Prostate MRI: PI-RADS Cutoffs Shift Cancer Detection Accuracy
Key Takeaways
- Among men with suspected prostate cancer across Chinese hospitals, a higher PI-RADS positive cutoff coincided with greater specificity but lower sensitivity for csPCa.
- Chinese-cohort csPCa detection rose from 5.3% in PI-RADS category 1 to 86.2% in category 5.
- Pooled PI-RADS diagnostic performance varied substantially between centers.
- Discrimination was generally higher with older age and higher PSA, although patterns differed by center.
Investigators in the multicenter PI-RADS study in Insights into Imaging retrospectively assembled 20,429 eligible men older than 50 with suspected prostate cancer from nine Chinese tertiary, district, and county institutions. MRI records dated from January 2002 to May 2025. A separately published worldwide meta-analysis of 70 studies and 13,330 patients served as a comparator, not a newly enrolled validation arm. The primary endpoint was clinically significant prostate cancer (csPCa), defined as International Society of Urological Pathology (ISUP) Grade Group ≥2 (Gleason score ≥3+4=7); any prostate cancer was secondary. Reference assessments included biopsy or prostatectomy histopathology, with a minimum three-year follow-up specified after an initially negative biopsy; not every cancer-negative classification required biopsy. Eight centers used PI-RADS scores from contemporaneous clinical MRI reports, whereas 1,478 historical Peking University First Hospital (PKUH) examinations predating PI-RADS were rescored retrospectively.
For csPCa, pooled sensitivity was 92.6% at PI-RADS ≥3 versus 81.5% at ≥4. Specificity was 51.5% versus 78.1%, respectively. The summary receiver operating characteristic (SROC) area under the curve (AUC) was 0.903 versus 0.826, respectively.
Between-center heterogeneity was substantial: pooled sensitivity I² ranged from 93.0–93.6%, and specificity I² from 98.1–98.8%. Chinese-cohort csPCa prevalence was 46.6% (9,527/20,429) versus 33% (4,399/13,330) in the previously published worldwide dataset. Detection rose across PI-RADS categories in both datasets, broadly following a similar pattern without establishing equivalent performance. Discrimination was generally higher with older age, higher prostate-specific antigen (PSA), and the later MRI-report era, although patterns varied by center and threshold. Excluding historical PKUH examinations left the primary performance pattern similar.
Retrospective referral and partial verification complicate interpretation: biopsy depended on clinical factors, and not every man undergoing MRI had histopathology or complete long-term follow-up. Radiologists had access to PSA and other clinical information, while acquisition protocols, reader practices, and reference standards differed across sites. Higher prevalence, a different PI-RADS category mix, verification procedures, and underlying methods prevent a direct equivalence test with the separately published international meta-analysis. Report period was only a proxy for PI-RADS version, not proof that version 2.1 caused later-era differences; age and PSA associations do not validate subgroup biopsy thresholds. Incomplete center-level covariates precluded formal meta-regression to identify causes of institutional variation. These data come from Chinese institutions, not a directly assessed U.S. cohort.
The authors interpreted the pooled PI-RADS operating points as a sensitivity–specificity trade-off. In their assessment, substantial institutional variation makes those estimates descriptive, context-dependent benchmarks rather than a uniform measure of performance at every hospital.
Clinician Questions
How were cancer-negative historical PKUH prostate MRI cases assessed when no biopsy was performed?
For cancer-negative cases without biopsy in the historical Peking University First Hospital prostate MRI cohort, investigators used structured clinical follow-up lasting 3–10 years, including serial PSA assessment, digital rectal examination, and repeat MRI when clinically indicated. No prostate cancer emerged during that follow-up; unlike biopsy-confirmed negative cases, this classification rested on surveillance and does not establish absence of cancer in all men undergoing MRI.
What biopsy approaches contributed to the Chinese PI-RADS prostate cancer accuracy estimates?
During the Chinese institutional prostate MRI study period, transrectal ultrasound-guided systematic biopsy predominated in institutional practice, while cognitive MRI-targeted biopsy became more common at tertiary centers from about 2015. Biopsy approach, core count, and targeting method were not consistently available at the individual-patient level, so the accuracy estimates do not reflect a uniform biopsy procedure.
This content is intended for healthcare professionals for educational purposes. The article and its image were developed with AI assistance and reviewed by members of the ReachMD team prior to publication. Please refer to the original source linked for the complete and most up-to-date information.