点击显示 收起
1 From the Baylor College of Medicine and Houston AIDS Research Team (HART), Houston, TX (FV); the Harlem Hospital and Columbia University, New York, NY (SR and WES); the School of Public Health, University of Minnesota, Minneapolis, MN (CMM and GB); the Body Composition Unit of St. Luke’s–Roosevelt Hospital, Columbia University, New York, NY (JW and DK); the George Washington University and Veterans Affairs Medical Center, Washington, DC (CLG); the Denver Community Programs for Clinical Research on AIDS, Denver, CO (JS); the University of California, San Francisco, and San Francisco Veterans Affairs Medical Center, San Francisco, CA (CG); and St Vincent’s Hospital, Sydney, Australia (AC)
2 Supported by grants no. 5U0142170-1 and 5U0146362-03 from the National Institutes of Health. 3 Reprints not available. Address correspondence to F Visnegarwala, Section of Infectious Diseases, Department of Medicine, Baylor College of Medicine, One Baylor Plaza, BCM 465 EC, Houston, TX 77030. E-mail: fehmidav{at}bcm.tmc.edu
ABSTRACT
Background: Data on associations of body composition with HIV disease characteristics are limited.
Objective: We compared sex-specific associations between HIV disease characteristics and body composition in an racially-ethnically diverse cohort of antiretroviral-naive patients.
Design: The study was a cross-sectional analysis of participants enrolled in a metabolic substudy of a multicenter trial. Regional fat was measured, and total body fat (TBF) was derived by using the Durnin-Womersley formula (DWF) and bioelectrical impedance analysis (BIA). Body cell mass (BCM) was measured by BIA.
Results: Among 422 participants, 22% were women, 60% were African American, and 36% had prior AIDS-defining illnesses. Mean (±SD) age was 38.2 ± 9.6 y, CD4+ count was 215 ± 184 cells/mm3, and HIV RNA log10 was 5.0 ± 0.8 copies/mL. On multivariate analysis, women with AIDS-defining illness had significantly (P < 0.005) lower regional body fat and TBF (BIA: –9.5 kg; DWF: –7.3 kg) but nonsignificantly lower BCM (–1.3 kg) than did women without such illnesses, whereas men with AIDS-defining illness had significantly (P < 0.005) lower BCM (–1.7 kg) but nonsignificantly lower TBF (BIA: –1.3 kg; DWF: –1.83 kg) than did men without such illnesses (P < 0.05 for sex differences in TBF). Significant negative associations of HIV RNA with BCM (–0.9 kg/log RNA; P = 0.03), TBF by BIA (–1.4 kg/log RNA; P = 0.05) and by DWF (–1.6 kg/log RNA; P = 0.01), and regional fat were observed in men only.
Conclusions: The effect of prior AIDS illness on body fat differed significantly between the sexes: women with prior AIDS-defining illness had significantly less fat than did women without such illnesses. An independent effect of HIV viremia on BCM and fat was seen in men. These distinctions may be due to inherent biological differences between the sexes.
Key Words: HIV infection • antiretroviral-naive patients • sex differences • body composition • anthropometric analysis • bioelectrical impedance
INTRODUCTION
Malnutrition characterized by decreases in body weight, body cell mass (BCM), and total body fat (TBF) is a central feature of advanced HIV infection and is associated with more rapid progression to AIDS or death (1, 2). Weight loss and depletion of BCM and body fat may precede the progression to AIDS, even in adults with normal CD4+ lymphocyte counts, which suggests that this depletion is related to HIV infection, independent of the additive effects of opportunistic illnesses (3).
There are few data on the effects of HIV viremia, the associated immune dysregulation, and the effect of opportunistic infections on overall body composition, particularly TBF and regional body fat, in the absence of antiretroviral therapy. The studies of body composition in the era before highly active antiretroviral therapy (HAART) reported a disproportionate loss of BCM and a greater preservation of TBF in white men with advanced HIV disease or AIDS (1). The data on women and nonwhite ethnic groups are limited. With the widespread use of HAART, a syndrome of HIV lipodystrophy has been described that is characterized by changes in body composition, particularly the loss of TBF and regional body fat but the preservation of BCM (4, 5).
We analyzed the sex-specific relation of total and regional body-composition measurements to the stage of HIV-1 disease, as defined by a history of AIDS-defining illness, CD4+ lymphocyte counts, and HIV RNA concentrations considered simultaneously, in antiretroviral-naive participants from diverse ethnic groups.
SUBJECTS AND METHODS
Patient population
The Community Programs for Clinical Research on AIDS (CPCRA) initiated a metabolic study (CPCRA 062) to evaluate the effects of 3 different antiretroviral treatment strategies on selected metabolic measures and on body composition in antiretroviral-naive patients. Baseline data were analyzed from participants enrolled in this study.
All participants gave written informed consent. The protocol was approved by the institutional review board of each institution involved in CPCRA 062.
Participants were enrolled if they had HIV-1 infection [positive results on enzyme-linked immunosorbent assay (ELISA) and Western blot or measurable plasma HIV-1 RNA], were 13 y old, and, if sexually active, were willing to use a barrier method of birth control throughout the course of the study. Patients were excluded if they were pregnant or breastfeeding, had any prior use of protease inhibitors or nonnucleoside reverse transcriptase inhibitors, or had a cumulative total of >4 wk of nucleoside reverse transcriptase inhibitor (NRTI) use or >1 wk of lamivudine (3TC) use.
Baseline assessments
Information on demographic characteristics, concomitant medications, prior AIDS-defining illnesses, CD4+ lymphocyte counts, and HIV RNA concentrations was obtained before study entry. Diagnoses of prior AIDS were based on the Centers for Disease Control and Prevention’s 1993 surveillance case definition of AIDS, which included the diagnoses of illnesses associated with severe immunosuppression in HIV patients, but they excluded the CD4+ cell count criteria. CD4+ lymphocyte counts were determined by local laboratories. The plasma HIV RNA was measured by using an HIV-1 polymerase chain reaction assay (Roche Amplicor version 1.0; Roche Diagnostics, Nutley, NJ) at a central laboratory.
Anthropometric and body-composition measurements
Height and weight were measured according to a standard procedure (6). BCM and TBF were calculated from bioelectrical impedance analysis (BIA) measurements obtained by using a BIA-10IQ analyzer (RJL Systems Inc, Detroit, MI). The formulas for computing BCM and TBF were provided by the manufacturer and described by Kotler et al (7). If the formula for TBF yielded a negative value (as it did in 5 cases), zero was used as the value for TBF in the computations.
Body circumferences (including midarm, waist, hip, and midthigh) were obtained by using a Dritz sewing tape as described by Lohman et al (6). Skinfold- thickness measurements (including triceps, subscapular, suprascapular, abdomen, and midthigh) were obtained by using a Lange caliper. A detailed description of the skinfold-thickness measurements for this study was given by Wang et al (8, 9).
The measurements of the skinfold thicknesses of the triceps, abdomen, and midthigh and of the circumference of the midarm, waist, and midthigh were used to calculate the subcutaneous fat compartments at these sites (8-10). Waist circumference and waist/hip ratio (WHR) were used as surrogates for visceral fat compartments (11, 12). In addition to BIA calculation, TBF was calculated with the Durnin-Womersley formula (DWF) (13) by using the triceps and subscapular skinfold-thickness measurements. All body-composition measurements were made by study nurses who received training and certification from a single trainer both before the study and annually thereafter, as described elsewhere (9).
Statistical analysis
All statistical analyses were performed with SAS software (version 8; SAS Institute, Cary, NC). Comparisons of the sex-specific means of baseline characteristics were performed by using t tests, and similar comparisons of percentages were done by using chi-square tests. Overall comparisons of race-ethnicity and sex-specific means of the measures of body mass were made by using the 2-way analysis of variance and the F test. Pairwise comparisons of the adjusted race-ethnicity and sex-specific means were performed by using t tests and their nominal significance probabilities. The adjusted means for an unbalanced design were obtained by using the linear model. To address the multiple comparison problems in testing racial-ethnic differences, Bonferroni’s procedure was used; in that case, the nominal significance of the t test must be <0.016 for the racial-ethnic difference in means to be significant at P < 0.5.
Sex-specific scatter plots and a smoothed regression line for each of the measures of body composition versus CD4+ lymphocyte counts and HIV RNA log10 were made for participants with and without prior AIDS-defining illnesses. A multivariate regression model for each measure of body composition was fitted with the following variables: age in 10-y groups, race-ethnicity (whites and nonwhites), smoking, history of intravenous drug use (IDU), sex, the 3 measures of HIV disease, and the interaction of each of the HIV disease terms with gender.
RESULTS
Baseline demographic and HIV disease characteristics
From 19 August 1999 to 1 January 2002, 422 participants were enrolled at 13 CPCRA units comprising 46 clinics and physicians’ offices throughout the United States (Table 1). Of these 422, 92 (22%) were women. Mean age was 38.2 ± 9.6 y; the frequency of prior IDU was 13.8%, the frequency of prior AIDS-defining illnesses was 36.0%, and the baseline mean CD4+ lymphocyte count (215 ± 184 cells/mm3) did not differ between women and men. However, significantly more women than men were African American (85% and 54%, respectively; P < 0.001), and significantly fewer women than men were white (3% and 34%, respectively; P < 0.05), but equivalent proportions of men and women were Latino (9% and 11%, respectively). Race or ethnicity was self-designated by each subject. Baseline HIV RNA concentrations were significantly lower in women than in men (4.8 ± 0.9 and 5.0 ± 0.7 HIV RNA log10 copies/mL, respectively; P = 0.05).
View this table:
TABLE 1. Baseline characteristics of the cohort by sex1
Total and regional body composition
The mean (±SD) measurements of body mass for the combinations of race-ethnicity and sex, including the results of the comparisons of the race-ethnicity and sex means, are shown in Table 2. Table 2 documents the greater heterogeneity of the measures of TBF, skinfold thicknesses, and skinfold fat areas as compared with BCM and other measures of body mass such as weight and the circumferences. In the case of the measures of fat, the CV is 0.5, whereas for the other measures of body mass, it is 0.2.
View this table:
TABLE 2. Race-ethnicity and sex-specific measures of body composition within groups by sex1
The significance of the adjusted race-ethnicity and sex means of the measures of body composition and of their comparisons is also shown in Table 2. Significant racial-ethnic differences were present for waist circumference, WHR, abdominal skinfold thickness, and midarm skinfold fat area. For these 4 measures, the means adjusted for the sex imbalance for Latinos are greater than those for the other 2 racial-ethnic groups. The means for the abdominal skinfold thickness and the midarm skinfold fat area are significant with Bonferroni’s procedure, as is the comparison of Latino with African American with respect to WHR. All of the differences of sex means adjusted for the racial-ethnic imbalances, except waist circumference, are significant.
Effect of prior AIDS-defining illness on total and regional body composition
The adjusted regression analyses between total and regional body-composition measurements versus prior AIDS-defining illnesses and the respective interaction of each factor with sex are shown in Table 3. Each model is adjusted for age, race-ethnicity (whites and nonwhites), cigarette smoking, prior IDU, hepatitis C, and the other 2 HIV disease characteristics. When adjusted for the other variables, the effect of a prior AIDS-defining illness in the men on measures of body composition was not significant, except for BCM (coefficient –1.7 kg; P < 0.005) and the suprascapular skinfold thickness (–2.17 mm; P = 0.02). However, weight (–10.59 kg; P < 0.005), BMI (–4.16; P < 0.005), TBF:BIA (–9.54 kg; P < 0.005), and TBF:DWF (–7.3 kg; P < 0.005) were significantly lower in the women with a prior AIDS-defining illness than in those without such an illness, whereas BCM (–1.3 kg; P = 0.23) did not differ significantly. In addition, all of the skinfold thicknesses and skinfold fat areas were significantly smaller in the women with a prior AIDS-defining illness than in those without such an illness (P < 0.05). Measurements of weight; BMI; 2 measures of TBF; hip circumference; WHR; triceps, subscapular, abdomen, and thigh skinfold thicknesses; and midarm and midthigh skinfold fat areas differed significantly (P 0.1) between the sexes by HIV disease status. Thus, women with a prior AIDS-defining illness had significantly less regional fat as measured by the 4 skinfold thicknesses and the 2 skinfold fat areas. For those sex-specific comparisons that were not significant at P 0.1, the effect of combined (both men and women) comparisons were significant at P < 0.05 for BCM and suprascapular skinfold thickness.
View this table:
TABLE 3. Adjusted regression analyses of body-composition measurements against HIV disease stage among antiretroviral-naive men and women1
The sex difference in the prevalence of prior AIDS-defining illnesses was not significant (P = 0.48, data not shown in tables). Most of the AIDS-defining illnesses in women were bacterial pneumonia; half of the 36 women with an AIDS-defining illness had bacterial pneumonia, whereas 22% (n = 26) of the 116 men with an AIDS-defining illness did so (P < 0.01). Most (52%) of the AIDS-defining illnesses in men were cases of Pneumocystis carinii pneumonia (PCP); the proportion in women was 39% (P = 0.18). In patients with a prior diagnosis of PCP, there was no significant sex difference in the time from diagnosis to enrollment in the study (P = 0.21) (data not shown).
Effect of HIV RNA concentration on total and regional body composition
Table 3 summarizes the regression analyses between total and regional body-composition measurements with HIV RNA concentration and its respective interaction with sex, adjusted for the other 2 HIV disease variables, and all the other variables listed above. It is interesting that, in men, all associations with body-composition measurements were negatively associated with HIV RNA. The associations with weight (–3.05 kg; P = 0.02), BCM (–0.92 kg; P = 0.02), TBF by DWF (TBF:DWF; –1.61 kg; P = 0.02), and suprascapular skinfold thickness (–1.40 mm; P = 0.04) were significant at P < 0.05, whereas TBF:BIA (–1.42 kg; P = 0.08), triceps (–1.12 mm; P = 0.07) and suprascapular (–1.2 mm; P = 0.06) skinfold thicknesses, and midarm fat area (–1.96 cm2; P = 0.07) were significant at P < 0.1. On the other hand, among women, 4 of the 5 associations of body mass—weight (increase of 1.15 kg; P = 0.63), BMI (increase of 0.82; P = 0.27), TBF:BIA (increase of 1.3 kg; P = 0.35), and TBF:DWF (increase of 0.78 kg; P = 0.5)—and 8 of the 14 associations of skinfold thickness and skinfold fat area were positively associated with HIV RNA concentrations. In women, none of these associations were statistically significant.
The differences in regression coefficients between men and women were considered significant at P < 0.1 in 5 [BMI, TBF (by BIA and DWF), waist, and suprascapular and subscapular skinfold thicknesses] of the 16 measures of body composition. For those sex-specific comparisons not significant at P 0.1, the effect of combined (both men and women) comparisons were significant at P < 0.05 for BCM.
Effect of CD4+ lymphocyte count on total and regional body composition
The adjusted analyses for CD4 cell count and body-composition measurements are presented in Table 3. Overall, the effect of CD4+ lymphocytes on weight, BMI, body fat, and cell mass was contrary to that of HIV RNA and prior AIDS-defining illness. However, as shown with prior AIDS-defining illnesses, this effect was more pronounced in women than in men, and none of the associations were significant in men. The following associations were significant for women: BMI (+0.87 kg; P = 0.02), TBF:BIA (+1.64 kg; P = 0.02), suprascapular skinfold thickness (+1.5 mm; P = 0.01), and midarm fat area (+2.0 cm2; P = 0.04).
The differences between the sexes were significant for TBF:BIA (P = 0.1), suprascapular skinfold thickness (P = 0.01), and skinfold fat area of the midarm (P = 0.06). For those sex-specific comparisons not significant at P 0.1, the effect of combined (both men and women) comparisons were significant at P < 0.05 for BMI, hip circumference, and waist fat area. The overall correlation between TBF:BIA and TBF:DWF was 0.90 (P = 0.0001) in men and 0.95 (P = 0.0001) in women (data not shown in tables).
DISCUSSION
To our knowledge, this is the first study that simultaneously evaluated the associations of HIV disease stage characteristics (CD4+ lymphocyte counts, HIV RNA concentrations and prior AIDS-defining illness) with total and regional body-composition values in a large, racially-ethnically diverse cohort of antiretroviral-naive patients and placed a particular emphasis on the sex differences. The main new finding was an independent effect of HIV RNA concentrations on body fat measurements in men. In addition, we observed a sex difference in the effect of AIDS-defining illnesses on body composition. Men with a prior AIDS-defining illness had significantly less BCM than did men without a prior AIDS-defining illness, whereas women with a prior AIDS-defining illness had significantly less fat than did women with such an illness, but the former had more preserved BCM than did the latter. The effects of a prior AIDS-defining illness on fat differed significantly between the sexes, whereas the effect on BCM did not.
The effect of race-ethnicity on body composition has been well described in HIV-uninfected populations (14-18). However, there are limited data in the HIV literature, and most studies have not included Latinos (19). In our cohort, the weight, BMI, BCM, and TBF did not vary by race-ethnicity. However, the Latinos had significantly greater central obesity as measured by skinfold thicknesses and fat areas in the abdomen than did the African Americans and whites. Similar distribution of increased central fat among Latinos has been described in HIV-uninfected populations (20, 21).
Our data suggest an indirect effect of HIV viremia leading to effects similar to AIDS wasting. The data on the independent effect of the HIV viral load on body composition are limited (22, 23). In most studies of AIDS wasting and those conducted in the pre-HAART era, patients had advanced disease. However, in our cohort, only one-third had an AIDS-defining illness, and more than half had CD4+ lymphocyte counts > 200 cells/mm3, which suggests that malnutrition due to the decreased oral intake and increased energy demands associated with an active opportunistic infection were not a contributing factor in most of our participants. A similar association between HIV RNA concentration and measures of body composition has been reported in 4 other studies. However, those studies were limited by the small number of participants, single-center design, and a predominant inclusion of white men (24-27). Moreover, only one of these studies adjusted the analysis for the CD4+ lymphocyte count as well as prior AIDS-defining illnesses. In contrast, the racial-ethnic diversity and the multicenter design of our study make our results more generalizable.
The mechanism of the effect of HIV viremia on the BCM and TBF is unclear. An increased energy expenditure, after correction for BMI, has been reported in participants with wasting syndrome (28, 29). It has been suggested that the effect may be due to an endogenous hypercortisolism related to the stimulation of glucocorticoid receptors that is mediated by the HIV virion-associated protein vpr (30). However, both central and peripheral fat losses in subjects in our cohort are unexplained by this mechanism. Higher HIV RNA concentrations are associated with higher cytokine concentrations, which have been related to the development of cachexia as described in both AIDS wasting and malignancy (31, 32). These factors may explain the association of higher HIV RNA concentrations with lower BCM. It is well known that lower body cell mass has been associated with lower bone mineral density (33, 34). Thus, our findings may partially explain the lower bone mineral density observed among subjects with advanced HIV disease than among their age- and sex-matched HIV-uninfected controls (35, 36).
Like others, we did not observe an independent effect of CD4+ lymphocyte count on body-composition measures. This lack of effect confirms the interrelation of the CD4+ count with other variables associated with HIV disease stage (24, 37).
The data on body composition among women in the pre-HAART era are limited (38). To our knowledge, this is the largest study of body-composition data among antiretroviral-naive women. We observed a dichotomy in the effect of the HIV RNA concentrations on fat in men and women. The association of HIV RNA concentrations with fat measures was in the opposite direction from that observed in men. These differences may be due to lower concentrations of HIV RNA at presentation in women than in men, a differential effect of HIV RNA due to the sex differences in body composition, a greater heterogeneity in fat measurements in women than in men, a lack of power in our study because there were fewer women than men, the different interrelation of HIV RNA with other HIV disease stage variables in women than in men, or racial-ethnic differences in the sex distribution of the cohort (ie, more female than male African Americans) or to all of these factors (17, 39-41). However, the latter is unlikely because Kotler et al did not find any effect of race on the sexual dimorphism in the body composition, which suggests that the effect we observed may not be related to the racial-ethnic composition of our cohort (19). In addition, the large size and the racial-ethnic diversity of our cohort confirm the findings of previous studies, which reported a greater loss of fat than of BCM in women with a prior AIDS-defining illness and AIDS wasting, whereas men have a greater loss of BCM than of fat (1, 19, 38, 42). The latter difference may be related to the greater premorbidity fat stores in women than in men, as well as to biological and hormonal differences that lead to a disproportionate reduction in body fat stores in women but a relative sparing of BCM, as described previously. In addition, the greater effect of a prior AIDS-defining illness on fat observed in women than in men may be due to delayed access to care in women with opportunistic infections as compared with men (43, 44), although we found no direct evidence for this from the available data on the time since PCP infection.
Our data confirm the associations of AIDS wasting and a lower BCM as measured by BIA reported earlier in the AIDS epidemic (1, 45, 46). The measurement of body fat by using BIA remains controversial. In cross-sectional analysis, the correlations between BIA, dual-energy X-ray absorptiometry (DXA), and skinfold thicknesses in measuring TBF are poor, and the results vary by the specific BIA formula used. In our study, the direction of association of the HIV disease stage with TBF:BIA was similar to that with TBF:DWF, and the correlations in both men and women were excellent, which suggests that these relatively easy-to-use techniques can be used for follow-up over time (47). Consistent with other studies, we found the sex differences in body fat to be greater by BIA than by DWF, despite the use of sex-specific formulas (48).
In conclusion, the effects of HIV viremia and prior AIDS-defining illnesses on body composition are different between men and women, perhaps because of the inherent biological differences between the sexes. Additional prospective and longitudinal studies are warranted to confirm our results. These data have important implications for our understanding of sex differences in the body-composition changes widely recognized as the HIV lipodystrophy syndrome. Characterization of the effects of HIV disease in antiretroviral-naive persons is essential to an understanding of the etiology of this syndrome, with its postulated complex interplay of host, disease, and treatment factors.
ACKNOWLEDGMENTS
We acknowledge all of the patients who participated in the study and all of the research nurses who carefully performed all of the anthropometric measurements.
FV was responsible for the concept, analysis, data interpretation, and drafting of the manuscript; CMM and GB had full access to all the data presented in this study, and they take responsibility for the integrity of the data and the accuracy of the data analyses. SSR, JW, GB, CLG, JS, and WES contributed to the study concept, design, and data collection. AC, DK, JW, GB, and CG contributed critical revision of the manuscript for intellectual content. FV and WES contributed to supervision and development. None of the authors had any personal or financial conflict of interest.
REFERENCES