Practical Properties of Some Structural Mean Analyses of the Effect of Compliance in Randomized Trials Krista Fischer-Lapp, PhD, and Els Goetghebeur, PhD Institute of Mathematical Statistics, University of Tartu, Tartu, Estonia (K. F.-L.) and Applied Mathematics and Informatics, University of Ghent, Ghent, Belgium (E.G.) ABSTRACT: We can use the structural mean model (SMM) to estimate the mean effect of dose- timing patterns of active treatment actually taken by patients in a randomized placebo- controlled trial. An SMM therefore models the expected difference between a patient’s potential response on the treatment arm and potential response on the placebo arm as a function of observed compliance on the treatment arm and baseline predictors. It accounts for the possibly selective nature of noncompliance without needing to model that aspect directly. It nevertheless enjoys the intention-to-treat property of protecting the  level when we are testing the hypothesis of no treatment effect. In the presence of selective compliance, classical regression methods lead to inconsis- tent and seriously biased estimates of the effects of treatment actually taken. The SMM is designed to reduce these problems. This paper studies selectivity and addresses some practical properties of the SMM estimator. Specifically, we use a blood pressure trial to explore the precision of the estimates in practical cases. We also compare mean squared errors (MSEs) of an SMM and the ordinary least-squares (OLS) estimator. We study the effect of baseline covariates on the precision of the SMM estimator and describe the potential role of a run-in period in this regard. Control Clin Trials 1999;20:531– 546  Elsevier Science Inc. 1999 KEY WORDS: Causal inference, estimating equations, noncompliance, placebo-controlled clinical trial, structural nested mean models 1. INTRODUCTION An intention-to-treat analysis, the standard approach for analyzing a ran- domized placebo-controlled clinical trial, estimates the effect of assigning pa- tients to the active treatment instead of placebo. This effect estimate is unbiased, assuming there are no missing data in the observed outcome variables and valid randomization. Given compliance data in such a trial, one may also want to estimate the expected effect of actually observed dose-timing patterns of active treatment. The difficulty is that noncompliance is a postrandomization Address reprint requests to: Krista Fischer-Lapp, PhD, Liivi 2, 50409 Tartu, Estonia; E-mail: kristal@ut.ee. Received May 22, 1998; accepted April 6, 1999. Controlled Clinical Trials 20:531–546 (1999)  Elsevier Science Inc. 1999 0197-2456/99/$–see front matter 655 Avenue of the Americas, New York, NY 10010 PII S0197-2456(99)00027-6