Keywords: transportability, generalizability, causal inference, clinical trials
A clinical trial often selects participants on the basis of covariates that are also determinants of the outcome and, on some scale, modifiers of the treatment effect. When that is the case, the average causal effect in the trial is not the same as the average causal effect in other populations in which the interventions may be applied. We consider methods for transporting the results of a trial to a target population for which only baseline covariate information is available. We show how a composite dataset, formed by appending a sample from the target population to the trial data, can be used to estimate the average treatment effect in the target population. We propose estimators based on modeling the expectation of the outcome, the probability of participation, or both (doubly robust). We discuss double robustness when the conditional outcome mean is modeled using linear exponential family quasi-likelihood functions and study the finite-sample performance of different estimators in simulations. Finally, we apply the methods using trial data from a study of antihypertensive treatments for chronic kidney disease and target population data from administrative healthcare databases.