Reporting and Interpreting Decision Curve Analysis: A Guide for Investigators

Area of Science:

• Urology
• Clinical Decision Making
• Health Services Research

Background:

• Clinical risk prediction models are crucial for urological decision-making.
• Traditional performance measures do not fully assess model utility for clinical decisions.
• Decision curve analysis (DCA) offers a method to evaluate prediction model utility.

Purpose of the Study:

• To provide recommendations for interpreting and reporting DCA in prediction model evaluations.
• To assess the understanding and application of DCA in urological literature.
• To illustrate DCA with a case study on high-grade prostate cancer prediction.

Main Methods:

• Informal review of urological literature to gauge understanding of DCA.
• Development of risk models for high-grade prostate cancer using patient data (n=31316).
• Explanation of risk thresholds, net benefit (NB), and test tradeoff within DCA.

Main Results:

• DCA can identify superior prediction models across various risk thresholds.
• In the prostate cancer case study, adding transrectal ultrasound (TRUS) predictors improved NB by 0.0114.
• The tradeoff analysis indicated that adding TRUS is worthwhile under specific patient-to-benefit ratios.

Conclusions:

• DCA is a valuable tool for assessing the clinical utility of risk prediction models.
• Proposed guidelines can enhance the understanding, application, and reporting of DCA.
• Appropriate interpretation and reporting of DCA can lead to better clinical decisions.