Fat Distribution in Women With HIV Infection

Study of Fat Redistribution and Metabolic Change in HIV Infection (FRAM)


J Acquir Immune Defic Syndr. 2006;42(5):562-571. 

In This Article


Study Population

Institutional Review Boards at all 18 participating sites (16 HIV and 2 control sites) approved the study protocol and consent process. Between June 2000 and September 2002, 1480 participants (1183 HIV-infected subjects and 297 healthy controls) were enrolled in FRAM. Of the 1480 participants, 33% were women (350 HIV-infected women and 142 controls). The recruitment and data collection procedures for the entire cohort have been described elsewhere.[38] Briefly, HIV-infected participants were selected from coded lists of patients seen in 16 HIV or infectious disease clinics or cohorts during 1999. Each list was then randomly ordered and provided back to the sites. Each site then contacted the subjects in consecutive order. Of 1058 women active in HIV or infectious disease clinics or cohorts, 269 did not answer multiple contact attempts; 71 were dead; 156 were ineligible (institutionalized, n = 4; communication difficulty, n = 13; primary care provider did not give permission, n = 13; or excluded by study criteria, n = 126 [age <18 years; and factors that prevented imaging: weight >136 kg, height >6'5", metal in body, and/or claustrophobia]). The remaining 562 women were approached for informed consent. Of these, 124 declined participation and 88 did not return for data collection. Therefore, examination was completed on 350 women (62.3% of 562 contacted and eligible).

Control subjects were recruited from 2 centers of the Coronary Artery Risk Development in Young Adults (CARDIA) study.[39,40] CARDIA subjects were originally recruited as a community-based sample of healthy 18- to 30-year-old white and African-American men and women from 4 cities in 1986 for a longitudinal study of cardiovascular risk factors. FRAM enrolled 142 control women in 2000 to 2001 from a CARDIA ancillary study, the Visceral Fat and Metabolic Risk in Young Adults Study.[41]

Our analyses comparing HIV-infected women and controls in the same 33 to 45-year age range included 183 HIV-infected women; 6 HIV-infected women were excluded who had an opportunistic infection or malignancy within the same or previous calendar month as the examination (to remove acute changes in fat). Our analyses of HIV-associated factors including antiretroviral therapy in HIV-infected women included 338 women between the ages of 19 and 70 years (mean, 41.5 ± 8.2 years), after excluding 12 women with recent opportunistic infection or malignancy.

Clinical Assessments

HIV-infected and control participants underwent identical standardized questionnaires and examinations. Each participant was asked to report whether she had any change in fat over the last 5 years in the cheeks next to the nose, lateral aspect of the face, legs, arms, buttocks (peripheral sites), back, chest, neck, or abdomen, or change in waist size (central sites). If participants answered yes, they were asked whether the change was an increase or decrease and to grade it as mild, moderate, or severe. Participants could also answer "no change" or "don't know"; "don't know" represented less than 10% of responses and was pooled with "no change." Research associates who were centrally trained were asked to determine whether participants had more or less fat compared with normal healthy people in each of those body areas before the participant filled out the self-administered questionnaire of body fat changes (to keep the research associate blinded to what the participants had reported). For the participants who were determined to have either more or less fat than normal, research associates rated the amount of fat using criteria from the HIV Outpatient Study: mild = only seen if looked for, moderate = easily seen, or severe = obvious immediately.[20]

Clinical lipoatrophy at each anatomical site was defined as concordance between participant report of loss of fat (mild, moderate, or severe) and examination finding of less fat than a normal healthy person at the same anatomical site. Lipohypertrophy was defined as concordance of self-report of increased fat and examination finding of more fat than normal at the same site.

Physical activity, alcohol intake, smoking, illicit drug use, education, and adequacy of food intake were assessed by standardized instruments.[40,41,42,43].Menopausal status and history of gynecologic surgery were assessed by interview. Women who reported at least 12 months of amenorrhea or history of bilateral oophorectomy were categorized as being menopausal. Research associates interviewed HIV-infected participants and reviewed medical charts to determine the dates of use of individual antiretroviral medications. Blood was drawn and sent to a central laboratory (COVANCE, Indianapolis, IN) for determination of CD4 cell counts and HIV RNA levels by polymerase chain reaction in HIV-infected women.


Height and weight were measured. Body composition was measured by MRI with participants in the supine position, arms extended over head.[37,44] 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 the reference point to both the hand (above head) and foot. MRI scans were segmented using image analysis software (Tomovision Inc., Montreal, Quebec, Canada). Tissue areas (cm2) were calculated 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 2 consecutive slices was calculated via a mathematical algorithm.[45] Volumes were normalized by dividing by height squared with summaries back-transformed to 1.75 m of height. We did not adjust for body mass index (BMI), as BMI is influenced by the very phenomenon being studied: quantity of fat. Using these methods, we quantified adipose tissue volume in the leg, lower trunk (abdomen and back), upper trunk (chest and back), and arm, and VAT.

Statistical Analysis

For comparisons of prevalence, P values were calculated by Fisher exact test. Numerical values were compared by Mann-Whitney test. Associations between dichotomous variables were quantified by odds ratios (OR) from logistic regression models and those between continuous variables by rank correlations.

Multivariate analysis was performed to determine whether factors unrelated to HIV infection and its therapies could account for observed differences in MRI measures between controls, HIV-infected women with peripheral lipoatrophy, and HIV-infected women without clinical peripheral lipoatrophy. Separate analyses were performed for each of the following 5 anatomical sites: visceral, legs, lower trunk, arms, and upper trunk. For each anatomical site, separate comparisons were made of control versus HIV-infected women with lipoatrophy and control versus HIV-infected women without lipoatrophy. These models were fitted to logarithmic transformations of MRI measures divided by height squared, analogous to BMI, to produce estimated percentage differences in height normalized quantity of adipose tissue. Variables controlled for in the models included the following: age, ethnicity, smoking, alcohol intake, illicit drug use (crack/cocaine, marijuana, heroin, and amphetamines/speed), adequacy of food intake, and level of physical activity. As the purpose of these analyses was to examine possible changes in the estimated HIV effects, we included a relatively more expansive set of variables than would be appropriate for building predictive models. Variables selected included those that had P < 0.05 in preliminary (unbootstrapped) multivariate models for any of the 5 anatomical sites considered, along with some that had high a priori plausibility as potential confounders. Confidence intervals (CIs) were determined using the bias-corrected accelerated bootstrap method,46 with P values defined as 1 minus the highest confidence level that still excluded zero. This was necessary because many outcome measures appeared to be non-Gaussian, even after normalization by height squared, log transformation, and controlling for multiple important predictors.

In HIV-infected subjects, multivariate analyses were separately performed to determine which factors were predictive of MRI-measured leg SAT or VAT. Confidence intervals were constructed for the estimated percentage differences from the multivariate models using the bias-corrected accelerated bootstrap model as described above. In addition to the predictors listed above, these models included HIV RNA level (log10), and CD4 count (log2) at the time of study visit. In multivariate models controlling for the above factors, we evaluated total duration of use of each individual antiretroviral drug (ARV), ARV class (nucleoside reverse transcriptase inhibitor [NRTI], nonnucleoside reverse transcriptase inhibitor [NNRTI], and protease inhibitor [PI]) and highly active antiretroviral therapy (HAART) as defined by: (a) 2 or more NRTIs in combination with at least 1 PI or NNRTI, (b) 1 NRTI in combination with at least 1 PI and at least 1 NNRTI, (c) a regimen containing ritonavir and saquinavir in combination with 1 NRTI and no NNRTIs, or (d) an abacavir-containing regimen of 3 or more NRTIs in the absence of both PIs and NNRTIs. We checked linearity by fitting more complex models using linear splines, finding that linearity appeared reasonable in all cases. Duration of each ARV, ARV class, and HAART were added to the model in a forward stepwise manner, and results for each ARV not reaching P < 0.05 are shown controlling for the specific ARVs that were statistically significant predictors.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.