| |
Antiretroviral Therapy is Associated With an Atherogenic Lipoprotein Phenotype Among HIV-1-Infected Men in the Multicenter AIDS Cohort Study
|
| |
| |
[Clinical Science]
JAIDS Journal of Acquired Immune Deficiency Syndromes:Volume 48(3)1 July 2008pp 281-288
Riddler, Sharon A MD, MPH*; Li, Xiuhong MAS; Otvos, James PhD; Post, Wendy MD, MS; Palella, Frank MD; Kingsley, Lawrence Dr.P.H*; Visscher, Barbara MD, Dr.P.H; Jacobson, Lisa P ScD; Sharrett, A R MD, Dr.P.H.
From the *Department of Medicine and Department of Infections Diseases and Microbiology, University of Pittsburgh, Pittsburgh, PA; Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD; LipoSciences, Inc, Raleigh, NC; Division of Infectious Diseases, Northwestern University, Chicago, IL; and the Department of Epidemiology, School of Public Health, University of California, Los Angeles, CA.
"In this study, we examined the lipoprotein profile of men with HIV-1 infection and the effect of antiretroviral therapy in the MACS. After controlling for numerous factors, we observed that the use of HAART was associated with an ALP (increased TG, increased VLDL-p and small LDL-p, with decreased large LDL-p and HDL-p). This pattern has been seen in non-HIV-infected patients with type 2 diabetes and obesity. The difference between the HAART-treated men and the HIV-seronegative men for total LDL-p did not achieve statistical significance, whereas the distribution between small and large LDL-p was significantly altered (19% lower large LDL-p and 13% higher small LDL-p for the HIV-infected HAART recipients). Proportionally, the largest differences were in total VLDL-p, which was 45% higher among HAART recipients as compared with the uninfected men. The observed differences between HIV-seronegative men and HIV-1-infected HAART-treated men persisted after adjustment for obesity (BMI), insulin, and diabetes and after exclusion of men with recent weight change. The effect of HAART was most pronounced among those men with a good clinical status (as defined by plasma HIV-1 RNA < 50 copies/mL and CD4+ T-cell counts greater than 350 cells/mm3). Conversely, men on HAART with a poor clinical status had the lowest levels of all lipoprotein particles, possibly due to low adherence to HAART, poor nutritional status, or declining health. HIV-1 infection without HAART treatment was associated with lower particle concentrations of small and large LDL-p and HDL-p. The lipoprotein particle pattern in the HAART-naive group was similar to that of HAART-treated men with poor clinical status."
Abstract
Background: Alterations in serum lipids and an increased risk of myocardial infarction have been associated with HIV-1 infection and its treatment.
Methods: Lipoprotein subclasses were measured by nuclear magnetic resonance spectroscopy in frozen plasma samples from participants in the Multicenter AIDS Cohort Study. The effects of HIV-1 infection, antiretroviral therapy, and other factors on median particle concentrations were examined using quantile regression.
Results: Fasted samples were tested from 1082 men, including 609 HIV-seronegative and 473 HIV-1-infected men. Compared with HIV-seronegative men, HIV-1-infected men on antiretroviral therapy had an atherogenic phenotype with higher numbers of very low density lipoprotein and small low-density lipoprotein particles and lower numbers of high-density lipoprotein and large low-density lipoprotein particles. HIV-infected, antiretroviral-naive men had significantly lower high-density lipoprotein and small low-density lipoprotein particle concentrations compared with the HIV-seronegative men. Among men on antiretroviral therapy, the atherogenic phenotype was most pronounced in men with a good clinical status.
Conclusion: Use of antiretroviral therapy in HIV-1-infected men was associated with an atherogenic lipoprotein phenotype.
INTRODUCTION
Abnormalities in serum lipid concentrations have been associated with both HIV-1 infection itself and its treatment with highly active antiretroviral therapy (HAART).1-5 In addition to lipid changes, HAART may result in metabolic disturbances such as insulin resistance, peripheral lipoatrophy, and central adiposity.5 Large prospective observational studies have suggested that use of HAART, particularly HIV-1 protease inhibitors (PIs), was associated with an increased risk of myocardial infarction that is partially mediated by serum lipid abnormalities.6-8
Nuclear magnetic resonance (NMR) spectroscopy can be utilized to quantify the number of particles of specific lipoprotein subclasses.9 Numerous studies have evaluated the relationship of lipoprotein subclasses and future cardiovascular events. Although not conclusive, the data suggest that smaller low-density lipoprotein (LDL) particle size, specifically a predominance of small dense LDL or a greater number of small LDL particles (LDL-p), is a predictor of coronary disease.10-13 Insulin resistance and hypertriglyceridemia are associated with a decrease in average LDL size, due to both an increase in the concentration of small LDL-p and a decrease in large LDL-p.14
Austin described an atherogenic lipoprotein phenotype (ALP) based on a predominance of small, dense LDL-p with accompanying elevated triglycerides and reduced high-density lipoprotein cholesterol (HDL-C).15 Obesity, diabetes, and insulin resistance are known to be associated with elevated triglycerides and reduced HDL-C, and by NMR, this pattern has been shown to be associated with substantially increased numbers of small LDL-p and decreased numbers of large LDL-p.14 In this paper, we will use ALP to describe a pattern of elevated triglycerides, reduced HDL-C, elevated small LDL-p, and reduced large LDL-p, which has been given increased emphasis in recently published guidelines of the National Cholesterol Education Program16 and which we observed with some consistency in the results presented here.
The objective of this study was to describe the distribution of selected lipoprotein subclasses among participants in the Multicenter AIDS Cohort Study (MACS), including HIV-seronegative men, HIV-infected men not on HAART, and HIV-infected men on HAART. The factors that predict alterations in the lipoprotein subclasses were assessed. In HIV-infected individuals treated with antiretroviral medications, differences in serum lipids have been observed between and within drug classes.17 We hypothesized that PI-containing regimens would have a more atherogenic profile [more total very low density lipoprotein particles (VLDL-p) and small LDL-p] than nonnucleoside reverse transcriptase inhibitor (NNRTI)-based regimens.18-20 Within the PI class, we predicted that ritonavir-containing regimens would have the highest total VLDL-p concentration and that atazanavir-containing regimens would have a less atherogenic profile compared with other PI regimens.21
DISCUSSION
In this study, we examined the lipoprotein profile of men with HIV-1 infection and the effect of antiretroviral therapy in the MACS. After controlling for numerous factors, we observed that the use of HAART was associated with an ALP (increased TG, increased VLDL-p and small LDL-p, with decreased large LDL-p and HDL-p). This pattern has been seen in non-HIV-infected patients with type 2 diabetes and obesity. The difference between the HAART-treated men and the HIV-seronegative men for total LDL-p did not achieve statistical significance, whereas the distribution between small and large LDL-p was significantly altered (19% lower large LDL-p and 13% higher small LDL-p for the HIV-infected HAART recipients). Proportionally, the largest differences were in total VLDL-p, which was 45% higher among HAART recipients as compared with the uninfected men. The observed differences between HIV-seronegative men and HIV-1-infected HAART-treated men persisted after adjustment for obesity (BMI), insulin, and diabetes and after exclusion of men with recent weight change. The effect of HAART was most pronounced among those men with a good clinical status (as defined by plasma HIV-1 RNA < 50 copies/mL and CD4+ T-cell counts greater than 350 cells/mm3). Conversely, men on HAART with a poor clinical status had the lowest levels of all lipoprotein particles, possibly due to low adherence to HAART, poor nutritional status, or declining health. HIV-1 infection without HAART treatment was associated with lower particle concentrations of small and large LDL-p and HDL-p. The lipoprotein particle pattern in the HAART-naive group was similar to that of HAART-treated men with poor clinical status.
The absence of an effect of adjustment for obesity, insulin, and diabetes suggests that the particular atherogenic profile described here may be a direct effect of the medications on lipoprotein metabolism that is not mediated by these metabolic abnormalities. Alternatively, it is possible that there was an association with insulin resistance that was masked by confounding between the numerous variables or that an association was not observed due to the modest sample size of the PI-containing HAART recipients among whom insulin resistance would be most likely. The association also seems to be unrelated to weight gain or loss.
Numerous reports have described the effects of HIV-1 infection and its treatment on traditional lipid measures. In general terms, HAART results in increases in total cholesterol and triglycerides with no significant effects on LDL-C and variable increases in HDL-C. Differences between and within antiretroviral drug classes have been observed.17 Only limited data are available regarding the distribution of lipoprotein subclasses in HIV-1-infected persons. The LDL subclass pattern B (small LDL) was observed to be 2.5-fold more frequent in AIDS patients than in age-matched controls in an observational study from the pre-HAART era.31 More recently, one small (N = 37) cross-sectional study has evaluated lipoprotein subclasses in persons with HIV infection.32 In this study, comparing 22 subjects receiving HIV-1 PI with 15 subjects who were receiving antiretroviral therapy without a PI, it was observed that the majority of subjects in both groups had a predominance of small LDL. The PI-treated individuals had significantly higher total cholesterol, triglycerides, and non-HDL-C and higher VLDL-p and HDL-p.
In the general population, the Veterans Affairs High-Density Lipoprotein Intervention Trial demonstrated that NMR measured LDL-p and HDL-p predicted coronary heart disease events better than lipoprotein cholesterol.11 In the Multi-Ethnic Study of Atherosclerosis, both large and small LDL-p were independently associated with carotid atherosclerosis.13
There are several limitations in this study. The MACS study population is all males, and similar analyses of mixed or female cohorts are needed. The analysis of the data was complicated by the clinical use of lipid-lowering medications in the cohort. The analysis was repeated after excluding men on lipid-lowering medication (Table 3); however, this may result in a biased estimate of the lipoprotein levels because these medications were more likely to have been used in men with the highest serum lipid levels. Adjustment for lipid-lowering medication in the model is less likely to introduce selection or indication bias. The HAART regimens in this observational cohort were heterogeneous, precluding detailed analyses of specific drugs. Furthermore, differences between regimens must be taken with caution as men with more atherogenic lipids may have been treated initially or switched to regimens with the least expected impact on lipids. In our data, PI-based regimens were, in general, associated with more ALP-than NNRTI-based regimens, but most of the other comparisons between regimen types were not significant. When these analyses were limited to men receiving the same regimen for 6 months or more, the differences are more pronounced, suggesting an indication bias for the HAART regimen switches in some cases. Analysis of lipoprotein subclasses in clinical trials with randomized treatment assignment, such as A5152s, which identified major lipoprotein changes within and between antiretroviral therapy regimens, would eliminate this bias and be useful in further elucidating the effect of specific regimens on lipoprotein subclasses.
In our study, we observed that, despite a similar LDL-C level, men with HIV-1 infection on HAART had significantly higher small LDL-p and a more atherogenic lipoprotein profile as compared with HIV-seronegative men. These findings provide additional data suggesting that HIV-1 disease and its treatment with HAART result in significant perturbations in lipoproteins. These abnormalities would be expected to result in increased risk of cardiovascular disease in future years.
RESULTS
Study Population
Fasting blood samples were tested from 1082 men who met the criteria for inclusion into these analyses, including 609 HIV-1-seronegative; 77 HIV-1-infected, HAART-naive; and 396 HIV-1-infected, HAART-treated participants. The characteristics of the study population at the time of the lipid measurements are provided in Table 1. Compared with the HIV-1-seronegative men, the HIV-1-infected, HAART-naive men were younger with a higher proportion of minorities and current smokers. As expected, these HIV-1-infected, untreated men also had lower total cholesterol, LDL-C, and HDL-C. Compared with HIV-1-seronegative men, the HIV-1-infected HAART-treated men used lipid-lowering medications more often. The total cholesterol for the HIV-1-infected, HAART recipients was comparable to HIV-1-seronegative men, but TGs were much higher and HDL-C was lower. Fasting insulin and glucose were both higher for the HIV-1-infected, HAART-treated men than for the HIV-1-seronegative men.
The proportion of men with stable weight (defined as less than 5% loss or gain in the prior 6 months) was 84.5%, 78.5%, and 82.5% for the HIV-1-seronegative, HIV-1-infected, HAART-naive, and HIV-1-infected, HAART-treated groups, respectively. The proportion with weight gain was comparable for the 3 groups, but the HIV-1-infected HAART-naive men were more likely to have had weight loss in the prior 6 months.
Of the 396 HAART-treated men, 190 (48%) were currently receiving a PI-containing regimen, 176 (44%) were on an NNRTI-containing regimen (without PI), and 30 (8%) used triple NRTI regimen. Among the 190 men on PI-HAART, 131 (69%) used ritonavir (any dose) and 40 (21%) used atazanavir, of whom 32 (80%) were also on ritonavir.
Lipoprotein Particles
Table 2 shows particle concentrations for the large and small LDL subclasses and for total VLDL and HDL. Intercepts in the Table are interpreted as the median concentrations for HIV-negative men with characteristics described in the footnote of the Table. Compared with HIV-1-seronegative men, HIV-1-infected men on HAART had a lipoprotein pattern consistent with the ALP, with significantly higher median particle concentrations of small LDL and total VLDL and lower numbers of large LDL and total HDL-p. HIV-1-infected HAART-naive men had significantly lower small LDL-p and total HDL-p compared with seronegative men and tended to have lower levels of large LDL-p as well. Important variables in the model included black race-associated with a less atherogenic profile; chronic hepatitis C virus-associated with significantly lower small LDL-p; and higher CRP-associated with more ALP. Compared with the HIV-1-seronegative men, total LDL-p were higher among the HIV-1-infected men on HAART [67 nmol/L, 95% confidence interval (CI) -3, 137; P = 0.06] and lower among the HIV-1-infected HAART-naive men (-110 nmol/L, 95% CI -171, -49; P < 0.001). This analysis was repeated after excluding men on lipid-lowering medication (Table 3). The patterns of lipoproteins were similar to the entire population.
TABLE 2. Particle Concentrations of Large LDL, Small LDL, Total VLDL, and Total HDL and Adjusted Median Differences (95% CI) Due to HIV Status, HAART Use, and Other Risk Factors Among All Participants
After adjusting for BMI, fasting insulin, and diabetes, there were minimal changes in the estimates of the model for each lipoprotein subfraction. In Figure 1, the percent differences in the estimates of the 4 categories of lipoproteins for the HIV-1-infected, HAART men compared with the HIV-seronegative men are displayed before and after further adjustment. For example, for large LDL-p, the percent difference between HIV+ HAART-treated men compared with seronegative men was -20.2% after adjustment for BMI, insulin, and diabetes compared with -20.9% before adjustment.
This analysis was repeated using data only from the men with stable weight for the prior 6 months. The restricted analysis did not result in any change in the primary findings in Table 2. Importantly, the increased total VLDL-p among the HIV-1-infected, HAART-treated men was not altered, indicating that the higher total VLDL-p observed in this group was not a factor of recent weight gain.
In Figure 2, the HIV-1-infected men on HAART were classified by current clinical status: good (n = 220), intermediate (n = 160), or poor (n = 16) as defined in the Methods. As shown in the Figure, the atherogenic pattern of higher small LDL-p and VLDL-p with lower large LDL-p was especially marked among HIV-1-infected HAART-treated men with a good clinical status. In contrast, HIV-1-infected men on HAART who had a poor clinical status had lower particle numbers of all lipoprotein classes. Median lipoprotein particle concentrations for the men with intermediate clinical status were intermediate between those of the men with good and poor clinical status. Adherence to the HAART regimen measured by the 4-day recall was 88%, 86%, and 75% for the good, intermediate, and poor clinical status groups, respectively. When the analysis was restricted to the men with stable weight, there were no changes in the effect of HAART among the HIV-1-infected men with good or intermediate clinical status. The poor status group was too small to draw significant conclusions regarding the impact of weight change.
Effect of ART Regimen on Lipoprotein Particle Distributions Among HIV-1-Infected HAART Recipients
The adjusted median differences in lipoprotein subclass particle concentrations for the HIV-1-infected HAART-treated men by antiretroviral drug class are provided in Table 4. Compared with HIV-seronegative men, both PI-containing and triple NRTI regimens were associated with ALP, that is, significantly higher small LDL-p and total VLDL-p and lower large LDL-p and total HDL-p, whereas NNRTI-based regimens were found to have more modest effects on total VLDL-p and HDL-p with no significant or substantial difference in large or small LDL-p.
PI-containing regimens were associated with significantly higher total VLDL-p and lower large LDL-p than NNRTI-containing regimens. The small LDL-p was higher as well, but this was not statistically significant (+124 nmol/L, 95% CI -10, 259). When restricted to men on the same regimen type for >6 months, these differences became more pronounced with significantly higher small LDL-p (+162 nmol/L, 95% CI 23, 301) and lower total HDL-p (-1.5 _mol/L, 95% CI -2.8, -0.2) among the PI recipients. Triple NRTI regimens had lower large LDL-p and tended to have higher small LDL-p and lower total HDL-p as compared with NNRTI.
Both ritonavir-containing and non-ritonavir-containing PI regimens were associated with significantly higher total VLDL-p and lower large LDL-p and a trend to higher small LDL-p and lower HDL-p when compared with HIV-1-seronegative men (Table 3). There were no significant differences between PI regimens with or without ritonavir. The results were similar when men on lipid-lowering medications were excluded. There were no substantial changes when the analysis was limited to men with stable HAART (on the same HAART regimen class at the prior visit 6 months earlier) or men with stable weight (data not shown). Additionally, adjustment for CD4 cell count and HIV-1 RNA did not substantially impact the results of the analysis.
Most men on atazanavir (32 of 40) were also taking ritonavir. Compared with HIV-1-seronegative men, the atazanavir recipients had significantly higher total VLDL-p and lower large LDL-p and total HDL-p. Compared with other PI-containing HAART, atazanavir regimens had lower concentrations for all lipoprotein particles, but only the difference in large LDL-p was significant (-76; 95% CI -140, -14). This finding (lower large LDL-p) became nonsignificant when the analysis was limited to men on the same regimen class at the prior visit. However, the sample size decreased from 40 to 18, indicating that more than half of the men had switched to atazanavir in the past 6 months.
No differences in lipoprotein particle concentrations were observed for stavudine-vs non-stavudine-containing regimens.
TABLE 4. Adjusted* Median Differences (95% CI) in Concentrations of Large LDL, Small LDL, Total VLDL, and Total HDL-p by HAART Regimen Class
METHODS
Sample
The MACS is an ongoing multicenter (Pittsburgh, PA; Baltimore, MD; Chicago, IL; and Los Angeles, CA) prospective cohort study of 6972 homosexual and bisexual men, of whom 4954 were enrolled between 1984 and 1985, 668 were enrolled between 1987 and 1991, and 1350 were enrolled between 2001 and 2003. Briefly, the MACS participants have study visits every 6 months. Each visit includes a detailed interview, physical examination, and collection of biological specimens. Adherence to each component of the HAART regimen was assessed by a 4-day recall of the number of doses of the medication taken. Institutional review boards at each site approved the study protocol, and each participant provided written informed consent.
For this analysis, we evaluated all men who had fasting (≥8 hours) blood specimens obtained at visit 41 (April-September 2004). Covariates included age, study site, race (white, black, Latino, other), smoking status (current, former, never), active hepatitis B, chronic hepatitis C, C-reactive protein (<1, 1-3, and >3 mg/L), and lipid-lowering medications (classified as those directed at triglyceride-rich lipoproteins: fenofibrate, gemfibrozil, and niacin; other: primarily statins; or none). Men with missing data for any of the covariates were not included in the analyses. Additionally, HAART initiators whose current treatment was not consistent with the HAART definition for at least 6 months were excluded.
There were 1465 men with fasting blood specimens tested for NMR lipids at visit 41, and 325 of the 1465 were excluded due to missing covariates. Of the remaining 1140 men with complete data, 2 men who used combination antiretroviral therapy in the last 6 months were excluded from HIV+ HAART-naive group and 56 men who did not use HAART in the last 6 months were excluded from the HAART initiator group. Finally, a total of 1082 men were included in the analysis: 609 HIV seronegative, 77 HIV-1 infected HAART naive, and 396 HIV-1 infected HAART treated.
Measurements
Blood samples were centrifuged within 6 hours of collection and frozen at -70°C until analyzed. Lipoprotein particle concentrations were measured by automated NMR spectroscopic assay (LipoScience Inc, N C) on freshly thawed plasma samples as previously described.9 Particle concentrations of lipoprotein subclasses of different sizes were directly obtained from the measured amplitudes of their spectroscopically distinct lipid methyl group NMR signals. Weighted average lipoprotein particle sizes are derived from the sum of the diameter of each subclass multiplied by its relative mass percentage based on the amplitude of its methyl NMR signal. Concentrations [nmol/L for LDL-p and VLDL-p and _mol/L for high-density lipoprotein particles (HDL-p)] of the following subclasses were analyzed for this study: small LDL (18.0-21.2 nm), large LDL (21.2-23 nm), total very low density lipoprotein (VLDL), and total high-density lipoprotein. These lipoprotein parameters have been shown to be negligibly affected by freezing at -70°C.9 The reproducibility of the NMR-measured lipoprotein particle parameters was determined by replicate analyses of plasma pools. The coefficients of variation (CVs) were <10% for the large and small LDL subclasses and <5% for total VLDL-p and HDL-p.9
Serum lipid measurements were performed by the Heinz Nutrition Laboratory at the University of Pittsburgh, Graduate School of Public Health. Total cholesterol was determined using the enzymatic method of Allain et al22 (CV = 1.3%). HDL-C was determined after selective precipitation by heparin/manganese chloride and removal by centrifugation of VLDL and low-density lipoprotein cholesterol (LDL-C) (CV = 2.1%).23 LDL-C was calculated using the Friedewald equation when the triglycerides (TG) level was less than 400 mg/dL.24 LDL-C was measured directly in samples with TG greater than or equal to 400 mg/dL using an automated spectrophotometric assay, LDL Direct Liquid Select, from Equal Diagnostics (CV = 2%).25 The TGs were determined enzymatically using the procedure of Bucolo and David26 (CV = 1.7%).26 CD4+ T-cell counts were determined by flow cytometry per National Institute of Allergy and Infectious Disease Flow Cytometry Quality Assessment Program-certified laboratories.27 Glucose and insulin measurement has been shown to be unaffected by freezing and long-term storage.28,29
The definition of HAART was guided by the DHHS/Kaiser Panel30 guidelines and defined as follows: (a) 2 or more nucleoside reverse transcriptase inhibitors (NRTIs) in combination with at least one PI or one NNRTI (89% of MACS observations classified as HAART); (b) 1 NRTI in combination with at least one PI and at least one NNRTI (6%); (c) a regimen containing ritonavir and saquinavir in combination with 1 NRTI and no NNRTIs (1%); and (d) an abacavir-or tenofovir -containing regimen of 3 or more NRTIs in the absence of both PIs and NNRTIs (4%), except for the 3 NRTI regimens consisting of: abacavir + tenofovir + lamivudine or didanosine + tenofovir + lamivudine. Combinations of zidovudine and stavudine with either a PI or an NNRTI were not considered HAART.
The clinical status of the HIV-1-infected men on HAART was categorized using CD4 cell count and HIV-1 RNA into 3 groups: good (n = 220), defined as plasma HIV-1 RNA < 50 copies/mL and CD4+ T-cell count >350 cells/mm3; poor (n = 16), defined as plasma HIV-1 RNA >1000 copies/mL and CD4 < 200 cells/mm3; or intermediate (n = 160), which included the remaining HAART-treated men who did not meet criteria for good or poor.
Statistical Analyses
The effects of HIV infection, HAART status, and other covariates on median lipid particle counts were examined using quantile regression (PROC QUANTREG, SAS software version 9; SAS Institute Inc, Cary, NC; http://support.sas.com/rnd/app/papers/quantreg.pdf ). This was because the model checking in ordinary least squares regressions showed that the residuals were not normally distributed and not constant, indicating that the ordinary least squares regression was not appropriate for use in these analyses. Unlike ordinary least squares regression, quantile regression makes no assumption about the distribution of error term in the model and is robust to outliers and provides regression coefficients for the conditional quantiles of an outcome variable. Therefore, we used multivariate median regression model to assess the relationship between the median lipid particle count and individual predictor variables.
The predictor variables that were included in the model were selected from prior knowledge of their association with lipids (age, ethnicity, smoking, hepatitis, and lipid-lowering medication) or because of their association with HIV status (C-reactive protein). All variables were retained in the model to obtain measures of the associations of interest without confounding by the other covariates. The associations of NMR lipids were assessed both without and with adjustment for body mass index (BMI), insulin, and diabetes to determine if these factors were potentially mediators of observed differences in lipoproteins. For the analyses of specific antiretroviral therapy types, the analysis was adjusted for CD4 cell count and HIV-1 RNA in addition to the other variables.
|
|
| |
| |
|
|
|
