To obtain improved accuracy in predicting extracapsular extension (ECE) and seminal vesicle invasion (SVI), we evaluated the variables affecting the predictability of staging magnetic resonance imaging (MRI, phased-array coil) and estimated their impact on accuracy between preoperative MRI staging and histological outcome. A total of 121 patients with localized or locally advanced prostate cancer who underwent robotic radical prostatectomy (RALP) were included. Following transrectal biopsy, all enrolled patients had undergone MRI for staging work-up. After RALP, only 43.8% (53/121) of the patients were matched with the MRI predicted stage. Compared to the matched group in the prediction of ECE, the unmatched group had significantly higher initial prostate-specific antigen (PSA, 12.8 ng ml⁻¹ versus 8.1 ng ml⁻¹, P = 0.048). In the prediction of SVI, initial PSA (8.1 ng ml⁻¹ versus 17.3 ng ml⁻¹, P = 0.009) and biopsy Gleason score (6.5 versus 7.6, P = 0.035) were significantly higher in the unmatched group. When applying clinical cutoffs of initial PSA of 10 and 20 ng ml⁻¹, the accuracy of MRI in the prediction of ECE was decreased in the group with PSA over 20 ng ml⁻¹ (75.6, 64.5 and 37.5%, P = 0.01), and this group had significantly decreased accuracy of MRI in the prediction of SVI (91.5, 77.4 and 37.5%, P<0.01). Applying the clinical cutoff of a Gleason score of 7, the accuracy of MRI in the prediction of SVI was decreased in the higher Gleason score group (93.9, 82.1 and 62.9%, P = 0.01). Thus, for these patient groups, to obtain margin negativity during radical prostatectomy, operative findings, rather than post-biopsy MRI images, may provide substantial information, implying a clinical advantage in conducting MRI before prostate biopsy.