We evaluated metabolic and clinical features of 71 HIV-infected patients with lipodystrophy by comparing them with 213 healthy control subjects, matched for age and body mass index, from the Framingham Offspring Study. Thirty HIV-infected patients without fat redistribution were compared separately with 90 matched control subjects from the Framingham Offspring Study. Fasting glucose, insulin, and lipid levels; glucose and insulin response to standard oral glucose challenge; and anthropometric measurements were determined. HIV-infected patients with lipodystrophy demonstrated significantly increased waist-to-hip ratios, fasting insulin levels, and diastolic blood pressure compared with controls. Patients with lipodystrophy were more likely to have impaired glucose tolerance, diabetes, hypertriglyceridemia, and reduced levels of high-density lipoprotein (HDL) cholesterol than were controls. With the exception of HDL cholesterol level, these risk factors for cardiovascular disease (CVD) were markedly attenuated in patients without lipodystrophy and were not significantly different in comparison with controls. These data demonstrate a metabolic syndrome characterized by profound insulin resistance and hyperlipidemia. CVD risk factors are markedly elevated in HIV-infected patients with fat redistribution.
HIV lipodystrophy syndrome is a recently recognized syndrome characterized by body composition changes, including the development of an enlarged posterior cervical fat pad (i.e., a buffalo hump), increased truncal adiposity, breast enlargement, peripheral fat loss, and facial fat atrophy [ 1–4 ]. The mechanism of the HIV lipodystrophy syndrome remains unknown. Recent studies suggest that up to 83% of patients treated with protease inhibitor (PI) therapy develop fat redistribution [ 5 ], but fat redistribution has also been reported among non–PI-treated patients [ 2 , 6 ]. Previous studies of the lipodystrophy syndrome often selected patients on the basis of medication status—for example, PI use, combination antiretroviral therapy, or specific medications [ 1 , 7–9 ]—and were not always limited to patients with fat redistribution.
Metabolic disturbances and fat redistribution have been characterized in HIV-infected patients receiving combination antiretroviral therapy according to drug exposure and patterns of fat redistribution [ 10 ]. In contrast, we investigated metabolic abnormalities in a well-characterized cohort of HIV-infected patients with clinical lipodystrophy in comparison with an established population of healthy control subjects with known cardiovascular disease (CVD) risk parameters. In order to establish the extent of lipid abnormalities, glucose intolerance, and increased CVD risk in HIV-infected men and women experiencing lipodystrophy, we evaluated metabolic and clinical parameters in comparison with healthy control subjects from the Framingham Offspring Study who were matched for age, sex, and body mass index (BMI). We investigated sex-specific differences in CVD risk parameters and the extent to which fat redistribution contributes to the metabolic abnormalities in HIV-infected men and women with lipodystrophy syndrome.
Patients and Methods
Case patients . Seventy-one patients (49 men and 22 women) with HIV infection who reported recent changes in body fat distribution (case patients) were prospectively evaluated from December 1998 through July 1999 at the Clinical Research Center of the Massachusetts Institute of Technology. Patients were recruited from respondents to community-based advertisements seeking HIV-infected patients with fat redistribution, or they were referred by their physicians for evaluation of fat redistribution. Patients were screened by telephone and asked if they had experienced any of the following: (1) loss of fat in the face, (2) increased fat under the chin or back of the neck, (3) increased abdominal girth, (4) increased chest or breast fat, or (5) loss of fat in the arms or legs. Patients who identified a change in fat distribution in ⩾1 body areas were invited to participate, and fat redistribution was confirmed by physical examination for all patients (see below).
Patients were excluded if they had changed antiviral medications within 6 weeks of the study; had a history of diabetes mellitus or previous treatment with antidiabetic agents; reported use of testosterone, estrogen, growth hormone, or other steroids in the past 6 months; were active alcohol or substance abusers; or were not 18–60 years of age. A subsample of the patients who participated in this evaluation were subsequently enrolled in a treatment study of metformin for HIV lipodystrophy. The results of this study have been published [ 11 ] but do not include any of the data presented here.
Case patient examination protocol . Each case patient underwent a complete medical history and physical examination to confirm fat redistribution. Fat redistribution was scored as absent (0), present (1), or severe (2) by a single investigator with regard to facial fat loss, increased neck fat (anterior or posterior), increased trunk or chest fat, and decreased leg and arm fat. Patients were categorized as having primary lipoatrophy (with no evidence of increased abdominal fat), primary lipohypertrophy (increased abdominal fat with or without increased neck fat and no evidence of peripheral fat atrophy), or mixed lipodystrophy (presenting with both increased abdominal fat and peripheral fat atrophy). Blood pressure, weight, height, BMI (expressed as weight in kilograms divided by height in meters squared), and waist (at the umbilicus), hip, midarm, and midthigh circumference were determined [ 12 ]. After a 12-h fast, a standard 75-g oral glucose tolerance test (OGTT) was performed according to World Health Organization (WHO) standards [ 13 ] with determination of blood glucose and insulin level, at 0 and at 1–2 h. In addition, fasting cholesterol, low-density lipoprotein (LDL), high-density lipoprotein (HDL), and triglyceride levels were obtained.
Antiretroviral therapy was characterized as to current or past use and duration of therapy with PIs, nucleoside reverse transcriptase inhibitors (NRTIs), and/or or nonnucleoside reverse transcriptase inhibitors (NNRTIs). Insulin resistance (HOMA IR) and pancreatic beta cell function (HOMA Beta) were calculated with use of the homeostatic model [ 14 ], a mathematical estimate of insulin sensitivity and resistance derived from values for fasting glucose and insulin concentration. Patients were classified as having significant hyperinsulinemia if their fasting insulin level was >18 µU/mL, the 90th percentile of the fasting insulin distribution among healthy glucose-tolerant Framingham subjects eligible for matching (see below).
HIV-infected control patients without lipodystrophy . Thirty HIV-infected patients (18 men and 12 women) without clinical evidence of fat redistribution were also evaluated. None of these patients had experienced fat redistribution of the face, neck, arms, legs, or trunk, and this was confirmed by physical examination in all cases. The remaining exclusion criteria were identical to the criteria used for HIV-infected patients with lipodystrophy.
Framingham Study control subjects . Control subjects for the HIV-infected patients with and without lipodystrophy were selected from the Framingham Offspring Study, a population-based observational study of risk factors for CVD described elsewhere [ 15 , 16 ]. Participants were primarily white. Control subjects were eligible for matching if they did not have diagnosed diabetes mellitus, were not on estrogen replacement therapy, and had complete information on all analytic covariates ( n = 2959). We matched each case patient and each patient without lipodystrophy to 3 control subjects on the bases of sex, age (within 5 years), and BMI (within 1 kg/m 2 ). Data on control subjects were obtained from January 1991 through September 1993, during the fifth Framingham Offspring Study examination cycle.
As with case patients, control subjects underwent a standardized medical history and physical examination, including measurement of height, weight, blood pressure, and waist (at the umbilicus), hip, midthigh, and midarm circumference. After measurement of levels of fasting glucose, insulin, and lipids, control subjects underwent a 75-g OGTT administered according to WHO standards, and 2-h postchallenge glucose and insulin levels were measured.
Biochemical and immunologic assays . Levels of glucose, insulin, cholesterol, triglyceride, and LDL were determined in both case patients and control patients by use of methods reported elsewhere [ 16 ]; the same methods were used in this study for both case patients and control subjects. Insulin levels were measured by radioimmunoassay in the same laboratory with the same kit (Diagnostic Product) to maximize comparability between the groups [ 17 ]. Insulin levels were measured in serum in the HIV-infected patients; plasma insulin levels measured in the Framingham subjects were calibrated to serum equivalents as previously described [ 16 ]. Insulin intra-assay and interassay coefficients of variation were 5.0%–10.0%, and cross-reactivity with proinsulin at the midcurve was at least 40%. HIV virus load was determined by use of the Amplicor HIV-1 Monitor test (Roche Molecular Systems) with a lower limit of detection of 400 copies/mL.
Statistical analysis . We made crude bivariate comparisons of clinical end points for HIV and for control subjects by use of a Student's t test. Insulin and triglyceride levels were log transformed before statistical comparison. We used a series of linear regression models, including covariates for waist-to-hip ratio and waist, hip, midthigh, or midarm circumference to control for the effects of regional fat distribution on differences between the groups. We also determined the effects of regional fat distribution on fasting insulin with a single model that included all covariates (waist, hip, midthigh, and midarm) stratified by group (HIV and control subjects). Analyses were stratified by sex and repeated. Logistic regression models with and without adjustment for waist-to-hip ratio were used to assess the risk of 2-h glucose levels of >140 mg/dL and >200 mg/dL [ 13 ], a fasting insulin level >18 µU/mL, systolic blood pressure >140 mm Hg, diastolic blood pressure >90 mm Hg [ 18 ], a cholesterol level >200 mg/dL, a triglyceride level >200 mg/dL, and an HDL level <35 mg/dL [ 19 ] between HIV-infected patients with lipodystrophy and Framingham control subjects and between HIV-infected patients without lipodystrophy and Framingham control subjects.
Clinical variables were compared within the HIV-infected patients with lipodystrophy by PI exposure and by pattern of lipodystrophy (lipoatrophy, lipohypertrophy, and mixed lipodystrophy). PI exposure was categorized as current use, past use, or PI naive. The effects of PI exposure on hyperinsulinemia were also analyzed by use a multivariate regression model predicting fasting insulin level, with age, BMI, waist-to-hip ratio, current use of PI therapy (yes/no), total duration of PI exposure, and total duration of NRTI therapy included in the model as predictors. Statistical analyses were performed by use of SAS software (SAS Institute, Cary, NC) [ 20 ]. Statistical significance was determined by use of 2-tailed tests; P ⩽ .05 was considered significant). Results are presented as mean ± SEM unless otherwise indicated.
Clinical characteristics . Clinical characteristics of the HIV-infected patients with lipodystrophy, HIV-infected patients without lipodystrophy, and their respective control subjects are shown in table 1 . Seventy-seven percent of the case patients were white, 11% were black, and 11% were Hispanic. Level of education was similar for case patients and for the Framingham control subjects (mean education, 15 years for each group). Of the female case patients, 86% were eumenorrheic. Seventy percent of case patients were receiving PI therapy at the time of the study, 100% were receiving a NRTI, and 30% were receiving a NNRTI. Eighty percent of the case patients were receiving highly active antiretroviral therapy (HAART), defined as the use of 2 NRTIs and either a PI or a NNRTI. Virus load was undetectable in 65% of the case patients.
Comparison of clinical variables between case patients and Framingham control subjects . HIV-infected patients with lipodystrophy demonstrated an increased waist-to-hip ratio, increased waist circumference, reduced hip circumference, and reduced midthigh circumference compared with the age- and BMI-matched Framingham control subjects ( table 2 ). Among case patients, levels of fasting insulin, 2-h insulin, 2-h glucose, cholesterol, triglyceride, the cholesterol-to-HDL ratio, and diastolic blood pressure were higher and HDL lower than they were among the control subjects ( table 2 ). Fasting glucose levels, systolic blood pressure, and LDL levels were not significantly different between the groups. HOMA IR and HOMA β were significantly higher in the case patients than in the Framingham control subjects. A subgroup analysis performed among the white HIV-infected ( n = 55) and matched Framingham control subjects ( n = 165) yielded similar results. There were significant positive associations between total lipodystrophy score and waist-to-hip ratio ( r = 0.33, P = .006) and fasting insulin level ( r = 0.26, P = .04).
Adjusting for waist-to-hip ratio did not substantially attenuate most differences between case patients and control subjects ( table 2 ). Differences in the levels of fasting insulin, 2-h insulin, 2-h glucose, cholesterol, and HDL remained significant between the groups ( table 2 ). The difference in diastolic blood pressure between the groups was no longer significant after adjusting for waist-to-hip ratio. Adjustment for waist, hip, midarm, or midthigh circumference instead of waist-to-hip ratio yielded similar results. In a single multivariate model that included hip, midthigh, waist, and midarm circumference measurements and was stratified by subject group, we found that waist ( P = .001) and thigh circumference ( P = .041) were significant predictors of the fasting insulin level among the case patients, whereas only the waist measurement ( P = .0001) predicted the fasting insulin level among the Framingham control subjects.
Risks among HIV-infected patients with lipodystrophy . HIV-infected patients with lipodystrophy were more likely than age- and BMI-matched Framingham control subjects to have impaired glucose tolerance (2-h glucose levels >140 mg/dL) (OR, 10.0; 95% CI, 4.6–21.7; P = .001), 2-h glucose levels >200 mg/dL (OR, 16.1; 95% CI, 1.8–140; P = .01), triglyceride levels>200 mg/dL (OR, 13.6; 95% CI, 7.0–26.5; P = .001), and HDL levels <35 mg/dL (OR, 4.1; 95% CI, 2.3–7.5; P = .001; table 3 ; figure 1 ). A greater number of case patients demonstrated diastolic blood pressure >90 mm Hg (11.3% vs. 5.6%) and LDL levels >160 mg/dL (21.8% vs. 14.1%), but these differences were not significant. In contrast to the significant increase in 2-h glucose level after glucose tolerance testing, the fasting blood glucose level was normal (<126 mg/dL) in all case patients.
Differences in the risk of impaired glucose tolerance and 2-h glucose levels >200 mg/dL, triglyceride levels >200 mg/dL, and HDL levels <35 mg/dL remained highly significant after controlling for waist-to-hip ratio ( table 3 ). A greater proportion of case patients than Framingham control subjects had cholesterol levels >200 mg/dL (57.1% vs. 41.8%, P = .03; unadjusted model), but the difference was not significant after adjusting for waist-to-hip ratio ( table 3 ).
Risks among HIV-infected patients without lipodystrophy . The proportions of subjects with insulin levels >18 µU/mL, 2-h glucose levels >140 mg/dL, cholesterol levels >200 mg/dL, triglyceride levels >200 mg/dL, and LDL levels >160 mg/dL were not different for the group of HIV-infected patients without lipodystrophy and the matched Framingham control subjects ( table 3 ). None of the HIV-infected patients without lipodystrophy or matched Framingham control subjects had a 2-h glucose level >200 mg/dL. The risk of impaired glucose tolerance (5.6% vs. 35.2%; OR, 0.11 [95% CI, 0.01–0.58]; P = .04), an insulin level >18 µU/mL (3.5% vs. 26.5%; OR, 0.10 [95% CI, 0.01–0.52]; P = .03), a cholesterol level >200 mg/dL (16.7% vs. 57.1%; OR, 0.15 [95% CI, 0.05–0.41]; P = .0005), and a triglyceride level >200 mg/dL (13.3% vs. 57.1%; OR, 0.12 [95% CI, 0.03–0.33]; P = .0002) was considerably lower among the HIV-infected patients without lipodystrophy compared with the patients with lipodystrophy. The risk of an LDL level >160 mg/dL was also lower (OR 0.27 [95% CI, 0.04–1.07]; P = .10) among patients without lipodystrophy than among patients with lipodystrophy. The proportions of subjects with HDL levels <35 mg/dL, systolic blood pressure >140 mm Hg, and diastolic blood pressure >90 mm Hg were not different between HIV-infected patients with and those without lipodystrophy.
Comparison of clinical variables by sex . Comparison of clinical variables between HIV-infected men and women with lipodystrophy demonstrated no differences in age, BMI, waist-to-hip ratio, and insulin, glucose, and lipid levels ( table 4 ). In contrast, significant differences in waist-to-hip ratio and HDL level were observed in the comparison of men and women in the Framingham Offspring Study ( table 4 ). The elevated waist-to-hip ratio among women with lipodystrophy is primarily due to a markedly greater waist circumference than among control subjects (96 ± 3 vs. 83 ± 1 cm, case patients vs. female Framingham control subjects, P < .001). Metabolic parameters, with the exception of HDL, remained significantly different between male case patients and matched male Framingham control subjects after adjusting for waist-to-hip ratio ( table 4 ). In contrast, differences in triglyceride and 2-h glucose levels, but not fasting insulin, 2-h insulin, or cholesterol levels, remained significant after adjusting for waist-to-hip ratio in the comparison of female case patients and matched female Framingham control subjects ( table 4 ).
Comparison of clinical variables by treatment status and fat redistribution pattern . Seven (10%) case patients were PI naive, 50 (70%) were currently receiving a PI, and 14 (20%) had a history of previous PI exposure. No differences in levels of glucose, insulin, cholesterol, and HDL were observed by PI treatment status. PI-naive case patients had lower LDL cholesterol concentrations than patients who were currently on PI therapy (95 ± 17 vs. 139 ± 7, P < .05). In a multivariate regression model predicting fasting insulin with age, sex, waist-to-hip ratio, BMI, current PI use, total duration of PI exposure, and duration of NRTI exposure, only duration of NRTI exposure ( P = .02) was significant. The fasting insulin level increased by 1.6 μU/mL for each year of NRTI exposure.
Metabolic parameters were also compared between case patients on the basis of fat redistribution patterns ( table 5 ). BMI and waist-to-hip ratio were greater among patients with lipohypertrophy (18%) and mixed lipodystrophy (66%) than among patients with lipoatrophy (15%). Patients with lipoatrophy had significantly lower fasting and 2-h OGTT insulin levels than did patients with mixed lipodystrophy. Patients with lipoatrophy had higher triglyceride concentrations and lower HDL levels than patients with mixed lipodystrophy. However, only the difference in fasting insulin level between patients with lipoatrophy and those with mixed lipodystrophy remained significant in a linear regression model that adjusted for differences in BMI (effect of group, lipoatrophy vs. mixed lipodystrophy, P = .01; effect of BMI P = .005).
In this study, we characterized metabolic abnormalities and CVD risk parameters in men and women with HIV lipodystrophy by contrasting their clinical characteristics with those of healthy participants from the Framingham Offspring Study cohort [ 15 , 16 ]. We were thus able to determine increased CVD risk for patients with lipodystrophy compared with the expected risk for healthy individuals of similar age and weight. We also compared CVD risk parameters in HIV-infected patients without lipodystrophy to matched Framingham control subjects to further distinguish the RR associated with the syndrome versus HIV infection alone. Our data demonstrate a clustering of metabolic abnormalities among HIV-infected patients with fat redistribution, characterized by hyperinsulinemia, hypercholesterolemia, hypertriglyceridemia, low levels of HDL, and truncal adiposity. These metabolic abnormalities indicate a significant insulin resistance syndrome [ 21 ] in HIV-infected patients with fat redistribution. Insulin resistance, dyslipidemia, truncal adiposity, and increased diastolic blood pressure are known to increase cardiovascular risk in patients who are not infected with HIV [ 22–24 ] and may similarly predispose HIV-infected patients with fat redistribution to accelerated CVD [ 25 ].
A number of previous studies have investigated metabolic and body composition parameters in patients with the HIV lipodystrophy syndrome. Carr et al. [ 9 ] demonstrated moderate hyperinsulinemia in patients selected for PI use in comparison with non-PI-treated patients and age- and BMI-matched control subjects. Our data suggest that fasting and postglucose-challenge hyperinsulinemia is much more marked in patients who present with clinical lipodystrophy. Walli et al. [ 26 ] demonstrated reduced insulin sensitivity in PI-treated patients but did not exclude patients with known diabetes or receiving diabetes therapy and did not select patients on the basis of clinical symptoms. More recently, Saint-Marc et al. [ 7 ] investigated 43 HIV-infected patients who were receiving dual NRTI therapy and did not show increased insulin levels. Abnormal lipid levels were shown in these studies, as well as in others [ 5 , 27–29 ], which have all selected patients on the basis of therapy, rather than by clinical manifestations.
Several studies have investigated body composition and metabolic abnormalities in HIV-infected women. Gervasoni et al. [ 8 ] evaluated a large number of HIV-infected women who were receiving dual NRTI therapy, only 10.5% of whom demonstrated fat redistribution. Dong et al. [ 1 ] investigated 21 women receiving HAART who were concerned about changes in body habitus, compared with patients receiving HAART who did not experience changes. No difference in fasting insulin levels was reported between groups; however, 29% of the patients without lipodystrophy had reported increased abdominal size. In these previous reports, antiretroviral therapy was a selection criteria for the investigation, and many patients did not exhibit symptoms of lipodystrophy.
Insulin resistance was demonstrated in the HIV-infected patients with lipodystrophy by significant fasting hyperinsulinemia in the setting of a normal fasting blood glucose level. Fasting insulin levels were not different between male and female case patients and were higher in each group than in Framingham control subjects. In 25% of case patients the fasting insulin level was above the 90th percentile expected for the matched control subjects. HOMA IR, an index of insulin resistance that incorporates the fasting insulin and glucose levels [ 14 ], was also significantly higher in patients with lipodystrophy. The fasting insulin level and homeostatic model are good single-sample epidemiologic markers for insulin resistance in patients without diabetes [ 24 ] that have been shown to predict elevated CVD risk [ 24 ] and the development of diabetes mellitus [ 30 ] and strongly correlate with results of euglycemic hyperinsulinemic clamp testing [ 31 ]. Analysis of our data suggests at least adequate pancreatic β cell function by use of the homeostatic model [ 32 ]. Further studies are necessary to better define the pathophysiologic mechanism of insulin resistance in HIV lipodystrophy, using direct methods to determine relative hepatic and peripheral insulin resistance.
Seven percent of the previously undiagnosed population we studied who had clinical evidence of fat redistribution had diabetes mellitus, and 35% demonstrated impaired glucose tolerance on the basis of WHO criteria [ 13 ]. Therefore, this population of HIV-infected patients is at high risk of developing clinically significant glucose abnormalities in addition to hyperinsulinemia. In contrast, rates of diabetes mellitus, impaired glucose tolerance, and significant fasting hyperinsulinemia were not higher among the patients without lipodystrophy than among the matched Framingham control subjects, which suggests that these metabolic disturbances are not simply a function of HIV infection. Glucose intolerance is associated with increased cardiovascular risk in non-HIV-infected populations [ 33 ] and may similarly increase cardiovascular risk in HIV-infected patients with fat redistribution. In our study, fasting glucose levels were not increased among the HIV-infected patients with fat redistribution. Analysis of these data suggests that, in this population, a fasting glucose level is unlikely to be abnormal and is not a good screening test of glucose abnormalities. Glucose response to standard oral glucose challenge is a more sensitive test to detect glucose abnormalities in HIV-infected patients with lipodystrophy.
The mechanism of insulin resistance among HIV-infected patients with fat redistribution is not known. Possible mechanisms include the following: direct metabolic effects of antiretroviral therapies [ 34 , 35 ]; metabolic dysfunction secondary to HIV disease itself, related cytokine and hormonal abnormalities, or both; an interaction between effects of antiviral therapy; and HIV disease. Fasting and 2-h insulin levels were uniformly increased among PI-naive patients as well as patients who were current and past recipients of PI therapy. Furthermore, duration of NRTI exposure and not PI exposure predicted fasting hyperinsulinemia in a multivariate regression analysis that controlled for age, sex, BMI, and waist-to-hip ratio. Although the differences between patients with and without lipodystrophy may be partially attributable to differences in exposure to antiretroviral therapy, analysis of these data suggests a complex pathophysiologic mechanism for the metabolic abnormalities associated with fat redistribution in HIV-infected patients, which is not due exclusively to the effects of PI therapy.
Our data demonstrate that hypertriglyceridemia, hypercholesterolemia, and low HDL levels are common in this population of patients. The ratio of cholesterol to HDL levels, an index of cardiovascular risk, was also significantly increased among the HIV-infected patients with lipodystrophy. Separate analyses showed significant differences in all parameters within the groups of male and female patients. These data indicate significant dyslipidemia in HIV-infected patients with fat redistribution. Hypertriglyceridemia is the most pronounced lipid abnormality in such patients; in our study, >50% of patients had triglyceride levels >200 mg/dL. Previous studies have identified hypertriglyceridemia in HIV-infected patients, which results from increased production and decreased clearance of very low-density lipoprotein by the liver [ 36 ]. In addition, recent studies have suggested that PIs have effects that specifically increase triglyceride levels.
Increased levels of cholesterol, LDL, and triglyceride were much less common among the HIV-infected patients without lipodystrophy than among the HIV-infected patients with lipodystrophy, but were not different in comparison with Framingham control subjects. HDL levels were low among the patients without lipodystrophy, which suggests that low HDL levels are a feature of HIV infection, independent of the lipodystrophy syndrome. These data are consistent with reports of low HDL levels in HIV-infected patients that were written before the recognition of the lipodystrophy syndrome [ 37 ]. Biochemical assays on case and control subjects were not run simultaneously because of the prior collection of samples in the Framingham Study. However, the same methods were used, and it is unlikely that technical variability would account for the findings in this study.
An important aim of this study was to investigate the sex-specific characteristics of the lipodystrophy syndrome. Stratification by sex demonstrated a similar pattern of fat redistribution, insulin resistance, and dyslipidemia among HIV-infected men and women with lipodystrophy. The significant and expected sex difference in waist-to-hip ratio seen in the control population was not seen in the comparison between male and female case patients. Waist-to-hip ratio and waist circumference were strikingly elevated in HIV-infected women with lipodystrophy. In the study by Rexrode et al. [ 38 ], a waist-to-hip ratio of ⩾0.88 conferred an increased CVD risk ratio of 4.47 in women <60 years old. The mean waist-to-hip ratio of women with HIV and fat redistribution was 0.96. Most of the metabolic abnormalities in female case patients were a function of fat redistribution, because few differences between HIV-infected women and control subjects persisted after adjusting for waist-to-hip ratio. Female patients affected by HIV lipodystrophy thus lose the expected advantage in fat distribution with respect to similarly aged healthy female subjects, and they develop a more android body habitus, which may confer significant excess cardiovascular risk [ 38 ].
Fat redistribution in patients with lipodystrophy can also be manifested by peripheral fat wasting [ 5 ], as demonstrated by the reduced midthigh circumference in the population we studied. We simultaneously controlled for waist, midarm, hip, and midthigh circumference in a regression model stratified by subject group in order to assess independent effects of peripheral fat loss on the fasting insulin level in patients with lipodystrophy in comparison with Framingham control subjects. Among the HIV lipodystrophy patients, but not the Framingham control subjects, midthigh circumference as well as waist circumference were significant predictors of fasting insulin levels. Peripheral fat loss may also be an important predictor of insulin resistance independent of central fat accumulation in HIV lipodystrophy.
In a subanalysis comparing patients with lipoatrophy to patients with lipohypertrophy and mixed lipodystrophy, there was less evidence of insulin resistance among patients with lipoatrophy alone. However, patients with lipoatrophy did have increased triglyceride concentrations and lower HDL levels, and most of the differences between fat redistribution subgroups was attributable to differences in BMI. One limitation of our study is that anthropometric measurements do not assess visceral and subcutaneous abdominal fat, which may be affected in patients with lipodystrophy [ 3 ] and may contribute to insulin resistance [ 39 , 40 ]. Further studies are necessary to investigate the metabolic affects of lipoatrophy in HIV-infected patients with fat redistribution and to determine whether lipoatrophy is clinically distinct or linked pathogenetically to the accumulation of truncal fat in such patients.
This study has significant clinical implications for the management of HIV-infected patients. HIV-infected patients with evidence of fat redistribution, including those who are PI-naive and who have peripheral fat loss, are at high risk for metabolic abnormalities, and their fasting lipid levels and glucose tolerance should be tested. Treatment of significant dyslipidemia and diabetes mellitus are indicated. Therapy for insulin resistance without overt glucose abnormalities may confer a cardiovascular benefit on such patients, but this remains investigational at the current time. Elucidation of the mechanisms by which HIV infection or HIV-related treatment strategies increase insulin resistance in this population may suggest novel therapies to improve insulin sensitivity in patients with HIV infection or in noninfected subjects at risk for complications of insulin resistance. Furthermore, prospective studies of the risk factors for the development of fat redistribution and associated metabolic abnormalities are needed.
In this study, we demonstrate a constellation of metabolic abnormalities, including hyperinsulinemia and dyslipidemia, suggestive of a significant insulin resistance syndrome among HIV-infected patients with fat redistribution. In comparison, cardiovascular risk parameters are not substantially increased in HIV-infected patients without clinical evidence of fat redistribution. Although affected patients have a significant increase in waist-to-hip ratio, the observed metabolic abnormalities persist after adjustment for increased waist-to-hip ratio in male patients, and to a lesser degree in female patients, and may result from loss of peripheral or subcutaneous fat. Diabetes mellitus and accelerated CVD may become important clinical problems for HIV-infected patients with fat redistribution and elevated cardiac risk factors.
We thank the nursing and dietary staff of the Massachusetts Institute of Technology General Clinical Research Center for their dedicated patient care; Gregory Neubauer for his help in the performance of the radioimmunoassays; Dr. Anne Klibanski and David Schoenfeld for helpful suggestions with the article; Drs. David Nathan and Daniel Singer for assistance in the collection of insulin levels in the Framingham Offspring Study; and Dr. Martin Hirsch, for his continued support.
- body mass index procedure
- waist-hip ratio
- insulin resistance
- diabetes mellitus
- cardiovascular disease risk factors
- glucose intolerance
- high density lipoproteins
- fasting blood glucose measurement
- hiv infection
- framingham offspring study
- overnight fasting insulin measurement