To explore whether elevated plasma glucose might progress to diabetes via a mechanism that involves plasma hypertonicity, we evaluated the independent and joint effects of these variables on diabetes risk. Community-dwelling older adults (70+years), who reported no previous diagnosis of diabetes and had glucose levels below 200 mg/dl in the 1992 Duke EPESE survey, were re-interviewed in 1996 for diabetes status (n = 979). Plasma tonicity at baseline was estimated from serum glucose, sodium, and potassium measures. In logistic regression models that controlled for glucose, as well as age, sex, race, weight status, activity level, serum creatinine, history of heart disease, stroke and cancer, plasma hypertonicity (> or = 300 mOsm/l) was independently associated with increased odds of developing diabetes (OR = 2.0, 95% CI: 0.9-4.2). Hypertonicity magnified the effects of elevated glucose (> or = 125 mg/dl), such that individuals with both exposures were over four times more likely to develop diabetes than those with elevated glucose only (OR = 4.9, 95%CI: 1.7-14.3), adjusting for all covariates. When tonicity was replaced by its determinant variables, glucose, sodium and potassium, in the multivariable models, independent effects of sodium were also observed. Further work is needed to pursue plasma hypertonicity as a factor in the progression of elevated plasma glucose to diabetes.