First-Year Infections Requiring Penicillin and Gut Bacterial Features at One Year: A Secondary Analysis of the Public ABIS Birth Cohort Dataset

Abstract Background Infections requiring antibiotics are common in infancy, and the first-year gut community is shaped by early exposures. Whether infection burden relates to it is unclear. We aimed to estimate the association between the ordinal burden of first-year infections requiring penicillin and the 230 publicly available gut bacterial features at one year, in a secondary analysis of the public ABIS dataset. Results Parent-reported first-year episodes of infection requiring penicillin (0, 1–2, >=3) were related to each of the 230 processed bacterial features in pre-specified, sequentially adjusted models (M0-M3). Per-feature linear regression used HC3 heteroskedasticity-robust standard errors, with one Benjamini-Hochberg false-discovery-rate (FDR) correction across all 230 features. Detection, sensitivity and multiple-imputation analyses (m = 20) were also performed. In the Model 3 complete-case analysis (n = 1,023), 3 of the 230 features reached q < 0.05, all negative: Coprococcus eutactus (beta = -0.283, 95% CI -0.403 to -0.164; q = 0.0008), Saccharimonadia unknown Species 103 (beta = -0.406; q = 0.0251) and Intestinimonas unknown Species 284 (beta = -0.191; q = 0.0486). Only the first two remained FDR-positive in the detection model; Intestinimonas unknown Species 284 did not and was least stable across sensitivity analyses. Imputation recovered the same three associations plus 21 exploratory signals. Conclusions A higher infection burden was associated with lower values for three of the 230 bacterial features at one year. The C. eutactus finding is concordant with a previously reported association in the same cohort, not independent. They do not support a causal role of penicillin or changes to prescribing.