External Validation of the Updated Partin Tables in a Cohort of North American Men

Purpose: The Partin tables were updated in 2007. However, to our knowledge their accuracy and performance characteristics have not been confirmed in an external validation cohort. Materials and Methods: We examined the discrimination and calibration properties of the 2007 Partin tables in 1,838 men treated with radical prostatectomy between 2001 and 2005 at Cleveland Clinic Foundation. The ROC derived AUC and the Brier score were used to quantify the discriminant properties of the predictions of the 2007 Partin tables for extraprostatic extension, seminal vesical invasion and lymph node invasion. Loess based calibration plots were used to examine the relationship between the predicted and observed rates of extraprostatic extension, seminal vesical invasion and lymph node invasion. Results: The rates of extraprostatic extension, seminal vesical invasion and lymph node invasion were 26.9%, 5.5% and 1.8%. The accuracy of extraprostatic extension, seminal vesical invasion and lymph node invasion prediction was 71%, 80% and 75% according to the AUC method, and 0.176, 0.051 and 0.037 according to the Brier score, respectively. Extraprostatic extension predictions between 0% and 25%, and lymph node invasion predictions between 0% and 5% correlated well with observed extraprostatic extension and lymph node invasion rates, respectively. Conversely a suboptimal correlation was recorded between predicted and observed seminal vesical invasion rates as well as between predicted and observed rates of extraprostatic extension and lymph node invasion for predicted extraprostatic extension and lymph node invasion values above 25% and 5%, respectively. Conclusions: In this examined validation cohort the overall accuracy (AUC) of the Partin tables was comparable to results reported in the original 2007 development cohort. However, performance characteristics indicate that predictions within specific probability ranges should be interpreted with caution.