Cardiovascular risk score change in HIV-1-infected patients switched to an atazanavir-based combination antiretroviral regimen

Authors


Dr Manuela Colafigli, Institute of Clinical Infectious Diseases, Catholic University of Sacred Heart, 1 Largo F. Vito, 00168 Rome, Italy. Tel: +39 0630154945; fax: +39 063054519; e-mail: manuela76@inwind.it

Abstract

Background

We aimed to establish whether the limited impact of atazanavir on the plasma lipid profile could translate into a reduction in the predicted cardiovascular risk in antiretroviral (ARV)-experienced patients switching to an atazanavir-containing regimen.

Methods

HIV-1-infected treatment-experienced patients, switched to atazanavir for whatever reason and without prior major cardiovascular events, were selected and followed for at least 1 month. An individual cardiovascular risk score (10-year risk of major cardiovascular events) based on validated events and measurable risk factors in Italian cardiovascular cohorts was calculated using software available online.

Results

A total of 197 patients were selected for inclusion in the study. After switching to atazanavir, the mean changes from pre-switch to last available measurement were −6.5% (P<0.001) for total cholesterol, −1.7% (P=0.029) for high-density lipoprotein (HDL) cholesterol, −11.3% (P<0.001) for non-HDL cholesterol and −8.6% (P<0.001) for triglycerides. The crude cardiovascular risk score was reduced from 3.43 to 3.38% (P=0.51); the analysis normalized by age showed a reduction from 3.43 to 3.14% (P<0.001). Subsets of patients with high baseline total cholesterol or triglycerides showed more marked reductions.

Conclusions

A treatment switch to atazanavir caused significant reductions in plasma lipids and a modest but significant reduction in the normalized-for-age cardiovascular risk score. Efforts should be made to concomitantly reduce the other preventable cardiovascular risk factors.

Introduction

The introduction of highly active antiretroviral therapy (HAART) for the treatment of HIV infection has dramatically improved the natural history of the disease and its prognosis [1]. Nevertheless, increased levels of triglycerides, total cholesterol and low-density lipoprotein (LDL) cholesterol, reduced levels of high-density lipoprotein (HDL) cholesterol, unfavourable changes in the HDL:total cholesterol ratio and fat redistribution are common side effects in patients receiving ARV drugs [2–6], even in the paediatric population [7]. The data collection on adverse events of anti-HIV drugs (DAD) study showed that, in addition to exposure to HAART, unfavourable cholesterol and triglyceride levels are independent predictors of myocardial infarction [8]. In particular, protease inhibitor (PI)-based therapy has been more often associated with unfavourable lipid profiles than PI-sparing regimens [9,10]. Moreover, both HIV infection per se and HAART could increase the cardiovascular risk by causing endothelial dysfunction, through an enhanced production of adhesion molecules and increased thrombogenesis [11–15].

Given this scenario, and in the context of increased confidence in the achievement of the primary goals of antiretroviral therapy (ART) (viral suppression, immune restoration and a decreased probability of AIDS-related events and death), there is growing concern regarding the possibility of increased cardiovascular risk in HIV-infected patients receiving ART. In a large prospective investigation on the relationship between the use of ART and cardiovascular risk, an increased overall estimated 3-year risk of myocardial infarction was shown in HIV-infected patients who had never had a myocardial infarction, and the increase in the risk was more evident in treatment-experienced than in treatment-naïve patients [16]. Moreover, the researchers found an adjusted relative cardiovascular risk rate of 1.26 per additional year of exposure to HAART [8]. The relative rate of myocardial infarction per year of exposure to PI-based therapy was found to be higher than the relative risk per year of exposure to nonnucleoside reverse transcriptase inhibitor (NNRTI)-based therapy, even after adjustment for other known cardiovascular risk factors not associated with HAART, and this result was not fully explained by lipid changes [17].

There is therefore a need to develop strategies to prevent a potential increase in cardiovascular risk in patients receiving ART. Together with lifestyle interventions and use of lipid-lowering agents [18,19], switching strategies to preserve the efficacy of HAART regimens while reducing their metabolic side effects are studied.

Atazanavir has been shown to have a less severe impact on lipid and glucose metabolism than several other PIs [9,10,20–26]. A previous observational study [25] and a randomized switch study (the SWAN study) [27] showed significant decreases in total, HDL and non-HDL cholesterol, a higher number of patients with lipid parameters above the consensus thresholds for pharmacological interventions and a lower number of patients receiving lipid-lowering agents, in patients treated with ritonavir (RTV)-boosted or unboosted atazanavir-based HAART.

The aim of our study was to evaluate the metabolic impact of switching to atazanavir-based regimens in clinical practice on plasma lipid levels and on predicted cardiovascular risk.

Methods

Patients' characteristics

Patients from a single tertiary care centre were selected according to the following criteria: being on HAART (three or more antiretrovirals in combination) for at least 6 months, switching therapy to an atazanavir-based regimen for whatever reason, age between 35 and 69 years, no prior history of cardiovascular events, and continuing the atazanavir-based regimen for at least 1 month with at least one laboratory follow-up. Regimens containing RTV-boosted atazanavir as well as unboosted atazanavir used at the recommended doses were considered. Patients who simply added atazanavir to their previous regimen for the purpose of intensification, as well as those who concomitantly took another PI at a therapeutic dosage and those who started the atazanavir-based regimen after a treatment interruption of more than 2 years, were excluded from the analysis.

The patient's age and sex, common cardiovascular risk factors such as smoking status, blood pressure, diabetes and concomitant medications (anti-hypertensive or glucose – or lipid-lowering agents), the antiretroviral regimen before and after switch to atazanavir, and viro-immunological and biochemical parameters were recorded as routine practice and collected for all patients at baseline (date of last laboratory values while receiving the previous HAART regimen) and at the last follow-up visit available while on an atazanavir-based regimen before 31 July 2006.

Cardiovascular risk score

The cardiovascular risk score was calculated using software, available online, produced by the Italian National Institute of Health (http://www.cuore.iss.it) based on the observation of 971 validated cardiovascular events during more than 170 000 patient-years of follow-up (PYFU) of initially event-free individuals belonging to multiple Italian cardiovascular cohorts (CUORE project). This score can only be used for women and men without any previous history of major cardiovascular events (myocardial infarction or strokes) and whose age is between 35 and 69 years, and allows the risk assessment of a first major cardiovascular event over the following 10 years based on age, gender, smoking status, diagnosis of diabetes, systolic blood pressure, use of anti-hypertensive drugs and plasma levels of total and HDL cholesterol. As it is not possible to calculate the score for extreme risk factors, total cholesterol plasma level must not be lower than 130 mg/dL or higher than 320 mg/dL, plasma HDL cholesterol must not be lower than 20 mg/dL or higher than 100 mg/dL, and systolic blood pressure must not be lower than 90 mmHg or higher than 200 mmHg.

The score was calculated at baseline, while on the pre-switch regimen (t0), and at the last available follow-up while on the atazanavir-based regimen (t1). In order to try to normalize the risk estimate for increasing age, the t1 score was calculated using two methods, the first considering the patient's actual age at t1 (crude score) and the second assuming that the patients' age is equal to that at t0 (normalized score).

Statistical analysis

The differences between variables at baseline and follow-up were analysed using the t-test for dependent samples in the case of continuous variables and the χ2 test in the case of categorical variables; the predictive factors associated with changes in the parameters were investigated by univariate and multivariate linear regression analysis.

Results

Patients' baseline characteristics

We collected data for 197 patients fulfilling the inclusion criteria. Their median age was 44 years [interquartile range (IQR) 41–48]; 62.9% were male, 75.3% were smokers, 12.8% had diabetes, 16.5% were prescribed anti-hypertensive therapy, and 14.7% concomitantly took at least one lipid-lowering agent. At baseline, the median viral load was 50 HIV-1 RNA copies/mL (IQR 50–830 copies/mL), the median CD4 count was 444 cells/μL, the mean plasma triglyceride level was 270 mg/dL [95% confidence interval (CI) 233–307], the mean total cholesterol level was 191 mg/dL (95% CI 185–198), the mean HDL-cholesterol level was 42 mg/dL (95% CI 40–43), and the mean non-HDL cholesterol level was 169 mg/dL (95% CI 161–176).

Previous therapy was PI-based for 56.9% of patients (50.5% boosted with RTV and 6.4% unboosted), NNRTI-based for 27.9% [of which 52.7% were treated with nevirapine (NVP) and 47.3% were treated with efavirenz (EFV)], NNRTI+PI-based for 6.6% of patients and nucleoside reverse transcriptase inhibitor (NRTI)-only-based for 8.1% of patients. The initial atazanavir regimen was boosted with RTV in 90.9% of patients and unboosted in 9.1%; during follow-up, 4.1% of patients were switched from a boosted to an unboosted atazanavir regimen and 3.6% from an unboosted to a boosted atazanavir regimen. Concomitant treatment with NNRTI was administered in 5.1% of patients (40% NVP and 60% EFV), while 58.9% changed at least one component of the nucleoside backbone at the moment of switch [in particular, 30.5% of all patients discontinued stavudine (d4T) and 22.3% switched from d4T to tenofovir]. Before starting the atazanavir-based regimen, 16% of patients experienced a drug holiday for up to 2 years for various reasons. The median time of observation for the atazanavir-based regimen was 14.8 months (IQR 8.8–19.3 months).

Changes in metabolic parameters and cardiovascular risk score

The pre-switch and the last available plasma lipid levels on the atazanavir-based regimen are shown in Fig. 1a. The absolute mean change in total cholesterol from baseline was −17.7 mg/dL [standard deviation (SD) 43.6; P<0.001], corresponding to a relative mean change of −6.5%; the absolute mean change in non-HDL cholesterol was −16.3 mg/dL (SD 44.4; P<0.001), corresponding to a relative mean change of −11.3%, and the mean change in HDL-cholesterol level was −1.6 mg/dL (SD 9.5; P=0.029), corresponding to relative change of −1.7%. The mean change in triglycerides was −74.8 mg/dL (SD 214; P<0.001), corresponding to relative change of −8.6%.

Figure 1.

 Lipid levels before switching to atazanavir and at the last follow-up on switch to atazanavir. Results are expressed in absolute values (a) and in proportion of patients overcoming the NCEP-ATPIII thresholds for intervention (b). TC, Total cholesterol; TG, triglycerides.

The proportion of patients with plasma lipid levels exceeding the thresholds indicated by the National Cholesterol Evaluation Program-Adult Treatment Panel III (NCEP-ATPIII) [28] for intervention at baseline and follow-up is shown in Fig. 1b. There was a significant decrease in patients with triglyceride plasma levels higher than 200 mg/dL (P=0.0045), with total cholesterol above 240 mg/dL (P=0.048) and with non-HDL cholesterol higher than 190 mg/dL (P=0.0059), whereas the proportion of patients with triglyceride plasma levels higher than 400 mg/dL showed a trend towards a significant reduction (P=0.052). The proportion of patients with HDL-cholesterol levels lower than 40 mg/dL did not change significantly.

After a median follow-up time of 14.8 months (IQR 8.8–19.3) on atazanavir-based therapies, the cardiovascular risk score (10-year estimated risk of cardiovascular events) did not significantly change from pre-switch (P=0.51). When the analysis was normalized by age, a significant reduction was observed (P<0.001; see Fig. 2).

Figure 2.

 Mean change of cardiovascular risk score (% of 10-year risk of major cardiovascular events) at baseline and at the end of the follow-up before (crude score) and after (adjusted score) adjusting for age.

Predictive factors for lipid changes and cardiovascular risk score change

Predictive factors for the magnitude of the change in the cardiovascular risk score are summarized in Table 1. In univariate analysis, the factors showing a significant predictive value with regard to reductions in the cardiovascular risk score on the atazanavir-based regimen were higher baseline total cholesterol and non-HDL cholesterol, higher baseline triglyceride levels, higher baseline cardiovascular risk score and the use of anti-hypertensive therapy, while the use of NRTI-only-based HAART regimens prior to the switch to atazanavir and female gender were associated with an increased cardiovascular risk score after switching to atazanavir. Age, smoking status, diagnosis of diabetes, use of PI- or NNRTI-based therapy prior to switch, change of the nucleoside backbone, switch from d4T to tenofovir, use of boosted or unboosted atazanavir in the follow-up and duration of exposure to the atazanavir-based regimen did not significantly predict differences in the cardiovascular risk score change (data not shown). Multivariate linear regression showed that a higher baseline cardiovascular risk score and higher baseline total cholesterol and triglycerides were the only independent predictors of a greater reduction in the cardiovascular risk score after switching to atazanavir. In a separate multivariable model, baseline total cholesterol levels were replaced by baseline non-HDL cholesterol levels, and this factor was also independently predictive of a greater reduction in the score (per 10 mg/dL higher, mean change −0.05%, 95% CI −0.08 to −0.02; P=0.004).

Table 1.   Factors predicting differences in change of cardiovascular risk score after switching to atazanavir-containing regimens
VariableUnivariable analysisMultivariable analysis
Mean (95% CI) difference
in cardiovascular risk score
change (%)
P valueMean (95% CI) difference
in cardiovascular risk score
change (%)
P value
  1. CI, confidence interval; HDL, high-density lipoprotein; NC, not calculated; NRTI, nonnucleoside reverse transcriptase inhibitor; TC, total cholesterol; TG, triglycerides.

Sex (female vs. male)+0.54 (+0.19; +0.89)0.003+0.11 (−0.22; +0.45)0.509
Anti-hypertensive therapy−0.59 (−1.06; −0.12)0.015NC 
TC at baseline (per 10 mg/dL higher)−0.09 (−0.12; −0.06)<0.001−0.042 (−0.077; −0.008)0.017
Non-HDL cholesterol at baseline (per 10 mg/dL higher)−0.08 (−0.11; −0.06)<0.001NC 
TG at baseline (per 100 mg/dL higher)−0.18 (−0.23; −0.12)<0.001−0.10 (−0.16; −0.036)0.002
NRTI only at baseline+0.65 (+0.03; +1.28)0.04+0.41 (−0.13; +0.94)0.137
Cardiovascular risk score at baseline (per 1% higher)−0.13 (−0.17; −0.08)<0.001−0.08 (−0.12; −0.03)0.002

Predictors of the magnitude of the change in lipid levels are summarized in Tables 2 and 3. The changes in total cholesterol plasma levels were significantly predicted in the univariate analysis by the baseline levels of total cholesterol, non-HDL cholesterol and triglycerides, prior NRTI-only-based HAART, the switch from d4T to tenofovir, and female gender. Age, use of boosted or unboosted PI- or NNRTI-based therapy prior to switch, changes in the nucleoside backbone, withdrawal of d4T, use of boosted or unboosted atazanavir in the follow-up and duration of exposure to the atazanavir-based regimen did not predict differences in plasma total cholesterol levels (data not shown). In multivariate analysis, baseline total cholesterol levels and triglyceride plasma levels, prior NRTI-only-based HAART and switch from d4T to tenofovir remained independently predictive of the magnitude of cholesterol change. Again, in a separate analysis in which baseline total cholesterol level was replaced by non-HDL cholesterol level, the latter was also independently predictive of total cholesterol changes (per 10 mg/dL higher, mean change −3.5 mg/dL; 95% CI −4.8 to −2.1; P<0.001).

Table 2.   Factors predicting differences in change in total cholesterol plasma levels after switching to atazanavir-containing regimens
VariableUnivariate analysisMultivariate analysis
Mean (95% CI) difference in
total cholesterol change (mg/dL)
P valueMean (95% CI) difference in
total cholesterol change (mg/dL)
P value
  1. CI, confidence interval; HDL, high-density lipoprotein; NC, not calculated; NRTI, nonnucleoside reverse transcriptase inhibitor; TC, total cholesterol; TG, triglycerides.

Sex (female vs. male)+16.1 (+3.5; +28.7)0.013+6.6 (−4.4; +17.5)0.238
TC at baseline (per 10 mg/dL higher)−4.2 (−5.2; −3.1)<0.001−3.1 (−4.3; −2.0)<0.001
Non-HDL cholesterol at baseline (per 10 mg/dL higher)−4.6 (−5.7; −3.4)<0.001NC 
TG at baseline (per 100 mg/dL higher)−6.6 (−8.7; −4.5)<0.001−2.9 (−5.2; −0.7)0.011
NRTI only at baseline+40.9 (+19.1; +62.6)<0.001+25.2 (+6.2; +44.1)0.010
Switch of stavudine to tenofovir−17.9 (−32.4; −3.4)0.016−14.0 (−26.2; −1.7)0.026
Table 3.   Factors predicting differences in change in triglyceride plasma levels after switching to atazanavir-containing regimens
VariableUnivariate analysis mean
(95% CI) difference in
triglyceride change
P valueMultivariate analysis mean
(95% CI) difference in
triglyceride change
P value
  1. CI, confidence interval; HDL, high-density lipoprotein; NC, not calculated; NRTI, nonnucleoside reverse transcriptase inhibitor; PI, protease inhibitor; PI/r, protease inhibitor boosted with ritonavir; TC, total cholesterol; TG, triglycerides.

Sex (female vs. male)+69.5 (+8.1; +130.9)0.027−13.5 (−48.2;+21.2)0.443
TC at baseline (per 10 mg/dL higher)−18.5 (−23.5; −13.4)<0.001−1.1 (−4.7; +2.5)0.556
Non-HDL cholesterol at baseline (per 10 mg/dL higher)−21.7 (−27.5; −15.9)<0.001NC 
TG at baseline (per 100 mg/dL higher)−67.3 (−73.5; −61.2)<0.001−66.3 (−73.5; −59.0)<0.001
PI at baseline−61.8 (−123.5; 0.0)0.050NC 
PI/r at baseline−64.4 (−124.1; −4.7)0.0352.5 (−31.8; +36.8)0.885
NRTI only at baseline+126.4 (+18.0; +234.8)0.02382.1 (+18.8; +145.4)0.011

Finally, in the univariate analysis, changes in triglyceride levels were associated with baseline total and non-HDL cholesterol, triglycerides, use of PI-based and RTV-boosted PI-based HAART regimens prior to the switch to atazanavir, NRTI-only-based HAART regimens prior to atazanavir and female gender; among these factors, only baseline triglycerides and use of NRTI-only-based therapy at baseline retained statistical significance in their capacity to predict the magnitude of triglyceride change in multivariate analysis (see Table 3). Age, NNRTI-based therapy prior to switch, changes in the nucleoside backbone, withdrawal of d4T, switch from d4T to tenofovir, use of boosted or unboosted atazanavir in the follow-up and time of exposure to the atazanavir-based regimen did not significantly predict different changes in magnitude in plasma triglyceride level (not shown).

Changes in the cardiovascular risk score in subgroups with high lipid levels or cardiovascular risk factors

In a separate analysis, we analysed the changes in the cardiovascular risk score in subgroups with high baseline lipid levels.

Among patients with a baseline plasma total cholesterol value higher than 240 mg/dL (n=37; median baseline cardiovascular risk score 3.8%; IQR 2.3–5.5%), the mean change in the predicted 10-year cardiovascular risk was −0.53% (SD 1.59%; P=0.05) using the crude score and −0.79% (SD 1.71%; P=0.008) using the normalized score. After selection of patients with baseline plasma triglyceride levels above 400 mg/dL (n=21; median baseline cardiovascular risk score 3.2%; IQR 1.8–7.6%), the crude score was reduced by a mean of 1.03% (SD 1.85%; P=0.019), while the normalized score showed a mean reduction of 1.40% (SD 1.99%; P=0.004).

Use of lipid-lowering agents

At baseline, 14.7% of patients (29 of 197) were receiving lipid-lowering agents concomitantly with ART; at the end of follow-up, the proportion of patients on lipid-lowering therapy was not significantly changed (12.2%; 24 of 197), as only six patients (3%) stopped lipid-lowering drugs and one patient (0.5%) was newly started on such therapy.

Discussion

Increasing attention is being paid by the HIV-treatment specialists and community to the metabolic side effects of HAART and to the potential increase in cardiovascular risk.

In this study, we aimed to investigate the potential protective role of atazanavir with regard to cardiovascular risk, based on its more favourable profile in terms of patients' lipid and glucose metabolism.

Our study confirms previous reports [20–27] indicating that switching a patient's HAART regimen to an atazanavir-based scheme results in a significant reduction in the plasma levels of total cholesterol, non-HDL cholesterol and triglycerides.

Results regarding lipid levels can also be considered comparable to those of the US Early Access Programme for atazanavir [25], although we found more limited reductions in triglycerides and total cholesterol, which is consistent with the fact that 91% of our patients received RTV-boosted atazanavir, that some concomitantly received NNRTI and that a proportion of patients were switched from an NRTI-based regimen.

The baseline estimated 10-year cardiovascular risk was 3.43%, which is very close to the 3.5 per 1000 patient-year incidence of myocardial infarction observed in the DAD study [8]. At the end of follow-up, the cardiovascular risk score of the patients was not significantly reduced. Nonetheless, patients were more than 1 year older than at baseline and this age-related effect could conceal the possible benefits of the therapeutic change. In order to try to avoid this bias, we calculated a ‘normalized’ score assuming that the patient's age at the end of follow-up was the same as at baseline and using all the end of follow-up parameters for the other components of the score. After this adjustment, the improvement in the cardiovascular risk score reached statistical significance, showing that the therapy switch could be effective in reducing the risk. The normalized-for-age cardiovascular risk score was reduced by 0.3%. The change was modest, but if we consider the ratio between this change and the mean cardiovascular risk score at baseline, it represents a relative change of 9%.

The minor improvement in the cardiovascular risk score, despite the positive effects on lipid metabolism, could reflect various difficulties: the analysed patients were quite young, and co-morbidities (such as diabetes and hypertension) were rare, so that individual cardiovascular risk scores were generally low at baseline; the risk score did not include triglyceride levels, so that the improvement in this parameter did not result in an improvement in the score. Moreover, 34.5% of the patients were switched from an NNRTI-containing regimen (27.9% without and 6.6% with a concomitant PI) and the role of NNRTIs in influencing cardiovascular risk is probably more limited [17].

In multivariate analysis, higher baseline total cholesterol and triglyceride plasma levels were significant predictors of more marked reductions in lipid plasma levels and adjusted-for-age cardiovascular risk score. Consistent with this, in these patients significant reductions in both the crude and the adjusted scores were observed. This finding underscores the importance of a switch to atazanavir for patients with significant dyslipidaemia in order to reduce their cardiovascular risk.

We also found that the switch of the NRTI backbone from a d4T- to a tenofovir-containing regimen was related to a significant reduction in plasma total cholesterol levels. This result is in agreement with previously published studies in the adult as well as in the paediatric population [29,30] and supports the concomitant switch of the backbone, when necessary and possible.

The present study has some limitations: because of its observational and retrospective design, we lacked data on lipodystrophy, and LDL-cholesterol values were available only in a minority of patients.

Finally, the cardiovascular risk score that we used was not designed for HIV-infected patients and this is the first attempt, as far as we know, to use it in such a population. The CUORE project cardiovascular risk score calculator for the assessment of cardiovascular risk was chosen as it has been created and validated in the Italian population: lifestyle in Italy is different from that in countries in which the Framingham and other internationally accepted risk calculators were developed, and a significant difference in the estimate of risk with a consequently different rate of prescription of lipid-lowering agents has been demonstrated, with the Framingham score tending to overestimate the risk in the Italian population [31,32]. Recently, the DAD study group developed a cardiovascular risk equation in its study population [33], but this equation still needs to be validated in an independent set of patients.

In conclusion, the results of our study support a treatment switch to an atazanavir-based regimen in order to try to reduce cardiovascular risk in patients with high lipidaemia while on HAART. The modest advantage conferred by this strategy was partially a result of the low baseline risk of the analysed population, but also emphasizes that this strategy should be accompanied by efforts to address other modifiable risk factors in order to try to obtain a greater reduction in cardiovascular risk.

Ancillary