Possible higher diabetes risk with raltegravir or PIs than NNRTIs
Starting antiretroviral therapy (ART) with an integrase inhibitor (INSTI) or protease inhibitor (PI) may confer a higher risk of diabetes than starting with a nonnucleoside (NNRTI), according to an analysis of 21,516 treated people in a North American cohort (NA-ACCORD). 
Heightened diabetes risk with INSTIs proved significant for raltegravir, but not for dolutegravir or elvitegravir.
Some (but not all) research suggests greater weight gains with INSTIs than with regimens based on NNRTIs or PIs, but the clinical import of these gains remains unclear. To address this question, NA-ACCORD investigators compared diabetes incidence (the new-diagnosis rate) in antiretroviral-treated people during the past decade.
Researchers selected antiretroviral-naive adults starting an INSTI, an NNRTI, or a PI in their cohort from January 2007 to December 2016. The outcome of interest was incident diabetes, defined as HbA1c above 6.5%, diabetes-specific medication, or a diabetes diagnosis plus diabetes-related medication. Follow-up ran from the day ART began until incident diabetes, virologic failure, regimen core switch, death, loss to follow-up, or the closing date of December 2016.
The 21,516 study participants included 10,553 (49%) starting an NNRTI, 6677 (31%) starting a PI, and 4286 (20%) starting an INSTI. Overall, 13% of participants were female, 39% men who have sex with men, and 58% nonwhite. The NNRTI, PI, and INSTI groups differed by age (median 42, 41, and 38), initial CD4 count (median 313, 262, and 360), first year of ART (2010, 2010, and 2014), and follow-up time (3.04, 2.35, and 1.64 years). Among people starting an INSTI, the raltegravir, dolutegravir, and elvitegravir groups differed in median age (44, 37, and 35), first year of ART (2011, 2015, and 2014), and follow-up time (2.64, 1.02, and 1.54 years).
To explore incident diabetes risk by antiretroviral class, the researchers used Cox proportional hazards regression adjusted for age, sex, race/ethnicity, HIV transmission risk, calendar year starting ART, and baseline weight, CD4 count, and viral load.
Diabetes incidence per 1000 person-years measured 9.24 with NNRTIs, 12.03 with INSTIs, and 11.82 with PIs. Adjusted hazard ratios (aHR) indicated about a 25% higher diabetes risk with INSTIs or PIs than with NNRTIs. Adjusted risk stopped short of statistical significant with INSTIs (aHR 1.22, 95% confidence interval [CI] 0.95 to 1.57) but not with PIs (aHR 1.25, 95% CI 1.05 to 1.49).
A separate analysis compared incident diabetes risk with raltegravir, dolutegravir, elvitegravir, or PIs versus NNRTIs. Incidence per 1000 person-years measured 16.22 with raltegravir, 15.57 with dolutegravir, 10.73 with elvitegravir, 13.74 with PIs, and 9.72 with NNRTIs. Starting a PI remained significantly associated with greater diabetes risk than starting with an NNRTI (aHR 1.22). Among individual INSTIs, starting raltegravir was independently associated with greater diabetes risk (aHR 1.50), but starting dolutegravir (aHR 1.14) or elvitegravir (aHR 0.96) was not.
The NA-ACCORD team proposed that starting ART with an INSTI or a PI versus an NNRTI “may confer increased risk of incident diabetes mellitus, though risk is heterogeneous among INSTIs.” Further analysis exploring direct versus indirect effects of antiretroviral classes or drugs on incident diabetes, they suggested, may help determine “if elevated risk observed in these analyses is attributable partially to weight gain.”
- Rebeiro PF, Jenkins C, Bian A, et al. The effect of initiating integrase inhibitor-based vs. non-nucleoside reverse transcriptase inhibitor-based antiretroviral therapy on progression to diabetes among North American persons in HIV care. IDWeek, October 2-6, 2019, Washington, DC. Abstract LB9.
http://natap.org/2019/IDWeek/IDWeek_13.htm (natap report)