A study of preradiation and postradiation, serial serum prostate-specific antigen (PSA) levels was performed in patients who had clinically localized prostate cancer. The prognostic value of the PSA in pretreatment evaluation and posttreatment follow-up was assessed. Sixty-three patients who presented with clinically localized prostate cancer and who were treated with external-beam radiation therapy were followed-up for a median of 25 months. A serum PSA and physical examination were performed at 3-month intervals, and a bone scan was done yearly. An increase in PSA triggered an additional metastatic workup. Prostate rebiopsy was performed for new, palpable nodules or for a serial increase in PSA in the context of a negative metastatic workup. Forty-one patients remained recurrence-free and 22 recurred clinically, 15 distantly and seven locally. The PSA was the strongest, independent, pretreatment prognostic indicator (P = .019) among pretreatment PSA, stage, and grade, but lost significance when the serum prostatic acid phosphatase (PAP) status was included. The initial rate of the PSA decrease after radiation (median half-life, 2.6 months) failed to predict outcome. Recurrence-free patients reached postradiation PSA levels that were equivalent to those reported in disease-free male populations; failure of the PSA to reach such normal levels was a multivariate predictor of subsequent failure (P less than .037). All clinicopathologic documentations of failure were preceded by an increase in PSA levels during follow-up. Delayed versus early PSA increase was associated with clinically localized versus metastatic first recurrence. The serum PSA is an independent pretreatment and posttreatment predictor of outcome. Additionally, for a median follow-up of 25 months, delayed PSA failure is associated with clinically localized rather than metastatic recurrence, a relationship that may help in selection for local salvage therapy.