Back grey_arrow_rt.gif
Low-density lipoprotein size and lipoprotein-associated phospholipase A2 in HIV-infectedpatients switching to abacavir or tenofovir
  Download the PDF here

Antiviral Therapy 2011; 16:459-468

"studies in the non-HIV-infected population have found an association between the predominance of sd-LDL particles and an increased risk of clinical and subclinical coronary and non-coronary disease [15,16]. sd-LDL particles promote inflammation in the subendothelial space by several mechanisms [32]. Smaller LDL particles can penetrate the vascular wall more easily than larger LDL particles and expose their cholesterol and other lipids to the pro-oxidant processes occurring there; oxidized lipids are potent proinflammatory agents [33].....Triglyceride concentrations are the major determinant of sd-LDL synthesis; accordingly, hypertriglyceridemic conditions such as insulin resistance, metabolic syndrome and obesity are associated with higher numbers of sd-LDL particles [15,16] increase in sd-LDL abundance is known to occur when a threshold TG concentration is exceeded, which has been described at 1.36-1.70 mmol/l"

Maria Saumoy1, Jordi Ordonez-Llanos2,3, Esteban Martinez4, Patricia Barragan1, Esteban Ribera5, Rosa Bonet2,3, Hernando Knobel6, Eugenia Negredo7, Montserrat Lonca4, Adrian Curran5, Josep Maria Gatell4, Daniel Podzamczer1,*, the Bicombo-met Substudy Team

1Infectious Disease Service, Hospital Universitari de Bellvitge, Hospitalet de Llobregat, Barcelona, Spain
2Biochemistry Department, Hospital Sant Pau, Barcelona, Spain
3Biochemistry Department, Universitat Autonoma, Barcelona, Spain
4Infectious Disease Service, Hospital Clinic, Barcelona, Spain
5Infectious Disease Service, Hospital Universitari, Barcelona, Spain
6Vall d'Hebron, Infectious Disease Service, Hospital del Mar, Barcelona, Spain
7HIV Unit, Lluita contra la Sida Foundation, Hospital Germans Trias i Pujol, Badalona, Spain
*Corresponding author e-mail:


Background: The aim of this study was to assess changes in the size and cholesterol content of low-density lipoproteins (LDL) and changes in lipoprotein-associated phospholipase A2 (Lp-PLA2) activity in HIV-infected patients switching to tenofovir + emtricitabine (TDF+FTC) or abacavir + lamivudine (ABC+3TC).

Methods: This was a substudy of a multicentre randomized trial comparing TDF+FTC with ABC+3TC-based regimens in patients with virological suppression. Fasting lipids and apolipoproteins (apo), LDL size and cholesterol content and Lp-PLA2 activity were measured at baseline and at week 48.

Results: A total of 62 patients, naive for the compared drugs, were included. At baseline, groups were comparable except for total Lp-PLA2 activity (P=0.047) and for a tendency towards the use of a major baseline thymidine analogue in the TDF+FTC arm (25 versus 18 patients; P=0.054). In the ABC+3TC arm a significant increase in total cholesterol (0.64 mmol/l; P=0.003), high-density lipoprotein cholesterol (HDL-c, 0.13 mmol/l; P=0.031), triglycerides (0.39 mmol/l; P=0.036), apo A-I (0.12 g/l; P=0.006), apo B (0.16 g/l; P=0.015) and non-HDL-c (0.50 mmol/l; P=0.009) concentrations was observed at week 48 compared with the TDF+FTC treatment arm. In addition, an increase in the cholesterol content of small, dense LDL subfractions (0.48 mmol/l; P=0.003) and a decrease in LDL size (-2.6 nm; P=0.011) was observed in the ABC arm without changes in the TDF patients. Total PLA2, LDL-PLA2 and HDL-PLA2 activity decreased in the TDF arm, but multivariate analysis showed baseline PLA2 values and previous use of thymidine analogues as the factors associated with these changes. Estimated cardiovascular risk did not change in either arm.

Conclusions: A more atherogenic LDL profile, including a decrease in LDL size, was found in the ABC group and not in TDF patients.


Cardiovascular disease (CVD) risk is increased in HIV-infected patients compared with the general population. The increased risk appears to be multifactorial, involving not only the traditional cardiovascular risk (CVR) factors but also the proinflammatory state associated with HIV infection and other coinfections and the use of some antiretroviral (ARV) drugs [1,2]. Two recent observational studies have related abacavir (ABC) use with an increased risk of ischaemic CVD [3,4]. To date, only one randomized trial (not designed for that purpose) has shown an increase in cardiovascular events in ABC recipients [5]. Tenofovir (TDF) has not been associated with increased CVR.

The mechanism by which ABC might increase the risk of CVD is unknown. Some randomized studies [5,6], including the general Bicombo study [7], have shown a less favourable lipid profile in pretreated patients switching to ABC compared with TDF. In ARV-naive patients, ABC was associated with a worse lipid profile compared with TDF in one study [8], but not in another [9]. Some authors have attributed the increased CVR to a proinflammatory and procoagulant mechanism promoted by ABC [4,10,11]. Nonetheless, the findings from other reports have not supported this idea in either ARV-naive patients [12] or in stable ARV-pretreated patients [13], including a substudy of Bicombo [14].

The CVR due to lipid disturbances might not be related solely to quantitative lipid changes; qualitative changes in the lipoproteins could also be implicated. To investigate this possibility, we analysed low-density lipoprotein (LDL) particle size and lipoprotein-associated phospholipase A2 (Lp-PLA2) activity, both of which have been associated with increased CVR in the general population [15-17]. LDL particles are heterogeneous, differing in size and density. LDL particles with a smaller size and higher density, the so-called small, dense LDL (sd-LDL) subfractions, are associated with increased CVR [15,16]. Lp-PLA2 is an enzyme with known proinflammatory properties: oxidizing LDL phospholipids turns them into potent inflammatory phospholipid derivatives. Circulating Lp-PLA2 is mainly found in association with LDL particles, particularly sd-LDL, and to a lesser extent with high-density lipoprotein cholesterol (HDL-c) particles [17]. Although some authors have attributed a pro- and anti-inflammatory role to the PLA2 associated with LDL cholesterol (LDL-c) and HDL-c, respectively, the relationship between Lp-PLA2 and CVD has been demonstrated for the total circulating activity [17,18]. To date, there is little information regarding the effects of ARV therapy on LDL subfractions [19-22], whereas Lp-PLA2 has not been evaluated.

The aim of this study was to evaluate qualitative lipid changes, including LDL size and subfractions, Lp-PLA2 activity and quantitative lipid changes occurring in stable HIV-infected patients randomly switching to abacavir + lamivudine (ABC+3TC) or tenofovir + emtricitabine (TDF+FTC). We hypothesized that if ABC use was associated with a more proatherogenic and proinflammatory LDL profile, the characteristics of LDL might have an influence on the possible association between ABC and increased CVR.


Study design and patient population

Bicombo-met is a substudy of Bicombo, a randomized, open-label, multicentre trial carried out in 18 Spanish hospitals. The inclusion criteria have been described elsewhere [7]. Patients were centrally randomized 1:1 to change both nucleoside reverse transcriptase inhibitors (NRTI) to the fixed-dose nucleoside analogue combination TDF+FTC or ABC+3TC, while continuing with the same protease inhibitors (PI) or non-nucleoside reverse transcriptase inhibitors (NNRTI). Patients taking ABC or TDF as part of the NRTI backbone were candidates for recruitment in the main study, but were excluded from this substudy. Patients for Bicombo-met were enrolled in five hospitals in Barcelona.

The primary endpoints of the study were the changes in lipid parameters, particularly changes in the size and cholesterol content of LDL-c and in Lp-PLA2 activity in the two arms after 48 weeks of treatment. The secondary endpoints were the changes occurring in the classical CVR factors in the same period and in the CVR estimated with the Framingham equation.

Clinical assessment

Clinical assessment and laboratory evaluations were performed at baseline and after 48 weeks on the assigned therapy. At each visit, we recorded the use of treatments other than ARV therapy, including lipid-lowering, antidiabetic and antihypertensive drugs. CVR factors were evaluated according to the National Cholesterol Education Program guidelines [23]. The estimated CVR over the next 10 years was calculated with the Framingham equation [24].

Anthropometric measurements, including height, weight and waist and hip circumferences, were evaluated at each visit. The body mass index (BMI) and waist-to-hip ratio were calculated.

Laboratory methods

Venous blood samples were obtained after an overnight fast in glass tubes containing gelose with no additives (BD Vacutainer SST, Franklin Lakes, NJ, USA). Tubes were immediately centrifuged and sent on dry ice to a central laboratory for lipid measurements, which were performed no later than 48 h after blood drawing. Total cholesterol (TC) and triglyceride (TG) concentrations, the latter corrected for free glycerol, and glucose were measured by standard enzymatic methods; apolipoproteins B (apo B) and A-I (apo A-I) were measured by immunoturbidimetry and HDL-c by a direct, homogeneous method. All assays were from Roche Diagnostic (Basel, Switzerland). TC:HDL and apo A-I:apo B ratios and non-HDL-c were calculated. LDL-c was calculated by the Friedewald formula when TG concentrations were <3.39 mmol/l or measured after separation of lipoproteins by ultracentrifugation at 105,000 g for 18 h at 4°C when TG ≥3.39 mmol/l. All measurements were controlled by national and international external quality programs and fulfilled the international recommendations for inaccuracy and total imprecision. LDL size (sd-LDL diameter <26.8 nm; large buoyant LDL particles ≥26.8 nm) was measured by gel electrophoresis in 3% polyacrylamide using a commercial system (Lipoprint Quantimetrix Co., Redondo Beach, CA, USA) [25]. The same method was used to separate nine LDL subfractions (LDL1 to LDL9) and to evaluate their respective cholesterol content. For the purposes of the study, the LDL subfractions were divided into two main subgroups: LDL1 to LDL3, roughly corresponding to large buoyant LDL, and LDL4 to LDL6, representing the smaller denser LDL fractions. Although the method permits measurement of smaller LDL fractions (LDL7 to LDL9), these were present in very small amounts in all samples, representing <5% of the cholesterol in LDL subfractions; thus, these fractions were excluded from the calculations.

PLA2 activity was measured in serum (total PLA2) and in lipoproteins using 2-thio-platelet-activating factor (Cayman Chemical, Ann Arbor, MI, USA) as substrate, as previously described [26]; the activity is expressed as μmol/min-1 xml. To determine the distribution of PLA2 among the lipoproteins, apo B-containing lipoproteins (very low-density lipoprotein, LDL-c and lipoprotein (a)) were precipitated from serum using dextran sulfate. PLA2 activity of the supernatant corresponded to the activity associated with HDL (HDL-PLA2). LDL-PLA2 activity was obtained by subtracting HDL-PLA2 from total PLA2.

Insulin and C-peptide were measured by immunochemiluminescent assays (DPC, Siemens Medical Solutions Diagnostics, Tarrytown, NY, USA). Insulin sensitivity was calculated with the homeostasis model assessment (HOMA) index as the product of the fasting glucose (mmol/l) and insulinaemia (mU/l), divided by 22.5.

Statistical analysis

All endpoints were analysed using an intention-to-treat analysis, which coincides with the as-treated analysis because no changes between arms occurred. Data from patients lost to follow-up were only included in the analysis of baseline characteristics. No sample size was calculated, as the biological variability of the biochemical measures established as the primary endpoint is unknown; therefore, a significance level of 0.05 was considered biologically relevant. Prior to the statistical analyses, the normality of distributions and homogeneity of variances were tested. The Student's t-test or Mann-Whitney U-test were used to compare continuous variables between arms. Qualitative variables were compared using the χ2 or Fisher exact test. Comparisons between baseline and follow-up in each arm were carried out with the paired t-test or Wilcoxon signed rank test. The median change of the difference between TDF+FTC and ABC+3TC and the confidence interval were calculated [27]. Continuous variables are expressed as the median and IQR. The Pearson correlation coefficient was calculated to estimate the strength of association between continuous clinical and laboratory characteristics. Potential predictors of changes in relevant lipidic and LDL size variables were further evaluated for statistical significance by multiple regression analysis. Analysis of covariance was used to test the main and interaction effects of categorical variables on Lp-PLA2 variables, controlling for the effects of the baseline Lp-PLA2 variables. Initially, univariate models were calculated and those variables found to be statistically significant (P-value <0.05) were entered in the multivariate analysis. Analyses were performed using SPSS, version 15.0 (Chicago, IL, USA).


Patient characteristics

A total of 62 patients were included, 31 in each treatment arm. The baseline characteristics of the two groups are described in Table 1. A trend to a more frequent baseline use of thymidine analogues (25 versus 18 patients; P=0.054) and a higher total and LDL-PLA2 activities (39.8 versus 32.1 μmol/min-1 xml, P=0.047; and 25.6 versus 18.4 μmol/min-1 xml, P=0.072, respectively) were found in the TDF+FTC arm compared with the ABC+3TC arm. Of note, only 12.9% in the ABC+3TC group and 6.5% in TDF+FTC received a PI during the study.

At week 48, six patients were lost to follow-up: one in the ABC+3TC group and five in the TDF+FTC group. Only one patient in the ABC+3TC arm required lipid-lowering therapy during the study.

At week 48, waist and hip circumferences decreased (-3.5 cm, P=0.006; and -2.5, P=0.028, respectively) in the TDF+FTC arm, whereas only a trend to a decrease in waist circumference was found in the ABC+3TC arm (-3 cm; P=0.053). No changes in weight, BMI or waist-to-hip ratio were found in either arm.

There were no significant changes in glucose or insulin concentrations or the HOMA index at week 48 in either the ABC+3TC or TDF+FTC arms or between arms (P=0.476, P=0.160 and P=0.625, respectively).

Changes in lipid parameters

The median change in lipid parameters in each therapy arm and between arms after 48 weeks of treatment is shown in Figure 1. Total cholesterol, LDL-c and apo B decreased in the TDF+FTC arm (-0.43 mmol/l, P=0.002; -0.18 mmol/l, P=0.040; and -0.09 g/l, P=0.003, respectively) at week 48; there were no changes in HDL-c, apo A-I or TG. By contrast, in the ABC+3TC arm, a significant increase was observed only in apo A-I (0.13 g/l; P=0.005). These quantitative changes were associated with a significant improvement in non-HDL-c (-0.37 mmol/l; P=0.007) in the TDF+FTC arm. There were no changes in the TC:HDL-c or apo A-I:apo B ratios in either therapy group.

In the analysis of the differences between arms at week 48, the median change of the difference between TDF+FTC and ABC+3TC was significant for TC (0.64 mmol/l; P=0.003), HDL-c (0.13 mmol/l; P=0.031), TG (0.39; P=0.036), apo A-I (0.12; P=0.006), apo B (0.16; P=0.015) and non-HDL-c (0.50; P=0.009) and marginally significant for LDL-c (0.31 mmol/l; P=0.06). In all cases, except for apo A-I and HDL-c, the changes in the TDF+FTC arm were more favourable than those in the ABC+3TC arm.

In the multivariate analysis (adjusted by age, gender and use of thymidine analogues at baseline) ABC+3TC treatment was independently associated with an increase in TC (β=0.643; P=0.003), HDL-c (β=0.206; P=0.003), non-HDL-c (β=0.443; P=0.016), apo B (β=0.092; P=0.031) and apo A-I (β=0.193; P=0.002), whereas a trend was observed with LDL-c (β=0.289; P=0.056) and TG (β=0.701; P=0.068).

In the analysis of LDL particles after 48 weeks of treatment, an increase in cholesterol content of the sd-LDL subfractions (LDL4-6) (+0.5 mmol/l; P=0.003) and a decrease in LDL size (-2.6 nm; P=0.011) were seen in the ABC+3TC arm. By contrast, no significant changes were found in the TDF+FTC arm (Figure 2). The change in LDL size correlated with the change in TG (r=-0.459; P<0.001), HOMA index (r=-0.359; P=0.037), weight (r=-0.352; P=0.021) and BMI (r=-0.322; P=0.035). In the multivariate analysis adjusted by gender, age and baseline thymidine analogue use, only a TG change (β= -2.273, 95% CI -4.402- -0.145; P=0.037) remained as a significant determinant of LDL size.

Regarding Lp-PLA2, at week 48, total, LDL and HDL-PLA2 activities decreased by 10% (P=0.001), 5.6% (P=0.052) and 24% (P=0.014), respectively, in the TDF+FTC arm, whereas no significant changes were observed in the ABC+3TC arm (0%, P=0.475; 11.5%, P=0.241 and 4.9%, P=0.819 for total, LDL- and HDL-PLA2 activity, respectively; Figure 3). Univariate and multivariate analysis of the factor determinants of the change in total, LDL- and HDL-PLA2 activity at week 48 are expressed in Table 2. In multivariate analysis, the baseline concentrations of total, LDL- and HDL-associated PLA2 were significant determinants of PLA2 changes, whereas the prior use of thymidine analogues was associated with the changes in total and LDL-PLA2. No association was found between Lp-PLA2 activity and LDL subfractions or LDL size.

Changes in CVR factors

When CVR factors were analysed separately, there were no significant changes during the study period. Likewise, the estimated CVR did not change during the study (P=0.758 in ABC+3TC and P=0.709 in TDF+FTC).


The most relevant finding of this study in HIV-infected patients was the quantitative and qualitative proatherogenic lipid changes observed in those who switched from the two NRTIs of a stable regimen to a fixed-dose combination of ABC+3TC, as compared with the favourable or neutral lipid profile in patients switching to TDF+FTC.

In the analysis of quantitative changes in each arm at week 48, TDF-treated patients exhibited a better lipid profile than at baseline, with significant decreases in TC, LDL-c and apo B, whereas the only change in ABC-treated patients was an increase in apo A-I. Although TC:HDL-c and apo A-I:apo B ratios did not change in either treatment arm, a significant decrease in non-HDL-c, a parameter summarising all the atherogenic cholesterol fractions, was observed in the TDF treatment arm. In the comparison between arms at week 48, TDF-treated patients showed significantly lower TC, LDL-c, TG, non-HDL-c and apo B concentrations than ABC-treated patients, who only showed an increase in apo A-I and HDL-c concentrations relative to the TDF group. These data indicate an overall favourable effect of TDF on lipid metabolism and a less favourable, mostly proatherogenic effect of ABC. These results are in agreement with the lipid changes observed in the Bicombo study, from which the current substudy was derived [7], and with other switching studies comparing these two non-thymidine nucleoside analogues [5,6].

An association between ABC and an increased risk of coronary artery disease has been recently reported in two large studies [3,4], although it was not found in others [28-30]. It has been suggested that ABC might induce coronary disease through a proinflammatory effect leading to arteriosclerotic plaque instability and rupture [31]. However, circulating proinflammatory biomarkers were found to be increased in some studies [4,10] but not in others [2,13], including a substudy of the Bicombo trial [14]. In that substudy there were no differences between ABC- and TDF-treated patients regarding several biomarkers of inflammation, endothelial dysfunction or hypercoagulability [14]. Thus, the findings of several studies seem to support the idea that lipid changes might represent a CVR factor associated with ABC use [5-7], whereas a proinflammatory effect of ABC remains under discussion. The present study in patients randomly switched to ABC- or TDF-containing regimens is the first to additionally evaluate qualitative lipid changes, including the LDL subfractions, and Lp-PLA2 activity; the predominance of sd-LDL subfractions and high activity of Lp-PLA2 have a high proinflammatory potential.

Studies in the non-HIV-infected population have found an association between the predominance of sd-LDL particles and an increased risk of clinical and subclinical coronary and non-coronary disease [15,16]. sd-LDL particles promote inflammation in the subendothelial space by several mechanisms [32]. Smaller LDL particles can penetrate the vascular wall more easily than larger LDL particles and expose their cholesterol and other lipids to the pro-oxidant processes occurring there; oxidized lipids are potent proinflammatory agents [33]. Our study demonstrated a more atherogenic pattern of LDL subfractions in ABC-treated patients, consisting of increased cholesterol content in sd-LDL fractions and decreased LDL size, whereas no significant changes were found in patients receiving TDF.

Triglyceride concentrations are the major determinant of sd-LDL synthesis; accordingly, hypertriglyceridemic conditions such as insulin resistance, metabolic syndrome and obesity are associated with higher numbers of sd-LDL particles [15,16]. In the present study, we did not observe any change in the insulin parameters at week 48, although waist and hip circumferences improved in TDF-treated patients and not in those receiving ABC. Thus, an improvement in LDL size would be expected in TDF patients; nonetheless, it did not occur. This could be because a more profound improvement in anthropometric measures might be needed to observe a change in LDL size or because the anthropometric measures used lacked specificity for this purpose. Few patients started lipid-lowering drugs or used PIs during the study. Both these factors can modify LDL size, but their influence on LDL particles could not be evaluated because of the small number of cases. Triglyceride concentration followed a decreasing pattern in the TDF arm and an increasing one in patients receiving ABC. As a result, at the end of the study, TDF-treated patients showed significantly lower TG values than ABC-treated patients. After adjusting for age and sex, TG concentration was the only parameter that negatively correlated with LDL size. The decreased LDL size in the ABC arm might be the result of an increase in TG concentrations that did not suffice to achieve statistical significance, although it was large enough to change LDL size. In this regard, an increase in sd-LDL abundance is known to occur when a threshold TG concentration is exceeded, which has been described at 1.36-1.70 mmol/l [34]. In the ABC arm, a median TG value of 1.47 mmol/l was observed at the end of the study; therefore, some patients exceeded the TG threshold and median LDL size decreased. These observations in the HIV population concur with previous findings in persons without HIV infection [15].

To date, few studies have analysed the qualitative changes occurring in LDL particles in HIV patients under HAART [19-22]. Negredo et al. [19] reported that in patients who switched from a PI-containing regimen to nevirapine, LDL-c concentrations and the number of circulating LDL particles decreased, without changes in LDL size. A lopinavir-containing regimen was associated with a reduction in LDL size [20]. In a recent study in patients receiving HAART, an atherogenic lipoprotein phenotype consisting of a higher number of small LDL particles and lower number of HDL and large LDL particles was described [21], a pattern close to the one observed in the ABC arm of our study.

The main physiological action of Lp-PLA2, an enzyme mainly produced by macrophages, is the hydrolysis of strongly inflammatory phospholipids, such as platelet-activating factor. Total circulating Lp-PLA2 activity has been shown to be an independent predictor of coronary heart disease and ischaemic stroke in the general population [17,35]. As Lp-PLA2 circulates mainly in association with LDL, but also with HDL, we analysed not only the total Lp-PLA2 activity but also the activity associated with LDL and HDL, to determine whether Lp-PLA2 response might differ depending on the associated lipoproteins. Patients on TDF showed significant decreases in total, HDL and LDL-Lp-PLA2 activity, whereas no significant changes were observed in the ABC arm. The Lp-PLA2 decreases in the TDF group and the absence of these improvements in the ABC group concur with the change (or its absence) occurring in TC and LDL-c, the lipids most strongly correlated with PLA2 activity [36].

Although some authors have attributed an anti-atherogenic role to HDL-associated Lp-PLA2, it has been shown that an increase of total activity, which is mainly due to LDL-associated Lp-PLA2, is positively related to an increase in adverse cardiovascular outcomes [37]. Thus, in the present study, the changes found in Lp-PLA2 activities indicate an amelioration of this proinflammatory activity in TDF-treated patients and a lack of a similar effect in the ABC-treated group. However, multivariate analyses showed that most PLA2 changes were driven by the baseline PLA2 concentrations and by the use of thymidine analogues at baseline, that is, by switching from these drugs to non-thymidine analogues. Unfortunately, the small sample size for the multivariate analysis does not enable us to conclude whether the observed effects were only due to thymidine analogue use or whether TDF therapy also played a significant role. This fact requires further study.

At the 48-week follow-up, there was no increase in the CVR estimated by the Framingham equation. This was probably because of the short observation time or the increase in HDL-c in the ABC arm.

Our study has some limitations. Firstly, the number of patients included is small. This was in part because a percentage of patients enrolled in the Bicombo study had received ABC or TDF previously and, for this reason, were not included in the present substudy. However, despite the limited sample size, we found significant changes in several closely related lipid parameters, all of which indicated a poorer lipid profile in the ABC arm. Secondly, visceral fat measurement by imaging methods was not carried out in this study. As fat distribution is a strong determinant of LDL size, the lack of refinement of these measurements might have impaired detection of correlations between fat distribution and LDL size.

In conclusion, when the changes in lipid and apolipoprotein concentrations, LDL size and cholesterol transported by sd-LDL were taken together, a favourable lipid profile was associated with TDF use. This was in contrast to the proatherogenic, proinflammatory changes that occurred with ABC use.

  icon paper stack View Older Articles   Back to Top