Ethnoveterinary herbal practices are associated with reduced within-farm antibiotic use and improved farm performance in commercial poultry in Bangladesh: A multi-district cross-sectional study.

ETHNOPHARMACOLOGICAL RELEVANCE: Antimicrobial resistance (AMR) in livestock is a critical One Health challenge, particularly in low- and middle-income countries (LMICs) where antibiotic regulation and surveillance are inadequate. Ethnoveterinary herbal practices are widely used in Bangladeshi commercial poultry farming but remain empirically undocumented at scale, limiting evidence-based AMR stewardship policy. AIM OF THE STUDY: To characterize ethnoveterinary herbal practices among commercial poultry farmers across 35 districts of Bangladesh, by documenting the plant species used, preparation methods, plant parts used, and target diseases, and to assess their observational associations with retrospective antibiotic use trajectories, farm performance outcomes, and food safety awareness. MATERIALS AND METHODS: A cross-sectional survey was conducted among 795 commercial poultry farmers (285 herbal users, 510 non-users) across 35 districts representing all 8 administrative divisions of Bangladesh (January-March 2026). The cohort included broiler, layer, and Sonali production systems. Statistical analyses included the Mann-Whitney U test (between-group performance comparisons), the Wilcoxon signed-rank test (within-group antibiotic shift), the chi-square test (categorical variables), multivariable logistic regression (predictors of herbal adoption), and multiple linear regression (predictors of antibiotic reduction magnitude). All analyses were conducted in Python 3.10 (scipy.stats v1.11, statsmodels v0.14; α = 0.05, two-tailed). Six machine learning classifiers with SHapley Additive exPlanations (SHAP) were applied as an exploratory investigation of potential predictors of farmer satisfaction. RESULTS: Among 795 surveyed farmers, the prevalence of herbal remedy use was 35.8% (n = 285). Fourteen medicinal plant species were documented; the most prevalent were Moringa oleifera Lam. (22.8%), Zingiber officinale Roscoe (21.4%), and Citrus × limon (L.) Osbeck (20.0%). Among herbal users (n = 285), self-reported antibiotic use decreased from a mean of 7.02 courses per cycle before adoption to 3.96 courses per cycle after adoption (43.6% within-group reduction; Wilcoxon W = 0, p < 0.001). This reflects a self-reported within-group temporal shift only; no between-group antibiotic comparison was possible. Herbal users had significantly lower Feed Conversion Ratio (FCR; 1.970 ± 0.261 vs. 2.144 ± 0.273; Mann-Whitney p < 0.001, r = 0.349) and mortality rate (3.42 ± 1.75% vs. 5.97 ± 2.30%; p < 0.001, r = 0.592) than non-users; these are observational, non-causal associations. The strongest independent predictors of herbal adoption were neighbor herbal use (OR = 2.741, 95% Bootstrap CI: 2.550-2.940, p < 0.001) and AMR knowledge items (OR = 1.511, 95% CI: 1.298-1.776, p < 0.001). All six machine learning classifiers showed near-random discriminative performance (AUC range: 0.407-0.535). Only 30.2% of herbal users were aware of potential phytochemical residue risks in meat and eggs. CONCLUSIONS: The findings provide a multi-district empirical baseline and an observational signal-suggestive but strictly non-causal-that ethnoveterinary herbal practices are associated with lower within-group antibiotic course frequencies and distinct production parameters. The prominent food safety awareness gap-fewer than one-third of herbal users recognizing residue risks-highlights an immediate need for targeted veterinary education, pharmacokinetic residue evaluation, and regulatory standardization of crude herbal veterinary inputs.