Development of Internet-Accessible Prediction Models for Prostate Cancer Diagnosis, Treatment, and Follow-Up.

摘要

The objective of the development of Internet-accessible prediction models is to enhance the diagnosis accuracy, treatment efficacy and prognosis for patients with carcinoma of prostate cancer (Cap). An Oracle database was created, and Internet-accessible data collection applications were developed. Program packages for daily data retrieval, standardization, and reorganization were built. The roles of variables (race/ethnicity, diagnostic age, labs - and treatment types) on the outcome of CaP patients were analyzed. The results show that CaP patients who chose watchful waiting tend to be older with lower serum PSA and lower Gleason score. The age at diagnosis, diagnostic PSA and clinical T-stage are the most significant predictors of secondary treatment in watchful waiting (submitted to J Urol). Pre-treatment testosterone level is a predictor of PSA recurrence (accepted for publication in J Urol). Post-treatment PSA doubling time less than 3 months is a surrogate for prostate cancer specific mortality following surgery or radiation therapy (submitted to J Urol). Biostatistical models with variables of race, pre-treatment PSA, clinical stage, pathological stage and Gleason sum for predicting PSA recurrence before and after radical prostatectomy were implemented on CPDR webpage (www.cpdr.org). Further data analysis and the development of the prediction models are in progress.