Evaluation of Prostate Cancer Risk Calculators for Shared Decision Making Across Diverse Urology Practices in...

Area of Science:

• Urology
• Oncology
• Biostatistics

Background:

• The Prostate Cancer Prevention Trial (PCPT) Risk Calculator is widely used for predicting prostate biopsy outcomes.
• Contemporary urology practice data may differ from the PCPT cohort, potentially affecting calculator accuracy.

Purpose of the Study:

• To develop and validate a new logistic regression model for predicting prostate biopsy outcomes using contemporary data.
• To compare the predictive performance of this new model against the PCPT Risk Calculator version 2.0.

Main Methods:

• A multinomial logistic regression model was developed using data from 11,809 first-time prostate biopsies (2012-2015) in the Michigan Urological Surgery Improvement Collaborative (MUSIC).
• The MUSIC model predicted the likelihood of high-grade cancer (Gleason score ≥7), low-grade cancer (≤6), or no cancer.
• Model performance was assessed using 10-fold cross-validation, comparing discrimination (multiclass AUC) and calibration with the PCPT Risk Calculator.

Main Results:

• The MUSIC model and PCPT calculator showed comparable discrimination (AUC 0.63 vs. 0.62).
• Calibration analysis revealed the MUSIC model more accurately predicted observed outcomes.
• The PCPT calculator overestimated no cancer likelihood and underestimated high-grade cancer risk in this cohort.

Conclusions:

• The PCPT Risk Calculator may not accurately predict individual prostate biopsy outcomes in contemporary urology settings.
• A new model developed with current data demonstrates superior calibration for predicting biopsy results.