In some studies of chronic viral infections where the objective is to estimate the distributions of time until viral eradication and viral resistance to treatment, patients must have treatment terminated in order to assess eradication status. Such patients then have their viral load continually monitored during a post-treatment period. If no virus is detected during this period, viral eradication is presumed to have occurred whereas detection of virus is interpreted to mean that the virus had been suppressed but not eradicated prior to treatment interruption. If the post-treatment period is long, as would be the case with diseases such as hepatitis C and HIV, there will be patients who have not completed the post-treatment period by the time the data are analysed. This paper proposes non-parametric and semi-parametric methods to incorporate partial post-treatment data in the estimation of the subdistributions of the time until eradication and resistance. The new methods extend previous methods for the analysis of eradication studies that do not account for incomplete post-treatment information, and are illustrated with data from a recent hepatitis C clinical trial.
Copyright 2003 John Wiley & Sons, Ltd.