The association between abacavir (ABC) use and myocardial infarction (MI) has been heavily debated since the Data Collection on Adverse Events of Anti-HIV Drugs (DAD) Study Group reported an increased risk of MI in patients with current or recent ABC use [1]. Data from other cohort studies have provided conflicting results [2–7], whereas analyses from clinical trials generally do not lend support to a causal relationship between ABC and MI [8–11]. In brief, the association between ABC and risk of MI was also found in a Danish cohort study [4] and in the SMART study dataset [7]. On the other hand data from highly active anti-retroviral therapy (HAART) naïve patients in pooled GSK-sponsored trials [6], the AIDS Clinical Trial Group (ACTG) A5001 [2], the Veterans Health Administration's Clinical Case Registry [3], and a case-control study from The French Hospital Database on HIV [5] did not find a significant association between ABC and MI.

In stronger evidence level data from several randomized controlled trials (RCTs) comparing ABC with tenofovir disoproxil fumarate (TDF), no increase in rate of MI was detected among patients randomized to ABC-based compared to TDF-based therapy [9–11]. Only in the STEAL-study, in which patients assigned to the ABC arm had a higher prevalence of cardiovascular risk factors, was an increased rate of cardiovascular events with ABC use detected [8]. Most recently, in a trial level meta-analysis of RCTs in which ABC was randomized as part of a HAART regimen, no association between increased risk of MI and ABC was found [12].

However, of the RCTs mentioned above, only two have comprised patients with suppressed HIV-infection [8, 9]. A feature of uncontrolled HIV replication is elevated levels of inflammatory markers that are associated with an increased risk of cardiovascular events [13–15], but initiation of HAART in treatment-naïve patients reduces the levels of these biomarkers [11]. Thus, some of the discrepancy between studies may have been caused by the confounding noise of uncontrolled HIV replication among HAART-naïve patients. Investigations of inflammatory biomarkers have not provided consistent evidence for increased coagulation, inflammation, or endothelial dysfunction during ABC treatment [7, 11, 16–20], but most studies have been limited by a lack of randomization and HLAB5701-screening, or included patients with uncontrolled HIV replication.

The Efficacy and Safety of Switching from Zidovudine to Tenofovir or Abacavir in HIV-infected Patients (SWAP) study was a randomized trial comparing switching from zidovudine (AZT) to ABC or TDF with parameters of renal function and bone metabolism as the pre-specified primary endpoints. However, the study provided an opportunity to perform a post hoc analysis to investigate biomarkers of cardiovascular risk as early as 4 weeks after treatment switch in patients with suppressed HIV-infection who were randomized to ABC or TDF treatment. We report here on the findings from the analyses of these cardiovascular biomarkers.

In this sub-study of the SWAP randomized clinical trial, the introduction of fixed-dose treatment with ABC/3TC lead to increases in plasma levels of E-selectin and sVCAM-1 compared with fixed-dose treatment with TDF/FTC, but the effect was short term and could not be detected after 12 and 48 weeks of follow up. No significant changes in the levels of sICAM-1, MPO, hs-CRP, or d-dimer were found.

As observed in previous randomized trials [8–10], we detected an increase in the level of total cholesterol and HDL in the ABC/3TC arm relative to the TDF/FTC arm, but there was no difference between total cholesterol/HDL ratios. It is therefore likely that the observed movement in fasting lipids will not significantly affect the 10-year cardiovascular risk, in line with what was seen in the STEAL study [8].

The strength of this study is its randomized design and that no further selection of participants was necessary, eliminating major sources of bias and confounding. Furthermore, the entry criteria required that patients be negative for the HLAB5701 allele, had not previously been exposed to ABC or TDF, and had been virologically suppressed for at least 3 months prior to randomization.

Other longitudinal studies did not find consistent patterns of increasing levels of cardiovascular biomarkers with ABC treatment. In a post hoc exploratory analysis of the HEAT study, changes in levels of CRP, sVCAM-1, and sICAM-1 were not significantly different among patients randomized to ABC/3TC or TDF/FTC [11]. It is noteworthy that participants in the HEAT study were HAART naïve at randomization and may not be comparable to the virologically suppressed patients in our study owing to the effect of uncontrolled HIV-replication on biomarker levels. The same applies for a recently published non-randomized study among viremic patients initiating abacavir or tenofovir, where no difference in changes of cardiovascular biomarkers, including early 4-weeks measurements, between treatment groups was found [24]. Similarly, a sub-study of the BIOCOMBO trial did not detect differences in markers of inflammation, endothelial dysfunction, hypercoagulability, or insulin resistance after treatment was switched to an ABC/3TC- or TDF/FTC-based regimen [18]. The STEAL study is particularly interesting because patients who switched to treatment with ABC/3TC, as compared to TDF/FTC, had a higher rate of cardiovascular events during follow up, but also a higher proportion of smokers and higher Framingham scores at baseline [8]. Nevertheless, an exploratory analysis did not show any consistent association between ABC/3TC and circulating levels of cardiovascular or inflammatory biomarkers, although a trend towards higher ICAM-1 levels in the ABC/3TC arm was observed [20].

ICAM-1 is the endothelial ligand for macrophage-1 antigen (Mac-1) that is expressed on the surface of leukocytes and mononuclear cells. Accumulation of these cells in the vessel wall is a hallmark of atherosclerosis and acute coronary syndrome, and is mediated by the interaction between adhesion molecules on endothelial and circulating cells. In an experimental system, Pablo et al. recently demonstrated that ABC but not AZT or 3TC induced significant and dose-dependent increases in rolling flux and adhesion in the leukocyte-endothelium interaction [25]; these effects were blocked by antibodies against ICAM-1 or Mac-1. Furthermore, treatment of neutrophils and monocytes with ABC increased the expression of Mac-1, whereas no increases in ICAM-1, VCAM-1, or E-selectin were seen following the treatment of endothelial cells [25]. Altogether, these results suggest that ABC treatment could be linked to increased expression of Mac-1 rather than endothelial adhesion molecules, which would explain the paucity of positive results in studies of the latter.

We found a transient effect of abacavir treatment on some, but not all of the investigated cardiovascular biomarkers. The main contribution of our analysis is the early evaluation of these at week 4; in other randomized trials such assessment was not done until week 12 [20] or 48 [11, 18]. However, whether a temporary increase in E-selectin and sVCAM-1 may translate into an increased risk of myocardial infarction within such a short time frame seems unlikely, and the clinical significance of this finding, if any, is uncertain. In addition, for both E-selectin and sVCAM-1, baseline levels were higher in the TDF/FTC arm than in the ABC/3TC arm, though only significantly so for E-selectin. Hence, we cannot rule out that the statistical phenomenon of regression towards the mean is contributing to the relative increase of these two biomarkers in the ABC/3TC arm compared with the TDF/FTC arm from baseline to week 4.

Both E-selectin and sVCAM-1 are associated with atherosclerosis, risk of cardiovascular events, and risk of death in patients with known cardiovascular disease [26–29] even though these associations have not yet been proven in HIV-infected patients. The association of ABC use and risk of MI in the DAD study data was no longer present 6 months after stopping ABC [1], but the temporal behaviour of the risk within this period is unknown. Our data suggest that the effect on vascular inflammation is fast acting, short-lived, and that no continuous pro-inflammatory effect is present. This could be relevant when considering switching treatment from ABC in patients with low cardiovascular risk who have been on ABC for several years. Such patients may not benefit from a therapy change with respect to cardiovascular risk.

Hs-CRP is a well-established marker of risk prediction of coronary events [14, 15, 23] and contains prognostic information with a cut-off as low as 1 mg/L. In a post hoc analysis, we tested for differences in the proportion of patients with hs-CRP above this limit. However, even though the proportion of patients with hs-CRP > 1 mg/L, as well as overall levels of hs-CRP, seemed to decrease with TDF/FTC treatment as compared with ABC/3TC treatment, the differences were not significant. Because of the low prevalence of detectable plasma HIV-RNA in our samples, any decrease in inflammatory markers is unlikely to be caused by improved virological control. We have recently shown that TDF down-regulates production of pro-inflammatory cytokines following stimulation of monocytes with lipopolysaccheride [30], and we speculate that TDF may have inherent anti-inflammatory properties that affect both upstream and downstream inflammatory markers to a modest degree.

Our study is limited by the small number of patients, which reduces the statistical power to detect differences between groups. Furthermore, the large number of comparisons made increases the likelihood that significant findings could have arisen by chance alone. However, the observed transient increases in biologically related endothelial biomarkers, including a tendency towards a similar transient increase in sICAM-1, speak against this possibility. All 5 participants who were excluded from the study (2 after baseline and 3 after week 12 measurements) were in the ABC/3TC arm, introducing a potential source of bias. However, in only one of these patients, who had pre-existing diabetes, was the reason for exclusion related to metabolic or cardiovascular causes. This was a sub-study that was not planned in the original protocol, but collection of plasma samples and lipid measurements were already established, and further selection of patients was not necessary. In the case of IL-6 measurements, 55% of samples had undetectable levels and were assigned the value of the lower detection limit to allow for statistical analyses; these assigned values may not reflect the true IL-6 concentration.

In conclusion, levels of total cholesterol and high-density lipoprotein (HDL) increased relatively during treatment with ABC/3TC compared with TDF/FTC, but no difference was found in total cholesterol/HDL ratio. We observed transient increases in the plasma levels of E-selectin and sVCAM-1 in patients randomized to ABC/3TC compared with TDF/FTC, but no difference was found in other biomarkers associated with endothelial dysfunction, inflammation, or coagulation. The clinical significance of these findings is uncertain.