Effectiveness of maternal immunisation with a three-component acellular pertussis vaccine at preventing pertussis in infants in the United States: Post-hoc analysis of a case-control study using Bayesian dynamic borrowing

Immunisation during pregnancy with a tetanus toxoid, reduced diphtheria toxoid and acellular pertussis (Tdap) vaccine can protect infants against pertussis between birth and paediatric vaccination. We aimed to estimate the vaccine effectiveness (VE) of third-trimester pregnancy immunisation with the three-component acellular pertussis (Td3ap) vaccine at preventing pertussis in infants <2 months in the United States (US), to support a label update. We performed a post-hoc sub-analysis of a case-control study conducted in six US Emerging Infections Program Network states between 2011 and 2014. Our analysis included only cases and controls whose mothers were either vaccinated with Td3ap or did not receive any Tdap vaccine. The association between Td3ap maternal immunisation and pertussis in infants was assessed for US data using a frequentist method with conditional logistic regression. A robustified analysis was conducted using Bayesian dynamic borrowing of non-US data, considering a mixing-weighted prior of 90% for historical non-US VE data, and of 10% for a vague prior. VE was estimated as (1-odds ratio) × 100%. Sensitivity analyses accounting for the impact of each non-US study, different mixing weights and missing/ambiguous data were performed. We included 108 cases and 183 controls. Based on US data, the estimated VE of third-trimester maternal immunisation with Td3ap at preventing pertussis in infants <2 months was 78.0% (95% confidence interval: −38.0; 96.5). VE estimated by Bayesian dynamic borrowing of non-US data (with a 90% weight for historical data) was 83.4% (95% credible interval: 55.7; 92.5); sensitivity analyses produced similar VE estimates. Effectiveness of third-trimester pregnancy immunisation with Td3ap at preventing infant pertussis in the US is very likely to be ≥ 50% and is most likely ∼ 80%. Bayesian dynamic borrowing of non-US VE data allowed overcoming the limited power (due to small sample size) of a brand-specific sub-analysis by considering additional evidence.