Semiparametric inference for relative heterogeneous vaccine efficacy between strains in observational case-only studies
The aim of this manuscript is to explore semiparametric methods for inferring subgroup-specific relative vaccine efficacy in a partially vaccinated population against multiple strains of a virus. We consider methods for observational case-only studies with informative missingness in viral strain type due to vaccination status, pre-vaccination variables, and also post-vaccination factors such as viral load. We establish general causal conditions under which the relative conditional vaccine efficacy between strains can be identified nonparametrically from the observed data-generating distribution. Assuming that the relative strain-specific conditional vaccine efficacy has a known parametric form, we propose semiparametric asymptotically linear estimators of the parameters based on targeted (debiased) machine learning estimators for partially linear logistic regression models. Finally, we apply our methods to estimate the relative strain-specific conditional vaccine efficacy in the ENSEMBLE COVID-19 vaccine trial.
READ FULL TEXT