Use of Human Immunodeficiency Virus-1 Protease Inhibitors Is Associated With Atherogenic Lipoprotein Changes and Endothelial Dysfunction
Background Human immunodeficiency virus protease inhibitors (HIV PIs) are associated with hyperlipidemia, hyperglycemia, and obesity; however, it is not known whether they increase risk of atherosclerotic vascular disease. The purposes of this study were to characterize the lipoprotein abnormalities associated with use of HIV PIs in individuals with HIV infection and to determine the pathophysiological significance of these changes by assessing their effect on endothelial dysfunction.
Methods and Results This was a cross-sectional study of 37 adults with HIV-1 infection who were receiving antiretroviral therapy. Twenty-two were taking HIV PIs (group 1); 15 were not (group 2). Lipids and lipoproteins were measured by enzymatic techniques and nuclear magnetic resonance spectroscopic analysis. Flow-mediated vasodilation (FMD) of the brachial artery was measured by high-resolution ultrasound. Subjects in both groups were similar in regard to age, time since diagnosis of HIV infection, and CD4 cell count. Group 1 subjects had higher total cholesterol (5.68 versus 4.42 mmol/L, P=0.007) and triglyceride (4.43 versus 1.98 mmol/L, P=0.009) levels, characterized by elevated levels of IDL and VLDL. Subjects in group 1 had impaired FMD (2.6±4.6%), indicative of significant endothelial dysfunction. Group 2 subjects had normal FMD (8.1±6.7%, P=0.005). In group 1, chylomicron, VLDL, IDL, and HDL cholesterol levels predicted FMD.
Conclusions Use of HIV PIs is associated with atherogenic lipoprotein changes and endothelial dysfunction. Because these metabolic and vascular changes predispose to atherosclerosis, monitoring and treatment of dyslipidemia in patients taking these medications is warranted.
Received March 27, 2001; revision received May 1, 2001; accepted May 2, 2001.
Human immunodeficiency virus protease inhibitors (HIV PIs) are key components of antiretroviral therapy in patients with HIV infection; however, many patients receiving these medications develop risk factors for coronary artery disease (CAD), such as hyperlipidemia, hyperglycemia, and central obesity.1–3⇓⇓ As survival of individuals with HIV infection increases, atherosclerotic vascular disease could become an important HIV-related complication, especially if therapy for HIV infection and the metabolic derangements associated with such treatment are atherogenic. Indeed, severe premature CAD has been reported in patients receiving these medications.4
Although consistent associations between use of HIV PIs, hypercholesterolemia, and hypertriglyceridemia have been identified, hypertriglyceridemia and atherogenic lipoprotein changes also have been described in patients with HIV infection who were not receiving these medications.1,5–8⇓⇓⇓⇓ The lipoprotein changes associated with use of HIV PIs have not been described in detail, and the effects of these medications on markers of atherosclerosis, such as endothelial function, have not been evaluated. The purposes of this study were to characterize the lipoprotein abnormalities associated with use of HIV PIs in individuals with HIV infection and to determine the pathophysiological significance of these changes by assessing their effect on endothelial dysfunction.
Experimental Protocol and Clinical Parameters
The University of Wisconsin Medical School Institutional Review Board approved this study, and all subjects provided informed consent. Consecutive patients with HIV-1 infection for ≥6 months who were receiving stable doses of antiretroviral medications were screened for participation. Individuals with CAD or recent opportunistic infection or who were taking anabolic steroids, immunomodulators, or lipid-lowering medications were excluded, as were subjects who smoked ≥1 pack of cigarettes daily. Of 76 eligible patients, 6 could not provide informed consent, 31 refused to participate because of time considerations, and 2 did not provide a reason for not enrolling in this study.
Subjects in group 1 (receiving HIV PIs) took stable doses of commercially available HIV PIs for ≥6 months. Subjects in group 2 (not receiving HIV PIs) underwent stable antiretroviral regimens that did not include an HIV PI. Clinical parameters are summarized in Tables 1 and 2⇓.
Fasting serum glucose levels were measured by the hexokinase method. CD4 cell counts were measured by flow cytometry. Plasma HIV RNA titers were measured by a bDNA hybridization assay (Chiron). Resting heart rate (HR) and systolic blood pressure (SBP) were measured by oscillometric sphygmomanometry.
Measurement of Lipid and Lipoprotein Parameters
After a 12-hour fast, 2.0 mL of venous blood was collected into EDTA specimen tubes and centrifuged at 2000 rpm for 15 minutes. A 1.0-mL aliquot was withdrawn, and total serum cholesterol, HDL cholesterol (HDL-C), and triglyceride levels were measured by enzymatic methods on a Hitachi 747 analyzer with standard reagents (Roche Diagnostics). Remaining plasma was frozen at −70°C and shipped within 60 days for lipoprotein analysis with nuclear magnetic resonance spectroscopy (LipoMed).9,10⇓ Sixteen measured lipoprotein subclasses were combined into 10 categories: chylomicrons (>200 nm), large VLDL and chylomicron remnants (60 to 200 nm), intermediate VLDL (35 to 60 nm), small VLDL (27 to 35 nm), IDL (23 to 27 nm), large LDL (21.3 to 23 nm), intermediate LDL (19.8 to 21.2 nm), small LDL (18.8 to 19.7 nm), large HDL (8.2 to 13 nm), and small HDL (7.3 to 8.2 nm). Levels of chylomicrons and VLDL subclasses were expressed in units of triglycerides. LDL and HDL subclass concentrations were expressed in units of cholesterol (mg/dL). Average diameters of VLDL, LDL, and HDL particles were determined by weighting the relative mass percentage of each subclass by their mean diameters.9
Measurement of Brachial Artery Ultrasound Parameters
Subjects who smoked cigarettes were excluded from this part of the study. Brachial artery (BA) reactivity studies were performed in the morning on the day of phlebotomy in the fasting state.11–13⇓⇓ Hyperemia was induced by inflating a tourniquet around the forearm to a suprasystolic pressure for 4.5 minutes. Flow-mediated vasodilation (FMD) of the BA was measured 1 minute after cuff deflation. Nitroglycerin-mediated vasodilation (NTGMD) was measured 3 minutes after administration of sublingual nitroglycerin (400 μg). BA diameters and blood flows were measured with a 7.5-MHz linear-array ultrasound transducer and a Hewlett-Packard 5500 Sonos ultrasound system. BA diameters were measured in triplicate with digital calipers (Secure Archive). Measurements were performed blinded to subject information, including group and study date. In this laboratory, reliability for measurement of the BA diameter is 0.987, reflecting an interclass correlation coefficient across all readings and conditions.11
All parameters were described by means and SDs. Between-group differences were compared by repeated-measures t tests and verified by permutation tests to calculate exact probabilities.14 Differences in categorical variables were evaluated by χ2 tests or Fisher exact tests. Pearson correlation coefficients described linear relationships between parameters. Predictors of FMD were determined by linear and stepwise regression analyses. Several regression equations were evaluated by standard statistical techniques to describe mediational effects of HIV PI therapy and other parameters on FMD.
Of the 37 subjects in this study, 22 were taking HIV PIs (group 1) and 15 were not taking these medications (group 2). The average age was 42.2±7.6 years. Differences between groups were not significant in regard to any nonlipid risk factors for CAD. Subjects in group 1 tended to have a higher body mass index (BMI) and waist-to-hip ratio; however, resting HR, SBP, and serum glucose levels were similar in both groups (Table 1). The average time since diagnosis of HIV infection was 84.5±42.0 months. Subjects had been taking antiretroviral therapy for 68.9±40.1 months. Subjects receiving HIV PIs had been taking these medications for 30.8±9.6 months. Indinavir was the most commonly used HIV PI. Only 2 subjects were taking ritonavir. Four subjects were taking 2 HIV PIs. All subjects were receiving nucleoside reverse transcriptase inhibitors (NRTIs), of which lamivudine, stavudine, and zidovudine were commonly used. More subjects in group 2 were receiving the non-nucleoside reverse transcriptase inhibitor (NNRTI) nevirapine than in group 1 (P=0.013). Otherwise, differences between the groups were not significant in regard to any of the clinical parameters related to HIV infection and therapy (Table 2). One subject in group 2 had used indinavir for 6 weeks but discontinued it >6 months before enrollment.
Lipid and Lipoprotein Parameters
Subjects in group 1 had mild hypercholesterolemia, with higher total serum cholesterol levels than subjects in group 2 (P=0.007). HDL-C concentrations were similar in both groups, but non–HDL-C levels were higher in group 1 (P=0.017). Group 1 subjects had marked hypertriglyceridemia, with higher total serum triglyceride levels than subjects in group 2 (P=0.009; Table 3).
Subjects in group 1 had elevated IDL and VLDL concentrations that were higher than in group 2 (P=0.023 and P=0.022, respectively). Elevated VLDL levels were due to higher concentrations of large VLDL particles, which include chylomicron remnants (P=0.012). Group 1 subjects tended to have higher levels of chylomicrons, larger VLDL particles, and small HDL particles; however, mean HDL particle diameters were similar in both groups. Although subjects in group 2 tended to have higher concentrations of intermediate-sized LDL particles, total LDL, LDL particle concentrations, and mean LDL particle diameters were similar in both groups. Subjects in both groups had predominantly small LDL (<20.5 nm, 73.6% versus 66.7%, P=0.717).
Significant correlations were not observed between any of the lipid and lipoprotein parameters and any of the clinical parameters related to HIV infection and therapy, including time since diagnosis of HIV infection and time on HIV PI therapy. Significant correlations were identified between several lipid and lipoprotein parameters, fasting serum glucose levels, SBP, waist-to-hip ratio, and BMI (data not shown). Subjects receiving lamivudine, stavudine, or zidovudine did not have significantly different lipid values than subjects not receiving these medications. Subjects receiving therapy only with NRTIs were similar to subjects receiving both NRTIs and NNRTIs in regard to lipid levels. Use of nevirapine did not affect lipid levels.
BA Ultrasound Parameters
Resting BA diameters and blood flow rates were similar in both groups. The increase in forearm blood flow was similar in both groups, indicating a similar stimulus to FMD of the BA. NTGMD was similar in both groups. Subjects in group 1 had markedly impaired FMD (2.6±4.6%), which was lower than the normal values observed in group 2 (8.1±6.7%, P=0.005; Table 4).
Significant correlations were not observed between FMD and any clinical, lipid, or lipoprotein parameter except SBP (r=0.404, P=0.014) and HR (r=−0.398, P=0.016). As expected, FMD and NTGMD correlated inversely with BA diameter (r=−0.460 and −0.457, P<0.010). Stepwise regression analysis identified use of HIV PIs, BA diameter, and HR as negative predictors of FMD. SBP was a positive predictor of FMD (P<0.0001). The final linear regression model identified significant contributions of these parameters and explained a significant amount of the observed variance in FMD (model 1, FMD=0.134−0.052×HIV PIs−0.002×HR+0.002×SBP−0.522×BA diameter, R2adjusted=67.1%, standard error of estimate [SEE]=0.034, P<0.001).
To clarify which metabolic abnormalities predicted FMD, data from subjects receiving HIV PIs were analyzed separately. In group 1, stepwise regression analysis again identified BA diameter (P<0.001) and SBP (P=0.037) as predictors of FMD. The linear regression model identified significant contributions of both parameters, as well as chylomicron, HDL-C, and glucose levels (P<0.050). In group 1, the final model also included VLDL, IDL, and HR (P<0.100) and explained a significant amount of the variance in FMD observed in subjects receiving HIV PIs (model 2, FMD=0.108−0.005×chylomicrons−0.001×VLDL−0.002× IDL−0.003×HDL-C−0.0004×glucose−0.002×HR+0.003 × SBP−0.514×BA diameter, R2adjusted=70.6%, SEE=0.026, P=0.018).
BA diameter, HR, and SBP contributed independently to both models; however, in subjects receiving HIV PIs, chylomicron, IDL, VLDL, HDL-C, and glucose levels helped predict FMD. To determine whether the additional parameters identified in model 2 explained the contribution of HIV PIs to the variance in FMD identified in all subjects, chylomicron, IDL, VLDL, HDL-C, and glucose levels were added back to model 1, in place of the term representing use of HIV PIs. The resulting model explained 63% of the variance in FMD observed in the present study, which was similar to the explanatory power of model 1 (SEE=0.038, P<0.001). When these parameters were directly added to model 1, use of HIV PIs remained a significant predictor of FMD (P=0.012); however, the marginal sum of squares, a measure of the unique predictive information contained in this variable, decreased from 0.033 to 0.008, which indicates that most of the effect of HIV PIs was accounted for by the metabolic parameters added back to the model.
In group 2, FMD did not differ between subgroups receiving and not receiving lamivudine, stavudine, or zidovudine. Subjects taking NRTIs alone or in combination with NNRTIs also had similar values for FMD. Nevirapine use did not influence FMD.
Subjects taking HIV PIs had atherogenic lipoprotein changes characterized by elevated levels of triglyceride-rich lipoproteins and their cholesterol-rich remnants. Subjects taking HIV PIs also had impaired FMD of the BA, a marker of severe endothelial dysfunction, whereas endothelial function in subjects with HIV infection but not taking these medications was normal. The primary determinant of impaired endothelial function was use of an HIV PI. In subjects receiving these medications, chylomicron, VLDL, IDL, and HDL-C levels predicted FMD. This is the first in vivo demonstration of the atherogenicity of HIV PIs.
HIV PIs and Atherogenic Lipoproteins
Hypercholesterolemia and hypertriglyceridemia have been associated with use of HIV PIs; however, specific lipoprotein abnormalities have not been described in detail.1,5–7⇓⇓⇓ HIV infection per se is associated with hypertriglyceridemia and small, dense LDL particles.8 In the present study, subjects who were not receiving HIV PIs had mild hypertriglyceridemia, elevated VLDL, and small LDL particles, as reported previously.8 Subjects receiving HIV PIs also had small LDL particles, but additional atherogenic lipoprotein changes were identified, including a disproportionate elevation in large VLDL and chylomicron remnants and elevated IDL, a remnant of VLDL metabolism. Subjects in group 1 also tended to have higher levels of small, less protective HDL particles.15 The observation that subjects taking HIV PIs had elevated IDL and large VLDL levels is consistent with a previous report that VLDL and IDL cholesterol and triglyceride levels increased in healthy adults who took ritonavir for 2 weeks.16
The triglyceride levels in group 1 were higher than in previous studies despite infrequent use of ritonavir; however, triglyceride levels in group 2 also were high, as were BMIs and waist-to-hip ratios.1,5–8⇓⇓⇓⇓ Although ritonavir raises triglyceride levels, other HIV PIs can cause severe hypertriglyceridemia.16,17⇓ In one study,17 60% of subjects taking HIV PIs had triglyceride levels >5.6 mmol/L, even though fewer than one third were taking ritonavir. The average BMI in that study (24.9 kg/m2) was higher than in previous studies (range 22.4 to 24.7 kg/m2) but lower than in our subjects (25.6 kg/m2). The triglyceride-raising effects of HIV PIs may be amplified in subjects with central obesity, especially if such obesity is mediated by insulin resistance.1,5,6,17⇓⇓⇓
Because apolipoprotein concentrations and activities of enzymes involved in lipid metabolism were not measured, the relative contributions of overproduction and impaired lipoprotein clearance to the abnormalities observed in the present study could not be discerned. Hepatic overproduction of VLDL, impaired action of apolipoprotein CIII, or modulation of the effects of HIV PIs by different apolipoprotein E isoforms may explain some of the lipoprotein abnormalities observed in the present study.16,18,19⇓⇓ Insulin resistance, which was not assessed formally in the present study, also may have affected lipoprotein metabolism.1,5,6⇓⇓
HIV PIs and Endothelial Dysfunction
Endothelial dysfunction of the brachial artery and endothelial dysfunction of the coronary artery are correlated with each other and predict future adverse cardiovascular events.20–23⇓⇓⇓ In the present study, subjects receiving HIV PIs had markedly impaired endothelial function. Other than BA diameter, which is the denominator of the formula for calculating FMD, use of HIV PIs was the primary predictor of FMD. In subjects receiving HIV PIs, triglyceride-rich lipoproteins and their cholesterol-rich remnants predicted FMD, consistent with previous observations that these lipoproteins are associated with endothelial dysfunction.24,25⇓ When these parameters and glucose levels were added back to the original model (model 1), they appeared to account for most but not all of the effect of HIV PIs on FMD. It is likely, therefore, that the lipoprotein abnormalities associated with use of HIV PIs contributed to the impaired endothelial function observed in patients taking these medications. Because glucose levels also contributed to the observed variance in FMD, it is possible that insulin resistance also affected FMD.26
Neither FMD nor any of the lipid or lipoprotein parameters correlated significantly with clinical parameters related to HIV infection, including CD4 cell count, HIV RNA titer, time since diagnosis of HIV infection, time on antiretroviral therapy, or time on HIV PI therapy. The lipid and lipoprotein changes observed in the present study, therefore, were not simply reflections of improved health, better immune status, or enhanced viral suppression.5 Indeed, CD4 cell counts and HIV RNA titers were similar in both groups.
Although this study was small, the BA ultrasound technique for evaluating endothelial function is very sensitive and reproducible.11–13⇓⇓ The sample size was based on published nomograms; however, the relatively small size could have limited detection of small differences between groups and among the HIV PIs, which may differ in their lipid effects.7,13⇓ Nonetheless, scientifically and clinically important differences were identified in regard to FMD and levels of atherogenic lipoproteins. The magnitude of these differences permitted development of regression equations that accounted for more than two thirds of the observed variation in FMD.
Because of the cross-sectional design, unmeasured baseline differences between subjects in each group may have affected outcome parameters. Both groups, however, were very similar in age, CAD risk factors, time since diagnosis of HIV infection, time on antiretroviral therapy, CD4 cell counts, and viral loads. Except for a difference in use of nevirapine, which did not affect FMD or lipid values, both groups were also similar in use of NNRTIs and NRTIs. Furthermore, consecutive subjects arriving in clinic were screened for participation, eliminating a potential source of selection bias. With the similarities between groups and the consecutive recruitment method used, it is unlikely that baseline differences accounted for the large differences in endothelial function, lipids, and lipoproteins observed in the present study. Given the effects of glucose levels in the multivariate models and the detrimental effects of insulin resistance on endothelial function, it is possible that insulin resistance also may explain some of the variance in FMD.26 Finally, because only a single time point in the course of a chronic disease was evaluated, longitudinal changes in end points could not be assessed.
Conclusions and Clinical Implications
HIV PIs are associated with atherogenic lipoprotein changes and endothelial dysfunction. These findings suggest that the metabolic changes associated with use of HIV PIs, including their adverse effects on triglyceride-rich lipoproteins and their cholesterol-containing remnants, are atherogenic. Patients receiving HIV PIs should be screened for hyperlipidemia and may be candidates for lipid-lowering therapies that improve endothelial function and prevent adverse cardiovascular events, depending on their long-term prognosis and risk of cardiovascular disease.27–29⇓⇓ The potential for drug interactions between lipid-lowering medications and HIV PIs should also be considered. Clinical decisions regarding initiation and intensification of drug therapies for patients with HIV infection should include their adverse effects on lipids, lipoproteins, and endothelial function.
This study was supported by the University of Wisconsin Cardiovascular Research Center and the General Clinical Research Center (NIH 3 M01RR03186-1551).
Dr Sosman serves as a consultant to, and receives research funding from, Agouron Pharmaceuticals, GlaxoWellcome Inc, and Merck and Co, Inc. Dr Otvos serves as a corporate officer and board member of LipoMed, Inc; he is also a stockholder of that firm.
- ↵Henry K, Melroe H, Huebsch J, et al. Severe premature coronary artery disease with protease inhibitors. Lancet. 1998; 351: Letter.
- ↵Periard D, Telenti A, Sudre P, et al, for the Swiss HIV Cohort Study. Atherogenic dyslipidemia in HIV-infected individuals treated with protease inhibitors. Circulation. 1999; 100: 700–705.
- ↵Otvos JD. Measurement of lipoprotein subclass profiles by nuclear magnetic resonance spectroscopy.In: Rifai N, Warnick G, Dominiczak MH, eds. Handbook of Lipoprotein Testing. 2nd ed. Washington, DC: AACC Press; 2000: 609–623.
- ↵Freedman DS, Otvos JD, Jeyarajah EJ, et al. Relation of lipoprotein subclasses as measured by nuclear magnetic resonance spectroscopy to coronary artery disease. Arterioscler Thromb Vasc Biol. 1998; 18: 1046–1053.
- ↵Stein JH, Keevil JG, Wiebe DA, et al. Purple grape juice improves endothelial function and reduces the susceptibility of LDL cholesterol to oxidation in patients with coronary artery disease. Circulation. 1999; 100: 1050–1055.
- ↵Sorenson KE, Celermajer DS, Spiegelhalter DJ, et al. Non-invasive measurement of human endothelium dependent arterial responses: accuracy and reproducibility. Br Heart J. 1995; 74: 247–253.
- ↵Lamarche B, Moorjani S, Cantin B, et al. Associations of HDL 2 and HDL 3 subfractions with ischemic heart disease in men: prospective results from the Quebec Cardiovascular Study. Arterioscler Thromb Vasc Biol. 1997; 17: 1098–1105.
- ↵Purnell JQ, Zambon A, Knopp RH, et al. Effect of ritonavir on lipids and post-heparin lipase activities in normal subjects. AIDS. 2000; 15: 51–57.
- ↵Lenhard JM, Croom DK, Weiel JE, et al. HIV protease inhibitors stimulate hepatic triglyceride stimulus. Arterioscler Thromb Vasc Biol. 2000; 20: 2625–2629.
- ↵Al Suwaidi J, Hamasaki S, Higano ST, et al. Long-term follow-up of patients with mild coronary artery disease and endothelial dysfunction. Circulation. 2000; 101: 948–954.
- ↵Schachinger V, Briton MB, Zeiher AM. Prognostic impact of coronary vasodilator dysfunction on adverse long-term outcome of coronary heart disease. Circulation. 2000; 101: 1899–1906.
- ↵Kugiyama K, Doi H, Motoyama T, et al. Association of remnant lipoprotein levels with impairment of endothelium-dependent vasomotor function in human coronary arteries. Circulation. 1997; 97: 2519–2526.
- ↵Lundman P, Eriksson M, Schenck-Gustafsson K, et al. Transient triglyceridemia decreases vascular reactivity in young, healthy men without risk factors for coronary heart disease. Circulation. 1997; 96: 3266–3268.
- ↵Balletshoffer BM, Rittig K, Enderle MD, et al. Endothelial dysfunction is detectable in young normotensive fist-degree relatives of subjects with type 2 diabetes in association with insulin resistance. Circulation. 2000; 101: 1780–1784.
- ↵O’Driscoll G, Green D, Taylor RR. Simvastatin, an HMG-coenzyme A reductase inhibitor, improves endothelial function within 1 month. Circulation. 1997; 95: 1126–1131.