Series 2-19-2009 Multiple Imputation Methods for Treatment Noncompliance and Nonresponse in Randomized Clinical Trials

L. Taylor,X. H. Zhou

2017-01-01

Abstract:This working paper is hosted by The Berkeley Electronic Press (bepress) and may not be commercially reproduced without the permission of the copyright holder. Summary. Randomized clinical trials are a powerful tool for investigating causal treatment effects, but in human trials there are oftentimes problems of noncompliance which standard analyses, such as the intention-to-treat or as-treated analysis, either ignore or incorporate in such a way that the resulting estimand is no longer a causal effect. One alternative to these analyses is the complier average causal effect (CACE) which estimates the average causal treatment effect among a subpopulation that would comply under any treatment assigned. We focus on the setting of a randomized clinical trial with crossover treatment noncompliance (e.g., control subjects could receive the intervention and intervention subjects could receive the control) and outcome nonresponse. In this article, we develop estimators for the CACE using multiple imputation methods, which have been successfully applied to a wide variety of missing data problems, but have not yet been applied to the potential outcomes setting of causal inference. Using simulated data we investigate the finite sample properties of these estimators as well as of competing procedures in a simple setting. Finally we illustrate our methods using a real randomized encouragement design study on the effectiveness of the influenza vaccine. 1. Introduction The focus of empirical studies in medicine is often to estimate the causal effect of treatments, where randomized clinical trials are considered the most acceptable tool for investigating these causal relationships. But in trials involving human subjects there are oftentimes problems of patient noncompliance, where the patient does not adhere to the treatment assigned. In addition, there is also the problem of nonignorable missing data, where the missing-data mechanism may depend on unobserved data. Standard methods and analyses either ignore these complications, which can lead to biased estimates of causal treatment effect, or account for them in such a way that the resulting estimand can no longer be considered a causal effect of treatment. A statistical framework for causal inference that deals with the issue of noncompliance is based on potential outcomes and was first introduced by Neyman (1923) for randomized studies and later developed by Rubin (1974, 1978) for nonrandomized studies and other forms of inference. Rubin's approach, sometimes referred to as the Ru-bin Causal Model (Holland, 1986), provides a framework for defining the parameters of interest and correctly attributing the …