BACKGROUND: Frequent use of antibiotics is thought to create selection pressure by clearing susceptible bacteria and allowing resistant bacteria to spread in a community. A cluster-randomized trial comparing 2 different frequencies of mass azithromycin distributions for trachoma provided a convenient experiment for determining the causal relationship between antibiotic consumption and antibiotic resistance. METHODS: Twenty-four communities were randomized to either annual or biannual mass azithromycin distributions for trachoma. Randomization was stratified on health catchment area and trachoma prevalence. Swabs were processed for the genetic macrolide resistance determinants ermB and mefA/E in a masked fashion from a random sample of 120 preschool children before treatment and another 120 children after 2 years of mass antibiotics. RESULTS: Macrolide resistance determinants were similar in the 12 annually and 12 biannually treated communities before treatment, with a median prevalence among preschool children of 20% (interquartile range [IQR], 10%-40%) in each group. By 24 months, macrolide resistance determinants were found more commonly in the biannually treated communities (median, 60% [IQR, 50%-80%]) than the annually treated communities (median, 40% [IQR, 20%-40%]; P < .001). Adjusting for baseline, the 24-month prevalence of macrolide resistance determinants in the biannual group was 29.4% higher than that of the annual group (95% confidence interval, 10.5%-56.7%). CONCLUSIONS: This randomized trial used direct genetic methods to confirm the causal relationship of community antibiotic consumption and antibiotic resistance. Communities randomized to less frequent use of antibiotics had a significantly lower prevalence of genetic antibiotic resistance determinants. CLINICAL TRIALS REGISTRATION: NCT00792922.