|Home | About | Journals | Submit | Contact Us | Français|
The aim of this study was evaluation of ultrasound (US) as a tool for the assessment of lipoatrophy in a population of HIV-infected patients. We enrolled a convenience sample of 151 HIV-infected Caucasian participants (males, 79%) who were treated for at least 1 year with combination antiretroviral therapy (CART) in Zagreb, Croatia. US measurements of subcutaneous fat thickness were done over the malar, brachial, and crural region. We determined sensitivity and specificity of US as a diagnostic tool for lipoatrophy using receiver-operating curves and concordant patient and clinician assessment as our reference for the presence of lipoatrophy. HIV was acquired through heterosexual contact in 50% of participants and by sex between men in 42%. The mean current CD4 cell count was 503.1cells/mm3 (standard deviation [SD]=250.8). Seventy-seven (51%) participants were treated with stavudine and 91 (64%) with a protease inhibitor for at least 6 months. Nineteen (13%) participants had lipoatrophy in at least one anatomic site. Sensitivity of US ranged from 67%–71%, specificity from 65%–71%, positive and negative predictive values ranged from 11%–20% and 96–97%, respectively. US-diagnosed lipoatrophy was more frequently found in patients with a history of stavudine treatment and in females. Patients with lipoatrophy had a longer duration of CART than those without lipoatrophy. US is a useful tool in ruling out the presence of clinical lipoatrophy in patients on CART. Using this objective measure of subcutaneous fat may be useful in helping clinicians make decisions about changing therapy.
The advent of combination antiretroviral therapy (CART) for HIV infection has led to improvements in both quality of life and life expectancy for HIV-infected patients.1 However, a number of unanticipated toxicities have emerged. Prominent among these is lipodystrophy, a syndrome that in early reports was characterized by increased fat (or lipohypertrophy) in intraabdominal, dorsocervical, and breast areas, and loss of fat (or lipoatrophy) in the face, limbs, trunk, and buttocks. These body fat changes can be stigmatizing and have been associated with decreased adherence to CART.1–3 Both lipoatrophy and lipohypertrophy can occur separately or together in an individual.4,5 Of the two, lipoatrophy is more common and more disfiguring.3,4
The underlying mechanisms of HIV-associated lipoatrophy remain unclear but are likely multifactorial.6 Intrinsic host factors, disease status, treatment duration and type, as well as other factors probably all play a role.7 Significant risk factors for lipoatrophy include exposure to and longer duration of thymidine analogues (particularly stavudine), older age, higher CD4 count, lower viral load, and white ethnicity.8
Two recent cross-sectional studies demonstrated that HIV-infected participants with and without clinical peripheral lipoatrophy had less subcutaneous fat than HIV-uninfected controls as measured by magnetic resonance imaging (MRI), suggesting that lipoatrophy becomes noticeable when subcutaneous fat loss is substantial.9,10 Therefore, imaging modalities could be used to detect fat loss early, before lipoatrophy becomes clinically apparent.
However, current imaging modalities that have been used for study of body fat changes including dual energy x-ray absorptiometry (DEXA), computerized tomography (CT), and MRI11 are expensive and impractical for use in resource-limited settings.12 Ultrasound (US) has been used less frequently than CT scan or DEXA for the measurement of regional body fat in HIV-infected persons,8 but may be a more practical alternative that also minimizes radiation exposure.13
Because of the relative ease of US and its wide availability in middle-income countries, we assessed the use of US to measure subcutaneous fat loss and to assess its concordance with patients' self report and physicians' diagnosis of fat lipoatrophy in HIV-infected patients in Croatia. This study will allow us to investigate the use of US to rule out lipoatrophy and inform providers in their decisions to change their patients' antiretroviral treatment regimens.
We conducted a cross-sectional study to compare US sensitivity and specificity in diagnosis of lipoatrophy in concordance with patient and physician reports of lipoatrophy.
Participants were selected using a convenience sample from among the HIV-infected individuals seeking treatment at the University Hospital for Infectious Diseases (UHID) in Zagreb, Croatia. Between 1985 and 2006 there were 608 documented cases of HIV infection in Croatia, of which 285 progressed to AIDS. The majority of Croatian AIDS patients are male and between the ages of 25 and 49 years.14 Croatia has a centralized system of care and all HIV-infected patients are treated at the HIV/AIDS centre at UHID.15 Participants were consecutively recruited and enrolled over a 15-month period beginning in July 2006. We included all patients older then 18 years with documented HIV infection (enzyme immunoassay and Western blot positive) and at least 1 year duration of CART. We excluded patients with AIDS-related dementia, lipodystrophies predating HIV infection, or current acute illness or current use of drugs that modify lipid and glucose metabolism (e.g., insulin and statins), diuretics, or corticosteroids.
We enrolled participants and performed all measurements at the same visit. A single radiologist, who was blinded to the clinical assessment and self-report, performed US measurements of thickness of subcutaneous fat in millimeters in three anatomic sites, over the right malar bone, approximately 10cm above the right elbow (brachial) and 6cm above the right lateral maleolus (crural), without pressing the underlying skin, using a SIEMENS Sonoline G-50 machine (Siemens Medical Solutions, Malvern, PA) in B mode. Each region was examined separately three times, according to the technique described by Martinez and collegues,16 and the measurements were averaged. We used a linear array probe (10MHz and 42mm) with the patient in the supine position, and positioned electronic calipers at the skin-fat (excluding skin) and fat–muscle interfaces.
Prior to the US, each patient completed a short questionnaire about observed fat changes in peripheral body sites, and at the same visit, the attending infectious diseases physician also assessed regional fat using a similar standardized questionnaire. The presence or absence of lipoatrophy or lipohypertrophy was determined using an instrument from the Fat Redistribution and Metabolic Change in HIV Infection Study (FRAM).1 We defined lipoatrophy and lipohypertrophy according to FRAM criteria, as participant-reported loss or gain of fat confirmed by clinical examination using a similar standardized questionnaire. The degree of lipoatrophy at each region was rated as absent and mild (score 0) or moderate and severe (score 1). Participants were blinded to physicians' assessments, and physicians were blinded to participants' self-assessments. Both patients and physicians had to report moderate/severe loss of fat in face and limbs to be considered to have clinically evident lipoatrophy. We abstracted additional data for each participant, including medication history, from medical records using a standardized abstraction form. Treatment with a particular antiretroviral drug or drug class was considered present if the drug was administered for more than 6 months.
We measured weight using a standard physician's office scale, rounded to the nearest 0.1kg and height with a wall-mounted stadiometer and rounded to the nearest 0.1cm. Those measurements were done three times and mean values were calculated.
We determined plasma HIV viral load using COBAS Amplicor HIV-1 Monitor Test, version 1.5 (Roche Diagnostic Systems, Basel, Switzerland) with a lower limit detection of 50 copies per milliliter (ultrasensitive method). We measured absolute CD4 T-cell counts (in cells/mm3) using Flow Count Fluorospheres (Beckman Coulter, Fullerton, CA). Four-color flow cytometry was performed using Cytomics FC500 flow cytometer (Beckman Coulter).
We calculated means, standard deviations (SD), and ranges for anthropometric, CD4 cell counts, duration of CART and US measurements. The χ2, Fisher exact test, or Mann-Whitney U test was used for comparison of categorical and ordinal variables and the Student's t test for comparison of continuous variables. The correlation between the atrophy severity score and US was assessed by Spearman's rank correlation coefficient (ρ). Concordance between participants' and physicians' reports was tested using the κ statistic.17 We examined the relationship between US measurements (in absolute values-millimeters) and concordant reports over a range of potential cutoff points. We used this series of points to plot receiver operating characteristics (ROC) curve. After plotting the curves, we chose a single cutoff point that maximized sensitivity and specificity (area under the curve). After choosing this cutoff point, we calculated the sensitivity, specificity, and positive and negative predictive values at that cutoff point. All analyses were conducted using SPSS statistical software version 13 (SPSS, Chicago IL).
All patients provided written informed consent prior to enrolment. The Ethics Committee of UHID and Committee on Human Research at the University of California, San Francisco (UCSF) approved the protocol.
We enrolled 151 participants. Information on selected characteristics of these participants are presented in Table 1. Thirty-two participants (21%) were females and 119 (79%) males. All participants were Caucasians. Fifty percent of patients acquired HIV infection through heterosexual contact, 42% through sex between men, 3% through transfusion of infected factor VIII, and 5% through injection drug use. The mean duration of CART was 4.4 years (SD=2.4). One hundred thirty-five (89%) participants received thymidine analoges, 77 (51%) a stavudine containing CART-regimen and 97 (64%) a protease inhibitor. Ninety-one participants (60%) had a prior diagnosis of AIDS at the time of the study.
Of the 151 patients, 17 (11%) reported facial lipoatrophy, 17 (11%) reported brachial lipoatrophy, and 25 (17%) reported crural lipoatrophy. Study physicians judged 23 participants (15%) to have lipoatrophy at the malar region, 19 (13%) in the brachial region, and 26 (17%) in the crural region. Altogether 52 patients had lipoatrophy at any site; 15 were reported by self-report only, 18 by physicians only and 19 by both self-report and physicians. Of these 52 participants, physicians reported lipoatrophy in 37 (71%) and participants self-reported lipoatrophy in 34 (65%). Concordance between participant and physician assessments varied by anatomical site. Concordance was 53% (κ=0.368; p<0.001) in the malar region, 59% (κ=0.496; p<0.001) in the brachial region and 56% (κ=0.457; p<0.001) in the crural region (Table 2). There was a weak correlation between US and participants-reported degree of fat loss (ρ=−0.19 in the malar region; ρ=−0.24 in the brachial region; and ρ=−0.21 in the crural region). The correlation between US and physician-reported degree of fat loss was somewhat stronger (ρ=−0.21 in the malar region; ρ=−0.32 in the brachial region; and ρ=−0.36 in the crural region).
The average subcutaneous fat thickness by US measurement was 3.2mm (SD=1.1) in the malar region, 4.1mm (SD=2.6) in the brachial region and 3.7mm (SD=1.6) in the crural region. Using concordant patient and clinician assessment as our standard for the presence of lipoatrophy, we calculated sensitivity and specificity for a series of US measurements of subcutaneous fat at each region and plotted ROC curves. The area under the ROC curve for the malar region was 0.753 (p=0.011), in the brachial region 0.696 (p=0.038) and, in the crural region 0.731 (p=0.004).
Using the ROC curves, we determined single cut-off values for subcutaneous fat measurements at all three regions that maximized sensitivity and specificity (Table 3). In the malar region, a cutoff value of 2.75mm yielded 67% sensitivity and a 65% specificity. In the brachial region, a cutoff value of 2.55mm yielded 70% sensitivity and 76% specificity. At the crural site, a cutoff value of 2.95mm had a sensitivity of 71% and a specificity of 71%. The positive predictive values were 11% for malar region, 17% for brachial region, and 20% for crural region. However, the negative predictive values were 97% (for the malar and brachial regions) and 96% for the crural region.
Nineteen (13%) participants had lipoatrophy in at least one site determined by US (Table 4). This group of patients did not differ from patients without lipoatrophy with respect to age, CD4 cell account or plasma HIV viral load. Lipoatrophy was more often found in females than in males (p=0.03). Participants who used stavudine for more than 6 months had more frequently lipoatrophy (p<0.001). The mean duration of therapy for patients with lipoatrophy was 5.8 years (SD=2.35) compared to 4.21 years (SD=2.42) for patients without lipoatrophy (p=0.011).
We demonstrated that US can be a clinically useful tool for diagnosing lipoatrophy in comparison to clinical assessment. Although the sensitivity and specificity of US were moderate (sensitivity ranged from 67% to 71% and specificity from 65% to 76% depending on anatomic region) the negative predictive value was high (96%–97%). This is due to the relatively low prevalence of lipoatrophy in the observed sample and suggests that US is a useful tool in ruling out the presence of clinical lipoatrophy in HIV-infected patients. It should be noted that these sensitivities and specificities were derived based upon subjective criteria of self-report confirmed by examination to determine a dichotomous outcome of lipoatrophy.
We were interested in studying the utility of US as a way to measure lipoatrophy because US has several advantages over other radiologic techniques particularly in low and middle income countries. US is readily accessible and available at the bedside, can be performed quickly during routine visits and, unlike imaging techniques that use ionizing radiation, possesses essentially no risk to the patient. MRI has been shown to be a highly accurate technique for measuring subcutaneous fat,11 but we believe that in Croatia the routine use of MRI for detecting lipoatrophy is impractical. The cost per measurement using US in Croatia is approximately 5% of the cost of measurement by MRI scan, and moreover, MRI scans are not readily available.
Our study showed moderate concordance between patient self-assessment of lipoatrophy and clinician assessment by κ statistics.17 This moderate concordance may be explained by several factors. Facial morphology, patient age, and skin quality may affect how subcutaneous fat is outwardly perceived.18 Patients and clinicians may perceive lipoatrophy differently, as patients may be less sensitive to small changes in facial or body shape over time. Clinicians, by contrast, see patients less frequently and may be more able to perceive changes. There are new and encouraging data on surgical treatment of facial lipoatrophy.19 However, such treatment is usually not available in low and middle income countries.
A potential limitation associated with the use of US may be the variation in the level at which the measurement was performed and external echoes from within adipose tissue that can make fascia difficult to visualize.12 The close proximity of tissue-muscle interfaces with bone can also produce multiple or confusing echoes, making measurement less accurate in severe lipoatrophy.12 Additionally, we were limited by our cross-sectional design, which did not allow for measurements of change in subcutaneous fat over time. Other limitations are that US cutoffs were based upon clinical report and examination, which are subjective but considered the standard in the field in most cohort studies.20,21
In summary, our findings suggest that US has potential utility for objectively ruling out lipoatrophy in HIV-infected and CART-treated patients. We are introducing US as a routine diagnostic procedure during regular follow-up visits for HIV-infected patients in our clinic which will allow us to examine the predictive value of US in a longitudinal fashion.
The authors kindly thank Professor Thomas Novotny of the University of California, San Francisco, California, for his guidance and support during all stages of this study. We would also like to acknowledge Dr. Phyllis C. Tien, from the Infectious Disease Section, Department of Veterans Affairs Medical Center San Francisco, California, for her helpful comments. Dr. Klaudija Viskovic was supported by the UCSF Center for AIDS Prevention Studies (US NIMH P30MH062246), the International Traineeships in AIDS Prevention Studies Program (US NIMH R25MH064712), and the C.V. Starr Foundation. We also thank the National AIDS Programmes of Southeast European countries; Andrija Stampar School of Public Health, Knowledge Hub on Capacity Building in HIV/AIDS Surveillance, Zagreb, Croatia.
No competing financial interests exist.