Evaluation of the Time-Varying Effect of Prognostic Factors on Survival in Ovarian Cancer.

Abstract

To explore the risk factors in ovarian cancer with respects of time-varying effects on recurrence and survival. Two hundred and ninety-eight patients with epithelial ovarian cancer in the Kaohsiung Veterans' General Hospital from January 1995 to the end of 2011 were included in the study. The assumption of the Cox proportional hazard model, i.e., the hazard ratio is a constant with time, was tested for available prognostic factors. An extended Cox model was then applied, and a statistical package was constructed to perform multivariate analysis in presence of both time-varying and time-independent factors. Most prognostic factors met the assumption of the Cox proportional hazard model (p > 0.05) except for cancer-associated antigen (CA) 125 nadir concentration during first-line chemotherapy (p = 0.02). Multivariate analysis, where CA125 nadir was allowed to change with time while other factors remained constant, showed that International Federation of Gynecology and Obstetrics (FIGO) stage, residual tumor, CA125 nadir, and age were independent risk factors for recurrence and death. The effect of CA125 nadir on recurrence and overall survival is not constant over time. It loses predictivity on recurrence and survival after 4.5 years. Awareness of the time-varying effects of the prognostic factors is beneficial to gynecologists in patient consultation and case evaluation.