BACKGROUND: With integrase strand transfer inhibitor (INSTI) use associated with increased body mass index (BMI) and BMI increases associated with higher diabetes mellitus (DM) risk, we explored the relationships between INSTI/non-INSTI regimens, BMI changes, and DM risk. METHODS: RESPOND participants were included if they had CD4, human immunodeficiency virus (HIV) RNA, and ≥2 BMI measurements during follow-up. Those with prior DM were excluded. DM was defined as a random blood glucose ≥11.1 mmol/L, hemoglobin A1c ≥6.5%/48 mmol/mol, use of antidiabetic medication, or site-reported clinical diagnosis. Poisson regression was used to assess the association between natural log (ln) of time-updated BMI and current INSTI/non-INSTI and their interactions on DM risk. RESULTS: Among 20 865 people with HIV included, most were male (74%) and White (73%). Baseline median age was 45 years (interquartile range [IQR], 37-52), with a median BMI of 24 kg/m2 (IQR, 22-26). There were 785 DM diagnoses with a crude rate of 0.73 (95% confidence interval [CI], .68-.78)/100 person-years of follow-up. ln(BMI) was strongly associated with DM (adjusted incidence rate ratio [aIRR], 16.54 per log increase
95% CI, 11.33-24.13
P <
.001). Current INSTI use was associated with increased DM risk (IRR, 1.58
95% CI, 1.37-1.82
P <
.001) in univariate analyses and only partially attenuated when adjusted for variables including ln(BMI) (aIRR, 1.48
95% CI, 1.29-1.71
P <
.001). There were no interactions between ln(BMI), INSTI, and non-INSTI use and DM (P = .130). CONCLUSIONS: In RESPOND, compared with non-INSTIs, current use of INSTIs was associated with an increased DM risk, which partially attenuated when adjusted for BMI changes and other variables.