Applying sample survey methods to clinical trials data

This paper outlines the utility of statistical methods for sample surveys i n analysing clinical trials data. Sample survey statisticians face a variet y of complex data analysis issues deriving from the use of multistage proba bility sampling from finite populations. One such issue is that of clusteri ng of observations at the various stages of sampling. Survey data analysis approaches developed to accommodate clustering in the sample design have mo re general application to clinical studies in which repeated measures struc tures are encountered. Situations where these methods are of interest inclu de multi-visit studies where responses are observed at two or more time poi nts for each patient, multi-period cross-over studies, and epidemiological studies for repeated occurrences of adverse events or illnesses. We describ e statistical procedures for fitting multiple regression models to sample s urvey data that are more effective for repeated measures studies with compl icated data structures than the more traditional approaches of multivariate repeated measures analysis. In this setting, one can specify a primary sam pling unit within which repeated measures have intraclass correlation. This intraclass correlation is taken into account by sample survey regression m ethods through robust estimates of the standard errors of the regression co efficients. Regression estimates are obtained from model fitting estimation equations which ignore the correlation structure of the data (that is, com puting procedures which assume that all observational units are independent or are from simple random samples). The analytic approach is straightforwa rd to apply with logistic models for dichotomous data, proportional odds mo dels for ordinal data, and linear models for continuously scaled data, and results are interpretable in terms of population average parameters. Throug h the features summarized here, the sample survey regression methods have m any similarities to the broader family of methods based on generalized esti mating equations (GEE). Sample survey methods for the analysis of time-to-e vent data have more recently been developed and implemented in the context of finite probability sampling. Given the importance of survival endpoints in late phase studies for drug development, these methods have clear utilit y in the area of clinical trials data analysis. A brief overview of methods for sample survey data analysis is first provided, followed by motivation for applying these methods to clinical trials data. Examples drawn from thr ee clinical studies are provided to illustrate survey methods for logistic regression, proportional odds regression and proportional hazards regressio n. Potential problems with the proposed methods and ways of addressing them are discussed. Copyright (C) 2001 John Wiley & Sons, Ltd.