The association between a prior diagnosis of diabetes mellitus and cancer was examined in a large multisite case-control study based on interviews of 8220 white males and 6690 white females aged 30-89 who were admitted to Roswell Park Memorial Institute between 1957 and 1965 for the diagnosis and treatment of cancer. A prior history of diabetes mellitus as diagnosed by a physician was examined as a risk factor for cancer by conducting sex- and site-specific comparisons with a common control group of 2363 white males and 2475 white females who were admitted to the same hospital during the same years, but were found not to have cancer. The findings are consistent with previous studies which have shown diabetes to be a risk factor for cancer of the uterine corpus. A positive association was also noted between diabetes and cancer of the vulva and vagina. The evidence for excess estrogen levels as the underlying mechanism for increased risk of uterine and vulvo-vaginal cancer is reviewed. No consistent association between diabetes and pancreatic cancer was observed in this study, except when those cases in which diabetes was diagnosed within a year before the diagnosis of pancreatic cancer were included. This suggests that diabetes may be an early presenting sign of pancreatic cancer rather than a risk factor. A positive association between prior diagnosis of diabetes was noted for kidney cancer and nonmelanoma skin cancer in females only. No association between type of treatment (i.e. insulin, oral agents) and cancer risk was noted.