A magnetic resonance imaging-based prediction model for prostate biopsy risk stratification

Dear Editor,

We reviewed the article by Mehralivand and colleagues¹ with great interest and commend the authors for closing a gap in the current literature by creating and validating a multiparametric magnetic resonance imaging (mpMRI) prediction model for prostate biopsy risk stratification.

MpMRI has become the imaging modality of choice to improve the accuracy of the prostate biopsy. An article from the National Institutes of Health demonstrated that targeted MR/ultrasound fusion biopsy diagnosed 30% more high-risk cancers and 17% fewer low-risk cancers in comparison with standard biopsy.² While mpMRI targeted biopsy demonstrates higher accuracy for detecting clinically significant cancers, the utility of mpMRI in prediction of outcomes is less established. Radtke and colleagues³ created a risk model incorporating mpMRI parameters and had an area under the curve (AUC) for clinically significant cancers of biopsy-naïve men of 83% and 81% for post-biopsy men; however, this model was not validated outside of its own cohort.

This is where the prediction model developed by Mehralivand and colleagues¹ contributes to the existing literature. The main outcome of this study was that the inclusion of MRI into a prediction model comprising clinical variables increased the AUC of the receiver operating characteristic curve from 72% to 84% in the developmental cohort and from 64% to 84% in the validation cohort for prediction of clinically significant cancer (Gleason 3 + 4 or higher).¹ Further, an MRI-based model decreased the false positive rate compared with the baseline model, without impacting the true positive rate. The MRI model increased the number of biopsies that could have been avoided from 6% by the baseline model to 38%.¹ This study helps to validate the prior research demonstrating the benefit of mpMRI-transrectal ultrasound (TRUS) fusion biopsy in detecting more clinically significant prostate cancer. This model also serves as an addition to the recent literature demonstrating the utility of mpMRI as a useful screening tool for prostate cancer.^4,5

However, this model was developed in patients who had a high suspicion of prostate cancer, that is, elevated prostate-specific antigen (PSA) or abnormal digital rectal examination and presence of suspicious enough lesions on MRI to warrant biopsy. This limits the applicability of this model to only patients with ‘visible’ lesions on MRI whereas a model based on nonimaging variables could potentially work on all patients and help identify clinically significant cancer in patients with MRI ‘invisible’ prostate cancer.⁶

Also, while the model predicts clinically significant cancer on biopsy, an accurate assessment of tumor aggressiveness or Gleason grading is difficult from MRI parameters alone. However, as the mpMRI continues to transition to a screening modality, determining the tumor aggressiveness of visible lesions would be an important secondary goal. That is where biomarkers and genomics can complement imaging findings. Prostate cancer gene 3 (PCA3), prostate health index (PHI), and 4-kallikrein scores have been shown to have a better AUC than PSA in determining prostate cancer, but these markers, to our knowledge, have not been incorporated into a model with MRI.⁷ In future models, incorporation of imaging and biomarkers might help determine the aggressive potential of MRI lesions and aid in their decision to biopsy and predict the oncological outcomes. Nevertheless, we look forward to seeing this model utilized at other centers for validation, and hope that this model serves to limit the unnecessary prostate biopsies. We congratulate Mehralivand and colleagues on the contributions made by this model and look forward to following it in the future.