3.2.1. Description of Studies

LIPO-010³ (N = 412) and CTR-1011⁵ (N = 404) were similarly designed multi-centre, DB, placebo-controlled, phase 3 RCTs (Table 3). Both trials enrolled participants from US and Canada, although CTR-1011 additionally enrolled participants from Europe. In both studies, tesamorelin (2 mg/day) was compared (in a 2:1 ratio) against placebo. Each trial consisted of a 26-week main phase, followed by a 26-week extension phase (the extension phase of CTR-1011 was denoted CTR-1012) (0). Across both trials, randomization was conducted using an interactive voice response system (IVRS), and stratified by testosterone use and impaired glucose tolerance (IGT)/diabetes condition in LIPO-010, and by diabetes status in CTR-1011. In each trial, the primary efficacy outcome was the reduction in VAT at week 26.

The study by Stanley et al. 2014 (N = 54) was a single-centre, DB, placebo-controlled RCT (Table 3).² The study exclusively enrolled participants from the US. Eligible participants underwent two independent randomization processes: a DB 1:1 randomization to tesamorelin (2 mg/day) versus placebo; and a 1:1 randomization to undergo hyperinsulinemic euglycemic clamp in addition to other study procedures (no details provided). The authors did not provide details about the randomization process, although they indicated that it was stratified by sex and, for men, by physiologic testosterone use. The trial was initially designed to evaluate the reduction in VAT at six months as the primary end point; however, prior to trial initiation, the study investigators added hepatic fat as a co-primary end point.

3.2.2. Populations

b) Baseline characteristics

Across all three studies, study participants were predominantly male and white/Caucasian (Table 4). The mean age of participants in LIPO-010 and CTR-1011 was approximately 47 years. Baseline demographic and anthropometric measurements of participants in LIPO-010 and CTR-1011 were generally well-balanced; one exception is that participants in Stanley et al. 2014 appeared to have the greatest mean levels of VAT at baseline, followed by those in CTR-1011, then those in LIPO-010. The clinical expert consulted by CDR for the purpose of this review hypothesized that the differences may be due to enrollment of participants from centres with more PI use or the studies may date to a time when there was a longer mean time of PI use, leading to more lipohypertrophy. The expert noted that the baseline VAT levels might “mean little” with respect to effects on treatment response.

Table 4

Summary of Baseline Characteristics (Main Phase).

In LIPO-010, participants assigned to the tesamorelin group appeared to have been receiving ART for approximately six months longer (on average) than those assigned to placebo, although the effects on study results would be inconsequential per the clinical expert. Across the three trials, there appeared to be some imbalances in the type of current ART regimens that participants were receiving. In LIPO-010, for instance, a smaller percentage of participants receiving tesamorelin reported taking an ART regimen that included a PI than those receiving placebo (55% versus 64%). Moreover, unlike in LIPO-010 and CTR-1011, fewer than half of participants in Stanley et al. 2014 were on a PI. Across all trials, more than 90% of participants reported taking a nucleoside reverse transcriptase inhibitor (NRTI); the expert noted that this trend is consistent with the standard of ART everywhere. In CTR-1011, the mean duration of PI in both treatment arms was approximately three years, although the degree to which this affects severity of the lipohypertrophy or treatment response is unclear. Last, in LIPO-010 and CTR-1011, all participants suffered from abdominal lipohypertrophy, while more than half also suffered from general lipoatrophy; the expert did not think that the presence or absence of lipoatrophy would affect treatment response. Overall, the expert noted that any observed inequities among and between trials were minor and unlikely to substantially affect treatment response.

3.2.3. Interventions

Across all three trials, tesamorelin 2 mg was compared against placebo. In LIPO-010 and CTR-1011, tesamorelin was administered in 2 mL volumes, and by subcutaneous injection in the abdomen.^3,5 The injections were self-administered, although there were no details provided about the degree to which participants were trained, or whether individuals who could not self-administer were screened out. Blinding was maintaining by using placebo that was identical in appearance to tesamorelin. Stanley et al. 2014 did not provide details about the treatment and blinding procedures, except that the study was DB.² In that study, additionally, the authors indicated that participants receiving tesamorelin or placebo were independently randomized to undergo a hyperinsulinemic euglycemic clamp. Specifically, these participants received (following a 12-hour fast) regular insulin (400 mU × m² × min⁻¹) for two minutes, after which (for 118 minutes) they received a continuous infusion of 80 mU × m² × min⁻¹ regular insulin.² The authors indicated that an infusion of 20% dextrose was adjusted to maintain plasma glucose concentration at 11 mmol/L (90 mg/dL). The consulting clinical expert noted that this procedure is a “research intervention and unavailable and unused in clinical medicine.”

3.2.4. Outcomes

3.2.5. Statistical Analysis

In LIPO-010 and CTR-1011, the sample size was based on a difference of 8% in the change in VAT at week 26 between tesamorelin and placebo, a standard deviation (SD) of 18.5%, power of 90%, significance level of 5%, and a distribution ratio of 2:1.^3,5 Dropout rates of 33% and 25% were assumed in LIPO-010 and CTR-1011, respectively; no rationale for the different expected dropout rates was provided. The threshold of 8% was considered to be the minimum difference needed to detect a clinically relevant difference between tesamorelin and placebo. The manufacturer notes that this difference was recommended (for LIPO-010) and confirmed (for CTR-1011) at a post–phase 2 meeting with the FDA.³ The choice of this cut-off is discussed further in section 0. Given the abovementioned assumptions, the required total sample sizes were 381 (254 tesamorelin, 127 placebo) and 340 in LIPO-010 and CTR-1011, respectively.^3,5

Across both trials, the primary efficacy outcome was the per cent change in VAT from baseline to week 26 in the intention-to-treat (ITT) population; missing data were imputed using the last observation carried forward (LOCF) method, whereby baseline values were carried forward into the treatment period. A secondary analysis was conducted in the per-protocol (PP) population and featured an observed case method, in which all non-missing observations were used.

In LIPO-010, the treatment effect was tested using an analysis of covariance (ANCOVA) model that featured treatment (study group) as a fixed effect and baseline VAT as a covariate.³ In CTR-1011, a similar ANCOVA model was used, except that it also included centre as a covariate.⁵ Across both trials, several pre-specified supportive analyses for the primary efficacy outcome were conducted. In LIPO-010, relevant analyses included a baseline interaction analysis, which used the same ANCOVA model as above but with a baseline-by-treatment interaction term, as well as a covariate analysis that evaluated the impact of the type of ART regimen followed by participants during the study. Specifically, the per cent change in VAT at week 26 was examined by several categories of ART regimen — NRTI/non-nucleoside reverse transcriptase inhibitor (NNRTI), NRTI/PI, NRTI/NNRTI-PI, NRTIs alone, and other — and the statistical significance of the ART regimen and treatment-by-ART regimen interaction were tested. The authors also appeared to conduct a post-hoc analysis examining the effects of NNRTI and the NNRTI-by-treatment interaction in a separate model after noting a statistically significant difference in the percentage of participants between the two treatment groups whose ART regimen included an NNRTI. In CTR-1011, an analysis examining the effect of any changes in ART was planned, although because most (87%) participants did not have any changes in their ART throughout the study, this analysis was not performed. Instead, an exploratory analysis evaluating the impact of the type of ART regimen followed during the study was conducted in a similar manner as in LIPO-010.

As with the primary efficacy outcome, analyses for secondary and other efficacy outcomes were primarily conducted on the ITT population using LOCF analysis, and secondarily on the PP population using observed case analysis.^3,5 To evaluate waist circumference in LIPO-010 and CTR-1011, an ANCOVA model adjusted for baseline waist circumference and treatment was used. Across both trials, the pre-specified statistical analysis plan for HRQoL and body image parameters indicated a parametric ANCOVA to analyze belly size evaluation and belly appearance distress, and a Mann-Whitney test for belly profile evaluation. Discussions between the manufacturer and the FDA, however, resulted in changes in the plan, with which the FDA agreed.¹⁸ In particular, both groups decided that for any ongoing phase 3 studies, (1) non-parametric (ranked) ANCOVAs for all three body image parameters could be the primary analyses; and (2) parametric ANCOVAs (belly appearance distress and belly size evaluation) and the Mann-Whitney test (belly profile evaluation) must be supportive analyses. Conversely, for the completed phase 3 study: (1) parametric ANCOVAs (belly appearance distress and belly size evaluation) and the Mann-Whitney test (body profile evaluation) must be the primary analyses for the first six months; and (2) non-parametric (ranked) ANCOVAs for all three body image parameters should be supportive analyses for the first six months.¹⁸ At the time of these discussions, the main phase of LIPO-010 had been completed, while CTR-1011 was ongoing. Given the bidirectional responses for body size evaluation, the non-parametric ANCOVA was performed on the change score calculated as the negative value of (absolute [end point] – absolute [baseline]), which yielded positive scores for participants moving toward improvement, negative scores for those moving toward worsening, and 0 for those who stayed the same distance from normal. The manufacturer highlighted that the change scores presented per the parametric ANCOVA models should be considered with caution as they did not take into consideration the direction of the change. Across LIPO-010 and CTR-1011, the manufacturer also conducted two “supportive” analyses for the body image parameters: an MID analysis and a responder analysis, the latter of which was “an anchor-based method based on each subject’s judgment of the smallest amount of improvement in belly profile that he/she would find to be beneficial (the ‘minimally important benefit’) in the profile analysis.”³ Values of the thresholds used to define responders are presented in Table 5.

Table 5

Responder Thresholds in LIPO-010 and CTR-1011.

In CTR-1011, treatment effects on the EQ-5D utility and visual analogue scale (VAS) scores were tested using linear mixed models that featured treatment and study visits time-point as fixed effects, and participant as random effects. Further, baseline EQ-5D utility and VAS scores, age, and gender were included in the model as covariates, and retained if significant. No details about the manner in which missing data were handled were provided.

Discussions with the FDA and the manufacturer also resulted in the creation of a “gatekeeper” strategy for the secondary efficacy outcomes to control the type 1 error in LIPO-010 and CTR-1011 (Table 6).¹⁸

Table 6

Gatekeeper Strategy for LIPO-010 and CTR-1011.

Rankings of the secondary efficacy outcomes in the gatekeeper strategy were modified between studies, based on requests from the FDA, thus explaining the differences in Table 6. Across both trials, the outcomes in the gatekeeper strategy were considered for analyses only if the primary efficacy outcome and the preceding secondary efficacy outcomes were found to be statistically significant. All statistical tests were two-sided, with alpha pre-specified at 0.05, and interactions were tested at alpha pre-specified at 0.10.

The study by Stanley et al. 2014 was planned to enrol 60 participants, with 48 individuals estimated to complete the study; this calculation assumed 80% power to detect a treatment effect of 16.5% change in VAT.² The authors did not provide a rationale for the anticipated treatment effect. However, they did indicate that study recruitment terminated prematurely as a result of drug supply issues, resulting in 43 participants completing the study. Post-hoc power calculations revealed that, with 43 participants and new data (regarding the SD of change in VAT) from a pooled analysis of LIPO-010 and CTR-1011, the study had 85% power to detect a treatment difference of 38.5 cm² in the change in VAT at a two-sided alpha level of 0.05. The co-primary end points of this study were change in VAT and hepatic fat at six months. Between-group treatment effect (related to the change in VAT) was evaluated using the Student t-test. Primary analyses were conducted using all available data, with missing data treated as missing. Sensitivity analyses, which featured an imputation method, were used to corroborate results from the primary analyses; specifically, for variables that were normally distributed, which was the case for the change in VAT, missing values were replaced with imputed values that were generated using longitudinal mixed-effects modelling. All analyses were two-sided, with alpha pre-specified at 0.05.

3.5.1. Internal Validity

All three trials were DB and placebo-controlled; LIPO-010 and CTR-1011 featured appropriate randomization and allocation concealment processes, while the study by Stanley et al. 2014 did not present relevant methodological details, thus precluding an assessment of the associated risk of bias, and leaving uncertain the validity of the results. In LIPO-010 and CTR-1011, the CT scans (for VAT) were centrally reviewed and analyzed in a blinded fashion, although no details were provided about the number of individuals who examined the images, as well as any training or calibration procedures they underwent; the uncertainty regarding the reproducibility of the findings might undermine the confidence in the results. The authors of Stanley et al. 2014 did not provide information about the manner in which the CT scans were reviewed, although they indicated that, for the measurement of VAT, previous research demonstrates that single-slice CT has an estimated correlation between repeat measurements of 0.99, with errors in precision estimated at 3.9%.⁵ If true, these values increase the confidence in the effects of tesamorelin on VAT.

Baseline characteristics were generally similar across treatment groups in all trials, with few differences. For instance, mean VAT levels were different across the three trials; however, discussions with the clinical expert suggested that these could be due to variations in clinical management of patients arising from temporal and/or geographical differences between the studies. Furthermore, compared with participants in CTR-1011, a smaller percentage of participants in LIPO-010 had undetectable viral loads, thus indicating that they were less healthy. It is plausible that treatment effects are artificially magnified in a less healthy population, thus making the treatment appear to be better than reality. Overall, the clinical expert consulted by CDR for the purpose of this review noted that any observed inequities between trials were minor and unlikely to substantially affect treatment response.

Both LIPO-010 and CTR-1011 used an ANCOVA model to evaluate the primary efficacy outcome; i.e., the per cent change in VAT between tesamorelin and placebo. The model included treatment as a fixed effect and baseline VAT as a covariate; in CTR-1011, an additional covariate (centre) was included in the model, although no rationale was provided. The results pertaining to this outcome presented in this report were extracted from the FDA statistical review, which appeared to generate the results (for both studies) using an ANCOVA model with treatment as fixed effect and baseline VAT as covariate. It was unclear why the FDA statistical review did not include centre as a covariate in the CTR-1011 analyses, although the results were consistent with those presented by the manufacturer.

LIPO-010 and CTR-1011 did not use a true ITT population; rather than including all randomized participants, the defined ITT analysis sets across both studies included only those participants who took at least one dose of the assigned study drug; i.e., a modified ITT population. Not adhering to the true ITT principle threatens the prognostic balance that a successful randomization process creates, thus leaving uncertain the validity of the results. Nevertheless, examining the participant disposition (Table 7) suggests that relatively few participants who were randomized did not contribute to the ITT analysis set: in LIPO-010, two participants (0.73%) in the tesamorelin group, and none in the placebo group; in CTR-1011, five participants (1.82%) in the tesamorelin group, and three participants (2.3%) in the placebo group. Missing data were imputed using the LOCF method, whereby baseline values were carried forward into the treatment period. However, carrying the last observation forward may have artificially stabilized VAT levels among participants who dropped out; conversely, observed data could also be biased if the probability of withdrawal is related to an increase in VAT levels. Although the number of participants for whom data were imputed was unclear in both trials, there did appear to be a large difference in the number of individuals contributing to the ITT versus PP analysis sets, which possibly suggests a substantive amount of imputation, and ultimately may limit the validity of the results (Table 7). Further, in CTR-1011, it was unclear how missing data were handled for the analysis of EQ-5D data, which was analyzed using a linear mixed model, and for which as much as 35% of participants did not contribute to the analysis — missing such a substantive portion of the trial sample leaves uncertain the robustness of the results, and increases the uncertainty of the magnitude of the treatment effects on EQ-5D. The study by Stanley et al. 2014 indicated using a modified ITT population among participants with available baseline and six-month follow-up data. Excluding participants from analyses may not preserve the integrity of randomization, especially if the percentage of exclusions is large, thus limiting the validity of the results. In Stanley et al. 2014, five participants (17.9%) randomized to tesamorelin, and six participants randomized to placebo (23.1%) did not contribute to the primary analysis. The authors of the study imputed data for participants with missing observations, and found the results to be consistent with the analysis of available data. Still, the authors did not account for four of the 26 participants (15.4%) who were excluded after randomization to placebo, of whom three declined participation prior to the baseline visit, and one developed an exclusionary pre-condition. Not accounting for these participants might mean that, unlike the randomized set of participants, the analyzed set is no longer prognostically balanced, which could undermine confidence in the results.

In LIPO-010 and Stanley et al. 2014, the rates of discontinuation appeared to be disproportional across the treatment groups, with more participants in the tesamorelin arm discontinuing from the study than those receiving placebo. As above, these imbalances might have disrupted the prognostic balance that successful randomization creates, thus leaving uncertain the validity of the results. Further, across both studies, the most common reason for withdrawal was AE. Due to the unique AEs associated with tesamorelin — e.g., injection-site reactions, arthralgia, and myalgia — participants may have been inadvertently unblinded, which itself could have affected their behaviour in the trials, and consequently impacted their responses to subjective outcomes, such as QoL.

LIPO-010 and CTR-1011 were powered based on a difference of 8% in the change in VAT at week 26 between tesamorelin and placebo, while the study by Stanley et al. 2014 was based on an estimated difference of 16.5%. The cut-off of 8% appears to have been derived from the 2004 Forum for Collaborative HIV Research, which, according to the manufacturer, established an “expected decline” of 8% in VAT for patients with HIV-associated lipodystrophy receiving an rhGH product in clinical trials of up to 26 weeks in duration.¹⁸ A Scientific Advisory Panel convened by Health Canada acknowledged that an 8% decrease in VAT is a clinically important short-term effect, although Health Canada noted that the cut-off was “largely arbitrary,” and “reportedly a ‘hybrid’ end point derived from the VAT loss seen in a previous [phase] 2 study of the rhGH product and the 5% total weight loss target the FDA has adopted for development of drugs for treating obesity.”²¹ The authors of Stanley et al. 2014 did not provide a rationale for estimating a treatment effect of 16.5%. Although all studies were adequately powered to evaluate the primary efficacy outcome, none of the trials were powered to assess secondary efficacy outcomes or for harms outcomes.

In LIPO-010 and CTR-1011, to help contextualize changes in body image, the manufacturer reports having derived an MID for each of the three parameters from previous lipodystrophy studies. However, it does not provide details about the specific studies to which it referred, thus leaving uncertain the validity of the chosen MID values, as well as the degree to which the results are clinically meaningful. It also reports that the instrument used to measure body image was validated,²³ although the FDA had several concerns about this claim.²⁴ In particular, the FDA’s Study Endpoint and Labeling Development (SEALD) team concluded that the instrument had “questionable content validity,” and the team raised concerns about the instrument’s ability to measure a clinically important treatment effect. The SEALD team specifically reported that the manufacturer did not address whether qualitative research was done to evaluate patient understanding of the final instrument. As a result, the instrument did not meet the standards for instrument development as recommended by the FDA, nor did it meet the standard for evidence saturation. The SEALD team also reported that internal consistency reliability data had limited relevance because single items were utilized as end points instead of summary scales. In addition, the team stated that the enrolment criteria in the trials did not pre-specify a minimum PRO score, which could have made it more difficult to demonstrate improvement. Further, with respect to the instrument used to measure QoL, the manufacturer provided only limited information regarding the interpretation of scores, and no information about the psychometric properties of the instrument, which leaves uncertain the validity of the results.

Across LIPO-010 and CTR-1011, several pre-specified supportive analyses were conducted, including testing for numerous covariate-by-treatment interactions, although no adjustments for multiple comparisons appear to have been made for these analyses. A further limitation is that, in CTR-1011, the manufacturer appeared to conduct a post-hoc analysis examining the NNRTI-by-treatment interaction; a post-hoc analysis decreases the credibility of the analyses and limits certainty in the results. Moreover, both trials used gatekeeping testing strategies that, per the manufacturer, were created for labelling purposes and to help control the experiment-wise type 1 error rate at a one-sided level of 0.05. This is a common and appropriate strategy to account for multiplicity. Only the changes in VAT and belly appearance distress were considered in the testing strategy, which limits the ability to interpret the analyses of outcomes outside the gatekeeping procedure. Without controlling for multiplicity, analyses of all outcomes other than change in VAT and belly appearance distress, as well as the subgroup analyses, should be considered hypothesis-generating and interpreted with caution. No adjustments for multiplicity were performed in Stanley et al. 2014, including for the co-primary end point, although the authors did not evaluate outcomes of interest other than change in VAT.

3.5.2. External Validity

Discussions with the clinical expert consulted by CDR for the purpose of this review highlighted that generalizability of the findings of the three trials is a major concern. Chiefly, the studies appear to have been conducted at a time when HIV patients were commonly receiving ART regimens that were associated with accumulation of visceral fat. The expert indicated that, in today’s clinical practice, HIV patients are substantially different from those who were enrolled in the three studies, and are much more likely to receive ART regimens that consist of backbones other than PIs, including integrase strand transfer inhibitors (INSTIs) and NNRTIs. To this end, of the six regimens recommended by the US DHHS to manage ART-naive patients, five are INSTI-based, and one is ritonavir-boosted PI (PI/r)–based:¹¹

INSTI-based regimens:

• Dolutegravir/abacavir/lamivudine (DTG/ABC/3TC)
• DTG plus tenofovir disoproxil fumarate (TDF)/emtricitabine (FTC)
• Elvitegravir (EVG)/cobicistat (COBI)/tenofovir alafenamide (TAF)/FTC
• EVG/COBI/TDF/FTC
• Raltegravir (RAL) plus TDF/FTC.

PI/r-based regimen:

• Darunavir/ritonavir plus TDF/FTC.

In Canada, three of the abovementioned preferred regimens are available as single-tablet regimens (STRs): Genvoya (EVG/COBI/FTC/TAF), Stribild (EVG/COBI/FTC/TDF), and Triumeq (DTG/ABC/3TC). The three remaining preferred regimens are available as multi-tablet regimens consisting of a two-drug backbone; e.g., FTC/TDF (Truvada) plus a third drug. STRs are preferred to multi-tablet regimens given their convenience advantage, which results in greater adherence and optimal treatment response. There are two additional STRs available, both of which are NNRTI-based, although they are listed as “alternative” regimens by the DHHS: Atripla (efavirenz/TDF/FTC), and Complera (rilpivirine /TDF/FTC). In LIPO-010, the number of participants receiving ART regimens that comprised integrase inhibitors was not reported, whereas fewer than 5% of participants in CTR-1011 were receiving INSTI-based therapies. In Stanley et al. 2014, the number of participants on integrase inhibitors was unclear, as the authors reported that seven individuals in the tesamorelin group (25.0%) and six in the placebo group (27.3%) were receiving entry inhibitors and integrase inhibitors.

Moreover, as a result of the large number of treatments available today, single or multiple substitutions of the components of an ART regimen offer optimal virologic suppression with fewer AEs, such as the accumulation of visceral fat. The clinical expert consulted by CDR for the purpose of this review indicated that fewer than 5% of patients in his HIV practice are affected by serious fat deposition.

Further, although the manufacturer tested the statistical significance of treatment-by-ART regimen interactions in LIPO-010 and CTR-1011, the results may be limited by the fact that individual drugs within a class were grouped and analyzed together.

Examination of the inclusion and exclusion criteria of the included studies reveals additional factors that may limit the generalizability of the findings. First, the authors of Stanley et al. 2014 provided limited demographic information about the participants enrolled in their study, thus precluding an adequate assessment of the generalizability of the study findings to a Canadian population. With respect to LIPO-010 and CTR-1011, the exclusion of certain subgroups of participants — those with type 1 diabetes and type 2 diabetes in LIPO-010, and those treated with oral antidiabetic drugs or insulin LIPO-010 and CTR-1011 — leaves uncertain the effects of tesamorelin in patients with HIV-associated lipohypertrophy who present with certain comorbidities. Diabetes is particularly important given concerns raised by the FDA that more patients treated with tesamorelin developed glucose intolerance, and had a higher risk of developing diabetes across the clinical development program.¹⁵

Finally, the durations of all three trials were insufficient (even after considering the 26-week extension phases in LIPO-010 and CTR-1011) to adequately capture some important safety outcomes, including the occurrence of diabetes and cancer, thus leaving uncertain the long-term safety profile of tesamorelin.

3.6.1. Visceral Adipose Tissue

Across all three trials, tesamorelin was associated with a statistically significantly greater reduction in VAT versus placebo (Table 8). Specifically, with respect to the per cent change in VAT at week 26, the LS mean differences (95% CI) of tesamorelin versus placebo were −19.6% (−23.7% to −15.3%) in LIPO-010, and −11.7% (−16.2% to −7.1%) in CTR-1011. In Stanley et al. 2014, the mean difference (95% CI) of the per cent change in VAT at six months was −16.6% (−30.6% to −2.6%).

Table 8

Key Efficacy Outcome — Visceral Adipose Tissue (Main Phase).

3.6.2. Waist Circumference

In LIPO-010 and CTR-1011, tesamorelin was associated with a statistically significantly greater reduction in waist circumference at 26 weeks versus placebo (Table 9). Specifically, the absolute differences (95% CI) of tesamorelin versus placebo were −1.8 cm (−2.8 cm to −0.9 cm) in LIPO-010, and −1.3 cm (−2.4 cm to −0.2 cm) in CTR-1011. Waist circumference was not evaluated in Stanley et al. 2014.

Table 9

Key Efficacy Outcome — Waist Circumference (Main Phase).

3.6.3. Body Image

In LIPO-010 and CTR-1011, there were no statistically significant differences between treatment groups with respect to change in belly size evaluation at week 26 (Table 10). Across both trials, the effects of tesamorelin versus placebo on change in belly appearance distress at week 26 were inconsistent: per the parametric ANCOVA models, there were no statistically significant differences between treatment groups; conversely, per the ranked (non-parametric) ANCOVA models, tesamorelin was associated with a statistically significantly greater reduction in belly appearance distress at week 26 versus placebo. Further, with respect to belly profile ratings, tesamorelin was associated with a statistically significantly greater reduction in belly dysmorphia versus placebo in LIPO-010 but not in CTR-1011. Body image was not evaluated in Stanley et al. 2014.

Table 10

Key Efficacy Outcome — Body Image (Main Phase).

3.6.4. Quality of Life and Health-Related Quality of Life

In LIPO-010 and CTR-1011, there were no statistically significant differences between treatment groups with respect to the change in mean overall QoL scores (item-wise) at week 26 (Table 11). Further, in CTR-1011, there were no statistically significant differences between treatment groups with respect to EQ-5D index scores and VAS and health-state ratings at week 26. QoL was not evaluated in Stanley et al. 2014.

Table 11

Key Efficacy Outcome — Quality of Life and Health-Related Quality of Life (Main Phase).

3.7.1. Adverse Events

Across all three studies, at least 70% of study participants in each trial experienced a treatment-emergent AE (Table 12). A greater percentage of participants in Stanley et al. 2014 experienced an AE (tesamorelin: 89.3%; placebo: 95.5%) than those in LIPO-010, followed by those in CTR-1011. Further, in LIPO-010 and CTR-1011, approximately 5% more participants receiving tesamorelin (LIPO-010: 82.8%; CTR-1011: 74.1%) experienced an AE than those on placebo (LIPO-010: 75.9%; CTR-1011: 69.8%).

Table 12

Harms (Main Phase).

3.7.2. Serious Adverse Events

Overall, a greater percentage of participants in Stanley et al. 2014 experienced an SAE (tesamorelin: 10.7%; placebo: 13.6%) than those in CTR-1011 (tesamorelin: 3.3%; placebo: 6.3%), followed by those in LIPO-010 (tesamorelin: 4.0%; placebo: 2.2%) (Table 12).

3.7.3. Withdrawals Due to Adverse Events

In LIPO-010 and Stanley et al. 2014, a greater percentage of participants receiving tesamorelin (9.5% in LIPO-010; 10.7% in Stanley et al. 2014) experienced AEs that led to early study discontinuation than those in the placebo group (2.9% in LIPO-010, 3.8% in Stanley et al. 2014) (Table 12). In CTR-1011, the percentage of participants who experienced AEs that led to early study discontinuation was similar between the tesamorelin (9.5%) and placebo (9.3%) groups.

3.7.4. Mortality

There were no deaths in LIPO-010 and Stanley et al. 2014. Two participants died in CTR-1011, one in each treatment group (Table 12).

3.7.5. Notable Harms

In LIPO-010 and CTR-1011, a greater percentage of participants receiving tesamorelin reported any injection-site condition (LIPO-010: 30.0% versus 24.1%; CTR-1011: 50.7% versus 21.4%), myalgia (LIPO-010: 7.7% versus 2.2%; CTR-1011: 3.7% versus 1.6%) than those in the placebo group (Table 12), or any fluid retention or edema (LIPO-010: 9.9% versus 5.8%; CTR-1011: 5.2% versus 0.8%). In Stanley et al. 2014, no participants receiving placebo experienced myalgia, while 10.8% of individuals in the tesamorelin group were affected. Further, in LIPO-010, one participant (0.4%) receiving tesamorelin (versus none receiving placebo) developed diabetes mellitus (recorded as a treatment-emergent AE). Moreover, in the same trial, a greater percentage of participants receiving tesamorelin versus placebo (2.9% versus 1.5%) developed a malignancy, whereas fewer participants receiving tesamorelin in CTR-1011 versus those receiving placebo (0.4% versus 3.2%) developed a malignancy.