Urine spermine and multiparametric magnetic resonance imaging for prediction of prostate cancer in Japanese men

Abstract

Objectives: To investigate the role of urine spermine and spermine risk score in predicting prostate cancer (PCa) diagnoses in combination with multiparametric magnetic resonance imaging (mpMRI).

Methods: Three hundred forty seven consecutive men with elevated prostate-specific antigen (PSA) with mpMRI examination were prospectively enrolled in this study. In 265 patients with PSA levels between 4 and20 ng/ml, pre-biopsy urine samples were analyzed for spermine levels with ultra-high performance liquid chromatography (UPLC-MS/MS). Transperineal image-guided prostate biopsies with 16-18 cores were performed. Logistic regressions were used to form different models for the prediction of the PCa, and the performances were compared using the area under the curve (AUC).

Results: The median serum PSA level and prostate volume were 7.4 ng/mL and 33.9 mL, respectively. PCa and high-grade PCa (ISUP group ≥2, HGPCa) were diagnosed in 66.0% (175/265) and 132/265 (49.8%) cases, respectively. The urine spermine levels were significantly lower in men with PCa (0.87 vs. 2.20, P < 0.001). Multivariate analyses showed that age, PSA, PV, urine spermine level, and Prostate Imaging Reporting and Data System (PI-RADS) findings were independent predictors for PCa. The Spermine Risk Score is a multivariable model including PSA, age, prostate volume, and urine spermine. Adding the Spermine Risk Score to PI-RADS improved the AUC from 0.73 to 0.86 in PCa and from 0.72 to 0.83 in high grade PCa (HGPCa) prediction (both P < 0.001). At 90% sensitivity for HGPCa prediction using Spermine Risk Score, 31.1% of unnecessary biopsies could be avoided. In men with equivocal MRI PI-RADS score 3, the AUC for HGPCa prediction was 0.58, 0.79, and 0.87 for PSA, PSA density, and Spermine Risk Score, respectively.

Conclusion: Urine Spermine Risk Score, including mpMRI could accurately identify men at high risk of HGPCa and reduce unnecessary prostate biopsies. Spermine Risk Score could more accurately predict HGPCa than PSA density in men with MRI showing equivocal PI-RADS 3 lesions.

Keywords: Biomarker; Multiparametric MRI; Prostate cancer; Urine spermine.

Fig. 1

Areas under the curve of the calculated probabilities of prediction of the prostate cancer.

Fig. 2

Internal validation with bootstrapping. a. any grade of prostate cancer predicted by PSA, normalized spermine, age, prostate volume, and PI-RADS score. b. high grade prostate cancer predicted by PSA, normalized spermine, age, prostate volume, and PI-RADS score.