Effect of covariate measurement error in randomized clinical trials

Prognostic factors often appear as covariates in clinical trials designed to estimate treatment effects. These factors are usually subject to measurement error and lead to decreased efficiency in the estimation of the treatment effect. We use Weibull regression models and asymptotic theory to derive the efficiency of estimation obtained by adjusting for one dichotomous and one continuous covariate measured with error. An application to a clinical trial involving advanced lung cancer patients illustrates the results.