In medical and epidemiological research, a common technique is to categorize continuous variables before evaluating its prognostic impact on the clinical trial outcome of interest. If well established cut points are not available, then statistical techniques are used to determine the cutoff point. This study was related to a clinical trials designed to test the effectiveness of a new treatment compared to a currently used treatment for prostate cancer. Research question was to find a dichotomous latent factor, labeled as asymptomatic and symptomatic, based on the patients’ observed PSA level and pain score. And, check whether the latent factor influenced the effectiveness of the new treatment. Different approaches have been considered to estimate the threshold for PSA- (i) maximum differences in treatment effect between symptomatic and asymptomatic patients, and (ii) maximum value for goodness of fit index (likelihood, c-index, concordance probability estimate). Predictive ability of these techniques have been checked through cross validation, while the first approach to estimate the threshold point gave over estimated results. Whereas, the threshold estimation based on goodness of fit measures provides consistent result in cross validation. Along with the classical estimation technique, Bayesian change point model has been used to estimate the PSA threshold. Cox (1972) model was used to estimate the effect of the covariates on patient’s survival time, and the effectiveness of the new treatment among symptomatic and asymptomatic patients. The result showed similar treatment effect among the symptomatic and asymptomatic patients.