We appreciate the comments of Stevens and Holman (1) regarding our analysis comparing the accuracy of the U.K. Prospective Diabetes Study (UKPDS) (2) and Framingham models in the prediction of 10-year risk for coronary artery disearse (CAD) in diabetic patients. In our letter, we mentioned that the diagnosis of CAD was established by coronary angiography. In addition, this study was performed in a cohort of patients with type 2 diabetes and included a 10-year follow-up. In the sample of cases with myocardial infarction who were examined during this analysis, fatal and nonfatal cases were included.
We agree that this analysis was retrospective with its disadvantages, and it is absolutely equitable that the variables used by both of these models should have been determined at the beginning of follow-up. However, it is obvious that this study was designed as a retrospective, since the UKPDS prediction model has recently constituted a useful cardiovascular disease (CVD) risk prognostic model. In the epidemiologic analysis of the UKPDS, Stratton et al. (3) showed that the effect of HbA1c values on microvascular complications over a range of 5.5–11% was nearly 10-fold, whereas the effect of HbA1c on myocardial infarction incidence, over the same range of HbA1c, was 2-fold.
Regarding the effect of HbA1c on myocardial infarction incidence, the range of HbA1c was from 5.5 to 11%, so we can say that 11% (upper limit) is two times 5.5% or 1 more time 5.5% (lower limit). Therefore 1 refers to one time and not 1%. The percent symbol was added by mistake.
Moreover, as the UKPDS study showed, improved glycemic control had no significant impact in cardiovascular outcomes in patients with type 2 diabetes. The Veterans Affairs Cooperative Study (4) also showed a nonstatistical significant deterioration of CVD events in intensively treated patients compared with those receiving standard treatment.
The exception for reducing mortality from myocardial infarction was in the overweight UKPDS cohort patients treated with metformin (5). However, metformin use was not associated with lower blood glucose levels compared with insulin or sulfonylureas, so it can be assumed that the cardioprotective effects of metformin could be interpreted by its well-known actions in the atherothrombotic risk profile (6) and blood pressure levels through nonglycemic pathways.
Summarizing the previously epidemiologic data, since coronary heart disease is a multifactorial disease and glycemic control is not significantly associated (7,8) with reduced CVD risk in diabetic patients, it can be assumed that the small contribution of chronic hyperglycemia (HbA1c) at the incidence of macrovascular complications has a respectively weak contribution on the calculation of coronary heart disease risk by the UKPDS mathematical model.
Finally, we absolutely agree with Stevens and Holman that prospective studies will be needed to determine the validity of the predictive value for both of these models.