Diabetes mellitus often develops in patients with hypertension. We investigated predictors of diabetes mellitus development in hypertensives at risk of developing the disease in the VALUE trial population. Among the 9995 non-diabetic hypertensive patients at baseline, 1298 patients developed diabetes mellitus during the average follow-up of 4.2 years. New-onset diabetes mellitus was defined from adverse event reports, information about new antidiabetic medication and/or a fasting glucose >or=7.0 mmol l(-1) at the end of trial. Twenty-five potential baseline predictors of new-onset diabetes mellitus were analysed by univariate logistic regression and 14 of 25 predictors were found to be statistically significant with a P-value <0.05. The predictors were in order of decreasing significance; glucose, body mass index (BMI), age, uric acid, non-Caucasian race, haemoglobin, heart rate, randomized study treatment, history of coronary heart disease (CHD), gender, total cholesterol, proteinuria, potassium and creatinine. Multivariate stepwise logistic regression analyses were used and potential baseline predictors of new-onset diabetes mellitus were considered significant by four different models (P-value <0.001). The final multivariate model selected included all patients, but not treatment group as a potential predictor, and the six significant predictors identified from this model were glucose, BMI, non-Caucasian race, age, heart rate and history of CHD. In conclusion, glucose and BMI were the most important predictors of new-onset diabetes mellitus in hypertensive patients at high cardiovascular risk, and easily accessible clinical characteristics strongly predict patients at risk of developing diabetes mellitus.