Regional adipose tissue measured by MRI over 5 years in HIV-infected and control participants indicates persistence of HIV-associated lipoatrophy
17 July 2010 - Volume 24 - Issue 11 - p 1717-1726
"In multivariable models, duration of stavudine use was strongly associated with smaller absolute amounts of leg SAT (-5.6% per year of exposure to d4T, P < 0.0001). However, time off stavudine was associated with an estimated increase of only 1.1% per year in leg SAT, which did not reach statistical significance (95% CI -0.65, 2.9; P = 0.22). Likewise, little gain was associated with discontinuation of zidovudine or other antiretroviral drugs (data not shown)."
Grunfeld, Carl; Saag, Michael; Cofrancesco, Joseph Jr.; Lewis, Cora Elizabeth; Kronmal, Richard; Heymsfield, Steven; Tien, Phyllis C; Bacchetti, Peter; Shlipak, Michael; Scherzer, Rebecca; for the Study of Fat Redistribution and Metabolic Change in HIV Infection (FRAM)
aUniversity of California, San Francisco and VA Medical Center, San Francisco, California, USA
bUniversity of Alabama, Birmingham, Alabama, USA
cJohns Hopkins School of Medicine, Baltimore, Maryland, USA
dDivision of Preventive Medicine, University of Alabama, Birmingham, Alabama, USA
eUniversity of Washington, Seattle, Washington, USA
fMerck & Co., Rahway, New Jersey, USA
gUniversity of California, San Francisco, California, USA.
Objective: Peripheral fat loss and visceral fat gain have been reported in HIV infection. There are limited data on long-term change in adipose tissue in HIV-infected patients vs. controls. Therefore, we determined change in regional adipose tissue from baseline examination to 5 years later among participants in the study of Fat Redistribution and Metabolic Change in HIV Infection.
Methods: Regional adipose tissue volume was measured using MRI at both examinations in 477 HIV-infected and 214 control men and women. Lipoatrophy was defined as leg subcutaneous adipose tissue (SAT) below the cutoff point marking the lowest decile (10%) of controls at each examination.
Results: HIV-infected and control participants showed similar adipose tissue gains. In men, all SAT depots and visceral adipose tissue started lower and remained lower on average in HIV-infected vs. controls. In women, leg and arm SAT also started lower and remained lower in HIV-infected vs. controls. Mean leg SAT of HIV-infected men was 67% of control men at baseline and 65% at follow-up; for women 83% and 77%. At baseline, 48% of HIV-infected participants had lipoatrophy; on average those with baseline lipoatrophy gained 0.96L of leg SAT compared with 1.23L gain for controls in the lowest decile (P = 0.16). At follow-up, 53% of HIV-infected participants had lipoatrophy. In multivariable models, discontinuation of stavudine appeared to produce little gain in leg SAT (-1.1%/year).
Conclusion: HIV-infected participants did not substantially recover SAT compared with controls, although both showed average gains. HIV-associated lipoatrophy persisted after 5 years of follow-up.
The introduction of combination antiretroviral therapy (ART) was followed by changes in fat distribution and metabolic abnormalities in HIV-infected individuals that may contribute to cardiovascular disease . Both peripheral fat loss (lipoatrophy) and central fat gain (lipohypertrophy) have been reported in HIV infection. A major finding of the first study of Fat Redistribution and Metabolic Change in HIV infection (FRAM) was that HIV-infected participants differed from controls in terms of lipoatrophy, but not lipohypertrophy, as defined by self-report confirmed by examination, as well as by direct measurement using total body regional MRI [2,3]. Participants with clinical lipoatrophy had less subcutaneous adipose tissue (SAT) in all depots than those without clinical lipoatrophy and controls. HIV-infected men without clinical lipoatrophy also had less SAT than healthy controls. Less SAT was associated with use of specific antiretroviral drugs. In contrast, the amount of visceral adipose tissue (VAT) was independent of SAT and not associated with specific antiretroviral drugs.
Little is known about what happens to adipose tissue over the long term in HIV-infected patients. Previous studies in uninfected individuals have found that younger and middle-aged adults gain 0.5-1.0 kg/year . Total body fat is known to increase with age (until age 55 in men and age 65 in women) .
Many studies have assessed the effects of switching antiretroviral drug regimens on adipose tissue in HIV-infected participants, but most lasted 1 year or less and those studies varied in results [6,7]. When participants were switched off protease inhibitors, loss of adipose tissue often continued. When participants were switched off stavudine or thymidine analogues, increases in leg or limb adipose tissue were usually small. In the few studies lasting up to 96-144 weeks, in which participants were switched off nucleoside reverse transcriptase inhibitors (NRTIs), gain in fat was more consistently found, ranging from 10 to 42% [8-13]. However, none of these studies compared changes in fat with the changes found in healthy controls.
Furthermore, these studies mostly used dual-energy X-ray absorptiometry or computed tomography scans, hence were limited in the regional depots studied. In the large observational studies that studied fat changes in HIV infection and included controls, measures were limited to the use of anthropometry [14,15]. Thus, no large study has compared changes over several years in whole body regional adipose tissue depots including VAT in a nationally representative, multiethnic cohort of both HIV-infected participants, and controls. A primary aim of the second FRAM study was to determine the changes in SAT and VAT using whole body MRI in both HIV-infected and control participants after 5 years of follow-up . We hypothesized that a well treated cohort of HIV-infected participants in the HAART era would not resolve their HIV-associated lipoatrophy over 5 years. We also sought to investigate the associations of ART use and discontinuation with changes in fat.
Body composition by MRI at baseline and follow-up were available on 691 participants whose characteristics at the follow-up examination are presented in Table 1. HIV-infected and control participants were similar in age, height, and percentage of whites and African Americans, but HIV-infected participants were more often male (68 vs. 53%) because of the design of the control study. Control participants weighed more and accordingly had higher BMI.
Comparison of fat gains in HIV-infected and control participants
Mean adipose tissue volumes at baseline and follow-up examination for each regional depot in HIV-infected and control men, restricted to the age range of controls (ages 33-45 at baseline), are shown in Fig. 1a. Mean change in regional adipose tissue volumes between examinations is shown in Fig. 1b. Corresponding values for women are in Fig. 2a and b. Statistically significant gains were seen in nearly every depot in both HIV-infected and control men and women (with most P < 0.0001, Figs 1b and 2b).
HIV-infected men had lower mean adipose tissue volumes in all depots (SAT and VAT) at the first examination than control men (Fig. 1a), although the difference in arm SAT did not reach statistical significance (P = 0.14). Mean 5-year gains were somewhat smaller in all SAT depots for HIV-infected men than control men (Fig. 1b); the difference only reached statistical significance for arm SAT (P = 0.0002). Hence, at the FRAM2 examination, adipose tissue volumes remained lower in HIV-infected men relative to controls in a pattern similar to the first examination. For example, mean baseline leg SAT in HIV-infected men was 3.3L and increased to 3.7L, whereas baseline leg SAT in control men was 4.9L and increased to 5.7L. Thus, mean leg SAT in these HIV-infected men was 67% that of control men at the first FRAM examination and 65% of control men at the second examination (Fig. 1a). Similar results were found for each of the other SAT depots. Average VAT gain was similar in HIV-infected and control men. As a consequence, mean VAT also remained lower in HIV-infected men at the FRAM2 examination.
HIV-infected women also had lower leg and arm SAT than control women at the first examination (Fig. 2a). Five year gains in leg SAT were somewhat smaller and in arm SAT were slightly larger in HIV-infected compared with control women, although the differences did not reach statistical significance (Fig. 2b); HIV women continued to have lower leg and arm SAT at examination 2 than control women (Fig. 2a). For example, mean baseline leg SAT was 8.6L at both examinations in HIV-infected women, whereas leg SAT in control women was 10.4L at baseline examination and 11.2L at the second examination (Fig. 2a). Thus, mean leg SAT in HIV-infected women was 83% of controls at the first examination and decreased to 77% at the second examination.
At the baseline examination, upper trunk SAT and VAT were higher in HIV-infected women than in controls (mean upper trunk SAT = 6.5L vs. 5.2L, P = 0.019; VAT = 1.6L vs. 1.3L, P = 0.023). However, the gains in upper trunk SAT and VAT were somewhat greater in control women than HIV-infected women with the differential gain in upper trunk SAT reaching statistical significance (Fig. 2b). As a consequence, at the follow-up examination (Fig. 2a), HIV-infected women had similar mean upper trunk SAT and VAT to control women (upper trunk SAT: HIV 7.1L, Control 7.0L, P = 0.91; VAT: HIV 2.5L, control 2.5L; P = 0.83).
We examined pooled models to determine whether the differences in 5-year change in adipose tissue between HIV-infected and control participants remained after multivariable adjustment for demographic and lifestyle factors. After adjustment, HIV-infected participants showed smaller increases than controls in all regional SAT depots and VAT (Table 2), with differences reaching statistical significance for leg SAT (-0.81L, P = 0.009), arm SAT (-0.33L, P = 0.0002), and upper trunk SAT (-0.85L, P = 0.0005). We observed statistically significant HIV by sex interactions for lower trunk (P = 0.028) and arm SAT (P = 0.014). In fully adjusted models, HIV infection in men was associated with smaller increases for lower trunk and arm SAT relative to controls (-0.98L, P = 0.016; -0.45L, P < 0.0001, respectively), whereas the changes in HIV-infected women relative to controls were small and did not reach statistical significance (0.16, P = 0.77; -0.15, P = 0.25).
Although HIV-infected and control participants gained adipose tissue on average, we found that the prevalence of any leg SAT loss was higher in HIV than controls (35 vs. 27%, P = 0.0013). Additionally, we found that HIV-infected participants were more likely than controls to lose VAT (17 vs. 5.1%, P < 0.0001).
Change in adipose tissue relative to baseline adipose tissue
In these participants, 48% of the HIV-infected had lipoatrophy at the baseline examination, defined as having leg SAT below the cutoff of the lowest 10% of controls. The HIV-infected participants in that category gained an average of 0.96L between examinations, whereas the respective controls gained 1.23L (P = 0.16). At the second examination, 53% of HIV-infected had leg SAT below the 10% cutoff of second examination control values. Among those HIV-infected who had lipoatrophy by this cutoff at the baseline examination, 82% had lipoatrophy at follow-up. When we instead defined lipoatrophy using the clinical definition [2,3], 33% of the HIV-infected had lipoatrophy at the baseline examination, defined as concordance between participant report of fat loss and examination finding of less fat than a normal healthy person. The HIV-infected participants in that category gained even less, an average of 0.86L between examinations.
For leg SAT and VAT, we examined the relationship of quartiled baseline adipose tissue with change in adipose tissue, using quartile cut points based on control participants, to allow direct comparison of HIV-infected and control participants at similar baseline adipose tissue. For leg SAT (Fig. 3a), in HIV-infected participants, the median change in those with the least baseline leg SAT was similar to that of controls in the same quartile and was significantly less than controls in the third and fourth baseline quartiles. By contrast, for VAT (Fig. 3b), there was little difference between HIV-infected and control participants in change in VAT across baseline VAT quartiles.
Discontinuation of antiretroviral drugs and gain of leg subcutaneous adipose tissue in HIV-infected participants
We also evaluated repeated measures models of amount of leg SAT at the two examinations to determine whether any of the gains in leg SAT could be attributed to discontinuation of stavudine, the key antiretroviral drug responsible for leg SAT loss. Seventy-four percent of the HIV-infected participants had a history of stavudine use, but only 9.4% were still taking stavudine at the time of the second examination (see Supplemental Table, http://links.lww.com/QAD/A36). Average duration of exposure to stavudine was 3.6 ± 2.7 (SD) years. Among those who had discontinued stavudine (65%), duration off stavudine was 4.3 ± 2.6 years. Among the 9.4% of patients who were still on d4T at year 5, average gain in leg SAT was 0.53 ± 3.0L. Somewhat smaller gains were seen in past d4T users (0.34 ± 3.1L) and never d4T users (0.39 ± 3.3L), although the differences did not reach statistical significance.
In multivariable models, duration of stavudine use was strongly associated with smaller absolute amounts of leg SAT (-5.6% per year of exposure to d4T, P < 0.0001). However, time off stavudine was associated with an estimated increase of only 1.1% per year in leg SAT, which did not reach statistical significance (95% CI -0.65, 2.9; P = 0.22). Likewise, little gain was associated with discontinuation of zidovudine or other antiretroviral drugs (data not shown).
In this 5-year follow-up study of regional adipose tissue volumes, we found that HIV-infected participants in the FRAM study gained less or similar adipose tissue on the average than control participants. Many HIV-infected participants started at lower levels due to HIV-associated lipoatrophy, and the results indicate that those with lipoatrophy had little recovery regardless of whether they discontinued the implicated antiretroviral drugs.
The diminished recovery was most obvious in leg SAT, the depot most affected by HIV-associated lipoatrophy. Over 5 years, control men and women gained on average 0.7L of SAT in the leg, whereas HIV-infected men gained only 0.5L and HIV-infected women showed no average gain. Thus, HIV-infected participants had even less leg SAT on average relative to controls at the year 5 examination. In the Multicenter AIDS Cohort Study, thigh circumference increased more over 4 years in control than HIV-infected men . In the Women's Interagency HIV Study, over 4 years thigh circumference increased in control women and decreased in HIV-infected women .
Some studies of the change in adipose tissue after switching antiretroviral therapy report mean percentage change, with longer term switch studies finding 10-42% increases after switching off NRTI [8-13]. However, the absolute gains were typically small in these switch studies. In the study with the largest percentage gain (42%), mean leg fat by dual-energy X-ray absorptiometry increased from 1.6 to 2.2 kg . Indeed, it may be that past articles reporting percentage increases led readers to believe there was more significant recovery. In our study, the geometric mean relative change in leg SAT was 28% in HIV-infected men compared with 3.9% in controls, which might suggest that substantial recovery was made in HIV-infected men. The percentage changes should be contrasted with the baseline and absolute changes. HIV-infected men started with a baseline average leg SAT of 3.3L, whereas control men started with a baseline of 4.9L. Despite starting lower, in HIV-infected men the actual mean gain in leg SAT was 0.47L vs. 0.65L in controls, which clearly conveys that HIV-infected men were not catching up. HIV-infected women started with less leg SAT than control women (8.7L vs. 10.4L). On average HIV-infected women did not gain leg SAT (-0.03L), whereas control women did gain (0.69L). Furthermore, in the two adipose tissue depots that were higher in HIV-infected women than controls at baseline (upper trunk SAT and VAT), controls gained more adipose tissue; therefore, at 5-year follow-up these depots were virtually the same in HIV-infected and control women.
Many of the longer term switch studies did not require the presence of lipoatrophy for entry [11-13]. When we examined HIV-infected participants who had lipoatrophy defined as having leg SAT below the cutoff of the lowest control decile, the HIV-infected participants had a 40% gain in leg SAT. However, the HIV-infected gained 0.96L, whereas the comparative controls gained 1.23L; the net result is that there is no relative recovery from lipoatrophy when HIV-infected participants are compared to controls.
In further support of this concept, multivariable modeling found that little gain could be attributed to years off stavudine or other antiretroviral drugs including zidovudine. Yet in the HIV-infected cohort, duration of exposure to stavudine was associated with lower leg SAT at both FRAM examinations. In the WIHS study, women who discontinued stavudine had smaller annual decreases in thigh circumference than those who continued stavudine .
The results with leg SAT should be contrasted to those with VAT, in which HIV-infected and control men saw similar gains, whereas HIV-infected women no longer had greater VAT than control women. These data, along with our previous studies of SAT and VAT in HIV-infection [2,3], add additional support for the concept that these depots are independent and are not affected by the same factors.
There are several limitations to our study. We did not require lipoatrophy as an entry criterion, because our study was originally designed to be representative of the spectrum of HIV infection. However, we found little recovery in those HIV-infected who started with the lowest baseline SAT relative to controls. Likewise, several of the longer term switch studies did not require lipoatrophy as an entry criterion [11-13]. Although our HIV-infected participants span a wide age range (19-76 years at baseline), comparisons with controls were restricted to a narrower age range (baseline 33-45 years), which limits our ability to generalize these results to older populations. Our study is observational. Although a randomized study of discontinuation vs. continuation of relevant NRTI would be optimal, a 5-year randomized study is neither feasible nor ethical. The longest NRTI switch study was 144 weeks. Finally, we cannot rule out accelerated gain of SAT after an even greater time off the drugs associated with lipoatrophy. However, in multivariable analyses the average 3.6 years of exposure to stavudine was associated with dramatically lower amounts of SAT, whereas in those who had been off stavudine for an average of 4.3 years, little of the adipose tissue gain could be attributed to the time off drug.
A major strength of our study is the comparison of change in adipose tissue in HIV-infected persons over 5 years to that of controls to account for the normal changes in aging. Our controls come from the Visceral Fat and Metabolic Rate in Young Adults Study (VIM) substudy  of the CARDIA cohort, in which the average BMI is similar to that of the nationally representative sample of National Health and Nutrition Examination Survey (NHANES).
In conclusion, 5 years after the first examination in the FRAM study, gain in adipose tissue was similar in HIV-infected and control participants. HIV-infected participants had significant subcutaneous lipoatrophy at the baseline examination and, 5 years later, relative lipoatrophy persisted compared with controls, even in those who discontinued antiretroviral drugs associated with lipoatrophy, such as stavudine, during this period. These data must be considered when studying the potential mechanisms underlying HIV-associated lipoatrophy. It remains to be determined whether there was destruction of adipose cells and precursors as proposed by some , or whether other factors continue to contribute to the persistence of lipoatrophy in HIV infection. Finally, as the presence of lipoatrophy has been associated with adverse metabolic effects [31,32] and with depression , the long-term consequences and treatment of persistent lipoatrophy need additional study.
The FRAM study was designed to evaluate the prevalence and correlates of changes in fat distribution, insulin resistance, and dyslipidemia in a representative sample of HIV-infected participants and controls in the United States. The methods of the FRAM study have been described in detail previously .
HIV-infected participants were recruited from 16 HIV or infectious disease clinics or cohorts in 1999. Control participants were recruited from two centers from the Coronary Artery Risk Development in Young Adults (CARDIA) study . A follow-up FRAM examination was conducted approximately 5 years later. The institutional review boards at all sites approved the protocols for both FRAM examinations.
Retention outcomes for participants enrolled in the first examination have been reported . The second examination included 581 HIV-infected and 241 controls recruited from those seen at the first examination. We report here on the subset of 477 HIV-infected participants and 214 controls that had measurements of adipose tissue depots at both FRAM examinations. The time between the two adipose tissue measurements averaged 4.9 ± 0.76 (SD) years. Because a greater percentage of HIV-infected participants did not have measured MRI at both examinations, we adjusted analyses as described in the following sections to address the concern of selection bias.
Magnetic resonance imaging
Whole body MRI was performed to quantify regional and total adipose tissue . Body composition was measured with participants in the supine position, arms extended over head, and analyzed as described in detail elsewhere [2,3,16,19]. In brief, using the intervertebral space between the fourth and fifth lumbar vertebrae as origin, transverse images (10-mm slice thickness) were obtained every 40 mm from hand to foot. MRI scans were segmented using image analysis software (Tomovision Inc., Montreal, Canada). A single image reading center (IRC) was used to read all scans; imaging techniques and anatomical sites (based on bone landmarks) were identical between HIV-infected and control participants using a standardized acquisition protocol, with IRC performing site visits to ensure protocol adherence. Scans were sent to the IRC at the Obesity Research Center, St. Luke's Roosevelt Hospital, New York, NY, which calculated tissue areas (cm2) by summing specific tissue pixels, then multiplying by individual pixel surface area. Volume per slice (cm3) of each tissue was calculated by multiplying area by thickness. Volume of each tissue for the space between two consecutive slices was calculated via a mathematical algorithm . Volumes were normalized by dividing by height2 with summaries back-transformed to 1.75 m of height. We did not adjust to BMI, as BMI is influenced by the phenomenon being studied: quantity of fat. Anatomic sites considered in this analysis were leg, lower trunk (abdomen and back), upper trunk (chest and back), arm, and VAT. The difference in results on repeated measurements in the IRC average 0.7% for skeletal muscle and 1.1% for adipose tissue . We calculated geometric mean relative change as log (year 5 measure/baseline measure/years x 5), where years denotes time between baseline and year 5; results were back-transformed to produce percentage effects.
Height and weight were measured by standardized protocols. Standardized questionnaires were used to determine demographic characteristics; medical history; risk factors for HIV; adequate food intake; physical activity; and use of alcohol, tobacco, and illicit drugs [21,22]. Total physical activity score was calculated using the validated CARDIA activity questionnaire and was quartiled for analysis. Alcohol was categorized as drinks per week. Tobacco and illicit drug use were categorized as current, past, or never use. Research associates interviewed participants and reviewed medical charts regarding ART medication use. A diagnosis of AIDS was made by history of opportunistic infection or CD4 cell count less than 200 cells/µl.
HCV RNA testing was performed on frozen sera using the Bayer Versant 3.0 branched DNA (bDNA) assay (Leverkusen, Germany) in the entire cohort. CD4 lymphocyte count and percentage, HIV RNA level in HIV-infected participants, and other blood specimens were analyzed in a single centralized laboratory (Covance, Indianapolis, Indiana, USA).
For each adipose tissue depot, baseline and 5 year follow-up examination were compared using a paired t-test within HIV-infected and control groups separately. A two sample t-test was used for comparisons of changes from baseline to follow-up in HIV-infected vs. control participants. Analyses that compared characteristics of HIV-infected participants with controls excluded HIV-infected individuals with recent opportunistic infection and were restricted to those between the ages 33 and 45 at baseline (n = 294), because the control population did not include participants outside this age range.
We analyzed changes in MRI-measured adipose tissue using multivariable linear regression with the 5-year change in adipose tissue as the dependent variable. Interactions of HIV status with sex, ethnicity, and age were assessed and included if they reached statistical significance.
Models were constructed for each outcome using HIV status, demographics (age, sex, and race), and lifestyle factors as predictor variables. Age, sex, race, and time between MRI examinations were forced to be included in every model. The linearity assumption was tested for continuous measures by adding quadratic terms to the models and by examining generalized additive models . Confidence intervals were determined using the bias-corrected accelerated bootstrap method , with P-values defined as one minus the highest confidence level that still excluded zero; this was necessary because the error residuals appeared to be non-Gaussian.
Candidate lifestyle factors included physical activity, smoking, alcohol use, adequate food intake, and illicit drug use. Candidate HIV-related factors (tested only for analyses that did not include controls) included AIDS diagnosis (by CD4 or opportunistic infection), reported HIV duration, HIV RNA level (log10), current and nadir CD4 cell count (log2), hepatitis C infection (by virus detection), days since last opportunistic infection, recent opportunistic infection status (last 100 days), and HIV risk factors. In multivariable models controlling for the above factors, we evaluated ever use, duration on, and duration off each individual ART drug and ART class: NRTI, protease inhibitor, nonnucleoside reverse transcriptase inhibitor, and HAART as previously defined .
Multiple imputation utilizing the Markov chain Monte Carlo (MCMC) method for arbitrary missing data was used to impute missing covariate values . Because HIV-infected participants were missing MRI more often than controls, we adjusted estimates using an inverse probability weighting approach  by modeling the participant's probability of having nonmissing adipose tissue using logistic regression analysis. The inverse of this probability was then used as a weight (applied to persons with known adipose tissue) in multivariable regression analyses.
We defined lipoatrophy as leg SAT below the 10th percentile of the control participants at each examination, with men and women done separately, as in previous analyses [27,28]. As an alternative, we defined lipoatrophy using the clinical definition, which is concordance of report of change in fat with abnormal looking fat on examination [2,3]. The prevalence of lipoatrophy at baseline and follow-up was compared by using McNemar's test.
All analyses were conducted using the SAS system, version 9.2 (SAS Institute, Inc., Cary, North Carolina, USA).