尿液代謝體在預測臨床有意義攝護腺癌的角色

陳忠信¹ , 黃祥博², 張凱雄³, 李明學⁴, 劉詩彬¹, 呂育全¹, 黃志賢⁵, 林子平⁵, 顧明軒⁵, 鍾孝仁⁵, 張延驊⁵, 廖俊厚⁶, 游智欽⁷, 鍾旭東⁸, 蔡燿州⁹, 吳佳璋¹⁰, 陳冠州¹⁰, 何承勳¹¹, 蕭培文¹², 蒲永孝¹

¹臺大醫院泌尿部; 台大醫學院²基因體暨蛋白體醫學研究所與⁴生物化學暨分子生物研究所; ³國家衛生研究院細胞及系統醫學研究所; ⁵台北榮民總醫院外科部泌尿科; ⁶輔仁大學及耕莘醫院泌尿部; ⁷台北慈濟大學外科部泌尿科; ⁸亞東醫院外科部泌尿科; ⁹台北醫學大學附設醫院泌尿部; ¹⁰台北醫學大學雙和醫院泌尿部; ¹¹台北醫學大學資訊管理研究所; ¹²中央研究院農業生物科技研究中心

The Role of Urine Metabolomics in Predicting Clinically Significant Prostate Cancer

Chung-Hsin Chen¹ , Hsiang-Po Huang², Kai-Hsiung Chang³, Ming-Shyue Lee⁴, Shih-Ping Liu¹, Yu-Chuan Lu¹, William-J Huang⁵, Tzu-Ping Lin⁵, Ming-Hsuan Ku⁵, Hsiao-Jen Chung⁵, Yen-Hwa Chang⁵, Chun-Hou Liao⁶, Chih-Chin Yu⁷, Shiu-Dong Chung⁸, Yao-Chou Tsai⁹, Chia-Chang Wu¹⁰, Kuan-Chou Chen¹⁰, Chen-Hsun Ho¹¹, Pei-Wen Hsiao¹², and Yeong-Shiau Pu¹

¹Department of Urology, National Taiwan University Hospital; Graduate Institute of Medical Genomics and Proteomics², and Biochemistry and Molecular Biology⁴, National Taiwan University College of Medicine; ³Institute of Cellular and System Medicine, National Health Research Institutes; ⁵Division of Urology, Department of Surgery, Taipei Veterans General Hospital; ⁶Department of Urology, Cardinal Tien Hospital and Fu-Jen Catholic University; ⁷Division of Urology, Department of Surgery, Taipei Tzu Chi Hospital, Taipei; ⁸Division of Urology, Department of Surgery, Far-Eastern Memorial Hospital; ⁹Department of Urology, Taipei Medical University Hospital, Taipei Medical University; ¹⁰Department of Urology, Taipei Medical University Shuang Ho Hospital; ¹¹Graduate Institute of Information Management, National Taipei University; ¹²Agricultural Biotechnology Research Center, Academia Sinica

Purpose: Accurate markers that predict potentially lethal prostate cancer (PC) before biopsy are lacking. We aimed to develop a non-invasive urine test that predicts clinically significant PC (sPC) in men with elevated risk. 

Materials and Methods: Totally 928 men from either those with an elevated risk but biopsy-naïve or confirmed PC but treatment-naïve donated spot urine samples for gas chromatography/mass spectrophotometry metabolomics profiling to construct four predictive models. Model I is to predict PC from benign. Models II, III, and GS are to predict sPC, which refers to National Comprehensive Cancer Network (NCCN) favorable intermediate risk (FIR) group and higher (Model II), unfavorable intermediate risk group and higher (Model III), and Gleason score (GS)≥7 PC (Model GS). Subjects were randomly assigned to the training or validation cohort. Logistic regression and Akaike information criterion (AIC) were used to select metabolite markers and build models. 

Results: For Models I, II, III and GS, 26, 24, 26, and 22 metabolites, respectively showed good area-under-curve (AUC) of the receiver-operating-characteristics curve analysis, ranging from 0.80 to 0.94 (training cohort, N=603). If marker panels were combined with serum prostate-specific antigen (PSA) levels, the combined models showed significantly improved AUCs [0.94 (95% confidence interval (CI) 0.92-0.96), 0.91 (95% CI 0.88-0.93), 0.89 (95% CI 0.87-0.92), and 0.84 (95% CI 0.81-0.87), respectively], except Model I. At 90% sensitivity (training cohort), 41%, 46%, 46%, and 35% unnecessary biopsies could have been avoided. These models validated successfully against the independent validation cohort (N=325). Decision curve analysis showed significant clinical net benefit at low threshold probabilities. Models II and III that predict sPC based on NCCN risk grouping were not only more robust but also more clinically relevant than Model GS that predicts GS≥7 PC. 

Conclusions: Models that combined urine metabolite markers and serum PSA levels may greatly enhance the power to predict sPC to inform biopsy and change our way of managing men with elevated PC risk. 

(Funded by Ministry of Science and Technology, Taiwan; Clinicaltrials.gov number NCT03237702)