Evaluation of ultrasound for assessing facial lipoatrophy in a randomized, placebo-controlled trial

We investigated the utility of ultrasonography for assessing facial lipoatrophy changes in HIV-infected adults receiving antiretroviral therapy who participated in a 48-week, randomized, placebo-controlled trial of rosiglitazone. Ultrasound was performed at weeks 0, 24 and 48 to determine the subcutaneous fat thickness over the malar bone. Changes in facial fat assessed by ultrasonography did not correlate significantly with more established objective measures of lipoatrophy severity. The measurement of malar fat using ultrasonography is not recommended.

The physical presentation of HIV lipodystrophy can be disfiguring and stigmatizing, and may result in reduced antiretroviral adherence, which increases the risk of treatment failure ^[1]. Although standard protocols using whole-body dual-energy X-ray absorptiometry (DEXA) and single or three-slice abdominal computed tomography (CT) have been developed to assess body composition objectively ^[2,3], facial fat is not quantitated. Sonography has been used in HIV lipodystrophy to quantify subcutaneous fat ^[4], and in intervention studies to quantify facial tissue thickness changes ^[5,6]. We undertook a study in patients participating in a randomized, 48-week, placebo-controlled trial of rosiglitazone for HIV lipoatrophy ^[7]. The primary objective was to determine the correlation between malar sonographic measures and body composition parameters assessed by DEXA and CT.

Although voluntary, access to one scanning site limited participation. All subjects provided written, informed consent after research ethics committee approval. Details of main study entry criteria, participants, interventions and assessments have been described previously ^[7]. An ultrasonographic assessment using a linear 7-10 MHz transducer was performed at weeks 0, 24 and 48. On each occasion a single longitudinal image was taken at the outer canthus of the right eye, with the upper margin defined by the malar bone. The malar fat thickness was the maximal distance between the skin surface (without skin depression) and the muscle plane inferior to the malar bone. All images and measurements were recorded and performed on one scanner.

Analyses were performed when all subjects had completed 48 weeks or were permanently lost to follow-up. The baseline characteristics of participants and non-participants were compared using descriptive statistics to assess representativeness. Bivariate associations between malar fat changes and body composition parameters were analysed using the Spearman correlation coefficient test. Analyses of change from baseline at week 48 were by intention-to-treat. Between-group differences were tested using non-parametric methods.

Fifty-six of the 58 individuals (97%) recruited in the scanning site vicinity gave consent, representing 52% of the main trial. Most participants were male (98%), and had a significantly greater mean limb fat mass (by DEXA) than non-participants (3.04 versus 2.32 kg; P = 0.006; Table 1). Twenty-five subjects (45%) received rosiglitazone, and 31 (55%) received placebo. During the trial, two subjects (one rosiglitazone) withdrew for personal reasons, and three (placebo) either did not attend the week 0 (one) or week 48 (two) ultrasound. One rosiglitazone and two placebo recipients ceased study treatment, one for an adverse event and two for personal reasons. Fifty-one subjects (91%) completed three ultrasounds.

Table 1

At baseline, malar fat did not correlate significantly with subcutaneous abdominal fat (ρ = 0.02; P = 0.90) or mid-thigh subcutaneous fat by CT (ρ = -0.03; P = 0.82) or limb fat mass assessed by DEXA (ρ = 0.02; P = 0.89). No significant associations were shown between a change in malar fat and a change in other objective body composition or lipodystrophy endpoints assessed in the main trial: limb fat mass, (ρ = 0.10; P = 0.44) (Fig. 1); mid-thigh subcutaneous fat (ρ = 0.10; P = 0.70); or the lipodystrophy case definition score (ρ = -0.20; P = 0.16) ^[8]. Also, no significant associations were shown with a change in subjective lipodystrophy severity scores assessed by physical examination (ρ = 0.04; P = 0.80) or patient questionnaire (ρ = 0.10; P = 0.60).

Fig. 1

At week 48, mean malar fat had decreased in both arms; from 4.6 to 3.0 mm (SD 1.9) in the rosiglitazone arm and from 4.7 to 2.8 mm (SD 1.9) in the placebo arm (mean difference 0.4; 95% confidence interval -0.9-1.6 mm; P = 0.53 by t-test). This decline contrasts with the 0.14 kg (SD 0.58) and 0.18 kg (SD 0.58) increases in limb fat mass observed with rosiglitazone and placebo, respectively, at 48 weeks ^[7].

These data suggest that ultrasonography is a poor measure of facial lipoatrophy, both cross-sectionally and longitudinally. The significant decrease in malar fat over 48 weeks is not easily interpretable. A previous cross-sectional, HIV lipodystrophy study ^[4], which used sonography found a mean malar fat thickness of 3.5 mm (SD 0.8), a measurement comparable to our mean baseline of 4.7 mm (SD 1.9). Sonography was employed in two intervention studies to quantify facial fat changes ^[5,6]. In the first, a prospective, 96-week study ^[5], measurements were made below the malar bone without specific landmarks, and as objective measures of lipoatrophy severity were not performed, comparison with our study is not possible. In the second ^[6], a randomized, prospective 24-week study, measurements were made over both malar bones. Based on graphical data, a change of approximately 1 mm over 24 weeks was observed at both sites. This change is within the error of the procedure and in keeping with our findings.

Milinkovic and colleagues ^[9] found a lack of association between sonography and more established and objective measures of body composition. In patients commencing antiretroviral therapy, correlation coefficients for subcutaneous abdominal fat using CT and sonography, and arm fat using DEXA and sonography, declined substantially over 12 months: from ρ = 0.50 at baseline to ρ = 0.10 for CT; and from ρ = 0.38 to ρ = 0.01 for DEXA ^[9]. In contrast, the association between objective techniques, DEXA and CT, remained high (ρ = 0.67; ρ = 0.79) over the same period.

Although sonography can detect density interfaces with 1 mm accuracy ^[10], measurement is subjective as extraneous echoes from within adipose tissue can make fascia difficult to visualize. The close proximity of adipose tissue-muscle interfaces with bone can also produce multiple or confusing echoes, thereby making measurement less accurate in severe lipoatrophy.

In adults with severe facial lipoatrophy, sonography may be subject to measurement error, correlates poorly with more established objective measures of lipoatrophy, and is therefore not recommended. This outcome is disappointing as there is a great need for non-invasive technology to quantitate facial lipoatrophy.

Acknowledgement

The authors would like to thank participants, investigators and coordinators at Sydney Rosey sites.

Sponsorship: Financial support for this work was received from Bristol-Myers Squibb, and study drug was provided by GlaxoSmithKline. The National Centre in HIV Epidemiology and Clinical Research is funded by the Australian Government Department of Health and Ageing, and is affiliated with the Faculty of Medicine, The University of New South Wales.