icon-    folder.gif   Conference Reports for NATAP  
Back grey_arrow_rt.gif
Effects of Contemporary ART Regimens on Inflammatory Markers
  Download the PDF here

Certain markers decreased after ART initiation, regardless of regimen, whereas changes in other markers varied by regimen.


HIV infection has been linked to an increased risk for cardiovascular disease (CVD), possibly mediated by increased inflammation, which may persist even after the initiation of effective antiretroviral therapy (ART). Increased CVD risk has also been linked to certain antiretrovirals, including some protease inhibitors and abacavir, although no consensus has been reached on the latter.

In a substudy of ACTG 5202, researchers evaluated changes in inflammatory and endothelial-activation markers following ART initiation among 244 patients who were randomized to receive tenofovir/FTC or abacavir/3TC plus either efavirenz or ritonavir-boosted atazanavir.

The key findings:

· From baseline to weeks 24 and 96, all treatment groups experienced similar, significant decreases in the soluble receptors of tumor necrosis factor (TNF)-α and in adhesion molecules.

· At weeks 24 and 96, abacavir/3TC was associated with a significant increase in high-sensitivity C-reactive protein (hsCRP) levels, particularly when paired with efavirenz. No significant change in hsCRP was seen with tenofovir/FTC or with boosted atazanavir.

· At week 24, tenofovir/FTC, but not abacavir/3TC, was associated with a significant decrease in interleukin (IL)-6 levels. By week 96, however, significant IL-6 decreases were seen in all treatment groups, with no differences according to nucleoside pair or third drug.

Comment: The observed decrease in TNF receptors and adhesion molecules after ART initiation has been seen in other studies and likely reflects a generally positive effect of ART on inflammation. The increase in hsCRP seen with abacavir/3TC + efavirenz is noteworthy, although this regimen is not commonly used and the small size of this substudy makes the finding difficult to interpret. In contrast to the parent ACTG 5202 study, this substudy showed higher rates of virologic suppression with abacavir/3TC than with tenofovir/FTC among patients who had baseline viral loads 100,000 copies/mL (64% vs. 55%). The only two participants in the substudy who had a myocardial infarction were both receiving tenofovir/FTC + efavirenz, underscoring the challenge of interpreting biomarker levels in modest-size studies. Even with data from much larger, longer studies, biomarkers are still not used to manage CVD risk in the general HIV-negative population.

- Keith Henry, MD


Inflammation markers after randomization to abacavir/lamivudine or tenofovir/emtricitabine with efavirenz or atazanavir/ritonavir: ACTG A5224 s, A5202 substudy

McComsey, Grace A.; Kitch, Douglas; Daar, Eric S.; Tierney, Camlin; Jahed, Nasreen C; Melbourne, Kathleen; Ha, Belinda; Brown, Todd T.; Bloom, Anthony; Fedarko, Neal; Sax, Paul E.


Background: The effect of specific antiretrovirals on inflammation is unclear.

Methods: A5224 s was a substudy of A5202, which randomized HIV-infected treatment-naive subjects to blinded abacavir/lamivudine (ABC/3TC) or tenofovir/emtricitabine (TDF/FTC) with open-label efavirenz (EFV) or atazanavir/ritonavir (ATV/r) in a factorial design. Our analysis compared changes in inflammation markers from baseline to week 24 between ABC/3TC and TDF/FTC. Secondary analyses included changes at week 96 and comparisons of EFV vs. ATV/r.

Results: Analyses included 244 subjects (85% male, 48% white non-Hispanic), median age 39 years, HIV-1 RNA 4.6 log10 copies/mL, CD4 240 cells/[mu]L. TNF-[alpha], sTNFR-I and -II, sVCAM-1 and sICAM-1 decreased significantly at weeks 24 and 96, without significant differences between components (p >= 0.44). At week 24, ABC/3TC had a greater hsCRP mean fold change than TDF/FTC (1.43 vs. 0.88, estimated mean fold change percent difference ([DELTA]) 61.5% [95% CI 13.6%, 129.5%]; p = 0.008). Similar results were seen at week 96 (p = 0.021). At week 24 (but not 96), EFV had a greater hsCRP mean fold change than ATV/r (1.41 vs. 0.88; [DELTA] = 60.2% [12.6%, 127.7%]; p = 0.009). IL-6 decreased significantly at week 24 with TDF/FTC but not with ABC/3TC (between-components p = 0.019). At week 96, IL-6 decreased significantly in both NRTI components (between-components p = 0.11). IL-6 changes were not significantly different between ATV/r and EFV at either time point (p >= 0.89).

Conclusions: Soluble TNF-receptors and adhesion molecules decreased following treatment initiation and did not differ by regimens. Differences were seen on hsCRP and IL-6 changes with ABC/3TC vs. TDF/FTC and on hsCRP with EFV vs. ATV/r.


HIV-infected patients experience high rates of non-AIDS complications, including cardiovascular disease (CVD), which has been linked to heightened inflammation.With effective antiretroviral therapy (ART), inflammation markers overall decrease but do not normalize [1-4]. Treatment with abacavir (ABC)-containing regimens has been associated with higher risk of CVD in some studies [5-10] but not others [11-14].

Several attempts have been made to understand this potential association, with one proposed mechanism a deleterious effect of ABC on inflammation. While one observational study found an association between ABC and higher inflammation markers [5], another did not [15]. Prospective randomized studies comparing ABC to non-ABC regimens also have not demonstrated a difference in these markers [16-19]. Because of the remaining uncertainty about the differential effect of ABC versus other nucleoside reverse transcriptase inhibitors (NRTIs) on inflammation, and the potential effect of the concomitant protease inhibitor (PI) or nonnucleoside reverse transcriptase inhibitor (NNRTI) therapy, we sought to compare changes in inflammation markers in the context of a large randomized trial.

At week 24, 171 (70%) had HIV-1 RNA<50 copies/mL [70% on TDF/FTC; 70% on ABC/3TC]. Among subjects who had screening HIV-1 RNA

100,000 copies/mL, 55% on TDF/FTC and 64% on ABC/3TC had HIV-1 RNA <50 copies/mL. Two subjects experienced a myocardial infarction during the study; one at week 15 and the other at week 164. Both subjects had screening HIV-1 RNA<100,000 copies/mL and were receiving their randomized regimen of TDF/FTC + EFV at the time of the event.


Subject characteristics

As previously detailed [21,22], 269 subjects were randomized to one of the 4 regimens and were included in A5224 s analysis. Of these 269 subjects, 244(91%) with available stored plasma from baseline and week 24 and/or 96 were included in this biomarker substudy. Among these 244, 61 were randomized to EFV +TDF/FTC, 64 to EFV +ABC/3TC, 57 to ATV/r +TDF/FTC, and 62 to ATV/r +ABC/3TC. Baseline characteristics are summarized in Table 1. Overall, 85% were male, 48% white non-Hispanics, and among 205 with available data, 41% were smokers. Median age was 39 years, CD4 240 cells/mL, and HIV-1 RNA 4.64 log10 copies/mL. None of the subjects had a prior history of myocardial infarction, and only one had a history of stroke. Baseline characteristics were balanced across arms, except that a larger proportion of women were randomized to EFV (18%) than ATV/r (11%), and a larger proportion of hypertensive subjects were randomized to ABC/3TC (21%) than TDF/FTC (11%), and similarly, EFV (22%) than ATV/r (11%). Baseline characteristics were similar between the 244 included in the biomarker substudy and the 25 A5224 s subjects not included (data not shown).

Overall, 26 (10.7%) subjects modified their randomized NRTI (4/118 [3.4%] TDF/FTC; 22/126 [17.5%] ABC/3TC) before week 24 and an additional 39 (16.0%; 13/118 [11.0%] TDF/FTC; 26/126 [20.6%] ABC/3TC) modified between weeks 24 and 96. Also, 22 (9.0%) modified their NNRTI/PI (19/125 [15.2%] EFV; 3/119 [2.5%] ATV/r) before week 24 and an additional 40 (16.4%; 17/125 [13.6%] EFV; 23/119 [19.3%] ATV/r) between weeks 24 and 96.

At week 24, 171 (70%) had HIV-1 RNA<50 copies/mL [70% on TDF/FTC; 70% on ABC/3TC]. Among subjects who had screening HIV-1 RNA 100,000 copies/mL, 55% on TDF/FTC and 64% on ABC/3TC had HIV-1 RNA <50 copies/mL. Two subjects experienced a myocardial infarction during the study; one at week 15 and the other at week 164. Both subjects had screening HIV-1 RNA<100,000 copies/mL and were receiving their randomized regimen of TDF/FTC + EFV at the time of the event.

Changes in Plasma TNF-a and Soluble TNF Receptors

(sTNFR-II co-Primary Endpoint)

At weeks 24 and 96, there was a statistically significant decrease in TNF-a, sTNFR-I, and sTNFR-II levels within all arms(p<0.001) (Fig. 1 and Table 2), without differences between ABC/3TC and TDF/FTC or ATV/r and EFV by intent-to-treat (p0.44) or as-treated analyses.

There was some evidence that the NRTI effect differed by screening HIV-1 RNA stratum for week 24 sTNFRII (p=0.069) and TNF-a (p1/40.093), and that the NNRTI/PI effect differed for week 96 sTNFR-I (p=0.048), thus analyses were conducted within each stratum for these markers. For sTNFR-II, within the low stratum the ABC/3TC mean fold change was marginally significantly smaller than TDF/FTC at 24 weeks (0.57 vs.0.65; D1/411.4% [95% confidence intervals (CI) 22.3%, 1.2%]; p=0.073), while within the high stratum it was not significantly different (0.50 vs. 0.47; D(delta)=7.4% [8.6%, 26.1%]; p1/40.38). As-treated analyses showed similar results. For week 24 TNF-a, in both the lowstratum and the high stratum, the estimated mean fold change was not significantly different between ABC/3TC and TDF/FTC (0.61 vs. 0.68; D1/410.9% [22.6%, 2.6%]; p1/40.11 within the low stratum and 0.57 vs. 0.53; D1/46.6% [8.1%, 23.8%]; p1/40.39 within the high stratum). As-treated analyses showed similar results. For week 96 sTNFR-I, within the low stratum the ATV/r estimated mean fold change was marginally significantly larger EFV (0.95 vs. 0.89); D(delta)=6.5% [0.6%, 14.2%]; p=0.074), but similar within the high stratum (0.81 vs. 0.86; D1/45.6% [14.9%, 4.8%]; p=0.28). As-treated yielded similar results.

Changes in Endothelial Activation Markers (sVCAM-1

co-Primary Endpoint):

At weeks 24 and 96, there was a statistically significant decrease in sVCAM-1 and sICAM-1 within all arms (p
Changes in C-reactive protein

As shown in Table 2 and Fig. 2, at week 24, within the ABC/3TC?EFV arm there was a statistically significant increase in hsCRP (estimated mean fold change 1.77 [95% CI 1.24, 2.52]; p=0.002) while within the TDF/FTC+ATV/r arm, there was a marginally significant decrease (0.69 [0.47, 1.00]; p=0.051). In TDF/FTC+EFV and ABC/3TC+ATV/r arms, hsCRP did not change significantly (1.11 [0.78, 1.61]; p=0.55 and 1.13 [0.82, 1.54]; p=0.45, respectively). At 96 weeks, hsCRP increased significantly for only ABC/3TC+EFV (1.54 [1.06, 2.23]; p=0.026).

Changes by NRTI components

At week 24 and 96 weeks, in the ABC/3TC arms (3rd drugs combined) there was a statistically significant increase in hsCRP (estimated mean fold change 1.43 [95% CI 1.12, 1.83]; p=0.004 at 24 weeks and 1.36 [95% CI 1.05, 1.75]; p=0.016 at 96 weeks) while within the TDF/FTC arms there was no significant change from baseline at either time points (p>/=0.35). At week 24 and 96, the change in hsCRP was significantly larger for ABC/3TC than TDF/FTC (1.43 vs. 0.88; D=61.5% [13.6%, 129.5%]; p=0.008 at week 24 and 1.36 vs. 0.88; D=53.5% [6.9%, 120.4%]; p=0.021 at week 96). As-treated analyses showed similar results.

Changes in IL-6

At week 24, there was a statistically significant decrease in IL-6 within the TDF/FTC +EFV (estimated mean fold change 0.74 [95% CI 0.58, 0.95]; p=0.020) and TDF/FTC +ATV/r arm (0.76 [0.62, 0.92]; p1/40.005) (Fig. 2 and Table 2). For ABC/3TC +EFV and ABC/3TC +ATV/r, the estimated mean fold change was 0.97 [0.80, 1.18]; p=0.75, and 0.95 [0.77, 1.18]; p=0.67, respectively. At week 96, there was a statistically significant decrease in IL-6 within all arms (p
Changes by NRTI Components

At week 24, there was a statistically significant decrease in IL-6 within the TDF/FTC arms (combining the 3rd drug) (estimated mean fold change 0.75 [95% CI 0.64, 0.87]; p<0.001) without significant change from baseline in the ABC/3TC arms (0.96 [0.83, 1.11]; p=0.59). The change in IL-6 was significantly larger for ABC/3TC vs. TDF/FTC (D=28.5% [4.2%, 58.4%]; p=0.019). As-treated analyses showed similar results. At 96 weeks, both ABC/3TC and TDF/FTC arms (3rd drugs combined) had a statistically significant decrease in IL-6 levels (both p<0.001), and there was no difference between the NRTI components in IL-6 change by intent-to-treat (0.61 vs. 0.75; D=22.2% [4.3%, 56.1%]; p1/40.11) but there was by as-treated (0.79 vs. 0.60; D1/432.6% [1.2%, 73.6%]; p=0.040).

At week 24 (but not 96), there was evidence of an interaction between the NRTI component and HIV-1 RNA stratum (p=0.012).Within the high stratum, there was a significantly different change in IL-6 for ABC/3TC vs. TDF/FTC (1.06 vs. 0.60; D=76.3% [31.3%, 136.8%]; p<0.001), but not within the low stratum (0.91 vs. 0.88; D(delta)=2.8% [22.9%, 36.9%]; p=0.85). As-treated analyses showed similar results.

Changes by NNRTI/PI Components

At weeks 24 and 96, by intent-to-treat and as-treated analyses, IL-6 decreased significantly (p/=0.80). There was no interaction between the NNRTI/PI component and HIV-1 RNA stratum at either time point.

At week 24 (but not 96), there was evidence of an interaction between the NRTI component and HIV-1 RNA stratum (p=0.054). Within the high stratum the ABC/3TC mean fold change was larger than TDF/FTC (1.82 vs. 0.74; D(delta)=142.5% [39.8%, 330.3%]; p=0.002), while within the low stratum no statistically significant difference was detected (1.22 vs. 1.00; D(delta)=21.5% [22.6%, 90.9%]; p=0.40). As-treated analyses showed similar results.

Association between baseline factors and

changes in biomarkers

Linear regression analyses assessed the association of baseline factors with changes in biomarkers (Table 3). The covariates were the same as those used in adjusted analyses, except for baseline marker level. For CRP and IL-6, by multivariable analysis, in addition to the ABC/3TC and ATV/r effects, lower CD4 count was independently associated with increased 24-week hsCRP change while greater HIV-1 RNAwas associated with decreased IL-6. Additionally, a significant interaction between the NRTI component and HIV-1 RNA was seen for hsCRP and IL-6. Specifically, per log10 copies/mL, HIV-1 RNA was associated with a 69.5% (1.7%, 182.6%) and 36.6% (-0.4%, 87.5%) larger mean fold change in ABC/3TC for hsCRP (1.19 vs. 0.70) and IL-6 (0.96 vs. 0.70), respectively. At 96 weeks, for both hsCRP and IL-6, male sex was associated with decreases. For the subset with smoking data (n1/4205), smoking was associated with 24 week increases in hsCRP, IL-6.


This report details changes in inflammation and endothelial activation markers among subjects randomized to one of four commonly-used ART regimens. As seen in prior studies, we demonstrated that ART initiation led to a decrease in markers of TNF-a activation and endothelial activation, but not hsCRP. We also found that on average with ABC/3TC therewere short-term (24 weeks) and longer-term (96 weeks) increases in hsCRP and short term differences in IL-6 compared to TDF/FTC, and that EFV induced increases in hsCRP compared to ATV/r at week 24. Adjusting for potential confounders/imbalances, including virologic efficacy, did not change these results and as-treated analyses yielded similar results to intent-to-treat.

Cardiovascular disease recently emerged as a major cause of morbidity and mortality in HIV. In addition to traditional CVD risk factors, inflammation appears to be a major component of this enhanced risk. So far, this has been shown in mostly cross sectional studies where higher antigen-specific T-cells responses [23], CRP [4,24,25], TNF-a4 and IL-6 [26] have been associated with carotid IMT, and higher CD4?CD38?HLA-DR? T-cells% with arterial stiffness [27].

Our study supports the findings of earlier studies that initiation of effective ARTresults in an overall decrease in inflammation markers, with the exception of hsCRP, which remains unchanged or even increases [2,17,28,29]. The reason hsCRP behaves differently from other markers remains elusive, with one potential explanation being HIV-associated subclinical hepatocyte dysfunction, as CRP is mainly produced by hepatocytes in response to IL-6. Results were unchanged after adjusting for several factors known to affect CRP, including hepatitis C [30]. Oral contraceptives have been shown to increase hsCRP [31,32], but excluding the 4 women who were on oral contraceptives (all on EFV; 3 on ABC/3TC) did not modify the results. Likewise, statins were shown to decrease hsCRP [33], but excluding the 23 statin users did not affect hsCRP or IL-6 results. Regardless, hsCRP appears to be an important marker, independently associated with CVD [4,24], HIV progression [34], and mortality [35], even after adjusting for CD4 and HIV-1 RNA. However, it is unclear whether the AHA CVD risk stratification cut-offs established for HIV-uninfected populations [36] are valid in HIV. Fig. 2 (c, d) shows the baseline and week 24 AHA hsCRP risk categories, without apparent shift to higher risk categories in either group.

Only limited data exist regarding the effect of specific ARTon inflammation, and most focus on ABC, because of its potential link to increased CVD in some studies [5-9]. Others [11-14] did not find such link, including A5001 [11] that included data from A5202. Studies investigating the effect of ABC on inflammation yielded conflicting results. The INSIGHT group found higher IL-6 and hsCRP, but similar TNF-a, with ABC regimens [5], while the MACS cohort found similar biomarker levels, including IL-6 and hsCRP, in ABC- vs. non- ABC-treated subjects [15]. Moreover, studies of virologically- suppressed individuals who switched their NRTIs to ABC or TDF found no differences in inflammation markers after switching [18,19]. Furthermore, the HEAT study, which prior to the current study was the only randomized study [16] that measured inflammation markers in subjects initiating their first ART regimen with ABC- versus TDF-based therapy, found no difference in these markers between the regimens. However, HEAT investigated only lopinavir/ritonavirbased therapies, whereas our study had both EFV and ATV/r as 3rd drugs, and EFV (but not ATV/r) was found to be associated with hsCRP increase.

We found an interaction between screening HIV-1 RNA stratum and the NRTI effect so that the differences seen between ABC/3TC and TDF/FTC were larger in the high screening viral load stratum than in the lower. This observation may explain the discrepancies seen in prior biomarker studies, and specifically in prior randomized studies of virologically-suppressed subjects where no difference in markers were found between ABC- and TDF-regimens [18,19].

The mechanism by which ABC affects inflammatory pathways is not clear. In one study, ABC induced dose-dependent increases in neutrophil adhesion through the activation of Mac-1, which interacts with its endothelial ligand ICAM-1 [37]. Some have suggested that analogous to other drug toxicity [38], ABC may interfere with purine signaling pathways leading to impairment of lymphocyte activation [39]. A recent study found that ABC up-regulates pro-inflammatory cytokine mRNA transcription by stimulation of Toll-like receptor 8 signaling [40]. Others have suggested that ABC can induce a low-level hypercoagulable state by increasing platelets aggregation due to hyperactive platelets [41], or increase fibrinolytic capacity by raising PAI-1 levels [42].

Although several studies have shown that CD4 cell recovery is similar after initiation of PI vs. NNRTI regimens [43,44], only one study found similar decreases in sTNFR-I and -II [45] with PI vs. NNRTI-regimens, but hsCRP and IL-6 were not measured. Our finding of higher CRP with EFV vs. ATV/r is consistent with a study showing that in adipocytes, EFV induces the release of higher levels of cytokines than PIs [46].

Our study has several limitations, including decreased sample size at week 96, possible selection bias of A5224s subjects who have available week 24 and/or 96 samples (possibly healthier subjects with lower inflammation), and the large number of analysis performed without adjustment for multiple comparisons, which may increase the risk of a type I error. Although the co-primary endpoints (sTNFR-II and VCAM-1) were chosen because our prior studies showed them to be associated with clinically-relevant outcomes, such as incident diabetes and carotid IMT [4,25,28], other markers chosen as secondary endpoints (e.g. IL-6 and hsCRP) have also been associated with mortality and CVD [24,26].

We have shown in a randomized study a differential treatment effect on some inflammation markers, with differences in hsCRP and IL-6 in those receiving ABC compared to TDF- and EFV- compared to ATV/rcontaining regimens. Further investigations are needed to duplicate these findings and clarify their clinical significance.