Mass azithromycin distribution and antibiotic resistance in the gut and nasopharynx: a cluster-randomized trial.

The Introduction cites prior MORDOR and AVENIR trials on azithromycin MDA and mortality, as well as WHO guidelines. The rationale follows logically: repeated azithromycin MDA reduces mortality but risks AMR selection, and the study aims to quantify that risk across different age-targeting strategies. The Discussion explicitly addresses how limitations of prior studies (e.g., lack of placebo arm, spillover) are being addressed in ongoing work.

Randomization method is described (response-adaptive, biostatistician-generated sequence in R) and the unit (community/grap). Blinding is comprehensive: all study personnel, participants, laboratory, and field workers masked; treatment concealment via identical packaging. Power analysis is reported for both nasopharyngeal and rectal swab primary endpoints with assumptions based on prior data. Inclusion/exclusion criteria are clearly prespecified at community and individual levels. Controls: both a placebo arm and an active comparator (infant-azithromycin) are used. No independent replication cohort is reported (common for a single large trial; this is the primary analysis). Outlier handling is not explicitly described but the pre-specified analysis approach (Wilcoxon) is non-parametric and robust. Replicate distinction (biological vs technical) is not explicitly defined but pooling at the village level is the unit of analysis, and the sample size accounts for community-level effects.

Table 1 provides complete demographics for both rectal and nasopharyngeal samples, including sex (%), age (median, IQR), and age categories (1–11 months vs 12–59 months). Sex is reported but not justified as a biological variable; the statistical analysis does not disaggregate by sex because the trial is cluster-randomized at the community level and analysis targets community effects. Demographics are adequate.

The ethics section names the University of California, San Francisco Committee for Human Research and the Comité National Éthique pour la Recherche en Santé in Niger as approving bodies. Informed consent is described: verbal consent from community leaders and written informed consent from guardians, documented electronically. Regulatory compliance is stated: "conducted according to Good Clinical Practice guidelines and adhered to the principles of the Declaration of Helsinki." No IACUC statement is needed (human study).

The investigational product (azithromycin oral suspension) is named, with manufacturer (Pfizer) and dosing specified. No antibodies, cell lines, or mycoplasma testing are used (n/a). Organisms are not applicable (human trial). Key reagents (DNA/RNA Shield, Zymo swabs, NEBNext Ultra II DNA Library Prep Kit, KAPA HyperExplore probes) are identified with vendors but not catalog numbers or lot numbers, which is adequate for the field. Software is well-identified: R (version 4.3.1), Kraken2 (version 2.1.2), Bracken (version 2.5), Burrows-Wheeler Aligner, Resistome Analyzer, DESeq2, Topconfects; custom code is not deposited in a public repo (data is available via BioProject, but code is not separately shared). The paper notes "requests for more information..." suggesting no public code repo.

The primary analysis uses Wilcoxon rank-sum tests with a gatekeeping procedure. P-values are reported exactly (e.g., P < 0.01, P = 0.66, P = 0.01). Effect sizes with 95% CIs are provided. Statistical software is identified (R version 4.3.1). Data presentation includes per-group n (50 communities per arm) and uses dot plots with error bars. Assumptions for parametric tests (t-distribution for CIs) are noted; the primary test is non-parametric, so normality/equal-variance verification is less critical. Mathematical plausibility is not applicable as the primary outcome is a log-transformed normalized read count.

Nonhost sequencing reads are deposited at BioProject ID PRJNA1337442. De-identified individual information is available via Dryad at 10.5061/dryad.p8cz8wb48. These are concrete access routes. The paper does not mention a repository for custom analysis code, stating only that requests for more information will be addressed. The code_sharing sub-criterion is thus reported_but_inadequate (no public repository).

Methods are very detailed, including sample collection, DNA extraction, sequencing, and statistical analysis. The trial is registered at ClinicalTrials.gov (NCT04224987). A reporting summary is linked (Nature Portfolio reporting summary). All prespecified outcomes are reported (primary AMR endpoints, secondary nonmacrolide endpoints, exploratory microbiome). Limitations are thoroughly discussed (genetic vs phenotypic resistance, cross-sectional design, pooling, generalizability, spillover). Conclusions are proportional: they state the findings (macrolide resistance selected in gut, not in nasopharynx, no co-selection) and do not overclaim to other classes or settings. Funding is clearly stated (Bill & Melinda Gates Foundation, NIH, Research to Prevent Blindness; Pfizer donated drugs; funders had no role). Conflicts are not explicitly stated but the paper includes a "Competing interests" section in the author contributions area (not fully visible in abstract, but the online content mentions competing interests are available). The acknowledgments thank Pfizer for donations, which could be seen as a potential conflict, but it is disclosed.