking medicine (BPQ100D) gives 7,605" }, { "choice": "covariate coding", "decision": "age continuous; RIDRETH1 five groups; DMDEDUC2 1-2, 3, 4-5; marital status in three groups; income-to-poverty ratio under 1.3, 1.3 to 3.5, 3.5 or more; smoking never, former, current; insurance HID010 (2001-2004) or HIQ011", "reason": "Table 1's categories, whose counts these codings give exactly" } ], "departures": [ { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The text excludes those with incomplete UI data and those lacking BMI, triglycerides, or fasting glucose, but its exclusion counts (6,365, 24,593, and 492 pregnant, leaving 18,751) and Table 1's counts are matched only if waist circumference is also required and the third leakage item (KIQ046, during nonphysical activities), which none of its outcomes uses, is also answered; without these two requirements 19,138 remain." }, { "kind": "reporting", "affects_headline": true, "followed": true, "detail": "The Results and Table 1 give Q4 as 287.23 to 679.46, but Table 1's equal quartile sizes (4,688, 4,687, 4,688, 4,688) put the Q3-Q4 boundary at 287.30 on the reproduced population, where its SUI, UUI, and MUI counts by quartile are matched exactly; at 287.23 six people move into Q4 (4,682 and 4,693), and Model 3 is 2.42 either way." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The Results give the sex split (49.45% women) as weighted proportions, but it is unweighted (9,273 of 18,751), as are Table 1's other shares (SUI 4,352 of 18,751, 23.21%)." } ], "projected_power": 0.914998575073168 }, "row249": { "id": "row249", "row": 249, "doi": "10.1186/s40795-024-00938-7", "measure": "OR", "contrast": "energy intake quartile 4 (> 33.29 kcal/kg/day) vs quartile 1 (< 17.73 kcal/kg/day)", "published": { "measure": "OR", "estimate": 0.61, "low": 0.53, "high": 0.69, "n": 21354, "events": 3154, "contrast": "energy intake quartile 4 (> 33.29 kcal/kg/day) vs quartile 1 (< 17.73 kcal/kg/day)" }, "published_b": [-0.49429632181478, -0.63487827243597, -0.371063681390832], "constants": null, "original": { "estimate": 0.611708769381904, "ci": [0.532757191821019, 0.70236052048347], "b": -0.491498976753489, "se": 0.0705066690008874, "df": null, "p": 3.14791584998127e-12, "n": 21354, "events": 3154, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 0.623487575880591, "ci": [0.545642276137302, 0.712438852849512], "b": -0.472426440475028, "se": 0.0680447449253054, "df": null, "p": 3.84188083643483e-12, "n": 21800, "events": 3269, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.99434075282814, "published_z": -7.34457472279713, "interval_ratio_original": 0.954532077544391, "ratio_harmonized": 0.961195165848684, "variants": [ { "label": "weighted (day-one dietary weight)", "estimate": 0.692152999800099, "ci": [0.557509276304492, 0.85931444640326], "b": -0.367948249826001, "se": 0.10866122779497, "df": 78, "p": 0.00111243699964906, "n": 21354, "events": 3154, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 152, "strata": 74 }, { "label": "unadjusted (published 0.74, 0.67-0.83)", "estimate": 0.744206657606935, "ci": [0.670079423590696, 0.826534183453438], "b": -0.295436517118048, "se": 0.0535328784914619, "df": null, "p": 3.41348965224578e-08, "n": 21354, "events": 3154, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1 (published 0.77, 0.69-0.86)", "estimate": 0.769829982613108, "ci": [0.688464332075496, 0.860811772693418], "b": -0.261585590315873, "se": 0.0569939947799889, "df": null, "p": 4.43874090380307e-06, "n": 21354, "events": 3154, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2 (published 0.76, 0.68-0.85)", "estimate": 0.766013853841569, "ci": [0.684601120385626, 0.857108185780779], "b": -0.266555023450044, "se": 0.0573295702209594, "df": null, "p": 3.32707292429632e-06, "n": 21354, "events": 3154, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 3 with activity and smoking, as the Methods list it", "estimate": 0.603142595876592, "ci": [0.524828995323819, 0.693141945666134], "b": -0.505601632804043, "se": 0.0709610793543006, "df": null, "p": 1.04038714070997e-12, "n": 21354, "events": 3154, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "energy as the two-day mean the Methods describe", "estimate": 0.653393097076416, "ci": [0.567009296022063, 0.752937460289009], "b": -0.425576344492314, "se": 0.0723499191384919, "df": null, "p": 4.04860540607733e-09, "n": 18816, "events": 2794, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "n-3 and n-6 PUFAs in mg/kg/day (Table 1's unit)", "estimate": 0.631811889978148, "ci": [0.543782970819882, 0.734091146171589], "b": -0.459163571579221, "se": 0.0765531877275445, "df": null, "p": 1.99801859171806e-09, "n": 21354, "events": 3154, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "smoking and activity not required (n 21,366)", "estimate": 0.6122152110316, "ci": [0.533241745251073, 0.70288470090051], "b": -0.49067140627247, "se": 0.0704650680524792, "df": null, "p": 3.32326011480804e-12, "n": 21366, "events": 3156, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "unadjusted, weighted", "estimate": 0.817109228152962, "ci": [0.694211374326625, 0.961763974812942], "b": -0.201982498868733, "se": 0.081872834219983, "df": 78, "p": 0.015815839138918, "n": 21354, "events": 3154, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 152, "strata": 74 } ], "left_out": "familial asthma (MCQ300B, a close relative ever told they had asthma): 2021-2023 does not ask it; it leaves the model and the complete-case step", "choices": [ { "choice": "weights", "decision": "unweighted", "reason": "the paper never mentions weights, strata, or PSUs; unweighted fits reproduce Table 2's unadjusted, Model 1, Model 2, and Model 3 quartile 4 estimates within 0.01, weighted ones do not (unadjusted 0.82 against 0.74)" }, { "choice": "energy intake", "decision": "the first day's recall (DR1TKCAL, reliable recalls) divided by measured weight (BMXWT)", "reason": "the Methods describe the mean of two recalls, but the authors' dataset holds only the first day's energy, and it matches this build person by person; the two-day mean is a variant" }, { "choice": "quartile cutpoints", "decision": "the published 17.73, 24.52, and 33.29 kcal/kg/day, each cutpoint opening the higher quartile", "reason": "Methods and Table 2; the Methods, Table 2's note, and Fig. 3 misprint Q3's lower bound as 21.52" }, { "choice": "study population", "decision": "aged 20 or older, asthma answered yes or no, first-day energy and weight known, and BMI, smoking status, and physical activity known besides the Model 3 covariates", "reason": "the paper drops anyone missing a covariate it describes; requiring these gives the published 21,354 people and 3,154 with asthma exactly" }, { "choice": "familial asthma", "decision": "MCQ300B yes or no, refused or don't know as missing", "reason": "unstated; complete-case analysis" }, { "choice": "income to poverty ratio", "decision": "the paper's three groups: 1.3 or less, over 1.3 to 3.5, over 3.5", "reason": "Methods" }, { "choice": "race and ethnicity", "decision": "RIDRETH1's five groups", "reason": "Table 1 shows five; the authors' dataset codes RIDRETH1" }, { "choice": "n-3 and n-6 PUFAs", "decision": "g/day; n-3 is 18:3, 18:4, 20:5, 22:5, and 22:6, n-6 is 18:2 and 20:4", "reason": "these sums equal the authors' n3 and n6 columns for every participant; Table 1 labels them mg/kg/day, but its medians are g/day (mg/kg/day is a variant)" }, { "choice": "Model 3 covariates", "decision": "Table 2's note: no physical activity or smoking", "reason": "the Methods build Models 2 and 3 on Model 1, which has both; the note's list gives Model 3's printed quartile 4 estimate (0.612 against 0.603 with them), though with them Model 2's (0.762 against 0.766 for a printed 0.76) fits better; both reproduce the headline, and the Methods' version is a variant" }, { "choice": "physical activity", "decision": "vigorous or moderate recreational activity (PAQ650, PAQ665) for 2009-2018; how often one does vigorous or moderate leisure-time activity (PAD810Q, PAD790Q) for 2021-2023", "reason": "the paper's three groups (sedentary, moderate, vigorous) without saying which items; it enters only Model 1 and the complete-case step" }, { "choice": "pregnancy", "decision": "pregnant women kept", "reason": "the paper names no such exclusion, and its sample size is reproduced without one" }, { "choice": "2021-2023 recall", "decision": "the first day's recall as released", "reason": "2021-2023 took both recalls by telephone, where 2009-2018 took the first in person at the examination; nothing else in the exposure changes" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods average energy intake over the two 24-hour recalls, but the authors' dataset (Supplementary Material 2) holds only the first day's energy, which matches this build person by person, and the first day's energy gives the paper's n of 21,354 and Table 2 (Model 3 0.612, 0.533-0.702); the two-day mean leaves 18,816 people and gives 0.653 (0.567-0.753)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "Table 1 and Fig. 2's list of adjustments give n-3 and n-6 PUFAs in mg/kg/day, but Table 1's medians (1.5 and 14.4) and the n3 and n6 columns of the authors' dataset are in g/day, and Model 3 is reproduced with them in g/day (0.612; in mg/kg/day, 0.632)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The Methods, Table 2's note, and Fig. 3 give Q3 as 21.52 to 33.29 kcal/kg/day, but Q2 ends at 24.52, which is also the second quartile boundary of the authors' dataset." } ], "projected_power": 0.866147459493095 }, "row256": { "id": "row256", "row": 256, "doi": "10.1038/s41598-024-66922-0", "measure": "OR", "contrast": "per unit of the CDAI (sum of the six intakes, each minus its mean over its SD, means and SDs from the paper's sample)", "published": { "measure": "OR", "estimate": 0.98, "low": 0.97, "high": 0.99, "n": 30788, "events": 25525, "contrast": "per unit of the CDAI (sum of the six intakes, each minus its mean over its SD, means and SDs from the paper's sample)" }, "published_b": [-0.0202027073175195, -0.0304592074847086, -0.0100503358535015], "constants": { "cdai": { "vitamin_a": { "mean": 599.996485290289, "sd": 554.672280270301 }, "vitamin_c": { "mean": 82.106765351285, "sd": 78.0821601024575 }, "vitamin_e": { "mean": 8.04491222523896, "sd": 5.23836805279439 }, "zinc": { "mean": 10.8486934206122, "sd": 6.07528057508883 }, "selenium": { "mean": 110.498910254026, "sd": 51.3159969984584 }, "carotenoids": { "mean": 8949.99633651951, "sd": 9777.14755118714 } } }, "original": { "estimate": 0.981688044226374, "ci": [0.969166283026363, 0.994371588297184], "b": -0.0184816950006108, "se": 0.00654980612077336, "df": null, "p": 0.00477674061166176, "n": 19096, "events": 16043, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 0.981577357481299, "ci": [0.969113199095698, 0.994201822469273], "b": -0.0185944528020732, "se": 0.00652022214739212, "df": null, "p": 0.00434706964931546, "n": 19179, "events": 16110, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.91481278771899, "published_z": -3.8803300298087, "interval_ratio_original": 0.794899937103803, "ratio_harmonized": 1.0061010530397, "variants": [ { "label": "stacked 2017-2018 and 2017-March 2020 files, as the paper did", "estimate": 0.981081961569095, "ci": [0.96934081899333, 0.99296531875739], "b": -0.0190992739067009, "se": 0.0061428336092494, "df": null, "p": 0.0018759720271156, "n": 21733, "events": 18221, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1, crude (published 0.97, 0.96-0.98)", "estimate": 0.969807541714, "ci": [0.960754828862379, 0.978945553756957], "b": -0.0306576377748202, "se": 0.00478497855606659, "df": null, "p": 1.4835385838763e-10, "n": 19096, "events": 16043, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1, crude, own sample", "estimate": 0.968987579595322, "ci": [0.960759690780713, 0.977285931560076], "b": -0.031503484928605, "se": 0.00435083401720189, "df": null, "p": 4.46071485822336e-13, "n": 22809, "events": 19019, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2, age and sex (published 0.98, 0.97-0.98)", "estimate": 0.977464455070937, "ci": [0.968030743153579, 0.986990100969905], "b": -0.0227933508825558, "se": 0.00494809187701693, "df": null, "p": 4.09516725790084e-06, "n": 19096, "events": 16043, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "weighted (dietary day-one weight)", "estimate": 0.984652825715358, "ci": [0.965376717678589, 1.00431382840956], "b": -0.0154661611417797, "se": 0.00998470425719564, "df": 119, "p": 0.124040665680583, "n": 19096, "events": 16043, "weighted": true, "weights": ["WTDRD1", "WTDRD1PP"], "dropped": [], "psu": 233, "strata": 114 }, { "label": "hyperlipidemia as stated, criteria not measured counted as not met", "estimate": 0.988295854587432, "ci": [0.97825098268764, 0.99844386919116], "b": -0.0117731780968747, "se": 0.00521225662283667, "df": null, "p": 0.0238990185451095, "n": 22405, "events": 16043, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "stacked, and hyperlipidemia as stated", "estimate": 0.987598116566492, "ci": [0.978203296555975, 0.997083165922321], "b": -0.0124794285942473, "se": 0.004876789747792, "df": null, "p": 0.0104990855121502, "n": 25533, "events": 18221, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "unknown cholesterol medicine as none (no medicine step)", "estimate": 0.979920918717566, "ci": [0.967767156401422, 0.992227314792214], "b": -0.0202834057597078, "se": 0.00636764547028486, "df": null, "p": 0.00144561280416556, "n": 21354, "events": 18301, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "two reliable recalls required", "estimate": 0.979636653578323, "ci": [0.965872478401119, 0.993596975268182], "b": -0.0205735377213883, "se": 0.00721948250613402, "df": null, "p": 0.0043757083448423, "n": 17000, "events": 14347, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": "physical activity as total MET-minutes a week over work, travel, and leisure (2021-2023 asks only about leisure-time activity)", "choices": [ { "choice": "cycles", "decision": "2005-2006 through 2015-2016 and the 2017-March 2020 files, each participant once", "reason": "the flow chart's 85,750, 33,914 under 18, and 51,836 adults are exactly the 2005-2018 files stacked with the 2017-March 2020 files, which repeat the 2017-2018 participants; a variant stacks them" }, { "choice": "missing dietary data", "decision": "no reliable first-day recall (DR1DRSTZ not 1)", "reason": "gives the flow chart's 6261 exactly; requiring both days would exclude 12,119" }, { "choice": "daily intakes", "decision": "mean of the two recalls, the first day's where the second is missing or unreliable, for the CDAI and energy", "reason": "the paper averages two days yet keeps everyone with a first-day recall; this gives Table 1's CDAI median and quartiles (-0.63, -2.68 to 1.94; published -0.61, -2.68 to 1.94) and its energy quartiles closely; a variant requires two recalls" }, { "choice": "carotenoids and vitamins", "decision": "carotenoids as alpha- plus beta-carotene, beta-cryptoxanthin, lycopene, and lutein with zeaxanthin; vitamin A as RAE; vitamin E as alpha-tocopherol", "reason": "unstated beyond 'total carotenoids'; the library's CDAI components, which reproduce Table 1's CDAI quartiles" }, { "choice": "standardizing means and SDs", "decision": "unweighted, over the paper's sample (in_population)", "reason": "the paper standardizes over its participants and reports no weights" }, { "choice": "missing hyperlipidemia data", "decision": "TC or HDL unmeasured, or cholesterol medicine unknown (BPQ080 unanswered, or told of high cholesterol with the medicine questions unanswered)", "reason": "gives 8492 against the flow chart's 8497; TC and HDL alone give 2505. A variant drops the medicine step" }, { "choice": "special diet and energy", "decision": "DRQSDIET = 1; mean energy under 500 or 5000 kcal or more", "reason": "gives 6061 and 351 against the flow chart's 5914 and 376" }, { "choice": "hyperlipidemia", "decision": "any criterion met (HDL 40 or less for men, 50 or less for women, as stated; medicine BPQ100D, BPQ101D in 2021-2023), and undetermined where a criterion not measured (TG and LDL outside the fasting subsample) could have been met, as the paper's computation coded it; the complete-case model leaves the undetermined out, where the paper imputed them", "reason": "the stated definition, with criteria not measured counted as not met, gives 70.7% against the paper's 82.91% (25,525 of 30,788); leaving undetermined those whom an unmeasured criterion could decide gives 83.4%, so the published estimate comes from that coding; the stated definition is a variant" }, { "choice": "weights", "decision": "unweighted", "reason": "the paper never mentions weights or design and its Table 1 counts are unweighted; a variant uses the dietary day-one weight" }, { "choice": "missing covariates", "decision": "complete cases", "reason": "the paper used unspecified multiple imputation; complete cases also drop every 18-19 year old, whom NHANES asks no adult education or marital status" }, { "choice": "marital status", "decision": "married or living with a partner against single", "reason": "2017-March 2020 groups living with a partner with married; that grouping gives Table 1's 59.86% married (60.5%), leaving partners out gives 54.8%" }, { "choice": "alcohol", "decision": "the paper's three groups, never and former drinkers as mild, without the library's daily binge (ALQ151) item", "reason": "the stated definition; it gives 65.9, 15.4, 18.7% against Table 1's 63.6, 16.3, 20.1 (the library's alcohol5 gives 62.7, 13.6, 23.6)" }, { "choice": "hypertension", "decision": "mean blood pressure 140/90 or more, told, or taking medicine (library hypertension_status)", "reason": "the stated definition; 'exceeding' read as at or above; gives 42.3% against Table 1's 41.85%" }, { "choice": "diabetes", "decision": "told, insulin, or diabetes pills", "reason": "the stated definition, with no glucose or HbA1c criterion" }, { "choice": "physical activity", "decision": "total GPAQ MET-minutes a week (library met_minutes) in 2007-2020; in 2005-2006 leisure activities (PAQIAF), travel, and home or yard tasks", "reason": "the stated definition; 2005-2006 asked no GPAQ, and leaving it missing would drop the cycle from a complete-case analysis. Table 1's median (2160) is about twice the GPAQ's (1100), which no stated scoring explains; leaving activity out changes the estimate by under 0.001" }, { "choice": "covariate forms", "decision": "age, BMI, MET-minutes, and energy continuous; income-to-poverty ratio in the paper's three groups (under 1.5, 1.5-3.5, over 3.5); education in three", "reason": "Table 1 reports age, BMI, activity, and energy as means or medians and the others as groups" } ], "departures": [ { "kind": "sample", "affects_headline": true, "followed": false, "detail": "The paper analyzes NHANES 2005-2020, but the 85,750 participants of its flow chart (Fig. 1; 33,914 under 18, 51,836 adults) are exactly the 2005-2018 files together with the 2017-March 2020 files, which hold the 2017-2018 participants a second time. Stacked that way the population here is 30,755 (published 30,788), 3,868 of them second copies, and Model 3 gives 0.9876 (0.9782-0.9971), against 0.9883 (0.9783-0.9984) with each participant once." }, { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The text gathers participants who had two dietary recalls and averages the two days, but its 6,261 excluded for missing dietary data are exactly those without a reliable first-day recall (requiring both recalls would exclude 12,119), so its population includes those with one; requiring two gives 19,895 people and Model 3 0.9866 (0.9756-0.9977)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text defines hyperlipidemia as any of five criteria, two of which (TG and LDL) NHANES measures only in its fasting subsample, and its population keeps those outside it (its 8,497 without hyperlipidemia data are matched by those without TC, HDL, or medicine status, 8,492 here). That definition gives 70.7% with hyperlipidemia here, against the paper's 82.91% (25,525 of 30,788); leaving undetermined those whom an unmeasured criterion could decide gives 83.4%, so the outcome was coded that way, with the undetermined filled in by the paper's imputation." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1 gives physical activity as a median of 2,160 MET-minutes a week (720 to 6,000), but the stated scoring gives 1,100 (0 to 3,840) over the population here, a quarter of whom report none, and 2,100 (724 to 5,700) among those with any activity." } ], "projected_power": 0.261055661684334 }, "row257": { "id": "row257", "row": 257, "doi": "10.1016/j.waojou.2024.100900", "measure": "OR", "contrast": "dietary zinc quartile 4 (11.96 mg/day or more) vs quartile 1 (5.68 or less)", "published": { "measure": "OR", "estimate": 0.71, "low": 0.53, "high": 0.95, "n": 4597, "events": 963, "contrast": "dietary zinc quartile 4 (11.96 mg/day or more) vs quartile 1 (5.68 or less)" }, "published_b": [-0.342490308946776, -0.63487827243597, -0.0512932943875506], "constants": null, "original": { "estimate": 1.40769406224119, "ci": [1.04306110839324, 1.8997952822933], "b": 0.341952948785907, "se": 0.152958517495609, "df": null, "p": 0.0253784616421539, "n": 4597, "events": 963, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 1.37037652259247, "ci": [1.02924500471468, 1.82457219133477], "b": 0.315085536089186, "se": 0.146053708137813, "df": null, "p": 0.0309802330337107, "n": 4597, "events": 963, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": false, "same_sign_original": false, "ratio_original": -0.998431020829403, "published_z": -2.30050016994777, "interval_ratio_original": 0.973312614212317, "ratio_harmonized": 0.921429504286735, "variants": [ { "label": "no asthma as the outcome", "estimate": 0.710381628240941, "ci": [0.526372504090478, 0.958716600545509], "b": -0.341952948785907, "se": 0.152958517495609, "df": null, "p": 0.0253784616421541, "n": 4597, "events": 3634, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude", "estimate": 1.3848773247217, "ci": [1.13083490582052, 1.69599045329851], "b": 0.325611561508214, "se": 0.103397485103208, "df": null, "p": 0.00163760369452858, "n": 4597, "events": 963, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, no asthma (published 0.72, 0.59-0.88)", "estimate": 0.722085618811729, "ci": [0.589625960485279, 0.884302381234341], "b": -0.325611561508216, "se": 0.103397485103208, "df": null, "p": 0.00163760369452851, "n": 4597, "events": 3634, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1, no asthma (published 0.76, 0.62-0.93)", "estimate": 0.755452246847383, "ci": [0.612965827158881, 0.931060218988082], "b": -0.280438706616604, "se": 0.106638380356915, "df": null, "p": 0.00854325359603811, "n": 4597, "events": 3634, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2, no asthma (published 0.74, 0.60-0.92)", "estimate": 0.740695067615643, "ci": [0.593583789246432, 0.924265778697627], "b": -0.300166253070197, "se": 0.112966689897928, "df": null, "p": 0.00788109661903766, "n": 4597, "events": 3634, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Q2 vs Q1", "estimate": 1.31142065811993, "ci": [1.04044772483407, 1.65296545082838], "b": 0.271111021445537, "se": 0.118093953822815, "df": null, "p": 0.0216917170026398, "n": 4597, "events": 963, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Q2 vs Q1, no asthma (published 0.78, 0.62-0.98)", "estimate": 0.762531834318985, "ci": [0.60497332203698, 0.961124692890728], "b": -0.271111021445537, "se": 0.118093953822815, "df": null, "p": 0.0216917170026398, "n": 4597, "events": 3634, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "weighted (dietary day-1 weight)", "estimate": 1.70724840933175, "ci": [1.15057863844891, 2.53324460733523], "b": 0.534882957144909, "se": 0.197955951663782, "df": 72, "p": 0.00858912002659542, "n": 4597, "events": 963, "weighted": true, "weights": ["WTDRD1", "WTDRD1PP"], "dropped": [], "psu": 140, "strata": 68 }, { "label": "complete cases (the paper's sensitivity analysis)", "estimate": 1.53944397107566, "ci": [1.081776326781, 2.1907372914447], "b": 0.43142129348542, "se": 0.18001190798806, "df": null, "p": 0.0165467846687764, "n": 3053, "events": 713, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "complete cases, no asthma", "estimate": 0.649585187112245, "ci": [0.456467329015311, 0.924405512714133], "b": -0.43142129348542, "se": 0.18001190798806, "df": null, "p": 0.0165467846687765, "n": 3053, "events": 2340, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": "family history of asthma (MCQ300B): 2021-2023 asks no family medical history", "choices": [ { "choice": "outcome coding", "decision": "asthma (MCQ010 yes) as the event, as the Methods define it", "reason": "the published ORs are those of not having asthma: Table 1's counts give crude ORs for asthma of 1.18, 1.21, 1.38 by quartile and for no asthma of 0.85, 0.83, 0.72, which Table 3 prints, and Table 3's n(%) column counts those without asthma; the variant with no asthma as the event shows that this explains every published estimate" }, { "choice": "missing covariates", "decision": "single imputation within the study population: round-robin regression of each incomplete covariate on all analysis variables, 10 rounds, predictions without random draws", "reason": "the paper's stated method (an iterative imputer with Bayesian ridge, scikit-learn's defaults); it keeps the paper's 4,597 and reproduces Models 1-3, which complete cases (3,053) do less well (variant)" }, { "choice": "weights", "decision": "unweighted", "reason": "unstated; the crude ORs equal the unweighted ones from Table 1's counts, and the unweighted Models 1-3 reproduce the published estimates and interval widths" }, { "choice": "zinc and dietary covariates", "decision": "first-day 24-hour recall (DR1T), reliable recalls only", "reason": "unstated; the first day's zinc gives the paper's quartile cutpoints (5.69, 8.37, 11.96) and group sizes (1,141, 1,156, 1,147, 1,153) exactly, and Table 1's energy, protein and fat medians" }, { "choice": "zinc quartiles", "decision": "the paper's cutpoints", "reason": "published" }, { "choice": "second-hand smoke", "decision": "anyone in the household smokes: SMD460 1 or more (2013 on), SMD410 yes (2011-2012)", "reason": "unstated; Table 1's 1,265 yes and 3,332 no are these codings' 1,232 and 3,264 plus the 101 missing values imputed; smoking inside the home (SMD470) gives far fewer" }, { "choice": "family history of asthma", "decision": "MCQ300B yes, don't know as missing", "reason": "unstated; Table 1's 1,644 and 2,953 are MCQ300B's 1,349 and 2,408 plus the 840 missing or don't know imputed" }, { "choice": "income to poverty ratio", "decision": "low 1.3 or less, medium to 3.5, high above 3.5", "reason": "the Methods' groups (Table 1 labels low as below 1.3)" }, { "choice": "eosinophils and hemoglobin", "decision": "LBXEOPCT (%) and LBXHGB (g/dL), continuous", "reason": "Table 1's rows for these two are swapped (eosinophils 2.5 (1.5, 4.1) %, hemoglobin 13.3 +- 1.3 g/dL in the data)" }, { "choice": "lower age bound", "decision": "none beyond BMDBMIC, which exists for ages 2-19", "reason": "the paper's flow (19,182 under 20, then 1,793, 3,396, 1,112 and 8,284 excluded, leaving 4,597 with 963 cases) is reproduced exactly" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": false, "detail": "The Methods define the outcome as asthma (MCQ010 yes), but Table 3's estimates are the odds of not having asthma: its crude ORs (0.85, 0.83, 0.72) are those of no asthma from Table 1's counts, and its n(%) column counts those without asthma (933, 916, 904, 881). With no asthma as the event Model 3 gives 0.710 (0.526-0.959); with asthma, as stated, 1.41 (1.04-1.90)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's eosinophil and hemoglobin rows are swapped: it gives eosinophils as 13.3 +- 1.3% and hemoglobin as 2.5 (1.5, 4.1) g/dL, which in the data are hemoglobin's mean and SD and the eosinophil percentage's median and IQR." } ], "projected_power": 0.171313590066731 }, "row275": { "id": "row275", "row": 275, "doi": "10.1097/MD.0000000000030539", "measure": "OR", "contrast": "per year of age (odds of being uninsured against insured)", "published": { "measure": "OR", "estimate": 0.95, "low": 0.93, "high": 0.96, "n": 1681, "events": 85, "contrast": "per year of age (odds of being uninsured against insured)" }, "published_b": [-0.0512932943875506, -0.0725706928348354, -0.0408219945202552], "constants": null, "original": { "estimate": 0.948412709898177, "ci": [0.933615269411033, 0.963444684086884], "b": -0.0529655234697008, "se": 0.00780760280250614, "df": 45, "p": 2.13324758833642e-08, "n": 1567, "events": 75, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 90, "strata": 45 }, "harmonized": { "estimate": 0.948015145511826, "ci": [0.935061849795797, 0.961147881625239], "b": -0.0533848005764375, "se": 0.00683073265579388, "df": 45, "p": 6.40802357556293e-10, "n": 1668, "events": 83, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 90, "strata": 45 }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.03260131957046, "published_z": -6.33304765139559, "interval_ratio_original": 1.00947706633036, "ratio_harmonized": 1.00791603819372, "variants": [ { "label": "unweighted", "estimate": 0.946730486541197, "ci": [0.931827992342182, 0.961871312637489], "b": -0.0547408234068894, "se": 0.00809515665615153, "df": null, "p": 1.35940298521389e-11, "n": 1567, "events": 75, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "age alone (published 0.94, 0.93-0.96)", "estimate": 0.944017361224194, "ci": [0.931188762446703, 0.957022694250668], "b": -0.0576107218784275, "se": 0.00679336924516957, "df": 45, "p": 6.96313060893152e-11, "n": 1681, "events": 85, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "age alone, unweighted", "estimate": 0.944666840552472, "ci": [0.931931537469424, 0.957576177819465], "b": -0.0569229633285672, "se": 0.00692510751422905, "df": null, "p": 2.03851526653535e-16, "n": 1681, "events": 85, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "income code 12 ('$20,000 and over') missing", "estimate": 0.948225081010896, "ci": [0.933353379426376, 0.963333742693159], "b": -0.0531633776818156, "se": 0.00784866167557378, "df": 45, "p": 2.20960929482942e-08, "n": 1516, "events": 75, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "years in US in three groups (harmonized coding) only", "estimate": 0.947577708497929, "ci": [0.932910721168071, 0.962475286507529], "b": -0.053846331142705, "se": 0.00774510470672428, "df": 45, "p": 1.19925294051866e-08, "n": 1567, "events": 75, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "harmonized plus income-to-poverty ratio (continuous)", "estimate": 0.953868189561252, "ci": [0.939260171236921, 0.96870340180469], "b": -0.0472297831580147, "se": 0.00766245785038934, "df": 45, "p": 1.78307112046002e-07, "n": 1506, "events": 76, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 90, "strata": 45 } ], "left_out": ["annual household income in five bands (INDHHIN2): 2021-2023 releases no household income in dollars, only the family income-to-poverty ratio (INDFMPIR), a different measure, so the harmonized model leaves income out, and with it the paper's exclusion of those with income unknown", "years in the US among the foreign-born split at 40 (20-39 against 40 or more): 2021-2023 records length of time in the US only up to '20 years or more' (DMDYRUSR), so the harmonized model codes US born, foreign-born under 20 years, and 20 years or more"], "choices": [ { "choice": "population", "decision": "adults 20 and older ever told they had cancer or a malignancy (MCQ220 1) who answered yes or no about health insurance (HIQ011 1 or 2)", "reason": "as stated; it gives the flow chart's 1,684 and 1,681 (85 uninsured) and Table 1's weighted totals (25,982,352; 1,129,421 uninsured) exactly" }, { "choice": "design", "decision": "interview weight WTINT2YR divided by 3 with SDMVSTRA and SDMVPSU; the cancer population is a domain of the full design, so cases with a missing predictor stay in the variance estimate", "reason": "the paper's weights and design, with SAS's NOMCAR option; whether it subset the data or used a domain is unstated, and here both give the same estimate and interval (45 design degrees of freedom). The published intervals are matched to the second decimal once SAS's default variance adjustment, (n-1)/(n-p) with n = 1,681, is applied, so the paper analyzed the 1,681 as a subset; the estimates are unchanged" }, { "choice": "model sample", "decision": "complete cases on the six predictors (1,567, 75 uninsured)", "reason": "SAS fits the model on complete cases; the paper reports the missing counts (education 3, income 104, US birth and length of stay 10), and these codings match each exactly" }, { "choice": "household income bands", "decision": "INDHHIN2 1-4 and 13 under $20,000; 5-7 and 12 ('$20,000 and over') $20,000-44,999; 8-10 $45,000-74,999; 14 $75,000-99,999; 15 $100,000 or more; refused and don't know missing", "reason": "unstated; only this grouping matches Table 1's counts (348, 526, 301, 123, 279, with 104 missing). Code 12 can't be placed in one band; leaving it missing gives 0.948 (variant)" }, { "choice": "US birth and length of stay", "decision": "US born (DMDBORN4 1); foreign-born by DMDYRSUS under 20 years (1-5), 20-39 (6-7), 40 or more (8-9); refused or don't know missing", "reason": "the paper's groups; Table 1's counts (1,400, 56, 98, 117, with 10 missing) match exactly" }, { "choice": "race, education, sex", "decision": "RIDRETH3 non-Hispanic White, Black, other (Asian and other or multiracial), Hispanic (Mexican American and other Hispanic); DMDEDUC2 college graduate, less than high school (1-2), high school or GED, some college; male as reference", "reason": "as stated; Table 1's counts match exactly" }, { "choice": "age", "decision": "years (RIDAGEYR, top-coded at 80), continuous, per year", "reason": "the Results read the unadjusted estimate per year ('6% lower in 1-year older participants'); with these codings every odds ratio in Tables 2 and 3 is reproduced to the second decimal" } ], "departures": [], "projected_power": 0.927016940365129 }, "row284": { "id": "row284", "row": 284, "doi": "10.1186/s12889-022-13524-y", "measure": "beta", "contrast": "workday sleep of 5 hours or less vs more than 7 up to 8 hours (serum albumin, g/L)", "published": { "measure": "beta", "estimate": -1, "low": -1.26, "high": -0.74, "n": 9973, "events": null, "contrast": "workday sleep of 5 hours or less vs more than 7 up to 8 hours (serum albumin, g/L)" }, "published_b": [-1, -1.26, -0.74], "constants": { "tp": 72, "alt": 19, "ast": 21, "creatinine": 74.26, "uacr": 7.56, "crp": 1.99, "glucose": 5.22, "bmi": 28.6 }, "original": { "estimate": -1.00105252788383, "ci": [-1.4190447328912, -0.583060322876467], "b": -1.00105252788383, "se": 0.204670147569042, "df": 30, "p": 3.16839455161209e-05, "n": 9973, "events": null, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 60, "strata": 30 }, "harmonized": { "estimate": -1.0001591850877, "ci": [-1.41908393966872, -0.581234430506672], "b": -1.0001591850877, "se": 0.205126771057639, "df": 30, "p": 3.30760865123865e-05, "n": 9973, "events": null, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 60, "strata": 30 }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.00105252788383, "published_z": -7.53832301746174, "interval_ratio_original": 0.62202116902017, "ratio_harmonized": 0.999107596483451, "variants": [ { "label": "examination weight, as the text says (WTMEC2YR)", "estimate": -1.00025708339358, "ci": [-1.40493250109734, -0.59558166568982], "b": -1.00025708339358, "se": 0.198149574242737, "df": 30, "p": 2.03442143479474e-05, "n": 9973, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "unweighted", "estimate": -0.51698105014777, "ci": [-0.759755623116408, -0.274206477179131], "b": -0.51698105014777, "se": 0.123866854127736, "df": null, "p": 2.99713997954392e-05, "n": 9973, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1 (published -1.09, -1.43 to -0.76)", "estimate": -1.09459482932884, "ci": [-1.48575292647573, -0.703436732181962], "b": -1.09459482932884, "se": 0.191530809682127, "df": 30, "p": 3.10013474054227e-06, "n": 9973, "events": null, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "Model 1, examination weight", "estimate": -1.10158357417022, "ci": [-1.49493206968006, -0.708235078660389], "b": -1.10158357417022, "se": 0.19260333962601, "df": 30, "p": 3.06150109383571e-06, "n": 9973, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "Model 2 (published -0.93, -1.24 to -0.61)", "estimate": -0.925439565359727, "ci": [-1.33661574386273, -0.514263386856729], "b": -0.925439565359727, "se": 0.201332675879926, "df": 30, "p": 7.26022236638168e-05, "n": 9973, "events": null, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "Model 2, examination weight", "estimate": -0.936736094952866, "ci": [-1.34434796993276, -0.529124219972973], "b": -0.936736094952866, "se": 0.199587412405355, "df": 30, "p": 5.53060346505318e-05, "n": 9973, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "creatinine and UACR log2-transformed, as Table 1 shows them", "estimate": -0.972865631629949, "ci": [-1.39586579774345, -0.549865465516444], "b": -0.972865631629949, "se": 0.207122298892284, "df": 30, "p": 5.47337754137401e-05, "n": 9973, "events": null, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "complete cases, no imputation", "estimate": -1.00802108540981, "ci": [-1.42661101301925, -0.589431157800372], "b": -1.00802108540981, "se": 0.204962822819218, "df": 30, "p": 2.93585911971495e-05, "n": 9675, "events": null, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "other Hispanic with other race, as the text's labels read", "estimate": -1.00360385540324, "ci": [-1.42080431573501, -0.58640339507147], "b": -1.00360385540324, "se": 0.204282469287846, "df": 30, "p": 2.97962415167409e-05, "n": 9973, "events": null, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "with a cycle term (2017-2018's analyzer reads albumin lower)", "estimate": -0.644002885154252, "ci": [-0.972161942438968, -0.315843827869536], "b": -0.644002885154252, "se": 0.160683289965652, "df": 30, "p": 0.000373640693843493, "n": 9973, "events": null, "weighted": true, "weights": "WTINT2YR", "dropped": [], "psu": 60, "strata": 30 } ], "left_out": "moderate work activity (PAQ620): 2021-2023 asks only about leisure-time activity", "choices": [ { "choice": "population", "decision": "adults 20 and older with workday sleep hours (SLD012), a usual workday sleep time (SLQ300 as 'HH:MM'), and serum albumin (LBDSALSI)", "reason": "the paper's steps; they reproduce its flow exactly: 11,288 adults, 79 without sleep hours, 35 without a sleep time (all 2017-2018 participants whose derived hours NCHS coded as under 3 or 14 or more and whose times it blanked; 2021-2023 blanks them too), 1,201 without albumin, 9,973 left, and Table 1's group sizes (694, 1,051, 2,248, 2,926, 1,890, 1,164)" }, { "choice": "weights", "decision": "interview weight, WTINT2YR over 2, with strata and PSUs (WTINT2YR in 2021-2023)", "reason": "the text says it combined the examination weight (WTMEC2YR/2), but the interview weight reproduces every printed number checked: Table 1's weighted means (sleep hours, age, albumin, laboratory values, BMI) and shares (sex, race, marital status, work activity, the three conditions) in all six groups, and Table 2's Models 1, 2, and 3 in all five sleep groups, intervals included (as weighted least squares, which the paper's intervals are; the survey design gives the same estimates with wider intervals). The examination weight misses Models 1 and 2 by 0.01 (-1.10, -0.94) and Table 1's shares by up to half a point; it is a variant, and it too gives -1.00 for the headline. The plan follows the computation the paper's numbers reveal" }, { "choice": "sleep groups", "decision": "5 hours or less, over 5 to 6, over 6 to 7, over 7 to 8 (reference), over 8 to 9, over 9; SLD012's codes for under 3 hours (2) and 14 or more (14) as values", "reason": "the paper's groups; they reproduce Table 1's sizes and mean hours exactly" }, { "choice": "race", "decision": "Mexican American with other Hispanic (RIDRETH1 1-2), non-Hispanic White, non-Hispanic Black, other race (5)", "reason": "the text names four groups without placing other Hispanic participants; Table 1's weighted shares are reproduced exactly only with them under 'Mexican American' (the text's labels, other Hispanic with other race, are a variant)" }, { "choice": "marital status and moderate work activity", "decision": "living alone if widowed, divorced, separated, or never married, otherwise married or living with a partner (no answer included); moderate work activity (PAQ620) yes, otherwise no (don't know included)", "reason": "unstated handling of missing answers; Table 1's weighted shares are reproduced exactly this way" }, { "choice": "hypertension, high cholesterol, cancer", "decision": "told yes, no, or unknown (refused, don't know), unknown as its own level", "reason": "the paper's three levels; Table 1's weighted shares are reproduced" }, { "choice": "laboratory covariates", "decision": "total protein (g/L), ALT, AST, creatinine (umol/L), UACR (mg/g), hs-CRP (mg/L), and serum glucose from the biochemistry profile (LBDSGLSI, mmol/L), continuous and untransformed", "reason": "Table 1's glucose means are serum glucose, not fasting plasma glucose; whether creatinine and UACR entered log2-transformed is unstated (Table 1 shows their logs), and untransformed values reproduce Model 3 in all five groups where log2 values give -0.97 (variant)" }, { "choice": "missing covariates", "decision": "each continuous covariate's missing values (BMI 127, UACR 158, hs-CRP 34, AST 16, a few others) filled with its unweighted median over the 9,973 participants on the paper's cycles, reused in 2021-2023; all 9,973 analyzed", "reason": "unstated; Table 1's weighted means and SDs of BMI, UACR, hs-CRP, and AST are reproduced exactly only with median filling, and so is Model 3. Complete cases (9,675) are a variant" }, { "choice": "albumin across analyzers", "decision": "values as released in both cycles, no crosswalk", "reason": "a measurement change, as the paper pooled them: NCHS's bridging study found 2017-2018's Roche Cobas 6000 reads albumin about 4.4% lower than 2015-2016's Beckman analyzer, while 2021-2023's Cobas 8000 differs from the Cobas 6000 by 1.5% with no adjustment recommended. Sleep hours also changed form between the two cycles (asked directly, then derived from times), so the pooled contrast mixes cycles; a variant adds a cycle term" } ], "departures": [ { "kind": "weighting", "affects_headline": true, "followed": true, "detail": "The text says it used the combined examination weight (WTMEC2YR/2), but Table 1's weighted means and shares and all three Table 2 models in all five sleep groups, intervals included, are reproduced to the printed digit only with the interview weight (WTINT2YR). With the examination weight Models 1 and 2 give -1.10 and -0.94 against the printed -1.09 and -0.93; the headline is -1.00 with either." }, { "kind": "weighting", "affects_headline": true, "followed": false, "detail": "The text says its weighted data were calculated according to the NHANES analytical guidelines, but its intervals are those of weighted least squares without the survey design: so fitted, Model 3 gives the published -1.00 (-1.26 to -0.74), while with the strata and PSUs the interval is -1.42 to -0.58." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text's race groups are Mexican American, non-Hispanic white, non-Hispanic black, and other race, but Table 1's weighted shares are reproduced exactly only with other Hispanic participants in the Mexican American group; with them in other race, as the labels read, Model 3 is -1.004 against -1.001." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The text and Table 1 give continuous variables as means +- SEs, but the values after +- are weighted SDs (age 49.01 +- 16.10 years in the 5 hours or less group)." } ], "projected_power": 0.998674690103737 }, "row295": { "id": "row295", "row": 295, "doi": "10.1186/s12889-024-19722-0", "measure": "OR", "contrast": "METS-VF quartile 4 vs quartile 1 (quartiles of the analytic sample)", "published": { "measure": "OR", "estimate": 2.33, "low": 1.65, "high": 3.28, "n": 7639, "events": 937, "contrast": "METS-VF quartile 4 vs quartile 1 (quartiles of the analytic sample)" }, "published_b": [0.845868267577609, 0.500775287912489, 1.18784342239605], "constants": { "mets_vf_quartiles": [6.45537149596948, 6.98186679540331, 7.36187719541174], "mets_vf_computed_quartiles": [8.74067756546741, 9.71381495965549, 10.646481808304] }, "original": { "estimate": 2.24286525750808, "ci": [1.56984304787124, 3.2044251622213], "b": 0.807754181289635, "se": 0.179209413934755, "df": 78, "p": 2.27728407389209e-05, "n": 7379, "events": 944, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 152, "strata": 74 }, "harmonized": { "estimate": 2.32532473157542, "ci": [1.64273440054621, 3.29154555080748], "b": 0.843859698781246, "se": 0.174547574956074, "df": 78, "p": 6.56558865409593e-06, "n": 7737, "events": 963, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 152, "strata": 74 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.954940872297853, "published_z": 4.82592993885111, "interval_ratio_original": 0.962877617790546, "ratio_harmonized": 1.04469864511746, "variants": [ { "label": "METS-VF as printed (mg/dL, with the outer log), Model 3", "estimate": 3.18071865000472, "ci": [2.07849012635041, 4.86746172244371], "b": 1.15710716182758, "se": 0.213710756173217, "df": 78, "p": 6.58781088538373e-07, "n": 7379, "events": 944, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 152, "strata": 74 }, { "label": "METS-VF as printed, crude", "estimate": 6.23166632777368, "ci": [4.50610147840702, 8.6180183484098], "b": 1.82964376537857, "se": 0.162850983368673, "df": 78, "p": 5.66190519204463e-18, "n": 7379, "events": 944, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 152, "strata": 74 }, { "label": "METS-VF as printed, age and sex", "estimate": 2.68518143009331, "ci": [1.81074169790281, 3.98190383579766], "b": 0.987748297407315, "se": 0.197911614805208, "df": 78, "p": 3.5727094915536e-06, "n": 7379, "events": 944, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 152, "strata": 74 }, { "label": "crude (published 3.90, 2.94-5.15)", "estimate": 3.46911428750157, "ci": [2.65630125075409, 4.53064347890925], "b": 1.24389931281868, "se": 0.134096018270271, "df": 78, "p": 3.14549990447775e-14, "n": 7379, "events": 944, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 152, "strata": 74 }, { "label": "Model 2, age and sex (published 2.27, 1.67-3.08)", "estimate": 2.10127995150781, "ci": [1.54608774182569, 2.855838847409], "b": 0.742546659777261, "se": 0.154114779551584, "df": 78, "p": 6.99518705938175e-06, "n": 7379, "events": 944, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 152, "strata": 74 }, { "label": "METS-VF as the paper computed it, per unit, Model 3 (supplement 1.23, 1.14-1.33)", "estimate": 1.22367112451254, "ci": [1.12682528945394, 1.32884044667778], "b": 0.201855458863005, "se": 0.0414151643195531, "df": 78, "p": 5.63673575003994e-06, "n": 7379, "events": 944, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 152, "strata": 74 }, { "label": "unweighted", "estimate": 1.98176086786475, "ci": [1.5231148592175, 2.57851606767052], "b": 0.683985776693728, "se": 0.134302615392031, "df": null, "p": 3.52683315248165e-07, "n": 7623, "events": 972, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "examination weight", "estimate": 2.37711893167254, "ci": [1.65335412882529, 3.41771573119113], "b": 0.865889221485884, "se": 0.182376234350564, "df": 78, "p": 9.16699757801569e-06, "n": 7623, "events": 972, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 152, "strata": 74 }, { "label": "2009-2010 arthritis type read with the later codes", "estimate": 2.07360706069298, "ci": [1.43231001485317, 3.00203601005789], "b": 0.729289632211003, "se": 0.185851089257648, "df": 78, "p": 0.00018631591157967, "n": 7379, "events": 879, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 152, "strata": 74 }, { "label": "energy from one recall where there is no second", "estimate": 2.24765983084017, "ci": [1.59633149651259, 3.16474036013775], "b": 0.809889599780663, "se": 0.171877299416614, "df": 78, "p": 1.05119714861319e-05, "n": 8278, "events": 1009, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 152, "strata": 74 } ], "left_out": ["alcohol intake as at least 12 drinks a year (ALQ101, asked through 2015-2016; 2021-2023 asks only whether one ever drank, how often, and how much)", "work activity (PAQ605, PAQ620; 2021-2023 asks only about leisure-time activity)"], "choices": [ { "choice": "headline", "decision": "quartile 4 vs quartile 1 in Model 3 (Table 2)", "reason": "the abstract gives no estimate; Table 2, the main text's only regression table, has quartiles only; the per-unit estimate (1.23, 1.14-1.33) is in Supplementary Table 2" }, { "choice": "METS-VF formula", "decision": "as the paper's numbers show it computed: METS-IR without its outer log and with glucose in mmol/L, then the printed METS-VF formula", "reason": "the formula as printed (mg/dL) gives Table 1 means of 6.73 and 7.15 against the published 9.54 and 10.21, and a crude OR of 6.23 against 3.90; the computed index gives 9.54 and 10.14, Supplementary Table 1's age by quartile, and the crude, Model 3, and per-unit estimates (3.47, 2.24, 1.224 against 3.90, 2.33, 1.23). The published estimate comes from that index, so the replication uses it; the printed formula is fitted as a variant" }, { "choice": "quartile cutpoints", "decision": "unweighted quartiles of the index in the Model 3 analytic sample of the paper's cycles, kept for 2021-2023", "reason": "the paper prints none" }, { "choice": "osteoarthritis", "decision": "told of arthritis (MCQ160A) and type osteoarthritis (MCQ191 = 2 in 2009-2010, MCQ195 = 1 from 2011); other types count as no; a type answered don't know or refused is missing", "reason": "the paper's question wording; with the computed METS-VF, osteoarthritis by quartile is 7, 13, 17, 20% against Supplementary Table 1's 6, 12, 16, 20%, and reading 2009-2010 with the later codes fits the published estimates less well (variant)" }, { "choice": "dietary energy", "decision": "mean of two reliable recalls, required; under 500 or over 5,000 kcal (women) or 8,000 (men) excluded", "reason": "the paper says all participants had two recalls and their average was used; requiring both gives 7,623 participants (7,379 with a positive fasting weight) against the published 7,639; allowing one recall gives 8,581 (variant)" }, { "choice": "weights", "decision": "fasting subsample weight (WTSAF2YR), each cycle's weight over five", "reason": "unstated; NCHS's guidelines use the smallest subsample's weight, and METS-VF needs fasting glucose and triglycerides. The paper divides the 2-year weights by 2, a constant that changes no estimate or interval" }, { "choice": "alcohol intake", "decision": "at least 12 drinks in any one year (ALQ101) through 2015-2016; in 2017-2018, past-year drinking days (ALQ121) times drinks a drinking day (ALQ130) of 12 or more", "reason": "the paper's definition; 2017-2018 no longer asks ALQ101. With this coding the share who drink (69.9% without osteoarthritis, 65.2% with) matches Table 1's (69.8%, 66.1%); ever having had a drink (ALQ111) would give about 77%, and the same estimate" }, { "choice": "work and leisure activity", "decision": "vigorous if any vigorous activity (PAQ605, PAQ650), moderate if any moderate (PAQ620, PAQ665), other otherwise, with missing answers as other", "reason": "Table 1's groups; the flow chart excludes no one for missing activity, and the groups' shares match Table 1. In 2021-2023 leisure activity is vigorous or moderate if done more than 0 times (PAD810Q, PAD790Q)" }, { "choice": "hypertension, diabetes, coronary heart disease", "decision": "self-reported diagnosis (BPQ020, DIQ010, MCQ160C), yes or no, with borderline diabetes missing", "reason": "unstated; the flow chart's counts of missing values (13, 224, 17) fit these items, and Table 1's diabetes share (11.8% and 21.8%) matches ours (11.7% and 22.1%)" }, { "choice": "other covariates", "decision": "education in three levels (DMDEDUC2 1-2, 3, 4-5); income to poverty ratio under 1.3, 1.3-3.5, over 3.5; smoking never, former, current; age, serum calcium, vitamin D and energy continuous", "reason": "Table 1's categories; it reports the continuous ones as means" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods print METS-IR as Ln[(2 x fasting glucose + fasting triglycerides) x BMI] / Ln(HDL-C), but Table 1's mean METS-VF (9.54 without osteoarthritis, 10.21 with) and Table 2's odds ratios are reproduced by METS-IR with no logarithm in its numerator and glucose in mmol/L (9.54 and 10.14; Model 3, quartile 4 vs 1, 2.24 against 2.33), not by the printed formula in mg/dL (6.73 and 7.15; 3.18)." }, { "kind": "weighting", "affects_headline": false, "followed": null, "detail": "Table 1's note calls its percentages survey-weighted, but they are the unweighted ones: non-Hispanic White 42.42% without osteoarthritis (41.4% unweighted here, 66.8% weighted) and diabetes 11.77% and 21.83% (11.7% and 22.1% unweighted, 8.6% and 16.3% weighted)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's Former and Never smoking rows are swapped: its 56.02% former and 23.89% never without osteoarthritis are 58.6% never and 22.8% former here, and its 46.64% former and 33.34% never with osteoarthritis are 48.5% never and 33.3% former." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The Results give Model 3's quartile 2 odds ratio as 1.14 (1.02-2.02), where Table 2 has 1.43 (1.02-2.01)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The text excludes 233 participants under 20 after those without osteoarthritis data, but the arthritis questions (MCQ160A) are asked only from age 20, so no one under 20 remains at that step." } ], "projected_power": 0.52367529105337 }, "row296": { "id": "row296", "row": 296, "doi": "10.1155/2024/4306797", "measure": "beta", "contrast": "per 1 cm of waist circumference; outcome the square root of total testosterone (ng/dL)", "published": { "measure": "beta", "estimate": -0.117, "low": -0.136, "high": -0.098, "n": 3359, "events": null, "contrast": "per 1 cm of waist circumference; outcome the square root of total testosterone (ng/dL)" }, "published_b": [-0.117, -0.136, -0.098], "constants": null, "original": { "estimate": -0.115510462802662, "ci": [-0.12962345667716, -0.101397468928164], "b": -0.115510462802662, "se": 0.00691043637735704, "df": 30, "p": 9.53683015817505e-17, "n": 3064, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, "harmonized": { "estimate": -0.113360426905851, "ci": [-0.126302205396393, -0.100418648415309], "b": -0.113360426905851, "se": 0.00633695002378903, "df": 30, "p": 1.49136443729032e-17, "n": 3269, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.987268912843264, "published_z": -12.0692519047993, "interval_ratio_original": 1.34627706700364, "ratio_harmonized": 0.981386656717981, "variants": [ { "label": "nonadjusted (published -0.117, -0.127 to -0.107)", "estimate": -0.115830104516769, "ci": [-0.126932293331022, -0.104727915702515], "b": -0.115830104516769, "se": 0.00543619377522268, "df": 30, "p": 1.14009559768719e-19, "n": 3064, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "age and race (published -0.120, -0.131 to -0.108)", "estimate": -0.118611036805864, "ci": [-0.130374054625341, -0.106848018986386], "b": -0.118611036805864, "se": 0.0057597691156162, "df": 30, "p": 2.97570524017679e-19, "n": 3064, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "nonadjusted, everyone in the study population", "estimate": -0.117757371078233, "ci": [-0.128131206735495, -0.107383535420971], "b": -0.117757371078233, "se": 0.00507955519120598, "df": 30, "p": 1.04446610784581e-20, "n": 3357, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "unweighted", "estimate": -0.113327991579797, "ci": [-0.122364792645787, -0.104291190513808], "b": -0.113327991579797, "se": 0.00461069751141895, "df": null, "p": 2.10016170869836e-133, "n": 3064, "events": null, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "activity: moderate work only", "estimate": -0.115526612774957, "ci": [-0.129642631047468, -0.101410594502445], "b": -0.115526612774957, "se": 0.00691191727575731, "df": 30, "p": 9.55615454971615e-17, "n": 3064, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "activity: leisure time instead of work", "estimate": -0.113612715354756, "ci": [-0.127604485427623, -0.0996209452818886], "b": -0.113612715354756, "se": 0.00685107906621225, "df": 30, "p": 1.18318597551987e-16, "n": 3063, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "harmonized covariates, paper's population", "estimate": -0.115854707886332, "ci": [-0.129841009877374, -0.101868405895289], "b": -0.115854707886332, "se": 0.00684840161648809, "df": 30, "p": 6.88125957225539e-17, "n": 3060, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 }, { "label": "glucocorticoids: systemic only (Multum 301)", "estimate": -0.114970601989995, "ci": [-0.128866828438951, -0.10107437554104], "b": -0.114970601989995, "se": 0.00680429607033093, "df": 30, "p": 7.10977275359321e-17, "n": 3076, "events": null, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 60, "strata": 30 } ], "left_out": ["exclusion of glucocorticoid users (2021-2023 prescription data name no drugs)", "exclusion of recent infection (flu, pneumonia, or ear infection in the past 30 days, HSQ520, not asked in 2021-2023)", "exclusion of HIV infection (no HIV test in 2021-2023)", "hepatitis C told by a doctor (HEQ030) in the liver disease exclusion (dropped from 2021-2023; any liver condition and hepatitis B told remain)", "physical activity (the paper's is time in work activity; 2021-2023 asks only leisure-time activity)", "session of the blood draw in three levels: 2021-2023 codes morning versus afternoon or evening, so the harmonized model uses those two levels", "nondrinkers as those who never had 12 drinks in any one year (ALQ101, not asked from 2017): the harmonized model counts those with no drink in the past year as nondrinkers, with the same drinking-day groups"], "choices": [ { "choice": "outcome scale", "decision": "square root of total testosterone in ng/dL", "reason": "the paper does not say which unit was square-rooted; Table 1 reports ng/dL and Table 2's quartile contrasts fit the ng/dL scale (the extraction's check)" }, { "choice": "weights", "decision": "examination weight (WTMEC2YR) halved for the two cycles; WTPH2YR from TST_L in 2021-2023", "reason": "the paper says only that it weighted; testosterone was measured on the examined sample" }, { "choice": "starting sample", "decision": "men 20 and older with total testosterone, estradiol, SHBG, waist, BMI, the 12-drinks-a-year item (ALQ101), and smoking status", "reason": "reproduces Figure 1's 5290 men and its missing counts (751, 210, 14, 244) exactly; the 3 men left out of 5290 are those who refused or did not know whether they smoked 100 cigarettes" }, { "choice": "cancer affecting hormones", "decision": "any reported cancer of the prostate, testis, thyroid, liver, or breast (MCQ230A-D)", "reason": "prostate, testis, thyroid, and liver reproduce Figure 1's 133 and the following 171 with thyroid problems exactly; no man in the paper's sample reported breast cancer, which is hormone dependent" }, { "choice": "thyroid disorder", "decision": "ever told of a thyroid problem (MCQ160M)", "reason": "reproduces Figure 1's 171" }, { "choice": "liver disease", "decision": "ever told of any liver condition (MCQ160L) or of hepatitis B or C (HEQ010, HEQ030)", "reason": "reproduces Figure 1's 211, 46 of them hepatitis B or C alone, and the 3556 remaining" }, { "choice": "glucocorticoid use", "decision": "any prescription medicine in the Multum glucocorticoid (301) or inhaled corticosteroid (296) category in the past month", "reason": "unstated; the paper counts 43, systemic plus inhaled gives 46 and systemic alone 31; the final sample is 3357 against the paper's 3359" }, { "choice": "recent infection", "decision": "flu, pneumonia, or ear infection in the past 30 days (HSQ520)", "reason": "unstated; gives 136 against the paper's 137" }, { "choice": "HIV/AIDS", "decision": "reactive HIV antibody test (LBDHI 2013-2014, LBXHIVC 2015-2016)", "reason": "unstated; gives the paper's 17" }, { "choice": "alcohol intake", "decision": "nondrinker if never 12 drinks in any one year (ALQ101); otherwise drinking days a month in the past year under 5, 5 to under 10, or 10 and more, past-year abstainers in the lowest group", "reason": "the paper's 'drinks/month' groups are reproduced only this way: 590/1816/330/621 against Table 1's 592/1816/331/620" }, { "choice": "physical activity time", "decision": "hours a week of vigorous plus moderate work activity (GPAQ PAQ605-PAD630) in Table 1's five groups", "reason": "unstated; of the GPAQ domains and their sums, work activity's univariate associations with testosterone match Supplementary Table 2 (-0.86, -0.33, -0.85, 0.31 against -0.88, -0.36, -0.96, 0.24) and its small groups match Table 1 (103, 308, 140 against 102, 297, 125), though it has 1718 inactive against 1907" }, { "choice": "diabetes", "decision": "told by a doctor, insulin or diabetes pills, HbA1c 6.5% or more, or fasting glucose 126 mg/dL or more; borderline counts as no", "reason": "the paper's definition; 574 against Table 1's 579" }, { "choice": "session, race, education, smoking", "decision": "PHDSESN in three levels; RIDRETH3 in six; DMDEDUC2 in five; never, former, current from SMQ020 and SMQ040", "reason": "the paper's categories; their counts match Table 1" }, { "choice": "missing PIR, cotinine, or work activity", "decision": "complete case", "reason": "unstated; Figure 1 excludes no one for them; the fully adjusted model keeps 3064 of the 3357 men, nearly all of the 293 dropped for missing PIR" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Covariates section and Table 1 group alcohol intake as nondrinker and 1-5, 5-10, and 10+ drinks a month (a drink being 12 oz of beer, a 5 oz glass of wine, or 1.5 oz of liquor), but Table 1's counts (592, 1,816, 331, 620) are reproduced only by drinking days a month over the past year, with nondrinkers those who never had 12 drinks in any one year and past-year abstainers in the 1-5 group (590, 1,816, 330, 621)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 2 gives the fully adjusted T/E2 estimate for waist quartile 2 as -0.932 (-0.150, -0.354), an interval that excludes its estimate." } ], "projected_power": 0.999999998714679 }, "row303": { "id": "row303", "row": 303, "doi": "10.3389/fendo.2023.1245199", "measure": "OR", "contrast": "per 100 units of SII (platelets x neutrophils / lymphocytes, counts in 10^3 cells/uL)", "published": { "measure": "OR", "estimate": 1.04, "low": 1.02, "high": 1.06, "n": 7877, "events": 1266, "contrast": "per 100 units of SII (platelets x neutrophils / lymphocytes, counts in 10^3 cells/uL)" }, "published_b": [0.0392207131532813, 0.0198026272961797, 0.0582689081239758], "constants": { "sii100_quartiles": [3.17142857142857, 4.515, 6.35928571428571] }, "original": { "estimate": 1.02690523781681, "ci": [1.00201451808568, 1.05241425989542], "b": 0.0265496558194979, "se": 0.012519191346305, "df": null, "p": 0.0339456660293864, "n": 5903, "events": 930, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 1.02702195461277, "ci": [1.00214566713097, 1.05251574681386], "b": 0.0266633081402914, "se": 0.0125104031803319, "df": null, "p": 0.0330651863535229, "n": 5903, "events": 930, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.676929450918887, "published_z": 3.99680881926385, "interval_ratio_original": 0.783837133457387, "ratio_harmonized": 1.0042807455421, "variants": [ { "label": "weighted (examination weight), as the paper says", "estimate": 1.01080632369503, "ci": [0.954777163091524, 1.07012344190717], "b": 0.0107483526411724, "se": 0.0276885593477307, "df": 25, "p": 0.701163233796664, "n": 5903, "events": 930, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, { "label": "crude, unweighted (published 1.04, 1.02-1.05)", "estimate": 1.03736245232189, "ci": [1.02199293188138, 1.05296311150241], "b": 0.0366813882584475, "se": 0.00761586059026007, "df": null, "p": 1.46137018089242e-06, "n": 7877, "events": 1266, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, weighted", "estimate": 1.04377998760251, "ci": [1.0120415362334, 1.07651378279817], "b": 0.0428487274073427, "se": 0.014993219541418, "df": 25, "p": 0.0084726620996254, "n": 7877, "events": 1266, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, { "label": "Model 2, unweighted (published 1.04, 1.02-1.05)", "estimate": 1.03634253471327, "ci": [1.01994560800113, 1.05300306293855], "b": 0.0356977211532, "se": 0.00813709889250931, "df": null, "p": 1.14907375136118e-05, "n": 7877, "events": 1266, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2, weighted", "estimate": 1.03118368128159, "ci": [0.997115243793549, 1.06641613510586], "b": 0.0307073475385641, "se": 0.016312524292774, "df": 25, "p": 0.0714655459967459, "n": 7877, "events": 1266, "weighted": true, "weights": "WTMECPRP", "dropped": [], "psu": 49, "strata": 24 }, { "label": "Model 3 without drink", "estimate": 1.02632471028143, "ci": [1.00171313113051, 1.0515409833407], "b": 0.0259841784345736, "se": 0.0123841627861033, "df": null, "p": 0.0358894205089629, "n": 6037, "events": 952, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 3 as Table 1 and the Covariates list: serum albumin, no drink", "estimate": 1.01900132060829, "ci": [0.993875111981973, 1.0447627462275], "b": 0.0188230502243358, "se": 0.0127383832099347, "df": null, "p": 0.139497736060675, "n": 6037, "events": 952, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 3 as the Methods list: albumins, no drink, WBC, platelets", "estimate": 1.01990666640983, "ci": [0.997668051878178, 1.04264099289232], "b": 0.0197111195897017, "se": 0.0112480592045959, "df": null, "p": 0.0797047361529162, "n": 5977, "events": 936, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "quartile 4 vs 1, Model 3 (published 1.31, 1.05-1.63)", "estimate": 1.19688684059357, "ci": [0.927194037369194, 1.54502515271854], "b": 0.179723886262069, "se": 0.130265814089312, "df": null, "p": 0.16768815433177, "n": 5903, "events": 930, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "borderline counted as no diabetes", "estimate": 1.0274059169865, "ci": [1.0026584608338, 1.05276418590342], "b": 0.0270370982192834, "se": 0.0124401086114097, "df": null, "p": 0.0297516383747135, "n": 6079, "events": 930, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": "vigorous work activity (PAQ605): 2021-2023 asks no work-activity questions", "choices": [ { "choice": "cycles", "decision": "2017-March 2020 pre-pandemic files only", "reason": "the flow chart starts from P_DEMO's 15,560 participants" }, { "choice": "diabetes", "decision": "told by a doctor (DIQ010 = 1) against not (2); borderline and don't know excluded", "reason": "the flow chart's 244 with missing diabetes data are exactly the 239 borderline and 5 don't know answers; this gives the published 7,877 and 1,266 cases, and the published mean SII (524.91, SD 358.90). Counting borderline as no changes the estimate by under 0.001 (variant)" }, { "choice": "weights", "decision": "unweighted, with the examination weight (WTMECPRP) as the paper's weight", "reason": "Table 1's weighted means match WTMECPRP exactly (age 61.09 and 46.55, SII 597.58 and 532.32), but only unweighted fits reproduce the crude and Model 2 estimates with their CIs (1.037, 1.022-1.053 and 1.036, 1.020-1.053 against 1.04, 1.02-1.05 each; weighted 1.044, 1.012-1.077 and 1.031, 0.997-1.066) and, within 0.01, the quartile rows of Models 1 and 2, as EmpowerStats fits them" }, { "choice": "Model 3 covariates", "decision": "the Table 3 footnote's list: age, sex, race, marital status, income, education, drink, smoke, BMI, high blood pressure, BUN, WBC, platelets, chloride, day-1 protein, carbohydrate, total sugars, total fat, cholesterol, vigorous work activity", "reason": "it describes the table that holds the headline, and its names match Table 1's rows. The Covariates paragraph and Table 1 have serum albumin in place of drink, and the Methods' Model 3 sentence has serum and urine albumin and no WBC, platelets, or drink (variants: 1.019 and 1.020). The footnote's list is also the closest to the published estimate" }, { "choice": "drink", "decision": "ever had a drink of alcohol (ALQ111), yes or no", "reason": "unstated; it is the yes/no question parallel to smoke (SMQ020, 100 cigarettes in life); leaving drink out changes the estimate by under 0.001" }, { "choice": "smoke, high blood pressure, vigorous work activity", "decision": "SMQ020, BPQ020, PAQ605, each yes against any other answer (refused and don't know as no)", "reason": "as the paper names them; coded so, Table 1's weighted percentages are reproduced exactly (51.14, 68.85, 20.53 with diabetes; 41.60, 27.14, 28.05 without)" }, { "choice": "categorical covariates", "decision": "race in RIDRETH1's five groups, marital status in DMDMARTZ's three, education in three (DMDEDUC2 1-2, 3, 4-5), income to poverty 1.5 or less, over 1.5 to 3.5, over 3.5, BMI under 25, 25 to under 30, 30 or more", "reason": "the groups Table 1 shows; its weighted percentages for BMI are reproduced exactly and for education and marital status within 0.05 points" }, { "choice": "continuous covariates", "decision": "age, BUN (mmol/L), chloride, WBC, platelets, and day-1 dietary intakes (reliable recalls) as linear terms; BUN and chloride as measured", "reason": "Table 1 reports them as means, which match ours for BUN, chloride, and the intakes" }, { "choice": "missing covariates", "decision": "complete cases (5,903 of 7,877)", "reason": "unstated, and the flow chart excludes no one for covariates; filling them in (missing groups for categorical covariates, medians for continuous ones, all 7,877) gives 1.025. No covariate set or missing-data handling tried reaches the published 1.04 (p = 0.0006); Models 1 and 2 are reproduced exactly" } ], "departures": [ { "kind": "weighting", "affects_headline": true, "followed": true, "detail": "The abstract reports weighted multivariate regression, and Table 1's means are those with the examination weight, but Table 3's Model 1 and Model 2 estimates and intervals are reproduced only unweighted (1.037, 1.022-1.053 and 1.036, 1.020-1.053 against 1.04, 1.02-1.05 each; weighted 1.044, 1.012-1.077 and 1.031, 0.997-1.066), and Model 3's interval has the unweighted width (ours 1.002-1.052 unweighted, 0.955-1.070 weighted)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods compute SII by dividing the lymphocyte count by the platelet count and multiplying by the neutrophil count, but the published mean SII (524.91, SD 358.90), reproduced here, is that of platelets x neutrophils / lymphocytes, the formula the Introduction gives." }, { "kind": "reporting", "affects_headline": true, "followed": true, "detail": "The abstract reads the Model 3 odds ratio (1.04, 1.02-1.06) as the change for each additional unit of SII, but Table 3 and the Results give it per 100 units (SII/100), the scale on which Models 1 and 2 are reproduced." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's income-to-poverty rows 1.5-3.5 and >3.5 repeat its BMI rows 25-30 and >30 in both columns (27.08 and 62.77 with diabetes, 32.94 and 38.99 without), so the income percentages with diabetes sum to 114.5." } ], "projected_power": 0.839462847104049 }, "row306": { "id": "row306", "row": 306, "doi": "10.3390/nu15051177", "measure": "OR", "contrast": "per 100 units of SII (platelets x neutrophils / lymphocytes, counts in 10^3 cells/uL)", "published": { "measure": "OR", "estimate": 1.03, "low": 1.01, "high": 1.05, "n": 6117, "events": 4265, "contrast": "per 100 units of SII (platelets x neutrophils / lymphocytes, counts in 10^3 cells/uL)" }, "published_b": [0.0295588022415444, 0.00995033085316809, 0.048790164169432], "constants": { "sii_quartiles": [310.916666666667, 439.888888888889, 621.5625] }, "original": { "estimate": 1.02979563501756, "ci": [1.00928862188394, 1.05071931547363], "b": 0.029360369950848, "se": 0.0102627508387806, "df": null, "p": 0.00422483631881336, "n": 6117, "events": 4182, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 1.02979563501756, "ci": [1.00928862188394, 1.05071931547363], "b": 0.029360369950848, "se": 0.0102627508387806, "df": null, "p": 0.00422483631881336, "n": 6117, "events": 4182, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.993286862942723, "published_z": 2.98323565643673, "interval_ratio_original": 0.965462668536491, "ratio_harmonized": 1, "variants": [ { "label": "weighted, examination weight (Table 1's weight)", "estimate": 1.06092683883881, "ci": [1.0204422003801, 1.10301764955355], "b": 0.0591429023427626, "se": 0.0192505630396896, "df": 40, "p": 0.00381280549068943, "n": 6117, "events": 4182, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 79, "strata": 39 }, { "label": "weighted, fasting subsample weight", "estimate": 1.06127070712329, "ci": [1.02062089843444, 1.10353953708534], "b": 0.0594669704655858, "se": 0.0193242689504146, "df": 40, "p": 0.00376107856560586, "n": 5920, "events": 4059, "weighted": true, "weights": ["WTSAF2YR", "WTSAFPRP"], "dropped": [], "psu": 79, "strata": 39 }, { "label": "unweighted, positive fasting subsample weight only", "estimate": 1.02900821642649, "ci": [1.00785599364622, 1.0506043682317], "b": 0.0285954416854243, "se": 0.0105972078325153, "df": null, "p": 0.00696749299461989, "n": 5920, "events": 4059, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, unweighted (published 1.04, 1.02-1.06)", "estimate": 1.05080935083166, "ci": [1.03065513110826, 1.0713576816019], "b": 0.0495606775635974, "se": 0.00988080803982257, "df": null, "p": 5.27987729082976e-07, "n": 6117, "events": 4182, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "crude, weighted", "estimate": 1.08688029300046, "ci": [1.04694786563747, 1.12833581316246], "b": 0.0833114760402199, "se": 0.0185210010540176, "df": 40, "p": 5.76643709383316e-05, "n": 6117, "events": 4182, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 79, "strata": 39 }, { "label": "Model 3, unweighted (published 1.02, 1.00-1.04)", "estimate": 1.01526958753598, "ci": [0.995268906734879, 1.03567219713221], "b": 0.0151541807111151, "se": 0.0101514625675359, "df": null, "p": 0.135487551162587, "n": 6117, "events": 4182, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 3, weighted", "estimate": 1.03848274341881, "ci": [1.00351553197356, 1.07466837733717], "b": 0.0377607473628582, "se": 0.0169471067687073, "df": 40, "p": 0.0315581137570503, "n": 6117, "events": 4182, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 79, "strata": 39 }, { "label": "outcome also counts told to take cholesterol medicine (BPQ090D)", "estimate": 1.02969157279959, "ci": [1.0087110415661, 1.05110848538779], "b": 0.0292593135153747, "se": 0.0105032512464548, "df": null, "p": 0.00534059023987565, "n": 6117, "events": 4272, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "quartile 4 vs 1, Model 2, unweighted (published 1.27, 1.08-1.50)", "estimate": 1.43846568167606, "ci": [1.22406293728654, 1.69042248917908], "b": 0.363577046704418, "se": 0.0823491889940714, "df": null, "p": 1.00979504381924e-05, "n": 6117, "events": 4182, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": [], "choices": [ { "choice": "cycles", "decision": "2015-2016 and 2017-March 2020", "reason": "the paper says 2015-2020; its 25,531 participants are the DEMO_I and P_DEMO totals" }, { "choice": "hyperlipidemia data required", "decision": "total cholesterol, triglycerides, LDL (Friedewald, LBDLDL) and HDL all measured", "reason": "this gives exactly the flow chart's 7,298 participants with SII and hyperlipidemia data, and 6,117 aged 20 or older" }, { "choice": "weights", "decision": "unweighted, with the examination weight (WTMEC2YR, WTMECPRP) as the paper's weight", "reason": "Table 1's weighted means and percentages match the examination weight (not the fasting weight); the crude and Model 2 estimates are reproduced only unweighted; the text says only that a weighting strategy was used. The fasting weight would also drop 197 participants with lipid values but a zero fasting weight, all in 2017-March 2020" }, { "choice": "lipid thresholds", "decision": "total cholesterol 200 mg/dL or more, triglycerides 150 or more, LDL 130 or more, HDL under 40 (men) or 50 (women)", "reason": "the paper prints no inequality signs; these are the NCEP ATP III thresholds it cites" }, { "choice": "cholesterol-lowering drugs", "decision": "now taking prescribed cholesterol medicine (BPQ100D; BPQ101D in 2021-2023), missing counted as no", "reason": "the paper says persons who reported using cholesterol-lowering drugs. This gives 4,182 events against the published 4,265; adding those told to take medicine (BPQ090D) gives 4,272 and the same estimate (variant)" }, { "choice": "age and race in Model 2", "decision": "age continuous; race in RIDRETH1's five groups", "reason": "Table 1 reports age as a mean and race in those five groups" }, { "choice": "Model 3 covariates (variant only)", "decision": "coded as Table 1's percentages show (see build)", "reason": "Model 3 is not the headline" } ], "departures": [ { "kind": "weighting", "affects_headline": true, "followed": true, "detail": "The Statistical Analysis says a weighting strategy was used, and Table 1's weighted means and percentages are those with the examination weight, but Table 3's estimates and interval widths are reproduced only unweighted (Model 2: 1.030, 1.009-1.051 unweighted and 1.061, 1.020-1.103 weighted, against 1.03, 1.01-1.05)." }, { "kind": "coding", "affects_headline": true, "followed": false, "detail": "The Methods compute SII as platelets x neutrophils / lymphocytes, but the Results' mean SII (459.54, SD 317.28) is below what that formula gives for the same 6,117 participants (507.7, SD 318.4), whose mean age (50.70) and share of men (48.08%) match the text exactly; Table 3's quartile estimates differ too (Model 2, quartile 4 vs 1: 1.27 published, 1.44 here), though the per-100 estimate agrees." }, { "kind": "reporting", "affects_headline": true, "followed": true, "detail": "The abstract attributes its estimate, 1.03 (1.01, 1.05), to a multivariate linear regression analysis, but it is Table 3's Model 2 odds ratio from logistic regression, as the Methods describe and as reproduced (1.030)." }, { "kind": "coding", "affects_headline": false, "followed": null, "detail": "The Covariates section and Table 1 give education as less than high school, high school, and more than high school, but Table 1's shares (5.37%, 8.42%, 86.20% with hyperlipidemia) are those of less than 9th grade, 9th to 11th grade, and high school graduate or more (DMDEDUC2 1, 2, and 3-5)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 4 gives the estimate for women with SII under 958.14 as 1.0006 (1.0002, 1.1010), an upper bound that the estimate and its lower bound rule out (on the log scale they put it near 1.0010)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 2's light alcohol consumption row repeats the never-smoker row's percentages (55.62, 58.51, 55.16, 52.74), so its drinking rows sum to less than 100% in every quartile." } ], "projected_power": 0.409844444030116 }, "row311": { "id": "row311", "row": 311, "doi": "10.1097/MD.0000000000039258", "measure": "OR", "contrast": "per unit of the TyG index, ln(fasting triglycerides x fasting glucose / 2), both in mg/dL", "published": { "measure": "OR", "estimate": 1.54, "low": 1.21, "high": 1.95, "n": 3225, "events": 364, "contrast": "per unit of the TyG index, ln(fasting triglycerides x fasting glucose / 2), both in mg/dL" }, "published_b": [0.431782416425538, 0.19062035960865, 0.667829372575655], "constants": null, "original": { "estimate": 1.53249416643178, "ci": [1.17692499503039, 1.99548686625251], "b": 0.426896582251184, "se": 0.133298897545575, "df": 117, "p": 0.00175480412112078, "n": 3111, "events": 359, "weighted": true, "weights": ["WTSAF2YR", "WTSAFPRP"], "dropped": [], "psu": 231, "strata": 114 }, "harmonized": { "estimate": 1.53249416643178, "ci": [1.17692499503039, 1.99548686625251], "b": 0.426896582251184, "se": 0.133298897545575, "df": 117, "p": 0.00175480412112078, "n": 3111, "events": 359, "weighted": true, "weights": ["WTSAF2YR", "WTSAFPRP"], "dropped": [], "psu": 231, "strata": 114 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.988684499441175, "published_z": 3.54678123151979, "interval_ratio_original": 0.903834112681984, "ratio_harmonized": 1, "variants": [ { "label": "unweighted", "estimate": 1.40922268203916, "ci": [1.18918950520939, 1.66996812440249], "b": 0.343038263040313, "se": 0.0866170385584347, "df": null, "p": 7.48238029943878e-05, "n": 3318, "events": 386, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "examination weight (WTMEC2YR), keeping those without a fasting weight", "estimate": 1.53427064132092, "ci": [1.19883103157501, 1.96356812496475], "b": 0.42805511588464, "se": 0.124571927271354, "df": 117, "p": 0.000817347898106275, "n": 3318, "events": 386, "weighted": true, "weights": ["WTMEC2YR", "WTMECPRP"], "dropped": [], "psu": 231, "strata": 114 }, { "label": "2017-2018 and 2017-March 2020 files stacked, as the paper's counts show", "estimate": 1.60783917618804, "ci": [1.26264720769606, 2.04740231533252], "b": 0.47489115094653, "se": 0.122178260407878, "df": 132, "p": 0.000160027382769183, "n": 3626, "events": 417, "weighted": true, "weights": ["WTSAF2YR", "WTSAFPRP"], "dropped": [], "psu": 261, "strata": 129 }, { "label": "borderline diabetes not counted", "estimate": 1.59189167949119, "ci": [1.21267222330856, 2.08969832946233], "b": 0.46492304458672, "se": 0.137391502397082, "df": 117, "p": 0.000972890799983072, "n": 2889, "events": 337, "weighted": true, "weights": ["WTSAF2YR", "WTSAFPRP"], "dropped": [], "psu": 231, "strata": 114 }, { "label": "HbA1c 5.29% or more counts as diabetes, as printed", "estimate": 1.48931209124323, "ci": [1.27481946789212, 1.73989381319284], "b": 0.398314329619246, "se": 0.0785363186439721, "df": 119, "p": 1.46644223089086e-06, "n": 11598, "events": 1037, "weighted": true, "weights": ["WTSAF2YR", "WTSAFPRP"], "dropped": [], "psu": 233, "strata": 114 }, { "label": "missing LDL or HDL cholesterol not excluded", "estimate": 1.33739564798909, "ci": [1.07347521256305, 1.66620253390816], "b": 0.290724176549217, "se": 0.110996589319075, "df": 117, "p": 0.00998029787349661, "n": 3216, "events": 370, "weighted": true, "weights": ["WTSAF2YR", "WTSAFPRP"], "dropped": [], "psu": 231, "strata": 114 }, { "label": "Model I (published 1.60, 1.29-1.99)", "estimate": 1.59850966257275, "ci": [1.23111339624123, 2.07554653303258], "b": 0.469071734274434, "se": 0.131865530960896, "df": 117, "p": 0.000542308582475511, "n": 3111, "events": 359, "weighted": true, "weights": ["WTSAF2YR", "WTSAFPRP"], "dropped": [], "psu": 231, "strata": 114 }, { "label": "Model II (published 1.64, 1.30-2.07)", "estimate": 1.6112265326517, "ci": [1.227200386415, 2.11542545802534], "b": 0.476995710480446, "se": 0.13747409633828, "df": 117, "p": 0.000730293267024677, "n": 3111, "events": 359, "weighted": true, "weights": ["WTSAF2YR", "WTSAFPRP"], "dropped": [], "psu": 231, "strata": 114 } ], "left_out": [], "choices": [ { "choice": "cycles", "decision": "2005-2016 and the 2017-March 2020 files, without the 2017-2018 files", "reason": "the paper's starting count (85,750) is the 2005-2018 files plus the 2017-March 2020 files, which repeat the 2017-2018 participants; stacked (variant) the estimate rises to 1.61 on 3,624, farther from the published, so the paper's final sample probably lost the duplicates" }, { "choice": "weights", "decision": "fasting subsample weight (WTSAF2YR; WTSAFPRP in 2017-March 2020), pooled by years", "reason": "the paper used 'the recommended weights', which NCHS gives as the fasting subsample weight for fasting triglycerides and glucose; the examination weight (variant) gives nearly the same estimate (1.53)" }, { "choice": "HbA1c cutoff for diabetes", "decision": "6.5%", "reason": "printed as 5.29%, which gives an analytic sample of 11,598 against the paper's 3,225 (variant); 6.5% is the standard cutoff its cited definitions use" }, { "choice": "borderline diabetes", "decision": "a self-reported borderline diagnosis (DIQ010 = 3) counts as self-reported diabetes", "reason": "the TyG quartile cutpoints in the Methods (8.61, 9.03, 9.47) and Table 1's mean TyG (9.07) and glucose (146.3) match only with it (8.60, 9.03, 9.47 over everyone with TyG values; 9.06; 147.3; without it 8.65, 9.06, 9.52; 9.10; 151.3), and Models I-III then match (1.60, 1.61, 1.53 against 1.60, 1.64, 1.54)" }, { "choice": "LDL and HDL cholesterol", "decision": "participants missing either are excluded", "reason": "the paper excludes those missing 'other study variables' and lists LDL and HDL among them; Table 1's triglyceride mean (142.7 mg/dL) is that of participants with LDL, which NHANES computes only for triglycerides of 400 or less (160 without the step, whose estimate is 1.34)" }, { "choice": "type 2 diabetes", "decision": "every diabetic adult; type 1 not separated", "reason": "the paper doesn't say how it separated type 1; leaving out those diagnosed before 30 who take only insulin (70) gives 1.49" }, { "choice": "depression", "decision": "PHQ-9 total of 10 or more, missing if any item is missing, refused, or don't know", "reason": "as stated; item nonresponse unstated" }, { "choice": "alcohol", "decision": "average drinks a week over the past 12 months (drinking days a year times drinks a drinking day over 52): none 0, moderate under 8, heavy 8 or more; never drinkers and past-year abstainers drink none", "reason": "the paper's categories (0, 1-8, 8 or more), which overlap at 8; Table 1's counts (1,204, 1,162, 368: 2,734) and Table 2's (1,204, 1,530, 491: 3,225) disagree, unresolved" }, { "choice": "smoking", "decision": "never, former, current from SMQ020 and SMQ040", "reason": "convention; Table 1's 'Never' (533) and 'Now' (1,664) labels look swapped (here 1,671 never and 556 current)" }, { "choice": "covariate coding", "decision": "age continuous; marital status in three groups (married or with a partner; widowed, divorced, or separated; never married); race in five (RIDRETH1); BMI under 25, 25 to under 30, 30 or more; heart failure (MCQ160B) and coronary heart disease (MCQ160C) as told; complete case", "reason": "Table 1's categories (its 'Divorced' holds widowed and separated, as its counts show); the paper excludes missing covariates" }, { "choice": "participants with TyG values but no fasting weight", "decision": "in the population, out of the weighted fit (206; they fasted too briefly)", "reason": "the paper's counts include everyone with TyG values; a fasting-weighted analysis gives them no weight" } ], "departures": [ { "kind": "model", "affects_headline": true, "followed": true, "detail": "Section 2.3 says its confounders, among them blood glucose, triglycerides, HDL, and LDL, were all included in the analyses, but Model III's estimate is reproduced without them, with the Table 3 footnote's covariates (1.53 against 1.54), and not with glucose and triglycerides added (3.05, 1.17-7.91)." }, { "kind": "reporting", "affects_headline": true, "followed": true, "detail": "The Methods count an HbA1c of 5.29% or more as diabetes, but that cutoff gives 11,598 participants against the published 3,225; with 6.5%, the sample gives the Methods' TyG quartile cutpoints (8.60, 9.03, 9.47 against 8.61, 9.03, 9.47) and Model III's estimate (1.53 against 1.54)." }, { "kind": "sample", "affects_headline": false, "followed": null, "detail": "Figure 1 starts from 85,750 participants, the 2005-2018 files plus the 2017-March 2020 files, which repeat the 2017-2018 participants, but the final sample and estimate fit the files without the repeats (weighted fits: 3,111 participants and 1.53 without them, 3,626 and 1.61 with them; published 3,225 and 1.54)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Tables 1 and 2 label triglycerides, HDL, and LDL in mmol/L, but the values are in mg/dL (triglycerides 142.66, against 142.6 mg/dL here)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Tables 1 and 2 swap the smoking labels Never and Now: their 533 'never' smokers are the current smokers (556 here) and their 1,664 'now' smokers the never smokers (1,671)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's alcohol counts (1,204, 1,162, 368) sum to 2,734 and their percentages to 84.9%, while Table 2 gives 1,204, 1,530, and 491 for the same 3,225 participants." } ], "projected_power": 0.225951521093319 }, "row315": { "id": "row315", "row": 315, "doi": "10.3389/fnut.2024.1396470", "measure": "OR", "contrast": "dietary selenium quartile 4 (over 0.144 mg/day) vs quartile 1 (0.072 mg/day or less)", "published": { "measure": "OR", "estimate": 0.77, "low": 0.63, "high": 0.95, "n": 6390, "events": 1523, "contrast": "dietary selenium quartile 4 (over 0.144 mg/day) vs quartile 1 (0.072 mg/day or less)" }, "published_b": [-0.261364764134408, -0.462035459596559, -0.0512932943875506], "constants": null, "original": { "estimate": 0.759716364855846, "ci": [0.620640810621233, 0.929956498432745], "b": -0.2748101194968, "se": 0.103161410762667, "df": null, "p": 0.00772439983272524, "n": 6738, "events": 1603, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 0.759716364855846, "ci": [0.620640810621233, 0.929956498432745], "b": -0.2748101194968, "se": 0.103161410762667, "df": null, "p": 0.00772439983272524, "n": 6738, "events": 1603, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.05144287680446, "published_z": -2.49434106318535, "interval_ratio_original": 1.01571982423802, "ratio_harmonized": 1, "variants": [ { "label": "weighted (WTDRD1)", "estimate": 0.68218168468598, "ci": [0.455907348247954, 1.02075970635134], "b": -0.382459256783995, "se": 0.197332346990081, "df": 30, "p": 0.0620664554447185, "n": 6738, "events": 1603, "weighted": true, "weights": "WTDRD1", "dropped": [], "psu": 60, "strata": 30 }, { "label": "crude (published 0.66, 0.56-0.78)", "estimate": 0.651822750348535, "ci": [0.554677696603535, 0.765981578984622], "b": -0.427982609308491, "se": 0.0823410290841158, "df": null, "p": 2.01787297468823e-07, "n": 6738, "events": 1603, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1 without age (published 0.77, 0.64-0.92)", "estimate": 0.755151395938193, "ci": [0.6319306895221, 0.902399013440316], "b": -0.280837025413999, "se": 0.0908886771992403, "df": null, "p": 0.0020022314133367, "n": 6738, "events": 1603, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2 with age, as the text describes", "estimate": 0.860835234518941, "ci": [0.698124591356249, 1.06146855470263], "b": -0.149852158118012, "se": 0.106892543964103, "df": null, "p": 0.160946505448875, "n": 6738, "events": 1603, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 1 with age, as the text describes", "estimate": 0.920264100493824, "ci": [0.764637268397829, 1.10756570423544], "b": -0.0830945843803843, "se": 0.0945217022570357, "df": null, "p": 0.379343869089977, "n": 6738, "events": 1603, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "CKD by MDRD with its race term", "estimate": 0.809227528670193, "ci": [0.657008987739401, 0.996712686398454], "b": -0.211675154655, "se": 0.106319517915198, "df": null, "p": 0.04648813171327, "n": 6738, "events": 1452, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "self-reported BMI (WHD010, WHD020)", "estimate": 0.791851549630472, "ci": [0.64532456204054, 0.971648862503375], "b": -0.233381342071547, "se": 0.10440015802734, "df": null, "p": 0.0253878151545718, "n": 6541, "events": 1562, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "supplement selenium in the recall's 24 hours (DS1TSELE)", "estimate": 0.758890906756433, "ci": [0.619896007434764, 0.929051649712728], "b": -0.27589724478079, "se": 0.103219396769607, "df": null, "p": 0.00751940613140868, "n": 6738, "events": 1603, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "adding smoking and income (Figure 2's adjustment)", "estimate": 0.777225112981044, "ci": [0.634515058388538, 0.952032372222277], "b": -0.252025249879102, "se": 0.103506498663018, "df": null, "p": 0.0148969890026605, "n": 6738, "events": 1603, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "hypertension also by measured pressure 140/90 or more", "estimate": 0.77150317669904, "ci": [0.629991112295869, 0.944802458383209], "b": -0.259414489625455, "se": 0.103387143404598, "df": null, "p": 0.0121019964401321, "n": 6738, "events": 1603, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": [], "choices": [ { "choice": "cycles", "decision": "2015-2016 and 2017-2018", "reason": "the paper's 19,225 participants are DEMO_I and DEMO_J exactly" }, { "choice": "weights", "decision": "unweighted", "reason": "the paper (SPSS) names no weights or design; its crude ORs are the unweighted ones of Table 2's counts (0.65 here against 0.66) and its CIs have unweighted widths; weighted variant" }, { "choice": "age in the models", "decision": "left out, as the paper's computation left it out", "reason": "the text says the models adjust for age, but its adjusted estimates are reproduced only without age: Model 1 0.76 (0.63-0.90) against 0.77 (0.64-0.92), Model 2 0.76 (0.62-0.93) against 0.77 (0.63-0.95), and per mg/day 0.22 and 0.20 against 0.23 and 0.23 (0.73 and 0.43 with age); the published estimate comes from the models without age, and the text's models are variants" }, { "choice": "CKD", "decision": "MDRD eGFR (175, standardized creatinine) under 60 without the equation's race term, or ACR 30 mg/g or more", "reason": "the paper cites the 2006 MDRD equation without its terms; without the race term 23.8% have CKD, as in the paper (21.5% with it), and the CKD shares of the four race groups (NH White 26.2%, Mexican American 19.6%, NH Black 32.0%, other 16.4%) match Table 1's (25.9, 20.2, 32.1, 16.4%), while with the term NH Black's is 21.1%; no creatinine correction, since 2015-2018 creatinine is standardized and the NHANES III correction the paper cites is for older surveys" }, { "choice": "exposure", "decision": "day-one recall selenium (DR1TSELE) in the paper's quartiles (up to 72, 103, 144 mcg, and over)", "reason": "the Methods describe one day's recall, and day-one quartile shares (26.2, 25.3, 25.0, 23.5%) match Table 2's (25.8, 25.4, 24.9, 23.9%) while the two-day mean's (22.6, 27.6, 28.8, 21.0%) do not" }, { "choice": "population", "decision": "aged 30 to 80 (80 is 80 and over), not pregnant (RIDEXPRG), reliable day-one recall, serum creatinine and urine ACR, and every listed covariate known", "reason": "the paper's exclusions for basic, laboratory, and survey information are unspecified; this gives 6,738 against its 6,390, with the same CKD share; leaving out age 80 gives 21.0% with CKD" }, { "choice": "Model 2 covariates", "decision": "Table 2's, without age (see above): sex, race, marital status, education, BMI, hypertension, diabetes, carbohydrate, selenium supplements, CRP, total cholesterol, triglycerides, uric acid", "reason": "the headline is Table 2's; Figure 2's model adds smoking and the income-to-poverty ratio (variant)" }, { "choice": "race", "decision": "four groups: Mexican American, NH White, NH Black, everyone else (other Hispanic and other races)", "reason": "Table 1 has four groups whose counts and CKD shares match these, though under rotated labels (its 'Non-Hispanic white' is other Hispanic and other races, its 'Non-Hispanic black' NH White, its 'Others' NH Black); the text's CKD patients 'more likely non-Hispanic whites' fits this reading" }, { "choice": "hypertension", "decision": "told by a doctor (BPQ020)", "reason": "Table 1's 35.1% and 65.5% fit told alone (35.5%, 65.3%), not told or measured pressure 140/90 or more (44%, 74%; variant)" }, { "choice": "diabetes", "decision": "told by a doctor (DIQ010 = 1; borderline counts as no)", "reason": "the paper's definition; Table 1's 13.2% and 34.7% against 13.0% and 33.4%" }, { "choice": "selenium supplements", "decision": "any selenium from supplements in the past 30 days (DSQTSELE above 0)", "reason": "unstated; Table 1's 10.1% and 12.0% fit neither this (22.5%, 25.4%) nor the recall's 24 hours (16.4%, 21.5%), which gives the same estimate (variant); 2021-2023 releases only the 30-day file" }, { "choice": "BMI", "decision": "measured (BMXBMI)", "reason": "Table 1's means (28.74, 29.59) are below measured BMI here (29.8, 31.0) and nearer self-reported BMI (29.0, 29.9), but neither matches; measured BMI is the examination's and gives the estimate nearer the published one (variant: self-reported, 0.89)" }, { "choice": "covariate coding", "decision": "age, BMI, carbohydrate (g), CRP (mg/L), total cholesterol, refrigerated-serum triglycerides, and uric acid continuous; living with a partner or alone; education in three levels", "reason": "Table 1's categories; its triglycerides (1.75, 1.90 mmol/L) are the refrigerated-serum ones (1.75, 1.91) on a sample too large for the fasting subsample; units do not change the selenium estimate" } ], "departures": [ { "kind": "model", "affects_headline": true, "followed": true, "detail": "The Methods and Table 2's note adjust Models 1 and 2 for age, but their estimates are reproduced only without age (quartile 4 vs 1: 0.76 and 0.76 against 0.77 and 0.77; with age 0.92 and 0.86), as are the per mg/day estimates (0.22 and 0.20 against 0.23 and 0.23; with age 0.73 and 0.43)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods compute eGFR with the 2006 MDRD Study equation (reference 26), which multiplies by 1.212 for Black people, but the paper's CKD share (23.8%) and Table 1's CKD shares by race are reproduced without that term (23.8%; non-Hispanic Black, the row Table 1 labels Others, 32.0% against 32.1%), not with it (21.5%; 21.1%)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods define hypertension by a doctor's diagnosis and systolic pressure of 140 mmHg or more or diastolic of 90 or more, but Table 1's shares (35.1% without CKD, 65.5% with) are reproduced by the diagnosis alone (35.5%, 65.3%), not with measured pressure added (44%, 74%)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's race labels are rotated: by their counts and CKD shares, its Non-Hispanic white row is other Hispanic and other races, its Non-Hispanic black row non-Hispanic White, and its Others row non-Hispanic Black." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The abstract says the risk of CKD falls 7.7% for every additional 0.1 mg of selenium, but Table 2's odds ratios per mg/day (0.23 in Models 1 and 2, reproduced per mg/day) correspond to 0.86 per 0.1 mg, 14% lower odds." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 2's per-SD row gives Model 1 as 0.93 (0.81-1.05) with p = 0.025, though its interval includes 1, and Model 2 as 0.93 (0.83-1.09) with p = 0.059, an interval not centered on its estimate on the log scale." } ], "projected_power": 0.378863047835739 }, "row319": { "id": "row319", "row": 319, "doi": "10.3390/antiox12091740", "measure": "OR", "contrast": "per SD of the CDAI (sum of the six intakes, each minus its mean over its SD, means and SDs from the paper's sample)", "published": { "measure": "OR", "estimate": 0.67, "low": 0.51, "high": 0.88, "n": 3109, "events": 453, "contrast": "per SD of the CDAI (sum of the six intakes, each minus its mean over its SD, means and SDs from the paper's sample)" }, "published_b": [-0.400477566597125, -0.673344553263766, -0.127833371509885], "constants": { "cdai": { "vitamin_a": { "mean": 609.883724670312, "sd": 572.008010715923 }, "vitamin_c": { "mean": 78.3106304277903, "sd": 68.5632643959913 }, "vitamin_e": { "mean": 7.69240752653586, "sd": 4.95580739943201 }, "zinc": { "mean": 8.63037793502734, "sd": 4.18880192427227 }, "selenium": { "mean": 91.7840141524606, "sd": 40.5372134889214 }, "carotenoids": { "mean": 8843.54583467353, "sd": 9773.92059625106 } }, "cdai_sd": 4.10573730839288, "cdai_sd_weighted": 3.95257095613415 }, "original": { "estimate": 0.867526052371815, "ci": [0.682671256999129, 1.10243611962234], "b": -0.142109736059536, "se": 0.118977070718924, "df": 45, "p": 0.238567904793855, "n": 2681, "events": 372, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 }, "harmonized": { "estimate": 0.889570911805353, "ci": [0.706793193764533, 1.11961520585021], "b": -0.117016054080942, "se": 0.1141952868061, "df": 45, "p": 0.310981236193932, "n": 2802, "events": 395, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 }, "reproduced": true, "same_sign_original": true, "ratio_original": 0.354850678071804, "published_z": -2.87774708713759, "interval_ratio_original": 1.1382263335918, "ratio_harmonized": 0.823420388536357, "variants": [ { "label": "Model A, crude (published 0.65, 0.53-0.80)", "estimate": 0.765933615483815, "ci": [0.613847897591909, 0.95569978431061], "b": -0.266659776856324, "se": 0.109899188731295, "df": 45, "p": 0.0193190029911477, "n": 2681, "events": 372, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "Model A, crude, own sample", "estimate": 0.784589022705559, "ci": [0.650119113027604, 0.946872537992784], "b": -0.242595236248276, "se": 0.0933439896177945, "df": 45, "p": 0.0125997012833794, "n": 3109, "events": 453, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "Model B (published 0.71, 0.58-0.87)", "estimate": 0.828251835518831, "ci": [0.662855423195267, 1.03491814207912], "b": -0.18843802166353, "se": 0.110600258509196, "df": 45, "p": 0.0953203508811169, "n": 2681, "events": 372, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "Model B, own sample", "estimate": 0.854222370929981, "ci": [0.708994515327399, 1.02919817180854], "b": -0.157563731582677, "se": 0.0925194593844851, "df": 45, "p": 0.0954602691902319, "n": 3072, "events": 441, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "unweighted", "estimate": 0.917426700492216, "ci": [0.782481094473154, 1.07564483886059], "b": -0.0861825927165243, "se": 0.0811764531470223, "df": null, "p": 0.288385591954167, "n": 2681, "events": 372, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "age and BMI continuous", "estimate": 0.867021181817079, "ci": [0.682763328800283, 1.10100484019899], "b": -0.142691871338464, "se": 0.118621082995379, "df": 45, "p": 0.235297575044157, "n": 2681, "events": 372, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "HDL-C continuous", "estimate": 0.869636000563412, "ci": [0.68600441470033, 1.10242260438847], "b": -0.139680545040088, "se": 0.11776489336302, "df": 45, "p": 0.241809073928092, "n": 2681, "events": 372, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "2017-2018 alcohol: monthly drinking, never drinkers missing", "estimate": 0.860836326615908, "ci": [0.679462386811189, 1.0906257588435], "b": -0.149850889470737, "se": 0.117472873027814, "df": 45, "p": 0.208635230085714, "n": 2555, "events": 360, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 }, { "label": "per weighted SD of the CDAI", "estimate": 0.8721374256282, "ci": [0.692462767297714, 1.09843261631187], "b": -0.136808269292971, "se": 0.114538578298283, "df": 45, "p": 0.238567904793856, "n": 2681, "events": 372, "weighted": true, "weights": "WTMEC2YR", "dropped": [], "psu": 90, "strata": 45 } ], "left_out": ["marital status in six groups (2021-2023 releases only DMDMARTZ's three groups)", "sleep disorders (2021-2023 does not ask SLQ050, whether one told a doctor about trouble sleeping)", "family history of heart attack or angina before 50 (2021-2023 does not ask MCQ300A)", "alcohol use as at least 12 drinks in any one year (asked as ALQ101 only through 2016; 2021-2023 does not ask it)"], "choices": [ { "choice": "postmenopausal", "decision": "RHD043 = 7 (menopause/change of life), or any woman aged 55 or older whatever her menstrual answers", "reason": "the only reading that gives the flow chart's 4057 exactly (2415 by RHD043 plus 1642 women 55 or older who did not answer menopause); 2021-2023 asks RHD043 by self-interview (ACASI) rather than face to face" }, { "choice": "missing dietary data", "decision": "no reliable first-day recall (DR1DRSTZ not 1)", "reason": "gives the flow chart's 544 exactly; requiring both days would exclude 891" }, { "choice": "missing ASCVD data", "decision": "any of MCQ160C, D, E, F not answered yes or no", "reason": "gives the flow chart's 34 exactly, and the final 3109 with 453 cases, whose age and race counts match Table 1 exactly" }, { "choice": "daily intakes for the CDAI", "decision": "mean of the two recalls, the first day's where the second is missing or unreliable", "reason": "the paper averages two days yet keeps everyone with a first-day recall; this gives Table 1's weighted CDAI medians (-0.26 overall, -1.04 with ASCVD, -0.09 without; published -0.18, -1.14, -0.05) more closely than first-day intakes (-0.41, -1.08, -0.27)" }, { "choice": "carotenoids and vitamins", "decision": "carotenoids as alpha- plus beta-carotene, beta-cryptoxanthin, lycopene, and lutein with zeaxanthin; vitamin A as RAE; vitamin E as alpha-tocopherol", "reason": "unstated; the library's CDAI components, and the closest fit to Table 1's CDAI medians" }, { "choice": "standardizing means and SDs", "decision": "unweighted, over the paper's 3109 women (in_population)", "reason": "the paper standardizes over its subjects; unweighted means and SDs fit Table 1's medians, weighted ones shift them by about 0.25" }, { "choice": "SD for the per-SD estimate", "decision": "unweighted SD of the CDAI over the same 3109 women", "reason": "unstated; a variant uses the weighted SD" }, { "choice": "weights", "decision": "WTMEC2YR / 3 with strata and PSUs", "reason": "Table 1's weighted sample size equals the sum of WTMEC2YR / 3 over the analytic sample, to the person" }, { "choice": "missing covariates", "decision": "complete cases", "reason": "the paper imputed covariates with under 10% missing by random forest; complete cases drop women missing waist, lipids, NLR, BMI, family history, alcohol, or another covariate" }, { "choice": "age, BMI, income", "decision": "Table S1's groups (age 40-49 to 80, four BMI groups, income-to-poverty ratio up to 1.00, 1.01-3.00, above 3.00)", "reason": "Table S1 says each was categorized; a variant enters age and BMI continuous" }, { "choice": "HDL-C", "decision": "two levels, 50 mg/dL or more against less", "reason": "Table 1 shows HDL-C as 0 or 1 (median 1.00, IQR 0.00-1.00) and the stratified analysis splits it at 50 mg/dL; a variant enters it in mg/dL" }, { "choice": "energy and PUFA intake", "decision": "first-day recall: energy in Table 1's four groups, PUFA in grams", "reason": "Table 1's energy counts (1452, 719, 580, 358) and PUFA quartiles (15.02, 9.64-22.02) are the first day's exactly; the two-day mean gives other counts" }, { "choice": "alcohol use in 2017-2018", "decision": "at least 12 drinks in the past year (ALQ121 days times ALQ130 drinks), never and former drinkers none", "reason": "2017-2018 does not ask ALQ101 and the paper does not say what it used; Table 1's 1457 users (56.0% weighted) fit several readings once its imputed values are allowed for; a variant uses monthly drinking with never drinkers missing" }, { "choice": "moderate to vigorous recreational activity", "decision": "yes to PAQ650 or PAQ665 (in 2021-2023 any moderate or vigorous leisure-time activity, PAD790Q or PAD810Q above 0)", "reason": "Table S1; gives 1232 against Table 1's 1179, which no stated reading matches" }, { "choice": "hypertension", "decision": "told of high blood pressure (BPQ020)", "reason": "the advice to take medicine (BPQ040A) is asked only of those told, so the stated definition reduces to BPQ020" }, { "choice": "diabetes", "decision": "told (DIQ010 = 1), insulin or pills, or HbA1c 6.5% or more, a missing part counting as not met", "reason": "Table S1; gives Table 1's 833 exactly" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": false, "detail": "On the paper's population, reproduced exactly, the CDAI as described gives Table 1's weighted medians (-1.04 with ASCVD and -0.09 without, against -1.14 and -0.05), but not Table 2: its quartile cutpoints (-1.04, 1.11, 3.72) lie above Table 1's quartiles (-2.27, -0.18, 2.46), and its crude odds ratio per SD (0.65, 0.53-0.80) is 0.78 (0.65-0.95) here under every CDAI construction tried." }, { "kind": "sample", "affects_headline": true, "followed": true, "detail": "The text adds women 55 or older with no menstrual information to those giving menopause as the reason for no periods, but its 4,057 postmenopausal women are reproduced only by adding every woman 55 or older whatever her menstrual answers (2,415 plus 1,642)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text computes intakes as the average of the two recalls, but the 544 excluded for missing dietary data are exactly those without a reliable first-day recall (requiring both recalls would exclude 891), so women with only a first-day recall were kept; averaging the recalls each woman has fits Table 1's CDAI medians more closely than first-day intakes." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1's counts for BMI 18.5-24.9 and 25.0-29.9 (1,297 and 356) do not fit their weighted percentages (25.4% and 29.7%), which the reproduced population gives from 715 and 923 women." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1 labels HDL-C in mg/dL but reports a 0/1 variable (median 1.00, interquartile range 0.00 to 1.00)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The Results give Model B's per-SD interval as 0.58-0.81, where Table 2 has 0.71 (0.58, 0.87)." } ], "projected_power": 0.343173294076359 }, "row327": { "id": "row327", "row": 327, "doi": "10.3389/fnut.2024.1415484", "measure": "OR", "contrast": "SII quartile 4 (625.74-28,397.28) vs quartile 1 (1.53-313.50); SII = platelets x neutrophils / lymphocytes (10^3 cells/uL)", "published": { "measure": "OR", "estimate": 1.47, "low": 1.24, "high": 1.74, "n": 14413, "events": 6518, "contrast": "SII quartile 4 (625.74-28,397.28) vs quartile 1 (1.53-313.50); SII = platelets x neutrophils / lymphocytes (10^3 cells/uL)" }, "published_b": [0.385262400790645, 0.215111379616945, 0.553885113226438], "constants": null, "original": { "estimate": 1.60104242976752, "ci": [1.31225421329458, 1.95338436405573], "b": 0.470654935704327, "se": 0.101485795344917, "df": null, "p": 3.52403992345664e-06, "n": 12089, "events": 5299, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "harmonized": { "estimate": 1.58075856434767, "ci": [1.29514493748321, 1.92935753091411], "b": 0.457904835843331, "se": 0.101676473491327, "df": null, "p": 6.68283979879448e-06, "n": 12089, "events": 5299, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, "reproduced": true, "same_sign_original": true, "ratio_original": 1.22164772564994, "published_z": 4.45784519420571, "interval_ratio_original": 0.851581825578256, "ratio_harmonized": 0.972909877505233, "variants": [ { "label": "crude, all 14,414 (published 1.62, 1.48-1.78)", "estimate": 1.58827247388156, "ci": [1.44593079632405, 1.7446266845571], "b": 0.462646931155096, "se": 0.04790581255649, "df": null, "p": 4.57198595831413e-22, "n": 14414, "events": 6243, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "Model 2, own sample (published 1.63, 1.48-1.80)", "estimate": 1.62360109981509, "ci": [1.46488610027936, 1.7995122834588], "b": 0.484646583375263, "se": 0.0524851946207714, "df": null, "p": 2.60788863984684e-20, "n": 13066, "events": 5665, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "weighted (fasting subsample weight)", "estimate": 2.15039678994846, "ci": [1.6154074759337, 2.86256342323033], "b": 0.765652378576075, "se": 0.144075431502794, "df": 94, "p": 7.18145083211112e-07, "n": 11431, "events": 5071, "weighted": true, "weights": "WTSAF2YR", "dropped": [], "psu": 184, "strata": 90 }, { "label": "MASLD exclusions as stated, excluded dropped", "estimate": 1.68074132187753, "ci": [1.35294566472202, 2.08795627549984], "b": 0.519234959109613, "se": 0.110691202261996, "df": null, "p": 2.7208280176689e-06, "n": 10696, "events": 4645, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "MASLD exclusions as stated, counted as non-MASLD", "estimate": 1.58845795118019, "ci": [1.35378955707103, 1.86380420020862], "b": 0.462763703606028, "se": 0.0815606641725973, "df": null, "p": 1.39616110088153e-08, "n": 12089, "events": 4645, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "USFLI as printed (insulin not logged)", "estimate": 1.58609338551688, "ci": [1.33590966439002, 1.88313049500174], "b": 0.461274002637413, "se": 0.0875840309503816, "df": null, "p": 1.38939478637358e-07, "n": 12089, "events": 10275, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "diabetes with 2-h OGTT glucose (text)", "estimate": 1.59548729072521, "ci": [1.30761256338545, 1.94673848060552], "b": 0.467179201004159, "se": 0.101520333423947, "df": null, "p": 4.18797246223504e-06, "n": 12089, "events": 5299, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "activity as printed (vigorous counted once)", "estimate": 1.60458253205775, "ci": [1.31524381596078, 1.95757248271421], "b": 0.472863618111987, "se": 0.101451638836345, "df": null, "p": 3.14713961546952e-06, "n": 12089, "events": 5299, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "high triglycerides at 150 mg/dL", "estimate": 1.60301754952121, "ci": [1.31275667421007, 1.95745739827925], "b": 0.471887821490885, "se": 0.101919507398921, "df": null, "p": 3.65657030684235e-06, "n": 12089, "events": 5299, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null }, { "label": "alcohol, income missing as a level (n near MI)", "estimate": 1.60047465999771, "ci": [1.33149285741388, 1.92379487657931], "b": 0.470300247748684, "se": 0.0938790137258182, "df": null, "p": 5.45313858156889e-07, "n": 14314, "events": 6215, "weighted": false, "weights": null, "dropped": [], "psu": null, "strata": null } ], "left_out": "physical activity at work and walking or cycling to get places (GPAQ, PAQ605-PAD645): 2021-2023 asks only about leisure time; replaced by leisure-time moderate-to-vigorous activity of 150 minutes a week or more, vigorous minutes counted twice", "choices": [ { "choice": "weights", "decision": "unweighted", "reason": "the paper mentions no weights, strata, or PSUs; Table 1 gives unweighted counts and the supplement's models are SPSS's unweighted logistic regression output. Weighting with the fasting subsample weight (WTSAF2YR) is a variant" }, { "choice": "study population", "decision": "adults 20 and older with an SII (platelet, neutrophil, and lymphocyte counts) and at least one of FLI or USFLI", "reason": "gives 14,414 against the paper's 14,413, with Table 1's ALT, AST, and GGT means (25.02, 25.35, 29.67 against 25.01, 25.35, 29.67) and SII mean and SD (514.04 and 394.52 against 513.93 and 394.28), and 56 lacking a blood count against the paper's 57; requiring both indices gives 14,051. The flow chart's 'missing liver data (n = 20,000)' does not add up (34,770 - 20,000 - 57 = 14,713; 34,770 - 14,413 - 57 = 20,300)" }, { "choice": "MASLD exclusions", "decision": "none applied: the outcome is steatosis by FLI or USFLI", "reason": "the paper's numbers show the definition's exclusions (hepatitis B or C, heavy drinking, drugs, liver cancer, autoimmune liver disease) were not applied: its n is everyone with index and blood count data (dropping the 1,556 with an exclusion would leave 12,858), and its 6,518 cases exceed the 6,243 with steatosis (counting the excluded as non-cases would leave 5,512); Table 1's drinking shares (55.1% of non-cases, 51.1% of cases) fit no reclassification (ours 54.1% and 49.6%; 55.8% and 46.2% with heavy drinkers made non-cases). The stated definition is a variant both ways; its drug step (our list, since the paper names no drugs: amiodarone, methotrexate, tamoxifen, valproate, systemic glucocorticoids; 339 users) can't be built in 2021-2023 but is not part of the chosen version" }, { "choice": "USFLI insulin term", "decision": "ln(insulin), as Ruhl and Everhart published the index", "reason": "the paper prints 1.1792 x insulin; computed that way 85.1% of the sample has steatosis (variant), against the paper's 45.2%; with ln(insulin) 43.3% do" }, { "choice": "index inputs", "decision": "fasting triglycerides (LBXTR; LBXTLG in 2021-2023) in mg/dL, GGT in U/L, waist in cm, fasting insulin in uU/mL (GLU file through 2011-2012, INS after), fasting glucose in mg/dL; steatosis if either available index meets its cutoff", "reason": "the published indices' units; fasting triglycerides give the paper's sample size (serum triglycerides would make FLI available for 29,639 adults)" }, { "choice": "case count", "decision": "kept as the published formulas give it (6,243 cases against 6,518)", "reason": "the paper's 275 extra cases fall mostly where USFLI decides (Table 1: BMI under 25, 182 cases against our 97; Mexican Americans 1,260 against 1,168; Black participants 1,282 against 1,269), which points to larger USFLI values in part of the data. Codings tried that change the count: FLI from serum triglycerides (6,406), insulin in pmol/L (7,033; in 2009-2012 alone exactly 6,518, but other pairs of cycles come as close and none gives Table 1's race and BMI splits), lower cutoffs; none is used" }, { "choice": "SII quartiles", "decision": "the published cutpoints, boundaries between the printed ranges (313.505, 440.005, 625.735)", "reason": "Figure 3 legend; quartiles of 3,604, 3,605, 3,601, and 3,604, with the legend's minimum (1.53) and maximum (28,397.28) exactly" }, { "choice": "diabetes", "decision": "told by a doctor, insulin or diabetes pills, HbA1c 6.5% or more, or fasting glucose 126 mg/dL or more; no OGTT", "reason": "the text also lists 2-hour OGTT glucose of 200 or more, but Table 1's 2,806 with diabetes (under its swapped labels) match the definition without it (2,807) and not with it (3,118); the text's version is a variant" }, { "choice": "physical activity", "decision": "work, walking or cycling, and leisure minutes a week (minutes a day times days), vigorous minutes counted twice, 150 or more as sufficiently active; no answer counts as none", "reason": "the paper prints moderate plus vigorous minutes without the doubling and names no domains; with all domains and the doubling 8,725 are active against Table 1's 8,732 (8,505 without the doubling, variant; 4,675 with leisure only)" }, { "choice": "triglycerides and HDL covariates", "decision": "fasting triglycerides 160 mg/dL or more; HDL cholesterol 90 mg/dL or more", "reason": "Table 1 gives only yes and no counts (3,018 and 424, no missing): 160 gives 3,042 (150 gives 3,555, variant) and 90 gives 414 (89 gives 446); Table S4(A)'s HDL OR (0.325) fits HDL high, not low" }, { "choice": "alcohol use", "decision": "12 or more drinks in the past 12 months (drinking days times drinks a day)", "reason": "the paper's 'more than 12 drinks in the past year'; 52.1% of those with data drink so, against Table 1's 53.26% (more than 12: 47.8%)" }, { "choice": "hypertension", "decision": "told by a doctor, taking medicine for it, or mean pressure 140/90 or more (auscultatory readings; oscillometric in 2021-2023)", "reason": "the paper's definition; 6,109 against the 6,007 Table 1 shows under its swapped labels" }, { "choice": "other covariates", "decision": "sex; age continuous; RIDRETH1's five groups; income to poverty ratio under 1.3; education below high school or not; marital status in three groups (DMDMARTZ's); health insurance (HIQ011); smoked 100 cigarettes (SMQ020); CVD as told of heart failure, coronary heart disease, angina, heart attack, or stroke; waist 102 cm or more (men) or 88 (women); BMI under 25, 25-29.9, 30 or more; ALT, AST, GGT continuous", "reason": "the paper's definitions; Table 1's counts are matched (education 3,590 against 3,593, marital groups within 2, insurance 11,257 against 11,273, tobacco 6,397 against 6,399, waist 8,204 exactly)" }, { "choice": "missing covariates", "decision": "complete case (n = 12,089)", "reason": "the paper imputed income, education, marital status, insurance, tobacco, alcohol, and BMI by chained equations (m = 5), which this pipeline can't; keeping missing alcohol and income as their own levels (n = 14,314) gives the same estimate (variant)" }, { "choice": "harmonized physical activity", "decision": "leisure-time moderate-to-vigorous minutes a week, vigorous counted twice, 150 or more", "reason": "2021-2023 asks only about leisure-time activity" } ], "departures": [ { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text defines MASLD as steatosis (FLI 60 or more, or USFLI 30 or more) without hepatitis B or C, excessive alcohol or drugs, liver cancer, or autoimmune liver disease, but the counts fit steatosis with none of these exclusions: the 14,413 participants are everyone with an index and a blood count (14,414 here; 12,858 with those excluded dropped), and the 6,518 cases exceed the 6,243 with steatosis here (5,512 with those excluded counted as non-cases)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The Methods print USFLI with a 1.1792 x insulin term, but the paper's share with steatosis (6,518 of 14,413, 45.2%) fits the published index's ln(insulin) (43.3%), not the printed term (85.1%)." }, { "kind": "coding", "affects_headline": true, "followed": false, "detail": "With USFLI as published, the paper's 6,518 cases still exceed the 6,243 that FLI and USFLI give for the same population (14,414 here, with Table 1's ALT, AST, GGT, and SII means matched), the extra cases falling where USFLI decides (182 cases with BMI under 25 in Table 1 against 97 here); no coding tried gives Table 1's race and BMI splits." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The text's diabetes definition includes a 2-hour OGTT glucose of 200 mg/dL or more, but Table 1's count with diabetes (2,806, under a swapped label) is that of the definition without it (2,807 here; 3,118 with it)." }, { "kind": "coding", "affects_headline": true, "followed": true, "detail": "The printed MVPA formula adds moderate and vigorous minutes, but Table 1's 8,732 sufficiently active participants are reproduced with vigorous minutes counted twice (8,725; 8,505 counted once)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "Table 1 swaps the labels of the diabetes, hypertension, and income rows (its 2,806 without diabetes, 6,007 without hypertension, and 67.38% poor fit those with diabetes, 2,807 here, those with hypertension, 6,109, and those not poor, 67.9%), and its cardiovascular disease row repeats the health insurance row's figures for those without MASLD." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The counts in the text and Figure 1 do not add up: 34,770 less the 20,000 and 57 excluded is 14,713, not 14,413, and 14,413 less 6,518 is 7,895 (Table 1), not 7,985; the text also gives the age exclusion as 2,572 where Figure 1 has 25,072." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The abstract's quartile 4 odds ratios for NLR (1.25, 1.04-1.49) and PLR (1.29, 1.09-1.53) differ from Figure 3's Model 3 estimates (1.29, 1.09-1.52; 1.13, 0.96-1.33)." }, { "kind": "reporting", "affects_headline": false, "followed": null, "detail": "The text reports SII quartile 4 as 1.62 (1.48, 1.78) in the propensity-matched sample, but that is Figure 3's unadjusted estimate in the full sample; Table S5(A) gives the matched estimate as 1.556 (1.277, 1.897)." } ], "projected_power": 0.398928378706859 } }