×

Bayesian approaches to joint longitudinal and survival models accommodating both zero and nonzero cure fractions. (English) Zbl 1144.62352

Summary: The trade-offs between survival benefits and therapeutic adverse effects on quality of life (QOL) is always an important clinical issue for cancer and AIDS patients. The International Breast Cancer Study Group (IBCSG) conducted a large clinical trial, IBCSG Trial VI, to examine the duration and timing of adjuvant therapy for advanced breast cancer patients after the initial removal surgery. We present a novel joint model for longitudinal and survival data to evaluate the relationship between QOL and breast cancer progression, and also assess issues associated with different therapeutic procedures and baseline covariates.
Multidimensional longitudinal QOL measurements are modeled in a hierarchical mixed effects model to account for psychological fluctuations and measurement errors, provide estimates for time points where QOL data are not available, and to explicitly allow for direct inferences about different dependence structures in the QOL data over time and over different QOL measures (indicators). A parametric survival model is also proposed for disease-free survival (DFS) to incorporate the underlying smooth QOL trajectories and prognostic factors. This survival model is attractive and capable of accommodating both zero and nonzero cure fractions. With advances in modern medicine, a positive cure fraction is often tenable for breast cancer patients since many are completely cured after surgery, and are no longer susceptible to relapse. A Bayesian paradigm is adopted to facilitate the estimation process and ease the computational complexity.

MSC:

62P10 Applications of statistics to biology and medical sciences; meta analysis
62F15 Bayesian inference
92C50 Medical applications (general)
62N02 Estimation in survival analysis and censored data
62J12 Generalized linear models (logistic models)
PDFBibTeX XMLCite