Skip to main content

20-12-2009 | Oncology | Article

CAPRA score and Stephenson nomogram superior for predicting biochemical recurrence


Free abstract

MedWire News: Both the Cancer of the Prostate Risk Assessment (CAPRA) score and Stephenson nomogram outperform the D’Amico risk stratification for predicting biochemical recurrence (BCR) after radical prostatectomy, say researchers.

BCR affects approximately 30% of patients following radical prostatectomy for localized prostate cancer, explain Pierre Karakiewicz, from the University of Montreal Health Center in Quebec, Canada, and colleagues.

Because the probability of BCR varies widely depending on baseline risk factors, researchers have found it most helpful to calculate individual risk.

Currently, three models are designed to generate individual probabilities of BCR, the D’Amico risk stratification scheme, the CAPRA score, and the Stephenson nomogram.

“To date, no direct comparison has been made among the three tools,” the researchers report in the journal European Urology.

They tested the accuracy (ie, Harrell’s concordance index) and calibration between predicted and observed BCR rates using data derived from 1976 patients treated with open radical prostatectomy for localized prostate cancer.

Analyses were conducted at 3 and 5 years after the procedure, and decision curves were used to evaluate the D’Amico, CAPRA, and Stephenson models.

The BCR rates at 3 and 5 years after radical prostatectomy were 80.2% and 72.6%, respectively.

Both CAPRA and Stephenson were more accurate at predicting 3-year BCR rates than the D’Amico model, with a concordance index of 74.3% and 75.2% versus 70.4%, respectively.

The results were similar for predicting 5-year BCR rates, with a concordance index of 72.9% and 73.5% versus 67.4%, respectively.

Karakiewicz et al also report that the D’Amico risk stratification model had a limited ability to distinguish BCR rates in high-risk patients, while both the CAPRA and Stephenson models were able to discriminate BCR among a wider range of patients.

The Stephenson model therefore appeared to outperform the other models when concordance index was used, while calibration results supported the use of either the CAPRA or Stephenson models. But the results also showed a small benefit for the CAPRA score relative to the Stephenson nomogram on decision curve analyses.

The researchers note, however, that their “findings apply to a specific population of German patients treated at a high-volume tertiary center.”

They therefore stress that “the rank order reported in this paper may not be the same in North American or other European cohorts.”

MedWire ( is an independent clinical news service provided by Current Medicine Group, a trading division of Springer Healthcare Limited. © Springer Healthcare Ltd; 2009

By Liz Scherer