This study was reviewed and given approval by the North West Multi-centre Research Ethics Committee. Reference number MREC 02/8/79, Protocol number GS-02-1008c.
Improvements in cheek volume in lipoatrophic individuals switching away from thymidine nucleoside reverse transcriptase inhibitors
Version of Record online: 11 MAR 2009
© 2009 British HIV Association
Volume 10, Issue 6, pages 351–355, July 2009
How to Cite
Benn, P., Sauret-Jackson, V., Cartledge, J., Ruff, C., Sabin, C., Moyle, G., Linney, A., Reilly, G. and Edwards, S. (2009), Improvements in cheek volume in lipoatrophic individuals switching away from thymidine nucleoside reverse transcriptase inhibitors. HIV Medicine, 10: 351–355. doi: 10.1111/j.1468-1293.2009.00694.x
This study was presented in part at the following conferences: 7th International Workshop on Adverse Drug Reactions and Lipodystrophy in HIV, 13–16 November 2005, Dublin (Abstract 8); 3rd International AIDS Society Conference on HIV Pathogenesis and Treatment, 24–27July 2005, Rio de Janeiro, Brazil (Abstract TuLB2.3C02).
- Issue online: 10 JUN 2009
- Version of Record online: 11 MAR 2009
- Accepted 15 January 2009
- 3D surface laser scans;
- facial lipoatrophy;
- tenofovir DF;
Thymidine nucleoside reverse transcriptase inhibitors (NRTIs) are associated with subcutaneous fat loss. Facial changes cannot be assessed by dual-energy X-ray absorptiometry (DEXA) scans. There are limited objective data on the reversibility of facial lipoatrophy.
We performed a facial volume substudy of a randomized thymidine NRTI replacement study carried out in HIV-infected subjects with moderate to severe lipoatrophy. Facial volume changes were assessed using validated 3D laser imaging. Changes in body composition were measured using DEXA scans. The association between changes in facial volume and body composition parameters at 48 weeks was measured using Spearman's rank correlation.
Forty-seven individuals (46 male), 11 receiving zidovudine and 36 receiving stavudine, switched to either tenofovir disoproxil fumarate (DF) (n=23) or abacavir (ABC) (n=24). Thirty-nine of these 47 patients (84.8%) reported facial lipoatrophy at baseline. The median volume increase in both cheeks from baseline was 1857.3 mm3. These volume changes and increases in limb fat at 48 weeks were similar in the two groups and correlated significantly (Spearman's r=0.41, P=0.004).
Facial volume in lipoatrophic individuals was found to increase after thymidine NRTI replacement. We demonstrated a significant correlation between improvements in facial and limb fat parameters. Switching from thymidine NRTIs in patients with facial lipoatrophy could potentially reduce the need for cosmetic interventions.
Since the introduction of combination antiretroviral therapy (ART) in 1996 there have been dramatic reductions in mortality and morbidity among HIV-infected individuals . However, thymidine nucleoside reverse transcriptase inhibitor (NRTI)-based ART has been shown to be associated with metabolic changes and subcutaneous fat loss or lipoatrophy. The first reported association with these changes was for protease inhibitors in 1997 [2,3]. Subsequently, an association between the use of NRTI therapy and fat loss, mainly in the arms, legs, buttocks and face, was reported .
Facial lipoatrophy has been shown to impact negatively upon an HIV-infected individual's quality of life and ability to adhere to ART [5–7]. Quality of life improvements have been demonstrated over the short-to-medium term following cosmetic interventions to improve the appearance of facial lipoatrophy, including collagen and poly-l-lactic acid (PLA) [8,9]. However, there are recent reports of complications of administration of PLA, including subcutaneous nodule and granuloma formation, in a subset of individuals [10,11].
Dual-energy X-ray absorptiometry (DEXA) scans have been pivotal in demonstrating a link between NRTIs and lipoatrophy; in particular, the thymidine NRTIs (stavudine and zidovudine) have been associated with the development of lipoatrophy . In the MITOX Study, individuals with lipoatrophy experienced improvements in limb fat (0.4 kg; 11%) following switching from thymidine NRTIs to abacavir, as measured using DEXA scans . These improvements occurred slowly, however, and at 96 weeks following switching the limb fat improvements were not clinically apparent .
The potential for long-term complications of semi-permanent facial fillers emphasizes the importance of assessing the impact of switching ART after diagnosis of facial lipoatrophy. However, DEXA scans do not reliably measure changes in facial fat. The case definition for lipodystrophy  does not include measures of facial lipoatrophy and there is no consensus as to how facial lipoatrophy should be monitored. Ultrasound, computed tomography (CT) and magnetic resonance imaging (MRI) have been used to monitor facial lipoatrophy with limited success [16–18].
Laser scanning is being increasingly used in research studies and clinical practice by maxillofacial and plastic surgeons. The laser scans produce a 3D image of the face which is downloaded into a computer. It is possible to detect small changes in the contours of the face over time or following an intervention by superimposing the scans using bony landmarks.
We and others have previously demonstrated that 3D facial laser scanning provides reproducible images and that they can be used to quantify changes. We found that, without any intervention, the maximum mean volume difference between two scans taken 1 week apart for tissue overlying the cheek areas in 17 patients with varying degrees of facial lipoatrophy was −132 mm3. By comparison, the mean volume difference overlying the cheek areas in 19 patients undergoing scans before and after collagen injections into both cheek areas was +3150 mm3. Furthermore, a longitudinal study demonstrated that changes in facial parameters as measured by 3D surface laser assessment of the face significantly correlated with volume changes in cheek fat as measured by MRI [19–22].
We postulated that switching from a thymidine NRTI to either tenofovir DF (TDF) or abacavir (ABC) would result in improvements in facial lipoatrophy. We therefore undertook a substudy of the RAVE Study  to determine the effect on facial lipoatrophy of switching ART using 3D surface laser assessment scans.
The RAVE study was a multicentre, open-label, randomized study conducted in a group of 105 individuals with moderate to severe lipodystrophy who were receiving effective ART including either stavudine or zidovudine .
Participating individuals were randomized to substitute their thymidine NRTI with either TDF or ABC. Inclusion and exclusion criteria for RAVE have been outlined previously .
We undertook a substudy on 47 of the 105 patients, all of whom were receiving effective ART (11 were receiving therapy with zidovudine and 36 therapy with stavudine). This was largely a pragmatic sample of patients willing and able to attend a central London location for assessment and imaging. Changes in body composition in this subgroup were measured by DEXA, and 3D facial laser imaging was performed at baseline and at 48 weeks to detect volume changes overlying the forehead and cheeks. Individuals and clinicians also subjectively assessed the severity of fat accumulation and fat loss at baseline and at 24 and 48 weeks. Individuals and clinicians were asked to grade the severity of fat accumulation or loss according to a four-point scale: none, mild (noticeable only when specifically sought), moderate (readily obvious to the patient) and severe (obvious to a casual observer).
The scans at baseline and at 48 weeks were performed in a standardized way as previously described . The baseline and 48-week facial scans were compared by superimposing the scans using bony landmarks, thus enabling quantification of volume change (mm3) in a nonvariable region (forehead) and in variable regions (both cheeks).
For the purposes of this study, individuals were required not to have facial hair growth at the time of the scans and were told not to use drugs with the potential to affect facial fat or to undergo administration of facial fillers. Comparisons of the changes in facial imaging parameters over the 48-week period between treatment groups were performed using the Mann–Whitney U-test. Associations between the facial imaging parameters and other measures of body composition at 48 weeks were quantified using Spearman's rank correlation.
The majority of the 47 patients were male, were white and had a homosexual risk for HIV infection. The median age of the study group was 42 years (range 31–70 years) and the median prior duration of ART was 4 years (range 2–11 years). The characteristics of patients participating in the substudy were broadly similar to those of the overall RAVE study population regarding age, sex, baseline therapy and limb fat . However, patients participating in the substudy were slightly more likely to be of white race and slightly more likely to have been receiving stavudine at the time of randomization. All individuals had lipoatrophy by self or clinician report with the majority, 39 (83%), having facial lipoatrophy. Overall, 13 patients reported mild facial lipoatrophy, 14 moderate and 10 severe. The severity was not recorded in two individuals. Individuals were randomized for substitution of the thymidine NRTI with either TDF (n=23) or ABC (n=24) (Table 1). All individuals had a baseline facial scan within 30 days of randomization (median 3 days before randomization; range −30 days to +27 days) and the median interval between the baseline and follow-up 3D facial laser scans was 48 weeks (range 46–56 weeks).
|Number of patients||23||24|
|Male sex [n (%)]||23 (100.0)||23 (95.8)|
|Age (years) [median (range)]||41 (31–55)||43 (33–70)|
|Risk group [n (%)]|
|Homo/bisexual||22 (95.7)||22 (91.7)|
|Other/unknown||1 (4.3)||2 (8.3)|
|Origin [n (%)]|
|UK||16 (69.6)||21 (87.5)|
|Other/unknown||7 (30.4)||3 (12.5)|
|Race [n (%)]|
|White||23 (100.0)||22 (91.7)|
|CDC stage at randomization [n (%)]|
|A||7 (30.4)||9 (37.5)|
|B||7 (30.4)||9 (37.5)|
|C||9 (39.1)||6 (25.0)|
|Self-reported fat wasting [n (%)]*||23 (100)||23 (100)|
|Self-reported fat accumulation [n (%)]*||15 (65.2)||11 (47.8)|
|Self-reported facial lipodystrophy [n (%)]*||20 (87)||19 (82.6)|
|Time since first starting NRTIs (years) [median (range)]||5.6 (1.5, 11.5)||4.1 (0.8, 8.1)|
|Number of antiviral drugs to which the patient was exposed [median (range)]||4 (2–11)||3 (2–7)|
|Thymidine NRTI at time of randomization [n (%)]|
|Zidovudine||4 (17.4)||7 (29.2)|
|Stavudine||19 (82.6)||17 (70.8)|
|DEXA measurement (kg) [median (range)]|
|Total fat||10.3 (5.7, 21.5)||10.2 (5.5, 17.9)|
|Limb fat||2.9 (1.2, 8.7)||2.8 (1.8, 5.5)|
|Truncal fat||6.9 (2.9, 12.3)||6.2 (2.7, 12.0)|
Changes in facial parameters
At baseline, 39 (83%) of the 47 individuals self-reported facial lipoatrophy (this was confirmed in 32 of these patients by a physician; three had no physician assessment available and four did not have facial lipoatrophy confirmed by a physician). The median overall change from baseline to week 48 in the volume of tissue overlying the forehead (a nonvariable region) was −119 mm3 (range −1205.3 to 457.7; approximately equivalent to a cube 4.8 × 4.8 × 4.8 mm3). This demonstrated, as expected, that very little change occurred in a nonvariable region. By contrast, the median overall change from baseline to week 48 in the volume of tissue overlying both cheek areas (variable regions) was greater: +1857.3 mm3 (range −8048 to 25 049; approximately equivalent to a cube measuring 12.3 × 12.3 × 12.3 mm3). There was no evidence that changes differed by treatment group in either of these regions (P=0.41 and 0.88, respectively; Mann–Whitney U-test).
There was no significant association between the change from baseline to week 48 in total cheek volume and patient age (Spearman's r=−0.17, P=0.28), duration of exposure to NRTIs (Spearman's r=0.11, P=0.50), protease inhibitors (Spearman's r=0.14, P=0.47) or nonnucleoside reverse transcriptase inhibitors (Spearman's r=0.05, P=0.76), the number of drugs an individual had previously received (Spearman's r=0.03, P=0.86) or the CD4 count at recruitment (Spearman's r=0.21, P=0.17). Furthermore, the change in total cheek volume did not significantly differ according to the thymidine NRTI that was used prior to randomization (P=0.96; Mann–Whitney U-test).
At week 48, 12 (25.5%) of the 47 patients subjectively reported an improvement in the severity of facial lipoatrophy, whereas 15 (31.9%) reported a deterioration and 20 (42.6%) reported no change. There were no differences in these proportions by treatment group (P=0.98; χ2 test). While the median change in total cheek volume at 48 weeks was greater among individuals reporting an improvement (1426.0 mm3) or no change (2662.4 mm3) in the severity of their facial lipoatrophy compared with those who reported an increase in severity (658.7 mm3), these differences did not reach statistical significance (P=0.29; Kruskal–Wallis test).
Association between facial volume and DEXA-measured body composition
Correlation of the facial scan parameters and other measures of body composition demonstrated that there were no significant associations between volume changes in tissue overlying the forehead area and whole body, trunk or total limb fat change at week 48 (Table 2). There were, however, significant correlations between the total cheek volume change and the whole body, trunk and total limb fat change at week 48 (Spearman's r=0.41, P=0.008) (Table 3, Fig. 1). This association remained significant after controlling for weight change (P=0.03) (Fig. 1).
|Change from baseline to week 48 [median (range)]|
|Overall (n=47)||Tenofovir DF (n=23)||Abacavir (n=24)||P-value for comparison between treatment arms|
|Forehead volume (mm3)||−119 (−1205, 458)||−138 (−1205, 356)||−51 (−785, 458)||0.41|
|Total cheek volume (mm3)||1857 (−8048, 25 049)||2449 (−8048, 25 049)||1656 (−5049, 8974)||0.88|
|*Whole body fat (g)||1222 (−3557, 5494)||1172 (−3557, 5494)||1282 (−2324, 4759)||0.93|
|*Limb fat (g)||357 (−1611, 2203)||476 (−1611, 1999)||313 (−279, 2203)||0.87|
|*Trunk fat (g)||711 (−2070, 3896)||662 (−1917, 3896)||854 (−2070, 2534)||0.94|
|Forehead volume difference||Total cheek volume difference|
|Whole body fat change||−0.17 (0.29)||0.35 (0.03)|
|Limb fat change||−0.30 (0.06)||0.41 (0.008)|
|Trunk fat change||−0.13 (0.44)||0.35 (0.03)|
One individual received anabolic steroids and one corticosteroids during the study period, on days 343 and 129, respectively. One individual commenced PLA injections into both cheeks on day 133 after randomization. After excluding these individuals the association between total cheek volume and total limb fat change at week 48 remained significant (Spearman's r=0.35, P=0.03).
This is the first study to demonstrate an improvement in facial volume in addition to limb fat mass following cessation of thymidine NRTIs. The changes observed in both cheek areas represented a median volume increase of 1857 mm3 (mean 2539 mm3) over the study period.
The changes in total cheek volume observed in this study significantly correlated with the increases in limb fat as measured by DEXA scans. This relationship remained significant after adjusting for weight changes and excluding individuals who had received either steroids or PLA injections during the study period.
Unlike DEXA scans of limb fat, the 3D facial laser scans do not allow calculation of the percentage volume increase in the cheek area compared with baseline, which is a limitation of this methodology. The baseline scan provides a map of the contours of the face and does not quantify a volume of tissue. Subsequent scans identify changes in the contour of the face from baseline which can be converted into a volume change. In addition, further studies are required in nonwhite and female groups of patients to validate this methodology in these patient groups.
Our findings suggest that switching from thymidine NRTIs results in noteworthy improvements in facial volume and contour. While we are unable to prove that this is attributable to an increase in facial fat, previous data demonstrate that improvements in facial contour correlate with increases in cheek fat as measured by MRI imaging .
This effect suggests that the use of semipermanent fillers may be a more appropriate treatment than permanent fillers. Furthermore, early treatment switching may reduce or avoid the requirement for facial fillers.
While this study includes relatively small numbers, it is the first study to demonstrate that facial lipoatrophy in addition to limb fat can improve on switching away from thymidine NRTIs. The role of 3D facial scans in the early detection of facial lipoatrophy, thus facilitating early switching away from thymidine NRTIs, clearly warrants further evaluation in longitudinal studies.
- 2Reports of Diabetes and Hyperglycaemia in Patients Receiving Protease Inhibitors for the Treatment of Human Immunodeficiency Virus. Washington, DC: FDA Public Health Advisory, 1997..