Do Plentiful Diabetes Risk Factors Include HIV?

Is HIV Itself a Diabetes Risk Factor?

Worthy statisticians who tried to disentwine HIV from other variables as a diabetes risk factor have reached two conclusions on whether HIV alone can lead to diabetes: Yes, it can. And no, it can't. One problem in any such analysis, no matter how large or refined, is residual confounding -- the unassailable possibility that factors not melded into the covariate mix are swaying results one way or the other. And HIV populations typically shoulder a fugal array of diabetes risk factors. So it may be impossible to say whether HIV infection -- independently of other risk factors common to people with HIV -- propels them toward a diabetes diagnosis.

Certainly one can imagine a rationale for why HIV, by itself, makes diabetes more likely. NIH metabolic mavens Colleen Hadigan and Sarah Kattakuzhy note that HIV-induced inflammation riles chemokines that regulate insulin and so could explain why HIV abets diabetes development.² As an example, they cite an AIDS Clinical Trials Group (ACTG) study that linked systemic inflammation after antiretroviral therapy (ART) began to new-onset diabetes.¹⁶ This case-control study compared 55 people diagnosed with diabetes within 48 weeks of starting their first antiretroviral regimen and 55 controls matched for baseline BMI and race/ethnicity who remained free of diabetes. After adjustment for 48-week BMI and CD4 count, baseline marker levels, and indinavir use, people with higher initial levels of inflammation signals hsCRP, sTNFR1, and sTNFR2 had higher odds of a diabetes diagnosis. After further adjustment for week-48 glucose, higher sTNFR1 remained an independent predictor of diabetes (highest versus lowest quartile aOR 23.2, 95% CI 1.28 to 423, P = 0.03).

A more recent analysis of 3695 antiretroviral-treated people in the SMART and ESPRIT studies buttressed these findings, discerning links between higher inflammatory marker levels and newly diagnosed diabetes.¹⁷ During an average 4.6 years of follow-up, diabetes developed in 137 people. A regression model adjusted for baseline diabetes- and HIV-related factors determined that every doubling of IL-6 or hsCRP, two inflammation markers, nudged up the risk of incident diabetes (aHR 1.29, 95% CI 1.08 to 1.55, P = 0.005; aHR 1.22, 95% CI 1.10 to 1.36, P < 0.001).

But a study of 214 antiretroviral-naive people in France found no significant links between these same inflammation markers and insulin resistance in an analysis adjusted for age, sex, geographic origin, BMI, and waist circumference.¹⁸ This French analysis differs from the ACTG and SMART/ESPRIT studies in focusing on antiretroviral-naive people and in choosing a softer endpoint -- insulin resistance rather than new-onset diabetes. But one would prefer concordant results in all three studies to nail down the HIV→inflammation →diabetes hypothesis.

And if you look at studies weighing the impact of HIV as an independent diabetes risk factor, results are all over the map: Some find HIV does boost diabetes risk,^10,19 others find it does not,^20-22 and others find a lower diabetes risk in antiretroviral-treated people with HIV than in comparison groups.^19,21,23 Two studies reached two of these three conclusions, depending on the study period and antiretroviral use.^19,21

Two big cohort studies -- one in Italy and one in Denmark -- tagged HIV infection as an independent diabetes risk factor in comparisons with general-population groups. A 2008 Italian study compared 4249 HIV-positive people attending a Milan infectious diseases clinic and 9148 healthy controls from 15 Italian regions (Table 1).¹⁰ About three quarters of both groups were men, median age was significantly younger in the HIV group (45.7 versus 46.6, P < 0.0001), and BMI was significantly lower (23.2 versus 25.3 kg/m², P < 0.0001). Median triglycerides were higher in the HIV group (126 versus 100 mg/dL, P < 0.0001). Diabetes prevalence proved significantly greater in the HIV group (4.1% versus 2.5%, P < 0.0001). After statistical adjustment for age, gender, BMI, triglycerides, and total, HDL, and low-density lipoprotein (LDL) cholesterol, chances of prevalent diabetes were 55% higher in people with HIV (aOR 1.55, 95% CI 1.02 to 2.31, P = 0.035).