Once-weekly semaglutide in people with HIV-associated lipohypertrophy: a randomised, double-blind, placebo-controlled phase IIb single-centred clinical trial

Summary

Background

HIV-associated lipohypertrophy remains problematic in people with HIV (PWH), characterized by abnormal accumulation of abdominal visceral adipose tissue (AT). Effective interventions are lacking, although it carries significant cardiometabolic co-morbidity risk. The primary aim of this trial was to investigate effects of the glucagon-like peptide 1 receptor agonist, semaglutide, on AT in HIV-associated lipohypertrophy.

Methods

This randomised, double-blind, placebo-controlled phase IIb clinical trial was conducted at a single US site (clinicaltrials.gov: NCT04019197). Key inclusion criteria included PWH ≥18 years old with controlled HIV-1, body mass index ≥25 kg/m², and lipohypertrophy but without diabetes. Participants were randomised 1:1 to receive 32 weeks of once-weekly subcutaneous semaglutide (8-week dose titration+24 weeks at 1·0 mg) or placebo; all personnel/participants remained masked to treatment assignment. Primary outcomes included changes at 32 weeks in AT quantity by body compartment. Analyses were performed using intention-to-treat principles. This trial is completed.

Findings

Between June 10, 2019 to July 28, 2022, 108 participants were randomised to receive semaglutide (n=54) or placebo (n=54). Eight (14·8%) in each group withdrew prematurely. Significant effects of semaglutide were seen over the 32-week study period in sex-adjusted multiplicative regression analyses for the primary outcome of interest, abdominal visceral AT [β (95% CI); estimated % change: −30·82 cm² (−50·13, −11·51); −30·6%]. Decreases were also seen in other key measures, including abdominal subcutaneous AT [−42·01 cm² (−75·49, −8·52); −11·2%] and total body fat [ln(−0·21) kg (−0·34, −0·05); −18·9%]. There were no statistically significant differences in possibly-related/related adverse events [absolute risk difference (95% CI): 0·1111 (−0·0727, 0·2869); however, one semaglutide-related grade 4 elevated lipase and two possibly-related cases of cholelithiasis (grades 1 and 2) were observed.

Interpretation

Semaglutide holds promise as an effective treatment for HIV-associated lipohypertrophy. Potential risk of serious adverse events deserves further scrutiny in larger trials in PWH.

Funding

National Institutes of Health

Introduction

HIV-associated lipohypertrophy remains problematic in antiretroviral therapy (ART)-treated people with HIV (PWH), primarily characterized by abnormal accumulation of abdominal visceral adipose tissue (VAT) and ectopic fat depots, particularly liver, heart, and skeletal muscle. HIV-associated lipohypertrophy is distinct from obesity and weight gain common in PWH after ART initiation (both characterized by generalized adipose tissue (AT) deposition). HIV-associated lipoatrophy, recognised by subcutaneous adipose tissue (SAT) loss in limbs, face, abdomen, and/or buttocks, is also a discrete entity; however, both phenotypes can coexist and carry similar cardiometabolic risks.^1,2

Importantly, even in the absence of HIV-associated lipohypertrophy, metabolic, fat, and muscle alterations are widespread in HIV due to viral and treatment effects. For example, PWH suffer from adipocyte dysfunction and poorer AT quality, higher rates of dyslipidaemia and insulin resistance, alterations in gut hormone secretion and gut epithelial barrier dysfunction, and sarcopenia (decline in muscle mass and function) compared to people without HIV.^3–5 These abnormalities lead to increased risk of co-morbidities like diabetes, cardiovascular disease (CVD), and frailty.^6,7 HIV-associated lipohypertrophy exacerbates these problems^2,8 and contributes to poorer body image, feelings of shame, and impaired psychosocial functioning.^9,10 Increasing obesity among PWH complicates the situation, not only by compounding co-morbidity risk but by masking the diagnosis of lipohypertrophy and need for HIV-specific interventions.

Currently, few interventions exist for HIV-associated lipohypertrophy. Lifestyle modifications with diet and structured exercise are the mainstay, but data are mixed on overall efficacy and effect on abdominal VAT. Antiretroviral switches have not been shown effective. Tesamorelin, a synthetic growth hormone-releasing hormone analogue, is the only FDA-approved medication for VAT reduction in PWH; however, its use is limited by need for reconstitution, daily subcutaneous (SC) injections, modest response rates, reversal of response with drug discontinuation, and possible increased risk of malignancy and glucose intolerance/diabetes.¹¹ Thus, there is an urgent need to identify novel therapeutics.

Glucagon-like peptide 1 receptor agonists (GLP1RAs) possess many properties suggestive of potential efficacy in HIV-associated lipohypertrophy; however, no data exist. GLP1RAs mimic effects of the physiological incretin hormone by stimulating glucose-dependent insulin release from pancreatic islets, inhibiting glucagon release, and delaying gastric emptying, among other important properties. They have proved effective for improving glycaemic control in diabetes, reducing visceral and liver fat, improving CVD outcomes in non-HIV populations, and weight loss in people considered overweight or obese.^12–21 Prospective studies are warranted to assess potential benefits and harms that may differentially affect this complex and unique HIV patient population. Therefore, the primary aim of this trial was to investigate effects of the GLP1RA, semaglutide, on quantity and distribution of AT in PWH with HIV-associated lipohypertrophy.

Methods

Study design

This is a randomised, double-blind, placebo-controlled phase IIb clinical trial (RCT). Participants with HIV-associated lipohypertrophy were enrolled at a single site (University Hospitals Cleveland Medical Center, Cleveland, Ohio, USA) and randomized to receive 32 weeks of once-weekly SC semaglutide or matching placebo. Findings presented here include the pre-determined 32-week main outcome measures. Institutional Review Board IORG000040 approved the study. Participants received study drug free of charge.

Participants

Eligible participants were ≥18 years old with documented HIV-1 infection, cumulative ART duration ≥1 year, receiving a stable ART regimen for ≥12 weeks, and HIV-1 RNA <400 copies/mL for ≥6 months prior to entry. In addition, participants required subjective observation of increased abdominal girth occurring after ART initiation, waist circumference (WC) and waist-to-hip ratio (WHR) of >95 cm and >0·94 cm, respectively, for men, and >94 cm and >0·88 cm, respectively, for women, and body mass index (BMI) of ≥25 kg/m². Known history of diabetes was exclusionary. A complete list of inclusion and exclusion criteria and details regarding trial design, methods, and procedures can be found in the protocol (appendix 2).

Potential participants were recruited from the institution’s HIV clinic or upon referral from other local HIV clinics. Collectively, recruitment efforts focused on creating a diverse, generalizable study population, including optimizing enrolment of under-represented racial, ethnic, and sex/gender groups. All participants signed a written informed consent.

Randomisation and masking

Study statistician (AS) provided randomisation allocation sequences generated from an online software program (sealedenvelope.com) with a 1:1 ratio and block size of 6 without stratifying factors directly to the investigational pharmacy that was maintained in a locked, secured location. Key personnel, study staff, participants, and participants’ medical providers were masked to treatment assignment. To preserve masking, investigational pharmacy assigned participants their treatment allocation at entry and dispensed all doses of semaglutide or matching placebo (saline) in identical-appearing, pre-filled syringes without any recognizable markings. All participants received identical information regarding potential side effects and underwent the same procedures. Statistical analyses and adverse event (AE) reviews were performed in masked fashion. Unmasking of treatment assignment was only allowed if warranted by a potentially study-related serious/unanticipated AE or related safety concerns; if indicated, investigational pharmacy was responsible for unmasking to the principal investigator only.

Procedures

Qualifying participants entered an initial 4-week lead-in screening phase consisting of weekly study visits (weeks −4, −3, −2, −1). At each visit, participants underwent targeted physical examinations, completed detailed dietary and physical activity questionnaires, and received standardized dietary and physical activity advice from a registered clinical dietician. This phase was designed to assess stability of lifestyle habits, anthropometric measurements, and ability to complete study requirements and adhere to weekly study visits.

Participants who successfully completed screening were enrolled into the interventional phase and randomised to receive semaglutide or matching placebo (week 0/entry). The interventional phase was comprised of two parts: titration phase (weeks 1–8) and full-dose phase (weeks 9–32). The titration phase, designed to minimize initial drug side effects for those randomised to receive semaglutide, consisted of weekly 0·25 mg SC semaglutide injections for 4 weeks, followed by weekly 0·5 mg SC semaglutide injections for another 4 weeks (or matching placebo). After the titration phase, participants received 1·0 mg SC semaglutide weekly for 24 weeks or matching placebo. Participants returned weekly to receive study drug (semaglutide or placebo); trained study staff administered all SC injections as per manufacturer’s instructions to maximize safety and ensure adherence.

At weeks 0/entry, 9, 14, 20, 26 and 32, participants underwent comprehensive clinical and laboratory assessments, including complete metabolic panel, lipase, lipid profiles, haematology panel, and urine pregnancy test (for those with reproductive potential). 12-hour fasting period was required prior to blood draws at weeks 0, 9, 20, and 32. Glycated haemoglobin (HbA_1c%) and glucose and insulin (fasting and 2 hours after consumption of 75 mg oral glucose) were measured at weeks 0, 9, and 32. Blood glucose was measured in real-time using AimStrip® Plus Blood Glucose Testing System (23–111-275); other blood values were measured from serum or plasma, as appropriate, at a central clinical laboratory. Homeostatic model assessment of insulin resistance (HOMA-IR) was calculated from the following equation: [glucose (mg/dL)×(insulin (mU/L)/405)]. HIV-1 RNA and CD4+ T-cell counts were obtained as part of standard of care practice every 12 weeks.

Body composition was assessed using whole-body dual-energy x-ray absorptiometry (DXA) and non-contrast helical computed tomography (CT) of chest and abdomen at weeks 0 and 32, as previously described.²² Measurements were read by a single radiologist (KAG) masked to treatment assignment and clinical information. DXA was performed with a standardized protocol in anteroposterior view using the same scanner (Lunar Prodigy Advance, GE Healthcare, Chicago, Illinois) for fat mass (total body, limb, trunk) and lean body mass (LBM) measurements. Single-slice at the L4-L5 level was used to estimate abdominal AT quantity (area in cm²) and quality (density in Hounsfield Units (HU)) delineated by compartment (VAT, SAT) and total AT (TAT). AT was identified by a mean attenuation of −190 to −30 HU, in which a more negative value represents a lower density and poorer AT quality.

Single-slice CT was also used to obtain a cross-sectional measure of total right psoas muscle area (evaluation of skeletal muscle quantity) and density (estimate of skeletal muscle ectopic fat; lower attenuation=greater fatty infiltration), as previously described and validated.^3,23 Liver fat content was estimated by measuring density, as previously validated.^24,25 Measurements were obtained in three circular regions of interest with an area of ≥2 cm² within the parenchyma, taking care to avoid vessels and bile ducts, in different axial slices within three areas (right anterior lobe, right posterior lobe, and left lobe); measurements from each area were then averaged and recorded as a single value to estimate liver fat content (lower attenuation=greater fatty infiltration). Electrocardiogram-gated cardiac CT scan was performed on a 64-slice multidetector scanner (Somatom Sensation 64; Siemens Medical Solutions, Malvern, Pennsylvania) to measure ectopic pericardial fat volume and density, as previously described.⁸ All fat within the pericardial sac (epicardial+non-epicardial fat) within a window of −195 to −45 HU was included. Pericardial fat density was measured as the mean attenuation on 3 consecutive axial slices through the center of the heart.

Additional clinical evaluations included standardized anthropometric measurements (weight/BMI, WC at the umbilicus, hip circumference at widest part of buttocks), physical activity status [International Physical Activity Questionnaire (IPAQ) short version (physical activity over previous 7 days)], and dietary assessments based on 24-hour food/supplement intake (obtained by an experienced nutrition core overseen by a registered dietician (AF); data analysed using Nutrition Data System for Research (NDS-R) software version 2018). One 24-hour food/supplement recall was obtained at the participants’ study visits; phone calls were placed on subsequent days to obtain a total of three 24-hour food/supplement recalls per time point. 10-year atherosclerotic cardiovascular disease risk (ASCVD) was estimated using American College of Cardiology’s ASCVD risk estimator plus.

Outcomes

Primary outcomes were changes in AT quantity by body compartment: 1) total body fat (TBF), 2) limb fat, 3) trunk fat, 4) abdominal TAT, 5) abdominal SAT, 6) abdominal VAT, and 7) pericardial fat. Secondary AT and skeletal muscle outcomes included changes in fat density in abdominal AT compartments (i.e., fat quality), total right psoas muscle area (i.e., muscle mass), total right psoas density (i.e., ectopic intermuscular fat), liver fat, and LBM. Additional secondary outcomes included changes in anthropometric measurements (weight, BMI, WC, WHR), glucose metabolism (HbA_1c%, fasting glucose, 2-hour oral glucose tolerance test (OGTT)), insulin resistance (fasting and 2-hour OGTT insulin and HOMA-IR), vital signs (heart rate, blood pressure), caloric intake, physical activity, and ASCVD risk estimate.

During the interventional phase, participants were evaluated weekly for possible side effects of study drug and AEs. At weeks 0/entry, 9, 14, 20, 26 and 32, participants underwent comprehensive laboratory safety monitoring. All collected laboratory values listed in the National Institute of Allergy and Infectious Diseases, Division of AIDS, corrected version 2.1 (2017) adverse events table were recorded and graded based on severity (1=mild; 2=moderate; 3=severe; 4=potentially life-threatening; 5=death). All listed clinical conditions in the table plus additional symptoms with known associations to semaglutide (e.g., cholelithiasis) or those that may indicate a possible drug side effect (e.g., “shakiness” from hypoglycaemia) were also recorded and graded similarly. Each AE’s relationship to study drug was designated as definitely, possibly, or not related by the study’s principal investigators. All safety data and AEs were reviewed by an independent data and safety monitoring board that met every 6 months and ad hoc as needed.

Statistical analyses

Sample size estimation and power analysis were based on significance level 0·05, 90% power, 20% dropout adjustment, and assumptions of expected % change in abdominal VAT (key outcome of interest) as described in the protocol. Conservatively assuming a mean difference of −34·33 cm² in abdominal VAT change between groups at 32 weeks, 43 participants per arm in a repeated measures design with two longitudinal measurements achieves >90% power. Sample size was inflated to 52 per arm to account for a conservative 20% loss-to-follow-up/missing data. Enrolment was further increased to 54 per arm during the COVID-19 pandemic to address a possible higher-than-expected loss-to-follow-up.

Analyses were performed using intention-to-treat principles based on randomised treatment assignment including all available data. Between-group comparisons were conducted using either two-sample t-tests or Wilcoxon rank-sum tests, as appropriate. Generalized estimating equations (GEE) were used to model effects of semaglutide on outcome variables, examining the interaction between time and treatment and controlling for covariates. We specified a Gaussian family, an identity link, and unstructured correlation for within-group repeated measurements in modelling an outcome using GEE. For models with significant interaction terms, appropriate linear combination of regression coefficients were assembled in an equation to extract the true effect size of semaglutide while considering the multiplicative effects of treatment and time.

Appropriate transformations were applied to outcome variables with skewed distributions prior to forming the GEE models. For highly-skewed variables where transformations could not create a symmetric distribution, simultaneous quantile (median) regressions were applied. Clustering methodology was used to capture correlations of repeated measurements from the same participant to build a median regression model, and bootstrapping (1000x) was applied to correct standard errors of the estimates. The β, linear combination of regression coefficients, represents estimated effect size. Model predictions of % change in outcome variables from semaglutide treatment were calculated from the final fitted models using the formulas 100(e^β−1) and 100(β/regression intercept) for natural logarithmic (ln)- and non-transformed variables, respectively.

Potential predictors of >5% loss in TBF and abdominal VAT were assessed using stepwise logistic regression; variables for inclusion in the models were chosen based on clinical significance. To estimate proportion and odds ratio of participants achieving >5% reduction in body weight compared to placebo, contingency table analysis and logistic regression modelling were used. Influence of sex within the semaglutide group was assessed using appropriate hypothesis tests.

AEs were tallied for each participant by type and severity and then further delineated by relationship to study drug. For participants with recurrence of the same AE and/or with varying degrees of severity, the one with the highest grade or longest duration was counted. Absolute risk differences of AE proportions between groups and 95% confidence intervals (CI) were calculated using point estimates (p̂_semaglutide-p̂_placebo) and Agresti-Coull method, respectively. Benjamini-Hochberg procedure was used to adjust for multiple testing in all analyses; given the phase II nature of our trial, we selected a false discovery rate of 0·20 for correction of p values. Missing data were random; thus, no adjustments were made for missingness in analyses. All analyses were performed using Stata 18·0 and R 4·3·0 statistical software. The study was registered on clinicaltrials.gov (NCT04019197).

Role of the funding source

The funding source had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

From June 10, 2019 to July 28, 2022, 154 potential participants were screened for eligibility, of whom 108 were enrolled and randomised to receive semaglutide or placebo (figure 1). Eight participants in each group prematurely withdrew from the study prior to the 32-week primary endpoint.

Figure 1: Trial profile.

Screening phase included the initial eligibility assessment plus a 4-week lead-in phase that consisted of weekly study visits designed to assess stability of each participant’s lifestyle habits, anthropometric measurements, and ability to complete study requirements and adhere to weekly study visits. Any participant who was randomised and received at least one study drug injection was included in the intention-to-treat analyses (n=108). 16 (14·81%) of randomised participants withdrew prematurely before the 32-week study endpoint. 1 (1·85%) participant of the total enrolled in the semaglutide group withdrew prematurely due to a study-related SAE (grade 4 elevated lipase value). CT=computed tomography. eGFR=estimated glomerular filtration rate. ULN=upper limit of normal. SAE=serious adverse event.

Baseline characteristics were similar between groups except for slightly more males in the semaglutide group (table 1). All participants had an HIV-1 RNA <400 copies/mL at entry except one in the semaglutide group (426 copies/mL). ART regimens included an integrase strand transfer inhibitor (INSTI) and protease inhibitor (PI) for 82% and 17% of participants, respectively. 74% of participants were on tenofovir alafenamide (TAF); 61% of participants were on INSTI+TAF. No participant was on metformin.

Table 1:

Baseline characteristics of the intention-to-treat population

	Semaglutide (n=54)	Placebo (n=54)
Demographics
Age, years	53 (40, 57)	53 (41, 57)
Sex
Male	38 (70·0%)	27 (50·0%)
Female	16 (30·0%)	27 (50·0%)
Race
Black	33 (61·1%)	34 (60·0%)
White	20 (37·0%)	18 (33·3%)
Bi-racial	0 (0·0%)	2 (3·7%)
Native American	1 (1·9%)	0 (0·0%)
Hispanic ethnicity	4 (7·0%)	5 (9·0%)
HIV variables
CD4+ T-cell count, cells/μL	826 (407, 1058)	793 (579, 994)
HIV duration, months	203 (118, 305)	226 (155, 282)
Antiretroviral therapy duration, months	168 (98, 230)	148 (98, 198)
Current INSTI use	45 (83·0%)	43 (80·0%)
Current protease inhibitor use	10 (19·0%)	8 (15·0%)
Current tenofovir alafenamide use	43 (80·0%)	36 (67·0%)
Anthropometric measurements
Weight, kg	97·5 (87·3, 110·6)	97·9 (83·6, 115·5)
Body mass index, kg/m2	32·9 (28·4, 36·0)	33·8 (29·9, 39·7)
Waist circumference, cm	107·0 (99·5, 113·0)	106·4 (101·8, 124·8)
Waist-to-hip ratio	0·96 (0·93, 1·02)	0·97 (0·94, 1·03)
Whole-body dual-energy absorptiometry measurements
Total body fat, kg	34·4 (27·4, 40·2)	35·5 (28·3, 49·7)
Total trunk fat, kg	18·6 (14·6, 23·0)	20·0 (15·3, 25·9)
Total limb fat, kg	15·6 (12·6, 19·2)	17·9 (11·6, 22·3)
Total lean body mass, kg	60·1 (50·7, 66·8)	56·9 (47·6, 65·6)
Computed tomography area and volume measurements
Abdominal TAT, cm2	416·1 (350·3, 475·7)	445·7 (384·8, 502·9)
Abdominal SAT, cm2	301·6 (225·8, 362·0)	332·6 (230·5, 389·4)
Abdominal VAT, cm2	101·2 (78·0, 143·2)	118·0 (83·5, 151·5)
VAT:TAT ratio	0·3 (0·2, 0·4)	0·2 (0·2, 0·4)
Pericardial fat, mL	78·6 (55·1, 115·6)	72·35 (47·5, 102·9)
Total right psoas muscle, cm2	14·7 (12·0, 18·6)	13·78 (10·4, 17·8)
Computed tomography density measurements
Liver, HU	58·7 (53·3, 63·6)	56·0 (51·6, 60·8)
Abdominal TAT, HU	−92·8 (−95·0, −90·1)	−91·1 (−94·1, −87·6)
Abdominal SAT, HU	−94·7 (−97·1, −91·5)	−92·7 (−96·8, −89·2)
Abdominal VAT, HU	−86·8 (−90·1, −81·8)	−87·5 (−90·6, −83·1)
Pericardial fat, HU	−109·5 (−110·9, −107·9)	−109·7 (−110·6, −108·6)
Total right psoas muscle, HU	47·1 (43·4, 51·0)	47·3 (43·2, 51·9)
Glucose metabolism and insulin resistance
HbA1c, %	5·5 (5·1, 5·8)	5·6 (5·3, 5·8)
Fasting glucose, mg/dL	101·0 (93·0, 105·0)	108·0 (98·0, 117·0)
Fasting insulin, uIU/ml	11·0 (6·0, 18·0)	15·00 (9·0–23·0)
Fasting HOMA-IR	2·5 (1·5, 4·7)	3·8 (2·5, 6·8)
2-hour OGTT glucose, mg/dL	105·0 (93·0, 129·0)	115·5 (92·0, 138·0)
2-hour OGTT insulin, uIU/ml	35·0 (21·0, 62·0)	38·0 (23·0, 90·0)
2-hour OGTT HOMA-IR	10·7 (4·9, 19·0)	11·1 (5·1, 36·3)
Lipoprotein profiles
Total cholesterol, mg/dL	181·50 (154·0, 203·0)	189·5 (154·0, 211·0)
High-density lipoprotein, mg/dL	46·6 (38·5, 56·3)	43·3 (36·7, 52·1)
Low-density lipoprotein, mg/dL	103·5 (82·0, 129·0)	108·5 (80·0, 133·0)
Very low-density lipoprotein, mg/dL	21·0 (16·0, 32·0)	26·0 (20·0, 36·0)
Triglycerides, mg/dL	103·5 (79·0, 159·0)	133·0 (104·0, 188·0)
Vital signs
Heart rate, beats/minute	76 (64, 82)	76 (71, 84)
Systolic blood pressure, mmHg	126 (120, 136)	126 (115, 137)
Diastolic blood pressure, mmHg	80 (76, 86)	80 (75, 84)
Lifestyle variables
Smoking status
Current	15 (27·8%)	23 (42·6%)
Past/never	39 (72·2%)	31 (57·4%)
Estimated daily dietary intake
Total calories, kcal	1604 (1170, 2083)	1905 (1192, 2113)
Estimated weekly physical activity
Low intensity, minutes	2520·0 (1260·0, 4200·0)	2130·0 (1260·0, 3360·0)
Moderate intensity, minutes	2100·0 (1260·0, 3360·0)	2730·0 (1680·0, 3480)
High intensity, minutes	195·0 (0·0, 720·0)	0·0 (0·0, 420·0)
Cardiovascular disease risk
10-year ASCVD risk estimate, %*	5·0 (2·6, 8·0)	5·4 (2·5, 8·8)

Data are presented as n (%) or median (interquartile range).

^*Participants outside the risk estimator’s parameters were assigned the minimum or maximum allowed values in the calculator [<40 years old (n=51); total cholesterol >300 mg/dL (n=1); total cholesterol <130 mg/dL (n=19); LDL <30 mg/dL (n=1). ASCVD=atherosclerotic cardiovascular disease. HbA_1c=glycated haemoglobin. HIV=human immunodeficiency virus. HOMA-IR=homeostatic model assessment of insulin resistance. HU=Hounsfield unit. INSTI=integrase strand transfer inhibitor. OGTT=oral glucose tolerance test. SAT=subcutaneous adipose tissue. TAT=total adipose tissue. VAT=visceral adipose tissue.

All participants maintained an HIV-1 RNA <400 copies/mL over the study period with four participants changing ART regimens. Of the three participants in the semaglutide group, one discontinued TAF, one changed INSTI regimens (elvitegravir/cobicistat/emtricitabine/TAF to cabotegravir/rilpivirine), and one changed from a PI-based regimen (darunavir/cobicistat/emtricitabine/TAF) to a nucleotide reverse transcriptase inhibitor (NRTI)-sparing, INSTI-based regimen (cabotegravir/rilpivirine). One participant in the placebo group changed from an INSTI-based regimen (bictegravir/emtricitabine/TAF) to a PI-based regimen (darunavir/cobicistat+emtricitabine/TAF).

Absolute and % changes by study group are shown in appendix p2. In the final GEE multiplicative regression models, semaglutide showed statistically significant effects at 32 weeks for all primary outcomes and anthropometric measurements except pericardial fat (table 2). Pericardial fat had an estimated 16·5% reduction but did not reach statistical significance. Abdominal compartments showed varying levels of decreases in area (VAT>TAT>SAT). Decreases in weight/BMI and WC were relatively proportional with smaller reductions in WHR. Regional fat masses showed the greatest loss within the trunk followed by comparable reductions in total body and limb fat.

Table 2:

Estimated effect sizes of semaglutide on outcome measures in people with HIV-associated lipohypertrophy treated with 32 weeks of once-weekly subcutaneous semaglutide or matching placebo

Outcome variable	βa	SE	95% CI of β	p-value	% changeb
Anthropometric measurements
ln(Weight, kg)	−0·11	0·04	−0·19, −0·03	0·0057*	−10·4%
Body mass index, kg/m2	−4·15	1·33	−6·77, −1·54	0·0018*	−11·2%
Waist circumference, cm2	−9·32	2·67	−14·55, −4·10	<0·0001*	−8·3%
ln(Waist-to-hip ratio)	−0·02	0·01	−0·05, 0·00	0·0854*	−2·3%
Whole-body dual-energy x-ray absorptiometry measurements
ln(Total body fat, kg)	−0·21	0·06	−0·33, −0·08	0·0009*	−18·9%
ln(Total limb fat, kg)	−0·19	0·07	−0·32, −0·05	0·0065*	−17·3%
ln(Trunk fat, kg)	−0·24	0·07	−0·37, −0·11	<0·0001*	−21·6%
Total lean body mass, kg	−2·98	1·96	−6·83, 0·86	0·13	−5·7%
Computed tomography area and volume measurements
Abdominal TAT, cm2	−72·07	19·29	−109·88, −34·26	<0·0001*	−15·1%
Abdominal SAT, cm2	−42·01	17·08	−75·49, −8·52	0·0138*	−11·2%
Abdominal VAT, cm2	−30·82	9·85	−50·13, −11·51	0·0017*	−30·6%
ln(VAT:TAT ratio)	−0·12	0·08	−0·27, 0·03	0·11	−11·5%
ln(Pericardial fat, mL)	−0·18	0·12	−0·41, 0·05	0·13	−16·5%
ln(Total right psoas muscle, cm2)	−0·05	0·05	−0·15, 0·04	0·28	−5·3%
Computed tomography density measurements
Liver, HU	3·55	1·97	−0·30, 7·41	0·0702*	6·2%
Abdominal TAT, HU	0·52	0·93	−1·30, 2·33	0·58	0·6%
Abdominal SAT, HU	−0·12	0.98	−2.05, 1.82	0·90	0·1%
Abdominal VAT, HU	3·70	1·25	1·25, 6·15	0·0031*	−4·4%
Pericardial fat, HU	0·80	0.57	−0·33, 1·93	0·17	−0·7%
Total right psoas muscle, HU	−0·51	1·28	−3·02, 1·99	0·69	−1·1%
Glucose metabolism and insulin resistance
HbA1C, %	−0·47	0·10	−0·67, −0·28	<0·0001*	−8·5%
ln(Fasting glucose, mg/dL)	−0·04	0·02	−0·09, 0·01	0·14	−3·6%
Fasting insulin, uIU/ml	0·02	0·14	−0·26, 0·29	0·91	1·6%
Fasting HOMA-IR	−0·02	0·15	−0·31, 0·27	0·89	−2·1%
2-hour OGTT glucose, mg/mL	−18·00	8·42	−34·50, −1·50	0·0324*	−15·5%
ln(2-hour OGTT insulin (uIU/ml))	−0·36	0·21	−0·76, 0·05	0·0819*	−30·1%
ln(2-hour OGTT HOMA-IR)	−0·56	0·25	−1·05, −0·07	0·0238*	−43·0%
Lipoprotein profiles
Total cholesterol, mg/dL	−7·17	7·67	−22·20, 7·85	0·35	−3·7%
ln(High-density lipoprotein, mg/dL)	0·08	0·05	−0·01, 0·17	0·0989*	8·1%
Low-density lipoprotein, mg/dL	−4·23	6·88	−17·72, 9·25	0·54	−3·7%
ln(Very low-density lipoprotein, mg/dL)	−0·20	0·10	−0·39, −0·01	0·0375*	−18·0%
ln(Triglycerides, mg/dL)	−0·23	0·10	−0·43, −0·03	0·0220*	−20·7%
Vital signs
Heart rate, beats/minute	−0·33	2·58	−5·39, 4·73	0·90	−0·4%
ln(Systolic blood pressure, mmHg)	−0·05	0·02	−0·10, −0·01	0·0203*	−5·2%
Diastolic blood pressure, mmHg	−1·20	1·77	−4·67, 2·26	0·50	−1·5%
Lifestyle variables
Estimated daily dietary intake
Total calories, kcal	−145·81	192·08	−522·28, 230·66	0·45	−10·8%
Estimated weekly physical activity
Low intensity, minutes	420·00	730·75	−1012·24, 1852·24	0·57	20.0%
Moderate intensity, minutes	−420·00	423·84	−1250·71, 410·71	0·32	−16·7%
High intensity, minutes	0·00	83·19	−163·04, 163·04	1·00	0.0%
Cardiovascular disease risk
ln(ASCVD risk estimate)	−0·04	0·19	−0·41, 0·32	0·81	−4·2%

^aβ coefficient from the final fitted linear combination of regression coefficients model estimating effect size of semaglutide at the 32-week time point on the outcome measure, adjusted for sex and multiplicative effects of treatment and time;

^bEstimated % change in outcome variables from semaglutide treatment calculated from the final fitted models using the formulas 100(e^β−1) and 100(β/regression intercept) for ln- and non-transformed variables, respectively.

^*Designates statistically significant effects of treatment on the respective outcome variable after Benjamini-Hochberg false discovery rate 0·20 correction. CI=confidence interval. HOMA-IR=homeostatic model assessment of insulin resistance. HU=Hounsfield unit. ln=natural logarithm. OGTT=oral glucose tolerance test. SAT=subcutaneous adipose tissue. SE=standard error. TAT=total adipose tissue. VAT=visceral adipose tissue.

Weight loss was similar between each measured time point over the study period (appendix p4). A greater number (%) of participants within the semaglutide group demonstrated a loss of >5% body weight [30 (65·2%) vs. 2 (4·4%)] with odds ratio of 41·3 (95% CI [8·8, 192·7]; p<0·0001). 39 (84·8%) and 30 (65·2%) participants within the semaglutide group lost >5% TBF and abdominal VAT area, respectively; however, no variable was predictive of these >5% losses (all p>0·10: baseline BMI, HbA_1c%, and CD4+ T-cell count; current TAF or INSTI use; age, sex, ART duration). There were no differences in main outcome measures within the semaglutide group by sex (appendix p5). In post hoc analyses, there were no differences at baseline or in unadjusted changes for weight/BMI over 32 weeks between participants within the semaglutide group stratified by current use of INSTI+TAF or by history of prior thymidine analogue NRTI use (data not shown).

Statistically significant treatment effects were seen for liver density (i.e., increased, suggestive of decreased liver fat). Psoas muscle density (i.e., ectopic intermuscular fat) did not change. Among fat quality measures, only abdominal VAT density showed a statistically significant increase (less negative=improved fat quality). Effects of semaglutide were not statistically significant for LBM or psoas muscle volume (i.e., muscle mass); however, both showed estimated decreases of 5·7% and 5·3%, respectively.

Statistically significant reductions were also seen in some measures of glucose metabolism and insulin resistance (HbA_1c%, 2-hour OGTT glucose, insulin, HOMA-IR), as well as decreases in systolic blood pressure, very low-density lipoprotein (VLDL) cholesterol, and triglycerides, and increases in high-density lipoprotein cholesterol. Diastolic blood pressure, heart rate, total cholesterol, low-density lipoprotein cholesterol, caloric intake, physical activity, and ASCVD risk estimates did not show any statistically significant changes.

AEs with known associations to semaglutide are depicted in table 3; remainder of reported AEs are shown in appendix p6. Majority of possibly- and study-related AEs were mild with no statistically significant differences between groups and/or were similar to entry values. Of note, one serious study-related AE occurred in a participant from the semaglutide group who was noted during routine safety monitoring to have grade 1 serum lipase (1·3x ULN) at week 9 that rose to grade 4 at week 14 (5·6x ULN). Repeat serum lipase and amylase were 6·0x and 2·4x ULN, respectively. Participant remained asymptomatic with a normal pancreatic ultrasound but was withdrawn from study as a cautionary measure; amylase and lipase returned to normal ranges within 2 weeks. Two possibly-related cases of cholelithiasis were also identified within the semaglutide group, one resulting in a prophylactic cholecystectomy; however, no cholecystitis cases were observed (including in these two participants and among the five participants with cholelithiasis detected on imaging at entry). 3 (5·56%) participants in the semaglutide group met criteria for diabetes at entry but did not experience any possibly- or study-related AEs.

Table 3:

Adverse events in people with HIV-associated lipohypertrophy treated with 32 weeks of once-weekly subcutaneous semaglutide or matching placebo

	Semaglutide (n=54)	Placebo (n=54)	Absolute risk difference (95% CI)
General adverse event data
≥1 adverse event	54 (100·00%)	53 (98·14%)	0·0185 (−0·0419, 0·0776)
Serious study-related adverse eventsa	1 (1·85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Possibly-related adverse events	21 (38·89%)	17 (31·48%)	0·0740 (−0·1056, 0·2484)
≥1 possibly- or study-related adverse event	24 (44·44%)	18 (33·33%)	0·1111 (−0·0727, 0·2869)
Adverse events or side effects leading to premature trial discontinuation (week of study withdrawal)	4 (7·41%)	1 (1·85%)	0·0556 (−0·0355, 0·1427)
Grade 4 elevated lipase (week 14)a	1 (1·85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Grade 1 elevated lipase, gastrointestinal symptoms (week 10)	1 (1·85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Grade 1 gastrointestinal and systemic symptoms (week 12)b	1 (1·85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Grade 1 memory impairment (week 17)	0 (0·00%)	1 (1·85%)	−0·0185 (−0·0776, 0·0419)
Weight loss (week 17)	1 (1·85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Safety areas of interest c
Study-related adverse events
Laboratory values
Elevated lipasea,d	1 (1·85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Clinical conditions
Injection site reactions (all grade 1)	4 (7·41%)	2 (3·70%)	0·0370 (−0·0594, 0·1309)
Possibly-related adverse events
Laboratory values
Elevated lipased	4 (7·41%)	2 (3·70%)	0·0370 (−0·0594, 0·1309)
Grade 1	2 (3·70%)	2 (3·70%)	−0·0185 (−0·1075, 0·0717)
Grade 2	2 (3·70%)	0 (0·00%)	0·0370 (−0·0327, 0·1041)
Elevated creatininee	1 (1·85%)	1 (1·85%)	0·0000 (−0·0687, 0·0687)
Clinical conditionsf
Cholelithiasis	2 (3·70%)	0 (0·00%)	0·0370 (−0·0327, 0·1041)
Any gastrointestinal disorder	18 (33·33%)	15 (27·78%)	0·0556 (−0·1178, 0·2250)
Nausea	6 (11·11%)	4 (7·41%)	0·0370 (−0·0787, 0·1501)
Vomiting	1 (1·85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Diarrhoea or loose stools	0 (0·00%)	7 (12·96%)	−0·1296 (−0·2230, −0·0270)
Abdominal pain	1 (1·85%)	2 (3·70%)	−0·0185 (−0·0943, 0·0586)
Constipation or irregular bowel movements	2 (3·70%)	2 (3·70%)	0·0000 (−0·0834, 0·0834)
Eructation or flatulence	5 (9·26%)	4 (7·41%)	0·0185 (−0·0923, 0·1280)
Gastroesophageal reflux disease	2 (3·70%)	1 (1·85%)	0·0185 (−0·0586, 0·0943)
Bloating or fullness	1 (1·85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Dyspepsia	1 (1·85%)	2 (3·70%)	−0·0185 (−0·0943, 0·0586)
Gastritis	0 (0·00%)	1 (1·85%)	−0·0185 (−0·0776, 0·0419)
Decreased appetite or changes in food cravings/tolerance	7 (12·96%)	1 (1·85%)	0·1111 (0·0034, 0·2109)
Fatigue	2 (3·70%)	2 (3·70%)	0·0000 (−0·0834, 0·0834)
Dysgeusia	3 (5·56%)	2 (3·70%)	0·0185 (−0·0717, 0·1075)
Dizziness	0 (0·00%)	3 (5·56%)	−0·0556 (−0·1294, 0·0223)
Headache	3 (5·56%)	0 (0·00%)	0·0556 (−0·0223, 0·1294)
Anxiety or nervousness	0 (0·00%)	1 (1·85%)	−0·0185 (−0·0776, 0·0419)
Irritability or moodiness	1 (1·85%)	2 (3·70%)	−0·0185 (−0·0943, 0·0586)
Sweating	2 (3·70%)	3 (5·56%)	−0·0185 (−0·1075, 0·0717)
Hunger	0 (0·00%)	2 (3·70%)	−0·0370 (−0·1041, 0·0327)
Generalized myalgias	0 (0·00%)	1 (1·85%)	−0·0185 (−0·0776, 0·0419)
Scalp alopecia	1 (1·85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Not study-related adverse events
Laboratory values
Elevated lipased	8 (14·81%)	3 (5·56%)	0·0926 (−0·0282, 0·2068)
Grade 1	7 (12·96%)	2 (3·70%)	0·0926 (−0·0197, 0·1983)
Grade 2	1 (1·85%)	1 (1·85%)	0·0000 (−0·0687, 0·0687)
Elevated creatininee	24 (44·44%)	13 (24·07%)	0·2037 (0·0238, 0·3691)
Grade 1	18 (33·33%)	9 (16·67%)	0·1666 (0·0012, 0·3202)
Grade 2	6 (11·11%)	3 (5·56%)	0·0555 (−0·0562, 0·1634)
Grade 3	0 (0·00%)	1 (1·85%)	−0·0185 (−0·0776, 0·0419)
Elevated total bilirubin (all grade 1)g	6 (11·11%)	1 (1·85%)	0·0926 (−0·0100, 0·1886)
Elevated alanine transaminaseh	5 (9·26%)	6 (11·11%)	−0·0185 (−0·1364, 0·1007)
Grade 1	3 (5·56%)	4 (7·41%)	−0·0185 (−0·1185, 0·0828)
Grade 2	2 (3·70%)	2 (3·70%)	0·0000 (−0·0834, 0·0834)
Elevated aspartate aminotransferasei	3 (5·56%)	4 (7·41%)	−0·0185 (−0·1185, 0·0828)
Grade 1	3 (5·56%)	3 (5·56%)	0·0000 (−0·0954, 0·0954)
Grade 2	0 (0·00%)	1 (1·85%)	−0·0185 (−0·0776, 0·0419)
Elevated glucosek	25 (46·30%)	39 (72·22%)	−0·2592 (−0·4262, −0·0738)
Grade 1	24 (44·44%)	38 (70·37%)	−0·2593 (−0·4274, −0·0726)
Grade 2	0 (0·00%)	1 (1·85%)	−0·0185 (−0·0776, 0·0419)
Clinical conditionsl
Any gastrointestinal disorder	35 (64·81%)	27 (50·00%)	0·1481 (−0·0385, 0·3242)
Nausea	15 (27·78%)	7 (12·96%)	0·1481 (−0·0068, 0·2925)
Vomiting	8 (14·81%)	3 (5·56%)	0·0926 (−0·0282, 0·2068)
Diarrhoea or loose stools	7 (12·96%)	10 (18·52%)	−0·0556 (−0·1922, 0·0851)
Abdominal pain	4 (7·41%)	1 (1·85%)	0·0556 (−0·0355, 0·1427)
Constipation or irregular bowel movements	16 (29·63%)	5 (9·26%)	0·2037 (0·0513, 0·3416)
Bloating or fullness	2 (3·70%)	2 (3·70%)	0·0000 (−0·0834, 0·0834)
Dyspepsia	11 (20·37%)	0 (0·00%)	0·2037 (0·0835, 0·3094)
Eructation or flatulence	10 (18·52%)	6 (11·11%)	0·0741 (−0·0640, 0·2068)
Gastritis	3 (5·56%)	0 (0·00%)	0·0556 (−0·0223, 0·1294)
Gastroesophageal reflux disease	14 (25·93%)	6 (11·11%)	0·1481 (−0·0019, 0·2876)
Decreased appetite or changes in food cravings/tolerance	8 (14·81%)	4 (7·41%)	0·0741 (−0·0504, 0·1932)
Fatigue	14 (25·93%)	6 (11·11%)	0·1481 (−0·0019, 0·2876)
Dysgeusia	4 (7·41%)	1 (1·85%)	0·0556 (−0·0355, 0·1427)
Dizziness	6 (11·11%)	1 (1·85%)	0·0926 (−0·0100, 0·1886)
Headache	12 (22·22%)	10 (18·52%)	0·0370 (−0·1161, 0·1876)
Anxiety or nervousness	5 (9·26%)	2 (3·70%)	0·0556 (−0·0466, 0·1538)
Irritability or moodiness	5 (9·26%)	4 (7·41%)	0·0185 (−0·0923, 0·1280)
Sweating	2 (3·70%)	3 (5·56%)	−0·0185 (−0·1075, 0·0717)
Hunger	3 (5·56%)	8 (14·81%)	−0·0926 (−0·2068, 0·0282)
Generalized myalgias	1 (1·85%)	1 (1·85%)	0·0000 (−0·0687, 0·0687)
Shakiness	2 (3·70%)	3 (5·56%)	−0·0185 (−0·1075, 0·0717)
Weakness	1 (1·85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Lightheadedness	0 (0·00%)	1 (1·85%)	−0·0185 (−0·0776, 0·0419)
Temporary neurosensory alteration	1 (1.85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Hiccups	1 (1·85%)	0 (0·00%)	0·0185 (−0·0419, 0·0776)
Diabetes mellitus	1 (1·85%)	2 (3·70%)	−0·0185 (−0·0943, 0·0586)

Data are number (%) of participants. Grade 1=mild, 2=moderate, 3=severe, 4=potentially life-threatening, 5=death.

^aGrade 4 lipase value in asymptomatic participant with no evidence of pancreatitis on abdominal ultrasound.

^bSymptoms did not resolve after discontinuation of study drug.

^cAll adverse events with known associations to semaglutide are included, regardless of relationship to study drug.

^dAll participants had lipase values within normal range at entry; among participants within semaglutide group with elevated lipase AEs, 11 (84·6%) were on an INSTI-based regimen, similar to the total study group’s % INSTI use; participant with grade 4 lipase level was not on an INSTI-based regimen.

^e12 (22·22%) and 1 (1·85%) participants in semaglutide group had grade 1 and grade 2 creatinine values, respectively, at entry; 7 (12·96%) and 2 (3·70%) participants in placebo group had grade 1 and grade 2 creatinine values, respectively, at entry.

^fAll grade 1 except 1 (1·85%) grade 2 cholelithiasis in semaglutide group; 5 (9·26%) and 4 (7·41%) additional participants in the semaglutide and placebo groups, respectively, had evidence of cholelithiasis on imaging at entry.

^g2 (3·70%) and 1 (1·85%) participants in semaglutide group had grade 1 and grade 2 total bilirubin values, respectively, at entry.

^h1 (1·85%) participant in semaglutide group had grade 1 alanine transaminase value at entry.

ⁱ2 (3·70%) participants in placebo group had grade 1 aspartate aminotransferase values at entry;

^k8 (14·81%) and 1 (1·85%) participants in semaglutide group had grade 1 and grade 2 glucose values, respectively, at entry; 17 (31·48%) and 7 (12·96%) participants in placebo group had grade 1 and grade 2 glucose values, respectively, at entry.

^lAll grade 1 except grade 2 abdominal pain/diarrhea in 1 (1·85%) participant in semaglutide group with history of diverticulosis. No adverse events were recorded for the following laboratory values and clinical conditions: pancreatitis, cholecystitis, subjective or documented elevated heart rate, diabetic retinopathy, documented hypoglycaemia. CI=confidence interval. INSTI=integrase strand transfer inhibitor.

Discussion

To our knowledge, this is the first RCT to investigate semaglutide effects in PWH and the first study to evaluate its effects on HIV-associated lipohypertrophy. Notably, there were considerable reductions in all primary outcome measures, including TBF, trunk/abdominal fat (in both visceral and subcutaneous abdominal compartments), limb fat, and pericardial fat. There were also improvements in liver fat. Estimated total fat loss was ~19% with an ~10% weight reduction that declined steadily over 32 weeks. Direct comparison to RCT in non-HIV populations is challenging given differences in patient populations, treatment doses, and study duration. However, weight loss was generally similar or greater compared to trials using 1·0 mg for ~32 weeks or longer.^13,17,18 Trials using 2·4 mg led to greater weight loss at ~32 weeks in some studies;¹⁴ yet, weight loss was only slightly greater or comparable in others,^12,15 including total weight loss over longer treatment durations.^12–14

Perhaps the most dramatic effect of semaglutide was seen in abdominal fat, an estimated total loss of ~72 cm² (−15%), comprised of both VAT and SAT. Data were corroborated by a corresponding ~22% and ~8% reduction in trunk fat mass and WC, respectively. Central obesity is associated with all-cause mortality in a dose-response manner, even after adjusting for BMI,²⁶ and consensus exists for utilizing WC as a marker of cardiometabolic risk.²⁷ Thus, such a measurable decrease in central obesity speaks to the positive effects of semaglutide in reducing central obesity and associated co-morbidity risk in PWH.

The visceral fat component in central obesity confers much of the risk,²⁸ as it plays key roles in the initiation and perpetuation of liver steatosis, systemic insulin resistance, and chronic inflammation. A disproportionate increase in visceral adiposity is the hallmark of HIV-associated lipohypertrophy, illustrated in our study population where ~25% of all abdominal fat was visceral with a ratio of VAT:SAT >1:3. The estimated proportion of VAT loss was ~2·7x greater than SAT loss with an estimated reduction in VAT of ~30%. Thus, such VAT decreases should arguably translate into significant long-term protective effects on cardiometabolic co-morbidity risk.

Similarly, semaglutide showed notable effects on ectopic fat depots, similar to GLP1RAs studies in non-HIV populations,^16,20 albeit less dramatic than abdominal fat losses. While not statistically significant, pericardial fat changes (epicardial+non-epicardial fat) were arguably clinically significant give our previous study showing total pericardial fat is positively correlated with insulin resistance and markers of monocyte activation in HIV.⁸ Likewise, effects of semaglutide were estimated to increase liver density by 6%, indicating a reduction in liver fat. It is difficult to determine precisely how this % increase translates to total fat content. A previous study showed a highly correlated linear relationship (r=0·99, p<0·001) between density and liver fat content assessed by liver biopsy.²⁴ Similarly, Kodama, et al determined attenuation values of 64·4, 59·1, 41·9, and 25·0 HU correlated with fatty infiltration degrees of 0%, 1–25%, 26–50% and >50%, respectively.²⁵

Difficulty in interpreting the clinical meaning of our liver density changes may suggest magnetic resonance imaging should be used to quantify liver fat; however, this alternative methodology often generates considerable artifact and is difficult to execute large-scale due to its expense and required machine time. Liver biopsies present equal practical challenges given their invasive nature and hepatic heterogeneity. Thus, we contend our methodology choice was sound and arguably more feasible for the research setting. As such, despite the aforementioned limitations, our data are still meaningful and serve to inform future trials.

Total right psoas muscle area and LBM also decreased ~5–6%; however, neither reached statistical significance. LBM is comprised of more than muscle (e.g., bone, organs, water), making it difficult to discern how much associated muscle mass loss occurred. A lack of statistical significance in psoas muscle area changes suggest limited impact on muscle mass; however, small sample size may have contributed to our findings. We also did not measure muscle strength or physical function, which would have aided in this investigation. Although studies in non-HIV populations have observed LBM reductions,^15,29 negative implications may be greater or atypical in PWH. For example, in addition to sarcopenia, PWH have increased rates of frailty, osteoporosis, and neurological decline.⁶ Moreover, excessive LBM losses are associated with fatigue, declines in neuromuscular function, increased risk for injury, and decreases in metabolism which can cause a paradoxical regain of fat.¹⁹ Accordingly, effects of GLP1RAs on LBM including assessments of muscle strength/physical functioning in PWH should be further investigated in larger RCTs. Until then, clinicians should consider individual risks vs. benefits and counsel patients on ways to preserve LBM during weight loss therapy.

It is important to note that participants were treated with 1·0 mg semaglutide weekly, the FDA-approved dose at the time of study initiation. Since then, recommended doses have increased to 2·0 and 2·4 mg for diabetes and weight loss, respectively. Despite using a lower dose, we observed clinically meaningful changes in many key outcome measures. Additional notable findings included relatively dramatic improvements in glucose metabolism and insulin resistance, especially 2-hour OGTT HOMA-IR. Improvements in lipid profiles were also noteworthy, likely related to decreases in central adiposity rather than dietary intake (given no change in total calorie consumption). Despite improvements in several variables that comprise the 10-year ASCVD risk estimate, it did not decrease appreciably. This is likely due to a relatively young population with low-risk estimates at entry.

1·0 mg weekly semaglutide for 32 weeks was generally well-tolerated with largely mild AEs similar to placebo, including GI side effects. Importantly, we observed one serious elevated lipase level that arguably would have progressed to pancreatitis had it not been detected early. Similarly, two participants developed cholelithiasis possibly related to semaglutide. PWH, especially those with cardiometabolic co-morbidities and/or are overweight or obese are at increased risk for conditions like cholelithiasis and pancreatitis that are also associated with semaglutide. Our rates of GI side effects and serious AEs (e.g., pancreatitis) were generally lower than studies using higher doses.^12–15 Yet, some trials using 2·4 mg observed rates of cholelithiasis comparable to ours using 1·0 mg.^12–15 Thus, while higher semaglutide doses arguably would produce greater positive effects, potential benefits should be weighed against added risk of serious AEs in PWH and require investigation in future RCTs with longer treatment periods.

Roughly two-thirds of weight loss in our study was attributed to fat decreases. While fat loss carries beneficial effects for PWH considered overweight or obese, for some, there are potentially negative consequences to consider. Specifically, there are considerable numbers of PWH with lipoatrophy, most often apparent in their faces and limbs. Further SAT loss may not only exacerbate societal stigma and psychosocial impairment but cause additional negative health effects. We excluded people with severe lipoatrophy and observed limb fat losses nearly proportional to TBF. We also did not observe any notable differences between participants with and without prior thymidine analogue NRTI use. Collectively, these data suggest clinicians should weigh potential benefits of fat loss, particularly VAT, against exacerbation of pre-existing lipoatrophy or mixed lipodystrophy picture.

Additional study limitations are worth mentioning. For example, we utilized density to measure fat quality which may fail to assess more subtle effects on adipocytes such as changes in macrophage infiltration or browning, as seen in other studies.¹⁹ Nevertheless, fat density has correlated with histologic assessments of fat quality in HIV.³⁰ Our intentional selection of PWH without known diabetes may also have limited results; however, such a study design allows us to elicit differential effects of semaglutide in this specific patient population that suffer from unique metabolic and adipocyte pathologies. Similarly, a comprehensive understanding of HIV-associated lipohypertrophy remains elusive, making interpretation of some findings challenging, especially as they compare to generalized obesity. Lastly, our study was limited by a relatively small sample size, short duration, and single-site enrolment.

This is the first RTC investigating use of GLP1RAs in PWH and HIV-associated lipohypertrophy. Our study showed beneficial effects on fat, particularly abdominal and ectopic AT after 32 weeks of semaglutide treatment. Semaglutide also demonstrated a positive safety profile, although decreases in LBM and risk of less common yet serious side effects require further investigation, particularly in RTC with higher doses and longer durations. In summary, semaglutide shows promise as an effective treatment for HIV-associated lipohypertrophy and co-morbidity risk in PWH.

Research in Context.

Evidence before this study

We searched PubMed for studies published from database inception until February 2024, using the terms [semaglutide] AND [hiv lipohypertrophy], and found 0 results. The search was repeated using the terms [semaglutide] AND [hiv], and there were 3 results, only one of which was an original article addressing weight loss in people with HIV (PWH) who were treated with semaglutide. This retrospective study reported that semaglutide was associated with an average weight loss [95% confidence interval (CI)] of 6·47 kg [−7·67, −5·18] and a reduction in HbA_1c of 1·07% [−1·64, −0·50] at 1 year. To understand the impact of treatment with glucagon-like-peptide 1 receptor agonists (GLP1RAs) on visceral adiposity, we searched PubMed using the terms [GLP-1 receptor agonist] AND [visceral adiposity] FILTER [clinical trial] and found 25 results. Seven studies addressed visceral adiposity as a primary outcome. All studies demonstrated reductions in visceral adipose tissue (VAT). One major phase 3 trial reported significant absolute reductions in liver fat content at the end of 52 weeks of tirzepatide 10 mg and 15 mg, which was significantly correlated with reductions in VAT (p=0·29). Another study reported the combination of metformin and exenatide significantly decreased VAT and the ratio of VAT to subcutaneous adipose tissue. Liraglutide also demonstrated significant reductions in VAT among women with polycystic ovarian disease (VAT decreased by 18%), although the effects of dulaglutide and calorie restriction on VAT were similar to calorie restriction alone (−0·97 cm² [−14·36, 12·42]; p=0·884). Consequently, GLP1RAs have gained significance as an effective treatment for VAT and ectopic fat accumulation in non-HIV populations. However, while abnormal ectopic and visceral fat deposition is a hallmark of HIV-associated lipohypertrophy, it is a distinct entity from obesity and weight gain common in PWH after antiretroviral therapy initiation. To date, there are no published studies using GLP1RAs for the indication of HIV-associated lipohypertrophy.

Added value of this study

Few interventions exist for HIV-associated lipohypertrophy with all demonstrating mixed results on efficacy, poor response rate, administration barriers and/or possibility of harm. This study is the first to examine the GLP1RA, semaglutide, as a novel treatment option for HIV-associated lipohypertrophy. We investigated the effects of semaglutide on the quantity and distribution of adipose tissue among PWH and HIV-associated lipohypertrophy.

Implications of all the available evidence

Effects of semaglutide demonstrated substantial decreases in adipose tissue, particularly abdominal VAT, total body fat, and ectopic fat after 32 weeks of treatment that are likely to have a beneficial impact on cardiometabolic co-morbidity risk related to HIV-associated lipohypertrophy. We also observed notable improvements in glucose metabolism, insulin resistance, and lipid profiles. Semaglutide demonstrated a positive safety profile at the studied 1·0 mg dose with largely mild side effects similar to placebo. However, one case of semaglutide-related grade 4 elevated lipase and two cases of possibly-related cholelithiasis (grades 1 and 2) were identified. These and other rare but potentially serious side effects deserve further investigation in this unique population, particularly in trials using higher doses and longer treatment durations. This study supports the use of semaglutide as a promising treatment option for HIV-associated lipohypertrophy in PWH.