Lichtenstein, Kenneth A.; Ward, Douglas J.a; Moorman, Anne C.b; Delaney, Kathleen M.b; Young, Benjamin; Palella, Frank J. Jrc; Rhodes, Philip H.b; Wood, Kathleen C.d; Holmberg, Scott D.*; and the HIV Outpatient Study Investigators
On 11 June 1997, the US Food and Drug Administration issued a warning and report of 83 HIV-infected patients who developed diabetes mellitus and hyperglycemia while taking protease inhibitor (PI) antiretroviral drugs . In the subsequent year, there were several more, usually anecdotal, reports of clinically disordered fat distribution – coined ‘lipodystrophy’ – manifest as dorsocervical fat accumulation (`buffalo hump'), central adiposity, peripheral fat wasting, and associated laboratory abnormalities, such as elevated triglycerides and cholesterol, and insulin resistance or type 2 diabetes mellitus [2–8]. It is still unclear whether fat atrophy, fat accumulation, and the laboratory abnormalities are manifestations of one or several distinct syndromes.
Due to the temporal relationship of the appearance of most of these complications with the use of PI drugs, these agents were presumed to be the cause of this syndrome. However, more recent reports of lipodystrophy in HIV-1 infected persons before they had taken PI  or who had taken nucleoside reverse transcriptase inhibitors (NRTI) but not PI  have cast doubt on this association as wholly explanatory. Data presented at recent international meetings further support the observations that PI may not be exclusively associated with lipodystrophy. In fact, several retrospective cohort studies presented at these meetings demonstrated associations with NRTI, most notably stavudine [11,12].
As a result, confusion exists about the relationship between antiretroviral drugs and lipodystrophy, which has sometimes prompted changes in therapy. Adding to the confusion is the lack of a consistent definition of ‘lipodystrophy’ and the lack of reproducible measurements with anthropometrics and computerized tomographic and DEXA scans.
Given these many considerations and concerns about lipodystrophy, we undertook a comprehensive survey of ambulatory HIV-1 infected patients seen in the HIV Outpatient Study (HOPS) [14,15] to identify the clinical factors associated with the presence of both fat accumulation and fat atrophy.
HOPS has been described elsewhere [14,15]. Briefly, this dynamic cohort includes HIV-1 infected persons seen at eight clinics specializing in the treatment of HIV in seven cities (Philadelphia, Pennsylvania; Tampa, Florida; Oakland, California; Washington, DC; Chicago, Illinois; Stony Brook, New York; and Denver, Colorado). The physicians at these sites routinely care for hundreds of HIV-1 infected patients every year and are considered well-trained and experienced HIV-caregivers. Data collected and electronically submitted for central processing and analysis comes from real-time, ongoing physician–patient interactions and include demographic characteristics, symptoms, diseases, treatments, and laboratory values. More than 4800 HIV-1 infected ambulatory (non-hospitalized) patients have been seen in approximately 82 000 outpatient visits to their HOPS clinicians since 1992. Data is entered into the HOPS database for each patient visit. Up-to-date, quality-reviewed data for visits to the end of 1998 were used for this analysis.
In 1998, 2640 patients came in for 17 502 visits at HOPS clinics, an average of 6.6 visits per patient per year. Most patients have routine scheduled visits at least every 3–4 months. All patients who came in for a visit to HOPS sites between 1 October and 31 December 1998 were asked to participate in a single point-in-time survey and examination for physical evidence of fat redistribution. Of 1210 HOPS patients who visited the clinics during this period, 1077 (89%) voluntarily completed a face-to-face survey and examination by their physicians on signs of fat maldistribution using standardized descriptive definitions.
Body areas evaluated included enlarged abdomen (`belly paunch'), dorsocervical fat accumulation (`buffalo hump’ or enlarged neck), thinning of the extremities, thinning of the hips/buttocks, sunken cheeks, or other facial changes (fat deposition over parotid glands, temporal wasting, etc.). All changes were graded, both subjectively (by report) and objectively (by examination), as ‘subtle’ (noticeable only if specifically looked for, no change in clothing fit), ‘moderate’ (easily noted by patient or physician, clothing has become tight or loose), or ‘severe’ (obvious to the casual observer, has required a change in clothing size). The final severity rating for each sign was based on patient and physician agreement.
To reduce the influence of subtle physical manifestations incorrectly being attributed to the syndrome, we defined a lipoatrophy group which we called ‘moderate/severe lipoatrophy’ based on the number and severity of the three body areas affected (signs): sunken cheeks, thinning extremities, and thinning hips or buttocks. Patients with any ‘severe’ signs were included in the group. Patients whose most severe sign was ‘moderate’ were only included if they had at least one additional sign (subtle or moderate). Patients with a single moderate sign, but no additional signs and patients with one, two or three subtle signs were assigned to the ‘none/mild’ group. Although the severity grading was necessarily subjective, the analysis group was selected to minimize false positive inclusions.
We combined patients with no signs of lipoatrophy together with patients whose signs were too mild for our ‘moderate/severe’ group. This method of dividing the sample into those with the syndrome and those without was designed to bias the analysis against finding differences in associated factors between the two groups.
An analogous group was constructed to analyze moderate/severe lipoaccumulation using identical methodology for assessments of visceral abdominal fat, facial changes, and dorsal-cervical fat pad.
Laboratory values (including CD4 cell counts and percentages and plasma HIV-1 RNA), height, weights, HIV/AIDS status, history of opportunistic infections and drug treatment history were obtained from the regular HOPS database. Most patients have many years of observational data with a median observation time of 2.9 years (interquartile range, 1.7–4.8 years). In the HOPS, all available treatment data has been collected, but regardless of the availability of detailed history, a specific data point for date of first antiretroviral treatment has been carefully collected. This was used to adjust all analyses of drug effects by duration of cumulative time on any antiretroviral therapy. Immunology and virology lab data is extensive. All measurements taken throughout the observational period are available, 81.9% of patients had CD4 values and 88.1% had viral load assessments within the 3 months prior to the survey visit. Patients were weighed at each visit, which enabled analysis of body mass index (BMI) changes.
Across the period of observation for each patient, values and their dates were calculated for these measurements of CD4 cell count, percentage, and BMI: first, nadir (lowest ever), pre-nadir maximum, post-nadir maximum, and most recent value. Loss from maximum to most recent value, gain from nadir to most recent value and change from maximum to nadir values were calculated. For viral burden measurements, log10 values were calculated for: first, peak (highest ever), pre-peak minimum, post-peak minimum, most recent, viral burden gain, viral burden loss, and change in viral burden.
Specific factors examined for their associations with the presence of dysmorphic changes included: demographic factors (age, HIV transmission risk group, race, gender, study site); past and current use of antiretroviral drugs both by class [NRTI, non-nucleoside reverse transcriptase inhibitors (NNRTI), and PI] and individually; HIV disease characteristics, such as time since HIV or AIDS diagnosis; immunologic factors of CD4 cell count and percentage; virologic factors, as measured by HIV-1 plasma RNA; and BMI.
Univariate analyses of potentially related factors were performed as an initial review. To assess associations with lipoatrophy and lipoaccumulation, chi-squared tests were performed for each categorical variable, and the non-parametric Kruskal–Wallis tests of medians were used for each continuous variable.
Among the significant associations, there were differences in the prevalence of lipoatrophy and lipoaccumulation by clinic site (2.4–29.1% for lipoatrophy; χ2, P < 001; 4.5–14.3% for lipoaccumulation, χ2, P = 0.029). Therefore, analyses were stratified by or controlled for site. After associated factors had been determined, site differences were examined more closely. Most of the apparent site differences could be accounted for by differences in the other associated factors from one site to another. When adjusted for other significantly associated clinical factors, the site differences were not significant (for lipoatrophy, Cochran–Mantel–Haenszel χ2 test, P = 0.165; for lipoaccumulation, P = 0.535); however, site was always included in multivariate analyses.
Distributions of values for explanatory variables measured as continuous (e.g. time since AIDS diagnosis, nadir CD4 cell count) were examined and corresponding variables were grouped categorically for the stratified analysis.
Stratified analyses were performed adjusting for site, age and cumulative time on antiretroviral medication. For the analysis of the relationships between each potential related factor and presence of fat maldistribution, Cochran–Mantel–Haenszel chi-squared tests for grouped data were performed, controlling for clinical site, age group (< 30, 30–40, 40–50 and ≥ 50 years), and cumulative time on antiretroviral therapy (< 3 years, 3–5 years, 5–7 years, ≥ 7 years.)
There were strong correlations among many of the variables that were shown, in the stratified analyses, to be related to lipoatrophy or lipoaccumulation. Logistic regression models were constructed to assess further the existence of associations found in categorical analyses while adjusting for a large number of correlated effects. Variables that appeared to have an association in the stratified analysis were entered into multivariate models: models with various combinations of variables were compared to evaluate the significance of each factor and the fit of the model. To estimate adjusted odds ratios for variables where prevalence increased across each level, categorical variables were used. For distinct comparisons, risk factors were categorized further into dichotomous groups when appropriate.
To determine whether an association existed between the use of an antiretroviral drug and lipoatrophy or lipoaccumulation, a set of logistic regressions was performed, with each model adjusting for cumulative time on antiretroviral drugs and for use of one other antiretroviral drug. Current use, as well as cumulative use, of each antiretroviral drug and each combination regimen was analyzed for associations with lipodystrophy. This enabled, for example, the evaluation of the effect of stavudine while controlling for previous use of zidovudine. Other models controlled for use of each antiretroviral class. In this sample, use of antiretroviral therapy (ART) and use of NRTI were synonymous, as virtually all antiretroviral regimens taken by these patients included a NRTI.
During the period of observation 1077 patients (89% of all patients seen during this period) agreed to interview and examination for evidence of physical signs of fat maldistribution, and their medical and database records were reviewed for treatment, virologic and immunologic data. There were no important differences in demographic profile between the survey population and the HOPS cohort as a whole (data not shown). In the population surveyed the median age was 41 years, 85% of patients were male, 72% of patients were white, non-Hispanic, 67% were gay or bisexual men and 9% were injecting drug users.
A total of 529 (49% of the study population) examined patients were found to have one or more signs of fat maldistribution. The most commonly noted body change was central adiposity, seen in 30% of the entire cohort, followed by thinning of the extremities (26.5%), thinning of the buttocks (23%), sunken cheeks (18.7%), dorsocervical fat pad (10.3%), and other facial changes (9.8%). The number of body areas affected with fatty changes was related to the particular anatomical sites affected. For example, individuals were found to have only central adiposity (and did not meet our definition of significant lipoaccumulation), whereas almost all of those with dorsal cervical fat pad also had several other body areas affected.
Of the patients 143 (13.3%) had only signs of peripheral fat atrophy, 142 (13.2 %) had only signs of fat accumulation, and 244 (22.7%) had both. To determine associated clinical factors, we analyzed patients with moderate to severe lipoatrophy and those with moderate to severe lipoaccumulation. Because mild signs of fat maldistribution could be confused with other clinical conditions and there is no agreed-upon case definition, we classified patients who had subtle lipoatrophy with those who had no signs and considered only those individuals with moderate or severe lipoatrophy or moderate or severe lipoaccumulation as having the syndrome. Of the entire cohort, 15.9% (171 patients) met our criteria for significant lipoatrophy, and 104 (9.7%) met our criteria for significant lipoaccumulation (Table 1). Forty-seven patients (4.4%) had both moderate/severe lipoatrophy and moderate/severe lipoaccumulation.
There were significant differences in the prevalence of lipodystrophy by geographic HOPS location due, in large part, to variations in other demographic and clinical factors such as age, duration of HIV-1 infection and use of certain drugs. While differences were of borderline significance upon adjustment by these other factors, all analyses stratified or controlled for site differences.
Significant, but different , associations were seen when assessing the prevalence of fat atrophy or fat accumulation with host, disease, and drug related factors (Table 2 and Table 3). Whereas age and BMI changes were associated with both atrophy and accumulation, measures of duration and severity of illness, race, and specific antiretroviral medications were associated with atrophy (Table 2). The presence of hemophilia, durable and successful viral suppression, and cumulative time on all ART were associated with fat accumulation (Table 3).
There were strong correlations among all factors associated with lipoatrophy and lipoaccumulation in the categorical analyses (P-values < 0.05) (Table 2 and Table 3). These demographic, immunologic and drug treatment associations were confirmed in several multivariate models that adjusted for these shared effects (colinearity) (Table 2 and Table 3).
In both the stratified and logistic regression analyses advancing age was incrementally and strongly associated with the prevalence of fat maldistribution, particularly in patients over 40 years of age (Table 2 and Table 3).
A change in BMI of > 2 kg/m2 increased the likelihood of both fat atrophy and fat accumulation, with BMI loss > 1 kg/m2 associated with atrophy, and a BMI gain of > 1 kg/m2 associated with accumulation (Table 2 and Table 3). However, overall change in BMI was a more important predictor of lipoatrophy than BMI loss.
The prevalence of fat atrophy was incrementally associated with several measurements of duration and severity of HIV infection (Table 2). The presence of and time since an AIDS diagnosis, nadir CD4 cell count or percentage, and most recent CD4 cell count and percentage were also significant factors. Of these measures, logistic regression models showed that most recent CD4 cell count < 100 × 106/l or percentage < 15 were the best single predictors. White patients were significantly more likely to develop atrophy than non-white pateints, and black patients were significantly less likely to do so than non-black patients, even when adjusted for age, HIV disease status, duration of ART and use of particular antiretroviral medications.
There was a progressive likelihood of fat accumulation with increasing total time on any regimen of antiretroviral medication, lower peak viral load, higher BMI throughout observation, and increasing time with an undetectable viral load (Table 3).
In analyses that examined the association of various antiretroviral drugs, or their class, with the prevalence of lipoatrophy, the use of stavudine and indinavir were strongly linked to lipoatrophy (Table 4). Use of indinavir for more than 2 years was associated with an increased prevalence of lipoatrophy. Any use of indinavir also appeared to be associated with the increased prevalence of lipoatrophy, but this association was strongly affected by those patients taking that PI for longer than 2 years. Only 58 patients had received other PI drugs for more than 2 years when compared to the 176 who had received indinavir. The prevalence of moderate/severe lipoatrophy in the entire cohort was 7.8% (14/180) if the patient had not received a PI and 17.5% (157/897) if they had received a PI. In all of our analyses, the use of stavudine was associated with an increased prevalence of lipoatrophy compared to all other NRTI, even when controlling for the use of PI or other NRTI. Associations of lipoatrophy with specific antiretroviral drugs were assessed in both the stratified and logistic regression analyses while controlling for time on ART and for concomitantly used antiretroviral medications.
To assess further the relative contributions of antiretroviral drug versus other non-drug factors, the likelihood of lipoatrophy was evaluated in relation to use of stavudine and indinavir, compared to each non-drug factor and compared to the sum of non-drug-related factors (Table 5). The likelihood of lipoatrophy increased markedly, within each stratum of duration of drug use, with the number of non-drug associated factors (reading across rows, Table 5). In addition, the likelihood of lipoatrophy increased incrementally in relationship to increasing duration of use of either drug (reading down columns, Table 5). Of 120 patients who had no other associated non-drug related factors, none had moderate or severe lipoatrophy, even among those who had taken stavudine or indinavir for long periods of time.
A similar analysis compared the contributions of non-drug factors to the length of time on ART in relation to prevalence of lipoaccumulation. While fat accumulation was not associated with use of the specific drugs indinavir and stavudine, it was associated with duration of use of any antiretroviral drugs. In the entire cohort the prevalence of moderate/severe lipaccumulation was 3.9% (7/180) for patients who had not received a PI and 10.8% (157/897) for those who had. As was seen with lipoatrophy (Table 5), the effect of additional non-drug factors increased the likelihood of lipoaccumulation (data not shown).
In this study, fat maldistribution was generally associated with increasing age and many measurements of duration and severity of disease. In addition, fat atrophy was associated with stavudine and with indinavir, and fat accumulation was associated with several correlates of immunologic recovery.
While there were clear associations between lipoatrophy and use of stavudine, indinavir or both, there were no dysmorphic changes unless other (non-drug) clinical risk factors were present. There were, however, incremental increases in the prevalence of atrophy with these medications when non-drug clinical factors were present (Table 5, reading down columns). This suggests that lipoatrophy is related to effective control of HIV-1 infection or use of these drugs in patients with more advanced HIV-1 disease.
In addition to the drugs, we found significant relationships between any lipodystrophy and age, duration and severity of HIV-1 disease, and duration of immune or virologic recovery following ART. Each factor added an additional incremental increase in the prevalence of lipodystrophy, and the overall likelihood of lipodystrophy was strongly related to the cumulative effect of the associated factors (Table 5, reading across rows). We do not contend that changes in HIV infection or other non-drug factors would be necessarily associated with lipodystrophy without the drugs, but that the drugs’ effects were modulated substantially by each patient's history and underlying condition.
Also, it would be premature to conclude that of the PI only indinavir, the first PI widely used, is associated with lipoatrophy. Perhaps after a greater length of time on treatment with the other PI  – or any newer antiretroviral drug – an association with lipoatrophy would be observed.
The degree of prior immune suppression may predispose to the development of the syndrome. Of the measures of duration of recovery associated with lipoaccumulation, the length of time since nadir CD4 cell count or percentage, the length of time the patient maintained an undetectable viral load, and changes in BMI were all markers of reversal of immune suppression associated with lipoaccumulation. Whether HIV-1 or the medications alter fat regulatory mechanisms or adipocytes directly or indirectly has yet to be established. The effects of dysregulated pro-inflammatory cytokine levels, associated with reversal of advanced HIV-1 disease, may be involved . Perhaps the syndrome is affected by long-term changes in metabolism due to pre-treatment HIV-1, not seen previously due to high mortality rates or to the treatment, itself.
A cross-sectional and retrospective cohort analysis such as this is not designed to determine cause-and-effect relationships, and unavoidably suffers from unmeasured biases. In particular, the morphologic findings were based on a subjective, if standardized, clinical assessment. (Still this clinical assessment can be performed quickly and inexpensively in a clinic setting with little training and no equipment, and changes can be tracked from visit to visit). Also, the duration of observation varied by patient, and, putatively, the longer the observation period, the greater the opportunity to evaluate lipodystrophy, thereby biasing measures of severity of illness. Nonetheless, analyses controlled for these potential biases, and the statistical associations reported here were usually quite strong and unlikely to be due entirely to chance.
In conclusion, HIV associated lipodystrophy occurs frequently in long-term HIV infected patients. Lipoatrophy was strongly associated with the use of the drugs stavudine and indinavir, more than 2 years of PI use, and underlying patient conditions: increasing age, BMI loss, and duration and severity of HIV infection. Fat accumulation was associated with age, BMI gain, and measurements of recovery such as duration of successful virus suppression and duration of antiretroviral, mainly NRTI, use. Proof of these associations as independent risk factors, the relative strengths of their associations with lipodystrophy, and the sequencing of the factors to suggest an etiology could not be determined. Yet the results of this observational study suggest that lipodystrophy is very probably more than a simple adverse drug reaction and more than one syndrome.
The authors thank A. Greenberg and R. Janssen, Division of HIV/AIDS Prevention-Surveillance and Epidemiology, CDC, and J. Chmiel, Northwestern University Medical School, for their review and many helpful suggestions.
1. Food and Drug Administration. Protease inhibitors may increase blood glucose in HIV patients. FDA Med Bull 1997, 27: No.2.
2. Bonnet E, Cumin L, Sailler L, et al. Associated lipodystrophy metabolic disorders due to protease inhibitor containing regimens. XII International Conference on AIDS. Geneva, June–July 1998 [abstract 12299].
3. 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.
4. Dubé MP, Johnson DL, Currier JS, Leedom JM. Protease inhibitor-associated hyperglycaemia [letter]. Lancet 1997, 350: 713–714.
5. Gharakhanian S, Salhi Y, Nguyen H, et al. Frequency of lipodystrophy and factors associated with glucose/lipid abnormalities in a cohort of 650 patients treated by protease inhibitors. Sixth Conference on Retroviruses and Opportunistic Infections. Chicago, January–February 1999 [abstract 642].
6. Lo JC, Mulligan K, Tai VW, Algren H, Schambelan M. `Buffalo hump’ in men with HIV-1 infection. Lancet 1998, 351: 867–870.
7. Miller KK, Daly PA, Sentochnik D. et al. Pseudo-Cushing's Syndrome in human immunodeficiency virus-infected patients. Clin Infect Dis 1998, 27: 68–72.
8. Schindler JT, Spooner KM, Decker CF. `Buffalo humps’ associated with protease inhibitors [letter]. Ann Intern Med 1998, 129: 164.164.
9. Glesby M, Desai A, Erbelding E, Kostman J. High prevalence of impaired glucose tolerance prior to initiating protease inhibitor therapy. Sixth Conference on Retroviruses and Opportunistic Infections. Chicago, January–February 1999 [abstract 650].
10. Saint-Marc T, Partisani M, Poizot-Martin I. et al. A syndrome of peripheral fat wasting (lipodystrophy) in patients receiving long-term nucleoside-analogue therapy. AIDS 1999, 13: 1659–1667.
11. Carr A, Miller J, Law M, Cooper DA. A syndrome of lipodystrophy, lactic acidaemia and liver dysfunction associated with HIV nucleoside analogue therapy: contribution to protease inhibitor-related lipodystrophy syndrome. AIDS 2000, 14: 25–32.
12. Saint-Marc T, Touraine JL. The effects of discontinuing stavudine therapy on clinical and metabolic abnormalities in patients suffering from lipodystrophy. AIDS 1999, 13: 2188–2189.
13. Forum for Collaborative HIV Research (FCHR). Metabolic Abnormalities in HIV Disease and Treatment (Monograph). The Need for an Interim Case Definition and a Study to Examine Prevalence. Washington, DC: Center for Health Policy Research, George Washington University Medical Center; April 1999.
14. Moorman AC, Holmberg SD, Marlowe SI. et al. Changing conditions and therapies of ambulatory HIV-infected patients: The HIV Outpatient Study (HOPS). Ann Epidemiol 1999, 9: 349–357.
15. Palella FJ Jr, Delaney KM, Moorman AC. et al. Declining mortality and morbidity in HIV-infected ambulatory patients. New Engl J Med 1998, 328: 853–860.
16. Purnell JQ, Zambon A, Knopp RH. et al. Effect of ritonavir on lipids and post-heparin lipase activities in normal subjects. AIDS 2000, 14: 51–57.
17. Ledru E, Nevena C, Olivier P. et al. Alteration of tumor necrosis factor alpha T-cell homeostasis following potent antiretroviral therapy: contribution to the development of human immunodeficiency virus-associated lipodystrophy syndrome. Blood 2000, 95: 3191–398.
The HOPS Investigators include the following investigators and sites: A. C. Moorman and S. D. Holmberg, Division of HIV/AIDS Prevention, National Center for HIV, STD, and TB Prevention (NCHSTP), Centers for Disease Control and Prevention (CDC), Atlanta, GA.; K. C. Wood and R. K. Baker, Health Research Network of APACHE Medical Systems, Inc., McLean, VA.; F. J. Palella, J. S. Chmiel, C. Chan, and J. Arnold, Northwestern University Medical School, Chicago, Ill.; K. A. Lichtenstein, K. S. Greenberg, B. Young, B. Widick, C. Stewart, and P. Zellner, Denver Infectious Disease Consultants, Rose Medical Center, Denver, CO.; B. G. Yangco, K. Halkias, and C. Lapierre, Infectious Disease Research Institute, Tampa, FL; D. J. Ward and C. Owen, Dupont Circle Physicains Group, Washington, DC; J. Fuhrer, L. Ording-Bauer, R. Kelly, and M. Nekola, State University of New York (SUNY), Stony Brook, NY; E. M. Tedaldi and S. Smith, Temple University Hospital, Philadelphia, PA.; J. B. Marzouk, R. T. Phelps, and M. Rachel, Adult Immunology Clinic, Oakland, CA.; R. E. McCabe and M. Rachel, Fairmont Hospital, Oakland, CA. Cited Here...
© 2001 Lippincott Williams & Wilkins, Inc.