Greater than 20.0% of American adults fulfill the National Institutes of Health (NIH) definition of obesity, with a body mass index (BMI) ≥30 kg/m2, whereas more than half of US adults are considered overweight, with a BMI ≥25 kg/m2.1-4 Beyond obesity, the metabolic syndrome, characterized by truncal obesity, hypertension, hypertriglyceridemia, insulin resistance, and concomitant increased cardiovascular risk, is extremely common in the United States, with an overall 22% prevalence.5 Studies that have directly addressed obesity in HIV-infected patients quantify the relatively slower progression to AIDS and the survival advantage afforded by elevated BMI in the era before successful chronic viral suppression.9-11 Now, in the highly active antiretroviral therapy (HAART) era, although some HIV-infected patients are still afflicted with wasting,12 we have observed that obesity is a significant issue in our HIV-infected population, especially among women.
The purpose of this cross-sectional study was to explore the prevalence of overweight and obesity in an HIV population receiving care at 4 affiliated hospitals in urban Philadelphia. To begin to evaluate the association between weight and metabolic abnormalities, we analyzed the relation between BMI and serum cholesterol, triglycerides, and glucose levels.
The University of Pennsylvania Center of AIDS Research Adult/Adolescent Database (Penn CFAR AAD) includes demographic, clinical, and HIV treatment-related data on patients receiving HIV clinical care at 1 of 5 sites: the Hospital of the University of Pennsylvania, Philadelphia Veterans Administration Medical Center, Pennsylvania Hospital, Presbyterian-University of Pennsylvania Medical Center, and the HIV Adolescent Clinic of Children's Hospital of Philadelphia. Patients were interviewed, and data were extracted from patients' charts at enrollment and subsequent 6-month intervals. The cohort mirrors the demographics of the reported AIDS cases in the Philadelphia area.
BMI, expressed as wt (kg)/[height (m)]2, was calculated for each patient using the most recent weight as of August 2003 and the height recorded at the initial CFAR encounter. Patients were characterized as obese (BMI ≥30 kg/m2) or overweight (BMI: 25-29.9 kg/m2) based on NIH definitions.4 HIV wasting was defined as a BMI ≤20 kg/m2, a value used in other epidemiologic studies in HIV-positive patients.13
The variables chosen for inclusion in the analysis were those known to correlate with BMI in the US population as determined by the Centers for Disease Control and Prevention's National Health and Nutritional Examination Survey (CDC-NHANES; available at: www.cdc.gov/nchs/nhanes.htm) and included sex, race, age, education, income, employment, and tobacco use. Other variables included were CD4 cell count, HIV viral load, receipt of HAART (defined as 3 or more HIV medications taken concomitantly and used together in common practice), receipt of a protease inhibitor (PI), being antiretroviral naive, and a history of intravenous (IVDU) and nonintravenous (NIVDU) illicit drug use. Data on sex, race, age, education, income, employment, tobacco use, drug use, HAART regimen, height, and weight were self-reported. Patients ≥18 years old were included. CD4 cell counts and viral loads within 60 days before or after the date of the reference weight were extracted from the patients' charts. If more than 1 value was recorded, that closest to the date was selected. Data on cholesterol, triglycerides, and glucose, obtained without regard to fasting state, were available for patients seen at 2 of 4 adult hospitals and are included in the analysis if obtained within 90 days of the weight used to calculate BMI.
The prevalences of wasting, normal weight, overweight, and obesity were determined as percentages of the cohort. In univariate analyses, odds ratios (ORs) and 95% confidence interval (CIs) for the association with obesity were calculated for all dichotomous variables. Age and BMI were assessed for correlation as continuous variables using the Pearson correlation coefficient.
Because of known differences in the rates and risk factors for obesity between men and women, all analyses were performed separately. We performed stratified univariate and multivariate analyses for men and women using all other variables. We used a forward stepwise multivariate binary regression model to determine which factors correlated with obesity and overweight. P > 0.1 was used to remove variables from the model.
Using available data on the prevalence of overweight and obesity in Philadelphia adults,14 we compared the age-adjusted race- and sex-stratified proportion of patients obese or overweight in our cohort with the proportion in the population at large using the χ2 test. For this analysis, age in the cohort was adjusted and stratified by sex to the Census 2000 data for the city of Philadelphia.
In a subset of patients in whom total, low-density lipoprotein (LDL), and high-density lipoprotein (HDL) cholesterol; triglycerides; and glucose levels were available within 90 days of the weight used to calculate BMI, we examined their relation with BMI. Because lipid and glucose levels were skewed in distribution, we used the Spearman rank test with a 2-tailed significance of P ≤ 0.05. Sex-stratified and non-sex-stratified tests were performed because of the small sample size of laboratory data available for women.
Statistical analyses were performed using SPSS 2000 for Windows (version 11.5.0; SPSS Inc., Chicago, IL) and MedCalc for Windows (version 7).
All subjects provided informed consent for participation in the database, and institutional review board (IRB) approval was obtained from the University of Pennsylvania and the Veterans Administration Medical Center.
The Penn CFAR-AAD included 1689 patients 18 years of age or older whose first encounter occurred before June 1, 2003, with BMI data available on 1669 patients. The database questionnaire was modified in October 2002 to include current antiretroviral medications; therefore, 1132 patients were included in the analyses of obesity and overweight and treatment.
Table 1 shows the overall characteristics of the study population. Most were male. Women were younger and more likely to be African American and had a higher median CD4 count than men (402 cells/μL, interquartile range [IQR]: 222-402 cells/μL vs. 378 cells/μL, IQR: 223-581 cells/μL; P < 0.001). BMI was normally distributed for each sex. The mean BMI in men was 24.9 kg/m2 (95% CI: 24.7-25.1 kg/m2), and in women, it was 27.5 kg/m2 (95% CI: 26.8-28.3 kg/m2). The prevalence of obesity in women was 28.3% versus 10.7% in men (OR = 3.3, 95% CI: 2.4-4.4; Fig. 1A). The combined prevalence of overweight and obesity in women was 58.2% versus 42.3% in men (OR = 1.9, 95% CI: 1.5-2.4). Obesity was most common in African American women, with a 30% prevalence, whereas the prevalence in African American men was 13.4% (OR = 2.8, 95% CI: 2.0-3.8). In contrast, the prevalence of overweight in African American women was 31.4%, and it was 29.7% in African American men (OR = 1.1, 95% CI: 0.81-1.5). Of non-African American women, 22.5% were obese compared with 7.5% of non-African American men (OR = 3.6, 95% CI: 1.9-6.6), whereas 25% were overweight compared with 33.9% of men (OR = 0.65, 95% CI: 0.38-1.1).
We performed stratified analyses to determine factors associated with obesity and overweight. In a univariate analysis, factors correlating with obesity and overweight in men were CD4 count ≥200 cells/μL (OR = 1.8, 95% CI: 1.4-2.5) and undetectable viral load (relative risk [RR] = 1.3, 95% CI: 1.02-1.6). Current cigarette smoking was protective (OR = 0.59, 95% CI: 0.47-0.74). Among women, being African American (OR = 1.8, 95% CI: 1.1-2.9) and having a CD4 count of ≥200 cells/μL (OR = 2.8, 95% CI: 1.6-4.9) were significantly associated with obesity, and cigarette smoking was also protective (OR = 0.65, 95% CI: 0.43-0.98). No significant correlation was found for employment status, education status, yearly income, history of IVDU or NIVDU, being antiretroviral naive, currently being on HAART, currently being on a PI, or, among women, HIV viral load. There was no significant correlation between BMI and age for men or women.
The multivariate models for obesity included potential confounding factors, such as age, employment status, education level, IVDU and NIVDU, naive status, current HAART use, current PI use, and HIV viral load in addition to factors found significant in the univariate models: sex, race, CD4 cell count, and smoking status. In the multivariate analysis in men, having a CD4 count ≥200 cells/μL remained associated with obesity and overweight (OR = 1.6, 95% CI: 1.04-2.40), whereas cigarette smoking remained protective (OR = 0.61, 95% CI: 0.44-0.83) (Table 2). In women, being African American (OR = 4.6, 95% CI: 1.9-11.1) and having a CD4 count ≥200 cells/μL (OR = 3.2, 95% CI: 1.1-8.8) were associated with obesity and overweight (see Table 2).
We compared the age-adjusted prevalence of obesity and overweight within our cohort with that of Philadelphia at large, stratified by sex, race, and age group (see Fig. 1B). In African American men, the city rate was 27% greater than in the HIV-infected cohort: 20% greater in non-African American men and 10.8% greater in African American women (all P < 0.001). In non-African American women, the rate was 48% compared with the 46% city rate (P > 0.5). In the age-stratified comparison with Philadelphia, the prevalence of obesity and overweight in the cohort was similar to that of the overall Philadelphia population in those aged 18 to 29 years (P = 0.43) and those aged 65 years or greater (P = 0.41). Obesity and overweight were less prevalent in the cohort than in the Philadelphia population in those aged 30 to 44 years and those aged 45 to 64 years (P = 0.0001 and P < 0.0001, respectively).
Data on lipids within 90 days of the weight used to calculate the BMI were available in 349 patients (20% of the cohort, 280 men and 69 women), and data on glucose in were available in 744 patients (44% of the cohort, 522 men and 222 women). Lipid data were more commonly available for patients who were nonsmokers on PIs having higher CD4+ cell counts and higher BMIs than the group as a whole (all P < 0.01). Glucose levels were checked more frequently in older women of low income (P < 0.01). There were weak positive correlations between BMI and total cholesterol, non-HDL cholesterol, triglycerides, and glucose overall (ρ = 0.14, 95% CI: 0.04-0.24; ρ = 0.17, 95% CI: 0.0-0.27; ρ = 0.14, 95% CI: 0.4-0.24; and ρ = 0.15, 95% CI: 0.08-0.22, respectively). Among men, there were also weak positive correlations between BMI and total cholesterol, non-HDL cholesterol, triglycerides, and glucose (ρ = 0.17, 95% CI: 0.06-0.28; ρ = 0.18, 95% CI: 0.07-0.29; ρ = 0.17, 95% CI: 0.06-0.28; and ρ = 0.16, 95% CI: 0.07-0.25, respectively). In women, probably because of the small sample size, there were no statistically significant correlations between BMI and cholesterol, non-HDL cholesterol, triglycerides, or glucose (ρ = −0.03, 95% CI: −0.27-0.21; ρ = 0.08, 95% CI: −0.16-0.32; ρ = 0.12, 95% CI: −0.08-0.32; and ρ = 0.13, 95% CI: −0.01-0.27, respectively).
Our data suggest that being overweight or obese is common in an urban HIV population, a reflection of the obesity epidemic in Philadelphia. Although the prevalences of overweight and obesity were not more common than in the general population, the prevalence of frank obesity was increased among HIV-infected women compared with men in our cross-sectional cohort. Female sex, CD4 cell count, and nonsmoking status correlated with obesity and overweight within the overall group. A CD4 count ≥200 cells/μL was the only factor significantly associated with obesity and overweight in men and women in the multivariate model. Among the subset of patients for whom lipid and glucose laboratory data were available, the positive correlation of BMI with cholesterol, triglyceride, and glucose levels suggests an elevated prevalence of the metabolic syndrome in our overweight population. Although not proving causation, because of the cross-sectional nature of the study design, these data suggest that overweight or obesity may contribute to the dyslipidemia and insulin resistance in our population of HIV-infected patients. Of note, 46% of our subjects were smokers. Given the potentially increased risk of vascular disease and malignancy in HIV, the high prevalence of smoking in the cohort may be of even more deleterious consequence than obesity.
Women far surpassed men in prevalence of obesity, even when factoring other risks, including the women's earlier stage of disease as assessed by CD4+ cell count, their lower overall income level, and their lower overall education level. Frank obesity was 2.6 times as common in women in comparison with the overall population of Philadelphia, in which obesity is 1.2 times as common in women. To our knowledge, this is the first study directly addressing risk factors for obesity in HIV-infected patients and directly comparing men and women. In young women, obesity can be associated with poor self-esteem and, subsequently, less negotiation of condom use,15 which could lead to an increased risk of HIV infection among overweight and obese women. This is a potential explanation for the HIV-infected cohort's dramatically wider gender disparity in obesity compared with the general population. Because of the association of AIDS with progressive and dramatic wasting, despite potential morbidities, some HIV-infected patients may favor maintaining elevated weight to serve as a protective “cushion” against future wasting or may believe that being overtly overweight masks their disease from friends or acquaintances. Further study may refine our understanding of this potential barrier to maintaining a healthy weight.
For the patients in whom specific antiretroviral regimens were known, no correlation was seen between BMI and being antiretroviral naive, being on HAART currently, or PI use. Because we did not examine longitudinal data in patients initiating treatment with specific agents, we were unable to assess whether certain antiretroviral regimens are associated with more weight gain than others. Although others have demonstrated changes in body composition in patients on PIs,16 increases in total fat mass specifically associated with certain drugs have not been published.
Given that our population is from a single US city with a high prevalence of obesity, our results may not be reflective of HIV patients throughout the United States, much less the world. A limitation in our analysis of the relation between BMI and lipid and glucose levels is our lack of data on the subjects' fasting state at the time of phlebotomy and additional anthropometric measurements; it is possible that the overweight and obese subjects might have been more likely to be nonfasting at phlebotomy than the others, which would influence their higher triglyceride and glucose levels. Longitudinal follow-up rather than cross-sectional data could elucidate associations not noted here for associations of obesity and overweight with particular antiretroviral regimens.
Because the major consequences of obesity take years to emerge, further longitudinal data are needed to quantify adverse sequelae within the HIV-infected population. Although BMI is an adequate tool to study obesity on the population-based level, differences in body habitus, particularly the presence of abdominal obesity, the degree of metabolic derangement, and an individual's family history, are better markers of risk in a given patient. In an obese patient with abnormal metabolic characteristics, lifestyle modification and weight loss interventions to achieve a healthy weight are likely to be beneficial. Optimum intervention strategies need to be established.
The authors thank the patients who were involved in this study.
1. Mokdad A, Ford ES, Bowman BA, et al. Prevalence of obesity
, diabetes, and obesity
-related health risk factors 2001. JAMA
2. Flegal KM, Carroll MD, Kuczmarski RJ, et al. Overweight
in the United States: prevalence and trends, 1960-1994. Int J Obes
3. Mokdad A, Bowman BA, Ford ES, et al. The continuing epidemics of obesity
and diabetes in the US. JAMA
4. National Institutes of Health. Clinical guidelines on the identification, evaluation, and treatment of overweight
in adults: the evidence report. Obes Res
. 1998;6(Suppl 2):51S-209S.
5. Ford ES, Giles WH, Dietz WH. Prevalence of the metabolic syndrome
among US adults. JAMA
6. Jones CY, Hogan JW, Snyder B, et al. Overweight
progression in women: associations HIV
disease progression and changes in BMI in women in the HIV
epidemiology research study cohort. Clin Infect Dis
7. Shor-Posner G, Campa A, Zhang G, et al. When obesity
is desirable: a longitudinal study of the Miami HIV
-1-infected drug abusers (MIDAS) cohort. J Acquir Immune Defic Syndr
8. Shuter J, Chang CJ, Klein R. Prevalence and predictive value of overweight
in an urban HIV
care clinic. J Acquir Immune Defic Syndr
9. Grinspoon S, Mulligan K. Weight loss and wasting in patients infected with human immunodeficiency virus. Clin Infect Dis
10. Jones CY, Hogan J, Snyder B, et al. Overweight
progression in women: associations between HIV
disease progression and changes in BMI in the HIV
Epidemiology Research Study (HERS) cohort [abstract 928]. Presented at: 10th Conference on Retroviruses and Opportunistic Infections CROI; 2003; Boston.
11. Household health survey data, Philadelphia Health Management Corporation's Community Health Database 2002. Southeastern Pennsylvania Household Health Survey. 2002.
12. Wingood GM, DiClemente RJ, Harrington K, et al. Body image and African American females' sexual health. J Womens Health Gend Based Med
13. McDermott A, Shevitz A, Knox T, et al. Effects of highly active antiretroviral therapy on fat, lean, and bone mass in HIV
-seropositive men and women. Am J Clin Nutr