PI-RADS category and biopsy-confirmed clinically significant prostate cancer in a selected Ecuadorian multiparametric MRI cohort: a cross-sectional diagnostic study

Multiparametric magnetic resonance imaging (MRI) and the Prostate Imaging Reporting and Data System (PI-RADS) may improve prebiopsy identification of clinically significant prostate cancer (csPC). We evaluated PI-RADS category, prostate-specific antigen (PSA), and PSA density in a selected Ecuadorian cohort. This retrospective, single-center, cross-sectional diagnostic study included 141 men evaluated for suspected prostate cancer from September 2022 through September 2023. The primary analytical cohort comprised 36 patients who underwent multiparametric MRI with PI-RADS assessment and histopathologic verification using combined systematic and MRI-targeted biopsy. The primary target condition was csPC, defined as International Society of Urological Pathology grade group 2 or greater; any prostate cancer was a secondary outcome. Detection was evaluated by PI-RADS category and at prespecified thresholds of 3 or greater and 4 or greater. Threshold analyses were repeated using systematic-core findings as a more index-test-independent reference standard. In the MRI cohort, 24 of 36 patients had csPC and 30 of 36 had any prostate cancer. Clinically significant prostate cancer was detected in 0 of 1 patient with PI-RADS 2, 0 of 4 with PI-RADS 3, 15 of 21 with PI-RADS 4, and 9 of 10 with PI-RADS 5. At a PI-RADS threshold of 4 or greater, all 24 observed csPC cases were classified as MRI-positive. Specificity was 5/12 (41.7%; 95% CI, 15.2%–72.3%), positive predictive value was 24/31 (77.4%; 95% CI, 58.9%–90.4%), and accuracy was 29/36 (80.6%). Only 5 patients were classified as MRI-negative, and none had csPC, yielding an observed negative predictive value of 5/5 (100.0%; exact 95% CI, 47.8%–100.0%). This estimate was therefore based on a very small denominator and was highly imprecise. Using the systematic-core csPC proxy, specificity was 5/13 (38.5%), positive predictive value was 23/31 (74.2%), and accuracy was 28/36 (77.8%). Areas under the receiver operating characteristic curve for csPC were 0.776 for PI-RADS, 0.816 for PSA, and 0.833 for PSA density. Higher PI-RADS categories were associated with greater detection of csPC in this selected, biopsy-verified MRI cohort, and a threshold of PI-RADS 4 or greater provided higher specificity, positive predictive value, and accuracy than a threshold of 3 or greater. The observed 100% negative predictive value was based on only 5 MRI-negative patients and had a wide confidence interval; it should not be interpreted as evidence that PI-RADS below 4 reliably excludes csPC. Similarly, the absence of observed false-negative csPC cases should not be interpreted as evidence of perfect sensitivity because of the small, highly selected cohort and incorporation of MRI-targeted sampling into the combined reference standard. These estimates should not be generalized to screening populations or unselected clinical settings.