Disproportionate hyperproinsulinemia is one manifestation of the B-cell dysfunction observed in non-insulin-dependent diabetes mellitus (NIDDM), but it is unclear when this abnormality develops and whether it predicts the development of NIDDM. At baseline, measurements of proinsulin (PI) and immunoreactive insulin (IRI) levels were made in 87 second-generation Japanese-American men, a population at high risk for the subsequent development of NIDDM, and, by using World Health Organization criteria, subjects were categorized as having normal glucose tolerance (NGT; n = 49) or impaired glucose tolerance (IGT; n = 38). After a 5-year follow-up period, they were recategorized as NGT, IGT, or NIDDM using the same criteria. After 5 years, 16 subjects had developed NIDDM, while 71 had NGT or IGT. Individuals who developed NIDDM were more obese at baseline, measured as intra-abdominal fat (IAF) area on computed tomography (P = 0.046) but did not differ in age from those who continued to have NGT or IGT. At baseline, subjects who subsequently developed NIDDM had higher fasting glucose (P = 0.0042), 2-h glucose (P = 0.0002), fasting C-peptide (P = 0.0011), and fasting PI levels (P = 0.0033) and disproportionate hyperproinsulinemia (P = 0.056) than those who continued to have NGT or IGT after 5 years of follow-up. NIDDM incidence was positively correlated with the absolute fasting PI level (relative odds = 2.35; P = 0.0025), even after adjustment for fasting IRI, IAF, and body mass index (relative odds = 2.17; P = 0.013).(ABSTRACT TRUNCATED AT 250 WORDS)