Numerous methods exist for jointly modeling distributions of correlated longitudinal and survival outcomes. These methods often require strong parametric assumptions and can be difficult to implement. In regulated clinical trials, investigators often prespecify either the survival or longitudinal outcome as the primary endpoint so that standard survival or longitudinal models may be used in the analysis. One potential drawback of such an approach is that investigators may not know in advance whether the survival or longitudinal outcome will be more sensitive to the treatment differences. In addition, the correlation present in the survival and longitudinal outcomes may cause bias in the analysis of either outcome separately. We propose a straightforward approach using multivariate time-to-event methods to evaluate the effect of a treatment or baseline predictor on both longitudinal and survival outcomes simultaneously. This method mainly addresses situations in which there exists attrition or non-reversible deterioration in the longitudinal process subject to censoring due to the 1 survival endpoint. We define cutpoints of interest in the longitudinal outcome and time-to-event endpoints based on time to reach a given cutpoint or the survival event, whichever comes first. We then use multivariate time-to-event methods on the resulting endpoints to evaluate the effect of the treatment or baseline predictor. We conduct simulation studies to compare the proposed strategy to existing approaches. The method is particularly attractive in clinical trial settings in which the primary analysis must be specified a priority. We illustrate the method on data from a study of chronic lung disease.


Clinical Trials | Longitudinal Data Analysis and Time Series | Multivariate Analysis | Survival Analysis