In vivo beta-cell function tests are used increasingly in humans during the preclinical phase of insulin-dependent diabetes mellitus (IDDM), but the severity of the beta-cell loss responsible for the abnormalities seen in these tests is unknown. We have measured several physiological beta-cell function tests--fasting plasma glucose, glucose disappearance constant, fasting insulin, acute insulin responses to arginine (AIRarginine) and glucose (AIRglucose), and glucose potentiation of AIRarginine (delta AIRarginine/delta G) and two direct objective measurements (pancreatic insulin content [PIC] and quantitative beta-cell mass)--in adolescent male baboons (Papio anubis/cyanocephalus). We have correlated in vivo measurements obtained within 3 days after the animals were killed with in vitro estimates of PIC and beta-cell mass in 15 animals, (2 nondiabetic requiring insulin treatment and 13 after varying doses of streptozocin to induce degrees of beta-cell damage ranging from normoglycemia to severe hyperglycemia). There was a strong linear correlation between beta-cell mass and PIC (r = 0.79, P less than 0.001). Physiological measures of beta-cell function were significantly correlated with both PIC and beta-cell mass. The correlations between physiological measures and beta-cell mass were linear and intercepted the beta-cell mass axis at 0.15-0.2 g, suggesting that in vivo measures of beta-cell function approach 0 when there is still approximately 40-50% of the beta-cell mass detectable histologically. With PIC, the linear correlations intercepted the axes close to 0. These findings provide considerable validity to the measurements of beta-cell function used in preclinical IDDM in humans.(ABSTRACT TRUNCATED AT 250 WORDS)