Analysis and design issues for studies using censored biomarker measurements with an example of viral load measurements in HIV clinical trials.

For many biomarkers, the range (L,R) over which they can be quantified is restricted by technical limitations, leading to some measurements that are left or right censored. However, despite the widespread availability of statistical methods for the analysis of censored data, many studies use an imputed value for censored measurements (for example, replacing a value <L by L, or by L/2). Commonly, an analysis that ignores such imputation is then used. In clinical trials, this leads to bias and a loss of power in evaluating treatment effects. In this paper, a review of appropriate statistical methods for parametric and non-parametric analysis of such measurements is presented. This includes methods for situations in which baseline measurements are available. New results concerning design issues such as sample size determination are also presented. The paper is illustrated using two examples of studies that included censored measurements of viral load in HIV-infected subjects.