We assessed the value of postsurgical computed tomography (CT) in detecting locally recurrent prostatic carcinoma and determined the most effective CT findings.We studied 13 patients with recurrent disease (prostate specific antigen [PSA] > 0.4 ng/ml) and 11 patients with no evidence of recurrence (PSA < 0.3 ng/ml). Pelvic CT scans were independently interpreted by four readers who were unaware of patient status. Readers measured tissue volume in the prostatic fossa and evaluated each scan for the presence of six potentially discriminating criteria. We determined sensitivity and specificity and developed mean and pooled receiver operating characteristic (ROC) curves for each criterion and for overall rating.The respective mean ROC curves, sensitivity, and specificity for each criterion were as follows: irregular tissue margins = .50, .67, and .32; inhomogeneous tissue density = .35, .76, and .11; asymmetric residual seminal vesicles = .68, .86, and .16; fat infiltration around seminal vesicles = .67, .69, and .43; infiltration of perirectal fat = .60, .71, and .40; and margins of the levator ani = .50, .78, and .09. The overall rating of whether a scan was normal or reflected recurrent cancer was .56, .75, and .32. Mean tissue volume in the prostatic fossa was 15.01 cm 3 for the positive cases and 11.06 cm 3 for the negative cases (p < .05), but because of a large overlap, this difference was not practically significant.CT scanning is not an effective technique for detecting recurrent prostate malignancy. Normally, there is a moderate amount of soft tissue in the prostatic fossa postoperatively that should not be confused for malignancy.