The Early Treatment Diabetic Retinopathy Study (ETDRS), conducted at 22 clinical centers during the period 1980 to 1989, collected baseline data on C-peptide levels after ingestion of Sustacal in 582 patients with diabetes mellitus, prior to enrollment in the trial. Data on several clinical factors associated with diabetes were also collected from all 3711 enrolled patients. C-peptide data were used to develop sets of clinical criteria for the classification of ETDRS patients and to compare and contrast definitions of type of diabetes used in previous studies. The distribution of C-peptide levels was strikingly bimodal, suggesting a division of study participants into two groups--those with levels at 80 pmol/L or less and those with more than 80 pmol/L of C-peptide after Sustacal ingestion. Constellations of clinical characteristics that could serve as proxies for C-peptide level were ascertained. The result was two sets of clinically developed definitions for type of diabetes in the ETDRS. According to the more restrictive set of definitions, three groups were identified, compared to two groups using the "broad" set of definitions. Discriminant analysis was also used to classify ETDRS patients, yielding similar results. A comparison of definitions of type of diabetes used in the ETDRS and in previous studies revealed that even in the absence of C-peptide data, clinically derived definitions provided good discrimination between type I and type II diabetes.