Background: treated prostate cancer (PCa) patients develop biochemical recurrence (BCR) in 27-53% of cases; the role of MRI in this setting is still controversial. In 2021 a panel of experts proposed a "Prostate Imaging-Recurrence Reporting" (PI-RR) score, aiming to standardize the reporting. The aim of our study is to evaluate the reproducibility of the PI-RR scoring system among readers with different expertise.
Methods: in this monocentric, retrospective observational study, the images of patients who underwent MRI with BCR from January 2017 to January 2022 were analyzed by two radiologists and a radiology resident. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were obtained. Interobserver agreement was calculated. The percentage of the PI-RR score of 3 was estimated to find out the proportion of uncertain exams reported among the readers.
Results: a total of seventy-six patients were included in our study: eight previously treated with RT and sixty-eight who underwent surgery. The accuracy range was 75-80%, the sensitivity 68.4-71.1%, the specificity 81.6-89.5%, PPV 78.8-87.1%, and NPV 72.1-75.6%. The inter-reader agreement using a binary evaluation (PI-RR ≥ 3 as positive mpMRI) demonstrated a correlation coefficient (k) of 0.74 (95% CI: 0.62-0.87). The percentage for the PI-RR score of 3 was 6.6% for reader one, 14.5% for reader two, and 2.6% for reader three.
Conclusion: this study confirmed the good accuracy of mpMRI in the detection of local recurrence of PCa and the good reproducibility of PI-RR score among all readers, confirming it to be a promising tool for the standardization of the assessment of patients with BCR.
Keywords: biochemical recurrence; multiparametric MRI; prostate cancer.