Coronary heart disease rates were estimated in three groups of people participating in the Sanza Survey - newly diagnosed non insulin dependent diabetics, subjects with impaired glucose tolerance and control subjects with normal glucose tolerance. The prevalence of Minnesota-coded ECG abnormalities showed a significant gradient with an approximately twofold excess in both the newly detected diabetic and impaired glucose tolerance cases over the subjects with normal glucose tolerance. The doubling of ECG ischemic changes found in subjects with impaired glucose tolerance and diabetes mellitus appeared to operate almost equally in the absence or presence of several other recognized risk factors for coronary ischemic damage. It is concluded that a relatively low degree of glucose intolerance alone may be important in determining coronary heart disease.