The present study was undertaken to compare the ability of magnetic resonance imaging (MRI) and transrectal ultrasound (TRUS) to identify the presence or absence of extracapsular tumor extension and seminal vesicle involvement in patients with clinically localized (stages A and B, T1-T2) prostatic cancer. Sixty-four patients (ages 44-77 years) underwent radical retropubic prostatectomy and pelvic lymphadenectomy after staging with both MRI and TRUS. Surgical specimens were step-sectioned and examined at multiple levels. Radiographic staging was correlated with pathological staging. Both MRI and TRUS demonstrated better staging accuracy than conventional digital rectal examination [42% for digital rectal examination (DRE), 63% by TRUS, and 67% by MRI], although they suffered from understaging (31% by TRUS and 22% by MRI). The positive predictive value (PPV) for stage B disease was disappointingly low (42% for DRE, 53% for TRUS and 59% for MRI). However, the PPV for stage C disease was higher (81% for TRUS and 77% for MRI). In the evaluation of extracapsular invasion, the PPV of TRUS (81%) was higher than that of MRI (77%). The negative predictive value (NPV), however, for both TRUS and MRI was low (58 and 56%, respectively). In the evaluation of seminal vesicle invasion, both TRUS and MRI demonstrated poor PPV (50 and 40%, respectively), but the NPV was excellent (90% for TRUS and 96% for MRI). MRI failed to detect metastatic lymph nodes in 4 of 6 patients while falsely predicting their presence in 4 patients.(ABSTRACT TRUNCATED AT 250 WORDS)