Validation of a prediction model for low volume/low grade cancer: application in selecting patients for active surveillance

PURPOSE: We previously demonstrated that assessment of the number of positive cores, tumor length in a core, Gleason score and prostate volume significantly enhanced the accuracy of a prediction model for low volume/low grade cancer in men who had undergone extended biopsy. To determine the validity of the model, we applied it to an independent population of men with prostate cancer. MATERIALS AND METHODS: The study group included 170 men who had undergone radical prostatectomy without neoadjuvant therapy. In all cases, prostate cancer was diagnosed on only 1 positive core of a 10-core extended biopsy. We assessed the accuracy of the model, which consists of tumor length less than 2 mm, Gleason score 3+4 or less and prostate gland volume greater than 50 cc in predicting the occurrence of low volume/low grade cancer (defined as tumor volume less than 0.5 cc, no Gleason grade 4 or 5 disease, and organ confined disease). RESULTS: Of the patients 101 (59.4%) had low volume/low grade cancer. Our model using all 3 previously mentioned variables had the highest performance, demonstrating a positive predictive value of 70.4% (88 of 125), a negative predictive value of 71.1% (32 of 45) and a diagnostic accuracy of 70.6% (120 of 170). This model performed better than a model based on tumor length only (positive predictive value, negative predictive value and diagnostic accuracy 68.1%, 57.9% and 64.7%, respectively) or a model based on tumor length and Gleason score (positive predictive value, negative predictive value and diagnostic accuracy 70.0%, 60.0% and 66.5%, respectively). CONCLUSIONS: This study validates that our model with a combination of tumor length, Gleason score and prostate volume is predictive for low volume/low grade cancer in an independent population of men who demonstrated only 1 positive core in an extended biopsy. This model can be used as a tool for selecting men for active surveillance.