PARAMETRIC SURVIVAL ANALYSIS OF TESTICULAR CANCER USING AN EXPONENTIAL MODEL WITH TYPE I CENSORED DATA

Abstract

Quantitative prognostic evaluation is essential for testicular cancer, yet standard analyses often overlook model assumptions. This study evaluates patient survival times using a parametric exponential model under a type I censoring structure (n = 134, from cBioPortal). Parameter estimation via Maximum Likelihood Estimation yielded 82 uncensored and 52 censored observations. The Anderson-Darling test (p = 0.067) indicated statistical adequacy for the exponential distribution while the Weibull distribution was rejected (p < 0,010), confirming the assumption of a constant hazard, estimating a Mean Time to Failure of 64,54 months with a constant hazard rate of 0.0155 per month. However, these model-based estimates face critical clinical limitations. The small sample size and the rigid assumption of a constant hazard fail to capture the dynamic, time-varying biological progression of cancer. While this parametric approach provides a simplified baseline, clinical conclusions must be drawn cautiously; future research requires larger cohorts and flexible, non-constant hazard models to ensure actual clinical generalizability.