We examine the structural bias for established estimators of vaccine effects on susceptibility and for newer estimates of vaccine effects on infectiousness. We then propose and analyse new bias corrections for vaccine effect estimators of both susceptibility and infectiousness, as well as their combined effect on infection transmission. Each estimator is evaluated empirically with computer simulations. Of the estimators examined in this paper, those with the least bias and root mean squared error are computed by adding one to the positive count in the placebo population. We also identify a source of bias for a standard Bayesian estimator of risk ratios.
Copyright 2001 John Wiley & Sons, Ltd.