Validity of the estimates of oral cholera vaccine effectiveness derived from the test-negative design

Abstract

BackgroundThe test-negative design (TND) has emerged as a simple method for evaluating vaccine effectiveness (VE). Its utility for evaluating oral cholera vaccine (OCV) effectiveness is unknown. We examined this method's validity in assessing OCV effectiveness by comparing the results of TND analyses with those of conventional cohort analyses. MethodsRandomized controlled trials of OCV were conducted in Matlab (Bangladesh) and Kolkata (India), and an observational cohort design was used in Zanzibar (Tanzania). For all three studies, VE using the TND was estimated from the odds ratio (OR) relating vaccination status to fecal test status (Vibrio cholerae O1 positive or negative) among diarrheal patients enrolled during surveillance (VE= (1−OR)×100%). In cohort analyses of these studies, we employed the Cox proportional hazard model for estimating VE (=1−hazard ratio)×100%). ResultsOCV effectiveness estimates obtained using the TND (Matlab: 51%, 95% CI:37–62%; Kolkata: 67%, 95% CI:57–75%) were similar to the cohort analyses of these RCTs (Matlab: 52%, 95% CI:43–60% and Kolkata: 66%, 95% CI:55–74%). The TND VE estimate for the Zanzibar data was 94% (95% CI:84–98%) compared with 82% (95% CI:58–93%) in the cohort analysis. After adjusting for residual confounding in the cohort analysis of the Zanzibar study, using a bias indicator condition, we observed almost no difference in the two estimates. ConclusionOur findings suggest that the TND is a valid approach for evaluating OCV effectiveness in routine vaccination programs.