To develop a triage strategy to reduce negative and indeterminate multiparametric MRI (mpMRI) scans in patients at risk for prostate cancer (PCa).
In this retrospective study, we evaluated 865 patients with no prior PCa diagnosis who underwent prostate mpMRI between 2009 and 2017. Age, prostate volume, PSA, and PSA density (PSAd) were assessed as predictors for positive mpMRI, defined as PI-RADSv2/Likert score ≥ 4. The patient cohort was split into training (n=605) and validation cohorts (n=260). The optimal threshold to rule out a positive mpMRI was chosen to achieve a negative predictive value (PV) of > 90%.
All clinical variables were significant predictors of positive mpMRI (p<0.05). PSAd (AUC=0.74) outperformed other parameters in diagnostic accuracy and yielded no significant difference compared to a multivariate model (AUC=0.75). At a PSAd > 0.078 ng/ml2 sensitivity, specificity, positive and negative PV were 94%, 29%, 22% and 95%, respectively, resulting in 25% (64/260) fewer scans. Using the multivariate model, sensitivity, specificity, positive and negative PVs were 85%, 32%, 22%, and 91%, respectively, resulting in 29% (75/260) fewer scans. Biopsies in men who would not have undergone mpMRI according to our proposed strategies revealed 2 clinically significant PCas using PSAd, and 1 using the multivariate model.
In patients at risk for PCa, applying a multivariate prediction model or using a PSAd cutoff of 0.078 ng/ml2 can result in 25-29% fewer mpMRI scans being performed while missing only a minimal number of csPCas. Further prospective validation is required.
The Journal of urology. 2019 Sep 03 [Epub ahead of print]
Dominik Deniffel, Yucheng Zhang, Emmanuel Salinas, Raj Satkunasivam, Farzad Khalvati, Masoom A Haider
Lunenfeld-Tanenbaum Research Institute, Sinai Health System, Toronto, Ontario, Canada., Department of Urology and Center for Outcomes Research, Houston Methodist Hospital, Houston, Texas.