A nomogram for predicting extraprostatic extension in prostate cancer based on extraprostatic extension grade and clinical characteristics

Abstract

To assess the efficacy of a nomogram model derived from extraprostatic extension (EPE) grade on magnetic resonance imaging (MRI) and clinical features in forecasting pathological EPE in prostate cancer. We conducted a retrospective analysis of the clinical data from 232 prostate cancer patients. Patients were categorized into EPE and non-EPE groups based on the presence of pathological EPE. Subsequently, they were randomly allocated into a training set (162 cases) and a validation set (70 cases) at a 7:3 ratio. We gathered clinical attributes and EPE grades for all patients. Three predictive models—clinic, magnetic resonance (MR) and clinic + MR—were developed within the training set. The clinic + MR model was visualized through a nomogram. The models’ performance was assessed using the receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA). Both univariate and multivariate logistic regression analyses identified the biopsy International Society of Urological Pathology (ISUP) category and biopsy maximum unilateral positive percentage as independent risk factors for EPE within the training set. The EPE grade exhibited consistent inter-observer agreement, evidenced by weighted Kappa values of 0.72 and 0.71 in the training and validation sets, respectively. Compared to the clinic and MR models, the clinic + MR model was the most effective in predicting pathological EPE, boasting area under the curves (AUCs) of 0.85 and 0.82 in the training and validation sets, respectively. Calibration curves from both sets demonstrated that the clinic + MR model provided accurate predictions for pathological EPE. Within the DCA, the clinic + MR model surpassed the clinic and MR models in terms of clinical net benefit in both sets. The clinic + MR model excels in predicting the pathological EPE of prostate cancer. Its superiority over the clinic model underscores its clinical relevance and the potential for broader implementation.

Keywords

• extraprostatic extension
• prostate cancer
• magnetic resonance imaging
• nomogram
• diagnostic performance