Background: Following treatment for localized prostate cancer, men are monitored with serial PSA measurements. Refining the predictive value of post-treatment PSA determinations may add to clinical management and we have developed a model that predicts for an individual patient future PSA values and estimates the time to future clinical recurrence.

Methods: Data from 934 patients treated for prostate cancer between 1987 and 2000 were used to develop a comprehensive statistical model to fit the clinical recurrence events and pattern of PSA data. A logistic regression model was used for the probability of cure, non-linear hierarchical mixed models were used for serial PSA measurements and a time-dependent proportional hazards model was used for recurrences. Data available up to February 2001 and September 2003 was used to assess the performance of the model.

Results: The model suggests that T-stage, baseline PSA, and radiotherapy dosage are all associated with probability of cure. The risk of clinical recurrence in those not cured by radiotherapy is most strongly affected by the slope of the long-transformed PSA values. We show how the model can be used for individual monitoring of a patient’s disease progression. For each patient the model predicts, based upon his baseline and all post-treatment PSA values, the probability of future clinical recurrence in the validation dataset and of 406 PSA measurements obtained 1-2 years after February 2001, 92.8% were within 95% prediction limits from the model.

Conclusions: This statistical model presented accurately predicts future PSA values and risk of clinical relapse. This predictive information for each individual patient, which can be updated with each additional PSA value, may prove useful to patents and physicians in determining what post-treatment salvage should be employed.


Disease Modeling | Epidemiology | Longitudinal Data Analysis and Time Series | Survival Analysis