Prospective, Intensive Study of Metabolic Changes Associated with 48 Weeks of Amprenavir-Based Antiretroviral Therapy

Abstract

To determine whether a 48-week course of amprenavir-based antiretroviral therapy is associated with metabolic alterations, 14 clinically stable human immunodeficiency virus (HIV)–infected, protease inhibitor–naive adults initiated amprenavir-based triple therapy. Twelve subjects (86%) achieved HIV RNA levels of <400 copies/mL at week 24. Fasting glucose and insulin levels did not change. Insulin sensitivity did not decrease in the first 24 weeks, but a trend toward a decrease appeared at week 48. Six subjects experienced onset or worsening of glucose tolerance by week 24. Levels of fasting triglycerides and low-density lipoprotein, high-density lipoprotein, and total cholesterol increased. Bone mineral content, lean tissue, total fat, trunk fat, limb fat, and the ratio of trunk to limb fat increased at week 48. Amprenavir-based therapy was associated with increases in serum lipid levels but no short-term decrease in insulin sensitivity. A trend toward insulin resistance appeared late in the study following weight gain, particularly of trunk fat, but without loss of limb fat

Potent antiretroviral therapy regimens containing HIV type 1 protease inhibitors have been associated with the development of new-onset diabetes mellitus [1–4]. Cross-sectional studies suggest that PI use is associated with reduced insulin sensitivity [5, 6]. Longitudinal studies have documented reductions in insulin sensitivity within several months of initiation of PI-based regimens in HIV-infected subjects [7, 8] and during indinavir monotherapy in healthy subjects [9]. We recently reported that administration of indinavir-based combination therapy for 8 weeks was associated with a 30% reduction in insulin sensitivity, as measured by minimal modeling of the intravenous glucose tolerance test (IVGTT) [7]

Laboratory studies suggest that the PI amprenavir may not cause some of the effects on lipid and glucose metabolism that have been observed with use of other PIs [10, 11]. Because of concern about potential increases among HIV-infected patients in cardiovascular risk attributable to dyslipidemia [12], insulin resistance [13–15], and fat redistribution [14], we conducted a pilot detailed metabolic and body composition assessment of PI-naive subjects, beginning treatment with a potent antiretroviral regimen that included amprenavir. We hypothesized that amprenavir would be associated with fewer metabolic effects than are other PIs

Subjects, Materials, and Methods

Study subjects Subjects were recruited from a university-based public hospital outpatient facility. Appropriate informed consent was obtained, and the guidelines for human experimentation of the University of Southern California were followed in conducting the clinical research. HIV type 1–infected patients were eligible if they had the following laboratory values: fasting glucose level, <110 mg/dL; CD4 cell count, ⩾100 × 10⁶ cells/L; HIV RNA level, >500 copies/mL. Patients who had received a PI for ⩾7 days previously, who had received a PI within the 60 days preceding the study, or who had received abacavir or both lamivudine and stavudine were excluded from the study. Twelve of the 14 subjects who were included in the analysis were antiretroviral naive, and 2 had only received nucleoside reverse-transcriptase inhibitors. Patients were ineligible if they had a history of diabetes mellitus or adrenal insufficiency, had used a rifamycin within the 7 days preceding the study, had an active opportunistic infection, had experienced a change in body weight of >5% within the 2 months preceding the study, or had used medications known to alter glucose tolerance within the 30 days preceding the study. No dietary or exercise interventions were offered. Reproducibility data indicated that inclusion of 12 subjects would allow detection of a clinically significant (30%) change in insulin sensitivity [16]. A target enrollment of 15 subjects was planned

Drug therapy After baseline studies, 12 the of 14 subjects who were included in the analysis began treatment with amprenavir (Agenerase; GlaxoSmithKline), 1200 mg, with abacavir (Ziagen; GlaxoSmithKline), 300 mg, and lamivudine (Epivir; GlaxoSmithKline), 150 mg, all taken orally twice daily. Two subjects who had a history of lamivudine therapy took stavudine (Zerit; Bristol-Myers Squibb), 40 mg twice daily, instead of lamivudine. One subject who experienced a possible hypersensitivity reaction to abacavir during the study was treated with didanosine (Videx; Bristol-Myers Squibb) instead, without interruption in antiretroviral therapy. Administration of other antiretroviral agents, immune modulators, lipid-lowering agents, hormonal agents, or agents known to affect insulin sensitivity or blood glucose levels was not allowed

Evaluations Subjects were evaluated at baseline and at 2, 8, 16, 24, 32, 40, and 48 weeks after treatment was initiated. At baseline and at 2, 8, 24, and 48 weeks, fasting body weight was measured and subjects underwent an oral glucose tolerance test (OGTT) and an IVGTT on consecutive days at the General Clinical Research Center of the Los Angeles County–University of Southern California Medical Center. Subjects were instructed to take their usual morning dose of medications with water at 6 A.M. and to report to the hospital after a 10–12-h overnight fast. All subjects were questioned by at least 2 study personnel regarding adherence to the fasting requirement and did not undergo evaluation if any other oral intake had occurred

OGTT Seventy-five grams of dextrose in water was administered orally over the course of <5 min. Blood samples were drawn from an indwelling intravenous catheter for plasma glucose and insulin determinations at 10 and 5 min before and 30, 60, 90, and 120 min after ingestion of dextrose

IVGTT Whenever possible, an antecubital vein catheter was used for administration of intravenous glucose. A second catheter placed retrograde in a dorsal hand vein was used for collection of samples, and the hand was warmed to 55°C during the procedure. At least 30 min after placement of the catheters, baseline samples were drawn, and dextrose (300 mg/kg of body weight) was given intravenously for 1 min. Tolbutamide (125 mg/m² of body surface area) was given as an intravenous bolus 20 min after the dextrose injection. Twenty-two samples were obtained over the course of 240 min and were assayed for glucose and insulin levels

Assays Specimens of blood for glucose and insulin determinations were collected in iced, heparinized tubes. Plasma was separated within 20 min of collection and stored at -70°C until the day of assay. Plasma glucose levels were measured by the glucose oxidase method. Plasma insulin was measured by RIA. Plasma HIV RNA levels were measured with the Amplicor Ultrasensitive assay (Roche Diagnostic Systems), unless the subject's HIV RNA level was >75,000 RNA copies/mL, in which case the standard Roche Amplicor assay was used. Total cholesterol, high-density lipoprotein (HDL) cholesterol, triglyceride, and free fatty acid levels were measured by enzymatic methods. Plasma lipoproteins were assayed by a nuclear magnetic resonance technique (NMR Lipoprofile; Liposciences) [17]

Dual-energy x-ray absorptiometry (DEXA) Subjects underwent whole-body DEXA on a QDR 4500W scanner (Hologic). Scans were analyzed for total body and regional composition by one of the authors (C.M.)

Food diaries At baseline and 8, 24, and 48 weeks, subjects completed a standardized 3-day food-intake record. Analysis was done with use of Nutritionist IV software (First DataBank)

Data analysis The primary end points of this study were changes in insulin sensitivity and fasting lipid levels. Insulin sensitivity was derived by minimal model analysis of the IVGTT insulin and glucose data [18]. Fasting plasma glucose and insulin levels were determined in duplicate, and the results were averaged. Insulin resistance was calculated by use of the homeostasis model assessment (HOMA-IR) [19]: HOMA-IR = (insulin in μU/mL × glucose in mM)/22.5. Total and incremental area under the curve (AUC) for the glucose and insulin levels measured by OGTT were calculated by the trapezoid method. Pancreatic B cell response to intravenous glucose (acute insulin response to glucose) was calculated as the incremental AUC for insulin during the first 10 min after glucose injection. The insulin response to oral glucose was calculated as the incremental ratio of insulin to glucose (change in insulin/change in glucose) at 30 min (ΔI₃₀/ΔD₃₀). Low-density lipoprotein (LDL) cholesterol levels were calculated by the Friedewald formula [20]

Descriptive results were reported as mean ± SE or median (range), depending on the distribution of outcome variables. Changes from baseline at each assessment were tested by the Wilcoxon signed rank test. The overall variation among values at baseline and at 2-, 8-, 24-, and 48-weeks was evaluated with repeated-measures analysis of variance

Results

Sixteen subjects were enrolled. One committed suicide before undergoing week 2 evaluations, and 1 was murdered before undergoing week 8 evaluations. The analysis is limited to the remaining 14 subjects who completed week 24 of study and 11 subjects who completed week 48. One subject in whom tuberculosis was diagnosed during the study and who received a rifabutin-containing antituberculous treatment regimen was removed from the study after week 24 because of virologic failure. Two subjects were removed from the study after week 24 because of poor adherence to study visits and relapses of substance abuse. None of the 14 subjects who were included in the analysis had received pentamidine. Demographic characteristics and metabolic responses to HIV disease treatment are shown in table 1

Table 1

Demographic characteristics and responses to treatment for 14 subjects included in a study of the association between amprenavir-based antiretroviral therapy and metabolic alterations

The parameters of glucose metabolism that were measured in this study are shown in table 2. Over the course of the 48-week study period, there were no significant changes in the mean fasting plasma glucose concentration. The fasting glucose concentration did not indicate impaired fasting glucose (defined as a plasma glucose level of 110–125 mg/dL) or diabetes mellitus (defined as a plasma glucose level of ⩾126 mg/dL) [21] in any subject. Mean insulin sensitivity did not change significantly during the initial 24 weeks of study but was lower at week 48. Insulin sensitivity was 4.9 ± 0.8 min^-1 per μU/mL × 10^-4 at baseline and 3.0 ± 0.4 min^-1 per μU/mL × 10^-4 at week 48 (P =.06, by the Wilcoxon signed rank test; P =.04, by repeated-measures analysis of variance). Measures of insulin secretion (B cell responsiveness to intravenous or oral glucose) did not change

Table 2

Results of measurement of glucose metabolism parameters at baseline and 2, 8, 24, and 48 weeks after initiation of amprenavir-based antiretroviral therapy for 14 subjects included in a study of the association between amprenavir and metabolic alterations

During OGTT, there was no change in the 30-min plasma insulin or glucose levels or the AUCs for insulin or glucose. There was a progressive increase in the 120-min plasma glucose level, from 114 ± 7 mg/dL at baseline to 146 ± 13 mg/dL at week 24 (P =.02 for the difference between baseline and week 24, by the Wilcoxon signed rank test), but at week 48, plasma glucose decreased to 122 ± 11 mg/dL (P =.35). Three subjects had impaired glucose tolerance (defined as a 2-h glucose concentration by OGTT of 140–199 mg/dL) at baseline. Two of these 3 subjects developed diabetes mellitus (defined as a 2-h glucose concentration by OGTT of >200 mg/dL), one at week 8 and the other at week 24. The third subject who had impaired glucose tolerance at baseline continued to display impaired glucose tolerance at weeks 8 and 24. Four subjects who had normal glucose tolerance at baseline developed impaired glucose tolerance by week 24. Three of these 4 also had impaired glucose tolerance at either the week 2 or the week 8 evaluation. A post hoc analysis of predictors of the new development of impaired glucose tolerance or diabetes mellitus found no significant predictive factor among any of the variables examined

Lipid levels are shown in table 3. Total cholesterol levels increased 35% (P <.001). Calculated LDL cholesterol levels increased 28% (P =.002). Triglyceride levels increased 90% (P =.009). Severe hypertriglyceridemia (>500 mg/dL) was not seen, and only 1 subject had triglyceride levels >400 mg/dL (447 mg/dL) at any time point. HDL cholesterol levels increased by 21% at week 48 (P =.003). Non-HDL cholesterol levels increased by 38% (P <.001). The ratio of total cholesterol to HDL cholesterol was significantly increased at week 2, and a trend toward an increased ratio of total cholesterol to HDL cholesterol was present at week 24 (P =.07), but the ratio was not increased significantly at week 48 (P =.64). Free fatty acid levels were significantly increased from baseline at weeks 2–24 but were not significantly increased at week 48 (P =.21). Significant increases in total cholesterol, calculated LDL cholesterol, non-HDL cholesterol, triglyceride, and free fatty acid levels were seen as early as week 2. Lipoprotein particle analysis revealed a small but statistically significant increase in estimated intermediate-density lipoprotein (IDL) cholesterol concentrations (P =.03). The mean LDL particle size did not increase. Significant increases in LDL particle concentration paralleled the increases in LDL cholesterol concentration estimated by nuclear magnetic resonance and calculated from the Friedewald equation. There were no significant shifts in the concentrations of any individual lipoprotein particle subclasses

Table 3

Fasting lipid levels before and 2, 8, 24, and 48 weeks after initiation of amprenavir-based antiretroviral therapy for 14 subjects included in a study of the association between amprenavir and metabolic alterations

DEXA results are shown in table 4. Body weight as measured by DEXA increased by 3.3 ± 1.3 kg at 24 weeks and by 4.9 ± 1.8 kg at week 48 (P =.03 for both comparisons with baseline). Total lean body mass increased by 1.5 ± 0.8 kg (P =.07), and bone mineral content increased by 0.04 ± 0.01 kg (P =.02) between baseline and week 48. Total body fat increased by 3.4 ± 1.2 kg (P =.03), and the body fat percentage increased by 3.2% ± 1.0% (P =.01). There was no association between the change in total body fat at week 48 and total body fat at baseline (P =.23), the change in total body fat at week 48 and baseline body mass index (P =.25), or the change in the percentage of total body fat at week 48 and the percentage of total body fat at baseline (P =.52). There was an increase of borderline significance in limb fat (1.2 ± 0.6 kg; P =.054) and a significant increase in trunk fat (2.2 ± 0.7 kg; P =.01). At week 48, trunk fat correlated with reduced insulin sensitivity (Pearson correlation coefficient, -0.72; P =.01). Only 1 subject experienced a decrease in trunk fat and limb fat during the study; all other subjects experienced an increase in both trunk fat and limb fat. The mean ratio of trunk fat to limb fat was higher at week 48 than at baseline (P =.005)

Table 4

Variables measured by dual-energy x-ray absorptiometry at baseline and 24 and 48 weeks after initiation of amprenavir-based triple therapy for 14 subjects included in a study of the association between amprenavir and metabolic alterations

Subjects' self-perceived body changes tended to parallel the DEXA results. At baseline, 6 (38%) of 16 subjects considered themselves to be underweight, 8 (50%) considered their weight to be “just right,” and 2 (13%) considered themselves to be somewhat or very overweight. At week 48, none of the 11 subjects considered themselves to be underweight, 5 (45%) considered their weight to be “just right,” and 6 (55%) considered themselves to be somewhat or very overweight. When asked “do you think that your appearance has changed since you began study treatment,” 3 (27%) of 11 subjects reported no change in belly size, whereas 8 (73%) had noticed an increase. When asked about a change in the amount of flesh in the face at week 48, 5 (45%) of 11 subjects responded that there was no change, and 6 (55%) had noticed a gain. Similarly, 7 (64%) of the 11 subjects stated that they had no change in leg size and 4 (36%) reported an increase. No subject complained of a decrease in the size of their legs or in the amount of flesh in the face since the initiation of study medications

Discussion

Administration of amprenavir-based combination antiretroviral therapy for 48 weeks was associated with a trend toward decreasing insulin sensitivity by the minimal model. However, there was no decrease in mean insulin sensitivity during the initial 24-week period of drug administration, although effects on oral glucose tolerance were seen (6 subjects developed new or had worsening glucose intolerance). Mean total body fat and trunk fat increased significantly within the first 24 weeks and continued to increase to the 48-week time point, as did the mean ratio of trunk fat to limb fat, suggesting that the decrease in mean insulin sensitivity may have been a consequence of central adiposity rather than a direct effect of the drug. Significant increases in various lipid levels occurred, without changes in dietary macronutrient intake. Treatment with the combination of amprenavir, abacavir, and lamivudine (stavudine was substituted for lamivudine for 2 subjects) was effective and well tolerated, with the exception of a single instance of a possible hypersensitivity reaction to abacavir

Administration of combination antiretroviral therapy including the PI indinavir for 8 weeks was associated with a significant decrease in mean insulin sensitivity by the minimal model [7], and similar findings have been reported for healthy subjects who received indinavir only for 4 weeks [9] or who received single doses of indinavir [22]. Cross-sectional [5, 6] and longitudinal [8] studies have also associated PI use with insulin resistance. Unlike our previous study, in which we investigated indinavir-based therapy in a similar population and used IVGTT [7], we did not demonstrate the development of insulin resistance in our subjects in the short term (⩽24 weeks in the present study). However, the development of new or worsening oral glucose intolerance, despite the absence of any changes in AUCs for glucose and insulin, as determined by OGTT, suggests that amprenavir is not entirely free of effects on glucose metabolism

The degree of weight and fat gain among our subjects was significant. Importantly, although most subjects experienced an increase in both limb and trunk fat, as determined by DEXA, the amount of trunk fat increase was greater, and the ratio of trunk fat to limb fat increased significantly. We suspect that this increase in adiposity may have been responsible for the decrease in mean insulin sensitivity observed at week 48. Mynarcik et al. [23] reported a close correlation between the degree of reduced limb fat and insulin resistance among HIV-infected subjects who had both peripheral lipoatrophy and central obesity. In the present study, however, loss of limb fat did not occur, but, rather, an increase occurred in all subjects but one, and the mean increase of 1.2 ± 0.6 kg was of borderline statistical significance (P =.054). Similarly, self-assessed facial thinning was not reported by subjects. This suggests that, although trunk adiposity did increase, the subjects in this trial did not experience the prevalent form of HIV treatment–associated lipodystrophy, in which a considerable majority of patients have lipoatrophy with or without obesity [24]. However, because of the relatively short follow-up period of 48 weeks, sufficient time may not have elapsed for subjects to have developed lipoatrophy [25]

Significant changes in lipid levels occurred during amprenavir-based combination therapy. Mean levels of fasting total cholesterol increased by 35%, of triglycerides by 90%, of HDL cholesterol by 21%, and of calculated LDL cholesterol by 28% over the course of 48 weeks. Changes in macronutrient intake did not occur. This suggests that at least part of the dyslipidemia is a direct effect of the drug regimen used. It is also possible that improved virologic control or improved immunity may contribute to lipid changes, but studies of healthy persons treated with monotherapy with the PI ritonavir have also documented dyslipidemia analogous to that experienced by HIV-infected subjects [26]. Nuclear magnetic resonance results tended to confirm the findings of standard lipid assays and documented no significant shift in lipoprotein subclasses. The increase in the mean total cholesterol level appeared to be predominantly due to an increase in the LDL fraction, with a smaller contribution from IDL and HDL. Importantly, there was no decrease in mean LDL particle size, which would be expected to further increase the atherogenic potential of dyslipidemia [27]. The use of other PIs was not associated with increases in HDL cholesterol in some studies [8, 28], but HDL cholesterol was found to have increased in other studies [29, 30]. An increase in HDL cholesterol during therapy may indicate a “return-to-health” phenomenon. In the present study, the significant increases in mean HDL cholesterol (6 mg/dL; P =.003) would tend to mitigate the proatherogenic tendency of the increased LDL cholesterol concentrations. This is confirmed by absence of a significant change in the mean ratio of total cholesterol to HDL cholesterol at week 48

The absence of acute effects on insulin sensitivity, as measured by minimal modeling of the IVGTT results, in the face of significant effects on triglyceride and free fatty acid levels, is intriguing. In both healthy [9] and HIV-infected subjects [7], indinavir has been shown to have no significant effects on lipid levels, but decreases in insulin sensitivity were seen. This paradox suggests that the effects of PIs on glucose metabolism and lipid metabolism may be the result of distinctly different pathogenic mechanisms and, further, that these mechanisms may be differentially affected by different PIs. It has been suggested that insulin resistance is a consequence of the lipid abnormalities induced by PI therapy [31]. However, these studies [7, 9, present study] provide evidence against this hypothesis

In summary, administration of potent triple-combination therapy including the PI amprenavir for 48 weeks was associated with increases in cholesterol and triglyceride levels, as well as increased adiposity, particularly in the trunk. Worsening oral glucose tolerance occurred, but a short-term reduction in insulin sensitivity was not seen. A trend toward reduction in insulin sensitivity, however, became apparent after weight gain and increased adiposity occurred. These observations suggest that PI-induced hyperlipidemia does not directly induce insulin resistance and that, unlike indinavir, amprenavir does not directly induce insulin resistance. Further study is needed to identify PIs that have minimal effects on metabolism

Acknowledgments

We are indebted to the staff of the Los Angeles County–University of Southern California Medical Center's General Clinical Research Center; Gina-Bob Dubé, for bibliographic support; Reid Gibson (Indiana University Endocrine Laboratory); and the patients of the 5P21-Rand Schrader Clinic who volunteered for this study, without whom this work would not have been possible. We also thank Jim Lenhard and the members of the GSK COL30309 study team: Toni Cates, Katherine Levesque, Ashwin Hirani, Julie Fleming, and Jaime Hernandez

Financial support: GlaxoSmithKline; National Institutes of Health (M01-RR-43)