| |
Abacavir-based therapy does not affect biological mechanisms associated with cardiovascular dysfunction
|
| |
| |
AIDS:
28 January 2010 - Volume 24 - Issue 3 - p F1-F9
Martinez, Esteban; Larrousse, Maria; Podzamczer, Daniel; Perez, Ignacio; Gutierrez, Felix; Lonca, Montserrat; Barragan, Patricia; Deulofeu, Ramon; Casamitjana, Roser; Mallolas, Josep; Pich, Judit; Gatell, Jose M; for the BICOMBO Study Team
aHospital Clinic-IDIBAPS, University of Barcelona
bHospital Universitari de Bellvitge, L'Hospitalet de Llobregat, Barcelona, Spain
cHospital General Universitario de Elche, Elche, Spain.
*Members of the BICOMBO Study Team are listed in the Acknowledgements.
The study was presented in part as oral communications at the 5th International AIDS Society Conference on HIV Pathogenesis, Treatment and Prevention, 19-22 July 2009, Cape Town, South Africa (abstract MOAB203) and at the 1st GESIDA (Grupo de Estudio del SIDA) Congress, 21-24 October 2009, Madrid, Spain (abstract O-4).
"Had any of the study compounds had any direct effect on the mechanisms of atherogenesis assessed, this effect would have emerged and it should have been detected because participants were continuously exposed to the drugs......In conclusion, abacavir/lamivudine increased total and LDL cholesterol compared with tenofovir/emtricitabine, but it did not cause inflammation, endothelial dysfunction, hypercoagulability, or insulin resistance in virologically suppressed HIV-infected patients.....Two major sources of bias and confounding not controlled for in previous studies, such as drug prescription and uncontrolled HIV infection, did not affect the BICOMBO study. Therefore, the BICOMBO study provided an excellent opportunity to assess the potential effects of abacavir, as ABC/3TC, on different mechanisms involved in the pathogenesis of myocardial infarction relative to TDF/FTC."
Abstract
Objective: To assess the effects of initiating abacavir-containing therapy on plasma lipids and cardiovascular biomarkers.
Design: Sub-study of the BICOMBO study in which participants were randomized to switch their nucleoside backbone to either abacavir/lamivudine or tenofovir/emtricitabine.
Methods: We assessed 48-week changes in fasting lipids and several biomarkers including serum high-sensitivity C-reactive protein (hsCRP), monocyte chemoattractant protein-1 (MCP-1), osteoprotegerin, interleukin (IL)-6, IL-10, tumor necrosis factor alpha (TNF-alpha), intercellular adhesion molecule-1 (ICAM-1), vascular adhesion molecule-1 (VCAM-1), selectin E and P, adiponectin, insulin, and D-dimer in otherwise healthy, virologically suppressed HIV-infected patients randomly switched to abacavir/lamivudine or tenofovir/emtricitabine with no history of cardiovascular disease, no prior abacavir or tenofovir use, and no virological failure or AIDS during follow-up.
Results: Eighty (46 abacavir/lamivudine and 34 tenofovir/emtricitabine) patients were included. Baseline characteristics were similar between groups and between patients in the sub-study vs. those not. There were no significant differences in baseline lipids and markers between groups. Although total (6.5 vs. -6.7%, P < 0.0001) and low-density lipoprotein (LDL) (8.6 vs. -9.1%, P = 0.004) cholesterol increased significantly in the abacavir/lamivudine group relative to the tenofovir/emtricitabine group, we found no significant changes in the biomarkers: CRP (-3.9 vs. 0.0%), MCP-1 (5.9 vs. 4.0%), osteoprotegerin (5.1 vs. -2.8%), IL-6 (0.0 vs. 0.0%), IL-10 (0.0 vs. 0.0%), TNF-alpha (0.0 vs. 0.0%), ICAM-1 (6.6 vs. 5.2%), VCAM-1 (0.02 vs. -0.01%), selectin E (-0.4 vs. 7.8%), selectin P (4.6 vs. 12.6%), insulin (-2.5 vs. 8.8%), adiponectin (-2.2 vs. 15.4%), and D-dimer (0.0 vs. 0.0%) (P ≥ 0.12 for all comparisons).
Conclusion: Abacavir/lamivudine increased total and LDL cholesterol compared with tenofovir/emtricitabine, but it did not cause inflammation, endothelial dysfunction, insulin resistance, or hypercoagulability in virologically suppressed HIV-infected patients.
Introduction
There is currently a huge debate on the use of abacavir and the risk of developing ischemic cardiovascular disease, particularly myocardial infarction [1-6]. Abacavir use has been identified as a marker of cardiovascular disease in several cohort studies [7-9], although it is currently unclear whether the role of abacavir is causative or not [10,11]. Potential pathogenetic mechanisms are also unclear. Abacavir has not been associated with insulin resistance [12] or lipoatrophy [12,13] and its lipid impact is lower than that of stavudine [13] or protease inhibitors [14], but higher than that of zidovudine [15] or tenofovir [16]. The D:A:D Study unexpectedly found almost a double rate for myocardial infarction in HIV-infected patients treated with abacavir for the previous 6 months. The risk associated with abacavir use was independent of traditional cardiovascular risk factors and it was no longer significant when abacavir had been stopped prior to the last 6 months, suggesting an 'on-off' mechanism directly induced by abacavir that might involve biological mechanisms associated with atherosclerosis [7,17]. A posthoc analysis of the SMART study also showed a higher risk of cardiovascular disease, including myocardial infarction, in patients treated with abacavir [8]. These patients had higher levels of inflammation and hypercoagulability biomarkers at the SMART study entry and it was speculated that abacavir could have induced those pathogenetic mechanisms.
By contrast, data from naïve patients in pooled GSK and ACTG [18,19] studies did not show any significant difference regarding myocardial infarction between abacavir-treated and nonabacavir-treated antiretroviral-naïve patients. In addition, a randomized study comparing between abacavir/lamivudine or tenofovir/emtricitabine in antiretroviral-naïve patients showed a similar decrease in inflammatory markers consistent with effective suppression of HIV replication [16]. In contrast with abacavir, tenofovir has not been associated with a higher risk for myocardial infarction in the D:A:D study [20]. To explain the discrepancies among studies, it has been suggested that the potential cardiovascular effects of abacavir might be less evident in antiretroviral-naïve patients because of the confounding noise due to uncontrolled HIV replication [21]. Several potential mechanisms affecting biological mechanisms associated with cardiovascular dysfunction have been suggested [22-25], but studies to date may have been subjected to bias and confounding because abacavir prescription was not randomized in any of these studies and there was a proportion of patients included with detectable plasma HIV-1 RNA in some of them.
Once-daily fixed-dose combinations of ABC/3TC and TDF/FTC are the most common pharmacologic presentations for using abacavir and tenofovir today. The BICOMBO randomized clinical trial primarily compared the 48-week efficacy and safety of replacing the nucleoside backbone by once-daily fixed-dose ABC/3TC or TDF/FTC in virologically suppressed HIV-1-infected patients treated with a protease inhibitor-based or a non-nucleoside reverse transcriptase inhibitor (NRTI)-based triple regimen [26]. Two major sources of bias and confounding not controlled for in previous studies, such as drug prescription and uncontrolled HIV infection, did not affect the BICOMBO study. Therefore, the BICOMBO study provided an excellent opportunity to assess the potential effects of abacavir, as ABC/3TC, on different mechanisms involved in the pathogenesis of myocardial infarction relative to TDF/FTC.
Results
Population
Among 203 (61%) patients with no prior ABC or TDF use in the BICOMBO study, 172 (52%) remained on the study drug assigned at 48 weeks but only 87 (26%) had serum samples available at baseline and 48 weeks. Of those, three patients in each group were excluded due to symptomatic cardiovascular disease (n = 2, one of them during the follow-up in the TDF/FTC group) or diabetes mellitus (n = 4, none of them during the follow-up) and one patient in the ABC/3TC group due to virological failure. Therefore, 80 (24%) (46 ABC/3TC and 34 TDF/FTC) patients were eligible and all agreed to participate in the sub-study. Baseline characteristics are shown in Table 1. Most of the patients (80%) were men. At baseline, median age was 43 years, median previous antiretroviral exposure was 3.9 years, and median Framingham score was 4.4. These characteristics were similar between groups and between patients participating in the sub-study vs. those remaining in the main study (Table 1).
Laboratory markers
The values of laboratory markers at baseline and at 48 weeks are shown in Tables 2 and 3, respectively. Markers both at baseline and at 48 weeks were detectable in most patients in each group, except for IL-6, TNF-alpha, IL-10, and D-dimer, which were detectable in between 35 and 47% of the patients in each group. Although there was a trend to higher baseline D-dimer in the TDF/FTC group relative to the ABC/3TC group (0.20 vs. 0.17 µg/l, P = 0.06), there were no significant differences in any marker at baseline or at 48 weeks between the two groups.
Eighteen (39%) and 11 (32%) patients in the ABC/3TC and TDF/FTC groups, respectively, showed an absolute hsCRP increase at 48 weeks (P = 0.62). Four (9%) and one (3%) patient in the ABC/3TC and TDF/FTC groups, respectively, showed hsCRP higher than 0.5 mg/dl at 48 weeks (P = 0.39). Sixteen (35%) and nine (26%) patients in the ABC/3TC and TDF/FTC groups, respectively, showed an increase in hsCRP at 48 weeks 25% higher than that at baseline (P = 0.61).
We found significant correlations between markers at baseline reflecting similar mechanisms: selectin E-selectin P (Spearman correlation coefficient 0.63, P < 0.0001), ICAM-1-VCAM-1 (Spearman correlation coefficient 0.51, P < 0.0001), and insulin-adiponectin (Spearman correlation coefficient -0.25, P = 0.03). We also found significant correlations between Framingham score and some markers at baseline such as OPG (Spearman correlation coefficient 0.31, P = 0.008), selectin E (Spearman correlation coefficient 0.29, P = 0.016), and selectin P (Spearman correlation coefficient 0.26, P = 0.032). Finally, there were significant correlations between 48-week changes in hsCRP and ICAM-1 (Spearman correlation coefficient 0.33, P = 0.008), ICAM-1 and VCAM-1 (Spearman correlation coefficient 0.34, P = 0.006), OPG and selectin-P (Spearman correlation coefficient -0.26, P = 0.037), selectin-E and D-dimer (Spearman correlation coefficient -0.36, P = 0.003), and adiponectin and VCAM-1 (Spearman correlation coefficient -0.33, P = 0.008).
Lipids
The values of fasting plasma lipids at baseline and at 48 weeks are shown in Tables 4 and 5, respectively. Lipids, both at baseline and at 48 weeks, were available in more than 90% of the patients in each group. Although there were no significant differences in any lipid parameter at baseline, significant increases in total (6.5 vs. -6.7%, P < 0.0001) and low-density lipoprotein (LDL) (8.6 vs. -9.1%, P = 0.004) were found in the ABC/3TC group relative to TDF/FTC group. There was also a trend to a higher total-to-high-density lipoprotein (HDL) cholesterol ratio in the ABC/3TC group relative to the TDF/FTC group (6.7 vs. -2.5%, P = 0.08).
There were no significant correlations between lipids and markers at baseline or between 48-week changes in lipids and markers. There were significant correlations between 48-week changes in total cholesterol and triglycerides (Spearman correlation coefficient 0.31, P < 0.0001), total cholesterol and LDL cholesterol (Spearman correlation coefficient 0.76, P < 0.0001), total cholesterol and HDL cholesterol (Spearman correlation coefficient 0.37, P < 0.0001), and LDL cholesterol and HDL cholesterol (Spearman correlation coefficient 0.15, P = 0.03).
Discussion
In this sub-study of the BICOMBO clinical trial, the introduction of ABC as the fixed-dose combination ABC/3TC for 48 weeks did not lead to any significant change in laboratory markers of inflammation, endothelial dysfunction, insulin resistance, or hypercoagulability as compared with the fixed-dose combination TDF/FTC. Changes in markers in both study groups were small and not clinically significant. The results were consistent in all markers measured as either absolute or percentage changes, and in the case of hsCRP with analysis of different outcomes. As a marker of quality, we found significant correlations between markers that measure similar mechanisms such as selectin E-selectin P, ICAM-1-VCAM-1, and insulin-adiponectin. Despite the negative findings on laboratory marker changes between the two study groups, we were able to detect significant increases in total and LDL cholesterol in ABC/3TC patients relative to TDF/FTC, in accordance with the lipid results of the parent study [26] and those of another sub-study that was specifically designed to assess lipid changes [30].
We took special care to avoid any apparent source of bias or confounding in the design of the sub-study. The randomized nature of the main BICOMBO study prevented prescription bias. Although not all the patients recruited in the parent study participated in the sub-study, the sub-study population was fully representative of that remaining in the parent study. By entry criteria of the main study, patients were virologically suppressed for at least the previous 6 months, and this fact prevented confounding due to HIV replication [31]. Participants began the sub-study exposed for the first time to either ABC/3TC or TDF/FTC and exposure to study drugs lasted for the whole follow-up. Had any of the study compounds had any direct effect on the mechanisms of atherogenesis assessed, this effect would have emerged and it should have been detected because participants were continuously exposed to the drugs. Finally, participants had no history of symptomatic cardiovascular disease or diabetes mellitus and no virological failure or AIDS events during follow-up, granting the avoidance of confounding effects due to any these problems. In fact, participants in this sub-study had low cardiovascular risk as shown by the Framingham score, their total and HDL cholesterol plasma values were within the normal range, baseline markers were low, and their 48-week changes were small [32]. We hypothesized that this neat background would have allowed any eventual drug-induced effect on markers of cardiovascular dysfunction to become apparent and hence detectable.
The STEAL study was a randomized clinical trial with a similar design as the BICOMBO study, except for participants being HLA-B*5701-negative and for follow-up being 96 weeks [33]. In contrast with the BICOMBO study results, ABC/3TC-treated patients from the STEAL study showed a higher rate of cardiovascular disease as compared with TDF/FTC-treated ones. Because cardiovascular outcome was not a preplanned end point and because the study was open label, it is unclear whether patients assigned to TDF/FTC in the STEAL study underwent similar clinical research to diagnose cardiovascular events as those patients assigned to ABC/3TC. Moreover, patients from the STEAL study were older, almost exclusively men, and with a higher baseline Framingham score than patients from the BICOMBO study. Finally, despite the randomized design, patients assigned to ABC/3TC in the STEAL study had a higher prevalence of cardiovascular risk factors than patients assigned to TDF/FTC. If the STEAL study investigators considered performing an analysis of cardiovascular biomarkers on stored sera, it would be of paramount importance that any potential confounding factor is controlled for as much as possible in order to obtain reliable results.
Our study had limitations. The sub-study was not preplanned, although the collection of serum samples for additional research purposes had been established in the original protocol. Only a portion of patients from the BICOMBO study could be included, but those included were balanced between treatment groups and they were representative of the rest of the population from the main study. Some markers had confirmed undetectable levels for a proportion of patients in both study groups; although an arbitrary value of the lower limit of detection was given for the purpose of computing changes, this was not an exact measurement. Finally, a number of different markers were studied, but there are other potentially important ones that were not assessed in this sub-study.
In conclusion, abacavir/lamivudine increased total and LDL cholesterol compared with tenofovir/emtricitabine, but it did not cause inflammation, endothelial dysfunction, hypercoagulability, or insulin resistance in virologically suppressed HIV-infected patients.
Methods
Patients
The BICOMBO clinical trial enrolled 333 HIV-1-infected adults with plasma HIV-1 RNA below 200 copies/ml for at least the previous 6 months and who were receiving triple antiretroviral therapy consisting of 3TC plus another NRTI (including ABC or TDF) plus either a protease inhibitor or a non-NRTI. Characteristics of patients and outcomes have been published previously [26]. Briefly, patients were randomized to switch from their NRTI backbone to either ABC/3TC or TDF/FTC. In accordance with standard clinical practice at the time of the study, HLA B*5701 screening before initiating ABC/3TC was not done. Only 17 (10%) and nine (5%) patients assigned to ABC/3TC or TDF/FTC, respectively, discontinued study drugs after 48 weeks of follow-up, and more than 80% of the discontinuations occurred in the first 2 months of the study. After 48 weeks of follow-up, only one patient (assigned to TDF/FTC) developed myocardial infarction.
Although the sub-study was planned after the parent study had been initiated in light of emerging information from the D:A:D study showing an association between recent abacavir use and the risk of myocardial infarction, the collection of serum samples for additional research purposes had been established in the original protocol. Three of the 18 centers from the BICOMBO study participated in this sub-study because they had baseline serum samples available.
For the purpose of the sub-study, eligible patients were those with no prior ABC or TDF use, no discontinuation of the fixed-dose NRTI assigned during follow-up, availability of serum samples at baseline and 48 weeks, no history of symptomatic cardiovascular disease or diabetes mellitus, and no virological failure or AIDS events during follow-up. All participants provided written informed consent. This sub-study was approved by the Ethics Committee at each participating center and by the Spanish Medicines Evaluation Agency.
Methods
Because the increased risk of myocardial infarction in the D:A:D study was associated with current use of abacavir or in the previous 6 months, we assumed that blood samples from eligible patients taken at 48 weeks would be capable to reflect any potential effect induced by abacavir. Due to the lack of knowledge regarding the pathogenesis linking abacavir and cardiovascular disease, we planned to investigate several mechanisms associated with atherogenesis to increase the sensitivity of the study. These mechanisms were inflammation, endothelial dysfunction, insulin resistance, and hypercoagulability [27,28]. We decided to assess several markers for each mechanism to increase the specificity of the results. Markers associated with inflammation included high-sensitivity C-reactive protein (hsCRP), monocyte chemoattractant protein-1 (MCP-1), osteoprotegerin (OPG), interleukin-6 (IL-6), interleukin-10 (IL-10), and tumor necrosis factor alpha (TNF-alpha). Markers associated with endothelial dysfunction included intercellular adhesion molecule-1 (ICAM-1), vascular cell adhesion molecule-1 (VCAM-1), selectin E, and selectin P. Markers associated with insulin resistance included adiponectin and insulin. D-Dimer was the only marker associated with hypercoagulability that could be studied. Some of these markers are known to be involved in more than one mechanism [27,28].
For each participant, data at baseline regarding age, sex, years of antiretroviral exposure, current use of protease inhibitors, and variables needed to estimate Framingham score were selected from the BICOMBO study database. Framingham score was estimated according to standard tools [29]. Data for each participant on fasting plasma lipid values at baseline and at 48 weeks were also collected from the main study database.
Measurement of laboratory markers
EDTA-plasma (Vacutainer System; Becton Dickinson, San Jose, California, USA) samples were stored at -80°C until analysis. Laboratory markers were measured by technicians blinded to treatment assignation. hsCRP was determined by particle-enhanced immunonephelometry (Dade Behring, Marburg, Germany). MCP-1, IL-6, IL-10, TNF-alpha, ICAM-1, VCAM-1, selectin E, and selectin P were measured using commercially available ELISA assays (R&D Systems, Minneapolis, Minnesota, USA). OPG was measured by ELISA (Biomedica, Wien, Germany). Adiponectin was measured by radioimmunoassay (Linco Research, St Charles, Missouri, USA). Insulin was measured by a monoclonal immunoradiometric assay (Medgenix Diagnostics, Fleunes, Belgium). D-Dimer was quantitatively measured using a commercially available latex-enhanced turbidimetric test (Dade Behring, Marburg, Germany). EDTA plasma D-dimer values were validated by statistical correlation with sodium citrate plasma D-dimer values in normal individuals (data not shown). Intra-assay coefficients of variation for the laboratory markers measured were hsCRP 3.1%, MCP-1 4.9-7.8%, OPG 4-10%, IL-6 1.6-4.2%, IL-10 1.7-5.0%, TNF-alpha 4.2-5.2%, ICAM-1 3.6-5.2%, VCAM-1 2.3-3.6%, selectin E 5.2-6.6%, selectin P 4.9-5.6%, adiponectin 3.8%, insulin 5.2%, and D-dimer 1.3-3.0%. Interassay coefficients of variation for the laboratory markers measured were hsCRP 2.5%, MCP-1 4.6-6.7%, OPG 7.0-8.0%, IL-6 3.3-6.4%, IL-10 5.9-7.5%, TNF-alpha 4.6-7.4%, ICAM-1 4.4-6.8%, VCAM-1 5.5-7.8%, selectin E 7.3-8.7%, selectin P 7.9-9.9%, adiponectin 5.5%, insulin 6.9%, and D-dimer 0.8-3.8%. The limits of detection for the laboratory markers measured were hsCRP 0.01 mg/dl, MCP-1 5.0 pg/ml, OPG 0.14 pmol/l, IL-6 0.7 pg/ml, IL-10 3.9 pg/ml, TNF-alpha 1.6 pg/ml, ICAM-1 0.096 ng/ml, VCAM-1 0.6 ng/ml, selectin E 0.009 ng/ml, selectin P 0.5 ng/ml, adiponectin 0.0001 µg/ml, insulin 4 mU/l, and D-dimer 0.17 mg/l.
Interpretation of the results
Samples disclosing undetectable levels of any marker were retested for confirmation. For patients with confirmed measurements of laboratory markers below the limit of quantification, we assumed the respective lower limit of quantification for data analysis. At least one of the markers should show a statistical difference in the change from baseline to 48 weeks between groups to consider that the mechanism associated with that marker was potentially caused by any of study drugs. However, we assumed that statistical differences in the 48-week change in the same direction for more than one marker associated with a given mechanism were required to consider that such a mechanism was definitively caused by any of study drugs.
Statistical analyses
Demographic characteristics at baseline between participants in the sub-study assigned to either ABC/3TC or TDF/FTC and between participants in the sub-study and nonparticipants from the main study were compared using the Wilcoxon rank-sum test or Fisher's exact test for continuous variables and categorical variables, respectively. Biomarkers and lipids at baseline and at 48 weeks and absolute and relative 48-week changes between participants in the sub-study assigned to either ABC/3TC or TDF/FTC were compared with the Wilcoxon rank-sum test. The punctual estimation and 95% confidence interval of difference in medians was estimated using the methodology of Hodges-Lehman using the distribution-free test of Moses. Because hsCRP is the marker with more widespread clinical use, we performed additional analyses. We compared the proportions of patients in each group showing an absolute hsCRP increase at 48 weeks, an hsCRP at 48 weeks higher than 0.5 mg/dl (upper limit of the normal reference range for the ultrasensitive CRP assay at our institutions), and an hsCRP increase at 48 weeks 25% higher than that at the baseline with the Fisher's exact test. Correlations between continuous variables were evaluated using Spearman's rank correlation test. All statistical analyses were carried out using the SAS software (SAS/STAT guide for personal computers, version 9.1.3; SAS Institute Inc., Cary, North Carolina, USA).
The study was supported in part by research grants from Gilead Sciences and GlaxoSmithKline, and grants from Fondo de Investigaciones Sanitarias (PI05255) and Red Tematica Cooperativa de Investigacion en SIDA (RIS G03/173), Ministerio de Sanidad y Politica Social (Spain).
|
|
| |
| |
|
|
|
