Background. Dyslipidemia alone does not fully explain the increase in cardiovascular events among patients receiving protease inhibitor (PI)-based treatment for human immunodeficiency virus infection. Some PIs, such as indinavir, directly induce endothelial dysfunction, an effect that may mediate that portion of the increase in cardiovascular events that is not attributable to dyslipidemia.
Methods. Endothelium-dependent vasodilation, insulin-mediated vasodilation, and whole-body and leg glucose uptake during use of a 1-h euglycemic hyperinsulinemic clamp (insulin infusion, 40 mU/m2/min) were measured in healthy men before and after 4 weeks of treatment with placebo (12 men), with 400 mg atazanavir per day (9 men), or with 400 mg lopinavir and 100 mg ritonavir twice per day (9 men).
Results. Median age (36 years) and mean body mass index ± SD (23.4±2.6; calculated as weight in kilograms divided by the square of height in meters) did not differ between groups. Endothelium-dependent vasodilation, expressed as the percentage change in the leg blood flow response to intrafemoral artery infusion of 15 ≪g/min of the endothelium-dependent vasodilator methacholine, did not change after 4 weeks of treatment in any group: mean percentage change ± SD, 154±102 from baseline and 242±254 at week 4 with atazanavir (P=.36), 76±62 and 86±79, respectively, with lopinavir-ritonavir (P=.68), and 111±86 and 127±153, respectively, with placebo (P=.63; for between-group differences, P=.55). The response to the endothelium-independent vasodilator nitroprusside was not different at week 4 for any group, nor was insulin-mediated vasodilation or leg or whole-body insulin-mediated glucose uptake (all within-group P values were >.1).
Conclusions. Unlike the dramatic impairment seen with indinavir, the newer PIs atazanavir and lopinavir-ritonavir do not induce endothelial dysfunction in healthy subjects. Thus, endothelial dysfunction does not appear to be a PI drug class effect. The cause of the non-lipid-mediated increase in cardiovascular events that are reported with PIs remains unclear.
Combination antiretroviral (ARV) therapy for HIV infection is associated with an increased risk of myocardial infarction, particularly with the use of HIV-1 protease inhibitors (PIs) [1, 2]. Prospective data suggest that ARV therapy-associated lipid disorders alone do not explain all of this increased risk . Endothelial dysfunction is a critical initial step of atherogenesis that contributes to the progression and clinical manifestations of atherosclerosis [3, 4]. PI-based ARV regimens have been associated with endothelial dysfunction . Thus, PI-related endothelial dysfunction may be responsible for increased cardiovascular events during PI therapy that are not merely due to PI-associated lipid changes.
Four weeks of administration of the HIV-1 PI indinavir significantly impaired endothelial function in healthy, nonobese, HIV-uninfected subjects in 3 different studies [6–8]. The magnitude of indinavir-induced endothelial dysfunction in these studies was great, comparable to the degree seen with type 2 diabetes mellitus. However, indinavir is now seldom used in clinical practice. In the United States, the newer PIs atazanavir and the fixed-dose combination of lopinavir and ritonavir account for nearly 70% of the total PI use as of August 2007 (IMS National Prescription Audit, 24 August 2007). We hypothesized that lopinavir-ritonavir would have an adverse effect on endothelial function, because of its adverse effects on serum lipid parameters [9, 10] or, as reported in some studies , insulin resistance. We further hypothesized that atazanavir, which does not cause adverse effects on lipids and insulin resistance [10, 11], would not affect endothelial function. In a randomized trial involving a group of healthy, nonobese subjects without HIV infection, we measured the effect of standard doses of atazanavir, lopinavir-ritonavir, and placebo, each given for 4 weeks, on endothelium-dependent vasodilation, insulin sensitivity at the level of the whole body and skeletal muscle, and insulin effects on the vasculature.
Materials and Methods
Subjects. Demographic characteristics of all participants are shown in table 1. All subjects (n=30) were men (mean age ± SD, 37±1 years) who were HIV seronegative; were healthy; were not obese (total body fat,≤27% by dual radiographic absorptiometry scan; Lunar DPX-L, system software 4.6b); were normotensive by cuff measurements, per Joint National Committee VI criteria ; had normal 75-g oral glucose tolerance tests, per American Diabetes Association criteria ; had normal lipid profiles, per National Cholesterol Education Program III criteria ; and were not taking any over-the-counter or prescription medications, including herbal supplements. None of the women who were screened for the study met the total body fat entry criterion. Studies were approved by the Indiana University-Purdue University and Clarian Health Partners institutional review boards, and all volunteers gave written informed consent. Subjects were instructed to maintain their usual dietary and physical activity habits during the study.
Study drugs. All vascular infusion drugs were diluted in normal saline, achieving concentrations of 25 ≪g/mL of methacholine chloride (Clinalfa) and 7 ≪g/mL of sodium nitroprusside (Roche Laboratories). Atazanavir capsules (200 mg; Reyataz; Bristol-Myers-Squibb) and soft-gel lopinavir-ritonavir capsules (133 mg lopinavir and 33.3 mg ritonavir; Kaletra; Abbott) were kindly provided by their respective manufacturers. Participants were randomly assigned, in a double-blind fashion, to receive 400 mg of atazanavir (two 200-mg capsules) per day plus lopinavir-ritonavir placebo (9 men), 400 mg/100 mg lopinavir-ritonavir (3 soft-gel capsules, each containing 133.3 mg lopinavir and 33.3 mg ritonavir) twice per day with food plus atazanavir placebo (9 men), or placebos for both drugs (12 men). Randomization was stratified by body mass index of >24 or≤24, calculated as weight in kilograms divided by the square of height in meters. Participant compliance was assessed by self-report and manual pill counts.
Procedures. Thirty healthy men were studied during November 2003-September 2006. Subjects were admitted to the Indiana University General Clinical Research Center 1 day before study onset and were fed a weight-maintaining diet. All vascular and metabolic studies (figure 1) were performed after an overnight fast and abstinence from smoking. The final doses of study drugs were administered at 7 am on the last day of the study, ∼1 h before methacholine infusion and ∼5 h before the final hour of the clamp study. Subjects had body weight, height, basal heart rate, and systolic and diastolic blood pressure measurements taken at baseline and after 4 weeks of study drug administration.
Vascular measurements were assessed as follows. Hemodynamic measurements (figure 1) were obtained while subjects were in the supine position, as described elsewhere . Leg blood flow was measured using a modified thermodilution technique. Baseline leg blood flow measurements were taken at least 30 min after catheter placement, and then leg blood flow was measured in response to graded intrafemoral artery infusions of 5, 10, and 15 ≪g/min of the endothelium-dependent vasodilator methacholine. After a washout period following the methacholine infusion, during which a return of leg blood flow values to baseline levels was documented, leg blood flow was then measured in response to graded infusions of 1.75, 3.5, and 7 ≪g/min of the endothelium-independent vasodilator sodium nitroprusside. Subsequently, a euglycemic hyperinsulinemic clamp study (described below) was initiated and continued for a minimum of 240 min. During the fourth hour of the insulin infusion, insulin-mediated vasodilation was assessed by measurement of leg blood flow and was compared with baseline leg blood flow values. The first 14 subjects received an infusion of NG-mono-methyl-l-arginine (L-NMMA) at the end of the clamp study (data not shown).
Metabolic measurements were assessed as follows. Immediately after baseline assessments of leg blood flow, a euglycemic hyperinsulinemic clamp study with an insulin infusion of 40 mU/m2/min was initiated, as described elsewhere . Participants were studied in the supine position, with a catheter inserted in an antecubital vein for infusion of glucose, insulin, and potassium. The femoral arterial and venous lines described above were used to obtain blood samples. The clamp technique was a modification of the method of DeFronzo et al. . Arterial plasma glucose values were determined at 5-min intervals, and the glucose infusion rate was adjusted to maintain normoglycemia (target plasma glucose level, 90 mg/dL). Venous plasma glucose values were measured at 20-min intervals for the first 3 h of the clamp study and at 5-min intervals for the last hour of the clamp study. The clamp procedure was continued for at least 240 min after initiation of the insulin and glucose infusions. To prevent hypokalemia, potassium was infused during the studies, at a rate of 0.0038 meq/kg/min.
Analytical methods. Glucose was assayed by the glucokinase method at the bedside with use of a Yellow Springs Instruments apparatus. Plasma adiponectin, resistin, and total plasminogen activator inhibitor 1 levels were measured by immunoassay (LINCOPlex assay; Millipore). Serum lipids were measured by standard enzymatic techniques. Low-density lipoprotein cholesterol levels were calculated using the Friedewald formula . Atazanavir, lopinavir, and ritonavir levels were measured by high-performance liquid chromatography at the University of Alabama at Birmingham Antiviral Pharmacology Laboratory.
Calculations. Changes in leg blood flow are expressed as absolute change and percentage change, to adjust for differences at baseline. Whole-body glucose uptake (M) was calculated as the mean glucose infusion rate normalized for body weight and corrected for changes in the glucose pool sizes, according to methods described by DeFronzo et al. , with use of data from a 1-h period late in the clamp period (160-220 min after initiation) before any vasoactive agents were infused. This period of time was used to avoid the effect of any L-NMMA infusion on glucose-uptake results. Arteriovenous glucose differences—arterial glucose level minus venous glucose level—were calculated every 5 min during the last hour of the clamp study. Leg glucose uptake was calculated as the product of leg blood flow measurement times the arteriovenous glucose difference during the last hour of the clamp study.
Statistical analysis. The prespecified primary study end point was endothelium-dependent vasodilation, expressed as the increase in leg blood flow from baseline (percentage change) in response to maximal methacholine (15 ≪g/min) . The study was powered to show a within-arm change of 1 SD by a paired Student's t test with 80% power to detect a difference with a nominal significance level of .05. In our data from a previous study of indinavir , the mean percentage change (±SD) from baseline in leg blood flow during maximal doses of methacholine was 145±80 from baseline values. In the current study, we had power to detect within-arm changes of 80% from baseline, a change that was considerably smaller than what was observed with indinavir. Results are given as the mean ± SD. A paired Student's t test was used to compare within-arm differences in metabolic and vascular variables before and after administration. To compare results across study arms, analysis of variance was used for continuous variables, and Fisher's exact test was used for categorical variables. Statistical significance was defined as P⩾.05. Analyses were performed with SAS, version 9.1 (SAS Institute).
Metabolic effects. Treatment had no effect on body mass index or fasting plasma glucose level (table 1). Plasma levels of adiponectin, resistin, and plasminogen activator inhibitor 1 did not change in any group (table 1). Minor but statistically significant increases in serum aminotransferase levels occurred only in the placebo group. A doubling of total bilirubin levels occurred with atazanavir treatment, and a slight but statistically significant increase in total bilirubin also occurred with placebo. Total cholesterol levels increased statistically significantly only with placebo, but in the overall analysis of variance, the change in cholesterol levels between groups was not statistically significant. Levels of triglycerides increased statistically significantly only with lopinavir-ritonavir (triglyceride level increase, from 77±29 mg/dL to 128±48 mg/dL; P=.002; for between-groups difference, P=.002). Low-density lipoprotein cholesterol and non-high-density lipoprotein cholesterol levels did not change in any group, whereas an increase in high-density lipoprotein cholesterol level occurred only in the placebo group (table 1).
Mean glucose concentrations during the clamp period (table 2) were similar before and after study treatment. Arteriovenous glucose difference, leg glucose uptake, and whole-body glucose uptake (M) were not affected by any study treatment. Subjects had respective mean predose and postdose plasma drug levels as follows: atazanavir, 0.49±0.69 ≪g/mL and 1.23±1.19 ≪g/mL; lopinavir, 5.20±5.17 ≪g/mL and 7.61±1.91 ≪g/mL; ritonavir, 0.33±0.35 ≪g/mL and 0.56±0.34 ≪g/mL.
Vascular effects. Mean arterial pressure decreased slightly with lopinavir-ritonavir treatment, from 94±6 mm Hg to 91±6 mm Hg (P=.03) (table 3). Leg blood flow increased in a dose-dependent fashion in response to the endothelium-dependent vasodilator methacholine before and after treatment with both study drugs and with placebo (table 3 and figure 2). The primary study end point of endothelium-dependent vasodilation—the increase in leg blood flow (percentage change) in response to maximal methacholine (15 ≪g/min)—did not change significantly after any of the study treatments. With lopinavir-ritonavir treatment (figure 2B) and with placebo (figure 2C), the dose-response curves were very similar before and after treatment. With atazanavir (figure 2A), the greater increase in leg blood flow after administration of the study drug was not statistically significantly different from baseline (P=.36). Leg blood flow in response to insulin during the final hour of the clamp increased similarly in all study arms, before and after study treatment (table 3).
The newer PIs atazanavir and lopinavir-ritonavir do not induce endothelial dysfunction in healthy subjects, in contrast to the dramatic impairment of endothelium-dependent vasodilation among healthy subjects seen with 4-week treatment with the older HIV-1 PI indinavir [6–8]. These observations suggest that endothelial dysfunction is not a PI drug class effect. Additionally, there was no change in insulin's effects on blood flow or insulin-mediated whole-body or leg glucose uptake with either study drug, indicating no impairment of insulin sensitivity.
Human data [5–8] and experimental models [18–22] have implicated PIs as a cause of endothelial dysfunction. The mechanism appears to be impaired nitric oxide bioavailability [8, 19]. Stein and colleagues  documented dyslipidemia and severe endothelial dysfunction in a cross-sectional observational study of subjects who received long-term PI-based ARV therapy (mean total duration, 70 months, which included 31 months of treatment with a PI) but not in those given treatment for HIV without a PI. Importantly, half of the subjects given PI treatment in that study received the PI indinavir . More contemporary studies, in which few subjects received indinavir, have failed to confirm a role for PI-containing ARV regimens in endothelial dysfunction . In fact, there was a statistically nonsignificant trend of better endothelial function measured by brachial flow-mediated dilation among subjects receiving PIs (predominantly nelfinavir and lopinavir-ritonavir) in that study . Similarly, use of the PI lopinavir-ritonavir was the strongest examined predictor of better endothelial function by brachial flow-mediated dilation in a small cross-sectional study .
Although not all PIs have been examined in this manner, in studies involving healthy subjects, only indinavir has been implicated as a cause of endothelial dysfunction [6–8]; thus, this effect may be agent specific. Consistent with our results, lopinavir-ritonavir administration to 6 healthy subjects for 4 weeks actually led to a statistically nonsignificant percentage increase in forearm blood flow during acetylcholine infusion, compared with baseline . In experimental models, only indinavir and ritonavir have been consistently implicated as causing endothelial dysfunction. In our study, only low-dose ritonavir (total daily dose, 200 mg) was used, which may have resulted in insufficient drug exposure to lead to endothelial dysfunction. As is the case with glucose and lipid metabolism effects , different PIs appear to have divergent effects on endothelial function.
No effects on glucose metabolism, including insulin-stimulated whole-body glucose uptake and leg glucose uptake, were observed with 4 weeks of study drug administration. This is in contrast to studies that demonstrated reduced whole-body glucose uptake, or insulin resistance, after 4 weeks of indinavir administration using a clamp insulin dose similar to that used in the present study [6, 27]. Similar negative results were seen with lopinavir-ritonavir administered for 4 weeks to healthy subjects , although 5 days of lopinavir-ritonavir administration caused insulin resistance in one study . We did not observe increases in adiponectin levels with either study drug, in contrast to previous reports that included indinavir [6, 28] and lopinavir-ritonavir .
We observed the expected increase in indirect bilirubin levels with atazanavir . Levels of triglycerides increased with lopinavir-ritonavir, whereas total and low-density lipoprotein cholesterol levels did not change, which are results similar to those of other studies involving healthy subjects [10, 28]. The cause of the increases in total and high-density lipoprotein cholesterol levels with placebo is not clear.
Limitations of this study include a relatively small number of subjects and a short observation time, which may not detect more-subtle long-term effects. However, significant endothelial function was documented with the PI indinavir in several studies that included fewer subjects studied for a similar period of time [6, 8]. It is also possible that delayed endothelial dysfunction may occur with PIs such as lopinavir-ritonavir because of long-term elevations in levels of triglycerides and potentially atherogenic lipoproteins. Because subjects without HIV infection were studied, our data cannot address the effect of HIV infection on endothelial function. Similarly, our data cannot address the effects of nucleoside reverse-transcriptase inhibitors  on endothelial dysfunction.
We conclude that the PIs atazanavir and lopinavir, when administered for 4 weeks to healthy, nonobese subjects, do not induce endothelial dysfunction or insulin resistance. The mechanisms behind the divergence of these results from the dramatic endothelial dysfunction seen with indinavir use [6–8] deserve further study. Because of the great metabolic and vascular diversity of existing PI agents, it will be important to study the effects of all PIs, both in common clinical use and in clinical development, for their metabolic and vascular properties.
We are indebted to the participants who volunteered for this study, to the staff of the Indiana University General Clinical Research Center, and to Kathy L. Flynn and Gina-Bob Dubá, for managing the references. Employees of Abbott and Bristol-Myers-Squibb reviewed the manuscript before submission and provided comments but were not involved in the approval of the final version or the decision about journal submission.
Financial support. National Institutes of Health (HL72711 and M01-RR00750) and drug gifts from Abbott Laboratories and Bristol-Myers-Squibb.
Potential conflicts of interest. M.P.D. has served as a consultant or received research support from Abbott, Bristol-Myers-Squibb, GlaxoSmithKline, and Merck. All other authors: no conflicts.
- body mass index procedure
- vascular flow
- endothelial dysfunction
- hiv protease inhibitors
- protease inhibitor
- nitroprusside sodium
- hiv infection
- cardiovascular event
- insulin sensitivity
- drug class effects
- glucose uptake
- infusion procedures