When to Biopsy the Seminal Vesicles: A Validated Multiparametric Magnetic Resonance Imaging and Target Driven Model to Detect Seminal Vesicle Invasion of Prostate Cancer.

Abstract

Current imaging and biopsy practices offer limited insight into preoperative detection of seminal vesicle invasion despite the implications for treatment decisions and patient prognoses. We identified magnetic resonance imaging features to assess the risk of seminal vesicle invasion and inform the inclusion of seminal vesicle sampling during biopsy. Patients underwent multiparametric magnetic resonance imaging and fusion targeted biopsy with or without seminal vesicle biopsy. Magnetic resonance imaging suspicion of seminal vesicle invasion, multiparametric magnetic resonance imaging of prostate base lesions of moderate or greater suspicion, extraprostatic extension, anatomical zone and biopsy data were used to generate multivariable logistic regression models. One model without and one with biopsy data were externally validated in a multi-institutional cohort. Decision curve analyses were done to determine net benefit of the 2 models. The training and validation cohorts comprised 564 and 250 patients, respectively. In the training cohort 55 patients (9.8%) had pathologically confirmed seminal vesicle invasion. In the prebiopsy model magnetic resonance imaging suspicion of seminal vesicle invasion (OR 9.5, 95% CI 4.0-22.4, p <0.001), multiparametric magnetic resonance imaging base lesions of moderate or greater suspicion with extraprostatic extension (OR 13.6, 95% CI 4.0-46.5, p <0.001), and a transition and/or central zone location (OR 11.6, 95% CI 3.5-38.3, p <0.001) showed strong correlations. In the post-biopsy model the risk of pathologically confirmed seminal vesicle invasion increased with the base Gleason Group (Gleason Group 5 OR 85.3, 95% CI 11.8-619.1, p <0.001). In the validation cohort the AUC of the prebiopsy and post-biopsy models was 0.84 and 0.93, respectively (p = 0.030). Magnetic resonance imaging evidence of seminal vesicle invasion or extraprostatic extension at the prostate base transition and/or central zone and high grade prostate cancer from the prostate base are significant features associated with an increased risk of pathologically confirmed seminal vesicle invasion. Our models successfully incorporated these features to predict seminal vesicle invasion and inform when to biopsy the seminal vesicles.