Thirty one studies (1,765 patients; range 20 to 220) were included in the review: 16 studies enrolled previously diagnosed men and used the prostatectomy specimen as the reference standard with correlation to a specific region of the prostate; and 15 enrolled men with suspected prostate cancer and used biopsy without correlation to a specific region of the prostate. Twelve studies were prospective, 10 retrospective and in nine the direction of data collection was unclear. Two studies used a case-control design. Ten of the 31 studies enrolled consecutive men, seven had blinded interpretation of magnetic resonance imaging/magnetic resonance spectroscopy images and 17 excluded uninterpretable results (proportion excluded ranged from 2% to 41% patients and 2% to 73% subparts).
Diagnosed prostate cancer (nine studies): Pooled sensitivity was 68% (95% CI 56% to 78%) and pooled specificity was 85% (95% CI 78% to 90%). Sensitivity was lower in low-risk studies (58%, 95% CI 46% to 69%; four studies) than in high-risk studies (74%, 95% CI 58% to 85%; five studies). Specificity was higher in low-risk patients (91%, 95% CI 86% to 94%) compared with high-risk patients (78%, 95% CI 70% to 84%)
Suspected prostate cancer (seven studies): Pooled sensitivity was 82% (95% CI 59% to 94%) and pooled specificity was 88% (95% CI 80% to 95%). Sensitivity was 75% (95% CI 39% to 93%) and specificity was 91% (95% CI 77% to 97%) in the low-risk group (five studies); there was insufficient data to pool studies for the high-risk group.