針對新的台灣攝護腺癌症風險評估做外部驗證

李昀叡、林仁泰、余家政、陳逸軒

高雄榮民總醫院泌尿外科

External validation of a new Taiwanese prostate cancer risk calculator within another medical centre

Yun-Jui Li, Jen-Tai Lin, Chia-Cheng Yu, I-Hsuan Chen

Division of Urology, Department of Surgery, Kaohsiung Veterans General Hospital, Kaohsiung, Taiwan

Purpose:

To validate a novel Taiwanese prostate cancer (PCa) risk calculator utilizing another contemporary cohort, comparing its predictive performance with updating methods.

Materials and Methods:

1309 and 875 men underwent prostate biopsies at two Taiwanese medical centers, respectively the Kaohsiung Veterans General Hospital (KSVGH) and Chang-Gung Memorial Hospital (CGMH) between 2012 and 2019. A novel Taiwanese PCa risk calculator was established using the 1309 patients from the KSVGH. Another cohort of 875 patients was introduced to assess its predictivity. Updating methods, including intercept and logistic recalibration, model refitting and revision, were calculated to compare their predictive probabilities of PCa and high-grade PCa (Gleason score ≧ 7) with adjustment of the cut point on the receiver operator characteristic (ROC) curves. Discrimination was analyzed using the area under the ROC curve (AUC). Calibration was graphically evaluated with the goodness-of-fit test. Decision-curve analysis was performed for clinical utility. At different risk thresholds to biopsy, the proportion of biopsies saved versus low- and high-grade PCa missed were presented.

Results:

Overall, 278/1309 (21.2%) and 305/875 (34.9%) patients were diagnosed with PCa, respectively at the KSVGH and CGMH; 181 out of 278 (65.1%) and 238 of 305 (78.0%) patients had high-grade PCa. Without updating, the area under the receiver-operating characteristics curves (AUCs) were 0.787 [95% confidence interval (CI): 0.756-0.818] and 0.83 [95% CI: 0.802-0.858] for predicting PCa and high-grade PCa, respectively at the CGMH. After comparing different updating methods to evaluate the predictive performance, adjustment of the cut point on the ROC curve demonstrated similar sensitivity, specificity and accuracy.

Conclusion:

Our new Taiwanese PCa risk calculator could be utilized in a different contemporary cohort. Instead of recalibrating, refitting or revising the model, adjustment of the cut-off point might be a simpler statistical technique to improve its predictability.