Estimating the population-level prevalence of antimicrobial-resistant enteric bacteria from latrine samples.

BACKGROUND: Logistical and economic barriers hamper community-level surveillance for antimicrobial-resistant bacteria in low-income countries. Latrines are commonly used in these settings and offer a low-cost source of surveillance samples. It is unclear, however, whether antimicrobial resistance prevalence estimates from latrine samples reflect estimates generated from randomly sampled people. METHODS: We compared the prevalence of antimicrobial-resistant enteric bacteria from stool samples of people residing in randomly selected households within Kibera-an informal urban settlement in Kenya-to estimates from latrine samples within the same community. Fecal samples were collected between November 2015 and Jan 2016. Presumptive Escherichia coli isolates were collected from each household stool sample (n = 24) and each latrine sample (n = 48), resulting in 8935 and 8210 isolates, respectively. Isolates were tested for resistance to nine antibiotics using the replica-plating technique. Correlation- and Kolmogorov-Smirnov (K-S) tests were used to compare results. RESULTS: Overall, the prevalence values obtained from latrine samples closely reflected those from stool samples, particularly for low-prevalence (< 15%) resistance phenotypes. Similarly, the distribution of resistance phenotypes was similar between latrine and household samples (r > 0.6; K-S p-values > 0.05). CONCLUSIONS: Although latrine samples did not perfectly estimate household antimicrobial resistance prevalence, they were highly correlated and thus could be employed as low-cost samples to monitor trends in antimicrobial resistance, detect the emergence of new resistance phenotypes and assess the impact of community interventions.

The impact of fecal sample processing on prevalence estimates for antibiotic-resistant Escherichia coli.

Investigators often rely on studies of Escherichia coli to characterize the burden of antibiotic resistance in a clinical or community setting. To determine if prevalence estimates for antibiotic resistance are sensitive to sample handling and interpretive criteria, we collected presumptive E. coli isolates (24 or 95 per stool sample) from a community in an urban informal settlement in Kenya. Isolates were tested for susceptibility to nine antibiotics using agar breakpoint assays and results were analyzed using generalized linear mixed models. We observed a <3-fold difference between prevalence estimates based on freshly isolated bacteria when compared to isolates collected from unprocessed fecal samples or fecal slurries that had been stored at 4°C for up to 7days. No time-dependence was evident (P>0.1). Prevalence estimates did not differ for five distinct E. coli colony morphologies on MacConkey agar plates (P>0.2). Successive re-plating of samples for up to five consecutive days had little to no impact on prevalence estimates. Finally, culturing E. coli under different conditions (with 5% CO2 or micro-aerobic) did not affect estimates of prevalence. For the conditions tested in these experiments, minor modifications in sample processing protocols are unlikely to bias estimates of the prevalence of antibiotic-resistance for fecal E. coli.