A preferred reporting items for systematic reviews and meta-analyses flow diagram showing the screening process and reasons for exclusion is presented in Fig. 1. A total of 24 studies were identified using PubMed, 77 were identified using Embase, and five were identified from hand searching the reference lists of identified articles. The major reason for excluding articles during the first screening round was that they were not a clinical or cohort study; during the second round of screening, not reporting an outcome of interest was the main reason for article exclusion. In total, ten studies met all criteria and were included in the qualitative analysis (Fig. 1). Among the included studies, the length of follow-up ranged from 1.5 to 4.8 years, and the studies varied greatly with respect to the ATV-containing and comparator regimens used (Tables 1, 2, 3).
Quality assessment of the included studies is shown in detail in the Supplementary Material. The included randomized-controlled trials were of moderate quality, mainly because they were either fully open-label studies [14, 23–25] or were only partially blinded to one component of the AR regimen  or to observers . In addition, one trial showed a significant elevation in triglycerides at baseline in the ATV arm  and two trials showed an imbalance in discontinuation rates between the ATV and comparator arms [14, 24]. The included cohort studies were of moderate quality, mainly due to the lack of information on patients lost to follow-up, on the representativeness of the study populations and on-treatment compliance.
Of the ten articles included, six reported CVD outcomes [24–26, 28–30], two reported data on atherosclerosis as assessed by the surrogate marker cIMT [14, 31], and two reported outcomes related to endothelial function [23, 27]. Data were insufficient and outcomes were too varied to conduct a quantitative meta-analysis or to make systematic comparisons between ATV and other ARTs. Therefore, data were compiled qualitatively.
Effect of ATV on CVD Outcomes
Of the studies reporting CV outcomes, three were cohort studies and three were randomized-controlled trials (Table 1). Four of the studies included only treatment-naïve patients [24–26, 28], one included both treatment-experienced and -naïve patients , and one included only treatment-experienced patients . The results of these studies are reported in Table 1.
The studies reporting the incidence of MI in HIV-infected patients showed that ATV was not associated with an increased risk of acute MI. In a cohort of >16,000 treatment-naïve patients, patients initially treated with both boosted and unboosted ATV (n = 543; inverse probability weighted hazard ratio: 1.12; 95% CI 0.35–3.62) or LPV/r (n = 654; 0.92; 95% CI 0.26–3.22) did not have an increased rate of MI compared with the initial treatment with a non-nucleoside reverse-transcriptase inhibitor, whereas patients initially treated with ABC (n = 611; 2.05; 95% CI 0.72–5.86) trended towards an increased rate of MI compared with tenofovir disoproxil fumarate (TDF) . Data from the treatment-naïve and -experienced patients included in the D:A:D (Data Collection on Adverse Events of Anti-HIV Drugs) study showed no relationship between ATV exposure (37,005 patient years) and an increased MI risk among HIV-infected patients compared with other AR agents, either overall (relative rate/year 0.95; 95% CI 0.86–1.04) or when ATV was given with or without ritonavir (0.99; 95% CI 0.90–1.08 and 0.80; 95% CI 0.61–1.03, respectively) . In the Durand et al. study, a cohort and nested control analysis of a public health insurance database included 7053 HIV-positive patients, and treatment-experienced patients exposed to ATV with and without ritonavir (number exposed to ATV unknown) did not have an increased risk of MI compared with controls without ATV exposure (p = 0.96; Table 1) . In this analysis, HIV-infected patients exposed to ABC, LPV, or ritonavir had an increased risk of acute MI (odds ratios 1.79, 1.98, and 2.29, respectively; p < 0.01 vs. matched controls) .
Other CV endpoints were similarly unaffected by treatment with ATV. The D:A:D study examined the risk of stroke among HIV-infected patients exposed to ATV and found no association between ATV and an increased stroke risk (overall 0.95; 95% CI 0.87–1.05; boosted 1.02; 95% CI 0.98–1.06; unboosted 0.80; 95% CI 0.61–1.03)  (Table 1).
Of the three randomized-controlled trials reporting a variety of CV endpoints (a composite of coronary artery disease, infarct, ischemia, angina, transient ischemic attack, cerebrovascular accident, or peripheral vascular disease  or unspecified CV adverse events [24, 25]), there was no increase in CV events among HIV-infected patients receiving ATV (n = 1141). In the ACTG study A5202, there was a low incidence of CV events overall in patients receiving ABC/lamivudine or TDF/emtricitabine (FTC) plus ATV/r or efavirenz, with no difference between treatment groups . In the studies reporting unspecified CV adverse events, there was one occurrence each of MI and cardiac failure among patients receiving ATV/r and one occurrence each of angina pectoris and myopericarditis among patients receiving nevirapine through 4 years of follow-up (Table 1) , and in the randomized open-label ARIES study (ClinicalTrials.gov identifier: NCT00440947), one occurrence of cerebrovascular accident and one of peripheral vascular disease in treatment-naïve patients treated with ATV with or without ritonavir in combination with ABC/lamivudine .
Effect of ATV on Markers of Atherosclerosis
Two studies investigated the effect of ATV on atherosclerosis using changes in cIMT. One was a prospective, randomized-controlled ACTG substudy (A5260 s) in treatment-naïve patients, while the other was a single-center, prospective matched-control cohort study in treatment-experienced patients (Table 2) [14, 31].
Overall, the results of the cIMT measurements showed that treatment with ATV was associated with beneficial effects on cIMT progression compared with non-ATV-based regimens (Table 2). In the ACTG 5260 s study, patients received either TDF/FTC plus either ATV/r (n = 109), DRV/r (n = 113), or RAL (n = 106) for 144 weeks. At study end, increases in cIMT at the common carotid artery (primary intent-to-treat analysis) had progressed more slowly in patients receiving ATV/r than in those receiving DRV/r or RAL; however, differences reached statistical significance in the comparison with DRV/r (p = 0.013) but not with RAL (p = 0.15) . In the on-treatment analysis and at the carotid bifurcation, ATV/r was associated with significantly slower cIMT progression compared with both DRV and RAL . In the cohort study, patients receiving regimens, including ATV/r (n = 33), had significantly decreased cIMT compared with matched controls not receiving ATV/r (n = 99) after 12 and 18 months of therapy (p ≤ 0.05 for both) . The differences in cIMT progression remained significant after adjustment for differences in baseline demographics.
Effect of ATV on Endothelial Function
Two randomized-controlled trials in treatment-experienced patients with HIV assessed endothelial function via flow-mediated dilation (FMD) of the brachial artery [23, 27] (Table 3). There was no impact of ATV therapy on endothelial function in these studies (Table 3). Both the randomized-controlled trials investigated switching from another PI to ATV (with or without ritonavir) with a trial duration of 24 weeks. In patients receiving unboosted ATV (n = 20), there was no change in FMD compared with the previous PI therapy (p = 0.40) , and patients receiving ATV/r (n = 26) did not demonstrate any significant change in FMD both from baseline and versus patients continuing on the previous PI/r therapy (p = 0.64 after adjustment for brachial artery diameter changes) . Endothelial function was also measured in the ACTG A5260 s study. FMD did not change significantly among patients treated with ATV/r, RAL, or DRV/r. In addition, despite the differences seen in cIMT progression, there were no differences in FMD change between treatment arms .