Abstract Body

While the use of older protease inhibitors (PIs) including indinavir (incidence rate ratio (IRR) 1.47/5 years) and lopinavir boosted with ritonavir (/r) (IRR 1.54/5 years) has been associated with excess risk of cardiovascular disease (CVD) it is unknown whether this also applies to contemporarily used PIs.

D:A:D study participants under follow-up (FU) after 01.01.2009 were followed to the earliest of CVD, last visit plus 6 months or 01.02.2016. CVD was defined as centrally validated myocardial infarction, stroke, sudden cardiac death or invasive cardiovascular procedures (coronary bypass, coronary angioplasty and carotid endarterectomy). Poisson regression models assessed associations between CVD and use of two contemporarily and frequently used PIs atazanavir (ATV/r) and darunavir (DRV/r).

Of 35,711 included persons, 1,157 developed CVD (IR 5.3/1000 PYFU [95%CI 5.0-5.6]) during 7.0 (IQR 6.3-7.1) years median FU. The crude CVD IR increased gradually from 4.91/1000 PYFU [4.59-5.23] in those never exposed to DRV/r to 13.67/1000 PYFU [8.51-18.82] in those exposed >6 years, while the changes with ATV/r were less pronounced (from 5.03 [4.69-5.37] to 6.68/1000 PYFU [5.02-8.35]), (figure). After adjustment for factors not considered on the causal pathway between PI/r use and CVD, cumulative exposure to DRV/r, but not ATV/r, was associated with excess CVD risk (IRR 1.59 [1.33-1.91] and 1.03 [0.90-1.18]/5 years respectively). Additional adjustment for time-updated dyslipidemia, and other factors potentially on the causal pathway to CVD, did not affect the association (DRV/r 1.53 [1.28-1.84] and ATV/r 1.01 [0.88-1.16]/5 years). The associations remained consistent for; myocardial infarction and stroke separately; after adjustment for bilirubin levels (associated with ATV/r use and potentially protective of CVD); when stratifying for whether DRV/r was used as the first ever PI/r containing regimen or not (p=0.29 for interaction); whether DRV/r was used with a non-nucleoside reverse transcriptase inhibitor or not (p=0.43 for interaction); and in those at high vs. low estimated 5 year CVD risk (p=0.12 for interaction).

In this large heterogeneous cohort of HIV-positive persons, cumulative use of DRV/r, but not ATV/r, was independently associated with a small, but gradually increasing CVD risk. Causal inference is limited by the observational nature of our data, but the findings call for further investigations into possible underlying mechanisms.