The Prostate Imaging for Recurrence Reporting (PI-RR) system has been recently proposed to promote standardisation in the MR assessment of prostate cancer (PCa) local recurrence after radical prostatectomy (RP) and radiation therapy (RT). This study aims to evaluate PI-RR's diagnostic accuracy, assess the inter-observer reliability among readers with variable experience, and correlate imaging results with anatomopathological and laboratory parameters. Patients who underwent a pelvic MRI for suspicion of PCa local recurrence after RP or RT were retrospectively enrolled (October 2017-February 2020). PI-RR scores were independently assessed for each patient by five readers with variable experience in prostate MRI (two senior and three junior radiologists). Biochemical data and histopathological features were collected. The reference standard was determined through biochemical, imaging, or histopathological follow-up data. Reader's diagnostic performance was assessed using contingency tables. Cohen's kappa coefficient (κ) and intraclass correlation coefficient (ICC) were calculated to measure inter-observer reliability. The final cohort included 120 patients (median age, 72years [IQR, 62-82]). Recurrence was confirmed in 106 (88.3%) patients. Considering a PI-RR score ≥ 3 as positive for recurrence, minimum and maximum diagnostic values among the readers were as follows: sensitivity 79-86%; specificity 64-86%; positive predictive value 95-98%; negative predictive value 33-46%; accuracy 79-87%. Regardless of reader's level of experience, the inter-observer reliability resulted good or excellent (κ ranges across all readers: 0.52-0.77), and ICC was 0.8. Prostate-specific antigen (PSA) velocity, baseline-PSA, and trigger-PSA resulted predictive of local recurrence at imaging. The PI-RR system is an effective tool for MRI evaluation of PCa local recurrence and facilitates uniformity among radiologists. This study confirmed the PI-RR system's good diagnostic accuracy for the MRI evaluation of PCa local recurrences. It showed high reproducibility among readers with variable experience levels, validating it as a promising standardisation tool for assessing patients with biochemical recurrence. • In this retrospective study, the PI-RR system revealed promising diagnostic performances among five readers with different experience (sensitivity 79-86%; specificity 64-86%; accuracy 79-87%). • The inter-observer reliability among the five readers resulted good or excellent (κ ranges: 0.52-0.77) with an intraclass correlation coefficient of 0.8. • The PI-RR assessment score may facilitate standardisation and generalizability in the evaluation of prostate cancer local recurrence among radiologists.
Read full abstract