Switch from PI to Abacavir: affect on fat, glucose
"Markedly Diminished Lipolysis and Partial Restoration of Glucose Metabolism, without Changes in Fat Distribution after Extended Discontinuation of Protease Inhibitors in Severe Lipodystrophic HIV+ Patients"
The Journal of Clinical Endocrinology & Metabolism July 2004, Vol. 89, No. 7 3554-3560
Marc van der Valk, Gideon Allick, Gerrit Jan Weverling, Johannes A. Romijn, Mariette T. Ackermans, Joep M. A. Lange, Berthe L. F. van Eck-Smit, Cornelis van Kuijk, Erik Endert, Hans P. Sauerwein and Peter Reiss
International Antiviral Therapy Evaluation Center (M.v.d.V., J.M.A.L.) and Departments of Endocrinology and Metabolism (G.A., H.P.S.), Clinical Epidemiology and Biostatistics (G.J.W.), Academic Medical Center, University of Amsterdam, Amsterdam 1105 AZ, The Netherlands; Department of Endocrinology (J.A.R.), Leiden University Medical Center, Leiden 2300 RC, The Netherlands; and Department of Clinical Chemistry (M.T.A., E.E.), Laboratory of Endocrinology and Radiochemistry, Departments of Infectious Diseases (J.M.A.L., P.R.), Tropical Medicine and AIDS, Nuclear Medicine (B.L.F.v.E.-S.), and Radiology (C.v.K.), Academic Medical Center, University of Amsterdam, Amsterdam 1105 AZ, The Netherlands
Treatment for HIV-1 infection is often complicated by a lipodystrophy syndrome associated with insulin resistance and an elevated rate of lipolysis (breakdown of fats into other compounds). In eight HIV-1 infected men with lipodystrophy syndrome, we studied the effects of replacement of protease inhibitor (PI) by abacavir on insulin sensitivity and lipolysis by hyperinsulinemic euglycemic clamp and on fat distribution assessed by dual-energy x-ray absorptiometry and computed tomography scan.
Glucose metabolism and lipolysis were assessed by tracer dilution employing [6,6-2H2]glucose and [2H5]glycerol, respectively. Data are expressed as mean ± SD or 95% confidence interval (CI), as appropriate.
There were no significant changes in fat distribution assessed by dual-energy x-ray absorptiometry and computed tomography scan at wk 36 and wk 96. The fasting total glucose production decreased from 16.1 ± 2.5 at study entry by 1.1 (range, --2.1 to --0.1) to 15.0 ± 1.5 µmol/kg·min after PI withdrawal at wk 36 (n = 8). In an analysis restricted to the patients on treatment at wk 96 (n = 6), the decrease was 0.9 (range, --2.1 to 0.3) µmol/kg·min. During insulin infusion, glucose oxidation (as percent of total glucose disposal) increased from 36.8 ± 12.7% by 11.0% (range, 1.3--20.8) to 47.9 ± 13.9% in the wk 36 analysis. In the analysis restricted to the patients on treatment at wk 96 (n = 6) the increase was 7.7 (--4.0 to 19.4)%. Fasting lipolysis decreased from 2.7 ± 0.6 µmol/kg·min by 0.9 (--1.6 to --0.2) to 1.8 ± 0.3 µmol/kg·min in the wk-96 analysis (n = 6).
The replacement of the studied PIs by abacavir in severe lipodystrophic HIV-1-infected patients results in a marked reduction of lipolysis. In contrast, fasting glucose production and insulin-stimulated glucose oxidation improve moderately, whereas insulin-stimulated glucose disposal and fat distribution do not change.
THERE HAVE BEEN numerous reports on the emergence of metabolic disturbances related to antiretroviral treatment of HIV-1 infection. These involve a lipodystrophy syndrome (LD) consisting of peripheral fat loss, with or without central fat accumulation; hyperlipidemia; and disturbances in glucose metabolism. Both nucleoside reverse transcriptase inhibitors (NRTIs) and the protease inhibitors (PIs) have been implicated in the pathogenesis of the syndrome. No consensus yet exists of which of these components constitute the LD. The body fat changes described above are associated with metabolic disturbances, regardless of the presence of HIV infection and/or antiretroviral therapy, as was clearly demonstrated in HIV-1-uninfected patients suffering from congenital LDs. Moreover, in a severely lipoatrophic insulin-resistant murine model, the sc implantation of autologous fat resulted in an almost complete reversal of insulin resistance. Therefore, it seems likely that some of the disturbances seen in HIV-associated LD are secondary to the changes in body fat. Moreover, differences in clinical phenotype might result in different patterns of metabolic disturbance.
Lipodystrophic HIV-1-infected patients using PIs have severe insulin resistance with respect to the peripheral uptake of glucose. Furthermore, fasting glucose production was increased, and there was hepatic insulin resistance with respect to the suppressive effects of insulin on glucose production, compared with both healthy volunteers and untreated HIV-1 infected patients. Administration of a single dose of the PI indinavir to healthy volunteers resulted in an acute, but transient, decrease in insulin sensitivity, indicative of a direct effect of this PI on glucose metabolism, independent of any changes in fat distribution. Moreover, the administration of the PI ritonavir to healthy volunteers resulted in a significant increase in triglycerides, indicative of the fact that certain PIs may exert direct effects on lipid concentrations as well.
Accordingly, in vitro experiments have undisputedly shown that certain of the PIs impair glucose transport by the inhibition of the intrinsic activity of the glucose transporter GLUT-4. The effect of the withdrawal of PIs on glucose metabolism in HIV-1-infected patients with LD who have been exposed long-term to PIs has not been precisely delineated. Two small studies using either fasting plasma glucose, insulin, or an iv insulin tolerance test have reported an improvement, whereas another study reported no changes in glucose metabolism. With respect to whole-body lipolysis, glycerol turnover in patients with lipodystrophy using PIs is increased to the same extent as in untreated HIV-1 infected patients. No studies have yet described the effects of PI withdrawal on lipolysis.
Therefore, an important question in HIV-associated lipodystrophy concerns the extent to which PIs per se contribute to the disturbances in glucose metabolism once treatment-induced changes in fat distribution have been established. To answer this question, we conducted a prospective study in HIV-1-infected patients with severe lipodystrophy who, at the time of inclusion, were using PI-based therapy. The different components of glucose metabolism and lipolysis were evaluated by hyperinsulinemic glucose clamp both before and 96 wk after the replacement of the PIs in their regimen by the NRTI abacavir. Furthermore, any changes in fat distribution were objectified by dual-energy x-ray absorptiometry (DEXA) and computed tomography (CT) scan.
The results of our study demonstrate that 2 yr of withdrawal of PI in HIV-1-infected patients with severe lipodystrophy does not result in an improvement of disturbed fat distribution, whereas glucose metabolism and lipolysis are partially restored and markedly diminished, respectively.
Although fasting glucose production decreased after PI withdrawal, it remained elevated when compared with values previously obtained both in healthy volunteers and in untreated HIV-infected patients. Furthermore, despite 2 yr of PI withdrawal, insulin-stimulated peripheral glucose disposal did not improve. In the wk-36 analysis, the degree of suppression of endogenous glucose production by insulin improved, indicating improved hepatic insulin sensitivity. However, in the wk-96 analysis, this decrease was no longer present in the current study. The rate of lipolysis, which was increased during fasting in HIV-1-infected patients with lipodystrophy, returned toward the levels we have previously measured in healthy volunteers. These metabolic changes occurred in the absence of significant improvements in body composition.
The mechanism by which PI acutely causes peripheral insulin resistance involves a direct inhibition of GLUT-4. Interestingly, in the severely lipodystrophic HIV-1-infected patients enrolled into our study, the extended withdrawal of PI did not result in significant improvement of insulin-stimulated peripheral glucose disposal. This may indicate that long-term treatment with PI irreversibly impairs the intrinsic activity of GLUT-4. Alternatively, this may be explained by direct or indirect changes in glucose metabolism induced by persisting lipodystrophy.
The trend for lipolysis to normalize after PI withdrawal occurred independent of any changes in concentrations of plasma insulin, a potent inhibitor of lipolysis, epinephrine and cortisol, both potent stimulators of lipolysis, and fat distribution. An explanation for the decrease in lipolysis might be a lowered tonus of the sympathetic nervous system (tonus: the normal state of a healthy muscle when resting in a state of slight contraction). However, this was not reflected by a decrease in norepinephrine concentrations. In HIV-1-infected lipodystrophic patients using PI, norepinephrine concentrations are elevated compared with untreated HIV-1-infected individuals, suggestive of an overall increased tonus of the sympathetic nervous system. Lipolysis nevertheless is similarly increased both in HIV-1-infected patients with lipodystrophy and in treatment-naïve HIV-1 patients when compared with healthy volunteers. This indicates that although the factor responsible for stimulating lipolysis in uncontrolled HIV infection has disappeared, it must have been replaced by another mechanism. Based on our findings, we propose that PI in an as-yet-unexplained fashion induces changes in the tonus of the sympathetic nervous system, resulting in increased whole-body lipolysis that gradually normalizes after PI withdrawal.
A potential bias of our study might be the fact that patients aged during the 2 yr of follow-up, which thereby may have counteracted any improvement PI withdrawal might have had on insulin sensitivity. Although aging is associated with a decrease in total glucose disposal, this decrease was demonstrated to be as low as 0.9 µmol/kg·min·decade of life. Furthermore, it seems that the age-related decrease in insulin sensitivity is mainly due to a reduction in insulin-stimulated glucose oxidation that did not change over time in our study. Moreover, glucose metabolism was also assessed at wk 12 of the study, and no acute improvements in insulin sensitivity were observed. (data not shown) Taken together, this makes it highly unlikely that aging explains why total glucose disposal did not improve significantly.
Another important limitation of our study is the small sample size and the fact that one patient died before completing 96 wk and that, in two of the eight patients, PI had to be reintroduced because of virological failure. This occurred in patient 5 at wk 36 and in one patient at wk 7. In both patients virological failure likely was the result of multiple preexisting zidovudine resistance conferring mutations, which were retrospectively demonstrated in stored specimens (data not shown). Unfortunately, at the time the study was designed and these patients were enrolled, the results of trials showing such patients to be at increased risk of virological failure when replacing PI by abacavir were not yet available. The small sample size of our study most likely explains why moderate improvements in fat distribution, which were reported after PI withdrawal in larger controlled trials, were not observed. Nevertheless, despite a limited sample size, we were able to demonstrate significant changes in lipolysis and glucose metabolism. The decrease in glucose production in the wk-36 analysis was statistically significant and remained lowered at wk 96, although no longer to a statistically significant degree, probably resulting from a decrease in statistical power. A similar pattern was seen for glucose oxidation expressed as percentage of total glucose disposal during the clamp, which improved in the wk-36 analysis but was no longer significant in the wk-96 analysis.
In summary, PI replacement by abacavir in severe lipodystrophic HIV-1-infected patients after 96 wk resulted in a clear trend for lipolysis to normalize. In contrast, fasting endogenous glucose production improved modestly, whereas insulin-stimulated glucose disposal and fat distribution did not change significantly. Taken together, this suggests that mechanisms in addition to inhibition of GLUT-4 activity are responsible for some of the changes in glucose metabolism seen in HIV-infected patients with established lipodystrophy. Finally, we propose that disappearance of PI-mediated increases in sympathetic nervous system activity may underlie the decrease in lipolysis after PI withdrawal.
SUBJECTS & METHODS
HIV-1-positive men with lipodystrophy were studied, who were included in the 'reverse' study. The main objective of this study was to assess the reversibility of the various components of the LD in HIV-1-infected patients after replacement of the PI-component in patients' antiretroviral regimens by the NRTI abacavir. Patients could be referred for the study if, in the opinion of their treating physician, they had lipodystrophy. Before being included, this had to be independently confirmed by two of us (M.v.d.V., P.R.) based on medical history and physical examination. Lipodystrophy was defined as the presence of peripheral lipoatrophy, central fat accumulation, or a combination of both. Furthermore, patients had to use a PI-containing regimen with a plasma HIV-1 RNA level having been less than 400 copies/ml for at least 6 months. Patients with diabetes mellitus, defined as having a fasting glucose concentration above 7.0 mmol/liter, were excluded.
Six weeks after having added abacavir (300 mg, two times daily) to their current regimen, patients were randomized to either discontinue their PI immediately or continue these for another 12 wk and then stop.
At five timepoints during the course of the study, an euglycemic hyperinsulinemic glucose clamp was performed (at the time of randomization = wk 0 and 12, 36, 72, and 96 wk after the randomization). Fat distribution was assessed at study entry and at wk 96.
The first participant was included in December 1999, and the last reached wk 96 in February 2003. All participants used a balanced diet, containing at least 250 g carbohydrates, 3 d before each metabolic study, and were instructed to try to maintain their current weight. The study was approved by the institutional review board of the Academic Medical Center in Amsterdam. Written informed consent was obtained from all participants before study entry.
Subjects were admitted to the metabolic clinical research center and studied in the supine position. After a 12-h fast, a catheter was inserted into the antecubital vein of each arm. One catheter was used for sampling of arterialized blood using a heated handbox (60 C). The other catheter was used for infusion of [6,6-2H2]glucose, glucose 20%, [2H5]glycerol, and insulin. At 0900 h (t = --2 h.), after drawing a blood sample for background enrichment of plasma glucose and glycerol, a continuous infusion of [6,6-2H2]glucose (>99% enriched; Cambridge Isotopes, Andover, MA) was started at a rate of 0.22 µmol/kg·min after a priming dose was administered that equaled 80 min of infusion. At 1000 h (t = --1 h), continuous infusion with [2H5]glycerol at 0.11 µmol/kg·min was started after a priming dose of 1.6 µmol/kg. At t = +0, 10, 20, and 30 min, blood samples were drawn for determination of basal endogenous glucose production and basal glycerol turnover. Subsequently, at t = +30 min, a primed continuous infusion of insulin (Actrapid 100 EH/ml; Novo Nordisk Farma B.V., Alphen ad Rijn, The Netherlands) was started for 2.5 h at a rate of 20 mU·m--2 body surface area·min--1. Plasma glucose concentration was measured every 5 min (Beckman glucose analyzer 2; Beckman Coulter, Palo Alto, CA), and glucose 20% was infused at a variable rate to maintain plasma glucose at 5.0 mmol/liter. [6,6-2H2]Glucose was added to the 20% glucose solution to achieve glucose enrichments of 2% to minimize changes in isotopic enrichment due to changes in the infusion rate of exogenous glucose, and thus to allow for accurate quantification of endogenous glucose production. The last hour of insulin infusion every 10 min, blood samples were drawn for determination of endogenous glucose production and glycerol turnover. During the study, subjects were allowed to drink only water.
Oxygen consumption (VO2) and CO2 production (VCO2) were measured by indirect calorimetry using a ventilated hood system (Sensormedics model 2900; Sensormedics, Anaheim, CA). VO2 and VCO2 were measured continuously during the final 30 min of both the basal and the hyperinsulinemic periods.
Total, as well as regional, fat mass was quantified in all patients by DEXA (Hologic QDR-4500W; Hologic, Inc., Bedford, MA; software version whole-body v8.26A: 5), providing a quantitative assessment of peripheral and truncal fat mass in kilograms. The ratio between peripheral fat mass (defined as the sum of arm and leg fat) and total fat mass (total fat mass minus head fat) was calculated to adjust for differences in body weight. A standardized single-slice abdominal CT scan through the level of the fourth lumbar vertebra was performed, from which the surface of total adipose tissue (TAT), visceral adipose tissue, and sc adipose tissue (SAT) was determined and expressed in square centimeters. The ratio between SAT and TAT was calculated to assess fat distribution.
Plasma insulin concentration was determined by an RIA (Insulin RIA 100; Pharmacia Diagnostic AB, Uppsala, Sweden; intraassay coefficient of variation (c.v.), 3--5%; interassay c.v., 6--9%; detection limit, 15 pmol/liter). Plasma samples for plasma catecholamine concentrations and enrichments of [6,6-2H2]glucose and [2H5]glycerol were determined as described before (12, 14). Cortisol was determined with a competitive chemiluminescent immunoassay (Immulite; Diagnostic Products Corporation, Los Angeles, CA).
Calculations and statistics
Endogenous glucose production and total glucose disposal were calculated by non-steady-state Steele equations as described previously (12). Steele's equations for steady-state conditions as adapted for the use of stable isotopes were used to calculate glycerol rate of appearance.
Glucose oxidation was calculated from VO2 and VCO2. Nonoxidative glucose disposal was calculated as the difference between total glucose disposal and glucose oxidation. Both glucose oxidation and nonoxidative glucose disposal are expressed as percentages of total glucose disposal, to adjust for differences in the total glucose disposal.
Patient characteristics at entry and patient disposition
The mean age of the patients included was 47 ± 8 yr. The mean weight was 76 ± 16 kg at study entry and modestly increased over time by 2.7 kg at wk 96 (n = 6). All patients had HIV-1 viral loads less than 50 copies/ml at the time of study entry. At study entry, the mean CD4 cell count was 540 ± 295 cells/mm3. Antiretroviral drug history, details of the regimens used, and body fat composition at study entry for each of the patients are shown in Table 1.
One patient was excluded from the entire analysis because of an increase in plasma HIV-1 RNA, 7 wk after the withdrawal of PI, necessitating reinstitution of PI-containing antiretroviral therapy (new regimen d4T/ddI/3TC/RTV/IDV) before wk 12. Patient 5 had to restart PI-containing therapy because of virological failure at wk 36 of the study (new regimen ddI/3TC/EFV/LPV/RTV). Soon after the reintroduction of PI, the plasma HIV-1 RNA load became undetectable again. Patient 6 died at home, possibly from an acute myocardial infarction, a few weeks after his wk-36 study visit. In none of the remaining patients was antiretroviral therapy changed during follow-up. These two patients were excluded from the wk-96 analysis, resulting in follow-up data of six patients at wk 96.
There were no statistically significant changes in basal insulin, glucose, catecholamines, and cortisol concentrations over time. In contrast, plasma total- and low-density lipoprotein cholesterol had decreased significantly after PI withdrawal.
The study physician (M.v.d.V.) and the patients noted no significant improvements in fat distribution over time. This was confirmed by whole-body DEXA scans that showed no improvement either in absolute total peripheral fat mass or peripheral fat expressed as a proportion of total fat. In addition, CT scans showed no significant changes in visceral adipose tissue and SAT surface area or in the ratio of SAT to TAT.
Endogenous glucose production
At study entry, the mean fasting glucose production was 16.1 ± 2.5 µmol/kg·min (n = 8). PI withdrawal resulted in a mean decrease of 1.1 (95% CI, --2.1 to --0.1) µmol/kg·min in fasting glucose production at wk 36. The analysis of wk-96 data (n = 6) showed a similar decrease of 0.9 (95% CI, --2.1 to 0.3) µmol/kg·min, resulting in a fasting glucose production of 15.0 ± 1.9 µmol/kg·min at wk 96. During the clamp, endogenous glucose production was 7.9 ± 2.7 µmol/kg·min at study entry and decreased modestly to 6.3 ± 1.8 µmol/kg·min at wk 36 (n = 8). In the wk-96 analysis (n = 6), endogenous glucose production decreased by 0.7 (95% CI, --2.7--1.3) µmol/kg·min, resulting in a wk-96 endogenous glucose production of 7.3 ± 0.7 µmol/kg·min.
At wk 12, no differences were observed in the changes in endogenous glucose production, both during the clamp and under fasting conditions, between the participants randomized to immediate or deferred withdrawal of PI (P = 0.98 and P = 0.55, during the clamp and fasting, respectively).
Peripheral glucose metabolism
During fasting at study entry, the glucose oxidation expressed as a percentage of total glucose disposal was 30.5 ± 10.4%. This did not change over time in both the wk-36 and wk-96 analyses. Of note, glucose oxidation expressed as a percentage of total glucose disposal during the clamp improved from 36.8 ± 12.7% at study entry by 11.0 (95% CI, +1.3--20.8%) resulting in 47.9 ± 13.9% in the wk-36 analysis. In the wk-96 analysis, the increase was 7.7% (95% --4.0--19.4%), resulting in a glucose oxidation of 44.9 ± 6.4% of total glucose disposal.
At wk 12, no differences were observed in the changes in glucose oxidation during fasting between the participants randomized to immediate or deferred withdrawal of PI (P = 0.88).
At study entry, the hyperinsulinemia during the clamp increased total glucose disposal by 32 (95% CI, 17--47) % from the fasting value. In the 36-wk analysis, this increase was 26% (95% CI, 0--39) and 20% (95% CI, --3--44) in the 96-wk analysis.
At study entry, total glucose disposal during the clamp was 21.3 ± 4.9 µmol/kg·min. Insulin-stimulated total glucose disposal during the clamp did not improve after PI withdrawal in both the wk-36 and wk-96 analysis. At wk 12, no differences were observed in either total glucose disposal or glucose oxidation during the clamp between the participants randomized to immediate or deferred withdrawal of PI (P = 0.26 and P = 0.62, for total glucose disposal and glucose oxidation, respectively).
Fasting glycerol turnover was 2.6 ± 0.6 µmol/kg·min at study entry. In the wk-36 analysis, fasting glycerol turnover decreased by 0.4 (95% CI, --0.9 to --0.1) to 2.2 ± 0.7 µmol/kg·min. The 96-wk analysis showed a similar pattern, with a decrease of 0.9 (95% CI, --1.6 to --0.2) to 1.8 ± 0.3 µmol/kg·min at wk 96. In the wk-36 analysis during the clamp, glycerol turnover was 1.8 ± 0.6 µmol/kg·min, which decreased by 0.4 (95% CI, --0.9 to --0.01) to 1.4 ± 0.5 µmol/kg·min. In the wk-96 analysis, the decrease was --0.6 (95% CI, --1.4--0.1). At study entry, the clamp decreased glycerol turnover by 32 (95% CI, 23--40). In the wk-36 analysis, the suppressive effect of insulin was 36% (95% CI, 21--51) and 37% (95% CI, 16--58) in the 96-wk analysis.
Both during the clamp and under fasting conditions at wk 12, no differences were observed in the changes in glycerol turnover between the participants randomized to immediate or deferred withdrawal of the protease inhibitor (P = 0.5 and P = 1.0 during the clamp and fasting, respectively).