The aim of the study was to demonstrate the noninferiority of polyacrylamide hydrogel (PH) vs. polylactic acid (PLA) for the treatment of facial lipoatrophy in HIV-infected adults.
The aim of the study was to demonstrate the noninferiority of polyacrylamide hydrogel (PH) vs. polylactic acid (PLA) for the treatment of facial lipoatrophy in HIV-infected adults.
A randomized, blinded, multicentre, noninferiority 96-week study was carried out. Patients with facial lipoatrophy were randomly assigned to receive intradermal injections with PH or PLA, and were blinded to the filler. The primary efficacy endpoint was patient satisfaction at week 48 assessed using a visual analogue scale score (VAS). Secondary efficacy end-points included cheek thickness and skin-fold, lipoatrophy grading and quality of life. Safety was assessed by the reporting of adverse events.
A total of 148 patients were included in the study; 93% were men, the median age was 47 years, the median CD4 count was 528 cells/μL, and the median duration of antiretroviral therapy was 12 years. Mean VAS increased from 2.8 at baseline to 7.1 and 7.5 in the PLA and PH arms, respectively, at week 48 (P = 0.0002 for noninferiority) and was sustained at week 96 (6.7 and 7.9 in the PLA and PH arms, respectively; P = 0.003 for noninferiority). Cheek thickness and skin-fold increases and lipoatrophy improvement were similar in the two arms. Quality of life remained unchanged or improved depending on the questionnaire used. In injected patients, subcutaneous nodules emerged in 28 (41%) and 26 (37%) patients in the PLA and PH arms, respectively (P = 0.73). Four patients in the PH arm developed severe inflammatory nodules, a median of 17 months after the last injection.
PH and PLA have similar efficacies in the treatment of facial lipoatrophy, but PH may be associated with more delayed inflammatory nodules.
Facial lipoatrophy is still a prevalent and stigmatizing complication of antiretroviral therapy that can severely affect quality of life and self-esteem, and may result in reduced antiretroviral compliance [1, 2]. The pathogenesis of lipoatrophy is multifactorial but the toxicity of thymidine analogue antiretrovirals is likely to play a major role . Current therapeutic options are limited. Reversibility of lipoatrophy after switching to thymidine analogue-sparing regimens is at best slow and limited [4, 5]. Use of thiazolidinediones, pravastatin or uridine has produced rather disappointing results, especially regarding the correction of facial lipoatrophy [6-8]. For these reasons, plastic surgery and dermal fillers are treatments of choice for facial lipoatrophy. Autologous fat transplantation (AFT) is a surgical procedure that needs harvestable fat in patients, and which can be associated with resorption or hypertrophy of transferred fat [9, 10]. Injections of facial fillers have therefore become the standard of care for patients with lipoatrophy. Three biodegradable filling compounds have received Food and Drug Administration (FDA) approval for correction of facial lipoatrophy in HIV infection: calcium hydroxyapatite gel, hyaluronic acid and poly-L-lactic acid (PLA) . However, most studies report an incomplete resolution of lipoatrophy in a number of patients and a decreasing benefit over time, prompting the need for re-injections [11-16]. More recently, the interesting efficacy of polyacrylamide hydrogel (PH), a nonbiodegradable aqueous gel, has been reported, with a favourable safety profile [17-20]. Current data are unfortunately not sufficient to guide the choice among these different therapeutic options, with only one partially randomized study available . In France, PLA is the only filler approved for the treatment of HIV-associated lipoatrophy and we wished to assess another alternative treatment for this condition. We thus conducted a prospective, randomized, multicentre study to compare the efficacy and safety of PLA and PH for the treatment of antiretroviral-induced facial lipoatrophy in HIV-infected patients.
Eligible patients were HIV-infected adults, with antiretroviral therapy-induced facial lipoatrophy and stable antiretroviral treatment for at least 3 months. The main exclusion criteria were previous surgical or cosmetic intervention for facial lipoatrophy, use of any fillers, concomitant opportunistic infection or major illnesses, current stavudine-based antiretroviral regimen, use of anti-aggregant therapy, a CD4 cell count < 100 cells/μL, a plasma HIV RNA level > 10 000 HIV-1 RNA copies/mL, a platelet count < 50 000 cells/μL, abnormal coagulation tests and pregnancy.
Written informed consent was obtained from all participants. The study was approved by the scientific board of the Agence Nationale de Recherches sur le SIDA et les Hépatites Virales (ANRS) and the Paris Saint-Louis ethics committee.
This was a prospective, randomized, patient-blinded, noninferiority study. Participants were randomly assigned, in a 1:1 ratio, to receive facial injections of PLA or PH. Randomization was stratified according to the severity of the lipoatrophy, as assessed using a baseline satisfaction visual analogue scale score (VAS) (< 3, between 3 and 6 or > 6), and by centre . A random allocation sequence was generated using a computer random number generator, and balanced by blocks of four. Patients were blinded to the filler and a sticker was put on the syringes filled with PH or PHA prior to the injections in order to mask the name of the filler.
For PLA injections, at least 2 h before each injection, 0.15 g of PLA (Newfill TM; Sanofi-Aventis, Strasbourg, France) in the form of a sterile dry powder was reconstituted with 4 mL of sterile water to produce a suspension. On the day of the injections, 1 mL of plain lidocaine was added to the vial and 1-mL syringes were prepared for the injections. PH (Eutrophill®; Procytech, Marcillac, France) was supplied as a ready-to-use gel in disposable 1-mL syringes. Facial injections in the two cheeks were performed by the same operator at each centre. After each injection, the cheeks were gently massaged in order to avoid bruises and nodules.
Patients were scheduled to receive a minimum of three sessions of injections (day 0 and weeks 4 and 8) and a maximum of seven. These sessions were scheduled every 4 weeks until week 24 according to the patient's request, with no injections allowed between weeks 24 and 48. Re-injections were only allowed after week 48 (primary endpoint), using the same filling compound, according to the patient's request.
At the screening visit, the severity of lipoatrophy was assessed by the patient (VAS) and also by the physician, the CD4 cell count and plasma HIV viral load were measured, and pregnancy and coagulation tests were performed. After screening, patients were randomized and seen every 4 weeks from baseline to week 28 (4 weeks after the last injection session), and then at weeks 48, 72 and 96. At each visit a physical examination was performed, VAS and pain scores were assessed, and cheek thickness and skin-fold were measured. Coagulation tests were performed before each session of injections. Quality of life assessments and photographs were also taken at baseline and at weeks 48 and 96.
The primary efficacy endpoint was patient satisfaction at week 48, assessed using a VAS ranging from 0 (very dissatisfied) to 10 (very satisfied), when patients were asked the following question: What is your degree of satisfaction regarding your face in relation to the lipoatrophy?, as previously reported . VAS was also assessed at week 96 to assess the long-term efficacy of both fillers. Secondary outcomes were assessed at weeks 48 and 96 and included cheek thickness and cheek skin-fold, measured using a skin-fold caliper , and quality of life, assessed using the Medical Outcomes Study HIV Health Survey (MOS-HIV) questionnaire  and the aNBC questionnaire derived from Carr et al. . This latter questionnaire assessed the severity of lipoatrophy as evaluated by the patients and the physician [from 0 (no severity) to 10 (maximum severity)] and discomfort as evaluated by the patients [from 0 (no discomfort) to 10 (maximum discomfort)], with summary scores ranging from 0 for no severity/discomfort to 5 for maximum severity/discomfort. Assessment of lipoatrophy by two dermatologists was also performed: photographs for 60 patients (30 from each arm of the study) were randomly selected at baseline, week 48 and week 96, were anonymized and then were randomly presented to the two investigators, blinded to the received filler. The two dermatologists independently scored the severity of lipoatrophy in each photograph using the following scale: 0, no lipoatrophy; 1, mild lipoatrophy; 2, moderate lipoatrophy; 3, severe lipoatrophy, according to the criteria of James et al. .
The safety of the injections was evaluated at each visit for patients who received at least one filler injection (the ‘safety population’). Adverse events during and after injections were reported. Pain during injection was scored from 0 (no pain) to 10 (maximum pain) using a VAS. The severity of adverse events was graded using the ANRS grading scale (available at: http://www.isped.u-bordeaux2.fr/RECHERCHE/GETSI/grilles_gravite/FR-GETSI-ANRS-099-grille-ANRS.htm). In addition, palpation of the cheeks at the site of injections was performed at each visit to detect the emergence of subcutaneous nodules.
The primary objective of the study was to show noninferiority of PH vs. PLA in VAS at week 48. The predefined noninferiority margin was set at 15% of the expected mean VAS at week 48 in the PLA arm (6.7 according to a previous study ) which corresponded to an absolute difference of 1 point or less in VAS for the PH group as compared with the PLA group. With a 2.5% one-sided type I error rate and a power of 80%, and assuming a 5% loss to follow-up, 70 patients had to be recruited in each group.
The main analysis was an intention-to-treat (ITT) analysis, where all randomized patients were analysed in the treatment group to which they were allocated, whether or not they received injections. Missing data were imputed using the last value carried forward (LOCF) methodology. Sensitivity analyses were performed where missing values were also imputed by linear interpolation between the last value before and the first value after the missing visit. An observed analysis restricted to available data was also performed where visits performed during a 5-week window before or after the week 48 scheduled visit or an 8-week window before and after the week 96 visit were used (evaluated cases). Per protocol (PP) analyses were then performed. The PP population was defined as randomized patients having received at least the three injections they were supposed to undergo. Further analyses adjusted on baseline VAS and randomization strata (baseline VAS < 3, between 3 and 6 or > 6) were also performed using analysis of covariance and analysis of variance, respectively. Rates of treatment failure were also compared between arms, with failure defined as a VAS at week 48 or week 96 below or equal to the baseline value. The noninferiority margin for the rate of failure was set at 12%.
Other endpoints were analysed using Student's t-tests, Wilcoxon rank-sum tests or Fisher's exact tests as appropriate, with a 5% two-sided type I error rate.
One hundred and forty-eight patients were enrolled in the study between 21 December 2006 and 29 September 2008 in 12 clinical sites in France. A flow chart for the patients included in the study is depicted in Figure 1. The number of patients who remained in the study at different time-points, as well as the number of patients who dropped out and the reasons for drop-out, are shown in the flow chart. However, among patients remaining in the study at week 48 (i.e. patients who had follow-up visits at week 48 or after), some were not evaluated within the protocol 5-week window around week 48 (i.e. between weeks 43 and 53). For instance, a patient who had visits 30 and 54 weeks after the first injection, but none in between for any reasons (e.g. failure to attend one visit), would not be considered as having a valid week 48 evaluation, but would still be considered to remain in the study.
The patients' characteristics are summarized in Table 1 and were well balanced across treatment arms. Most patients were male (93%), the median age was 47 years, the median duration of antiretroviral therapy was 12 years, plasma HIV RNA level was < 50 copies/mL in 75% of patients and the median CD4 cell count was 528 cells/μL. The severity of lipoatrophy as assessed by the physician was moderate or severe in 75% of patients, and 53% of patients scored their lipoatrophy with a VAS < 3, with a median score of only 2.8.
|PLA (n = 73)||PH (n = 75)|
|Age (years) [median (range)]||48 (30–72)||47 (34–61)|
|Male [n (%)]||70 (96)||68 (91)|
|Body mass index (kg/m2)[mean (SD)]||21.9 (3.1)||21.0 (2.6)|
|HIV transmission risk [n (%)]|
|Heterosexual||14 (19)||15 (20)|
|Homo/bisexual||37 (51)||38 (51)|
|IDU||14 (19)||12 (16)|
|Other||9 (11)||11 (13)|
|Duration of HIV infection (years) [median (range)]||17 (3–25)||16 (8–24)|
|History of AIDS [n (%)]||28 (39)||28 (38)|
|CD4 count (cells/μL) [median (range)]||538 (175–1994)||528 (200–1395)|
|Plasma HIV RNA < 50 copies/mL [n (%)]||52 (71)||59 (79)|
|Antiretroviral exposure (years) [median (range)]||12 (1–20)||13 (2–19)|
|Current antiretroviral regimen [n (%)]|
|Protease inhibitor||41 (57)||50 (67)|
|Nucleoside reverse transcriptase inhibitor||69 (96)||69 (92)|
|Nonnucleoside reverse transcriptase inhibitor||28 (39)||28 (37)|
|Patients on zidovudine [n (%)]||9 (12)||10 (13)|
|Cumulated exposure (months) [median (range)]|
|Stavudine||37 (0–158)||44 (0–170)|
|Didanosine||38 (0–136)||38 (0–123)|
|Zidovudine||19 (0–126)||18 (0–128)|
|Tenofovir||11 (0–79)||8 (0–82)|
|Abacavir||0 (0–101)||0 (0–84)|
|Lamivudine/emtricitabine||70 (5–151)||83 (0–150)|
|Indinavir||0 (0–107)||2 (0–109)|
|Time since last stavudine intake (months) [median (range)]*||56 (1–121)||60 (4–126)|
|Severity of lipoatrophy† [n (%)]|
|Mild||16 (22)||21 (28)|
|Moderate||37 (51)||32 (43)|
|Severe||20 (27)||22 (29)|
|Visual analogue scale satisfaction score (VAS) [n (%)]|
|< 3||43 (59)||44 (59)|
|3–6||21 (29)||25 (33)|
|> 6||9 (12)||6 (8)|
Over the first 48 weeks of the trial, the median number of injection sessions per patient was 5 (range 0–7) and 5 (range 0–7) and the median volume of filler injected into both cheeks per session was 9 mL (range 0–12 mL) and 4 mL (range 0–30 mL) in the PLA and PH groups, respectively.
Primary outcome results are shown in Table 2. In the ITT LOCF analysis at week 48, the noninferiority of PH vs. PLA was demonstrated. Indeed, the mean VAS at week 48 reached 7.1 (SD 2.6) and 7.5 (SD 2.0) in the PLA and PH arms, respectively, with a difference (PH – PLA) of +0.4 [95% confidence interval (CI) –0.4 to +1.1; P-value for noninferiority = 0.0002]. Other analyses confirmed the noninferiority of PH vs. PLA whether we used an ITT analysis with only evaluated cases or a linear interpolation analysis for missing data. PP analyses also showed noninferiority and adjustment for baseline VAS did not affect the results (data not shown).
|PLA (n = 73)||PH (n = 75)||Treatment difference PH – PLA (95% CI)||P-value for noninferiority|
|Patients' satisfaction||Mean (SD)||Mean (SD)|
|Baseline||2.8 (2.3)||2.8 (2.3)|
|LOCF||7.1 (2.6)||7.5 (2.0)||+0.4 (−0.4; 1.1)||0.0002|
|Interpolation||7.1 (2.6)||7.5 (2.0)||+0.4 (−0.4; 1.1)||0.0002|
|Evaluated cases||7.5 (2.2)||7.7 (1.9)||+0.1 (−0.7; 1.0)||0.006|
|Per protocol population|
|LOCF||7.6 (2.1)||7.9 (1.7)||+0.2 (−0.4; 0.9)||0.0002|
|Interpolation||7.6 (2.1)||7.8 (1.7)||+0.2 (−0.5; 0.9)||0.0002|
|Evaluated cases||7.5 (2.2)||7.7 (1.9)||+0.2 (−0.7; 1.1)||0.004|
|LOCF||6.7 (2.6)||6.9 (2.4)||+0.2 (−0.6; 1.0)||0.003|
|Interpolation||6.6 (2.6)||6.9 (2.4)||+0.3 (−0.6; 1.1)||0.001|
|Evaluated cases||7.3 (2.1)||7.1 (2.4)||−0.2 (−1.0; 0.7)||0.036|
|Failure rates||n (%)||n (%)|
|LOCF||7 (10)||6 (8)||−2% (−11; 8)||0.004|
|Interpolation||8 (11)||7 (9)||−2% (−12; 9)||0.005|
|Evaluated cases||1 (2)||1 (2)||+0% (−10; 10)||0.008|
|LOCF||9 (12)||10 (13)||+1% (−10; 12)||0.028|
|Interpolation||11 (15)||9 (12)||−3% (−14; 8)||0.005|
|Evaluated cases||2 (4)||3 (6)||+2% (−9; 12)||0.020|
After week 48, re-injections were performed in 10 patients (14%) in the PLA arm and in eight patients (11%) in the PH arm (P = 0.62). At week 96, the ITT LOCF analysis also demonstrated noninferiority of PH vs. PLA. Mean VAS at week 96 were sustained at 6.7 (2.6) and 6.9 (2.4) in the PLA and PH arms, respectively, with a difference (PH –PLA) of +0.2 (95% CI –0.6 to +1.0; P-value for noninferiority = 0.003).
Consistent with the primary outcome, failure rates were noninferior in the PH arm vs. the PLA arm at both weeks 48 and 96 (Table 2). Other secondary endpoints are shown in Table 3. Changes from baseline in cheek skin-fold and thickness were similar in the PH and PLA arms at weeks 48 and 96, with the exception that the increase in cheek skin-fold at week 96 was slightly greater with PH. Also, quality of life scores using the MOS-HIV were similar in the two treatment arms up to week 96, without any significant improvement from baseline. However, the aNBC questionnaires demonstrated a significant and similar benefit of the injections in both arms.
|PLA (n = 73)||PH (n = 75)||Mean treatment difference PH – PLA (95% CI)||P-value|
|Cheek skin-fold (mm) [mean (SD)]|
|Baseline||8.7 (2.4)||8.0 (2.5)|
|Δ week 48 vs. baseline||2.1 (3.4)||2.9 (4.0)||+0.8 (−0.4; +2.0)||0.21|
|Δ week 96 vs. baseline||1.5 (3.2)||2.7 (3.9)||+1.2 (0.1; 2.4)||0.038|
|Cheek thickness (mm) [mean (SD)]|
|Baseline||7.2 (2.7)||7.1 (3.3)|
|Δ week 48 vs. baseline||2.1 (3.6)||2.3 (4.3)||+0.1 (−1.2; +1.4)||0.85|
|Δ week 96 vs. baseline||1.7 (3.6)||2.0 (4.5)||+0.3 (−1.0; +1.6)||0.63|
|MOS-HIV scores (0–100) [median (range)]|
|Physical component summary score|
|Baseline||51.5 (8.7)||49.7 (8.1)|
|Week 48||50.3 (9.6||50.7 (9.6)||0.87|
|Week 96||51.8 (7.8)||50.8 (7.1)||0.51|
|Mental component summary score|
|Baseline||47.3 (9.5)||47.5 (9.8)|
|Week 48||47.8 (10.9)||48.7 (10.4)||0.70|
|Week 96||48.3 (9.5)||46.4 (10.6)||0.36|
|Baseline||49.4 (7.9)||48.6 (7.4)|
|Week 48||49.0 (8.4)||49.7 (8.7)||0.74|
|Week 96||50.1 (7.4)||48.6 (7.6)||0.35|
|aNBC questionnaire (0–5) [median (interquartile range)]|
|Severity as evaluated by the physician|
|Baseline||1.6 (1.0–2.1)||1.6 (0.8–2.6)|
|Week 48||0.6 (0.3–1.1)||0.6 (0.3–1.2)||0.82|
|Week 96||0.6 (0.5–0.8)||0.6 (0.4–1.0)||0.77|
|Severity as evaluated by the patient|
|Baseline||2.6 (1.8–3.5)||2.8 (2.1–3.5)|
|Week 48||1.1 (0.7–2.5)||1.2 (0.5–2.2)||0.83|
|Week 96||1.5 (0.9–2.6)||1.7 (1.1–2.8)||0.49|
|Discomfort as evaluated by the patient|
|Baseline||2.5 (1.7–3.8)||2.6 (2.0–3.3)|
|Week 48||0.7 (0.3–1.5)||0.8 (0.4–1.2)||0.74|
|Week 96||1.1 (0.6–2.0)||1.3 (0.5–2.2)||0.63|
The assessment of lipoatrophy severity in randomly presented facial photographs by two independent investigators was performed in 65 patients (32 in the PLA arm and 33 in the PH arm) [the kappa score, providing an assessment of the agreement between the two assessors, was very high, at 0.98 (95% CI 0.94-0.99)]. The mean severity score was not significantly different between the two groups at baseline (P = 0.78), week 48 (P = 0.51) or week 96 (P = 0.50), although there was a significant improvement from baseline in the severity of lipoatrophy in both arms at week 48 and at week 96 (P < 0.0001 for both).
Both treatments were overall well tolerated and no adverse event led to premature discontinuation of the injections. Four patients died during the study, all in the PLA group, but none of these deaths was related to the injections. One of the patients died before receiving any injections. Two patients died during follow-up and one 2 weeks after completing the protocol follow-up. Causes of death were lung cancer (two patients), hepatocellular carcinoma and mesenteric ischaemia. Serious adverse events were also reported in 26 patients in the safety population (i.e. those who received at least one injection) (Table 4) but were related to the injections in only four patients, who presented with facial inflammatory lesions following PH injections (see below).
|PLA (n = 69)||PH (n = 70)||P-value|
|Early adverse events (during injections)|
|Pain [n (%)]||66 (100)||68 (100)||> 0.99|
|Pain score (0–10) [mean (range)]||3 (0; 7)||4 (1; 8)||0.001|
|Bleeding [n (%)]||4 (6)||7 (10)||0.53|
|Grade 3–4||0 (0)||0 (0)||> 0.99|
|Haematoma [n (%)]||6 (9)||8 (11)||0.78|
|Grade 3–4||0 (0)||0 (0)||> 0.99|
|Vagal hypertonia [n (%)]||2 (3)||4 (6)||0.68|
|Grade 3||0 (0)||1 (1)||> 0.99|
|Grade 4||0 (0)||0 (0)||> 0.99|
|Other [n (%)]||4 (6)||6 (9)||0.74|
|Late adverse events (following injections)|
|Oedema [n (%)]||2 (3)||4 (6)||0.68|
|Haematoma [n (%)]||7 (10)||11 (16)||0.45|
|Pain [n (%)]||7 (10)||11 (16)||0.45|
|Subcutaneous nodules [n (%)]||28 (41)||26 (37)||0.14|
|Maximal size (mm) [mean (min; max)]||4 (0.5; 15)||4 (1; 30)||0.089|
|Serious adverse events (SAEs)|
|Late-onset inflammatory lesions [n (%)]||0||4 (6)||0.12|
|At least one SAE unrelated to injection [n (%)]||13 (19)||9 (13)||0.36|
|Acute vascular event||3||2|
|Death [n (%)]||3 (4)*||0||0.12|
Bleeding and haematoma at the injection site, vagal hypertonia during injections and oedema post-injections were the most frequently reported adverse events, and were observed in the two arms at similar rates (Table 4). Most events were mild or moderate, with only one grade 3 (vagal hypertonia). Injections of PH were significantly more painful than injections of PLA, which was mixed with lidocaine. Subcutaneous nodules at the injection site were reported by the investigator in 41% and 37% of cases in the PLA and PH arms, respectively (P = 0.14). These nodules were small, with mean maximal size 4 mm in both arms. The nodules were reported by patients at similar rates (46% and 40% in the PLA and PH groups, respectively) (data not shown). In four patients, all randomized in the PH arm, large inflammatory lesions at the injection site, of 20 to 40 mm in diameter, developed after a median of 17 months (range 15 to 23 months) after the last injection. None of these four patients had any dental or invasive procedure in the days or weeks before the occurrence of inflammatory lesions. Three patients received antibiotics, with surgical drainage in two cases and spontaneous drainage in one. Massage alone was used in the fourth case, and eventually resolution of these inflammatory lesions was reported in all cases. Of note, no injection-related infections were reported in this trial.
In this prospective, randomized, multicentre study we demonstrated the noninferior efficacy of PH as compared with PLA dermal injections for the treatment of lipoatrophy in HIV-infected adults receiving antiretroviral therapy. In this study, patients were blinded to the allocated filler. Despite the difference in injected volumes and pain related to the injections between the two groups, as a result of the different textures of the fillers and the use of lidocaine in the PLA group only, the blinding was unlikely to have been affected as patients were naïve to treatment. Efficacy was assessed using patients' satisfaction with a VAS as previously described . Patients' satisfaction increased from a mean VAS of 2.8 at baseline to a mean score of 7.1 and 7.5 in the PLA and PH groups, respectively, at week 48. The treatment difference (PH – PLA) was +0.4 (95% CI –0.4 to +1). Noninferior efficacy of PH was demonstrated, as the CI of the treatment difference between arms was greater than the predefined threshold of –1. Noninferiority at week 48 was also demonstrated using a number of sensitivity analyses, in both ITT and PP analyses. Similarly, failure rates, which remained < 10%, were also noninferior in the PH arm as compared with the PLA arm (Table 2). Secondary efficacy endpoints (change from baseline in cheek skin-fold and thickness measures, and severity of lipoatrophy assessed using facial photographs) also confirmed the similar efficacies of the two fillers. Only the difference in cheek skin-fold at year 2 may suggest that PLA was somewhat inferior, as cheek skin-fold declined in this group between weeks 48 and 96 but was more stable in the other group (Table 3). However, the reproducibility of skin-fold and cheek thickness evaluation is uncertain, precluding any definitive explanation for this single difference between arms at 2 years. Also, cheek thickness did not vary between the two groups in the second year of follow-up. PH could therefore represent a suitable alternative to PLA for the treatment of facial lipoatrophy in patients with HIV infection. However, PH was not superior to PLA, and although noninferiority criteria were also met at week 96 whatever the endpoint, there was a trend towards a decreased efficacy with both fillers over time (Table 2). Indeed, in the two arms similar proportions of patients needed re-injections after week 48 (14 and 11% in the PLA and PH arms, respectively), and the rates of failure slightly increased from week 48 to week 96. These data suggest that other alternatives should be assessed for patients with lipoatrophy and in particular for those failing either PLA or PH treatment. Unfortunately, there have been very few randomized studies such as ours to assess treatment strategies for HIV-related lipoatrophy. The efficacy of PLA for the treatment of lipoatrophy in HIV-infected individuals has been reported in several studies [14-16, 26-30]. The experience with PH is more limited, but observational studies have reported a lower rate of subcutaneous nodules and need for re-injections as compared with PLA [17-20, 31, 32]. Only one previous randomized study has compared PLA with PH . In that study, Guaraldi et al. reported that the efficacies of PLA and PH were similar in improving patients' aesthetic satisfaction, with a good safety profile and fewer patients developing subcutaneous nodules with PH as compared with PLA (none vs. nine, respectively). However, only 50 patients were enrolled in that study with a short follow-up time of only 24 weeks. That study therefore had inadequate power to show noninferiority and detect rare safety events. Other studies, such as that by Negredo et al., also reported similar efficacies for AFT, PH and PLA, with PH providing the longest lasting benefit . However, that study was not randomized and the number of patients in each group was unbalanced (AFT, n = 8; PLA, n = 25; PH, n = 105), and therefore conclusions based on the findings of that study must be made with caution .
Our trial also provided the opportunity to assess the short- and long-term safety of these two fillers. Although both treatments were generally well tolerated with no treatment discontinuation because of an adverse event, a small proportion of patients presented with mild to moderate adverse events at the time of facial injections, with no significant difference between arms. Interestingly, subcutaneous nodules at the injection site were a frequent complaint and were detected at similar rates in patients in the PLA and PH arms (41% and 37%, respectively; P = 0.14). These nodules were detected only by digital palpation and were not symptomatic. Such nodules have been reported in most studies using dermal fillers, with lower rates usually being found among those receiving PH [14-16, 26-30]. Of note, the long-term follow-up of these patients allowed the detection of late-occurring inflammatory nodules at the injection site. Although these nodules were only reported in four patients (6%) after a median follow-up time of 17 months, they all occurred in patients receiving PH. Such inflammatory lesions have previously been reported in a few patients receiving PH, up to 6 years after the injections [17, 33-37]. These data call into question the long-term safety of these dermal fillers and justify the reporting of these adverse events to pharmacovigilance to obtain a better idea of their frequency.
This study has, of course, several limitations. First, there is no gold standard for assessing the efficacy of treatment strategies for HIV-associated lipoatrophy. In this study we used patients' satisfaction based on a VAS to quantify the efficacy of treatments as reported in previous studies [14, 15, 18, 19]. Although not a hard endpoint, we feel this is a relevant assessment of patients' satisfaction in particular, as patients were blinded to the fillers they received. Different endpoints have been used in other studies, but here we used a number of secondary efficacy endpoints (rate of failure, changes in cheek skin-fold and thickness, and lipoatrophy scores using photographs) that all confirmed the efficacies of both treatments and the noninferiority of PH as compared with PLA. Quality of life assessments using the MOS-HIV questionnaire remained unchanged during the course of the study, probably because this questionnaire does not capture appropriately the change in facial lipoatrophy induced by filler injections, as previously reported . The aNBC questionnaire, however, showed improved patient satisfaction with both fillers herein discomfort only as stated in the text. Other quality of life questionnaires, such as the Adult AIDS Clinical Trial Group Assessment of Body Change and Distress (ABCD), might be more appropriate and have shown the benefit of filler injections in a number of studies [18, 19].
Finally, one of the major weaknesses of the study was the high number of patients who were lost to follow-up or who did not attend study visits on schedule. It was a challenge in this trial to obtain good patient compliance with scheduled visits, particularly when injections were not performed. This explains why only 87 of the 148 patients were evaluated at week 48 (59%) and 104 (70%) at week 96 (Fig. 1). To reduce potential bias, we performed a number of sensitivity analyses for missing values using last observation carried forward analyses, interpolation between two visits, and analyses on evaluated cases only, both in ITT and in PP analyses. It was therefore reassuring that all analyses were consistent with each other and all found noninferiority of PH vs. PLA.
In summary, in this randomized multicentre study, we assessed two treatments for patients with HIV-related lipoatrophy, PLA and PH. Both fillers improved patients' satisfaction, which was sustained up to week 96, and we demonstrated that the efficacy of PH was noninferior to that of PLA, suggesting that PH represents a suitable alternative to PLA. The safety of both treatments was acceptable overall, although a sizable proportion of patients developed small subcutaneous nodules at the injection site. Of concern is the late occurrence of large inflammatory nodules at the injection site in a few patients receiving PH, which in some cases required a surgical procedure. More studies are therefore needed to assess other treatment strategies, in particular in patients failing either PLA or PH treatment, and to monitor the long-term safety of these treatments.
We are grateful to all the patients who participated in the trial. We are also grateful to the participating centres: Hôpital Ambroise Paré, Boulogne (Prof. P. Saiag and Dr M.-L. Bafounta), Hôpital Avicenne, Bobigny (Prof. O. Bouchaud, Dr M. Bentata and Dr D. Bakhos), Hôpital Bicêtre, Le Kremlin Bicêtre (Prof. J.-F. Delfraissy and Dr F. Blanc), Hôpital de Brabois, Nancy (Prof. T. May and Dr E. Allamagny), Cannes Hospital, Cannes (Dr Bodokh), Hôpital Cochin/Tarnier, Paris (Prof. N. Dupin and Dr M. Burnouf), Hôpital Edouard Herriot, Lyon (Prof. A. Claudy and Dr E. Carbonnel), Hôpital Henri Mondor, Créteil (Prof. Y. Levy and Dr E. Domergue), Hôpital La Grave, Toulouse (Dr L. Virben and Dr F. Lucas), Hôpital Paul Brousse, Villejuif (Prof. D. Vittecoq and Dr P. Minoui), Hôpital Saint-Antoine, Paris (Prof. P.-M. Girard, Dr M. Volstein and Mr J.-L. Lagneau), and Hôpital Saint-Louis, Paris (Prof. J.-M. Molina, Dr M. Lafaurie, Dr M. Dolivo, Ms B. Loze and Ms C. Waquet). The study was funded and sponsored by the Agence Nationale de Recherche sur le SIDA et les Hépatites Virales (ANRS). We would like to thank in particular Mrs M.-J. Commoy and Ms S. Couffin-Cadiergues for their help in conducting this trial. We thank Mrs H. Ghorbel and Ms C. Bonfil-Praire from the Saint-Louis Biostatistics Department for monitoring the study. We also thank Sanofi and Procytech for graciously providing polylactic acid and polyacrylamide hydrogel for the study. We thank the members of the Data Safety Monitoring board for trial supervision (Mr F. Rodenburg, Dr D. Costagliola, Dr C. Lebbe and Dr C. Lascoux).
Conflicts of interest: MD is currently a consultant for Sanofi. All remaining authors declare no conflict of interest.