The original Whickham Survey documented the prevalence of diabetes and lipid disorders in a sample of 2779 adults aged 18 years and over, which matched the British population structure. The aim of the 20-year follow-up study was to determine the incidence and natural history of diabetes. Outcomes in terms of morbidity and mortality at follow-up were determined in over 97% of the original population. Ninety-four subjects had been identified and treated for diabetes since the first survey, including 17 subjects identified as having a fasting plasma glucose > or = 7.8 mmol l-1 at follow-up. The incidence of diabetes for the total population was 2.2 1000-1 year-1 (95% confidence interval 1.8, 2.6). The risk factors identified at first survey were corrected for age, cut-off at the 95 centile and entered into a log linear model. Those which strongly predicted development of diabetes in the total population were fasting blood glucose (odds ratio (OR) (with 95% confidence intervals) = 2.3 (1.5, 3.5)) and body mass index (OR = 2.2 (1.5, 3.3)) in men, and fasting blood glucose (OR = 2.6 (1.7, 4.1)) and fasting serum triglyceride (OR = 2.8 (1.8, 4.4)) in women. A logit model has enabled the calculation of the probability of developing diabetes 20 years later. It was the characteristics of becoming older such as obesity, hypertriglyceridaemia, and raised fasting blood glucose, rather than age itself, which were associated with the development of diabetes.