Background: Some high-risk prostate cancer (PCa) patients may show more favorable Gleason pattern at radical prostatectomy (RP) than at biopsy. Objective: To test whether downgrading could be predicted accurately. Design, setting, and participants: Within the Surveillance, Epidemiology and End Results database (2010–2016), 6690 National Comprehensive Cancer Network (NCCN) high-risk PCa patients were identified. Outcome measurements and statistical analyses: We randomly split the overall cohort between development and validation cohorts (both n = 3345, 50%). Multivariable logistic regression models used biopsy Gleason, prostate-specific antigen, number of positive prostate biopsy cores, and cT stage to predict downgrading. Accuracy, calibration, and decision curve analysis (DCA) tested the model in the external validation cohort. Results and limitations: Of 6690 patients, 50.3% were downgraded at RP, and of 2315 patients with any biopsy pattern 5, 44.1% were downgraded to RP Gleason pattern ≤4 + 4. Downgrading rates were highest in biopsy Gleason pattern 5 + 5 (84.1%) and lowest in 3 + 4 (4.0%). In the validation cohort, the logistic regression model–derived nomogram predicted downgrading with 71.0% accuracy, with marginal departures (±3.3%) from ideal predictions in calibration. In DCA, a net benefit throughout all threshold probabilities was recorded, relative to treat-all or treat-none strategies and an algorithm based on an average downgrading rate of 50.3%. All steps were repeated in the subgroup with any biopsy Gleason pattern 5, to predict RP Gleason pattern ≤4 + 4. Here, a second nomogram (n = 2315) yielded 68.0% accuracy, maximal departures from ideal prediction of ±5.7%, and virtually the same DCA pattern as the main nomogram. Conclusions: Downgrading affects half of all high-risk PCa patients. Its presence may be predicted accurately and may help with better treatment planning. Patient summary: Downgrading occurs in every second high-risk prostate cancer patients. The nomograms developed by us can predict these probabilities accurately.
File in questo prodotto:
Non ci sono file associati a questo prodotto.