Abstract
BACKGROUND: There are some limitations in the commonly used methods for the detection of prostate cancer. There is a lack of nomograms based on multiparametric magnetic resonance imaging (mpMRI) and (68)Ga-prostate-specific membrane antigen (PSMA) positron emission tomography-computed tomography (PET-CT) for the prediction of prostate cancer. The study seeks to compare the performance of mpMRI and (68)Ga-PSMA PET-CT, and design a novel predictive model capable of predicting clinically significant prostate cancer (csPCa) before biopsy based on a combination of (68)Ga-PSMA PET-CT, mpMRI, and patient clinical parameters. METHODS: From September 2020 to June 2021, we prospectively enrolled 112 consecutive patients with no prior history of prostate cancer who underwent both (68)Ga-PSMA PET-CT and mpMRI prior to biopsy at our clinical center. Univariate and multivariate regression analyses were used to identify predictors of csPCa, with a predictive model and its nomogram incorporating (68)Ga-PSMA PET-CT, mpMRI, and the clinical predictors then being generated. The constructed model was evaluated using receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis, and further validated with the internal and external cohorts. RESULTS: The model incorporated prostate-specific antigen density (PSAd), Prostate Imaging Reporting and Data System (PI-RADS) category, and maximum standardized uptake value (SUVmax), and it exhibited excellent predictive efficacy when applying to evaluate both training and validation cohorts [area under the curve (AUC): 0.936 and 0.940, respectively]. Compared with SUVmax alone, the model demonstrated excellent diagnostic performance with improved specificity (0.910, 95% CI: 0.824-0.963) and positive predictive values (0.811, 95% CI: 0.648-0.920). Calibration curve and decision curve analysis further confirmed that the model exhibited a high degree of clinical net benefit and low error rate. CONCLUSIONS: The constructed model in this study was capable of accurately predicting csPCa prior to biopsy with excellent discriminative ability. As such, this model has the potential to be an effective non-invasive approach for the diagnosis of csPCa.