Abstract
OBJECTIVE: To assess the effect of histopathological lesion characteristics on the sensitivity of magnetic resonance imaging (MRI) for per-lesion identification of extracapsular extension (ECE) of prostate cancer. PATIENTS AND METHODS: The study included 176 patients (median age 58.9 years, range 38-77) who underwent endorectal MRI before radical prostatectomy between January 2001 and July 2004, had no previous treatment and had whole-mount step-section pathological specimens showing at least one capsule-abutting lesion. The likelihood of ECE of capsule-abutting lesions was retrospectively scored from 1 to 5 based on radiologists' prospective MRI interpretations. Generalized estimating equation regression models were used to determine the effect of the following histological variables on the sensitivity of MRI for identifying ECE of capsule-abutting lesions: maximum diameter, largest perpendicular diameter (LPD), bi-dimensional diameter product, Gleason grade, and zonal extent. RESULTS: On histopathology, 339 capsule-abutting lesions were found, including 54 with ECE. MRI correctly identified ECE in 36/54 capsule-abutting lesions, including nine of 18 with focal ECE and 27/36 with established ECE, giving sensitivities (95% confidence interval) of 67 (53-78)%, 50 (27-73)% and 75 (58-87)%, respectively. MRI incorrectly identified ECE in 27/285 (9%) capsule-abutting lesions without ECE. MRI sensitivity for per-lesion ECE identification was significantly associated only with histopathological LPD (P = 0.009). Fifty-one patients (29%) had ECE. MRI had a sensitivity (95% confidence interval) of 69 (54-81)% and specificity of 90 (83-94)% for per-patient ECE identification. CONCLUSIONS: The sensitivity of MRI in per-lesion identification of prostate cancer ECE is significantly associated with the lesion LPD at histopathology.