Study
Opening the study…
Study
Opening the study…
Where the work did something other than what it follows, its registered plan or the claim it replicates, as its author says.
From its pre-registered plan, under plan/
Done differently
code/analyze.R: fixed a bug in choosing an estimate for an age band. The registered code let an estimate stand for a band whenever its own band started no later and ended no earlier, so a 1 to 59 month estimate stood for 12 to 59 months; now its band must start within a month below and end within a month above. Only the secondary analysis of infants against children 12 to 59 months changed: with the registered code it pooled 3 comparisons (0.939, 95% CI 0.674 to 1.306), using AVENIR's and CHAT's 1 to 59 month estimates in place of their older-child estimates; with the fix it pools the 2 comparisons that print both bands (MORDOR in Tanzania, from its narrower bands, and CHAT). P1, P2, P3 and every other result are unchanged.
Done differently
code/analyze.R: R1, R2 and R3 also write each comparison's estimate as the analysis used it (`each`), for the paper's Table 1. Output only; no estimate changed.
Bears on C1, C2, and C3
Not stated
The plan didn't say what to do with a comparison whose sources print each arm's mortality rate with its 95% interval but no ratio and no counts. The registered rules leave such comparisons out: MORDOR II (so P1 pools 5 comparisons from 3 trials, not the 6 from 4 the plan listed), AVENIR's child arm against its infant arm on infant mortality (the registered robustness check of P3, which therefore couldn't be run and P3 is reported as not robust), and AVENIR's child arm against placebo at 12 to 59 months. code/summarize.R, added after extraction, estimates each from the two rates (log ratio, with a standard error from each rate's interval, the arms taken as independent) and reports them as exploratory, with P1 recomputed with MORDOR II added; and combines the two ratios AVENIR prints for infants with and without an older sibling.
Bears on C1 and C3
Done differently
Added after extraction (code/summarize.R): a common-effect estimate for the two individually randomized infant trials, because the registered random-effects estimate with Knapp-Hartung intervals from two trials has an interval too wide to inform (0.147 to 6.225); also with NAITRE's 6-month primary outcome in place of its 12-month estimate, which the registered code takes as its whole follow-up.
Bears on C5
Not stated
The plan said PRET's comparison would be reported as printed but not which of its printed age bands. No PRET entry is marked as the trial's primary analysis (its primary outcomes were trachoma), so the registered code takes the whole-follow-up estimate with the smallest variance, for children 6 months to under 12 years; the paper also reports the band PRET's report names as its mortality outcome, 6 months to under 5 years.