A Predictive Model Based on Biparametric Magnetic Resonance Imaging and Clinical Parameters for Improved Risk Assessment and Selection of Biopsy-Naïve Men for Prostate Biopsies

It has been shown that prostate cancer risk prediction models that combine clinical parameters with blood and/or urine genetic and protein biomarkers improve individualized pre-biopsy risk assessment but have limited discriminatory power in detecting and ruling out significant disease. In addition, multiparametric MRI is increasingly used to differentiate between men with enhanced risk of significant cancer who require invasive prostate biopsies and men with high probability of benign conditions or insignificant cancer, who might safely avoid them. Still, significant Gleason grade group ≥2 tumors are missed by multiparametric MRI. Therefore, additional clinical predictors are needed to supplement MRI as a triage test to rule out significant cancers and spare many men from unnecessary invasive prostate biopsies. However, it is unclear how risk models and MRI should be used conjointly. Furthermore, state-of-the-art multiparametric MRI is time-consuming and expensive and would place a significant financial and resource burden on any healthcare system if used before biopsy on all men. A simpler, more rapid biparametric MRI method that uses fewer scan sequences, no intravenous contrast media and maintains high diagnostic accuracy would decrease costs and could facilitate a more widespread clinical implementation of pre-biopsy prostate MRI. Therefore, the objective of this study was to develop a multivariable predictive model based on biparametric MRI findings and clinical parameters to detect and rule out significant prostate cancer in biopsy-naïve men.

We included 876 biopsy-naive men with clinical suspicion of prostate cancer (prostate-specific antigen, <50 ng/mL; tumor stage, <T3) that all underwent pre-biopsy prostate biparametric MRI (T2- and diffusion-weighted) followed by 10-core standard biopsies (all men) and MRI-transrectal ultrasound fusion targeted biopsies of biparametric MRI-suspicious lesions (suspicion score, ≥3). We then created four prediction models based on MRI scores and clinical parameters (age, tumor stage, prostate-specific-antigen [PSA] level, prostate volume, and PSA density) for significant prostate cancer detection (defined as any biopsy-core with Gleason grade-group, ≥2) and compared them by analyzing the areas under the curves and decision curves (Figure 1).

Figure 1:

BTA 2019 predictive model fig 1

To the best of our knowledge, we here present the first predictive model for significant prostate cancer estimation that combines a prospectively derived biparametric MRI score with easily obtainable clinical parameters and uses results from advanced biopsy techniques (standard plus MRI targeted biopsy) as standard reference. Our study shows that the biparametric MRI-derived score was the strongest single explanatory predictor of significant cancer and the discriminatory accuracy was significantly in the best predictive model when the MRI scores were combined with age, clinical stage, and PSA density. This model had the highest discriminatory power (area under the curve, 0.89), showed good calibration on internal bootstrap validation, and resulted in the greatest net benefit on decision curve analysis for clinically relevant biopsy thresholds >5% in balancing detection of significant cancer (benefits) against the risk of undergoing unnecessary biopsies (harms).

Overall, we believe that our model provides an individualized actual probability of having Gleason grade group ≥2 cancer on prostate biopsy and can be used to counsel men considering the option of avoiding invasive biopsies. Such clinical tools that can inform patients on a personal level are needed in this era of clinical shared decision-making.

Furthermore, the use of pre-biopsy MRI is rapidly evolving and is now recommended in the European Association of Urology guideline for all men under suspicion of prostate cancer. A simpler and more rapid biparametric MRI approach (decreased scan time, lower cost, no intra-venous contrast-media) holds the potential for more widespread clinical implementation of pre-biopsy MRI to benefit all men under suspicion of prostate cancer. However, we acknowledge that our risk models were developed using data from a Scandinavian population where systematic PSA screening is not performed and the models might, therefore, perform differently in a PSA-screened or more ethnically heterogeneous cohort. Thus, future external validation in other cohorts will be needed as more biparametric MRI data become increasingly available.

Written by: Lars Boesen, M.D, Ph.D., Departments of Urology and Urological Research, Copenhagen University Hospital Herlev, Herlev, Denmark

Read the Abstract

Login