Summary Although insulin resistance and decreased insulin secretion are characteristic of established non-insulin-dependent diabetes mellitus (NIDDM), which of these metabolic abnormalities is the primary determinant of NIDDM is still controversial. A disproportionate increase in the proinsulin to insulin ratio has been proposed as a marker of compromised insulin secretion. We examined the association of fasting immunoreactive insulin (which cross-reacts with proinsulin), specific insulin (which does not cross-react with proinsulin), total immunoreactive proinsulin (or insulin precursors), and the fasting proinsulin/specific insulin ratio to the risk of developing NIDDM in the 3.25-year follow-up of the Mexico City Diabetes Study. These measurements were made in 85 subjects who subsequently converted to NIDDM (prediabetic subjects) and in 85 age and gender matched subjects who remained non-diabetic at follow-up (control subjects). Immunoreactive insulin, proinsulin and the proinsulin/specific insulin ratio were significantly higher in prediabetic than in control subjects. However, the relation between specific insulin and the development of NIDDM was weaker than for proinsulin or immunoreactive insulin. After further adjustment for obesity, body fat distribution and glucose tolerance status, proinsulin and the proinsulin/specific insulin ratio, but not specific or immunoreactive insulin, predicted conversion to NIDDM. A high proinsulin/specific insulin ratio predicted conversion to NIDDM both in subjects with normal and those with impaired glucose tolerance at baseline. We conclude that in prediabetic subjects increased proinsulin, a marker of islet cell distress or compromised insulin secretion, is associated with rapid conversion (within 3.25 years) to NIDDM even in obese populations. [Diabetologia (1997) 40: 830–837]