Two-hundred and four men with a defined degree of impaired glucose tolerance derived from the Whitehall Survey and its pilot study were enrolled in a therapeutic trial and followed for ten years. For the first five years of the trial approximately half the group received 50 mg phenformin daily and the other half an identical placebo. For the whole ten years of the trial approximately half the group were recommended a diet in which carbohydrate intake was limited to 120 g/day, while the other half was recommended a qualitative limitation of sugar intake. 60 men (29.4%) worsened to diabetes during the follow-up period. The major independent predictor of worsening was the baseline blood glucose level (glucose tolerance). High baseline plasma triglyceride levels and low baseline systolic blood pressure levels were also independent predictors of worsening, though of lower significance (0.01 less than p less than 0.05). Obesity, measured as body mass index, was not a significant risk factor. Thirty-nine men died, with 19 deaths ascribed to coronary heart disease (CHD). Baseline blood pressure significantly predicted both all causes and CHD mortality. All causes mortality rates were higher in the less obese. Allocation to drug or diet therapy did not significantly affect mortality rates or the proportions worsening to diabetes.