Relationship between total bilirubin, endothelial function, inflammation and oxidative stress in HIV-infected adults on stable antiretroviral therapy

Abstract

Objective

Enhanced inflammation is evident in HIV, even with virologic suppression. Outside HIV, studies show an independent association between higher total bilirubin and better endothelial function as well as lower prevalence of coronary heart disease possibly due to the anti-inflammatory and antioxidant effect of bilirubin.

Methods

A cross-sectional study was performed in HIV-1 infected adults on stable antiretroviral therapy (ART) to determine if a relationship exists between total bilirubin and endothelial function (flow mediated dilation (FMD) of the brachial artery), inflammation (interleukin-6 (IL-6), soluble tumor necrosis factor receptors, C-reactive protein, adhesion molecules), coagulation markers (fibrinogen and D-Dimer) and oxidative stress (F₂-Isoprostanes). Endpoints were compared based on total bilirubin levels and atazanavir status using distributionally-appropriate, two-sample tests. Correlation coefficients were determined between total bilirubin and end points. Linear regression was used to model the relationship between total bilirubin (and atazanavir status) and FMD.

Results

98 adults were included. Total bilirubin was higher in atazanavir group (median (IQR) 1.8 (1.1–2.6) vs. 0.6 (0.4–1.4) mg/dL; p<0.01) as was insulin, HOMA-IR and fibrinogen. Total bilirubin was positively correlated with fibrinogen and was not correlated with other outcomes. After adjustment, neither total bilirubin nor atazanavir status was associated with FMD.

Conclusions

In virologically-suppressed, HIV-infected adults on stable ART, neither total bilirubin nor atazanavir use was associated with improved endothelial function as measured by FMD, inflammation or oxidative stress as measured by biomarkers.

Introduction

The important role of inflammation in atherosclerosis and atherothrombosis is increasingly recognized(1), and in HIV-infected patients, it may be the principle driver of increased risk of subclinical atherosclerosis(2) and cardiovascular events.(3) This has spurred interest in the development of anti-inflammatory therapeutics to reduce cardiovascular risk.

Bilirubin, an endogenous product of hemoglobin catabolism, has antioxidant and anti-inflammatory properties that attenuate endothelial activation and dysfunction in response to pro-inflammatory stress.(4) It has been shown to prevent oxidation of low-density lipoproteins and to inhibit vascular cell adhesion molecule-1 (sVCAM-1)-dependent migration of leukocytes into the endothelium(5). Epidemiologic studies in HIV-uninfected populations have associated elevated serum bilirubin levels with better endothelial function(6) and lower prevalence of coronary heart disease(7), stroke(8) and lower-extremity peripheral arterial disease.(9)

Gilbert’s syndrome is a common cause of mildly elevated indirect bilirubin which occurs in patients with decreased activity of hepatic bilirubin uridine diphosphate-glucuronosyltransferase (UGT1A1). A specific mutation in this gene (UGT1A1*28) has been associated with a lower risk of cardiovascular disease.(10) The protease inhibitor atazanavir (ATV) inhibits UGT1A1 activity, which results in mild hyperbilirubinemia similar to Gilbert’s syndrome. As such, ATV may have a beneficial effect on inflammation, oxidative stress and cardiovascular risk which is independent of its favorable metabolic profile.

Studies have been conflicting with regard to the effect of ATV on endothelial function. In a small, randomized, placebo-controlled trial in diabetics, three days of ATV 300 mg twice daily improved endothelial function measured by venous occlusion plethysmography.(11) However, in another small, randomized, placebo-controlled trial in healthy, adult men, four weeks of ATV 400 mg daily did not affect methacholine-induced endothelium dependent vasodilation of the femoral artery.(12) In HIV, two randomized trials that switched patients to unboosted(13) or boosted(14) ATV failed to show short term improvements in endothelial function measured by flow-mediated dilation (FMD) of the brachial artery. These two studies focused on whether improvement in lipid profiles would restore endothelial function. There is no report on the relationship between serum bilirubin and endothelial function.

The primary objective of our study was to examine the relationship between total bilirubin level and endothelial function measured by FMD of the brachial artery among ATV users and non-users. We additionally assessed the relationship between total bilirubin and markers of inflammation, coagulation, oxidative stress and lipid levels.

Methods

Study design

This was a retrospective, cross-sectional study designed to evaluate the relationship between total bilirubin levels and FMD of the brachial artery as well as markers of inflammation, coagulation and oxidative stress and lipid levels. All HIV-1 infected adults on stable antiretroviral therapy (ART) for at least 12 weeks with HIV-1 RNA <400 copies/mL who had FMD of the brachial artery performed by ultrasound as part of entry into a study through the HIV Metabolic Research Center at Case Western Reserve University were eligible for inclusion in this study. Exclusion criteria were active infection, inflammatory condition or malignancy, uncontrolled diabetes mellitus, creatinine clearance < 50 mL/min, alanine aminotransferase (ALT) or aspartate aminotransferase (AST) > 2X the upper limit of normal within 6 months, pregnancy, lactation, regular use of anti-inflammatory or anti-oxidant medication, intravenous drug use or daily alcohol use. No selection criteria regarding specific ART regimens were imposed for any of the studies. Participants evaluated for enrollment into the HIV Metabolic Research Center studies were recruited from the John T. Carey Special Immunology Unit at University Hospitals Case Medical Center in Cleveland, Ohio. All individuals signed written informed consent for participation in the HIV Metabolic Research Center trials and also to have their blood stored for use in future HIV-related metabolic research. This study was approved by the University Hospital Case Medical Center Institutional Review Board with a waiver for further informed consent.

All data collected, demographics, HIV and cardiovascular characteristics, laboratory values and stored samples, were obtained on the date FMD was performed. The primary outcome of this study was endothelial function determined by FMD of the brachial artery. Secondary outcomes of interest included markers of inflammation (interleukin-6 (IL-6), soluble tumor necrosis factor receptors-I and –II (sTNFR-I and –II), high sensitivity C-reactive protein (hs-CRP), soluble intercellular adhesion molecule-1 (sICAM-1), soluble vascular cell adhesion molecule-1 (sVCAM-1), coagulation (D-Dimer and fibrinogen), oxidative stress (F2-Isoprostanes), lipoprotein levels and insulin resistance estimated by the homeostasis model assessment of insulin resistance (HOMA-IR).

Evaluation of endothelial function

Endothelial function was evaluated by measuring FMD of the brachial artery with ultrasound(15) as previously described.(16) Participants were instructed to come fasting, to hold anti-hypertensive medications and to not use tobacco or caffeine-containing products for 12 hours before the study. All studies were performed by a single technologist (C.W.) using a Phillips iU22 Ultrasound and a L10-7 MHz linear array transducer and a 5 minute occlusion time. Images were read using Brachial Artery Analyzer software (MIA Inc.), a semi-automated, border interfacing program. For FMD determination, brachial artery diameters before and after confirmed reactive hyperemia were measured in triplicate and averaged from a 1 cm segment of the artery. FMD is expressed as a percent change from baseline brachial artery diameter to brachial artery diameter after reactive hyperemia.

Evaluation of biomarkers

Plasma from each participant was previously stored at −70°C immediately after processing. Stored samples were then batched and tested for the markers of inflammation, coagulation and oxidative stress outlined above. Interleukin-6, sTNFR-I and –II, sICAM-1 and sVCAM-1 were determined by quantitative sandwich ELISAs (R&D Systems, Minneapolis, MN). Inter-assay variability ranged from 2.02%–15.36%, 3.66%–5.77%, 2.13%–3.79%, 3.43%–7.37% and 4.76%–8.77%, respectively. High sensitivity C-reactive protein and fibrinogen were determined by particle enhanced immunonephelometric assays on a BNII nephelometer (Siemens). Inter-assay variability ranged from 3.01%–6.46% and 3.42%–7.59%, respectively. D-Dimer was determined by immuno-turbidometric assay on a STA-R Coagulation Analyzer (Diagnostica Stago). Inter-assay variability ranged from 1.54%–9.03%. All above biomarker assays were performed at the Laboratory for Clinical Biochemistry Research under the direction of Dr. Russell Tracy, Department of Pathology, University of Vermont. F₂-Isoprostanes were measured in the Eicosanoid Core Laboratory at Vanderbilt University. Briefly, F₂-Isoprostanes were quantified using gas chromatography-mass spectrometry after Sep-Pak and TLC purification as pentafluorobenzyl ester, trimethylsilyl ether derivatives utilizing stable isotope dilution techniques with [²H₄]-15-F_2t-IsoP (Cayman Chemical, Ann Arbor, MI) as an internal standard. The precision of this assay is ±4%, the accuracy ±95% and inter-assay variability is less than 8%.

Data analysis

Important demographic, HIV and cardiovascular factors are described for the group overall, by ATV status (currently taking ATV vs. not) and by total bilirubin level (≥75^th percentile vs. <75^th percentile). Median and interquartile range (IQR) are reported for continuous variables and frequency and percent for categorical variables. All demographic, HIV and cardiovascular factors, as well as endpoints are compared based on ATV status and total bilirubin level using unpaired t-tests or Wilcoxon Rank Sum tests as distributionally-appropriate for continuous variables, and by Chi-Square tests, Fisher’s Exact tests or Pearson Exact Chi-Square tests as appropriate for categorical variables.

Spearman correlation coefficients were determined between total bilirubin as a continuous variable and end points. All above statistical tests were two-sided and considered significant with p<0.05. No corrections for multiple comparisons were made in this exploratory study.

Next, in order to explore the relationship between FMD and total bilirubin in this sample, univariable followed by multivariable linear regression was performed. In the univariable analysis, all demographic, HIV and cardiovascular factors, inflammation, coagulation and oxidative stress markers as well as ATV status and total bilirubin as a dichotomized variable by ≥75^th percentile compared to <75^th percentile and a continuous variable were modeled with FMD as the outcome. In the first multivariable modeling approach, those variables with p<0.25 were included in 3 separate multivariable models with ATV status or total bilirubin as a categorical or continuous variable as the independent variable of interest. In addition, a second multivariable modeling approach including clinically-relevant variables regardless of statistical association was undertaken. Age, sex, race, BMI, CD4⁺ cell count, HIV-1 RNA level, whether on an anti-hypertensive or cholesterol-lowering medication, smoking status and brachial artery diameter were included in 3 separate models with ATV status or total bilirubin as a categorical or continuous variable as the independent variables of interest as above. In the multivariable models, variables with p<0.05 were considered statistically significant. Final models were checked to be sure that the assumptions of linear regression were met.

All analyses were performed using SAS v. 9.2 (The SAS Institute, Carey, North Carolina, USA).

Results

98 individuals met eligibility criteria. Table 1 shows demographic, cardiovascular and HIV characteristics overall, by ATV status (ATV vs. no ATV) and by total bilirubin level (≥75^th percentile vs. <75^th percentile). Comparing participants on ATV to those not, the groups were similar except for total bilirubin level, insulin and HOMA-IR. Total bilirubin was higher in the ATV group (1.8 (1.1–2.6) vs. 0.4 (0.3–0.5) mg/dL; p<0.01) as expected. Insulin level and HOMA-IR were also higher in the ATV group (10 (6–17) vs. 7 (4–14) µIU/mL; p=0.05 and 2.1 (1–4) vs. 1.4 (0.9–2.8); p=0.05, respectively). More patients in the highest quartile of total bilirubin were on PIs (96% vs. 37%; p<0.01) compared to those in the lowest three quartiles. For all other characteristics the groups were similar.

Table 1.

Demographic, Cardiovascular and HIV Characteristics of by Atazanavir Use and by Total Bilirubin Level

Variable	Overall (N=98)	Atazanavir (n=36)	No Atazanavir (n=62)	P*	Total Bilirubin ≥1.4 mg/dL (n=25)	Total Bilirubin <1.4 mg/dL (n=73)	P*
Age, years	47.5 (43–52)	47 (44–51.5)	48 (43–53)	0.7	49 (45–53)	47 (43–52)	0.44
Men	86 (88)	31 (86)	55 (89)	0.75	23 (92)	63 (86)	0.73
Caucasian African American Latino Other	41 (42) 52 (53) 4 (4) 1 (1)	13 (36) 21 (58) 2 (6) 0 (0)	28 (45) 31 (50) 2 (3) 1 (2)	0.75	10 (40) 14 (56) 1 (4) 0 (0)	31 (42) 38 (52) 3 (4) 1 (1)	0.94
Never smoked Current smoker Past smoker	30 (31) 46 (47) 22 (22)	8 (22) 19 (53) 9 (25)	22 (35) 27 (44) 13 (21)	0.42	6 (24) 11 (44) 8 (32)	24 (33) 35 (48) 14 (19)	0.38
Systolic BP, mmHg	120 (112-128)	116.5 (112-122)	120.5 (106-130)	0.3	115 (112–120)	120 (106–130)	0.33
Diastolic BP, mmHg	80 (74–86)	80 (76–86)	80 (72–85)	0.68	78 (76–85)	80 (74–86)	0.65
On anti-HTN drug	29 (30)	10 (28)	19 (31)	0.76	6 (24)	23 (32)	0.48
Total cholesterol, mg/dL	178 (151–209)	165.5 (147–190.5)	186.5 (156–210)	0.24	154 (139–214)	181 (156–208)	0.61
HDL, mg/dL	40.5 (36–51)	40 (35.5–46.5)	41.5 (36–55)	0.29	40 (36–45)	41 (35–52)	0.87
Non-HDL, mg/dL	131.5 (110-166)	123.5 (103-154)	139 (116–170)	0.36	119 (102–166)	134 (116–165)	0.57
Triglycerides, mg/dL	124.5 (84–198)	137.5 (94–211.5)	115.5 (81–198)	0.27	126 (87–182)	123 (83–202)	0.78
On lipid-lowering drug	26 (27)	6 (17)	20 (32)	0.09	6 (24)	20 (27)	0.74
Glucose, mg/mL	84 (78–90)	86 (78.5–92)	84 (78–89)	0.19	84 (78–92)	84 (78–90)	0.62
Insulin, µIU/mL	8 (5–15)	10 (6–17)	7 (4–14)	0.05	8 (6–15)	8 (4–15)	0.61
HOMA-IR**	1.8 (0.9–3.2)	2.1 (1–4)	1.4 (0.9–2.8)	0.05	1.8 (1–3.2)	1.9 (0.9–3)	0.6
Body mass index, kg/m2	26 (23.3–30)	28.3 (23.2–30.6)	25.2 (23.3–29)	0.1	27.8 (23.2–30)	25.5 (23.3–29.6)	0.68
Total bilirubin, mg/dL	0.6 (0.4–1.4)	1.8 (1.1–2.6)	0.4 (0.3–0.5)	<0.01	—	—	—
AST, U/L	22 (17–29)	20.5 (16–30)	22 (17–28)	0.62	21 (16–28)	22 (17–29)	0.77
ALT, U/L	40 (32–51)	39 (32–53.5)	40 (32–50)	0.85	40 (36–50)	39 (30–51)	0.41
CD4+, cells/mm3	578.5 (431–789)	556 (451.5–739.5)	588 (413–834)	0.9	529 (431–667)	608 (455–864)	0.22
Nadir CD4+, cells/mm3	130 (32–238)	114 (32–231)	140 (34–240)	0.47	130 (12–200)	130 (38–246)	0.24
HIV duration, years	11.3 (7.2–16.2)	11.3 (8.8–15.2)	10.1 (6.5–17.5)	0.86	11.3 (8.8–14.3)	11.2 (7.2–17.5)	0.81
On PI	51 (52)	—	—	—	24 (96)	27 (37)	<0.01
On NNRTI	47 (48)	—	—	—	1 (4)	46 (63)	<0.01

Continuous variables reported as median (interquartile range); categorical variables reported as frequency (column percent).

BP, blood pressure; HTN, hypertension; HDL, high density lipoprotein; HOMA-IR, homeostasis model of insulin resistance; PI, protease inhibitor; NNRI, non-nucleoside reverse transcriptase inhibitor.

^*Unpaired t-test or Wilcoxon Rank Sum test as distributionally appropriate for continuous variables; Chi-Square test, Fisher’s Exact test or Pearson Exact Chi-Square test as appropriate for categorical variables.

^**HOMA-IR was calculated using the following formula: HOMA-IR = fasting glucose (mg/dl)×fasting insulin (µU/ml)/405.

Results of FMD analysis and inflammation, coagulation and oxidation marker levels overall, by ATV use (ATV vs. no ATV) and by total bilirubin level (≥75^th percentile vs. <75^th percentile) are shown in Table 2. There were no differences between groups with regard to the baseline brachial artery diameter. Median (IQR) FMD for the overall group was low, 3.29% (1.58%-6.17%), compared to healthy adults(17) and there were no between-group differences with regard to FMD in this study. There were no significant differences between groups divided by ATV use or by bilirubin level with regard to inflammation markers, D-Dimer or F₂-Isoprostanes. However, fibrinogen was higher in the ATV group. Total bilirubin level as a continuous variable was not correlated with FMD or any inflammation or oxidation markers, with the exception of fibrinogen (Spearman correlation coefficient was 0. 2285; p=0.02).

Table 2.

FMD, Inflammation, Coagulation and Oxidation Markers Overall, by Atazanavir Use and by Total Bilirubin Level

Variable	Overall (N=98)	Atazanavir (n=36)	No Atazanavir (n=62)	P*	Total Bilirubin ≥1.4 mg/dL (n=25)	Total Bilirubin <1.4 mg/dL (n=73)	P*
Brachial artery diameter, mm	4.68 (4.09–5.01)	4.61 (4.08–4.98)	4.7 (4.28–5.01)	0.57	4.58 (4.13–5.02)	4.7 (4.05–5)	0.94
FMD, %	3.29 (1.58–6.17)	3.15 (0.69–5.91)	3.33 (1.8–6.25)	0.67	2.79 (0.66–4.82)	3.7 (2.06–6.41)	0.08
IL-6, pg/mL	2.44 (1.52–4.08)	2.82 (1.79–5.34)	2.2 (1.37–3.59)	0.18	2.57 (1.65–4.29)	2.36 (1.52–3.78)	0.64
sTNFR-I, pg/mL	1286.19 (1074.2–1581.1)	1273.24 (1067.51–1556.79)	1296.71 (1074.2–1581.1)	0.82	1303.4 (1052.32–1686)	1284.48 (1076.29–1459.62)	0.81
sTNFR-II, pg/mL	2694.95 (2269.6–3291.7)	2673.55 (2335.2–3389.55)	2707.96 (2198.2–3050.67)	0.83	2717.9 (2152–3672.67)	2682.5 (2296.5–3027.99)	0.76
hs-CRP, μg/mL	1.33 (0.8–3.8)	1.78 (0.56–4.65)	1.28 (0.86–2.98)	0.8	1.08 (0.47–2.58)	1.34 (0.86–3.8)	0.20
sICAM-1, ng/mL	226.89 (150.27–310.65)	254.32 (153.75–317.82)	223.34 (150.27–296.37)	0.51	251.87 (145.15–310.65)	225.67 (161.5–296.37)	0.88
sVCAM-1, ng/mL	687.41 (566.83–884.36)	694.3 (555.19–879.79)	677.61 (574.47–891.21)	0.76	756.62 (629.98–884.36)	669.17 (566.83–836.11)	0.4
D-Dimer, µg/mL	0.17 (0.11–0.28)	0.18 (0.13–0.29)	0.14 (0.11–0.25)	0.19	0.14 (0.11–0.31)	0.17 (0.11–0.25)	>0.99
Fibrinogen, mg/dL	372.5 (322–439)	411.5 (354.5–501.5)	350.5 (300–418)	<0.01	406 (331–493)	365 (318–421)	0.06
F2-IsoPs, ng/mL	0.05 (0.03–0.21)	0.06 (0.04–0.20)	0.04 (0.03–0.22)	0.26	0.06 (0.04–0.21)	0.05 (0.03–0.2)	0.23

All variable reported as median (interquartile range)

FMD, flow mediated dilation; IL-6, interleukin-6; sTNFR-I, soluble tumor necrosis factor receptor-I; sTNFR-II, soluble tumor necrosis factor receptor-II; hs-CRP, high sensitivity C-reactive protein; sICAM-1, soluble intercellular adhesion molecule-1; sVCAM-1, soluble vascular cell adhesion molecule-1; F₂-IsoPs , F₂-Isoprostanes

^*Unpaired t-test or Wilcoxon Rank Sum test as distributionally appropriate.

In univariable analysis, total bilirubin, age, BMI, AST, HIV-1 RNA, IL-6, D-Dimer and brachial artery diameter had p<0.25 and were entered into the first multivariable model. Neither total bilirubin (as a categorical or continuous variable) nor ATV status were independently associated with FMD in this multivariable model. In a second modeling approach adjusting for clinically relevant variables, i.e., age, sex, race, BMI, CD4+ cell count, HIV-1 RNA level, whether on an anti-hypertensive or cholesterol-lowering medication, smoking status and brachial artery diameter, did not change this result. Parameter estimates and significance levels for the variables of interest are shown in Table 3 for both univariable and multivariable analyses.

Table 3.

Univariable and Multivariable Analysis with FMD as Dependent Variable

	Parameter Estimate	95% Confidence Interval	P
Univariable Approach
Atazanavir Use (Yes=1)	−0.2898	−1.6206-1.041	0.67
Total Bilirubin (>75th percentile=1)	−1.2987	−2.7482-0.1508	0.08
Total Bilirubin (mg/dl)	−0.4366	−1.1165-0.2434	0.21
First Multivariable Approach*
Atazanavir Use (Yes=1)	−0.686	−1.9342-0.5623	0.28
Total Bilirubin (>75th percentile=1)	−1.2697	−2.604-0.065	0.06
Total Bilirubin (mg/dl)	−0.3099	−0.9464-0.3366	0.34
Second Multivariable Approach**
Atazanavir Use (Yes=1)	−0.5744	−1.9078-0.7589	0.39
Total Bilirubin (>75th percentile=1)	−1.329	−2.7427-0.0848	0.07
Total Bilirubin (mg/dl)	−0.3531	−1.0284-0.3221	0.3

^*Parameter estimates shown are the variables of interest adjusted for all variables with P<0.25 in univariable analaysis, ie, age, BMI, AST, HIV-1 RNA level, IL-6, D-Dimer and brachial artery diameter

^**Parameter estimates shown are the variables of interest adjusted for clinically-relevant variables, ie, age, sex, race, BMI, CD4+ count, HIV-1 RNA level, whether on an anti-HTN or lipid lowering medication, smoking status and brachial artery diameter, regardless of statistical association of these variables with the outcome, FMD

Discussion

To our knowledge this is the first study to explore the relationship between total bilirubin and endothelial function, inflammation, coagulation and oxidation in virologically-suppressed, HIV-1-infected adults on stable ART. With regard to our primary endpoint, neither total bilirubin nor ATV status were significantly associated with FMD of the brachial artery. This is consistent with both the study by Flammer AJ, et al and the SABAR study by Murphy RL, et al which showed that switching to ATV from another PI, whether boosted with ritonavir or not, did not improve endothelial function measured by FMD after 24 weeks despite significant improvement in lipid levels.(13, 14) It is conceivable that the effect of a modest increase in bilirubin in this population is masked by the ongoing heightened inflammation due to chronic HIV infection. Indeed, with potent ART, inflammation and endothelial dysfunction do improve(18, 19), but not to normal levels, when compared to HIV-uninfected individuals.(2, 19) Further, participants included in this study did not have extremely elevated total bilirubin levels (median 1.8 mg/dl; IQR 1.1–2.6 mg/dl; minimum 0.3 mg/dl and maximum 5 mg/dl). Although seeing an effect with extreme hyperbilirubinemia would be mechanistically intriguing, this would have uncertain clinical relevance. Another consideration is that the anti-oxidant effect of elevated bilirubin was outweighed by the oxidative stress induced by ART.(20)

Although a different method for measuring endothelial function was used, our results are incongruent with the study by Dekker D, et al in which ATV-induced hyperbilirubinemia did improve endothelium-dependent vasodilation measured by forearm blood flow response to acetylcholine in participants with type II diabetes mellitus after 3 days.(11) In our study, perhaps an effect would have been seen if FMD had been measured earlier after ATV was initiated (median (IQR) duration on ATV in our study was 28.5 (16.8–47.7) months). The clinical implication of a solely transient acute effect would also be questionable, however. Another consideration is that endothelial dysfunction is more pronounced in subjects with diabetes mellitus than in our HIV population and is why an effect was seen in this potentially higher risk group. To better assess if the degree of endothelial dysfunction played a role in the association between total bilirubin level and FMD in our study, the correlation between total bilirubin level and FMD in those with the lowest FMD (FMD less than the median FMD of 3.3%) was determined. Total bilirubin was not correlated with FMD in this subgroup (Spearman correlation coefficient=0.13; p=0.38).

With regard to our secondary endpoints, neither total bilirubin nor ATV status were associated with markers of inflammation, coagulation or oxidation with the exception of fibrinogen. Fibrinogen levels were higher among participants taking ATV. This result is consistent with a study by Madden E, et al where PI use was associated with elevated fibrinogen levels.[18] In our study, median (IQR) fibrinogen was 411 (49–411) mg/dL in those on a PI vs. 334 (294–334) mg/dL in those not on a PI (p<0.01). Because most participants in our study on a PI were on ATV (36/51), it stands to reason that one is a marker for the other.

The strength of this study is the large number of participants allowing for adequate power to address the study question. There are limitations, however. Because ART was not randomized in this study, unmeasured confounding or confounding by indication could be the reason for the results obtained. Cardiovascular risk may have contributed to the decision to prescribe an ATV-based regimen. If this were true, FMD may have been impaired to a greater extent in patients receiving ATV and may have masked the effect of bilirubin. However, cardiovascular risk factors were balanced between the participants, including those not modifiable, ie age, sex and race. Also, adjusting for cardiovascular risk factors did not change the results qualitatively. In addition, we were unable to control for time on ATV or prior ART exposure. As suggested above, an effect may have been seen if participants had recently been started on ATV; however, the clinical benefit of a transient effect of this intervention would be questionable. Another limitation is the lack of adjustment for multiple statistical tests which could have increased the likelihood of finding statistical significance due to chance alone. Finally, due to the cross-sectional design, it is not possible to attribute cause to effect. Given the negative results of this study, these last two points are less important, but should be taken into account in the design of future studies.

In conclusion, neither ATV use nor higher total bilirubin levels were statistically associated with better endothelial function or lower inflammation and oxidation in virologically-suppressed, HIV-1 infected adults on stable ART. It is possible that the anti-oxidant and/or the anti-inflammatory effect of bilirubin is transient, observed only with very high levels of bilirubin, or that it is not sufficiently potent to overcome other causes of endothelial dysfunction in this population.