According to epidemiological and angiographic studies (1,2), diabetic patients present a two to four times greater risk for coronary artery disease (CAD) than nondiabetic individuals.
The Framingham model (3) estimates 10-year CAD risk based on the traditional risk factors, including age, sex, HDL and LDL cholesterol, hypertension, and smoking. In addition to these risk factors, the U.K. Prospective Diabetes Study (UKPDS) model, designed for people with type 2 diabetes (4), incorporates more specific variables, such as HbA1c, age at diabetes diagnosis, and diabetes duration.
The aim of this analysis was to compare the accuracy of these models in the prediction of 10-year risk for CAD in diabetic patients.
Clinical information related to the above factors was retrieved from our diabetic outpatient database with 10-year clinical follow-up. Of the 339 participants (53% women and 47% men), 108 (32%) presented with CAD. We did not observe any statistically significant differences in their demographic characteristics. The diagnosis of CAD was established by coronary angiography. Diabetic patients without a history of CAD were not considered to have CAD; however, if they had presented symptoms of CAD, they would undergo a treadmill test and myocardium scanning with Th 201 (single photon emission computerized tomography).
Receiver operating characteristic curves were constructed for both models to evaluate their accuracy in the prediction of 10-year CAD risk (measured by the area under the receiver operating characteristic curve, range 0.5–1).
Areas under the curves were 0.61 (P < 0.01) and 0.65 (P < 0.01) for UKPDS and Framingham, respectively. The comparative analysis showed similar sensitivity (56 vs. 55%) between these models. A higher specificity in the Framingham model (65 vs. 56%) was noted. We also noted higher positive and negative predictive values of Framingham, 43 and 75%, respectively, compared with 37 and 73% in the UKPDS.
According to the results of this analysis, the Framingham model seems to be more appropriate for the prediction of CAD risk in diabetic patients. An explanation for this could be the small contribution of UKPDS variables (HbA1c, age at diabetes diagnosis, and diabetes duration) to the 10-year risk for CAD.
Regarding the relation of HbA1c to the development of CAD, data from the UKPDS 23 (5) indicated that for each 1% increment in HbA1c there was a 1.11-fold increased risk of CAD, whereas for each 1-mmol/l increment in LDL concentration there was a 1.57-fold increased risk. It should also be noted that an HbA1c increment from 6.5 to 11% (6) just doubles the risk of myocardial infarctions, whereas an HbA1c increment of 1% multiplies the risk of microangiopathic incidents by 10. Additionally, recent studies (7,8) have confirmed the significant role of traditional risk factors in the prediction of CAD in contrast to the poor prognostic value of blood glucose concentration. In conclusion, comparing the above data, the association of HbA1c with the risk for CAD is considered rather weak.
Regarding the contribution of age at diabetes diagnosis and diabetes duration to the 10-year risk, it is well known that in the early stages of the disease, when the symptoms are not apparent, diabetes is frequently underdiagnosed. Moreover, in some cases, although diabetes complications (micro- and/or macrovascular) have already presented, the existence of diabetes is still ignored by the patient. However, it is difficult to determine the exact age of diabetes onset and duration. Therefore, the evaluation of the exact duration of diabetes is potentially inaccurate, which may result in underestimation of CAD risk.
Future efforts should focus on the ultimate estimation and evaluation of the prognostic value of both models using randomized, prospective, comparative studies.