Protease inhibitors (PI) in combination with nucleoside reverse transcriptase inhibitors (NRTI) are now considered to be the standard of care for optimal antiretroviral therapy[1,2]. These new therapies are usually well tolerated; however, with their widespread use previously unrecognized side-effects are becoming more evident as availability of the drugs improves and the duration of treatment increases.
Recently, a syndrome of peripheral fat wasting (lipodystrophy), central adiposity, hyperlipidaemia and insulin resistance has been identified in patients treated with PI-containing regimens[3,4].
The term lipodystrophy refers to the loss of subcutaneous fat tissue from the facial fat pads (Bichat or preauricular fat pads), arms and legs resulting in a cachectic appearance, prominent zygomata and sunken eyeballs.
Two different theories have been proposed, both relate this syndrome with the use of PI[5,6]. However, precise aetiologic mechanisms resulting in clinical fat distribution changes seen in the syndrome still remain unclear.
Since the early days of the AIDS epidemic, there has been clear clinical evidence that wasting is a fundamental feature of HIV disease progression. Unique to patients with HIV infection is that wasting has been shown to be an unintentional loss of predominantly lean body mass (protein losses), whereas body fat was relatively spared[7-9]. Short-term weight loss was usually accompanied by decreased caloric intake and was often due to secondary infection. Peripheral fat wasting as a consequence of any drug therapy in patients with normal daily caloric and nutrient intake has never been reported.
In one recently published study, Kotler et al.  reported anectodal cases of peripheral lipodystrophy in patients who had never taken a PI, suggesting that such fat wasting in HIV-infection may not be a direct side-effect of PI. Fat wasting and its potential relationship to NRTI has not been thoroughly investigated in clinically stable NRTI-experienced patients.
The purpose of this study was to compare body composition, body fat distribution and insulin secretion in a group of patients taking long-term nucleoside analogue therapy including stavudine or zidovudine; a group of 15 therapy-naive HIV-infected patients were evaluated as controls.
From March 1998 to November 1998, 43 consecutive HIV-infected patients were studied. This group was composed of seven females and 36 males on long-term NRTI therapy, aged 27-66 years (mean, 41 years). The control group consisted of 15 therapy-naive HIV-infected patients (four females and 11 males; mean age, 39.7 years; range, 30-57 years). These 58 patients were recruited among the 156 patients participating in the Inter-LIPOCO study-a French prospective observational cohort study of risk factors for the development of changes in body fat distribution in unselected patients receiving antiretroviral therapy [NRTI, non-nucleoside analogue reverse transcriptase inhibitor (NNRTI) or PI-based combination drug regimens], currently being performed in five clinical centres in France.
As of 30 November 1998, 56 patients (13 therapy-naive and 43 on NRTI therapy) were recruited from Lyon and two therapy-naive patients were recruited from Strasbourg and Marseille. The study was designed as a prospective study with follow-up and assessment performed every 2 months. The research protocol was approved by the Institutional Ethic Committee of the University of Lyon and informed consent was obtained from all of the patients.
To be eligible for inclusion in the study patients had to be aged 18 years or over, have biologically documented HIV-1 infection, have been receiving two NRTI agents during at least the previous 6 months and be under active follow-up. No other selection criteria for inclusion in the study were used.
Patients were excluded if they had active opportunistic infection or malignancy within 3 months prior to entry and/or fever (temperature >37.8°C). No patient used any other drug known to influence glucose metabolism or fat distribution [anabolic hormones or systemic glucocorticoids, recombinant human growth hormone, appetite stimulants]. In addition, no patient had been treated with drugs known to regulate triglyceride and cholesterol levels.
Patients were evaluated as ambulatory outpatients. All investigations were performed after a 12 h overnight fast and at least 15 min after the placement of a peripheral intravenous catheter.
Fat wasting was identified by patient self-report of relevant changes in their face (loss of Bichat and/or preauricular fat pads), arms, legs and subcutaneous abdominal tissue, and confirmed clinically by qualitative physical examination.
Anthropometric measurements and body composition
The patients wore light indoor clothes and no shoes when the anthropometric measurements were performed. Measurements included body weight (kg) and height (cm). Data on weight collected before the start of the last NRTI therapy were obtained retrospectively from medical files. Body mass index (BMI, weight in kg divided by the square of the height in m) was calculated. Mid-arm circumference (cm) was measured to the nearest millimetre using a cloth tape, mid-way between the acromial process of the shoulder and the tip of the elbow, with the right arm hanging relaxed at the side.
Body composition was measured by tetrapolar bioelectrical impedance analysis (BIA; RJL-Akern BIA 109; Data-Input, Frankfurt, Germany) in the supine position as described previously. This method allows calculation of the following indexes of body composition: body fat, lean body mass and extracellular water. Fat in absolute mass and as a percentage of body weight was calculated.
Body fat composition was also estimated using caliper measurements (Lange skinfold caliper; Cambridge Scientific Industries, Cambridge, Massachusetts, USA) of skinfold thickness at four sites (scapular, biceps, triceps and suprailiac) taken on the right side. All measurements were made in triplicate and the means calculated.
Percentage of body fat was calculated by the method of Durnin and Womersley. Most measurements of body composition were made by the same dietician.
Subcutaneous and visceral fat distribution
Regional fat distribution was measured by computed tomography (CT) performed at two body levels to measure abdomen and mid-thigh subcutaneous fat areas as well as intra-abdominal fat area. Patients were imaged on a SR7000 CT-scanner (Philips Medical Systems, Best, The Netherlands). An unenhanced helical abdominal acquisition was obtained with the following parameters: Kvp 120, mAs 250, Collimation 10mm, table speed 10mm/s, gantry rotation period 1 s (pitch 1:1), 10-20 rotations. Imaging data were reconstructed every 10mm.
The slice passing through the umbilicus (fourth lumbar vertebra), reported as a valid predictor of total abdominal fat in men and women, and that passing through the middle of the thigh were transferred in an Easyvision (Philips Medical Systems) workstation. The areas of intra-abdominal adipose tissue, i.e. visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) expressed in cm2, were calculated by summing the area of pixels in the slice with CT values from -150 to -50 Hounsfiel units, according to the method described by Miller et al. .
The ratio between the subcutaneous and the visceral fat areas (SAT:VAT) was calculated for each CT scan as an indicator of the predominance of the subcutaneous or visceral accumulation of fat. The ratio between the visceral and total adipose tissue (TAT) was also calculated. The technicians who performed the CT fat measurements were blind to the patients treatment details.
Plasma lipid profile, endocrine and biochemical measurements
Serum total cholesterol and triglycerides, measured by enzymatic methods with the same automated analyser, serum high-density lipoprotein cholesterol, isolated by chemical precipitation, serum low-density lipoprotein cholesterol, apolipoprotein A1 and B, uric acid, and free fatty acids were measured on fasting venous blood.
Basal morning evaluation of gonadal and adrenal function included the measurement of plasma testosterone, follicle stimulating hormone, luteinizing hormone and plasma cortisol concentrations.
CD4 cell count (×106 cells/l) and quantitative HIV-1 RNA (viral load, log10 copies/ml) were also measured on venous blood.
Oral glucose tolerance test (OGTT)
An OGTT was used to assess glucose tolerance and to measure the insulin response to oral glucose. After a 12 h overnight fast, the patients were given a 75g oral glucose load in 250ml water. The glucose solution was consumed within a 3 min period. Venous blood samples were taken 15 min before the glucose was administered, at the time of administration, and 30, 60, 90 and 120 min thereafter for determination of plasma glucose and insulin concentrations. The glucose tolerance status of each patient was categorized according to the criteria of the World Health Organisation.
Analysis of the dietary history over the 3 days preceding study entry was performed. All patients completed prospectively a self-administered questionnaire. For each meal, patients were asked to detail fully everything they had consumed, including information on the time of meal intake. For each food item recorded, patients were asked to describe how it was prepared (i.e., raw, fried, boiled), list all the major ingredients, and record amounts in ounces, tablespoons or units. The macro- and micronutritional contents of each diet were calculated by using FoodPerfect, an interactive nutritional analysis computer program based on the Canadian Nutrient File database. Total caloric, protein, fat, carbohydrate and cholesterol intakes were evaluated.
Statistics were performed using the statistical analysis system (SAS Institute Inc, Cary, North Carolina, USA). Comparison of qualitative variables was determined using the χ2Pearson test. The significance of the differences in continuous variables was evaluated using the Mann-Whitney-Wilcoxon non-parametric test. Logistic regression was used to establish multivariate models to assess odds ratio (OR) of developing lipodystrophy according to antiretroviral drugs. Adjustment was made for sex, age, HIV-1 RNA level, global time of exposure to antiretroviral therapy and time on current antiretroviral therapy.
Time to development of lipodystrophy was estimated by the Kaplan-Meier method.
The main demographic characteristics and laboratory data are shown in Table 1. The treated patients were divided into two groups according to NRTI. Three groups were studied. Twenty-seven patients were currently receiving a stavudine-containing combination therapy (four women and 23 men; mean age, 41.8; range, 27-66 years), for a mean time of exposure of 1.35±0.7 years [14 (32.6%) were taking stavudine plus lamivudine and 13 (30.2%) were taking stavudine plus didanosine]; 16 patients were currently receiving a zidovudine-containing combination therapy (three women and 13 men; mean age, 39.9 years; range, 30-62 years) for a mean of 2.1±0.85 years [13 (30.3%) were taking zidovudine plus didanosine, two (4.7%) were taking zidovudine plus lamivudine and one (2.3%) was taking zidovudine plus zalcitabine]; and 15 were therapy-naive HIV-infected patients.
The three groups were well balanced with respect to age, sex, weight, BMI, CD4 cell counts and daily caloric intake. Duration of HIV infection was significantly higher in the treated groups (stavudine and zidovudine groups) as compared with the therapy-naive patients (P<0.05), whereas HIV viral load was significantly lower (P<0.01).
Baseline CD4 cell counts, HIV-1 RNA, duration of HIV infection and mean time of exposure to NRTI did not differ between the stavudine and the zidovudine groups (2.70±1.6 years versus 2.43±1.5 years, non-significant).
Seventeen (63%) patients in the stavudine group had been pretreated prior to receiving stavudine (five with zidovudine monotherapy and 12 with a double- nucleoside therapy including zidovudine) and two (12.5%) patients in the zidovudine group had been pretreated with zidovudine monotherapy.
No significant differences were observed in total cholesterol, low-density lipoprotein cholesterol, apolipoprotein A1, apolipoprotein B and uric acid levels between each treated group and controls.
High-density lipoprotein cholesterol was significantly lower in the stavudine group as compared with the zidovudine group but it did not differ between the zidovudine and the control groups. Triglyceride levels were significantly higher in the stavudine group than in the controls (P<0.05), but it did not differ between the stavudine and the zidovudine group or between the zidovudine and the control group. There was a trend for higher mean serum free fatty acid levels in the stavudine group as compared with the zidovudine and control groups, although the difference was not statistically significant.
Serum cortisol, testosterone, follicle stimulating hormone and luteinizing hormone levels did not differ significantly between the three groups.
Post-absorptive plasma glucose concentrations measured during the OGTT were not significantly different, and no differences were found in both fasting and glucose-stimulated plasma insulin and C-peptide levels between the three groups (data not shown).
Body composition and anthropometric measurements
The results of measurements of body composition and anthropometry are shown in Table 2. The lean body mass and fat mass parameters of body composition were similar among the groups when expressed in absolute terms. However, when expressed as percentage of body weight, a moderate but significant decrease in FM was observed in the stavudine group as compared with the zidovudine and control groups (P<0.05), whereas this parameter did not differ between the zidovudine and the control groups. Biceps and scapular skinfold thickness was significantly lower in the stavudine group than in the zidovudine group (P<0.05) and, although the difference was not statistically significant, triceps and suprailiac skinfold showed a tendency to be lower in the stavudine group.
Regional fat distribution
Subcutaneous fat as measured by CT scan at abdominal and mid-thigh levels was significantly lower in the stavudine group as compared with the zidovudine group and the controls (P<0.05 and P<0.01, respectively) (Table 3). Although the zidovudine group had smaller values of abdominal and midthigh subcutaneous fat area, as compared with the controls, the difference did not reach statistical significance.
Visceral adipose tissue was significantly higher in the stavudine group (P<0.05 for the comparison with the zidovudine group and the controls). Abdominal SAT:VAT ratio was significantly lower in the stavudine group as compared with the zidovudine and control groups (P<0.01), while abdominal VAT:TAT ratio was significantly higher (P<0.05).
Caloric intake was similar in the stavudine, zidovudine and control groups (Table 4). However, patients receiving stavudine had higher mean intake of fat and cholesterol as compared with the zidovudine and control groups (P<0.01).
Peripheral fat wasting
Peripheral lipodystrophy was clinically observed in 17 (63%) patients in the stavudine group, in three (18.75%) patients in the zidovudine group and none in the control group. Lipodystrophy of the ‚Bichat‚ or preauricular fat pads was seen in the stavudine group only. In this group, lipodystrophy occurred in all body regions including subcutaneous abdominal fat areas whereas, in the zidovudine group, two patients experienced loss of subcutaneous fat in the legs and one in the face without fat wasting in the ‚Bichat‚ fat pads.
Kaplan-Meier analysis estimated the median time to lipodystrophy to be 14 months (range, 3.1-22.0 months) (Fig. 1). The crude OR calculated for peripheral lipodystrophy development in the stavudine group as compared with the zidovudine group was 7.37 (95% CI, 1.68-32.31; P=0.005) whereas the OR calculation showed a decreased risk in the zidovudine group as compared with the stavudine group (OR, 0.14; 95% CI, 0.03-0.59; P=0.005). In the stavudine group, 52% (14 out of 27) of patients concomitantly used lamivudine and 48% (13 out of 27) used didanosine, whereas in the zidovudine group 13% (two out of 16) and 81% (13 out of 16) patients concomitantly used lamivudine and didanosine, respectively. However, fat wasting did not significantly correlate with lamivudine (OR, 1.87; 95% CI, 0.53-6.53; P=0.324) or didanosine use (OR, 0.44; 95% CI, 0.13-1.52; P=0.191). The risk of lipodystrophy for patients who received stavudine was independant of the use of lamivudine (OR, 0.59; 95% CI, 0.12-2.89; P=0.516), or didanosine (OR, 1.69; 95% CI, 0.35-8.22; P=0.516). Differences in development of fat wasting according to antiretroviral drugs remained highly significant after adjustment for confounding variables (Table 5).
Body weight data at the time of evaluation and 3 months earlier were available for all treated patients. During this interval, the stavudine group had a mean weight loss of 2.1kg whereas the zidovudine group had a mean weight increase of 1.3kg (P<0.05). Among patients in the stavudine group, 13 (48%) were losing weight at the time of study as compared with two in the zidovudine group. The percentage of patients with a mean weight loss of 3kg or more was 29.6% and 5.8% in the stavudine and the zidovudine groups, respectively. The double nucleoside therapy including stavudine was discontinued in 12 of the 17 patients who reported lipodystrophy and they were switched to a triple regimen consisting of zidovudine plus lamivudine and nevirapine (six patients), zidovudine plus lamivudine and efavirenz (three patients), zidovudine plus didanosine and efavirenz (three patients). A major improvement was noted in two patients as demonstrated by an increase in the percentage of total body fat measured by BIA as well as an increase in the subcutaneous abdominal adipose tissue area; a mild improvement was noted in three patients and no change in six patients, all complaining of severe lipodystrophy. The improvement became noticeable about 2 months after switch but it was clearly marked after 4 months.
Triglyceride levels which were at least twice the normal values in five of the 12 patients decreased by 50% during the first month after switch, and decreased further in four patients or normalized in eight patients after 3 months. Simultaneously, free fatty acid levels, which were up to 1.5 times the normal values in seven out of the 12 patients, returned to normal in nine patients after therapy was interrupted. All of the patients had free fatty acids within the normal range values before the occurrence of lipodystrophy and all but three had normal triglyceride levels. The plasma levels of both triglycerides and free fatty acids remained elevated or unchanged in the patients who reported lipodystrophy but who did not change therapy.
Data collected from the 43 patients of our study further support the occurrence of abnormal fat disposition in HIV-infected patients treated with non-PI-containing regimens. Peripheral fat wasting occurred in clinically stable asymptomatic HIV-infected patients and may be related to long-term nucleoside analogue therapy, particularly that including stavudine.
Although no difference in mean BMI was found between the three groups, patients taking a stavudine-containing combination therapy tended to have a marked decrease in body fat content.
In the current investigation, the BIA data, anthropometric measurements and abdominal and mid-thigh CT scans all support the presence of preferential loss of fat in patients receiving stavudine, whereas no statistically significant differences were noted between each treated group and controls in terms of lean body mass and mid-arm circumference.
No differences were observed between patients in the zidovudine group and the non-treated patients in terms of body composition and fat distribution, suggesting that the observed alterations of the nutritional state are not a direct consequence of HIV infection but are most probably related to antiviral therapy.
The syndrome of lipodystrophy differs notably from the well-known malnutrition or wasting syndrome reported in HIV infection: patients experienced a predominant and rapid loss of subcutaneous fat tissue, whereas lean body mass remained unchanged. Fat wasting in the ‚Bichat‚ or preauricular fat pads might be the most specific noticeable disfigurement; until the beginning of 1996, such disfigurement had never been reported, even in those patients experiencing markedly diminished fat mass secondary to active opportunistic infections and/or reduced caloric intake.
Caloric intake was similar in the stavudine, zidovudine and control groups, suggesting that short-term fat wasting observed in the stavudine-experienced patients was not due to decreased caloric intake. It is of note that the stavudine group had higher mean daily intake for fat and cholesterol: differences in nutrient intake may be a consequence of underlying metabolic changes resulting in an unconscious compensatory mechanism for fat wasting.
The metabolic changes commonly seen in patients with lipodystrophy and PI therapy were not apparent. First, in contrast with the study reported by Carr et al. , the great majority of our patients had low fasting plasma glucose, insulin and C-peptide concentrations as well as normal amounts of insulin in response to oral glucose load. Furthermore, in the post-absorptive state, plasma insulin concentrations were invariably in the low-normal range, indicating an increase in insulin sensitivity.
In our whole population, three patients showed insulin resistance without diabetes (two in the stavudine group and one in the zidovudine group) and one patient had a history of non-insulin-dependent diabetes prior to becoming infected by HIV.
In the stavudine group, patients with lipodystrophy had glucose-stimulated plasma insulin levels similar to, or slightly greater than those of therapy-naive patients or non-lipodystrophic patients. Thus, insulin resistance does not appear as a consistent finding in patients experiencing lipodystrophy and it is therefore highly unlikely that the syndrome of lipodystrophy is due to the inhibition of the antilipolytic effect of insulin (a result of insulin resistance caused by the inhibition of at least one insulin-degrading enzyme by PI ). Secondly, although triglyceride levels tended to be higher in those receiving stavudine therapy, there was no statistical difference in this respect between the stavudine and the zidovudine groups Our data, therefore, do not support a role for triglycerides in the pathogenesis of lipodystrophy in these patients. However, interestingly an increase in both triglyceride and free fatty acid levels appeared to precede short-term fat wasting. This finding would suggest that triglyceride and free fatty acid concentrations might be useful in predicting future lipodystrophy, as suggested recently by Carr et al..
Since completion of our study in November 1998, three other non-PI-experienced patients have presented to us with peripheral lipodystrophy. Two patients were taking a double nucleoside analogue therapy (patient 1, stavudine plus lamivudine; patient 2, stavudine plus didanosine) and another patient was taking stavudine plus lamivudine and efavirenz. The three patients experienced short-term weight loss of 8, 6 and 3kg without central obesity as confirmed by CT scan. Interestingly, lactate and pyruvate levels in these three patients were found to be elevated (above the upper limit of normal), suggesting mitochondrial dysfunction.
The pathogenesis of peripheral lipodystrophy remains unexplained. Potential mechanisms may include decreased fat cell number (hypoplasia), perhaps via prolonged adipocyte apoptosis or decreased fat cell size (hypoatrophy) via stimulation of lipolysis. In few patients, reversal of symptoms was obtained after cessation of the drug. However, in those patients with severe fat wasting, toxicity persisted despite stavudine therapy discontinuation, suggesting that the lipodystrophy syndrome might be due to fat cell death and that the latter may be irreversible.
Thus, although it has been established that the development of peripheral lipodystrophy is related to the use of PI, the current observations suggest that patients receiving a nucleoside analogue therapy including stavudine, unlike PI-treated patients, may develop a syndrome of short-term fat wasting associated with increased plasma free fatty acid and triglyceride levels, and increased daily fat intake without insulin resistance. Duration of exposure to stavudine greatly increases the risk of this effect.
Our findings are consistent with those of preliminary reports evaluating the effect of PI cessation and substitution with a NNRTI over a 6 month period. In particular, Carr et al. demonstrated that peripheral fat mass continued to decline in patients who switched from PI- to NNRTI-based combination drug regimens, suggesting that factors other than PI contribute to fat loss.
A variety of metabolic abnormalities have been described in patients treated with highly active antiretroviral therapy (HAART) and because these conditions (lipodystrophy, hyperlipidaemia, central adiposity, insulin resistance) often cluster in the same patients, there has been speculation that a common mechanism could be responsible for these pathological states. Finally, we speculate that at least two syndromes might be distinguished: (i) a syndrome of fat depletion (peripheral lipodystrophy) which could be a direct drug effect; and (ii) a syndrome of fat accumulation (central adiposity) with and without loss of subcutaneous fat tissue from the legs and buttocks associated with insulin resistance, probably related to the magnitude of viral load decline as suggested in an earlier report. This latter syndrome might be a manifestation of chronic HIV infection exacerbated by the successful effect of HAART including a PI or a NNRTI regimen.
The syndrome of peripheral lipodystrophy is probably a complex process. Causality between fat wasting and the use of stavudine therapy needs to be clarified, particularly before decisions can be made on whether to switch to other drugs in affected patients. Randomized clinical trials are required to confirm our findings and to determine long-term consequences of these dysmorphic changes in the context of prolonged asymptomatic survival.
Large studies should be performed to determine which factors play a role in the predisposition to develop this toxicity. The role of apoptosis in the pathogenesis of lipodystrophy resulting in cell death and depletion of adipocyte cells should be investigated and, in addition, the potential role of mitochondrial dysfunction should be explored.
The authors thank G. Griffin, for critical reading of the manuscript, F. Bruno for assistance with the data analysis and the patients for their ongoing participation.