Cardiovascular disease risk factors in HIV patients - association with antiretroviral therapy. Results from the DAD study

Objective: To determine the prevalence of risk factors for cardiovascular disease (CVD) among HIV-infected persons, and to investigate any association between such risk factors, stage of HIV disease, and use of antiretroviral therapies.

Design: Baseline data from 17 852 subjects enrolled in DAD, a prospective multinational cohort study initiated in 1999.

Methods: Cross-sectional analyses of CVD risk factors at baseline. The data collected includes data on demographic variables, cigarette smoking, diabetes mellitus, hypertension, dyslipidaemia, body mass index, stage of HIV infection, antiretroviral therapy.

Results: Almost 25% of the study population were at an age where there is an appreciable risk of CVD, with those receiving a protease inhibitor (PI) and/or non-nucleoside reverse transcriptase inhibitor (NNRTI) tending to be older. 1.4% had a previous history of CVD and 51.5% were cigarette smokers. Increased prevalence of elevated total cholesterol (≥ 6.2 mmol/l) was observed among subjects receiving an NNRTI but no PI [odds ratio (OR), 1.79; 95% confidence interval (CI), 1.45-2.22], PI but no NNRTI (OR, 2.35; 95% CI, 1.92-2.87), or NNRTI + PI (OR, 5.48; 95% CI, 4.34-6.91) compared to the prevalence among antiretroviral therapy (ART)-naive subjects. Subjects who have discontinued ART as well as subjects receiving nucleoside reverse transcriptase inhibitors had similar cholesterol levels to treatment-naive subjects. Higher CD4 cell count, lower plasma HIV RNA levels, clinical signs of lipodystrophy, longer exposure times to NNRTI and PI, and older age were all also associated with elevated total cholesterol level.

Conclusion: HIV-infected persons exhibit multiple known risk factors for CVD. Of specific concern is the fact that use of the NNRTI and PI drug classes (alone and especially in combination), particularly among older subjects with normalized CD4 cell counts and suppressed HIV replication, was associated with a lipid profile known to increase the risk of coronary heart disease.

Introduction

The combination of three or four drugs from any of the three available classes that can inhibit the replication of HIV [nucleoside analogue reverse transcriptase inhibitors (NRTI), non-nucleoside reverse transcriptase inhibitors (NNRTI) and protease inhibitors (PI)], has lead to a dramatically improved outcome from this chronic infection [^1-6].

While the benefits of highly active antiretroviral therapy (HAART) have revolutionized the care of HIV-infected patients, frequent and sometimes severe treatment-associated metabolic side effects have been observed [⁷]. Several well known important risk factors for cardiovascular disease (CVD) can be induced and/or enhanced by PI-containing HAART [⁸]. These include increases in serum total cholesterol (particularly an increase in the atherogenic non-high density lipoprotein (non-HDL) cholesterol [⁹]) and triglycerides, as well as impaired glucose tolerance/overt diabetes mellitus associated with increased insulin resistance [¹⁰], and possibly arterial hypertension [^11,12]. However, whether and how soon these antiretroviral therapy-induced abnormalities may result in a clinically detectable increased risk of CVD remains unknown, as does the impact of the underlying HIV infection per se. The available data are largely limited to case reports of myocardial infarctions in young PI-treated HIV- infected patients [^13-17]. A meta-analysis of the immediate risk of myocardial infarction in randomized trials comparing PI and non-PI containing regimens, demonstrated no significant differences between the regimens [¹⁸], and presented retrospective studies have provided conflicting evidence [^19-24].

To gain further insight into the risk of treatment-associated CVD, a multinational, tri-continental collaboration between ongoing HIV cohort studies was initiated in December 1999 (the DAD study, Data collection on Adverse events of anti-HIV Drugs [²⁵]) with the objectives of detecting the incidence of myocardial infarction and stroke, and of identifying whether exposure time to the agents contained in antiretroviral drug regimens is independently associated with the risk of developing these cardiovascular events. The working hypothesis of the study is that anti-HIV drugs may accelerate the atherosclerotic process and, by doing so, increase the risk of CVD including myocardial infarction. The study is powered to detect a twofold increased risk of myocardial infarction, and will follow a cohort of more than 20 000 HIV-infected patients at various stages of infection and therapy prospectively for a minimum of 2 years.

The objectives of the present analyses were to determine the proportion of patients with an elevated risk profile for CVD at the time of inclusion into the DAD study, and to identify factors associated with these increased risk profiles, particularly with regards to the type and duration of antiretroviral therapy.

Methods

Design

The DAD study is an observational study formed by the collaboration of previously established HIV cohorts. Eleven cohorts [^26-36] participate and contribute data on more than 20 000 HIV infected patients followed at 188 clinics in 20 countries situated in Europe, USA and Australia (Table 1 and Appendix).

Table 1

Data collection

Patients are followed prospectively during visits to out-patient clinics scheduled as part of regular medical care. Eligible patients are all under active follow-up at the time of initiation of the DAD protocol, irrespective of antiretroviral treatment status. Patients were enrolled into DAD consecutively as they were seen in the clinic from the time the DAD study was implemented in each of the participating cohorts. The first cohorts started to include patients in December 1999, and all patients were included prior to 1 April 2001.

At enrolment and at least every 8 months thereafter standardized data collection forms are completed at the sites providing information from physical examination, patient interview and patient case notes, concerning family history of coronary heart disease, patients' prior history of CVD and diabetes, cigarette smoking, blood pressure, therapy for diabetes mellitus, lipid-lowering and anti-hypertensive therapy, the presence of clinical signs of lipodystrophy and serum lipid levels (including total- and HDL-cholesterol, triglycerides, and information on fasting conditions). Further, all cumulative data characterizing the patient's underlying HIV infection since inclusion in any of the individual cohorts are collected, including information on demography, antiretroviral therapy, CD4 cell counts and HIV viral loads. Dates of diagnosis of all AIDS-defining diseases are recorded, using the 1993 clinical definition of AIDS from the Centers for Disease Control and Prevention [³⁷]. All collected information is transformed into a standardized format and merged into a central data-set.

Variables

HIV laboratory parameters

In various analyses, CD4 cell count was stratified in strata of 100 × 10⁶ (cells/l) or assessed as a continuous variable (log₂ transformed). Similarly, HIV RNA was stratified in strata of: ≤ 500, 501-10 000, 10 001-100 000, and > 100 000 copies/ml, and also assessed as a continuous variable (log₁₀ transformed).

Antiretroviral therapy

Six categories were predefined based on current use of antiretroviral therapy regimen at the time of enrolment into the DAD study. These are: (i) naive; (ii) treatment-experienced, but not currently receiving antiretroviral therapy; (iii) currently receiving only NRTI; (iv) currently receiving NNRTI and NRTI but not PI; (v) currently receiving PI and NRTI but not NNRTI; or (vi) currently receiving PI, NNRTI and NRTI.

Previous antiretroviral therapy exposure was modelled as cumulative time spent using each of the three drug classes.

CVD risk factors

The grouping of the risk factors assessed was defined prior to the initiation of the analysis. CVD risk factors were assessed as dichotomous categorical variables, where the cut-off levels chosen were conservative estimates of 'high risk' based on levels used for risk scoring in the background population [^38-41]. The specification of risk factors is as follows. (i) Dyslipidaemia: defined as elevated total cholesterol ≥ 6.2 mmol/l (240 mg/dl), and/or decreased HDL-cholesterol ≤ 0.9 mmol/l (35 mg/dl), and/or elevated triglycerides ≥ 2.3 mmol/l (200 mg/dl). [The cut-offs applied are based on cut-offs for high risk for CVD in the NCEP guidelines.] (ii) Older age: age ≥ 45 years for men and ≥ 55 for women. (iii) Family history of coronary heart disease: first-degree relative with myocardial infarction before age 50. (iv) Previous CVD: patients' own previous experience of myocardial infarction and/or stroke. (v) Hypertension: elevated systolic blood pressure ≥ 150 mmHg and/or elevated diastolic blood pressure ≥ 100 mmHg, or usage of anti-hypertensive drugs. (vi) Diabetes: history of diabetes or usage of anti-diabetic therapy. (vii) Body mass index (BMI) was stratified in four categories: underweight (BMI, < 18 kg/m²), normal weight (BMI, 18-26 kg/m²), overweight (BMI, 26.1-30 kg/m²) and obesity (BMI, > 30 kg/m²). Obesity was considered a cardiovascular risk factor. (viii) Smoking: current cigarette smoking at inclusion in the DAD study. (ix) Presence of clinical lipodystrophy was defined as either characteristic fat loss (from the face and/or extremities), central fat gain (abdominal and/or cervicodorsal) or mixed (at least one sign each of fat loss and central fat gain), as judged by the treating physician.

Statistical analyses

Prevalence of single risk factors was calculated for the groups of patients for which data were available. Based on the observed prevalence, dyslipidaemia, diabetes, hypertension and lipodystrophy were further assessed as outcome variables in logistic models.

Univariable chi-squared and Kruskal-Wallis tests were used to compare categorical and continuous baseline demographic, clinical and laboratory characteristics between antiretroviral regimen categories. Association of CVD risk factors with antiretroviral therapy, demographic, clinical and laboratory parameters were tested in univariable logistic regression models. Multivariable logistic regression was then performed to identify parameters independently associated with the presence of CVD risk factors. The multivariable model included all parameters significantly associated with the risk factor assessed, at a level of P < 0.05 in the univariable model. For the main associations, the outcomes of the logistic models were tested in linear models, where the outcome variables were modelled as continuous variables.

From the literature it is known that total cholesterol levels are not significantly influenced by fasting status and HDL-cholesterol is influenced only slightly [^42-44]. Although triglycerides are influenced by fasting, their daily fluctuation does not have a simple relationship with intake of meals [⁴⁵]. In order to consider the impact of fasting status on the results, sensitivity analyses were repeated separately for fasting and non-fasting triglyceride measurements. Similarly, and for all lipid measurements, sensitivity analyses were performed to assess associations among the cohorts with less missing data. The associations of the primary analyses were reproduced by the sensitivity analyses; these results have generally not been included in this report.

All analyses were performed using Statistical Analysis System (SAS) version 6.12 (SAS Institute Inc, Cary, North Carolina, USA).

Results

Demographics

By April 2001, the central database contained information on 17 852 patients enrolled in DAD from nine of 11 participating cohorts. The patient characteristics are shown in Tables 1, 2 and 3. Seventy-six percent were male, the median age was 39 years [inter quartile range (IQR), 34-45], 25% previously had AIDS. Mode of HIV acquisition was homosexual contacts in 43%, heterosexual contacts in 28% and injecting drug use in 23%. The median CD4 cell count was 430 × 10⁶ cells/l (IQR, 270-621 × 10⁶ cells/l) and median plasma HIV RNA was below 500 copies/ml (IQR, < 500-4800 copies/ml). These variables varied by cohort (Table 1).

Table 2

Table 3

Two cohorts (CPCRA and the Brussels St. Pierre cohort) were included later in the DAD study and their patient characteristics at baseline not analysed in this manuscript.

Antiretroviral therapy

At enrolment, 13% of the study population were antiretroviral therapy naive, 6% were previously exposed, but not currently receiving any antiretroviral therapy, 11% were receiving a regimen containing NRTI only, 20% were receiving NNRTI-based therapy, 43% PI-based therapy and 7% were on a regimen containing all three drug classes. Overall, 72% of the study population had at any one time been exposed to PI with a median exposure time of 2.5 years (IQR, 1.5-3.2 years), 36% had ever been exposed to NNRTI with a median exposure time of 0.9 years (IQR, 0.5-1.5 years) and 87% had ever been exposed to NRTI with a median exposure of 3.2 years (IQR, 2.0-4.7 years) (Table 2).

CVD risk factors and association with antiretroviral drugs and duration of therapy

CVD risk factors were prevalent in the study population (Table 3). Almost 25% of the study population was in an age group constituting a CVD risk factor by our definition, with the highest prevalence among patients receiving PI, NNRTI or both of these drug classes. 11.4 % had a family history of coronary heart disease with no significant difference between the antiretroviral therapy groups, and 1.4% had a previous history of CVD, with the highest prevalence in the group of patients receiving a regimen containing both PI and NNRTI. More than half of the study population were current cigarette smokers, with the highest prevalence among the naive patients and patients not currently receiving antiretroviral therapy.

Hypertension

More than 8% of the study population had hypertension. In a univariable logistic model, using the antiretroviral therapy-naive group as reference, regimens containing NNRTI, PI or both drug classes were associated with an increased risk of being hypertensive (Table 4). But after adjustment for other factors which univariably were associated with the presence of hypertension, the associations with antiretroviral therapy disappeared or were reversed (Table 4). This was explained mainly by a strong correlation of hypertension with other factors (age, sex and BMI).

Table 4

Diabetes

The overall prevalence of diabetes was 2.5%. In a univariable model, all regimens were associated with an increased risk of diabetes when compared with naive patients (Table 4). After adjustment for other factors, current therapy with a regimen containing NNRTI or NNRTI + PI remained marginally independently associated with the presence of diabetes.

Serum total cholesterol

The association of antiretroviral therapy with lipid levels is shown in Table 3. Assessed from median cholesterol levels (Table 3) and in univariable models (Table 4), patients currently using regimens containing NNRTI + NRTI, PI + NRTI or all three drug classes combined were at increased risk of having a high total cholesterol when compared with naive patients, with the highest risk among patients receiving a regimen containing all three drug classes. This pattern remained unchanged after controlling for other risk factors (Table 4). Subjects receiving NRTI only as well as subjects who had discontinued antiretroviral therapy have similar total cholesterol levels to naive subjects (regardless of duration of previous exposure to any of the drug classes), the latter suggestive of a reversible drug effect on total cholesterol level.

We further examined the effect of duration of exposure to the drug classes. As current and previous antiretroviral therapy exposures were highly correlated, these parameters were fitted in separate models. In a univariable logistic model for cumulative antiretroviral therapy exposure time, the OR for elevated total cholesterol was 1.00 (IQR, 0.98-1.02; P = 0.81), 1.39 (IQR, 1.31-1.47; P < 10^-4) and 1.42 (IQR, 1.38-1.47; P < 10^-4) per year of exposure to NRTI, NNRTI and PI, respectively. After controlling for other risk factors for dyslipidaemia, these associations remained essentially unchanged (data not shown; the model included sex, age, smoking, family history of coronary heart disease, previous cardiovascular disease, BMI, HIV transmission category, CD4 cell count, HIV RNA and duration of NRTI, NNRTI and PI therapy).

Among patients who currently or previously were exposed to antiretroviral therapy, level of immunodeficiency and plasma HIV RNA were independently associated with elevated total cholesterol after adjustment for other factors, including duration of antiretroviral therapy. The association was present within each antiretroviral therapy regimen group (Fig. 1). Overall, the adjusted risk of having elevated total cholesterol increased by 24% per twofold increase in CD4 cell count [OR, 1.24 (IQR, 1.18-1.30) per log₂CD4, P < 10^-4], thus the highest risk of elevated cholesterol is among patients with preserved or regained immunity (Fig. 1a). Of note, there was no association of CD4 cell count with total cholesterol in treatment-naive patients. In all antiretroviral therapy groups, and also in the group of antiretroviral therapy-naive patients, higher HIV viral load was associated with a decreased risk of elevated total cholesterol (Fig. 1b), overall the adjusted OR was 0.70 (IQR, 0.65-0.75), P < 10^-4, per 1 log₁₀ increase in HIV RNA.

Fig. 1

Serum triglycerides

The prevalence of elevated triglycerides was 28.4% among patients with fasting values and 35.4% for the non-fasting measurements (36% of measurements were fasting values, 24% non-fasting and the remaining lacked information regarding fasting status). The associations of antiretroviral treatment with elevated triglycerides resembled the associations seen with total cholesterol (Table 4), and were also similar within each group, when fasting and non-fasting measurements were tested separately (data not shown).

In a univariable logistic model for cumulative antiretroviral drug exposure time, the OR for elevated triglycerides was 1.05 (IQR, 1.03-1.07), 1.28 (IQR, 1.21-1.35) and 1.38 (IQR, 1.34-1.42) per year of exposure to NRTI, NNRTI and PI respectively (all, P < 10^-4). In the multivariable model these associations remained essentially unchanged (data not shown).

The association with CD4 cell count and HIV viral load differed between regimens. Among patients who were antiretroviral therapy-naive, previously exposed, but not currently receiving any antiretroviral therapy, or currently receiving a regimen containing NRTI only, the adjusted risk of elevated triglycerides increased with increasing HIV RNA [OR, 1.18 (IQR, 1.07-1.31) per 1 log₁₀ increase; P = 0.001], whereas there was no significant association with CD4 cell count [OR, 1.06 (IQR, 0.99-1.13) per twofold increase; P = 0.12].

Among patients receiving NNRTI, PI or a regimen containing both drug classes, the risk of elevated triglycerides increased with increasing HIV viral load [adjusted OR, 1.13 (IQR, 1.06-1.21) per 1 log₁₀ increase; P < 10^-4] and also increased with increasing CD4 cell count [OR, 1.20 (IQR, 1.15-1.26) per twofold increase in CD4 cell count; P < 10^-3].

Serum high density lipoprotein (HDL)-cholesterol

All regimens were associated with an increased risk of low HDL-cholesterol except regimens containing NNRTI, when compared to naive subjects (Table 4). In a univariable logistic model for cumulative antiretroviral drug exposure time, the OR for decreased HDL-cholesterol per year of exposure to NRTI, NNRTI and PI respectively, was 1.08 (IQR, 1.05-1.11; P < 10^-4), 0.87 (IQR, 0.80-0.95; P < 0.002) and 1.01 (IQR, 0.97-1.06; P = 0.53). The multivariable model showed similar associations.

The associations of CD4 cell count and HIV viral load were similar for the absolute value of HDL-cholesterol as for total cholesterol, i.e., the risk of having decreased HDL-cholesterol is highest among patients with low CD4 cell count and high HIV viral load.

Body composition

In all regimen groups there were few obese patients (Table 3), with a slightly higher prevalence among treatment-naive (4.8%) patients than in other groups. As would be expected, antiretroviral therapy was highly associated with the presence of clinical lipodystrophy, with the highest risk among patients receiving a regimen containing all three drug classes (Tables 3 and 4). Patients exposed for a longer time to the antiretroviral drug classes tended to have a higher prevalence of lipodystrophy at baseline (data not shown). Using the composite definition of lipodystrophy, there was no association between BMI and lipodystrophy (data not shown).

When assessed as an explanatory variable, lipodystrophy was associated with the presence of several of the CVD risk factors discussed above. In a multivariable model including the total study population, and adjusting for co-variables as listed in Table 4, the adjusted OR for the association of lipodystrophy with elevated total cholesterol was 1.56 (IQR, 1.41-1.72; P < 10^-4), elevated triglycerides 2.16 (IQR, 1.98-2.37; P < 10^-4) and decreased HDL 1.53 (IQR, 1.35-1.73; P < 10^-4). The presence of lipodystrophy was associated with an increased risk of hypertension and diabetes [OR, 1.34 (IQR, 1.17-1.54) and 2.05 (IQR, 1.63-2.58), respectively; both P < 10^-4].

Discussion

In the DAD population we have observed a high prevalence of multiple risk factors for CVD, particularly among patients currently receiving an antiretroviral therapy regimen containing all three drug classes. DAD has the strength of having included more than 20 000 patients with details concerning CVD risk factors and thereby is by far the largest study conducted to date which addresses CVD risk factors in HIV infection. A novel finding was that regimens containing drugs from both the PI and NNRTI classes were associated with the highest prevalence of dyslipidaemia, suggestive of a possibly additive effect of combinations of drugs from these drug classes. Furthermore, we observed that hypercholesterolaemia was associated with a higher CD4 cell count (in antiretroviral-treated people), a lower HIV plasma viral load, the presence of clinical signs of lipodystrophy and older age.

Dyslipidaemia was most strongly correlated with antiretroviral regimens currently being used, and less with a history of previous exposure to the different drug classes. This finding corresponds with previous reports, in which the dyslipidaemia associated with PI occurred shortly after beginning therapy (in the fraction of patients prone to develop this adverse effect) [^8,46] and the rate of increase in lipid levels abated within months of initiation of the drugs. It is also consistent with studies that have shown that switch from PI to NNRTI-based or NRTI-only regimens is associated with attenuation or resolution of dyslipidaemia [⁴⁷] within a short period of time (i.e., a few months). However, the cumulative time of exposure to various antiretroviral drug classes (with the concomitant risk of raised lipid levels) is likely to be relevant when predicting the risk of future CVD.

The association between PI therapy and elevated levels of total cholesterol and serum triglycerides has been noted previously in smaller cohort studies. The average increases in lipid levels in the largest series published to date [^9,10,48,49], comparing levels during PI therapy with either pre-therapy levels or levels in PI-naive HIV-infected patients, were 28% for total cholesterol and 96% for triglycerides. We observed no difference in risk of low HDL-cholesterol among patients treated with PI, NRTI or not currently receiving antiretroviral therapy, and patients in these groups all had lower HDL-cholesterol levels than treatment-naive individuals. Consistent with most other studies, duration of PI therapy did not influence the level of HDL-cholesterol [⁴⁹], whereas duration of NRTI was associated with a higher risk of low HDL-cholesterol.

An increase in total cholesterol with no increase in the HDL fraction is of particular concern, because it implies an elevation of the atherogenic non-HDL-cholesterol. The risk of elevated total cholesterol was increased per additional drug included in the regimen, and for longer exposure time to PI. Additional analysis to assess possible differences between individual PI is underway to examine these associations in more detail [⁵⁰].

In contrast with previous observations of an association of PI use and dyslipidaemia, our finding of an association between NNRTI-containing regimens and dyslipidaemia has not been widely investigated. Phase I studies of efavirenz in HIV-uninfected subjects revealed increases in total cholesterol levels of 10-20% in some subjects [⁵¹], and no differences between different NNRTI were reported in HIV-infected individuals [⁵²]. In concordance with our results, an increase in (protective) HDL-cholesterol with NNRTI has recently been reported [^53,54]. More detailed analyses to assess possible differences between individual NNRTI are ongoing [⁵⁵].

Consistent with previous reports, NRTI-only therapy was not associated with elevated cholesterol [^52,56]. With regards to triglycerides, recent studies have indicated differences among drugs in the NRTI drug class, with a higher propensity for high triglyceride levels associated with stavudine use [^56,57]. Future analyses from the DAD study will further assess differences between individual drugs within the NRTI drug class.

So far, few studies have examined factors which predispose HIV-infected patients to treatment-associated lipid abnormalities [^58-60]. We have identified several factors that are significantly associated with the presence of dyslipidaemia in HIV infected subjects receiving antiretroviral therapy. We found a strong association between elevated total cholesterol level and higher CD4 cell counts, which was present within each treatment category but not in the antiretroviral therapy-naive group. Nevertheless, within each CD4 cell count stratum, the effect of antiretroviral therapy was clearly observed, which indicates that the effect of antiretroviral substances certainly cannot solely be explained by a reversal to 'normal' pre-disease cholesterol levels as a result of improved cellular immunity. The CD4 cell count level remained independently associated with elevated total cholesterol also after adjustment for duration of treatment. This does not rule out, however, that the observed association may still - at least in part - be due to residual confounding of the effect of antiretroviral therapy, either directly via a dose-response effect (i.e., higher CD4 cell count and lower plasma viral load are surrogates of better adherence and hence higher exposure to causative drugs) or indirectly via lowering HIV-RNA levels (see below).

The level of HDL-cholesterol, although to a lesser extent, likewise increased with more conserved cellular immunity, consistent with observations in the pre-HAART era [⁶¹], while no clear association was observed for levels of triglycerides after adjusting for therapy.

For total cholesterol, the association with HIV viral load was the inverse of the association with CD4 cell count. Thus we found increasing levels of total cholesterol with lower HIV RNA, and similarly for HDL-cholesterol. The latter has also been reported from other studies [⁶²]. Conversely, overall and after adjustment for other factors, levels of triglycerides increased with increasing HIV viral load, consistent with the findings in the pre-PI era of elevated triglycerides linked to HIV disease progression [⁶³].

Subjects already exposed to other risk factors for CVD are likely to have an accelerated course of atherosclerosis and the clinical complications hereof, given the known synergistic effects of different CVD risk factors [³⁸]. In the DAD study, we have observed a high prevalence of other known and potential CVD risk factors among patients receiving either PI or NNRTI, including cigarette smoking, diabetes, hypertension and altered body composition.

The overall prevalence of diabetes mellitus in the DAD study was 2.5% and varied between regimens, from 1.1% in patients not currently receiving antiretroviral therapy to 4.3% in patients receiving PI and NNRTI. This is consistent with other studies which have shown an association between impaired glucose tolerance, diabetes mellitus and use of PI [^8,9,48]. The prevalence of diabetes in PI-treated HIV patients has been reported to be in the range of 2-8%, with the highest detection rate in studies based on performance of oral glucose tolerance testing. In the setting of our observational study, in which oral glucose tolerance testing is not mandated and the diagnosis will mainly rely on measuring repeated elevated fasting blood sugar, the prevalence of diabetes is likely to be underestimated.

Data on the prevalence of hypertension in HIV patients are limited. A few studies have reported an increased prevalence of hypertension in PI treated patients [^11,12] or in conjunction with lipodystrophy [⁶⁴]. In our study, the associations between antiretroviral drug regimens and hypertension in univariable logistic models were no longer present after adjustment for other factors associated with hypertension. Thus, our data do not support a concern that HIV treatment per se is likely to induce hypertension.

Consistent with current knowledge [¹⁰], the use of antiretroviral therapy was strongly associated with the presence of lipodystrophy. Furthermore, there was a marked association between dyslipidaemia and several of the other CVD risk factors on the one hand and clinical lipodystrophy on the other. Such associations do not necessarily suggest a particular aetiology of lipodystrophy, but rather describe a clinical phenomenon, as the definition of lipodystrophy in the present analyses includes all clinical presentation of the syndrome, when in fact the various fat re-distribution patterns may represent separate entities with different aetiology [⁶⁵]. Clinical lipodystrophy - i.e., fat redistribution - is associated with presence of several known metabolic risk factors for CVD, and therefore the lipodystrophy syndrome would also be expected to be associated with an increased risk of CVD.

Whether the presence of fat redistribution in itself - by way of abnormal fat loss and/or gain - represents an independent risk for CVD remains unresolved. A large collaboration is ongoing with the purpose of establishing a case definition for the lipodystrophy syndrome [⁶⁶], which will facilitate the evaluation of the different clinical patterns and their possible influence on risk of CVD.

Compared with antiretroviral therapy-naive subjects, those treated with antiretroviral drugs tended to be less obese. Whether this observation is causally related to adverse events caused by the antiretroviral drugs should be investigated further. Obesity is an independent risk factor for CVD in the background population [^40,67].

Strengths and limitations

The strength of our results is primarily related to the substantial size of the study population. The diversity of the study population, including patients from a variety of geographical areas and a substantial number of women and minorities, ensures that the study is representative of the HIV infected population in industrialized countries.

The limitations are mainly related to the observational design of the study and the cross-sectional nature of the current analyses. Firstly, the results presented are only associations from which no conclusions regarding causality can or should be drawn. Secondly, due to the observational design of the study, many measurements are expected not to be always conducted in a uniform manner. This includes measurement of blood pressure and laboratory analyses of lipid levels. However, even in the absence of uniform standards for this study, national and international standardization of serum lipid measurements have been accomplished through collaboration of the Centers for Disease Control and Prevention and the World Health Organization, and comparable results can be obtained globally because of these standardization efforts [⁶⁸]. Thirdly, the relatively high proportion of missing data should be noted (Table 3), which amongst other things implies that the prevalence of the individual risk factors is imprecise. Measures have been taken to complete the collection of pending baseline data during follow-up. Finally, information concerning certain other potential risk factors for CVD was not collected in our study, including genetic factors, physical activity, diet and alcohol consumption.

Perspectives

The present analysis shows that, especially among older patients, the use of potent antiretroviral therapy resulting in more profound virus suppression and more preserved immunity, was associated with a high both relative and absolute risk of exhibiting risk factors for coronary heart disease.

Using these results, work is in progress to model the estimated risk of CVD based on validated algorithms [⁶⁹]. Such projections assume that the induced risk factors can be directly transposed, which is likely to be a simplification as there presumably will be a time lag from when factors known to accelerate the atherosclerotic process are induced (i.e., when PI and/or NNRTI are started) and until clinical manifestations of atherosclerotic vascular disease will occur. As many of these factors are likely to act synergistically, and as the underlying HIV infection itself and its various manifestations may also contribute, current knowledge does not permit reliable assessment of the duration of the above mentioned time-lag. However, comparison of the estimated expected with the observed CVD event rate in the DAD study may provide some understanding.

The question as to whether antiretroviral therapy-associated metabolic disorders contribute to premature cardiovascular disease is of major importance for the way HIV infection is clinically managed. If current treatment of HIV infection would indeed be shown to be associated with an increased risk of CVD, such risk of course would need to be balanced against the proven major benefits of therapy. It would likely have implications for considerations concerning the composition of regimens, the timing of initiation of therapy, as well as for the evaluation and use of various pharmaceutical and non-pharmaceutical measures directed at reducing CVD risk. Last but not least it stresses the continued need for developing less toxic and better tolerated effective treatments for HIV infection.

Acknowledgements

Sponsorship: The ATHENA study was supported by a grant (CURE/97-46486) from the Health Insurance Fund Council, Amstelveen, the Netherlands. The Aquitaine Cohort was supported by a grant from the 'Agence Nationale de Recherches sur le SIDA' (ANRS, Action Coordonnée no.7, Cohortes). The BASS study was supported by grants from the 'Fondo de Investigación Sanitaria' (FIS 99/0887) and 'Fundación para la Investigación y la Prevención del SIDA en Espanã' (FIPSE 3171/00). The EuroSIDA study was supported by grants from the European Commission BIOMED 1 (CT94-1637) and BIOMED 2 (CT97-2713) programs, from Pharmacia & Upjohn, GlaxoSmithKline, Roche and Merck. The ICONA network was supported by an unrestricted educational grant from Glaxo Wellcome, Italy. The Swiss HIV Cohort Study was supported by a grant (3345-062041) from the Swiss National Science Foundation. Support for the DAD study was provided by the 'Oversight Committee for The Evaluation of Metabolic Complications of HAART', a collaborative committee with representation from academic institutions, the EMEA, the FDA and all pharmaceutical companies with licensed anti-HIV drugs in the US marked, i.e., Abbott, Agouron, Boehringer Ingelheim, Bristol-Myers Squibb, GlaxoSmithKline, Merck, Pfizer, Pharmacia & Upjohn, Hoffman-La Roche.