The Run-In Period in Clinical Trials The Effect of Misclassification on Efficiency Erica Brittain, PhD Biostatistics Research Branch; Division of Epidemiology and Clinical Applications; National Heart, Lung, and Blood Institute; Federal Building, Room 2All; Bethesda MD 20892 Janet wittes, PhD Cooperative Studies Program Coordinating Center; Department of Veterans Affairs ABSTRACT: This article considers the effect of misclassification of potential participants during the run-in period preceding randomization in a clinical trial. We present a simple mathematical model of adherence that allows for misclassification. Simulations based on this model assess the impact of a run-in period on statistical power. The run-in period is most effective when there is a high proportion of poor adherers and a low rate of misclassification. In situations with either a high degree of adherence or sub- stantial misclassification, the run-in period may reduce the efficiency of the trial, particularly when the cost of recruitment is high. We also discuss the early adherence in two recently conducted trials that had no run-in periods in order to estimate the extent of misclassification and consequent effect on power that would have been observed had a run-in occurred. KEY WORDS: Adherence, compliance, run-in, misclassification INTRODUCTION A basic tenet of experimental design is that statistical analyses must con- sider each experimental unit to be a member of its assigned treatment. In a randomized clinical trial, failure of many randomized individuals to receive their assigned treatment is likely to attenuate the treatment effect and, hence, reduce the power of the trial. Because so many participants in clinical trials fail to adhere completely to their assigned therapies, strategies to assure high adherence, and hence to increase the efficiency of the study, are attractive [1,2]. For trials that investigate the effect of drugs, the so-called run-in period is one approach for enhancing adherence that has recently gained some pop- ularity [3]. Probstfield [4] provides a bibliography of trials that have used run- ins. A run-in is a period prior to randomization during which potential partic- ipants who have met all entry criteria for a clinical trial are assigned the same Address correspondence to Erica Brittain, PhD, 1630024th Avenue, North Plymouth, MN 55447. Received November22, 1989; revised July 18, 1990. Controlled ClinicalTrials 11327-338 (1990) 327 Published by ElsevierSciencePublishingCo., Inc. 1990 0197-2456/90/$0.00 655 Avenueof the Americas,New York, New York 10010