Reproducibility of individualized coronary heart disease risk calculations in patients with diabetes mellitus

Aims To determine prospectively, the reproducibility of individualized coronary heart disease (CHD) risk estimations in a high-risk (diabetic) population. Methods One hundred and three patients attending a hospital diabetes clinic who were in the primary prevention category for CHD had measurements of cholesterol, HDL-cholesterol and systolic blood pressure (SBP) performed in one of 13 general practices and then 2 weeks later in the hospital clinic. The data were combined with age, sex, smoking history and diabetic status data to produce a 10-year CHD risk estimate for each occasion using the Framingham algorithm. Results The coefficients of variation for cholesterol, HDL and SBP were 6.0%, 9.4% and 7.0%, respectively. When classified by treatment thresholds of 15% and 30% 10-year CHD risk, 88% of patients were classified in the same category on both occasions. Kappa values for the 15% and 30% risk thresholds were 0.7-1 and 0.82. This indicates good interobserver agreement for the estimation of CHD risk. The use of a single BP rather than the mean of two, resulted in seven of 206 estimations of CHD risk performed in 103 patients crossing a risk threshold, with 6/7 being placed in a higher risk category. Conclusions Estimation of CHD risk on a single occasion is sufficient to make robust treatment decisions based on risk thresholds. Use of a single BP measurement rather than the mean of two overestimates the risk category in around 3% of cases.