Purpose: Previous analyses have demonstrated that CHD risk prediction models underestimate probability of a CHD event in those with Type 1 Diabetes (T1D). This underestimate can alter the modeling of important outcomes in decision models of diabetes interventions. Our objective was to examine risk factors that may account for underestimation.

Methods: Data were from the Pittsburgh Epidemiology of Diabetes Complications Study (EDC), a prospective cohort study of 658 subjects with childhood onset T1D, diagnosed between 1950 and 1980. Baseline exams took place 1986-1988 (mean age 28, diabetes duration 19 yrs). The cohort has been followed biennially since. The Framingham Risk Engine equation was applied to the EDC data to generate the probability of risk for MI or CHD death. Probabilities were then split in to deciles. Expected and observed events were compared using the Hosmer and Lemeshow goodness of fit statistic.

Results: The cohort experienced 123 CHD events during the follow-up period, with a mean age at onset of 40. When observed and expected probabilities were compared, there was significant lack of calibration (p<0.001). The chi-square comparing observed to expected probabilities was highest in the top three deciles. Framingham risk factors (age, smoking, cholesterol/HDLc, systolic blood pressure) and other risk factors previously found to be associated with CHD in T1D were compared within the top three deciles and within the other seven deciles comparing CHD incident cases to non-cases.

Women in the top three deciles, who had an event, had significantly higher estimated glucose disposal rate (eGDR-calculated using a formula derived from euglycemic-hyperinsulinemic clamp studies), fibrinogen, White blood cell count (WBC), albumin excretion rate (AER), waist/hip ratio, and Beck Depression Inventory Score at baseline than those that did not have an event. For those women in the lower seven deciles, fibrinogen and WBC were significantly different. Men from the top three risk deciles had a higher HbA1, AER, fibrinogen, WBC and eGDR. Those with a history of antihypertensive use also had more events. In the lower seven deciles, only AER was significantly different.

Conclusion: The Framingham risk equation does not adequately predict the probability of a CHD event in T1D. Risk factors including renal disease, WBC and insulin resistance, not considered by the risk equation, may account for the lack of fit.