Recent Development in the Joint Modeling of Longitudinal Quality of Life Measurements and Survival Data from Cancer Clinical Trials

Abstract

In cancer clinical trials, longitudinal Quality of Life (QoL) measurements and survival time on a patient may be analyzed by joint models, which provide more efficient estimation than modeling QoL data and survival time separately, especially when there is strong association between the longitudinal measurements and the survival time. Most joint models in the literature assumed classical linear mixed model for longitudinal measurements and Cox’s proportional hazards model for survival times. The linear mixed model with normal distributed random components may not be sufficient to model bounded QoL measurements. Moreover, when some patients are immune to recurrence of relapse and can be viewed as cured, the proportional hazards model is not suitable for survival times. In this paper, we review some recent developments in joint models to deal with bounded longitudinal QoL measurements and survival times with a possible cure fraction. One of such joint models assumes a linear mixed tt model for longitudinal measurements and a promotion time cure model for survival data, and the two parts are linked through a latent variable. Another joint model employs a simplex distribution to model the bounded QoL measurements and a classical proportional hazard to model survival times, and the two parts share a random effect. Semiparametric estimation methods have been proposed to estimate the parameters in the models. The models are illustrated with QoL measurements and recurrence times from a clinical trial on women with early breast cancer.