Context: Patients with HIV may have increased risk of atherosclerotic cardiovascular disease owing to multiple biological mechanisms.
Objective: To evaluate the evidence for subclinical atherosclerosis among patients with HIV.
Design: Systematic review of observational studies.
Data sources: We searched Medline, Cochrane DSR, ACP Journal Club, DARE, CMR, HTA, NHSEED, Embase and the Cochrane Controlled Trials Register for studies published before November 2008.
Study selection: Eligible studies were cross-sectional, cohort, or case–control studies reporting carotid ultrasound intima-media thickness (CIMT), focal plaque incidence, or coronary artery calcium (CAC), as determined by HIV positivity or protease inhibitor (PI) exposure.
Data extraction: Two independent reviewers abstracted data using a standardised form. The primary outcome was weighted mean difference (WMD) for CIMT comparing HIV positive versus negative patients. Other outcomes included WMD by PI exposure and the odds ratio (OR) for a focal carotid plaque or CAC. Data from six cross-sectional, seven case–control and 13 cohort studies were included, involving 5456 HIV positive and 3600 HIV negative patients.
Results: The weighted mean CIMT was 0.04 mm thicker among patients with HIV than among non-HIV patients (95% CI 0.02 to 0.06; p<0.001). HIV positivity was not associated with carotid plaque or CAC. PI exposure did not significantly affect CIMT, carotid plaque, or CAC. There was evidence of publication bias and stratified analysis and meta-regression showed outcomes were influenced by study design, age, gender and smoking. However, HIV positivity slightly increased CIMT even after sensitivity analyses.
Conclusions: HIV infection and PI exposure are not strong independent risk factors for subclinical atherosclerosis. Confounding may contribute to overestimation of the risk associated with HIV and PI exposure.
Statistics from Altmetric.com
Patients with HIV may have increased risk of atherosclerotic cardiovascular disease owing to multiple biological mechanisms. First, infection with HIV provokes a chronic, systemic inflammation. Inflammation is increasingly accepted as having an important role in the pathogenesis of atherosclerosis and acute cardiovascular events.1 Second, metabolic derangements resulting from HIV may result in dyslipidaemia (especially low high-density lipoprotein (HDL)), endothelial dysfunction, hypercoagulability, platelet activation and impaired fibrinolysis.2 3 Patients with HIV have been described as having histologically unique features of coronary artery disease, including a rapid progression of diffuse, circumferential arterial lesions with proliferation of smooth muscle cells, elastic fibres and endoluminal protrusions.4 Similarly, patients with HIV differ epidemiologically from non-HIV patients with acute coronary syndrome (younger, lower HDL, more smoking, less angiographically apparent coronary artery disease and more post-percutaneous intervention stent restenosis).5 Third, treatments of HIV, especially protease inhibitors (PIs), while clearly associated with important mortality reductions from HIV, may accelerate atherosclerosis especially because of metabolic side effects including dyslipidaemia and insulin resistance.6 7 Finally, patients with HIV may have a greater burden of atherosclerotic disease because of higher rates of established traditional risk factors independent from the disease process of HIV itself, including male gender, smoking, hyperlipidaemia and glucose intolerance.
Numerous observational studies have investigated the potential association of HIV positivity, PI exposure and atherosclerosis.8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 However, any association of HIV with atherosclerosis remains vague owing to discrepant findings among these studies and the likelihood of bias affecting some study results. Investigators have sought to clarify a potential relationship of HIV with atherosclerosis burden through cross-sectional studies using high-frequency carotid ultrasound to measure carotid intima-media thickness (CIMT) or carotid plaque and cardiac CT (electron beam or multidetector) to quantify coronary artery calcium (CAC). CIMT, carotid plaque and CAC are well-established direct measures of subclinical atherosclerosis burden and predict the risk of future cardiovascular events.42 43 However, studies of subclinical atherosclerosis among patients with HIV have had disparate results with odds ratios ranging from protective to harmful.8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 Cohort studies that have quantified the risk of coronary events among patients with HIV and those exposed to PIs have similarly yielded inconsistencies.34 35 36 37 38 39 40 41
Previous reviews have established that patients with HIV have increased cardiovascular risk factors.44 The association of HIV and PIs with atherosclerosis was previously reviewed but an update including more recent evidence is warranted.45 46 We conducted a systematic review of published literature and carried out a meta-analysis to further investigate the potential association of HIV positivity, PI exposure and atherosclerosis.
We searched Medline, Cochrane DSR, ACP Journal Club, DARE, CMR, HTA, NHSEED, Embase and the Cochrane Controlled Trials Register published for relevant articles published in any language from the beginning of electronic indexing through 31 October 2008. We used the text words and related Medical Subject Headings terms: HIV, AIDS, acquired immunodeficiency syndrome, HIV seropositivity, carotid, tunica intima, IMT, plaque, coronary calcium, EBCT, electron beam computed tomography, atherosclerosis. Search results were limited to observational studies (cross-sectional, case–control, or cohort) in adults with HIV infection (age >18 years). References of reviewed articles were also searched for relevant titles. We executed and reported our findings according to the guidelines of the Meta-analysis Of Observational Studies in Epidemiology group.47
Two reviewers (one doctor and one doctor/epidemiologist) independently conducted the literature search and extraction of relevant articles. The title and abstract of potentially relevant studies and review articles were screened for appropriateness before retrieval of the full articles. We included observational studies in adults comparing the presence and degree of subclinical atherosclerotic lesions by carotid ultrasound or CAC among HIV positive patients with HIV negative patients, or patients with HIV exposed to PI drugs with patients with HIV not exposed to PIs. We defined subclinical atherosclerosis as evidence of atherosclerotic cardiovascular disease as detected by screening in asymptomatic patients using: CAC >0 or carotid ultrasound with a focal structure that encroaches into the arterial lumen of at least 0.5 mm or 50% of the surrounding CIMT value.42 43 48 We defined HIV positivity in accordance with current the United States Centers for Disease Control definition: positive HIV antibody assay, positive HIV nucleic acid testing (eg, RNA PCR), p24 antigen test, viral isolation, or diagnosis by a doctor or qualified medical provider based upon laboratory criteria and documented in a medical record.49
Using a standardised abstraction sheet, we recorded the following characteristics of the study: author, year, country, design, duration, sample size, method of measuring CIMT or carotid plaque or CAC and prevalence of important confounders, including age, gender, smoking, hypertension, diabetes mellitus, mean low-density lipoprotein (LDL), mean HDL, body mass index (BMI), mean CD4 count, duration of HIV, duration of PI exposure and mean systolic blood pressure. Outcomes abstracted included the carotid intima-media thickness in mm, incidence of carotid plaque and incidence of CAC. Missing data were imputed by best subset regression using Stata (version 8.2, College Station, Texas, USA). Study authors were contacted for additional data when relevant. Two reviewers (EH and JM) independently abstracted data and disagreements were resolved by consensus.
Two reviewers independently rated study quality using the Newcastle–Ottawa scale for the assessment of the quality of observational studies.50 This instrument is recommended for use by the Cochrane Collaborative Review Group on HIV Infection and AIDS.51 Assessment of quality is graded by the description of patient selection (four criteria), study–control group comparability (one criterion) and outcome assessment (three criteria). Based upon previous recommendations, studies meeting five or more criteria were considered to be high quality.
Our principal abstracted measures of effect were the CIMT in mm and odds ratios of carotid plaque or CAC. Univariate unadjusted outcomes were used; when authors reported multivariate adjusted outcomes, these were included in the sensitivity analysis but not the primary meta-analysis. CIMT were combined with the weighted mean difference (WMD). Using the raw data extracted for incidence of carotid plaque and CAC, 2×2 tables were constructed for calculation of unadjusted odds ratios, which were then pooled to compare the incidence of carotid plaque and CAC. Outcomes were pooled using the Der Simonian and Laird (random effects) model.
Heterogeneity was assessed by the I2 statistic. The I2 statistic provides an estimate of the amount of variance due to heterogeneity rather than chance and is based on the traditional measure of variance, the Cochrane Q statistic. We conducted stratified analysis to assess for potential confounders’ contribution to heterogeneity, including age, gender, smoking, LDL, HDL, hypertension, diabetes mellitus, BMI, mean CD4 count, site of carotid ultrasound measurement, study design and study quality (⩽ or > the median overall Newcastle–Ottawa score as well as individual component analysis). Meta-regression was performed using the method of moments. To exclude the possibility that any one study was exerting excessive influence on the results, we conducted a sensitivity analysis by systematically excluding each study at a time and then rerunning the analysis to assess the change in effect size. Publication bias was assessed using Egger’s method. All analyses were performed with Stata. All p values were two sided with an α of 0.05.
Among 258 studies identified, we excluded 200 that were not relevant to observational study of subclinical atherosclerosis among patients with HIV. Of 58 articles selected for detailed evaluation, 13 were excluded owing to missing data or comparison groups not relevant to our study question, 20 were duplicate data and 25 were included from the initial search.8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 After the addition of a study identified from the references of included articles a total of 26 articles were included.33 Several cohort studies of adverse cardiovascular event rates were identified but not included in this analysis of imaging rates of subclinical atherosclerotic lesions.34 35 36 37 38 39 40 41
Study characteristics and quality
Table 1 shows the study characteristics. Six studies were cross-sectional, seven case–control and thirteen were cross-sectional data derived from prospective cohort studies. There were 9056 patients in 26 studies (5456 HIV positive and 3600 HIV negative). Demographics across all 26 studies were mean (SD) age 45 (6) years, 78 (19)% male, 9.3 (1.9) years’ duration of HIV infection, 50% of all patients with HIV exposed to PIs and 27.8 (8) months’ average duration of PI exposure. The mean CD4 count was 481×106 cells/l. Studies used differing sites and techniques to measure carotid intima-media thickness. Four studies measured unilateral IMT of the common carotid, nine studies used the average of multiple sites in the bilateral common carotid and 13 studies averaged the IMT of multiple sites at the bilateral common carotid in combination with sites from the carotid bifurcation, internal carotid, or external carotid. Definition of focal carotid plaque ranged from 1 mm to 2.0 mm. No study defined plaque based upon the Mannheim criteria of a focal structure that encroaches into the arterial lumen of at least 0.5 mm or 50% of the surrounding CIMT value.43 Positive coronary calcium was defined as CAC >0, although one study19 used CAC >5 and one study >10.18 Tables 2 and 3 describe the rates of covariates among HIV positive versus negative patients and PI exposed versus non-exposed patients. Twenty-one (81%) studies reported data regarding age, 21 (81%) gender and 19 (73%) smoking. Not all studies reported other covariates in the same manner (eg, some report mean LDL, others report incidence of hypercholesterolaemia). Twenty-three (88%) controlled for age and at least one other important confounder; three studies did not control for confounding.19 28 30 Eleven studies (42%) stated there was blinded assessment of the study outcome (ultrasound or CAC result).8 10 13 15 19 20 21 23 25 32 33 The median Newcastle–Ottawa score was 6. One study received < 5 on the Newcastle–Ottawa scale.28
Thirteen studies compared HIV positive with HIV negative patients. All studies showed a trend toward increasing CIMT thickness with HIV, except two.17 27 Standard deviations for mean carotid IMT were imputed for three studies.10 12 16
The WMD for CIMT comparing HIV positive with HIV negative patients was 0.04 mm (95% CI 0.02 to 0.06, p<0.001) (fig 1). There was significant heterogeneity by I2 = 86.5%, p<0.001. Odds ratio for carotid plaque (fig 2) was 1.50 (95% CI 0.95 to 2.39, p = 0.084); I2 = 67%, p = 0.009.
Twelve studies compared CIMT for PI exposed versus non-exposed patients. Most studies showed a trend toward thicker CIMT among PI exposed patients, except four studies.12 21 24 32 The WMD for CIMT (fig 3) was not significantly different by PI exposure, WMD = 0.02 mm (95% CI −0.01 to 0.06, p = 0.190); I2 = 93%, p<0.001). Odds ratio for carotid plaque (fig 4) was not significantly increased by PI exposure, OR = 1.71 (95% CI 0.90 to 3.28, p = 0.103); I2 = 52%, p = 0.059.
Five studies reported incidence of CAC among HIV positive versus HIV negative patients.8 18 20 28 30 The pooled odds ratio for raised CAC by cardiac CT was not significant comparing HIV positive with HIV negative patients: OR = 0.95 (95% CI 0.55 to 1.65, p = 0.851); I2 = 65%, p = 0.024. The odds ratio of CAC for three studies comparing PI exposure was also not significant, OR = 1.38 (95% CI 0.53 to 3.59, p = 0.506); I2 = 65%, p = 0.058.
Sensitivity and subgroup analysis
For studies reporting CIMT by HIV positivity, there was evidence of confounding by male gender using meta-regression. Figure 5 demonstrates increasing differences in CIMT with increasing confounding by gender. Stratified analysis identified three outliers that were excluded from a sensitivity analysis.15 29 32 Patients with HIV in the study by Seminari et al were 21% more likely to be current smokers and 14% higher male gender.29 Patients with HIV in Hsue et al had a 30% increased rate of smoking, 11% higher rate of dyslipidaemia and 12 mg/dl lower HDL.15 Patients with HIV in Yaldizli et al differed by 22% higher smoking, 40% higher male gender, 40% higher dyslipidaemia, 4 years of age, 6% higher diabetes and 6% higher hypertension. Removing these three studies in a sensitivity analysis, the WMD for carotid IMT decreased in magnitude but remained significantly different among HIV versus non-HIV patients (WMD = 0.02, 95% CI 0.006 to 0.038, p = 0.007). When studies with age disparity of greater than 1 year were excluded, there was no significant difference in CIMT by HIV exposure.9 14 20 28 30 31
For studies comparing CIMT by PI exposure, there was no confounding evident by meta-regression. By, stratified analysis, one study29 was an outlier for age (3 years older age among PI exposed patients) and systolic blood pressure (PI exposed: systolic blood pressure 4 mm Hg higher). However, that study did have a reverse bias in smoking (PI: 12% lower rate of smoking). One study32 had significant differences in hypertension (PI: 7% more hypertensive), dyslipidaemia (PI: 7% greater), diabetes (PI: 3% higher), mean HDL (PI: 14 mg/dl lower), although reverse bias from age (PI: 2 years younger) and gender (PI: 4% fewer male subjects). Although the WMD for CIMT by PI exposure was not significant, exclusion of these two biased studies29 33 moved the WMD even closer to the null (WMD = 0.00, 95% CI −0.033 to 0.36, p = 0.942).
Comparing odds ratios of carotid plaque, sensitivity analysis shows that exclusion of the study by Seminari et al would change the odds ratio of carotid plaque by HIV positivity from non-significant to a statistically significant 1.65 (95% CI 1.07 to 2.55). Excluding the Seminari paper from the calculation of odds ratio of carotid plaque by PI exposure, the odds ratio would be significant, 1.93 (95% CI 1.11 to 3.37).
There were insufficient studies for subgroup analysis of CAC data. There was inadequate reporting of statin therapy, c-reactive protein, specific type of PI, or viral load to evaluate for influence upon carotid ultrasound or CAC. Potential covariates that were not significant sources of heterogeneity in this study included diabetes, BMI, CD4, LDL, duration of HIV and duration of PI exposure.
We stratified all outcomes by method of measuring carotid ultrasound (unilateral, bilateral and whether common carotid only versus combined measurements along the common, internal, or external carotid). Method of measuring carotid ultrasound was not a significant source of heterogeneity and did not systematically affect the results of this study.
When stratified by study design (fig 6), data from case–control and nested within prospective cohort studies resulted in lower magnitude effect sizes than cross-sectional studies.
Publication bias assessment
For studies comparing CIMT by HIV positivity, there was evidence of publication bias by funnel plot asymmetry, Egger’s p = 0.001. Smaller studies were more likely to conclude that HIV patients have thicker CIMT. Larger studies were clustered about the null hypothesis. For studies comparing CIMT by PI exposure, there was no significant publication bias, Egger’s p = 0.395.
The preponderance of observational data demonstrates that patients with HIV have increased rates of cardiovascular events. Owing to varying methodologies of controlling confounding factors and inherent limitations of observational studies, the degree of risk that HIV or PI treatment independently might increase a patient’s risk of atherosclerosis is difficult to quantify with the available evidence.
We demonstrate in this analysis that HIV positivity is associated with a small increase in CIMT (0.04 mm), a difference that decreased to 0.02 mm when performing a sensitivity analysis that removes studies comparing patients with significantly different risk factor profiles. Additionally, there was a non-significant trend towards increased prevalence of carotid plaque among HIV positive patients. Exclusion of one statistical outlier (Seminari et al) would make the carotid ultrasound data more strongly significant, with the magnitude of the effect size for the odds ratio 1.65 for HIV and 1.93 for PI exposure. Treatment with PI versus PI sparing highly active antiretroviral therapy (HAART) regimens showed a non-significant trend towards increased CIMT and focal carotid plaque but not CAC. Interestingly, the slight trend towards risk demonstrated by carotid ultrasound (CIMT and focal plaque) was not similarly shown by CAC incidence on cardiac CT. This discrepant result is of importance since CAC has been shown to have a greater predictive value for future coronary heart disease events than carotid ultrasound.42 Alternatively, one must view these results within the age group of the population studied. CIMT may identify generalised atherosclerosis at an earlier stage than calcified atherosclerosis as quantified by cardiac CT. The cohort in our analysis was relatively young (mean (SD) age 45 (6) years) and hence, one would expect the majority of patients in this age group to have low rates of calcified coronary atherosclerosis regardless of risk factor profile. It has been suggested that CAC is a later finding in the atherosclerotic process. However, the clinical effect potentially imparted by a CIMT difference of 0.04 mm is of unclear prognostic significance. For example, a difference of this degree in CIMT would not significantly affect the percentage quartile rank (eg, 25th, 50th, >75th) of most patients as compared with published normal values.53 54
Three prospective cohort studies (table 4) described the rate of CIMT progression among patients with HIV. These studies found differing outcomes. Hsue concluded that HIV positivity increased CIMT progression whereas Currier and Mercie concluded the opposite. Hsue, but not Currier or Mercie, found that PI exposure was associated with CIMT progression. Mercie analysed by HAART exposure and not PI itself. Currier evaluated specifically ritonavir and concluded that this drug specifically was significantly associated with CIMT progression (for p<0.10) with a p value of 0.06. Hsue and Currier concluded that a low baseline CD4 count was associated with faster IMT progression, whereas Mercie concluded the opposite. One study55 was not included because the CIMT was combined with femoral IMT. This study described an accelerated progression of CIMT (0.045 mm/year) that was not associated with PIs but was associated with low CD4 count.
It is important to note that, results from meta-analyses of observational data should be interpreted with caution. Biases inherent to the study design of the original cross-sectional and cohort studies must still be regarded as significant in the final analysis. Our analysis demonstrates that bias may contribute to overestimation of the independent risk of HIV and PI exposure. Although the power to detect publication bias was low, there was evidence of such bias. This reached statistical significance for the carotid IMT data by HIV positivity but not PI exposure.
Our study was limited by significant heterogeneity. Significant heterogeneity is a common limitation of meta-analyses of observational studies and we considered potential sources of heterogeneity during data extraction and used stratified analyses and meta-regression to investigate heterogeneity among the carotid IMT studies. However, meta-regression and stratified analyses were limited by between-study differences in reporting of covariates (eg, some studies report continuous covariates such as mean LDL, others report categorical data such as incidence of hypercholesterolaemia). We also performed sensitivity analysis to identify statistical outliers and whether any one study predominantly influenced the pooled outcomes.
We found evidence that gender was the strongest predictor of heterogeneity, although four of 26 studies15 29 32 33 had significant differences in other important potential confounders that impacted the pooled results after sensitivity analysis. Although most studies individually performed stratified analyses or multivariate analyses to control for confounding, the studies with the strongest effect sizes tended to have the greatest baseline demographic differences among comparison groups, which raise the question of residual confounding.29 32 Excluding such studies, the WMD for carotid IMT remained significant, although the effect size was reduced from 0.04 mm to 0.02 mm. There was insufficient data to perform subgroup analysis by type of PI (some PIs have been reported to cause less frequent metabolic side effects).56 57 Other limitations include the inability to control for selection biases, such as the possibility that patients with cardiovascular risk might have been less likely to be treated with PIs, thereby spuriously increasing risk of non-PI regimens, patients receiving PI might have started treatment with lipid-lowering treatment such as statins which reduce cardiovascular events, and there was infrequent reporting of rates of statin use and use of differing definitions of “carotid plaque” (for this reason, a universal definition such as that defined by the Mannheim consensus is of benefit).43
Despite cohort studies of increased rates of clinically manifest coronary heart disease among patients with HIV, HIV infection and PI exposure do not appear at the population level to be strong independent risk factors for subclinical atherosclerosis by CIMT or CAC. CIMT and CAC are well-established predictors of cardiovascular events in non-HIV patients.42 One possible explanation for the discrepancy among cohort studies of clinical outcomes and imaging studies is that increased myocardial infarction (MI) among patients with HIV might be a spurious outcome caused by difficult to control confounding and challenges to defining exposure to anti-retroviral therapy owing to frequent cross-over of drugs. Alternatively, HIV positive patients might be at increased risk of cardiovascular events through mechanisms not fully explained by differences in CIMT or CAC, such as platelet aggregation, inhibition of fibrinolysis, endothelial dysfunction, or arteritis mediated by direct viral infiltration of the endothelium or co-infection with another virus.58 59 Our result is similar to that of Hsue and colleagues, who reported lower rates of angiographically significant coronary stenosis among patients with HIV with acute coronary syndrome.15 Based upon the available population studies, it appears that increased risk of MI among patients with HIV in cohort studies is not wholly explained through subclinical atherosclerosis.
Any increased risk of atherosclerosis caused by treatment with PIs should be balanced against their proven marked reduction in morbidity and mortality for patients with HIV. As discussed by Stein in response to the publication of the DAD (Data collection on Adverse events of anti-HIV Drugs) cohort results, a relative risk of 1.16 increased MIs is small proportionally and should be weighed against the risk of increasing age (1.39), male gender (1.91), smoking (2.83), or history of cardiovascular disease (4.3).60 These considerations are important owing to the improved outcomes with effective treatment of HIV and many patients continuing to lead healthy lives for decades after diagnosis. The current evidence suggest that clinicians should focus on treatment of traditional cardiovascular risk factors as HIV positivity and PI exposure of themselves are not strong independent risk factors, although they are associated with increased rates of traditional risk factors especially smoking, low HDL and insulin resistance.
Data from randomised controlled trials of HAART drugs may further clarify cardiovascular adverse effects but are difficult to assess because the HIV outcomes trials may not be powered to assess cardiovascular outcomes, owing to the complex combinations of regimens and to the frequent changeover of drug regimens because of viral resistance or adverse effects.
Observational data demonstrate that HIV infection is independently associated with a small increase in subclinical atherosclerosis measured by CIMT but no increase in focal carotid plaque or CAC. Treatment with PI versus PI sparing HAART regimens showed a non-significant trend towards increased CIMT and focal carotid plaque but not CAC. HIV positivity or PI exposure are not strong independent risk factors for subclinical atherosclerosis. The available data probably overestimate risk due to bias. Future studies of patients with HIV receiving PIs and other HIV treatments should continue to track cardiovascular risk factors and outcomes. Future studies of carotid atherosclerosis should conform to uniform guidelines such as the Mannheim criteria43 to allow for generalisability.
We thank Dr Jeffrey Jackson for assistance with preparation of this manuscript.
Competing interests None.
Provenance and peer review Not commissioned; not externally peer reviewed.
The opinions and assertions contained herein are the authors’ alone and do not represent the views of the Walter Reed National Military Medical Center, the US Army, or the Department of Defense.
Data sharing A technical appendix, statistical code and dataset are available from the corresponding author.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.