ABSTRACT
OBJECTIVES: Endorectal coil MRI (endoMRI) of the prostate is useful to evaluate tumor localization. There is little evidence on patient characteristics affecting its diagnostic performance. We evaluate the influence of clinical and histological parameters on the accuracy of endoMRI. METHODS: Sixty-nine patients with prostate cancer were included. After virtually dividing the prostate into pixels of 1 cm2, results of endoMRI were compared with those from prostatectomy specimens' whole-mount sections. Univariate and multivariate analyses were performed to calculate the impact of clinical and histological parameters on the number of appropriately described pixels. RESULTS: In 9, no tumor could be demonstrated by endoMRI. 48.3% of patients were staged correctly, 23.3% were over- and 28.3% understaged. Mean rates of correctly labeled pixels were 0.44 (± 0.04 SEM) for tumor and 0.90 (± 0.01) for benign segments. In univariate analysis, the rate of correctly labeled tumor segments showed significant positive correlations with Gleason score ≥7 and negative correlations with prostate weight and multifocality. The rate of correctly labeled benign segments showed significant negative correlation with tumor weight. All factors were independent variables in multivariate analysis. CONCLUSIONS: The reliability of endoMRI depends on clinical parameters. Higher Gleason scores, unifocal tumors and smaller prostate volumes ameliorate endoMRI performance.
