Critical analysis of the ability of the endorectal coil magnetic resonance imaging scan to predict pathologic stage, margin status, and postoperative prostate-specific antigen failure in patients with clinically organ-confined prostate cancer

Abstract

Purpose: To determine whether there is a role for endorectal coil magnetic resonance imaging (erMRI) in the prediction of pathologic stage, margin status, and/or postoperative prostate-specific antigen (PSA) failure in patients with clinically organ-confined prostate cancer.

Patients and methods: Using erMRI, the radiologic-pathologic correlation of extracapsular extension (ECE) and seminal vesicle invasion (SVI) was evaluated in 445 surgically managed patients. Logistic regression multivariable analysis was applied to the clinical stage, PSA, biopsy Gleason grade, and erMRI findings to assess the outcomes of ECE, SVI, positive surgical margins (PSM), and postoperative PSA failure.

Results: The accuracy of erMRI to predict for ECE and SVI numerically decreased with both increasing PSA and biopsy Gleason score because of the increasing false-negative scans in cases of microscopic transcapsular or seminal vesicle disease. Of patients who could not be categorized into low or high risk for postoperative PSA failure on the basis of clinical stage, preoperative PSA, and biopsy Gleason score, a negative or positive erMRI for ECE or SVI stratified these patients into groups with a 78% versus 21% (P < .0001) 3-year rate of actuarial freedom from PSA failure. In this subgroup, the overall accuracy of the erMRI was 70% +/- 6% and 94% +/- 2% for ECE and SVI, respectively. The most significant predictor on multivariable analysis of PSM was the erMRI finding of ECE (P = .0001).

Conclusion: This initial report suggests that a preoperative erMRI can identify clinically organ-confined prostate cancer patients at high risk for having ECE, SVI, and PSM that otherwise would be missed on the basis of the clinical stage, preoperative PSA, and biopsy Gleason score. Confirmatory studies are needed.