icon-folder.gif   Conference Reports for NATAP  
  9th International Congress on Drug Therapy in HIV Infection
November 9-13, 2008
Back grey_arrow_rt.gif
D:A:D Cohort Group Offers Tool to Predict Short-Term Diabetes With HIV
  9th International Congress on Drug Therapy in HIV Infection, November 9-13, 2008, Glasgow
Mark Mascolini
D:A:D cohort investigators devised and validated a diabetes prediction tool that works better than the Framingham equation in people with HIV infection [1]. The D:A:D calculator probably outperforms the Framingham predictor because it incorporates both traditional and HIV-related risk factors.
The analysis involved 13,609 HIV-infected D:A:D cohort members with a complete diabetes risk profile. In that group 251 had new-onset diabetes mellitus during a median follow-up of 3.5 years (interquartile range 1.36 to 6.16). D:A:D investigators split the 13,609-person group into a "training set" to calculate the influence of risk factors on diabetes diagnosis and a "validation set" to test the accuracy of the prediction model. The training set included 8990 people with 170 new diabetes diagnoses in a median 3.51 years of follow-up, and the validation set included 4619 people with 81 new diabetes diagnoses in a median 3.47 years of follow-up.
Traditional diabetes risk factors considered included age, gender, ethnicity, glucose, blood pressure, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein cholesterol, triglycerides, body mass index, and family history of coronary heart disease as a surrogate for diabetes mellitus. HIV-related covariates were HIV exposure category, duration since HIV-positive test, prior AIDS, CD4 count, viral load, lipodystrophy, duration of antiretroviral exposure, antiretroviral class exposure, and hepatitis B or C status.
The model did not consider individual antiretrovirals. The D:A:D team excluded race from their calculations because records did not specify the race of many cohort members. They excluded smoking because it appeared to protect against diabetes. Finally, the investigators figured 8-year probability of diabetes with the Framingham model, then converted that prediction to the shorter D:A:D follow-up interval.
Seven factors predicted diabetes in the training set:
⋅ Glucose above 5.6 mmol/L (f) (100 mg/dl) or 7.8 mmol/L (nf) (140 mg/dl): incidence rate ratio [IRR] 12.22, 95% confidence interval [CI] 9.0 to 16.59, P < 0.001
⋅ Body mass index (BMI)
· BMI 25 to 30: IRR 2.19, 95% CI 1.56 to 3.07, P < 0.001
· BMI at or above 30: IRR 3.6, 95% CI 2.28 to 5.69, P < 0.001
⋅ Male: IRR 1.66, 95% CI 1.05 to 2.61, P < 0.029
⋅ Age (per 5 years older): IRR 1.23, 95% CI 1.14 to 1.33, P < 0.001
⋅ HIV exposure category
· Injecting drug use: IRR 2.2, 95% CI 1.42 to 3.39, P < 0.001
· Heterosexual: IRR 1.52, 95% CI 1.02 to 2.26, P = 0.038
· Other: IRR 2.52, 95% CI 1.47 to 4.32, P < 0.001
⋅ HDL at or above 1.034 mmol/L (40 mg/dl): IRR 0.57, 95% CI 0.42 to 0.78, P < 0.001
⋅ Triglycerides at or above 1.693 mmol/L (150 mg/dl): IRR 1.19, 95% CI 1.10 to 1.28, P < 0.001
Compared with the Framingham algorithm, the D:A:D model provided a slightly closer fit with actual diabetes diagnoses in the validation set (AROC curves of 0.77 for Framingham and 0.81 for D:AD). Plotting incidence of new-onset diabetes by risk level for both the D:A:D and Framingham equations, the investigators found that their model outdid Framingham in predicting diabetes in low-risk patients and also in patients at the highest risk.
The Framingham equation tended to overestimate diabetes incidence in people under 40 years old and to underestimate incidence in people over 50. The D:A:D formula overestimated diabetes incidence in people over 50 but fit closely with actual diagnoses in the under-40 group and the 40-to-49 group. Framingham overestimated incidence in people with a BMI under 25 and underestimated incidence in those with a BMI between 25 and 30. The D:A:D equation closely predicted actual diagnoses in these groups. Both models slightly overestimated diabetes incidence in people with a BMI at or above 30.
1. Petoumenos K, Fontas E, Worm SW, et al. Predicting the short-term risk of diabetes in HIV-infected patients in the D:A:D cohort: the D:A:D study group. 9th International Congress on Drug Therapy in HIV Infection, November 9-13, 2008, Glasgow. Abstract O314.