No effect of rosiglitazone for treatment of HIV-1 lipoatrophy: randomised, double-blind, placebo-controlled trial
The Lancet, Volume 363, Number 9407 07 February 2004
Andrew Carr presented the results of this study in briefer fashion in an oral session at Retrovirus, where I asked him if HIV+ individuals with diabetes were studied in this trial and he said no. The reason for asking about this is that its possible that rosiglitazone might have benefit regarding body changes for HIV+ diabetics because it shows benefit in HIV- diabetics regarding lipoatrophy. But this was not studied here. So the possibility that it might be effective in diabetics with HIV has not officially been addressed. A pilot study has been conducted in persons with more advanced insulin resistance, and this study suggested there might be benefits. Jules Levin
Andrew Carr, Cassy Workman, Dianne Carey, Gary Rogers, Allison Martin, David Baker, Handan Wand, Matthew Law, Katherine Samaras, Sean Emery, David A Cooper, for the Rosey investigators
St Vincent's Hospital, Sydney (A Carr MD, Prof D A Cooper DSc); AIDS Research Initiative, Sydney, Australia (C Workman MBBS); National Centre in HIV Epidemiology and Clinical Research, University of New South Wales, Sydney (D Carey MPH, A Martin MSc (Med), H Wand PhD, M Law PhD, S Emery PhD, Prof D A Cooper); Department of General Practice, University of Adelaide (G Rogers MBBS); 407 Doctors, Sydney (D Baker MBBS); and Garvan Institute for Medical Research, Sydney (K Samaras PhD).
Background: Lipodystrophy commonly complicates antiretroviral therapy of HIV-1 infection. Thiazolidinediones such as rosiglitazone promote subcutaneous fat growth in type 2 diabetics and adults with congenital lipodystrophy, and can prevent HIV-1 protease inhibitor toxicity to adipocytes in vitro. We postulated that rosiglitazone would improve HIV lipoatrophy.
Methods: 108 HIV-1-infected lipoatrophic adults on antiretroviral therapy were randomised to rosiglitazone 4 mg twice daily (n=53) or matching placebo (n=55) for 48 weeks. The study had 80% power to detect a 0.5 kg difference in changes in limb fat (using dual-energy X-ray absorptiometry) between groups at week 48 by intention-to-treat analysis, and a 0.7 kg difference within each protease inhibitor stratum.
Findings: Limb fat increased by 0.14 kg in the rosiglitazone group and 0·18 kg in the placebo group (mean difference -0·04 kg [95%CI –0.29 to 0.21]; p=0·74 by t test), with three participants (one on rosiglitazone and two controls), lost to follow-up. Rosiglitazone had no significant benefit on any other measure of lipodystrophy, despite large relative increases in plasma adiponectin (4.2 mmol/L [102%]; p<0·0001) and in three markers of insulin sensitivity (p=0·01 to 0·02). Six participants ceased study drug in each group, four participants (three on rosiglitazone and one control) for related adverse events. The main adverse effects, which seem to be almost unique to this population, were asymptomatic hypertriglyceridaemia (mean relative increase 0.9 mmol/L at week 48; p=0.04) and hypercholesterolaemia (1.5 mmol/L; p=0·001).
Lipodystrophy (peripheral lipoatrophy, central fat accumulation, and lipomatosis) affects at least 50% of HIV-1-infected adults receiving antiretroviral therapy, can be stigmatising and painful, and leads to suboptimal adherence to antiretroviral treatment. The associated dyslipidaemia and type 2 diabetes increase cardiovascular disease risk proportional to the duration of antiretroviral therapy, although these metabolic factors may not fully account for this increased risk.
Lipodystrophy is caused by some drugs within two of the four antiretroviral classes, namely, nucleoside analogue reverse transcriptase inhibitors and protease inhibitors, especially when both classes are combined.3,4 Lipoatrophic fat has reduced expression of the adipocyte differentiation factor peroxisome proliferator activator receptor-gamma (PPARG).5 Protease inhibitors may contribute to lipoatrophy by inhibition of sterol regulatory enhancer binding protein 1 (SREBP1)-mediated activation of the adipocyte retinoid-X-receptor:PPARG heterodimer. Nucleoside analogues may inhibit adipocyte mitochondrial DNA polymerase-( and so reduce the generation of mitochondrial proteins needed for oxidative phosphorylation.
The only proven intervention for HIV lipoatrophy is switching from a thymidine-based nucleoside analogue (particularly stavudine) to the nucleoside analogue abacavir. This strategy, however, led to only a moderate improvement in limb fat mass after 2 years. For reasons unknown, protease inhibitor cessation does not seem to improve lipoatrophy.
Thiazolidinediones are PPARG agonists that are effective for the treatment of type 2 diabetes, in part by partitioning fatty acids and glucose within adipocytes. In vitro, thiazolidinediones promote adipogenesis, even in the presence of an HIV-1 protease inhibitor.6 Thiazolidinedione treatment in diabetics and adults with congenital lipodystrophy increased peripheral fat, decreased visceral fat, and improved glycaemic abnormalities over 24 weeks, and might have had similar effects in autoimmune lipoatrophy. Five studies, two of which were randomised, assessed rosiglitazone or the related pioglitazone for HIV lipoatrophy, with variable outcomes. These studies were not powered adequately, however, to detect changes in lipoatrophy.
We undertook a randomised, placebo-controlled, 48-week trial to test the hypothesis that rosiglitazone would improve lipoatrophy in adults receiving antiretroviral therapy. The primary objective was to determine the effect of rosiglitazone 4 mg twice daily over 48 weeks on limb (arm plus leg) fat with dual energy X-ray absorptiometry (DEXA) (which cannot quantitate facial fat). Secondary endpoints comprised other body composition measures, metabolic measures and safety.
Interpretation: Rosiglitazone for 48 weeks did not improve lipoatrophy in HIV-1-infected adults receiving antiretroviral therapy. Use of less toxic antiretroviral treatment is necessary to prevent lipoatrophy.
Between December, 2001, and June, 2002, 149 patients were screened and 108 (72%) participants were randomised (figure 1). All patients had lipoatrophy on physical examination and self-report. Almost all (98%) participants were male, 18 (17%) had AIDS, and 66 (61%) were receiving a protease inhibitor. More participants in the rosiglitazone group than in the placebo group were receiving stavudine or zidovudine. The mean durations of all previous nucleoside analogue and protease inhibitor treatments, however, were similar. Six participants (four on rosiglitazone and two controls) were receiving lipid-lowering treatment at baseline.
Six (11%) rosiglitazone recipients interrupted therapy (four permanently), two for adverse events and four because of patient choice. Six (11%) placebo recipients ceased treatment, one for an adverse event and five out of their own choice. Dose of study drug was not reduced in any participant. The only lipid-lowering therapy commenced was gemfibrozil in one rosiglitazone participant.
14 antiretroviral drugs were stopped in each group (eight [15%] participants receiving rosiglitazone and nine [16%] participants receiving placebo). The most common drugs ceased were stavudine (three participants per group [weeks -2, 24, and 38 in the rosiglitazone group; weeks 24, 44, and 48 in the placebo group) and lamivudine (two participants per group). Five (9%) and three (5%) participants in the rosiglitazone and placebo groups, respectively, stopped taking a protease inhibitor. No participant died or developed a new AIDS-defining or cardiovascular event.
There were 35 protocol violations (18 in the rosiglitazone group, 17 in the placebo group) in 25 participants (13 and 12 participants, respectively), the most common being non-study lipid or glycaemic measurements (10 participants per group). No participant commenced lipid-lowering therapy as a result, however.
Rosiglitazone had no beneficial effect on limb fat (figure 2). The mean changes in limb fat mass at week 48 were 0.14 (SD 0·58) kg with rosiglitazone and 0.18 (0.68) kg with placebo, a difference of –0.04 (95% CI –0.29 to 0·21) kg (p=0.74 by t test). There was no significant effect on any other body composition or lipodystrophy endpoint (table 4). No difference in exercise intensity (data not shown), energy intake, or fat intake between the groups explained the lack of benefit. There was no significant effect noted in any subgroup at week 48 (table 5). Rosiglitazone caused a greater increase in limb fat at week 24 in those not receiving stavudine or zidovudine (rosiglitazone 0.48 kg, placebo 0.19 kg; p=0·06) than in those receiving stavudine or zidovudine (rosiglitazone –0.06 kg, placebo 0.09 kg; p=0.31; p value for interaction 0.05), but this difference was not maintained at week 48.
Rosiglitazone improved fasting insulin levels (figure 3), HOMA scores, and glucose to insulin ratios, even though no participant was diabetic or developed diabetes. Rosiglitazone increased plasma adiponectin at week 24 (3.6 mmol/L [+93%]; p<0·0001) and week 48 (4.2 mmol/L [+102%]; p<0·0001), but not plasma leptin at either timepoint. Alanine aminotransferase and alkaline phosphatase levels fell with rosiglitazone after 4 weeks and these declines were sustained.
Triglycerides, total cholesterol, and LDL cholesterol levels increased from week 4. 30 (57%) participants receiving rosiglitazone and 20 (36%) receiving placebo developed grade 3 or 4 hypertriglyceridaemia (p=0·0001), and 11 (21%) and four (7%) participants, respectively, developed grade 3 or 4 hypercholesterolaemia (p=0·0001). Increases in triglycerides over the first 12 weeks of therapy were significantly associated with use of rosiglitazone (odds ratio 7·3 [95%CI 2·0-27·0]; p=0·003) and higher baseline triglycerides (odds ratio 5·0 [2·6-9·9]; p<0·0001).
Haemoglobin concentrations fell significantly with rosiglitazone, but no participant developed grade 2 or greater anaemia. There was no grade 2 or higher increase in alanine or aspartate aminotransferase levels with rosiglitazone, and rosiglitazone had no significant effect on CD4-positive lymphocyte counts or plasma HIV viral load. There were 12 serious adverse events, seven in the rosiglitazone group (four participants) and five in the placebo group (five participants). Of these 12 events, one was definitely associated with rosiglitazone, two possibly associated with rosiglitazone, one possibly associated with placebo, and one remotely associated with rosiglitazone.
Rosiglitazone 4 mg twice daily did not improve lipoatrophy in HIV-infected adults receiving antiretroviral therapy. This result is in contrast to the sustained improvements in lipoatrophy that are seen after switching from a thymidine nucleoside analogue reverse transcriptase inhibitor (mainly stavudine), and in contrast to the benefit of troglitazone in congenital lipodystrophy.10-13,19 No sustained, significant trend was recorded in the five subgroups in which we postulated that rosiglitazone might be more active. The only possible benefit was seen in participants that were not receiving a thymidine nucleoside analogue, but this was seen only at week 24. As such, this result most likely represents a type 1 error or perhaps a true effect that is not sustained for 48 weeks and so of no clinical relevance. Although insulin sensitivity appeared to improve, hyperlipidaemia was observed, also reported by Sutinen and colleagues.
Our findings are in keeping with those of two small randomised studies that yielded negative outcomes, despite our study being substantially longer and, unlike the investigation done by Sutinen and colleagues, being powered for far less than a 40% change in subcutaneous fat mass. The possible beneficial effect reported by Hadigan and co-workers was not significant; it is not known whether their finding was affected by inclusion of only lipoatrophic patients with insulin resistance or by changes in antiretroviral treatment, diet, or exercise after baseline. The modest increases in limb fat seen in both groups in our study may be the result of thymidine nucleoside analogue cessation before baseline, but we cannot prove this. That almost all individuals in the study reported by Sutinen and co-workers were receiving stavudine might be relevant.
Since HIV lipoatrophy is associated with some residual PPARG expression in adipocytes, the lack of response to rosiglitazone was unexpected. There are several possible reasons for this. Perhaps the most likely is that stimulation of residual PPARG cannot adequately promote adipogenesis in the presence of antiretroviral treatment. Specifically, although rosiglitazone can reverse the lipoatrophic effect of protease inhibitors in isolated cell lines, whether rosiglitazone can reverse nucleoside analogue-induced lipoatrophy in vivo is unknown. Although the rosiglitazone group had a greater proportion of participants receiving stavudine or zidovudine at baseline than did the placebo group, the absence of a sustained, significant difference in outcome between the subgroups receiving or not receiving stavudine or zidovudine argues somewhat against such a possibility. Second, recovery of key genes that are also depleted in lipoatrophy such as glucose transporter 4 might be necessary; PPARG does not always upregulate these genes in vitro. Third, PPARG depletion in lipoatrophic fat and in vitro could be a by-product and not part of the pathogenetic pathway, even though rosiglitazone can both reverse and prevent adipocyte toxicity and PPARG deficiency with protease inhibitors.6 Fourth, PPARG agonists stimulate pre-adipocyte growth but can arrest growth in mature adipocytes in vitro. Poor absorption or cellular activity of rosiglitazone were less likely as seen by the low rate of participant withdrawal and its multiple metabolic effects. Lastly, it is remotely possible, but in our view unlikely, that the outcome was specific to rosiglitazone and that another thiazolidinedione such as pioglitazone might prove effective.
One explanation for the discordant effects of rosiglitazone on body fat and insulin sensitivity may lie in the finding that thiazolidinediones affect insulin sensitivity not only in adipocytes but also in skeletal muscle and the liver; PPARG expression in these latter organs is not known to be reduced in HIV lipodystrophy. Thiazolidinediones can also have discordant effects on adipocyte insulin sensitivity and growth in vitro. Our data argue against the suggestion that peripheral fat is critical for the insulin-sensitising effect of thiazolidinediones.
Mechanisms for the discordant leptin and adiponectin responses to rosiglitazone are unknown, although factors controlling the expression of each gene might differ. It is also possible that leptin but not adiponectin secretion is dependent upon adipocyte growth. No adipokine specific to subcutaneous adipocytes has been identified that could be measured to test whether rosiglitazone was active in the remaining lipoatrophic subcutaneous fat. Ongoing analysis of subcutaneous fat biopsies from some participants might clarify this.
Hypertriglyceridaemia is rare in HIV-uninfected diabetics treated with thiazolidinediones. PPARG is expressed in hepatocytes, the site of triglyceride synthesis, albeit to a lesser extent than in fat. Lipoatrophic mice receiving rosiglitazone have increases in hepatic triglyceride synthesis and steatosis, whereas HIV-infected and uninfected lipoatrophic adults receiving thiazolidinediones have reductions in liver fat. In our investigation, rosiglitazone treatment also reduced the marginally elevated alanine aminotransferase levels, a common feature of hepatic steatosis. Collectively, these data suggest that rosiglitazone increases hepatic triglyceride synthesis and secretion in human beings, and that diabetics with normal fat and those with congenital lipoatrophy, but not those with HIV lipoatrophy, are able to store this triglyceride within adipocytes.
Rosiglitazone cannot be recommended for the treatment of HIV lipoatrophy in adults receiving antiretroviral therapy, even though it has insulin-sensitising effects in this population (despite paradoxically increasing triglyceride levels). Its efficacy for lipoatrophy in participants receiving neither a thymidine nucleoside analogue nor a protease inhibitor remains unknown, as does any effect in women or children, and studies in these populations could be useful. Any use for insulin resistance or type 2 diabetes in lipoatrophic HIV-infected patients will need to take into consideration the adverse lipid effects we noted.
Our results have substantial implications for antiretroviral therapy of HIV infection. If lipoatrophy can only be improved partially and only by modifying antiretroviral treatment, then avoidance of implicated drugs will be critical for those patients wishing to
Participants were recruited at 17 HIV primary care (n=13) or hospital outpatient (n=4) sites in Australia. Eligibility criteria included documented HIV-1 infection, age greater than 17 years, and limb fat less than 20% of limb tissue or limb fat percentage at least 10% less than truncal fat percentage by DEXA. These DEXA criteria were chosen so that participants would be similar to those recruited in two earlier randomised HIV lipodystrophy studies. Participants had to have been receiving combination antiretroviral therapy unchanged for 12 weeks, and have no screening lipid and glycaemic indices needing treatment, apart from diet or exercise, and a negative pregnancy test. Women of childbearing potential had to be using contraception. Contraceptive, sex hormone replacement, and lipid-lowering treatments used in the preceding 6 months could be continued.
Patients were ineligible if they had HIV wasting syndrome, heart failure of New York Heart Association class 2 or greater, uncontrolled hypertension, any serious medical condition such as active AIDS, pancreatitis, or hepatitis within the previous 6 months, or if they were breast-feeding. Insulin, oral anti-diabetic agents, anabolic steroids (except testosterone replacement for proven hypogonadism), oral glucocorticosteroids at greater than replacement dose (prednisolone 7.5 mg per day or equivalent), growth hormone, other agents intended to stimulate appetite or weight gain, immune modulators, hydroxyurea, and cimetidine were not permitted. Laboratory exclusion criteria were serum transaminases, bilirubin and lactate greater than 2.5 times the upper normal limit, serum creatinine above the upper normal limit, haemoglobin less than 95 g/L, fasting glucose greater than 7.0 mmol/L, and fasting triglycerides greater than 15.0 mmol/L.
All participants provided written, informed consent after approval by the local human research ethics committee.
In the absence of an objective case definition when the study commenced, lipodystrophy was defined subjectively by presence of lipoatrophy or fat accumulation, or both, in the face, dorso-cervical spine, arms, breasts, abdomen, buttocks, or legs by use of standardised physical examination criteria. The objective lipodystrophy case definition and severity scoring system developed subsequently were incorporated before study completion as secondary endpoints. Virological failure was previously defined.
The study was designed to randomise equally 100 eligible participants to rosiglitazone 4 mg twice daily or matching placebo for 48 weeks. Rosiglitazone was chosen for its safety profile and lack of interaction with antiretroviral-metabolising cytochrome P450 isozymes. Participants continued all antiretroviral therapy when possible and were advised to maintain their current diet and exercise pattern. Randomisation was done by statisticians at the National Centre in HIV Epidemiology and Clinical Research (NCHECR), Sydney, and minimised by current protease inhibitor use, current global lipoatrophy severity on combined physical examination and patient report, and study site. Minimisation by protease inhibitor use was used since we were mindful of the possible different pathogenesis of lipoatrophy with protease inhibitors compared with nucleoside analogues. Lipoatrophy severity was a surrogate for limb fat mass, because we postulated that more severe fat loss might be associated with less fat gain. Minimisation based on limb fat mass would have needed central reading of each baseline scan within 4 weeks, something that could not be achieved for all 17 sites nationwide. Each lipoatrophy severity score was the sum of patient and physician scores for the face, arms, legs, and buttocks, a score known to correlate with peripheral fat mass; in each region, a score of 0 for nil, 1 for mild, 2 for moderate, or 3 for severe was assigned, and so the maximum lipoatrophy score was 24. Participants were minimised based on a final score less than or equal to 12 or greater than 12. Minimisation by site was used so that subjective assessments of lipodystrophy (a secondary endpoint) would be less prone to bias.
To ensure masking, only the study statistician had access to treatment allocation; lipid and glycaemic parameters, which could have been affected by active rosiglitazone, were measured at a central laboratory and were unavailable to participants or doctors for the duration of the trial. Emergency procedures were established to unmask study drug in the event of severe adverse reactions, but this did not happen.
Management guidelines for elevated hepatic transaminases, hyperglycaemia, hypoglycaemia, and anaemia included the potential toxicity profile of thiazolidinediones (including troglitazone, a drug withdrawn because of hepatotoxicity). These guidelines are not shown since none of these events occurred. Permanent cessation of study drug was mandatory for other grade 3 or 4 adverse events, pregnancy, or use of proscribed therapy. Grade 1 or 2 study drug-related adverse events were managed according to the treating clinician's assessment. Substitution of antiretroviral drugs was mandatory for related and recurrent grade 3 or 4 adverse events and optional for other adverse events or for virological failure.
Demographical characteristics, previous antiretroviral treatment, and concomitant therapies were recorded at screening. Participants were seen at weeks 0, 4, 8, and 12 weeks, and then every 6 weeks to week 48, including nine of the 12 participants who ceased randomised treatment but agreed to follow-up.
Safety assessments included clinical adverse events, use of concomitant medications, physical examination, full blood count, biochemistry (electrolytes, liver enzymes, urea, creatinine, creatine phosphokinase, amylase), plasma HIV-1 RNA and CD4-positive lymphocyte count, and serum ß-human chorionic gonadotropin (pregnancy test) in women. Quality of life was self-assessed with a visual analogue scale.
Fasting metabolic indices included glucose and insulin (including 2 h after 75 g oral glucose loading), C-peptide, estimated insulin resistance (by homoeostasis model assessment [HOMA]), total cholesterol, direct low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, triglycerides, and lactate. Participants and clinicians were masked to all lipid and glycaemic results after screening unless fasting triglycerides rose above 15.0 mmol/L (because of the small risk of pancreatitis) or fasting glucose rose above 7.0 mmol/L (diagnostic of diabetes) or fell below 3·0 mmol/L. Plasma adiponectin and leptin, markers of adipocyte function, were measured by radioimmune assays (Linco Research, St Charles, USA) at baseline, week 24, and week 48.
Body composition was quantified at screening, week 24, and week 48 by DEXA and CT. DEXA measured total and regional body fat and lean tissue. CT measured the mean intra-abdominal and subcutaneous fat areas at the intervertebral spaces immediately below the L2, L3, and L4 vertebrae and subcutaneous fat area at the right mid-thigh. We did not undertake CT of the face because there are no validated landmarks for this measurement. Quality assurance programmes were instituted for DEXA and CT, and all scans were analysed centrally. Patients' physical activity and diet (food frequency questionnaires) were recorded to determine whether either of these measures affected body composition.
Other objective body fat measures included the lipodystrophy case definition score, total and trunk fat with DEXA, weight, body-mass index, and waist and hip circumferences. Subjective lipodystrophy presence and severity overall were measured by clinicians and participants every 24 weeks (every participant was assessed by the same clinician at all timepoints) with a standardised scoring system; a score of 0 for nil, 1 for mild, 2 for moderate, or 3 for severe was assigned for every region.