# Analysis plan: do bodily load, structural limitation and slow transit map onto specific somatic depressive symptoms?

Registered before any association between the measures below and depressive symptoms was computed on unshuffled data.

## Hypotheses under test

An account of depression as accumulated physical load distributed across three body systems (structural support, connective-tissue force distribution, visceral transport) predicts that: (1) measures of each system track the symptoms it is said to produce, so structural limitation tracks psychomotor slowing more than slow bowel transit does, and slow transit tracks sleep, energy and appetite more than structural limitation does; (2) bodily measures relate to somatic symptoms beyond what cognitive-affective symptoms explain, and more than to cognitive-affective symptoms beyond somatic ones; (3) socially distributed load (low income) loads onto somatic more than cognitive-affective symptoms. We test these as stated, without endorsing them.

## Data

NHANES 2005-2006, 2007-2008 and 2009-2010 (US public domain), files DEMO, DPQ, BHQ, PFQ, BMX, BPX, CRP, GHB, TCHOL, HDL, BIOPRO, fetched by digest. Adults aged 20 or over with all nine PHQ-9 items answered (0-3).

## What we saw before registering

Codebooks; counts of adults and of non-missing values per measure; a run of the full pipeline on real data with PHQ-9 answers randomly permuted across participants (which destroys every association), to test the code. No association on unpermuted data was computed.

## Measures (`code/analyze.py`)

- Symptoms: somatic = mean of PHQ items 3 (sleep), 4 (energy), 5 (appetite), 8 (psychomotor); vegetative = mean of items 3-5; psychomotor = item 8; cognitive-affective = mean of items 1, 2, 6, 7, 9.
- Structural limitation: mean difficulty (0 none to 3 unable) over seven load-bearing tasks (PFQ061 B walk quarter mile, C ten steps, D stoop/kneel, E lift/carry, I stand from chair, M stand long, T push/pull), at least five answered; "do not do" excluded. Adults under 60 who screened negative on every limitation question (and so were not asked) score 0; respondents needing equipment to walk score the two walking items as unable.
- Slow transit: Bristol stool type 1-2 or fewer than three bowel movements a week (one answer suffices when it indicates slow transit).
- Load: body mass index.
- Allostatic load: number of nine markers in the adverse sample quartile (systolic and diastolic pressure, waist, HbA1c, CRP, total cholesterol, albumin and HDL in the lowest quartile, creatinine), scaled to nine when at least seven are available.
- Low income: negative family income-to-poverty ratio.

## Tests

All associations are partial Spearman correlations (rank-residualised), adjusted for age, sex and survey cycle; bootstrap 95% CIs and two-sided percentile p-values (2,000 resamples, seed 20261007); unweighted.

- **H1 (double dissociation).** H1a: partial r(structural, psychomotor | transit) minus partial r(transit, psychomotor | structural) > 0. H1b: partial r(transit, vegetative | structural) minus partial r(structural, vegetative | transit) > 0. Supported only if both differences are positive with CIs excluding zero.
- **H2 (beyond mood, and specificity).** For each of structural limitation, slow transit, BMI and allostatic load: partial r with somatic adjusting for cognitive-affective (Holm across four), and that minus partial r with cognitive-affective adjusting for somatic (Holm across four). Supported for a measure if both are positive after adjustment.
- **H3 (social load).** Same specificity difference for low income.
- **Secondary.** Partial correlation of each measure with each PHQ item.

## Limits stated in advance

Cross-sectional; no course or treatment response. PHQ items 3 and 5 are bidirectional (too much or too little), so melancholic and atypical profiles cannot be separated. No measure of connective-tissue force distribution is available. Unweighted analyses are not nationally representative estimates. Somatic depressive symptoms overlap physical illness, which can produce associations without any shared load.