The effect of changing from the original to the new National Diabetes Data Group diagnostic criteria for diabetes mellitus was to delete 16.5% of the original diabetes incidence cohort described among Rochester, Minnesota, residents In 1945–1969, to shift the clinical spectrum at diagnosis toward more severe disease, to reduce relative survival, and to increase the risk of developing macro- or microvascular complications. The changes in apparent natural history were unexpectedly modest in magnitude, however, and should have little practical effect on comparisons of diabetes prognosis under the two different sets of diagnostic criteria.