Changes in prostate-specific antigen (PSA), used to estimate PSA doubling times, may reflect prostate cancer growth. To determine if PSA doubling time prior to diagnosis predicted outcome in men with prostate cancer, we evaluated 16 men with prostate cancer who had (1) serial PSA determinations (mean 9.9) on frozen sera from twelve to twenty-six years before diagnosis; (2) at least five years of follow-up in those subjects without metastatic disease (range 5.5-12.3 years); and (3) archival material from diagnosis available for pathologic evaluation. PSA doubling time prior to diagnosis was investigated with relation to patient outcome (regardless of treatment) and the known predictors of tumor behavior, Gleason score and nuclear morphometry. In 5 of 16 men who had evidence of metastatic disease at diagnosis, metastasis developed or they died of prostate cancer during follow-up (group 1). Eleven of 16 had no evidence of metastatic disease during follow-up (group 2). Both Gleason score and variance of nuclear roundness (VNR) were significantly greater for group 1 (p < 0.05). There was no significant difference between the two groups with respect to PSA doubling time, and the PSA level at diagnosis did not correlate with the development of metastatic disease. One of 5 men with no PSA level greater than 4.0 ng/mL prior to diagnosis died within two years of diagnosis. These data suggest that (1) a normal PSA at diagnosis does not exclude an aggressive cancer, and (2) changes in PSA that occur before the diagnosis of prostate cancer may not always predict outcome. Since PSA level is influenced by tumor grade, an inability to correct PSA for tumor grade could have influenced the results.