Risk factors for the HIV-associated lipodystrophy syndrome in a closed cohort of patients after 3 years of antiretroviral treatment


Dr Stefan Mauss, Grafenberger Allee 128a, D-40237 Duesseldorf, Germany. Tel: +49 211239 5520; fax: +49 211239 55210; e-mail: mausst@compuserve.com


Objective To identify prevalence and risk factors associated with the HIV-associated lipodystrophy syndrome (HIVLD) after 3 years of antiretroviral therapy, to investigate the diagnostic value of anthropometric measures and to assess the impact of HIVLD on quality of life.

Design and methods A prospective, cross-sectional, multicentre, observational, cohort study was performed in 27 German teaching hospitals, nonacademic hospitals and private practices. A total of 221 HIV-positive patients commencing antiretroviral therapy between July and September 1996 were studied. The main outcome measure was lipodystrophy, defined as otherwise unexplained truncal fat accumulation and/or fat loss in face or extremities. The analysis consisted of multiple logistic regression models, receiver operating characteristics (ROC) curves for anthropometric measures and visual analogue scales for quality of life.

Results The prevalence of HIVLD after 3 years was 34%. The following variables were independently associated with HIV-LS: stavudine use > 12 months [odds ratio (OR) 2.1, 95% confidence interval (CI) 1.1–3.9], CD4 count nadir < 200 cells/µL (OR 2.2, CI 1.1–4.6), hypertriglyceridaemia (OR 2.3, CI 1.3–4.2) and nonnucleoside reverse transcriptase inhibitor (NNRTI) intake > 12 months (OR 0.2, CI 0.04–0.87). No cut-off point was found for anthropometric indices with a sensitivity and specificity of ≥ 0.8. The mean visual analogue ratings for impaired quality of life, on a scale of 0–10, were: 5.2 (self-esteem), 2.9 (social contacts), 4.2 (sexuality) and 3.5 (daily activities).

Conclusions These findings suggest a multifactorial aetiology for HIVLD. Stavudine use and a CD4 count below 200 cells/µL may be associated with an increased risk for the development of HIVLD. In contrast, NNRTI treatment may be associated with a reduced risk. Anthropometric indices were found to be insufficient as a diagnostic tool. Quality of life was severely affected by HIVLD.


The HIV-associated lipodystrophy syndrome (HIVLD), also known as the fat redistribution syndrome, is characterized by sustained changes in body fat distribution, often accompanied by metabolic disturbances such as hypertriglyceridaemia and impaired blood glucose tolerance [1,2]. Altered body shape is regarded as the core element of the syndrome. The most widely used diagnostic criterion is an otherwise unexplained central fat accumulation or peripheral fat loss as assessed by the patient or the physician [3].

The reported prevalence of HIVLD ranges from 1% to 80%[4]. This wide range may at least partially be due to differences in case definitions, ascertainment methods and patient populations. However, most observers agree that the prevalence of the syndrome is rising. There is concern that HIV-LS may adversely affect adherence to antiretroviral treatment, secondary to a diminished quality of life.

Recent observational studies have tried to identify risk factors. Many of these studies are retrospective and most are based on patient populations with large differences in the cumulative time on antiretroviral therapy, but only a few of the studies adjust for this in analysis. Thus, the results may be confounded, since duration of antiretroviral therapy itself has been reported to be a risk factor.

Standardized diagnostic tools have not yet been determined. Computed tomography and magnetic resonance imaging scans have been proposed, but these methods are costly, and data from reference populations are lacking. In contrast, anthropometric measures are simple and inexpensive, but have not been validated [5].

The study presented here is based on a closed prospective cohort of HIV-positive patients who all started antiretroviral treatment between July and September 1996. The presence of HIVLD was assessed after 3 years of antiretroviral therapy. The study focuses on three aspects of HIVLD and includes: (1) a determination of prevalence and identification of risk factors for the occurrence of HIVLD; (2) investigation of the diagnostic value of simple anthropometric measures; and (3) an assessment of the impact of HIVLD-associated body changes on different domains of quality of life.

Patients and methods


Seventy-nine centres of the German, closed, multicentre, cohort study ART 96 followed a total of 828 HIV- seropositive patients who first started antiretroviral therapy between July and September 1996 [6]. Twenty-seven of these centres participated in this study (DAGNAE LipART). They were selected to reflect the cohort as a whole on the basis of regional representation (in order to follow a geographically diverse population), of clinic and practice representation (in order to have diverse practice settings), of clinician experience (must follow at least 10 HIV-infected patients) and of internet availability (in order to allow online data and information transmission). All available patients with complete follow-up data, who gave informed consent, were included.

For all patients, baseline data included demographics, time since first positive HIV-antibody test, risk behaviour, Centers for Disease Control classification from 1993, CD4 cell count and HIV-1 RNA. Follow-up data included detailed records of antiretroviral therapy, including reasons for treatment changes, AIDS-defining events and quarterly measurements of CD4 cell counts and viral load.

Evaluation of lipodystrophy syndrome

A purely atrophic type of lipodystrophy was defined as a loss of fat in at least one of the following body compartments: cheeks, buttocks, arms and legs. A purely hypertrophic type of lipodystrophy was defined as an increase of fat in at least one of the following body compartments: neck, breast and abdomen. A mixed type of lipodystrophy was defined as the presence of at least one change consistent with the atrophic and lipodystrophic type as defined above.

All patients were seen for the assessment of lipodystrophy between April and December 1999 after written informed consent was obtained. Data were obtained using two physician and one patient standardized questionnaire; these were pretested in a pilot project. Physicians were asked to state whether or not lipodystrophy was present. To rate the extent of changes in body habitus separately for the six body compartments (cheeks, neck, breast, abdomen, buttocks and legs), visual analogue scales were used ranging from −5 to +5.

The circumference of the waist, hip and right thigh was measured according to World Health Organization (WHO) recommendations [7]. Detailed graphic and written instructions on measurement points were provided on the reporting forms to ensure reliability. The data were used to compute two anthropometric indices (waist:hip ratio and waist:thigh ratio).

Before each patient was seen by the doctor, he/she was asked to complete a standardized questionnaire. All patients were asked to rate body changes using the same visual analogue scales as the physicians. The patients also rated the impact of body changes on the quality of life domains of social contacts, sexuality, self-esteem and performance of daily activities, using visual analogue scales ranging from 0 (no impact) to 10 (maximum impact).

Information was recorded on each patient's history of cardiovascular events, presence or absence of diabetes mellitus, use of lipid lowering drugs, body weight and height.

Blood samples were analysed for serum cholesterol, triglycerides and blood glucose.

Data analysis

Demographic, clinical, immunological, virological and treatment-related variables were tested for an association with lipodystrophy in univariate analyses using χ2 techniques. For several variables, different categories for testing purposes had been predefined.The level of viral load detection was set to 500 HIV-1 RNA copies/mL, due to historical detection limits of the tests used in the early years of the study. The duration of drug intake was categorized and tested as ever taken or taken for ≥ 6, ≥ 12 or ≥ 18 months. CD4 count categories tested were ≤ 500, ≤ 200 and ≤ 100 cells/µL.

Multivariate analyses were performed using multiple logistic regression models with P < 0.1 for entry into the model and P < 0.05 for remaining in the model. Associations between quantitative ratings of body changes, domains of quality of life and other variables were assessed by Spearman's rank correlation analysis. Test properties of anthropometric indices (waist:hip ratio, waist:thigh ratio) as a diagnostic tool for lipodystrophy were analysed using receiver operating characteristic (ROC) curve analyses with clinicians' diagnoses as the standard. All analyses were performed using the Statistical Analysis System [8].


A total of 221 patients were enrolled in the study (DAGNAE LipART). Demographics and baseline characteristics are shown in Table 1. Baseline characteristics of patients included in the LipART therapy sample did not differ significantly from the ART 96 cohort from which the sample was drawn [6].

Table 1.   Demographics and baseline characteristics of the study population (n=221) at the start of antiretroviral therapy
  1. Values are medians (ranges) or absolute numbers (%). IDU=injecting drug user.

Age (years)38(20–70)
Mode of HIV transmission
HIV-1 RNA (copies/mL)58 000(500–1900 000)
CD4 count (cells/µL)276(4–936)
CD4 nadir < 200 cells/µL77(35%)
AIDS at baseline28(13%)
Baseline antiretroviral therapy
 Double combination152(69%)
 Triple combination62(28%)

The use of antiretroviral drug classes and characteristics at month 36 are summarized in Table 2. Almost two-thirds of all subjects had received treatment with protease inhibitors (PIs) for over 1 year. Indivinar was taken by 79 patients (36%), nelfinavir (NFV) by 72 (33%), ritonavir (RTV) by 67 (30%), saquinavir (SQV) by 59 (27%) and amprenavir (APV) by one patient. The mean cumulative time on each drug was 603 days for indinavir, for 517 days for NFV, 345 days for RTV, 447 days for SQV and 87 days for APV. Nonnucleoside reverse transcriptase inhibitor (NNRTI) use was distributed equally between nevirapine (n=39) and efavirenz (EFV; n=38), with only two patients treated with delavirdine.

Table 2.   Characteristics and history of antiretroviral treatment after 3 years of follow-up
  1. Values are means ± SD or absolute numbers (%). 3TC=lamivudine; ZDV=zidovudine; d4T=stavudine; ddI=didanosine; NNRTIs=nonnucleoside reverse transcriptase inhibitors; PIs=protease inhibitors.

Body weight (mean ± SD kg71(± 13.3)
Body mass index (kg/m2)22.6(± 3.6)
Steroid use33(15%)
Hypertriglyceridaemia (> 200 mg/dL)92(42%)
Hypercholesterolaemia (> 200 mg/dL)113(52%)
Diabetes mellitus (> 140 mg/dL)12(5%)
Cardiovascular events9(4%)
Antiretroviral historyEver> 12 month
 3TC209 (95%)191 (86%)
 ZDV200 (90%)140 (63%)
 d4T137 (62%)98 (44%)
 ddI81 (37%)38 (17%)
 NNRTIs73 (33%)18 (8%)
 PIs158 (72%)136 (62%)

Diabetes mellitus was infrequent. Cardiovascular events consisted of three cases of newly diagnosed symptomatic coronary heart disease, including two myocardial infarctions and one right ventricle failure. Three were reported as hypertensive crisis and may not be considered true cardiovascular events.

The prevalence of HIV-LS was 34% (n=76/221) by physician's clinical diagnosis. Isolated lipoatrophy was found in 14 of these patients (18%) and isolated lipohypertrophy was reported in six patients (8%). The mixed type, combining atrophic and hypertrophic features, was most prevalent, reported in 56/76 patients (74%).

In univariate analyses, the following parameters were significantly associated with the presence of lipodystrophy: intravenous drug use, CD4 nadir < 200 cells/µL, stavudine (d4T) use (ever and > 12 months), hypertriglyceridaemia (> 200 mg/dL) and hypercholesterolaemia (> 200 mg/dL). No association was found between between PI use, as a class or as individual agents, and lipodystrophy. In contrast, NNRTI use for > 12 months was significantly associated with a decreased risk of lipodystrophy (Tables 3 and 4). The associations of hypertriglyceridaemia and hypercholesterolaemia with HIVLD were stronger when measured in fasting than nonfasting patients [hypertriglyceridemia: odds ratio (OR) 3.8 vs. 2.8; hypercholesterolaemia: OR 2.7 vs. 1.8].

Table 3.   Univariate analysis of associations with HIV-associated lipodystrophy
VariableOdds ratio95% Confidence intervals
  1. Demographic categories based on median age, median weight and median body mass index in the study population. *P < 0.05;**P < 0.01.

Age > 36 years0.71 0.41–1.25
Mode of transmission Homosexual1.11 0.61–2.01
IDU2.6 0.98–6.89*
Male0.51 0.24–1.08
CD4 nadir < 200 cells/µL1.92 1.08–3.42*
HIV-1 RNA < 500 copies/mL ever1.21 0.69–2.14
HIV-1 RNA < 500 copies/mL > 12 months1.0 0.54–1.84
Baseline AIDS1.42 0.60–3.37
Weight < 70 kg0.86 0.49–1.50
Baseline body mass index < 23 kg/m21.035 0.53–1.81
Smoking1.60 0.73–3.52
Hypertriglyceridaemia2.78 1.57–4.92**
Hypercholesterolaemia1.77 1.02–3.14*
Diabetes mellitus0.96 0.28–3.29
Cardiovascular events1.00 0.243–4.12
Table 4.   Univariate analysis of treatment history after 3 years and associations with HIV-associated lipodystrophy
 Odds ratio95% Confidence intervals
  1. PI=protease inhibitor; d4T, stavudine; ZDV=zidovudine; ddI=didanosine; 3TC=lamivudine; NNRTI=nonnucleoside reverse transcriptase inhibitor. *P < 0.05;**P < 0.01.

Baseline double vs. triple therapy0.840.45–0.59
PI ever1.610.85–3.06
PI > 12 months1.440.81–2.58
d4T ever1.831.01–3.32*
d4T > 12 months2.321.32–4.08**
ZDV ever1.050.41–2.73
ZDV > 12 months0.760.43–1.35
ddI ever1.690.95–2.99
ddI > 12 months1.940.95–3.94
3TC ever0.720.22–0.59
3TC > 12 months0.640.23–1.83
NNRTI ever0.830.45–1.50
NNRTI > 12 months0.220.05–0.97*

In multivariate logistic regression models, a CD4 count nadir of < 200 cells/µL (P=0.03), d4T use for > 12 months (P=0.02) and hypertriglyceridaemia (P=0.0007) were identified as being independently associated with HIV-LS. NNRTI use for > 12 months (P=0.03) was independently associated with a decreased probability of HIVLD (Table 5).

Table 5.   Multivariate analysis of risk factors associated with HIV- associated lipodystrophy: adjusted odds ratios
VariableOdds ratio95% Confidence intervals
  1. d4T=stavidine; NNRTI=nonnucleoside reverse transcriptase inhibitor. *P < 0.05;**P < 0.01.

d4T > 12 months2.10 1.1–3.9*
CD4 count nadir < 200 cells/µL2.21 1.1–4.6*
NNRTI > 12 months0.22 0.042–0.87*
Hypertriglyceridaemia2.3 1.3–4.2**

Based on these variables, high risk and low risk profiles were computed. A patient with a CD4 count nadir of < 200 cells/µL, d4T intake for > 12 months, elevated triglycerides and no NNRTI use > 12 months had a 72% probability (OR 10.8) of presenting with HIVLD. In contrast, a patient with a low risk profile (CD4 count never < 200 cells/µL, no d4T use for > 12 months, no hypertriglyceridaemia and treatment with NNRTIs for > 12 months) had a 4% probability (OR 0.17) of presenting with HIV-associated lipodystrophy.

Anthropometric measures and indices (waist:hip ratio, waist:thigh ratio) showed poor test characteristics as diagnostic tools for HIVLD. In the ROC analysis, no cut-off point could be determined where both sensitivity and specificity were=0.8 (Fig. 1).

Figure 1.

 Receiver operating characteristic curve for anthropometric indices as diagnostic tests for HIV-associated lipodystrophy.

All HIVLD-positive patients perceived their body changes as a health disturbance. Also, all these patients reported that at least one of four quality of life domains were affected by the lipodystrophy. Impact on social contacts was reported by 63%, on the performance of daily activities by 68%, on sexuality by 68% and on self-esteem by 83% of the patients. By visual analogue scales, the mean impact of HIVLD on social contacts was 2.86 (SD 3.05), on sexuality 4.20 (SD 3.51), on self-esteem 5.23 (SD 3.36) and on performance of daily activities 3.54 (SD 3.30).


In this cohort of 221 patients, the prevalence of clinical HIVLD was 34% after 3 years of antiretroviral therapy. Previous studies have identified the total duration of antiretroviral treatment as an independent risk factor for HIVLD [3,9,10]. In a population with variable durations of antiretroviral therapy, information on the effects of other factors, in particular those of individual antiretroviral agents, may be skewed unless the data are adjusted for total time on antiretroviral treatment with sufficient test power. In the present study, this issue was controlled by the study design, since all patients started antiretroviral treatment over the same time frame (July–September 1996). A limitation of this study is the absence of an assessment for HIVLD before the commencement of antiretroviral therapy. At that time, however, lipodystrophy had not been established as a possible adverse reaction of this therapy.

Multivariate analysis showed that prolonged treatment with d4T, a history of marked cellular immunosuppression and elevated blood triglyceride levels were associated with an increased risk of HIVLD. Prolonged periods of treatment with NNRTIs appeared to be associated with a reduced risk. PI treatment itself was not independently associated with an increased risk of HIVLD, and an analysis of the individual PIs did not change this finding. The impact of PIs may have been overestimated in earlier reports, if not controlled sufficiently for the use of specific nucleoside analogue reverse transcriptase inhibitors (NRTI), total duration of antiretroviral treatment and sequential treatment strategies starting with NRTI regimens and adding PIs [1,11–13]. Our finding is in agreement with the results from other studies that also failed to demonstrate a strong effect of PIs on the risk of HIVLD [14,15]. Small to moderately synergistic effects of other NRTIs than d4T and the PIs cannot be ruled out by this study and could become detectable with longer observation periods or larger study populations. The present study emphasizes the dominant role of d4T intake as a risk factor for HIVLD, as reported in other recent cohort studies [10,16–18]. Mitochondrial toxicity of d4T and other NRTIs has been proposed as the underlying mechanism [19,20]. Additional analyses of early vs. delayed use of d4T and the temporal sequence of d4T and zidovudine did not change the results. The reduced risk for HIVLD associated with NNRTI treatment that is suggested by our data is consistent with preliminary findings reported from a cohort of patients treated with EFV or indinavir in combination with two NRTIs and with the results of a recently published cohort study [10,21].

A history of a low CD4 cell count was also found to be a risk factor in an Australian cohort study [3]. Cellular immunosuppression is associated with disturbances in lipid metabolism and cytokines, which may promote the development of HIVLD [22–24]. However, the mechanism underlying a direct contribution of HIV disease to HIVLD remains hypothetical.

Elevated serum triglycerides were associated with lipodystrophy. The association was even stronger in the subgroup of patients for whom fasting samples were available. An association between hypertriglyceridaemia and lipodystrophy has been reported previously [3]. It has also been observed frequently in HIV-positive patients on antiretroviral therapy without lipodystrophy [25]. Factors contributing to hypertriglyceridaemia are reported to include antiretroviral medication, cytokine disturbances, particularly elevated endogenous interferons and loss of subcutaneous fat as a depot for triglycerides [22,23,26–28].

From a clinical perspective, it may be useful to define a high risk patient profile. However, our models have not been tested in an independent, prospective validation set, and therefore their properties as a prediction tool are uncertain.

The effects of antiretroviral treatment should be assessed prospectively, simultaneously considering efficacy, tolerance and risk of HIVLD. Clinical research on HIVLD is still being hampered by the absence of a generally accepted case definition and the lack of standardized diagnostic tools. Computed tomography, magnetic resonance imaging techniques and dual energy X-ray absorption (DEXA) have been proposed, but have not been widely used in routine patient care. This is likely due to high costs and the lack of validated clinical standards. Standardized indices based on anthropometric measures are, in theory, a simple and low-cost alternative. However, anthropometric indices like the waist:hip ratio and the waist:thigh ratio did not perform well in this study.

As may have been expected, body changes due to lipodystrophy adversely affected self esteem, social contacts, sexuality and daily performance in the majority of patients. Semiquantitative severity ratings using visual analogue scales revealed a serious impact on quality of life. These findings lend further support to the hypothesis that lipodystrophy may lead to impaired adherence to antiretroviral treatment and emphasize the need for further preventive and therapeutic efforts against HIVLD.


We thank the members of the LipART study group: L Weitner, H-J Stellbrink and A Stoehr, Hamburg; W Becker, G Gorriahn, E Jägel-Guedes and Z Lichtenstein, Munich; W Brust and W Schuster, Mannheim; S Dupke and A Moll, Berlin; P Gehring, Dortmund; L Locher, P Gute and S Prister, Frankfurt am Main; B Kuhlmann and P von Wussow, Hannover; F Mosthaf, Karlsruhe; G Schmutz, Düsseldorf; X Pfeil, Leipzig; J Rockstroh, Bonn; E Schnaitmann and A Ulmer, Stuttgart; S Staszewski, Frankfurt; K Starke, Wiesbaden; W Wiesel, Köln.