Introduction
The phenomenon of progressive, selective wasting of subcutaneous fat from the face and limbs of HIV-infected patients being treated with antiretroviral therapy first emerged in 1997 and was more widely reported in diverse populations throughout 1998 and 1999 [1-3]. Fat accumulation (intra-abdominal, dorsocervical, subcutaneous lipomata), dyslipidaemia and insulin resistance were found to be variably associated with subcutaneous fat wasting, and the term lipodystrophy syndrome was proposed to describe these abnormalities collectively [2]. The lipodystrophy syndrome was ascribed to a unique cumulative toxicity of HIV protease inhibitor (PI) therapy [2,4] having been noted only 2 years after the introduction of PI into routine clinical practice. By comparison, it was 11 years since the first use of nucleoside analogue reverse transcriptase inhibitors (NRTIs). However, stavudine had not been used for as long as other NRTIs and was being used more frequently in the PI era. In addition, the earliest studies of metabolic abnormalities associated with antiretroviral therapy were designed a priori to compare use of a PI with no PI in a cross-sectional manner and, therefore, could not show the contribution or interactions of concurrent NRTI [2].
The development of subcutaneous fat wasting, intra-abdominal obesity and buffalo humps in PI-naive patients has now been documented within several cohorts [5,6]. As NRTIs have been used sequentially and fat wasting is a progressive, cumulative phenomenon, there are potential bias against agents used more recently, such as stavudine and PI. We have sought to determine the relative contribution of individual NRTIs and PIs to the development of subcutaneous fat wasting, with due consideration of multiple confounding factors.
Methods
The study assessed 277 participants of the Western Australian HIV Cohort Study (described in [7]) for clinical evidence of subcutaneous fat loss between December 1997 and May 1999. Patients were evaluated at least every 3 months. The case definition for clinical fat wasting required loss of fat in the face, arms or legs (in some cases giving a perception of increased muscle definition or prominent veins) detectable by both the patient and their clinician. The date of onset of clinically apparent fat wasting was recorded. In addition, the patient and their clinician completed a questionnaire for the Australian Lipodystrophy Prevalence Study between November 1998 and February 1999. The following clinical information was collected: diastolic and systolic blood pressure, smoking history, alcohol consumption, anabolic steroid use, exercise activity, personal or family history of ischaemic heart disease, personal or family history of diabetes mellitus, and use of lipid-lowering drugs. An assessment was made by both the patient and their physician of changes in appearance of the following areas since commencement of antiretroviral therapy (or during the preceeding 12 months if antiretroviral naive): face, arms, legs, abdomen, breast and buttocks. The development of abdominal distension, buffalo humps or lipomatosis was also recorded. The severity of these changes was graded as nil, mild, moderate or severe. The following data are routinely collected on all participants of the Western Australian HIV Cohort Study [7] : weight, body mass index, HIV- and non-HIV-related illnesses, antiretroviral drug use, prophylactic medications, CD4 cell count, HIV RNA concentration (HIV Amplicor Roche, Branchburg, USA; limit of detection 400 copies/ml), serum electrolytes, urea and creatinine, full blood picture, liver function tests, fasting triglyceride, cholesterol, high and low density lipoproteins, Apo B, lipoprotein (a), insulin, insulin resistance index [8], C-peptide and glucose (all recorded at least 3 monthly), creatinine kinase, lactate, uric acid, serum testosterone, Apo E genotype, age, race and height. The Ethics Committee of Royal Perth Hospital approved the protocol, and all patients provided written informed consent.
Patients who had acute symptomatic mitochondrial cytopathy characterized by recent onset of weight loss, acute symptomatic hepatic steatosis or lactic acidosis were excluded from the analysis. No patients in the cohort had features of HIV-wasting syndrome (unexplained involuntary weight loss of greater than 10% of baseline weight plus chronic diarrhoea, chronic weakness or fever).
Whole body dual energy X-ray absorptiometry (DEXA) scans (Hologic QDR-4500A Hologic, INc, 590 Lincoln St, Waltham, MA 02154, USA) were conducted by a single operator in 161 randomly selected patients. There was no significant difference in the demographic, clinical or laboratory characteristics of those who had DEXA scans and those who did not (data not shown). The percentage fat at the arms, legs and central abdomen (calculated from the mass of fat versus lean and bone mass) as well as total lean body mass in kilograms was recorded. A group of 70 patients had two or three consecutive DEXA scans, on average 6 months apart, during the study period (without any change to therapy between scans).
Statistical analyses
Statistical analyses were carried out using Cox proportional hazards regression and multiple linear regression as appropriate, using standard statistical software (SAS statistical package, SAS Institute, Cary, North Carolina, USA).
A comprehensive analysis of the times from receipt of triple therapy (dual NRTI with either PI or nevirapine) to the onset of fat wasting was carried out using Cox regression with time-dependent covariates to take account of changing and cumulative treatment effects. This framework allowed comparison of all patients, whether or not they received PI/nevirapine concurrently with dual NRTI therapy; it also allowed appropriate assessment of prior dual therapy on the rate of progression from receipt of PI or nevirapine. A dominant increase in the rate of progression to fat wasting resulting from the addition of PI to dual NRTI was illustrated via Cox regression by comparing the times to onset of fat wasting in those individuals receiving dual therapy only with those whose treatment regimen also included a PI.
Percentages of fat in either the leg or abdominal region on whole-body DEXA scans were analysed using multiple regression on the log scale for better approximation of the standard assumptions.
In all cases, potential predictive variables were assessed using standard Wald and likelihood ratio tests to determine those which were significant.
Results
Patient characteristics and antiretroviral treatment history
The characteristics of patients are shown in Table 1 and the distribution of antiretroviral use in the total cohort at the time of assessment in Fig. 1. Typically the patients were taking zidovudine or stavudine with a second NRTI (didanosine or lamivudine) and either a PI or nevirapine. The PI used was indinavir (61%), nelfinavir (17%), ritonavir/saquinavir (15%) and ritonavir (3%). The median duration of nelfinavir use (5 months) was shorter than the median durations of indinavir use (16 months), ritonavir use (11 months) and saquinavir use (13 months). Dual NRTI therapy (which included zidovudine in most cases) with or without nevirapine was taken by 56% of PI-treated patients prior to commencement of PI. There were no patients in the cohort who had ever taken stavudine, lamivudine, nevirapine or PI monotherapy, and 47 patients were not taking any antiretroviral therapy.
Subcutaneous fat wasting occurs in PI-naive, NRTI treated patients
The clinical case definition for subcutaneous fat wasting was met by 112 of 201 (54%) PI-treated patients compared with 9 of 71 (13%) antiretroviral-treated PI-naive patients. The incidence of clinical fat wasting in PI-treated patients was 52/100 person-years of PI treatment compared with 2.3/100 person-years of non-PI antiretroviral therapy. Two patients reported buffalo humps and ten reported subcutaneous lipomatosis. The characteristics of the nine PI-naive patients who experienced fat wasting are presented in Table 2.
Prolonged past dual NRTI therapy and prolonged use of stavudine increases risk of fat wasting in recipients of combination triple therapy
A comprehensive Cox regression analysis was carried out on time from commencement of triple antiretroviral therapy (either dual NRTI with nevirapine or dual NRTI and PI) to onset of fat wasting, incorporating both fixed and time-dependent covariates. To enable a direct comparison of the relative effects of stavudine versus zidovudine, patients were excluded from the analysis during any time at which their therapy did not contain one of these two drugs. Variables initially considered in the model included age, race, sex, lowest recorded CD4 cell count, highest recorded plasma HIV RNA concentration, duration of HIV infection, duration of AIDS, opportunistic infections, prior duration of dual NRTI therapy and the cumulative times on individual drugs since commencement of triple therapy. The factors found to increase the rate of progression to fat wasting significantly were age (relative risk = 1.052 per year;P  < 0.0001), white race (relative risk = 3.9;P  = 0.023), longer duration of dual NRTI therapy prior to commencement of triple therapy (relative risk = 1.021 per month;P  = 0.0046) and increased cumulative time on a regimen containing stavudine compared with time on a regimen containing zidovudine (relative risk = 1.085 per month;P  < 0.0001). Stavudine increased the risk of fat wasting by 265% per year compared with zidovudine. Increased cumulative time on nevirapine compared with time on a PI as a component of triple therapy was associated with a slower rate of progression (relative risk = 0.943 per month;P  = 0.022) (Table 3).
Stavudine is associated with increased risk of fat wasting compared with zidovudine in PI recipients
To illustrate simply the comparison between stavudine and zidovudine, PI recipients were stratified according to whether the first NRTI in use at the time of commencement of PI was zidovudine or stavudine. No adjustment was made for subsequent changes in therapy, giving a type of intention-to-treat analysis. Adjustments were again made for age and race as well as prior time on dual NRTI. The rate of progression from commencement of PI to onset of fat wasting was significantly higher for those on concurrent stavudine than for those on zidovudine (relative risk = 1.9;P  = 0.003) (Fig. 2). The median time from commencement of a PI to clinically apparent fat wasting was 18.5 months in patients on a stavudine-containing regimen compared with 26 months in those on a zidovudine-containing regimen.
Lamivudine or didanosine therapy
A similar analysis comparing lamivudine and didanosine was carried out but did not show any significant differences in rate of progression to clinically apparent fat wasting (data not shown).
Addition of PI to dual NRTI therapy is associated with shorter time to onset of clinically apparent fat wasting
To investigate the effect of dual NRTI therapy alone relative to dual NRTI therapy with PI, the times from commencement of dual NRTI to the onset of fat wasting was compared in those individuals receiving only dual NRTI therapy and those who started NRTI and PI concurrently. Individuals whose initial antiretroviral therapy contained nevirapine rather than a PI were excluded from the analysis and times were censored if treatment changed from dual therapy, with the inclusion of a PI. Older age and white race were found to be significant predictors of wasting (P  = 0.002 and P  = 0.02, respectively). After adjusting for these factors, the times from dual therapy to fat wasting were significantly shorter in those individuals whose treatment also included a PI (relative risk = 13.1, 95% confidence interval 5.9-29.2;P  < 0.00005). Hence the rate of progression to fat wasting is dominated by the addition of a PI into the treatment regimen.
Assessment of regional fat by DEXA scans
The relationship between antiretroviral treatment and regional fat percentage as assessed at the first DEXA scan was examined in a multiple linear regression model. In view of the strong effect of gender on body composition, the analysis was restricted to men (Table 4). Longer cumulative times on NRTI monotherapy, NRTI dual therapy and triple combination therapy containing stavudine were associated with lower percentage fat in the leg (0.29, 1.1 and 2.0% relative fat loss per month, respectively;P  = 0.037, 0.0003 and < 0.0001, respectively).
Analysis of fat on the arms revealed the same associations (data not shown); similar findings were observed for central abdominal fat, though the association with NRTI monotherapy and dual NRTI therapy did not reach statistical significance (Table 4).
Sequential DEXA scans show non-linear decline in regional fat
Change in regional fat percentage was analysed over time in 70 patients. In order to avoid possible confounding effects of changes in treatment, patients were only considered if their therapy remained unchanged between scans. Analyses of arm and leg fat were done separately and gave similar results. The percentage changes per month in leg fat from initial scan to final scan were calculated and analysed by multiple regression for associations with age at initial scan, time between scans, the percentage fat on the initial DEXA scan and the time the patient had been taking PI treatment prior to the initial scan. No significant association was found with age, time between scans or prior time on PI. However there was an approximately linear relation between the percentage change and the logarithm of the initial DEXA, estimated as percentage change = 4.0 - 1.711 log(percentage fat on initial DEXA) (P  = 0.008). Therefore, the average rate of change is higher when the percentage fat measured by the initial DEXA is high, decreasing to approximately zero for DEXA values near 10% fat, indicating a non-linear decline in regional fat over time.
Discussion
Patients who have been treated with dual NRTI therapy without a PI may develop wasting that is clinically indistinguishable from that seen in patients on combined NRTI and PI. We have shown that the risk of developing fat wasting is greater in patients who have had prolonged exposure to any dual NRTI therapy and that the risks associated with individual NRTI differ. We did not find that any virological or immunological marker related to HIV infection itself predicted risk of fat wasting. Finally, the time to wasting associated with use of combination NRTI and PI is relatively short compared with that associated with dual NRTI alone. Taken together, these findings suggest that NRTI as a class independently contribute to risk of fat loss, but the combination of NRTI and PI is synergistic in leading to clinically apparent fat wasting.
As stavudine was the last NRTI to be licensed prior to use of PI, current use of stavudine is a potential surrogate for prior zidovudine treatment and longer treatment in general. If fat wasting is a duration-dependent class toxicity of NRTI, there is a potential confounding bias against stavudine compared with zidovudine. However, we have found that the difference between stavudine and zidovudine in predicting the risk of fat wasting is significant even after appropriate adjustment for the potentially confounding factors such as age, duration of past NRTI therapy, duration of HIV infection, duration of AIDS, opportunistic infections, CD4 cell count and HIV viral load. Moreover, we have relied on time-to-event analysis in which the duration of current and past therapies are taken into account. As the use of stavudine and zidovudine is mutually exclusive within the cohort, differences between these agents are readily observed. While recent studies have suggested that lamivudine causes fat wasting [3], this was not observed in our study. However, as lamivudine and didanosine are almost always used as the `second NRTI' (with either zidovudine or stavudine) in our cohort (Fig. 1), a real but equal effect on fat wasting cannot be excluded. Indeed the contribution of lamivudine and didanosine to fat wasting is implied by the finding that dual NRTI therapy predicts greater fat loss than does NRTI monotherapy (Table 4).
A number of factors may explain why subcutaneous fat loss was not associated with use of NRTI until 1997, despite up to 11 years of widespread experience with these agents. In our analysis, the risk of fat wasting is primarily dependent on cumulative exposure to NRTI therapy. A duration-dependent drug toxicity causing slow, insidious fat loss would have been difficult to detect prior to the advent of PI, when other causes of wasting such as opportunistic infections and uncontrolled HIV infection were more prevalent. The use of PI therapy was accompanied by the increased use of stavudine, which has the strongest effect on fat wasting of all the NRTI. Our study shows that combined PI and dual NRTI leads to a dramatically faster fat loss compared with that seen with dual NRTI therapy without PI. Hence, while PI therapy does clearly exert a strong independent effect on risk of fat loss, their use has also revealed a previously unrecognized predisposition to fat wasting caused by long-term NRTI therapy. Our analyses show that the risk is duration dependent, but we cannot distinguish between insidious `slow burn' in fat, intermittent stepwise fat loss or precipitous losses after some threshold of exposure to NRTI is reached.
It is unlikely that the findings are confounded by acute symptomatic lactic acidosis or hepatic steatosis, which are already well-recognized toxicities of all NRTI (except lamivudine) and in which poor caloric intake, vomiting and rapid weight loss may contribute to loss of both fat and lean body mass [9-12]. All cases of acute symptomatic mitochondrial toxicity (which generally resulted in permanent cessation of NRTI) were excluded; however, such exclusion may have led to exclusion of cases of NRTI-induced subcutaneous fat wasting and introduced a conservative bias in the analyses.
Almost all known serious adverse effects of NRTI therapy are organ-specific manifestations of mitochondrial toxicity, via effects on DNA polymerase-γ[13-18]. There is evidence that NRTIs potentially select for mitochondrial DNA deletions in vitro &[17] and abnormalities of fat distribution such as multiple symmetric lipomatoses and buffalo humps are described in rare syndromes associated with inborn heteroplasmic mitochondrial DNA deletions [18]. Mitochondrial depletion syndromes associated with other inborn genetic defects of nuclear DNA and mitochondrial DNA deletions are characterized by failure to thrive [18] and thin body habitus [19]. As mitochondrial DNA defects would be expected to increase with age [14,15], our finding of age as a predictor of wasting is consistent with this hypothesis. The strong and consistent effect of race on risk of fat wasting is unexplained but may, in part, account for the geographic variation in reported incidence of fat wasting. The highest rates have been reported in Australia, where white Caucasians predominate.
In this study, stavudine therapy was associated with more fat wasting and higher serum lactate concentration than zidovudine therapy (not shown), suggesting a link between mitochondrial metabolic failure and fat wasting. It is possible that stavudine exerts a direct effect on β-oxidation as, unlike other NRTI, it is phosphorylated to its monophosphate, diphosphate and triphosphate derivatives within isolated mitochondria [20].
While it has been hypothesized that PI may induce peripheral fat cell apoptosis by binding to a region within the cytoplasmic retinoic acid-binding protein type 1 (CRABP-1) (which has a degree of amino acid homology with HIV-1 protease [4]), crystallographic models of CRABP-1 do not support binding to PI [21]. However, reduced adipogenesis and increased lipolysis have been induced by PI in vitro in a mesenchymal stem cell line [22]. The thiozolidinedione drug troglitizone has been shown to stimulate transcriptional activity of peroxisome proliferator-activated receptor gamma (PPAR-γ) within these cells, which mediates differentiation into mature adipocytes. Activation of PPAR-γ also leads to increase in mitochondrial mass [23]. Troglitizone is associated with increases in subcutaneous fat when used clinically. Therefore, inhibition of PPAR-γ by PI may explain both reduced adipogenesis in the subcutaneous compartment and depletion of mitochondria and exacerbation of NRTI-induced mitochondrial toxicity. However, other possible interactions involving cytochrome P450 enzymes or the overloading of mitochondrial function by excess free fatty acids or other substrates cannot be excluded, and these require further investigation in vitro and in vivo.
Rather than a single pathogenetic mechanism involving PI and NRTI effects on mitochondria, subcutaneous fat wasting may be the cumulative result of two separate and unrelated class effects. An alternative mechanism not related directly to drug toxicity, such as an immune restoration phenomenon, is less likely given the lack of relationship between CD4 cell count, HIV viral load or any known opportunistic infection and risk of fat wasting [24]. However, the possibility that any drug that causes highly potent viral suppression may potentiate NRTI-associated subcutaneous fat wasting cannot be excluded.
Further insights into the pathogenesis of subcutaneous fat wasting may be gained by prospective, randomized fat-sparing or reversibility studies. Our findings indicate that consideration should be given to switching stavudine and other NRTI as well as PI in future reversibilty studies.
This analysis has focused on subcutaneous fat wasting because this was the dominant clinical problem for patients in our cohort. Fewer patients reported abdominal obesity, and increased waist circumference associated with insulin resistance was not common (data not shown). Our findings suggest that visceral obesity, lipomatoses, dyslipidaemias, insulin resistance and overt diabetes mellitus all need to be analysed as separate clinical endpoints, taking into account all prior antiretroviral treatment exposures and not just PI therapy. Until the varied and complex associations of these abnormalities with different antiretroviral agents are better understood, it may be premature to study the lipodystrophy syndrome as a single clinical outcome.
As the efficacy, convenience and short-term tolerability of individual antiretroviral agents improve, the long-term cumulative toxicities will increasingly determine their utility in the treatment of HIV-infected patients. In addition to subcutaneous fat wasting being a distressing cosmetic concern for many patients, the associated metabolic derangement and potential risk of vascular diseases in such subjects has not been fully defined. Improved understanding of the pathogenesis of subcutaneous fat wasting may help to predict other unexpected potential long-term adverse effects of antiretroviral therapy.
Acknowledgements
We are grateful to Mr Geoff Roff for performing all the DEXA scans for the study, Dr Brendan Adler for reporting abdominal computed tomographic scans and Drs Phillipa Lamont, James McCarthy and David Nolan for helpful comments. This study has been facilitated in part by the ongoing Australian Lipodystrophy Prevalence Study coordinated by the Australian National Centre in HIV Epidemiology and Clinical Research. We are indebted to Associate Professor Martyn French and the staff of the Department of Clinical Immunology, Royal Perth Hospital and to the patients who participated in this study.
References
1. Viraben R, Aquilina C. Indinavir associated lipodystrophy. AIDS 1998, 12: F37 -F39.
2. Carr A, Samaras K, Burton S. et al. A syndrome of peripheral lipodystrophy, hyperlipidaemia and insulin resistance in patients receiving HIV protease inhibitors. AIDS 1998, 12: F51 -F58.
3. Gervasoni C, Ridolfo AL, Trifiro G. et al. Redistribution of body fat in HIV-infected women undergoing combined antiretroviral therapy. AIDS 1999, 13: 465 -472.
4. Carr A, Samaras K, Chisholm DJ. et al. Pathogenesis of HIV-1 protease inhibitor-associated peripheral lipodystrophy, hyperlipidemia, and insulin resistance. Lancet 1998, 351: 1881 -1883.
5. Saint-Marc T, Poizot-Martin L, Partisani . et al. A syndrome of peripheral fat wasting (lipodystrophy) in patients receiving long-term nucleoside analogue therapy. AIDS 1999, 13: 1659 -1667.
6. Madge S, Kinloch-de-Loes, Mercey D. et al. Lipodystrophy in patients naive to HIV protease inhibitors. AIDS 1999, 13: 735 -737.
7. Mallal S. The Western Australian HIV Cohort Study, Perth, Australia. J Acquir Immune Defic Syndr Hum Retrovirol 1998, 17 (Suppl. 1): S23 -S27.
8. Matthews DR, Hosker JP, Rudenski AS. et al. Homeostasis model assessment: insulin resistance and β-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 1985, 28: 412 -419.
9. Stein DS. A new syndrome of hepatomegaly with severe steatosis in HIV seropositive patients. AIDS Clin Care 1994, 6: 19 -20.
10. Fortgang IS, Belitsos PC, Chaisson RE. et al. Hepatomegaly and steatosis in HIV infected patients receiving nucleoside analogue antiretroviral therapy. Am J Gastrol 1995, 90: 1433 -1436.
11. Olano JP, Boruki MJ, Wen JW. et al. Massive hepatic steatosis and lactic acidosis in a patient with AIDS who was receiving zidovudine. Clin Infect Dis 1995, 21: 973 -976.
12. Lenzo NP, Garas BA, French MA. Hepatic steatosis and lactic acidosis associated with stavudine treatment in a HIV patient: a case report. AIDS 1997, 11: 1294 -1296.
13. Brinkman K, ter Hofstede HJM, Burger DM. et al. Adverse effects of reverse transcriptase inhibitors: mitochondrial toxicity as a common pathway. AIDS 1998, 12: 1735 -1744.
14. John M, Mallal S, French M. Emerging toxicity with long-term antiretroviral therapy. J HIV Ther 1998, 3: 58 -62.
15. Lewis W, Dalakas MC. Mitochondrial toxicity of antiviral drugs. Nat Med 1995, 1: 417 -421.
16. Fry M, Loeb LA. Animal Cell Polymerases 221. Boca Raton, FL: CRC Press; 1986.
17. Wang H, Lemire BD, Cass CE. et al. Zidovudine and dideoxynucleosides deplete wild-type mitochondrial DNA levels and increase deleted mitochondrial DNA levels in cultured Kearns-Sayre syndrome fibroblasts. Biochim Biophys Acta 1996, 1316: 51 -59.
18. Schapira AH. Inborn and induced defects of mitochondria. Arch Neurol 1998, 55: 1293 -1296.
19. Hirano N, Silvestri D, Blake DM. Mitochondrial neurogastrointestinal encephalomyopathy (MNGIE); clinical, biochemical and genetic features of an autosomal recessive mitochondrial disorder. Neurology 1994, 44: 721 -727.
20. Cui L, Locatelli L, Xie MY. et al. Effect of nucleoside analogues on neurite regeneration and mitochondrial DNA synthesis in PC-12 cells. J Pharmacol Exp Ther 1997, 280: 1228 -1234.
21. Stevens GJ, Chen M, Greko R et al. Investigations into proposed mechanisms of HIV-associated peripheral lipodystrophy, hyperlipidemia and insulin resistance. First International Workshop on Adverse Drug Reactions and Lipodystrophy in HIV. San Diego, June 1999 [abstract 058].
22. Lenhard JM, Weiel JE, Paulik MA et al. HIV protease inhibitors block adipogenesis and increase lipolysis in vitro. Sixth Conference on Retroviruses and Opportunistic Infections. Chicago, January, 1999 [abstract 665].
23. Lenhard JM, Kliewer SA, Paulik MA. et al. Effects of troglitizone and metformin on glucose and lipid metabolism: alterations of two distinct molecular pathways. Biochem Pharmacol 1997, 54: 801 -808.
24. Mallal SA, James IR, French MAH. Detection of subclinicalMycobacterium avium intracellularecomplex infection in immunodeficient HIV-infected patients treated with zidovudine. AIDS 1994, 8: 1263 -1269.
© 2000 Lippincott Williams & Wilkins, Inc.