A lower risk of prostate cancer among diabetics has been suggested by several but not all studies. However, the studies have not always accounted for time since diagnosis of diabetes mellitus, or have not examined confounding factors such as diet and diagnostic bias. We thus examined this relationship in the Health Professionals Follow-Up Study from 1986 and 1994, in which 1,369 new cases of non-stage A1 prostate cancer were documented in 47,781 men. A prior history of a diagnosis of diabetes (mostly adult-onset) was associated with a reduced risk of prostate cancer (multivariate relative risk [RR] = 0.75; 95 percent confidence interval [CI] = 0.59-0.95) controlling for age, body mass index (wt/ht2) at age 21, and, in 1986, race, vasectomy, and intakes of total energy, total fat, calcium, fructose, and lycopene. After excluding the first year of follow-up after the diagnosis of diabetes, the RR was 0.63 (CI = 0.54-0.89). Prostate cancer was not reduced in the first five years after diagnosis (RR = 1.24, CI = 0.87-1.77), but was lower in the next five years (RR = 0.66, CI = 0.39-1.10) and lowest after 10 years (RR = 0.54, CI = 0.37-0.78); P-value for trend across time = 0.004. Similar associations were noted for advanced cases. Detection bias was unlikely to account for our findings. The basis of this relationship is unclear but may reflect hormonal changes related to diabetes, perhaps low testosterone levels.