PARAMETRIC APPROACHES TO QUALITY-ADJUSTED SURVIVAL ANALYSIS

We present a parametric methodology for performing quality-of-life-adjusted survival analysis using multivariate censored survival data. It represents a generalization of the nonparametric Q-TWIST method (Quality-adjusted Time without Symptoms and Toxicity). The event times correspond to transitions between states of health that differ in terms of quality of life. Each transition is governed by a competing risks model where the health states are the competing risks. Overall survival is the sum of the amount of time spent in each health state. The first step of the proposed methodology consists of defining a quality function that assigns a ''score'' to a life having given health state transitions. It is a composite measure of both quantity and quality of life. In general, the quality function assigns a small value to a short life with poor quality and a high value to a long life with good quality. In the second step, parametric survival models are fit to the data. This is done by repeatedly modeling the conditional cause-specific hazard functions given the previous transitions. Covariates are incorporated by accelerated failure time regression, and the model parameters are estimated by maximum likelihood. Lastly, the modeling results are used to estimate the expectation of quality functions. Standard errors and confidence intervals are computed using the bootstrap and delta methods. The results are useful for simultaneously evaluating treatments in terms of quantity and quality of life. To demonstrate the proposed methods, we perform an analysis of data from the international Breast Cancer Study Group Trial V, which compared short-duration chemotherapy versus long-duration chemotherapy in the treatment of node-positive breast cancer. The events studied are: (1) the end of treatment toxicity, (2) disease recurrence, and (3) overall survival.