Longitudinal Changes in Body Composition, Adaptive Thermogenesis, and Muscle Strength in Patients with Obesity Treated with Semaglutide

Abstract

Introduction

Obesity management increasingly emphasizes not only total weight reduction but also the quality of weight loss, reflecting the balance between fat and lean tissue changes, preservation of strength capacity, and changes in energy expenditure. Semaglutide is effective for weight management, yet its integrated real-world effects on body composition, resting energy expenditure (REE), adaptive thermogenesis (AT), and muscle strength in adults with obesity without diabetes remain incompletely characterized.

Methods

We performed a single-center, retrospective observational study in routine care at the Obesity Treatment Unit (UTIO), Hospital Universitari Vall d’Hebron (Barcelona, Spain), January 2024–December 2025. Adults (≥18 years) with obesity (BMI ≥ 30 kg/m²) and without diabetes were included if baseline (V1) and 12-month (V2) data were available. Patients were classified as lifestyle educational program (n = 39) or lifestyle educational program plus semaglutide (n = 44), requiring ≥70% exposure during follow-up. Anthropometry, dominant handgrip strength, body composition by bioelectrical impedance analysis, and REE by indirect calorimetry were assessed at V1 and V2. Musculoskeletal remodeling was evaluated using continuous, age-, and BMI-dependent Skeletal Muscle Index Standardized Deviation Scores (SMI-SDS) validated against gold-standard MRI data. Between-group differences were tested using ANCOVA or rank ANCOVA with the absolute change from baseline (delta [Δ]) as the dependent variable, adjusted for the baseline age, gender, and body weight; partial eta squared (η²p) was reported. A study-specific predictive equation was derived to estimate REE and calculate AT from measured versus predicted REE at baseline and 12 months, with an external sensitivity analysis using the Ostendorf equation.

Results

Semaglutide group was associated with significantly greater 12-month reductions in body weight (Δ −11.0 kg, p = 0.001), BMI (Δ −4.0 kg/m², p = 0.001), and waist circumference (Δ −7.0 cm, p = 0.002) versus lifestyle group. WtHR showed a small, nonsignificant increase with lifestyle group (Δ + 0.01) versus a significant decrease in semaglutide group (Δ −0.04); the between-group difference was statistically significant after covariate adjustment (p < 0.001). Bioelectrical impedance analysis showed significant reductions in adiposity (including fat mass and indicative visceral fat area trends) in semaglutide group; skeletal muscle mass (SMM) and fat-free mass (FFM) also decreased in the semaglutide group (SMM = 29.9 to 28.7 kg; FFM = 53.8 to 52.7 kg; p < 0.001). Crucially, reference-based muscle evaluation revealed a highly significant structural improvement in SMI-SDS selectively in semaglutide group (delta 0.52 vs. 0.09 in lifestyle group; adjusted p < 0.001), demonstrating qualitative muscle mass optimization despite absolute lean tissue loss. Absolute handgrip strength was preserved in both groups, while relative handgrip strength improved with semaglutide (adjusted p = 0.003; η²p = 0.110). REE decreased in both groups; the adjusted analyses identified a p value at the limit of statistical significance (adjusted p = 0.050; η²p = 0.049). The AT analysis showed that part of this decline exceeded what would be expected from changes in FM, FFM, age, and gender. However, the magnitude of AT did not differ significantly between semaglutide group (delta AT: −210.2 kcal/day) and lifestyle group (delta AT: −373.4 kcal/day). RQ showed no differences between groups in terms of preference for fat oxidation.

Conclusion

In routine obesity care, semaglutide was associated with reductions in total adiposity and improvements in central anthropometric markers. Absolute handgrip strength was preserved in both groups. Crucially, reference-based muscle evaluation revealed a highly significant structural improvement in SMI-SDS selectively in semaglutide group. AT in both groups showed a decline in REE beyond that expected from FFM, FM, age, and gender, with no significant between-group differences. Comprehensive monitoring with body composition, handgrip strength, and indirect calorimetry may enhance follow-up of patients receiving obesity management medication.

Introduction

Obesity management is commonly followed using body weight and other anthropometric index, yet these measures provide limited insight into how weight change is partitioned between fat and lean compartments during obesity management. In clinical practice, bioelectrical impedance analysis (BIA) is a practical, noninvasive method to estimate body composition (BC) repeatedly and to complement anthropometric tracking with compartment-level information [1, 2].

Semaglutide, a glucagon-like peptide-1 receptor agonist (GLP-1RA), has demonstrated substantial efficacy for weight management in adults with overweight or obesity without type 2 diabetes. In the STEP 1 randomized trial, once-weekly subcutaneous semaglutide 2.4 mg combined with lifestyle education produced sustained, clinically meaningful weight reduction, and STEP 5 showed that weight loss can be maintained through 104 weeks with continued treatment [3, 4]. Importantly, subgroup analyses from STEP 1 have also characterized improvements in BC with semaglutide measured with dual-energy X-ray absorptiometry (DXA), including reductions in fat mass (FM) and visceral fat alongside changes in lean mass proportion [5]. However, beyond overall weight loss, there is ongoing interest in how semaglutide influences BC compartments when assessed using feasible tools such as BIA in specialized obesity care. Real-world reports have also highlighted the utility of BIA to characterize BC changes with semaglutide treatment, across oral and subcutaneous formulations, supporting the clinical relevance of compartment-level monitoring beyond weight alone [6, 7].

Weight loss is frequently accompanied by reductions in resting energy expenditure (REE). Notably, much of the evidence for this phenomenon, and its magnitude, is derived from long-term lifestyle education programs and post-bariatric surgery outcomes, where it is often discussed as adaptive thermogenesis (AT) in relation to changes in BC [8, 9]. Indirect calorimetry is currently the gold-standard method for the assessment of REE and can therefore complement BC assessment when characterizing metabolic changes during weight loss. Beyond energy expenditure, indirect calorimetry also determines the respiratory quotient (RQ), providing a critical assessment of substrate oxidation (the ratio of carbohydrate to lipid metabolism). Evaluating the RQ is essential to understanding how weight loss interventions influence metabolic flexibility and whether a shift toward fatty acid oxidation is sustained over time [10–12].

Weight loss is also accompanied by AT, a metabolic adaptation characterized by a reduction in REE that exceeds what would be expected from changes in BC alone [9, 13, 14]. Recent work has operationalized AT at the level of REE as a significantly lower measured versus predicted REE after weight loss and suggests that the magnitude of this response may be broadly comparable across weight-loss modalities, including very-low-energy diets and bariatric procedures [15]. Longitudinal evidence following intensive lifestyle-induced weight loss shows that measured REE can remain markedly suppressed years after the intervention, with persistent reduction in REE after adjustment for BC and age [9]. Importantly, comparable REE dynamics are also clinically relevant after bariatric surgery: in individuals with severe obesity, measured REE by indirect calorimetry decreases early after surgery and may remain reduced at 12 months, and the magnitude of early REE reduction has been reported to predict subsequent weight-loss outcomes and weight regain [8]. The work by Ostendorf and colleagues further proposed practical predictive equations to estimate AT in adults with obesity, using both normal-weight controls and controls with overweight/obesity as reference groups [14]. However, while these observations support the clinical relevance of AT across different weight-loss settings, data characterizing changes in REE and AT following GLP-1 receptor agonist therapy, such as semaglutide, remain notably scarce. Together, these studies underscore that AT may contribute to the difficulty of maintaining weight loss and highlight the value of indirect calorimetry when evaluating metabolic responses to obesity treatment.

Beyond the quantification of REE, AT, and tissue compartments, evaluating the preservation of muscle strength is critical in obesity management. While weight loss often involves a reduction in lean mass, fat-free mass (FFM), or skeletal muscle mass (SMM), it is essential to determine if such changes compromise physical performance [16, 17]. Muscle strength, specifically assessed via handgrip strength, serves as a reliable and practical proxy for overall musculoskeletal health and functional status in clinical settings [18–20]. Integrating functional strength assessments alongside BIA and indirect calorimetry allows for a comprehensive evaluation of weight loss, ensuring that adiposity reduction occurs without a concomitant decline in physical strength.

Furthermore, while tracking absolute changes in compartments such as FFM, FM, and SMM provides a descriptive overview, these raw parameters often fail to capture qualitative tissue remodeling and can look misleadingly preserved due to the non-muscular connective tissue expansions associated with progressive adiposity. To overcome these limitations and enhance clinical accuracy, it is critical to implement skeletal muscle reference data based on MRI-validated BIA estimates rather than relying on nonspecific or static metrics [21]. As highlighted by Enderle and colleagues [21], normal continuous reference ranges for BC – particularly the Skeletal Muscle Mass Index (SMI) – are dynamic and fundamentally co-dependent on continuous variations in both age and body mass index (BMI). Relying on unindexed cutoff values can mask underlying sarcopenic obesity as individuals with severe adiposity naturally require higher absolute muscle mass to support excess body weight [21]. Therefore, shifting from simple descriptive tracking to evaluating standard deviation scores against a reference framework validated by gold-standard magnetic resonance imaging (MRI) offers a significantly more robust approach to analyze true muscle preservation during pharmacologically assisted weight management [21].

On this basis, we conducted a real-world retrospective observational study in patients with obesity and without type 2 diabetes to compare 12-month changes in anthropometry, handgrip strength, BC, muscle mass, REE, and AT between patients managed with lifestyle educational program alone versus lifestyle educational program plus semaglutide.

Materials and Methods

Study Design, Setting, and Study Period

We conducted a single-center, retrospective observational study using routinely collected clinical data from the patients with obesity who attended the multidisciplinary Obesity Unit (UTIO), Vall Hebron University Hospital, Barcelona, Spain. The center is accredited as an EASO-COM (European Association for the Study of Obesity Collaborating Center for Obesity Management). The study period was January 1, 2024, to December 31, 2025. Assessments were analyzed at two time points: baseline (V1; prior to treatment initiation) and follow-up at 12 months (V2).

In UTIO routine clinical practice, we implement a collaborative approach where patients and clinicians jointly decide on lifestyle, medical, or surgical interventions for chronic obesity management [22]. All patients undergo a standardized baseline morphofunctional assessment, including BIA, indirect calorimetry, and muscle strength evaluation (handgrip). Baseline calorimetry and morphofunctional measurements are used to individualize the dietary and nutritional plan. All patients are also enrolled in a structured educational lifestyle program consisting of eight monthly sessions (one session per month over 8 months) of nutritional education, physical activity education, and lifestyle counseling delivered by the multidisciplinary (physician, nutritionists, psychologists, physiotherapists and nurses) UTIO team [22]. In accordance with UTIO clinical protocols, attendance at all eight sessions is a mandatory requirement for all patients (both the lifestyle and semaglutide treatment group) to remain in the treatment and follow-up program; consequently, all participants in both groups completed the full educational curriculum ensuring a 100% session attendance rate, the details of which are provided in online supplementary Table 1 (for all online suppl. material, see https://doi.org/10.1159/000553510).

Study Population and Eligibility Criteria

Eligible participants were adults (≥18 years) with obesity (BMI ≥ 30 kg/m²) without any type of diabetes, treated at UTIO who had available baseline (V1) and 12-month (V2) BIA measurements. Patients undergoing bariatric surgery during follow-up were excluded. The absence of diabetes was confirmed by clinical records, including the absence of a formal diagnosis based on standard clinical criteria, laboratory results below the diagnostic thresholds for diabetes during routine care, and no current use of glucose-lowering medications.

Exposure Definition and Group Assignment

According to the objectives of our study, participants were classified into two groups based on the treatment strategy received during follow-up: (i) lifestyle educational program alone (lifestyle group) and (ii) lifestyle educational program plus semaglutide (semaglutide group). Because this was a real-world observational study, treatment allocation was not randomized and reflected routine clinical decision-making guided by shared goals between clinicians and patients. Management options at UTIO are individualized and include a lifestyle educational program as the foundation for all patients, with the addition of obesity management medication (OMM) or metabolic surgery when clinically indicated and accepted by the patient. While the specific reasons for treatment selection were not documented, clinical experience at UTIO suggests that patients choosing exclusively the lifestyle modification program may face barriers to medication cost. Consequently, both study groups represent standard-of-care pathways within a real-world clinical environment.

To minimize medication exposure misclassification, assignment to the semaglutide group required documented semaglutide exposure for at least 70% of the planned 12-month follow-up period, as determined from clinical records; patients assigned to semaglutide who did not meet this exposure threshold were not included in the study. Semaglutide treatment included both oral and subcutaneous formulations. Oral daily doses used in the cohort were 7 mg and 14 mg, and subcutaneous weekly doses were 0.5 mg, 0.75 mg, 1.0 mg, 1.7 mg, and 2.4 mg; dosing and titration were individualized as part of routine care and were not protocolized in this retrospective study (online suppl. Table 2).

Variables and Measurements

Anthropometry and Clinical Variables

General variables included age, gender, and obesity treatment category (lifestyle group vs. semaglutide group). Anthropometric variables included body weight (kg), height (cm), BMI (kg/m²), waist circumference, and waist-to-height ratio (WtHR), collected at V1 and V2 during routine clinical visits.

Strength Outcomes

Muscle function outcomes included dominant handgrip strength and relative dominant handgrip strength (handgrip strength normalized to body weight), measured at V1 and V2 as part of UTIO routine clinical assessment.

Body Composition by BIA

BC was assessed using BIA with the InBody 970 (InBody Co., Ltd., Seoul, Republic of Korea). BIA-derived variables included adiposity and lean mass compartments (e.g., percent body fat, FM, FFM, SMM and related indices) and phase angle, as available in the UTIO database (definitions of BIA variables and their formulas are presented in online supplementary Table 3).

It must be noted that because the trunk contributes less than 12% to total body resistance compared to limbs, the device-derived visceral fat area (VFA) represents an indirect estimation based on cross-components mathematical inter-correlations rather than a direct volumetric measurement [23–25]. Consequently, VFA was analyzed strictly as a complementary descriptive trend alongside direct standard anthropometric markers.

Reference-Based Muscle Mass Evaluation

To determine whether longitudinal weight loss was associated with a clinically significant depletion of muscle mass beyond descriptive changes, individual skeletal muscle outcomes were calculated and evaluated using the continuous, age-, and BMI-dependent regression equations derived by Enderle and colleagues [21]. Importantly, these multifrequency BIA reference equations were calibrated and cross-validated using gold-standard imaging modalities, specifically full-body MRI for total SMM quantification, alongside DXA and air-displacement plethysmography within a comprehensive four-compartment reference model [21]. For each participant, three distinct variables were calculated at baseline (V1) and 12 months (V2):

• observed SMI by BIA (kg/m²): calculated directly from the BIA as SMM indexed to height squared (SMM/height²);

• predicted Skeletal Muscle Index (Predicted SMI, kg/m²): the expected continuous reference value for an individual’s specific age and BMI was calculated using the sex-specific multivariable regression models:

  • -for women: predicted SMI = 0.0274 x age − 0.000559 x age² + 0.1391 x BMI + 3.89;
  • -for men: predicted SMI = 0.0399 x age − 0.000668 x age² + 0.1740 x BMI + 4.79;

• Skeletal Muscle Index Standardized Deviation Score (SMI-SDS): to quantify the deviation of each patient’s observed muscle mass from their age- and BMI-matched predicted mean, the Standardized Deviation Score (or Z score) was computed using the formula:

  • -SMI-SDS = (observed SMI − predicted SMI)/SEE,

    • where SEE represents the standard error of estimation from the reference model (0.52 for women and 0.55 for men). A lower or more negative SMI-SDS reflects a higher degree of muscle deficit relative to an identity-matched healthy profile.

Indirect Calorimetry

Indirect calorimetry was used to assess longitudinal changes in REE at V1 and V2. REE was measured by indirect calorimetry using a Q-NRG metabolic monitor (COSMED, Rome, Italy) in canopy mode. Indirect calorimetry variables extracted from the Q-NRG report included REE, RQ, fat oxidation percentage, and REE normalized to fat-free mass (REE/FFM).

Study-Specific Prediction of REE and AT

To distinguish the expected reduction in REE attributable to BC changes from metabolic adaptation beyond predicted requirements, a study-specific predictive equation for REE was derived using baseline data from the complete-case analytic sample required for model construction and longitudinal AT assessment. The analytic sample comprised participants with complete data on age, gender, FM, and FFM, as well as measured REE by indirect calorimetry at both baseline (V1) and 12 months (V2).

A multivariable linear regression model was fitted with measured baseline REE as the dependent variable. Candidate predictors were selected from clinically relevant baseline demographic and BC variables, namely, FFM, FM, age, and gender (coded as 1 for men and 0 for women). The final study-specific equation included these four variables and showed a coefficient of determination (R2) of 0.561 and an adjusted R2 of 0.536. The final equation was as follows:

• Predicted REE = −76.048 + 21.814 × FFM + 17.842 × FM − 4.516 × age + 140.151 × gender.

In this model, FFM and FM were significant predictors of baseline REE, and the model explained 56.1% of the variance in measured baseline REE. Predicted REE was then calculated at baseline (V1) and 12 months (V2) by applying the same equation to the corresponding visit-specific variables. AT was defined as measured REE minus predicted REE at each time point. Accordingly, baseline AT was calculated as measured REE at V1 minus predicted REE at V1, and 12-month AT was calculated as measured REE at V2 minus predicted REE at V2. Delta AT was defined as AT at V2 minus AT at V1. Because delta AT variables were non-normally distributed, data were summarized as median [Q1–Q3]; unadjusted between-group comparisons were performed using Mann-Whitney U tests and adjusted analyses using rank-transformed ANCOVA (rank-ANCOVA) with treatment group as fixed factor and baseline age, gender, and body weight as covariates.

As an external sensitivity analysis, AT was also recalculated using the Ostendorf predictive equation as an external reference model [14]. This equation was selected because it was derived from a large reference sample that included adults with overweight/obesity, used indirect calorimetry and BC-related predictors aligned with the present study, and showed excellent predictive performance (R2 = 0.92). Compared with other published equations, it provided a more robust external reference for our cohort [9]. Predicted REE according to the Ostendorf equation was calculated as predicted REE = 650.8 + 16.9 × FFM + 7.6 × FM −4.5 × age +132 × gender, where male gender was coded as 1 for men and 0 for women. Results derived from Ostendorf equations are reported separately (online suppl. Table 5).

Study Objectives and Outcomes

The primary objective was to compare baseline-to-12-month changes in anthropometry, BC, SMM, REE, AT, and strength outcomes between patients of lifestyle group versus semaglutide group. We hypothesized that treatment with semaglutide would improve BC and could mitigate reductions in SMM, REE, AT, and strength during weight loss compared with lifestyle educational program alone.

Sample Size

A formal sample size calculation was not performed due to the exploratory and retrospective nature of the analysis.

Statistical Analysis

Statistical analyses were performed using IBM SPSS Statistics, version 20.0 (IBM Corp., Armonk, NY, USA). Continuous variables were summarized as mean ± standard deviation for normally distributed data and as median [interquartile range] for non-normally distributed data. Descriptive statistics were computed using an available-case approach, with missing data handled by pairwise deletion to maximize the use of available observations.

Normality of change scores (delta Δ = V2 − V1) was assessed using the Shapiro-Wilk test within each group, and inferential tests were selected accordingly. Within-group comparisons (V1 vs. V2) used paired t tests for normally distributed variables and Wilcoxon signed-rank tests for non-normally distributed variables. Unadjusted between-group comparisons (lifestyle group vs. semaglutide group) were performed on change scores using independent t tests (normal Δ) or Mann-Whitney U tests (non-normal Δ). All tests were two tailed, and statistical significance was set at p < 0.05.

To account for baseline clinical heterogeneity, adjusted between-group analyses were performed using ANCOVA (general linear model), with the absolute change from baseline (delta Δ) as the dependent variable and treatment group as the fixed factor. Covariates included the baseline value of age, gender, and body weight. The use of age and gender as covariates minimizes the impact of potential confounders. This ensures that the changes observed between groups are independent of these baseline characteristics. For outcomes not meeting parametric assumptions, rank-ANCOVA was applied. Partial eta squared (partial η²) was reported to quantify the magnitude of adjusted group effects (exploratory models of ANCOVA are described in online supplementary Table 4).

Results

Study Population and Baseline Characteristics

A total of 83 patients were included: 39 managed with lifestyle educational program alone (lifestyle group) and 44 managed with lifestyle educational program plus semaglutide (semaglutide group). The patient selection flow is shown in Figure 1. Baseline age, gender distribution, body weight, and BMI were comparable between groups (Table 1). No statistically significant between-group differences were observed at baseline for age, gender distribution, body weight, or BMI (all p > 0.05), indicating comparable baseline characteristics prior to intervention (Table 1).

Fig. 1.

Diagram cohort selection.

Table 1.

Study population and baseline characteristics

​	Lifestyle intervention group (n = 39)	Semaglutide group (n = 44)	p value
Age, years	43.4±12.76	46.9±15.45	0.261
Sex (female/male)	F (N = 31): 79.5%/M (N = 8): 20.5%	F (N = 37): 84.1%/M (N = 7): 15.9%	0.586
Body weight, kg	106.9 [97.5, 121.8]a	103.5 [94.02, 116.5]a	0.247
BMI, kg/m2	40.8±6.3	39.3±5.7	0.236

No statistically significant differences were observed between groups at baseline (p > 0.05 for all comparisons).

^a Data presented as mean ± SD or median [Q1–Q3].

Anthropometric and Strength Outcomes

Lifestyle group showed no significant change in body weight (median 106.9 kg at V1 vs. 108.8 kg at V2; p = 0.353) or BMI (40.8 ± 6.3 vs. 41.0 ± 6.3 kg/m²; p = 0.551) after 12 months, whereas waist circumference increased (114.0–117.0 cm; p = 0.026) and WtHR showed a nonsignificant upward trend (0.68–0.71; p = 0.091) (Table 2). In contrast, the semaglutide group experienced significant reductions in body weight (median 103.5 to 95.6 kg; p < 0.001), BMI (39.3 ± 5.7–35.2 ± 6.3 kg/m²; p < 0.001), waist circumference (111.0–108.0 cm; p = 0.002). Between-group analyses of delta scores (Δ = V2–V1) showed significantly greater improvements with semaglutide versus lifestyle educational program alone, in body weight (Δ −11.0 kg vs. + 1.1 kg; unadjusted p < 0.001), BMI (Δ −4.0 vs. + 0.1 kg/m²; unadjusted p < 0.001), waist circumference (Δ −7.0 vs. + 0.5 cm; unadjusted p < 0.001), and WtHR (Δ −0.04 vs. + 0.01; unadjusted p < 0.001) (Table 2; Fig. 2).

Table 2.

Anthropometric and strength outcomes

Variable	Lifestyle intervention group (n = 39)			Semaglutide group (n = 44)			Delta (Δ)
	baseline (V1)	follow-up (V2)	p value	baseline (V1)	follow-up (V2)	p value	lifestyle group	semaglutide group	p a (unadjusted)	p b (adjusted)	partial η2
Body weight, kg	106.9 [97.5, 121.8]a	108.8 [98.8–119.9]a	0.353	103.5 [94.02, 116.5]a	95.6 [77.1, 105.8]a	<0.001	1.1 [−1.9, 3.7]a	−11.0 [−13.9, −6.0]a	<0.001	<0.001	0.580
BMI, kg/m2	40.8±6.3	41.0±6.3	0.551	39.3±5.7	35.2±6.3	<0.001	0.1±1.7	−4.0±2.2	<0.001	<0.001	0.527
Waist circumference, cm	114.0 [103.7, 117.1]a	117.0 [110.0–124.0]a	0.026	111.0 [101.5, 128.0]a	108.0 [92.5, 122.0]a	0.002	0.5 [−0.5, 5.8]a	−7.0 [−10.3, −3.3]a	<0.001	<0.001	0.420
WtHR	0.68 [0.62, 0.73]a	0.71 [0.65, 0.76]a	0.091	0.67 [0.62, 0.78]a	0.67 [0.56, 0.74]a	0.001	0.01 [−0.01, 0.04]a	−0.04 [−0.07, −0.02]a	<0.001	<0.001	0.434
Dominant handgrip strength, kg	28.0 [22.0, 36.0]a	28.0 [21.0, 34.0]a	0.416	24.5 [20.5, 31.7]a	25.0 [20.0, 30]a	0.090	−2.0 [−3.0, 2.0]a	−1.5 [−3.7, 1.0]a	0.521	0.641	0.003
Relative dominant handgrip strength, kg/kg	0.26 [0.21, 0.32]a	0.25 [0.20, 0.30]a	0.247	0.25 [0.21, 0.29]a	0.26 [0.20, 0.33]a	0.003	0.00 [−0.03, 0.20]a	0.02 [0.00, 0.05]a	0.004	0.003	0.110

p: intra-group significance (V1vs. V2) via Wilcoxon or paired T test. p^a (unadjusted p value): calculated using independent-samples T test or Mann-Whitney U test on the raw deltas. p^b (adjusted p value): calculated via ANCOVA (general linear model) or rank-transformed ANCOVA, using baseline age, gender, and body weight as covariates. Delta (Δ): V2-V1.

^a Data presented as mean ± SD or median [Q1–Q3].

Fig. 2.

Major changes in BC.

Analysis was performed after adjustment using ANCOVA (or rank-ANCOVA when appropriate), with delta as the dependent variable, treatment group as fixed factor, and covariates including baseline value of age, gender, and body weight. Between-group differences remained significant for body weight (adjusted p < 0.001; partial η² = 0.580), BMI (adjusted p < 0.001; partial η² = 0.527), and waist circumference (adjusted p < 0.001; partial η² = 0.420). The adjusted between-group effect for WtHR was also statistically significant (adjusted p < 0.001; partial η² = 0.434).

Regarding muscle function, dominant handgrip strength did not significantly change within either group (lifestyle group: p = 0.416; semaglutide group: p = 0.090), and no between-group difference in change was observed after adjustment (unadjusted p = 0.521; adjusted p = 0.641; partial η² = 0.003). Relative dominant handgrip strength increased in the semaglutide group (p = 0.003) but not in the lifestyle group (p = 0.247); importantly, the between-group difference in change remained significant after adjustment (adjusted p = 0.003; partial η² = 0.110).

Bioelectrical Impedance Outcomes

In the lifestyle group, BIA-derived adiposity and lean compartments did not change significantly from V1 to V2 (Table 3). In the semaglutide group, adiposity markers decreased significantly from V1 to V2, including percent body fat (median 48.5%–44.8%, within-group p < 0.001), body FM (50.0–42.1 kg, p < 0.001), body FM index (18.7 ± 4.1–15.6 ± 4.8 kg/m², p < 0.001), and indirect estimates of VFA (238.7–203.1 cm², p < 0.001). Between-group comparisons of change (Δ = ΔV2–V1) showed significantly greater reductions in adiposity with semaglutide versus lifestyle group (all p^a < 0.001), and these differences remained statistically significant after covariate adjustment (all adjusted p < 0.001; partial η² = 0.408–0.566 for PBF/body fat mass index/body fat mass and 0.400 for VFA). Muscle and lean mass compartments also decreased in the semaglutide group (e.g., SMM 29.9–28.7 kg; FFM 53.852.7 kg; both within-group p < 0.001), with greater negative Δ versus lifestyle group in unadjusted analyses (p^a < 0.001) and persistent differences after adjustment (adjusted p < 0.001; partial η² = 0.251–0.265 across SMM/SMI/FFM/FFMI/BCM) (Table 3).

Table 3.

Bioelectrical impedance outcomes and reference-based muscle mass evaluation outcomes

Variable	Lifestyle group (n = 39)			Semaglutide group (n = 44)			Delta (Δ)
	baseline (V1)	follow-up (V2)	p value	baseline (V1)	follow-up (V2)	p value	lifestyle intervention group	semaglutide group	p a (unadjusted)	p b (adjusted)	partial η2
Bioelectrical impedance outcomes
PBF (%)	48.4 [43.7, 50.9]a	48.6 [44.9, 51.1]a	0.411	48.5 [44.2, 51.1]a	44.8 [38.9, 48.3]a	<0.001	0.0 [−1.0, 1.6]a	−3.1 [−4.7, −1.2]a	<0.001	<0.001	0.408
BFM, kg	51.9 [43.4, 59.5]a	53.4 [43.7, 58.4]a	0.285	50.0 [40.5, 58.2]a	42.1 [30.9, 51.5]a	<0.001	0.3 [−1.9, 4.3]a	−8.1 [−10.5, −4.7]a	<0.001	<0.001	0.566
BFMI, kg/m2	19.1±4.9	19.3±4.9	0.261	18.7±4.1	15.6±4.8	<0.001	0.2±1.5	−3.2±1.7	<0.001	<0.001	0.496
VFA, cm2	249.5 [214.0, 263.3]a	239.9 [217.7, 259.0]a	0.316	238.7 [208.5, 263.2]a	203.1 [151.6, 239.9]a	<0.001	−4.1 [−14.1, 6.7]a	−32.4 [−45.4, −19.3]a	<0.001	<0.001	0.400
FFM, kg	57.2 [51.3, 67.0]a	56.7 [50.9, 66.1]a	0.421	53.8 [47.5, 58.4]a	52.7 [46.6, 55.5]a	<0.001	0.0 [−1.8, 1.2]a	−2.3 [−3.7, −0.7]a	<0.001	<0.001	0.251
FFMI, kg/m2	21.5 [19.9, 23.1]a	21.4 [19.8, 23.2]a	0.706	20.3 [18.5, 22.6]a	19.6 [17.9, 21.2]a	<0.001	0.1 [−0.7, 0.5]a	−0.9 [−1.4, −0.3]a	<0.001	<0.001	0.272
Body PhA, °	5.5±0.6	5.5±0.6	0.492	5.3±0.6	5.1±0.6	<0.001	−0.03±0.30	−0.17±0.30	0.054	0.132	0.029
BCM, kg	36.9 [33.1, 44.0]a	36.7 [33.2, 43.2]a	0.264	35.0 [30.9, 39.5]a	33.8 [30.0, 35.9]a	<0.001	0.0 [−1.4, 0.8]a	−1.7 [−2.6, −0.7]a	<0.001	<0.001	0.266
SMM, kg	31.6 [28.2, 38.1]a	31.4 [28.2, 37.4]a	0.273	29.9 [26.1, 33.9]a	28.7 [25.4, 30.8]a	<0.001	−0.1 [−1.4, 0.6]a	−1.6 [−2.4, −0.6]a	<0.001	<0.001	0.258
SMI, kg/m2	8.6 [7.9, 9.3]a	8.6 [8.1, 9.2]a	0.900	8.2 [7.3, 9.2]a	8.0 [7.1, 8.7]a	<0.001	0.1 [−0.1, 0.2]a	−0.3 [−0.5, −0.0]a	<0.001	<0.001	0.295
Reference-based muscle mass evaluation outcomes
Observed SMI by BIA, kg/m2	9.92 [8.71, 10.70]a	9.61 [8.72, 10.74]a	0.264	9.43 [8.57, 10.59]a	9.21 [8.35, 9.94]a	<0.001	−0.11 [−0.49, 0.17]a	−0.38 [−0.64, −0.16]a	<0.001	<0.001	0.155
Predicted SMI, kg/m2	10.16±1.48	10.19±1.50	0.900	9.70±1.44	9.11±1.38	<0.001	0.03±0.25	−0.60±0.37	<0.001	<0.001	0.520
SMI-SDS	−2.77±1.59	−2.68±1.49	<0.001	−2.65±1.72	−2.13±1.60	<0.001	0.09±0.51	0.52±0.50	<0.001	<0.001	0.148

p: intra-group significance (V1vs. V2) via Wilcoxon or paired T test. p^a (unadjusted p value): calculated using independent-samples T test or Mann-Whitney U test on the raw deltas. p^b (adjusted p value): calculated via ANCOVA (general linear model) or rank-transformed ANCOVA, using baseline age, gender, and body weight as covariates. Delta (Δ): V2-V1.

^a Data presented as mean ± SD or median [Q1–Q3].

Body phase angle (PhA) showed a borderline between-group difference in delta (Δ = V2−V1) in unadjusted analyses (lifestyle group: Δ−0.03 ± 0.3° vs. semaglutide group: Δ−0.17 ± 0.3°; p^a = 0.054), which did not remain statistically significant after covariate adjustment (adjusted p = 0.132; partial η² = 0.029). Major changes in BC are represented in Figure 2.

Reference-Based Muscle Mass and Sarcopenic Profile

Evaluation of muscle mass compartments against the continuous external reference model revealed distinct trajectories between treatment groups. At baseline, both groups exhibited a marked muscle deficit relative to their age- and BMI-matched healthy peers, as indicated by severely depressed baseline SMI-SDS scores (lifestyle group: −2.77 ± 1.59; semaglutide group: −2.65 ± 1.72).

Over the 12-month follow-up, the expected continuous requirement for muscle mass (predicted SMI) shifted downward exclusively in the semaglutide group (from 9.70 ± 1.44 to 9.11 ± 1.38 kg/m²; within-group p < 0.001) as a physiological consequence of substantial total weight and BMI reductions. In contrast, predicted SMI remained unchanged in the lifestyle group (10.16 ± 1.48 vs. 10.19 ± 1.50 kg/m²; p = 0.900).

Crucially, when evaluating the standardization of true muscle preservation, the semaglutide group experienced a statistically significant and clinically favorable increase in the Standardized Deviation Score (SDS), moving closer to the reference average (SMI-SDS at 12 months: −2.13 ± 1.60; within-group delta: 0.52 ± 0.50, p < 0.001). Conversely, the lifestyle group demonstrated unmeaningful changes in its muscle deficit profile (SMI-SDS at 12 months: −2.68 ± 1.49; delta: 0.09 ± 0.51). Independent-samples T test on the unadjusted scores indicated a significantly greater improvement in the SMI-SDS with semaglutide versus lifestyle alone (p^a < 0.001). After multivariable covariate adjustment using ANCOVA, the between-group difference in the longitudinal change of SMI-SDS remained highly statistically significant (adjusted p < 0.001), demonstrating a moderate-to-large group effect size (partial η² = 0.148) favoring the semaglutide group.

Indirect Calorimetry Outcomes

REE decreased significantly from V1 to V2 in both groups (lifestyle group: 1,935.0–1,543.0 kcal/day, p = 0.001; semaglutide: 1,805.5–1,235.0 kcal/day, p < 0.001). The between-group comparison of ΔREE was not statistically significant in unadjusted analyses (p^a = 0.117). The adjusted between-group difference in REE was at the limit of statistical significance (adjusted p = 0.050; partial η² = 0.049) (Table 4).

Table 4.

Indirect calorimetry outcomes

Variable	Lifestyle intervention group (n = 39)			Semaglutide group (n = 44)			Delta (Δ)
	baseline (V1)	follow-up (V2)	p value	baseline (V1)	follow-up (V2)	p value	lifestyle intervention group	semaglutide group	p a (unadjusted)	p b (adjusted)	partial η2
REE, kcal/day	1,935.0 [1,677.0, 2245.0]a	1,543.0 [1,358.0, 1,903.0]a	0.001	1,805.5 [1,465.5, 2,144.5]	1,235.0 [1,036.7, 1567.5]	<0.001	−329.0 [−527.0, 0.0]	−398.0 [−724.7, −139.7]	0.117	0.050	0.049
RQ (adimensional)	0.71 [0.67, 0.76]a	0.68 [0.64, 0.73]a	0.049	0.72 [0.67, 0.75]a	0.68 [0.65, 0.72]a	0.018	−0.01 [−0.07, 0.03]a	−0.02 [−0.09, 0.01]a	0.591	0.719	0.002
Fat oxidation, %	99.0 [80.0, 100.0]a	100.0 [85.0, 100.0]a	0.317	93.5 [83.5, 100.0]a	100.0 [96.2, 100.0]a	0.102	0.0 [0.0, 6.0]a	0.0 [0.0, 12.0]a	0.313	0.404	0.009
REE/FFM, kcal/kg FFM/day	34.0 [30.0, 38.6]a	26.8 [24.0, 32.0]a	0.001	32.5 [30.2, 37.0]a	24.3 [21.7, 27.8]a	<0.001	−6.4 [−10.0, 0.6]a	−7.5 [−13.0, −2.7]a	0.171	0.094	0.040

p: intra-group significance (V1 vs. V2) via Wilcoxon or paired T test. p^a (unadjusted p value): calculated using independent-samples T test or Mann-Whitney U test on the raw deltas. p^b (adjusted p value): calculated via ANCOVA (general linear model) or rank-transformed ANCOVA, using baseline age, gender, and body weight as covariates. Delta (Δ): V2-V1.

^a Data presented as mean ± SD or median [Q1–Q3].

REE/FFM decreased significantly from V1 to V2 in both groups (lifestyle group: 34.0–26.8 kcal/day/kg FFM, p = 0.001; semaglutide group: 32.5–24.3 kcal/day/kg FFM, p < 0.001). The between-group difference in REE/FFM change was not statistically significant in either unadjusted or adjusted models (adjusted p = 0.094; partial η² = 0.040). For secondary calorimetry outcomes, RQ decreased modestly within both groups (lifestyle group p = 0.049; semaglutide group p = 0.018), but neither ΔRQ nor ΔFat oxidation differed between groups in adjusted analyses (adjusted p = 0.719 and adjusted p = 0.404, respectively) (Table 4).

AT Based on the Study-Specific Predictive Equation

Using the study-specific equation, baseline AT was close to neutral in both groups, with median values of 22.2 kcal/day in the lifestyle group and 7.0 kcal/day in the semaglutide group. At 12 months, AT shifted downward in both groups, reaching −346.7 kcal/day in the lifestyle group and −312.7 kcal/day in the semaglutide group (Table 5). The median delta AT was −373.4 kcal/day in the lifestyle intervention group and −210.2 kcal/day in the semaglutide group, indicating a longitudinal reduction in measured REE beyond that predicted from BC, age, and sex in both treatment groups (Table 6). However, between-group differences were not statistically significant in either unadjusted or adjusted analyses (adjusted p = 0.961). Similar results were observed with the Ostendorf equation as an external reference model, which also showed no significant between-group difference in delta AT (adjusted p = 0.274). These results are presented in Tables 5 and 6 and online Supplementary Table 5.

Table 5.

REE and AT derived from the study-specific predictive equation, by treatment group

Variable, kcal/day	Lifestyle group	Semaglutide group
Measured REE at baseline (V1)	1,935.0 [1,606.0, 2,250.0]a	1,833.0 [1,497.3, 2,144.5]a
Measured REE at 12 months (V2)	1,543.0 [1,358.0, 1,903.0]a	1,235.0 [1,036.8, 1,567.5]a
Predicted REE at baseline (V1)	1,923.5 [1,717.8, 2,284.0]a	1,757.9 [1,548.4, 2,091.3]a
Predicted REE at 12 months (V2)	1,923.3 [1,716.5, 2,163.4]a	1,578.2 [1,324.1, 1,848.9]a
AT at baseline (V1)	22.2 [−195.2, 226.5]a	7.0 [-88.5, 251.5]a
AT at 12 months (V2)	−346.7 [−587.8, −8.9]a	−312.7 [-519.6, −36.9]a
Delta AT	−373.4 [−558.7, −12.5] a	−210.2 [-584.2, −23.9] a

Measured REE corresponds to REE obtained by indirect calorimetry at baseline (V1) and 12 months (V2). Predicted REE values were calculated using the study-specific predictive equation. AT was defined as measured REE minus predicted REE at each time point; longitudinal AT (or delta AT) was calculated as AT at V2 minus AT at V1. Analyses were restricted to participants with complete data required for the predictive model and AT calculation (baseline age, sex, FFM, FM, and measured REE at V1 and V2. Complete-case analytic sample for regression models: n = 75 (lifestyle group, n = 35; semaglutide group, n = 40).

^a Data presented as median [Q1–Q3].

Table 6.

AT delta outcomes

Variable	Lifestyle intervention group	Semaglutide group	p a unadjusted	p b adjusted	Partial η2
Delta (V2-V1) AT, study-specific equation, kcal/day	−373.4 [−558.7, −12.5]a	−210.2 [−584.2, −23.9]a	0.588	0.961	0.000
Delta (V2-V1) AT, Ostendorf equation, kcal/day	−394.7 [−561.5, −17.9]a	−321.6 [−668.9, −126.6]a	0.559	0.274	0.017

p ^a (unadjusted p value): calculated using independent-samples T test or Mann-Whitney U test on the raw deltas. p^b (adjusted p value): calculated via ANCOVA (general linear model) or rank-transformed ANCOVA, using baseline age, gender, and body weight as covariates. Delta (Δ): V2-V1.

^a Data presented as median [Q1–Q3].

Discussion

To the best of our knowledge, this is the first real-world study to evaluate 12-month longitudinal variations in BC compartments, metabolic adaptation, and functional strength in adults with obesity without diabetes managed with a lifestyle educational program plus semaglutide versus a lifestyle educational program alone, integrating BIA and indirect calorimetry. According to our findings, the semaglutide treatment was associated with substantial reductions in total adiposity and significant improvements in central anthropometric indices, alongside a contextual preservation of absolute musculoskeletal strength. Crucially, while absolute FFM and SMM expectedly declined in parallel with total weight reduction, normalization against an external continuous reference population (SMI-SDS) confirmed that this tissue loss represents a physiological structural realignment to a lower physical load requirement rather than a detrimental wasting process. Furthermore, this structural adaptation was functionally supported by the maintenance of absolute muscle strength and a significant optimization of the relative strength-to-weight profile.

Semaglutide group was associated with greater improvements in anthropometric outcomes including reductions in body weight, BMI, WtHR, and waist circumference, with these differences remaining significant after multivariable adjustment despite the reduction in REE that was evidenced with the indirect calorimetry and the presence of AT. The direction and magnitude of these changes are broadly consistent with randomized evidence from the STEP program, in which once-weekly semaglutide 2.4 mg combined with lifestyle educational program produced clinically meaningful weight loss and sustained benefit with continued treatment [3, 4]. When interpreting these efficacy and BC findings, it is crucial to acknowledge the pharmacological heterogeneity of our semaglutide group. As detailed in online supplementary Table 2, patients received a mix of oral and subcutaneous formulations across different doses, and only a subset reached the maximum 2.4 mg subcutaneous dose. Therefore, this group should not be interpreted as a homogeneous cohort receiving maximal therapy. However, this heterogeneity accurately reflects real-world clinical practice, where dose titration could be frequently limited by gastrointestinal tolerability, patient preference, or other socioeconomic factors. The observation that semaglutide induced highly significant improvements in adiposity compartments and central obesity highlights the real-world effectiveness of this therapy beyond the tightly controlled conditions of randomized clinical trials.

Our results describe not only the trajectory of absolute REE during weight loss but also the degree to which the observed reduction exceeded that expected from BC and demographic characteristics. It is important to distinguish between the expected reduction in REE and true AT, which refers to a decrease in REE beyond that predicted by BC changes [26, 27].

In the present study, a cohort-specific predictive equation based on FFM, FM, age, and gender allowed this distinction to be explored in the complete-case analytic sample. Using this approach, AT shifted significantly downward at 12 months, reaching −346.7 kcal/day in the lifestyle group and −312.7 kcal/day in the semaglutide group. These negative values robustly confirm the clinical presentation of an active longitudinal metabolic slowing in both treatment groups, demonstrating that a substantial portion of the longitudinal decline in REE was not fully explained by concurrent variations in measured tissue compartments. However, between-group differences in delta AT were not statistically significant (adjusted p = 0.961), and similar findings were obtained when the Ostendorf equation was used as an external reference model (adjusted p = 0.274). Accordingly, while these results do not support a clearly distinct or exaggerated AT response in the semaglutide group compared to lifestyle group, they definitively highlight that both strategies trigger a significant biological energy conservation penalty during real-world management.

This finding is clinically relevant because it indicates that the metabolic response to weight loss includes a component of energy conservation beyond the expected tissue-related decline in REE. Such adaptation may contribute to the biological pressure toward weight regain and may help explain why weight-loss maintenance remains challenging even when substantial adiposity reduction is achieved. Importantly, from a clinical safety standpoint, there has been an ongoing concern that the substantial and rapid weight loss achieved with potent anti-obesity medications like semaglutide could trigger an unproportionately severe metabolic counter-regulation, leading to an AT thermogenesis response. However, our data uniquely challenge this concern, demonstrating that despite a major ∼11 kg weight reduction in the semaglutide group, the delta AT was not exacerbated and remained statistically equivalent to – and numerically lower than – that of the lifestyle group.

This preservation of metabolic efficiency aligns with emerging evidence, which suggests that long-term semaglutide therapy effectively reduces FM while preserving metabolic efficiency and stabilizing lean mass compartments over time [28]. In the lifestyle group, this interpretation is particularly notable because results showed a median delta AT of −373.4 kcal/day – despite limited group-level changes in body weight and most BIA-derived compartments. This raises the possibility that unmeasured behavioral changes (such as subtle dietary restriction), neuroendocrine counter-regulations, or qualitative tissue-level changes influenced energy expenditure beyond what was captured by the available BC measures. Conversely, the semaglutide group experienced a comparable reduction in energy expenditure (median delta AT: −210.2 kcal/day) but achieved a highly significant ∼11 kg weight loss. This suggests that while semaglutide does not prevent the physiological onset of AT, its central mechanisms of action are effective enough to sustain a calorie deficit that successfully overrides this concurrent metabolic drag, achieving superior clinical efficacy without compounding the expected metabolic penalty.

At the same time, the AT analysis should be interpreted cautiously. Because BC was assessed by BIA rather than by reference methods such as DXA, predicted REE and AT estimates should be considered clinically informative rather than definitive mechanistic measures. Potential variations in hydration status or the inherent assumptions of BIA prediction equations in severe obesity could subtly affect the FFM and FM inputs used in our regression models. Therefore, future prospective studies are needed to validate these observations in external cohorts with more detailed BC phenotyping.

A major strength of our analysis is the compartment-level characterization of weight loss using BIA, which enables interpretation of tissue redistribution dynamics beyond total kilograms lost. In contrast to the lifestyle group, which showed no significant BIA changes over 12 months, semaglutide was associated with robust reductions in adiposity markers (percent body fat, FM, FMI, and indirect estimates of VFA), and all between-group differences remained significant after adjustment with moderate-to-large effect sizes, noting that regional trunk outputs represent indicative trends due to the biophysical limits of torso impedance measurements. This profile is consistent with semaglutide BC findings reported in STEP 1 ancillary analyses, where weight loss was largely attributable to FM reduction, including favorable changes in regional/visceral adiposity [5, 17]. Real-world evidence has similarly highlighted that semaglutide (oral or subcutaneous) is associated with improvements in BC assessed by BIA, supporting the clinical relevance of compartment monitoring in routine practice [6].

Lean compartments also reduced in the semaglutide group (SMM, SMI, FFM, FFMI, BCM), which is physiologically expected during weight loss and has been emphasized in systematic reviews as a key aspect to monitor during GLP-1RA therapy [16, 29]. Because SMM, FFM, BCM, and phase angle were derived from BIA, their absolute changes and modest between-group differences should be viewed as indicative of trends rather than as precise quantitative estimates of organ or tissue-level remodeling.

Nevertheless, the BIA absolute reductions in SMM, FFM, BCM, and phase angle observed in the semaglutide group indicate that clinically relevant losses of lean tissue and cellular integrity can still occur during pharmacologically assisted weight loss, particularly when weight loss is large or prolonged. These findings underscore the importance of monitoring lean compartments over time and of combining GLP-1 receptor agonists with resistance exercise and protein-adequate nutritional strategies to mitigate sarcopenic risk [30].

When evaluating these lean mass dynamics, a key technological and biological distinction must be considered regarding how BIA algorithms partition active tissue. Commercial multifrequency BIA devices traditionally utilize regression equations calibrated against DXA reference metrics rather than direct anatomical gold standards like MRI or computed tomography. While DXA serves as a valid clinical standard, it fundamentally quantifies total lean body mass or FFM, which encapsulates non-muscular elements such as extracellular water, connective tissues, and visceral organs, rather than isolated SMM. This differentiation is of paramount physiological interest during active weight reduction; whereas SMM contributes to only approximately 45%–50% of total FFM under baseline conditions, longitudinal evidence demonstrates that nearly 100% of the subsequent weight-loss-induced decrease in FFM is structurally explained by a net reduction in actual skeletal muscle tissue [31]. Consequently, relying solely on crude, DXA-linked BIA lean outputs may fail to capture true muscular compartmentalization, potentially masking localized tissue changes.

A pivotal finding of our study emerges from the continuous, reference-based standardization of muscle mass index (SMI-SDS), which directly addresses the clinical ambiguity of absolute muscle loss during pharmacologically induced weight reduction. When evaluating crude BIA metrics, the absolute decline in SMM and SMI within the semaglutide group could be easily misinterpreted as detrimental muscle wasting. However, by standardizing these values against an external continuous framework adjusted for age and BMI, we demonstrated a significant upward shift in the SMI-SDS toward the healthy population mean [21]. This physiological adaptation reflects that the absolute reduction in muscle mass is not a sign of tissue degradation but rather a structural realignment to a lower total body mass and decreased physical load requirements. Conversely, the lack of significant SMI-SDS remodeling in the lifestyle group underlines that conservative approaches fail to correct the deep baseline muscle-to-weight mismatch characteristic of severe obesity. Consequently, incorporating continuous reference standard scores offers a more robust and realistic diagnostic tool for monitoring muscle health in active weight management than static cutoff values.

Semaglutide group had significant reductions in waist circumference, WtHR, and VFA compared with lifestyle group, consistent with a preferential improvement in central adiposity. Central adiposity is strongly linked to cardiometabolic risk, and STEP 1 similarly reported larger waist reductions with semaglutide compared with placebo alongside favorable BC shifts. From a clinical standpoint, these findings support using both waist measures and compartment estimates to capture risk-relevant changes during obesity treatment [3, 5]. However, in alignment with the physical principles of BIA, trunk-level measurements exhibit inherent baseline limits due to low resistance relative to limbs [23–25]. Therefore, BIA-derived visceral fat tracking does not offer clinical superiority over direct anthropometric estimates like waist circumference and should be contextualized purely as a surrogate reflecting broader inter-correlated tissue changes. Additional clinical trials, specifically SELECT and OASIS 2, have further substantiated the efficacy of semaglutide in achieving superior reductions in waist circumference and visceral adipose tissue compared to placebo or lifestyle education alone. These findings remain consistent across both subcutaneous and oral formulations [32–34].

Both groups experienced within-group reductions in REE, and adjusted analyses identified a between-group difference in REE at the limit of statistical significance at follow-up, whereas adjusted difference in RQ was not significant. Decreases in REE during weight loss are frequently discussed in contemporary reviews that emphasize such energy-expenditure changes can be clinically relevant for long-term weight maintenance [27]. Clinically, the reduction in REE observed in both groups highlights the metabolic challenge of weight-loss maintenance. In the lifestyle group, REE decreased even in the absence of significant weight loss, potentially creating a lower energy threshold that makes further weight reduction through lifestyle alone increasingly difficult. In contrast, the substantial weight loss achieved with semaglutide suggests that its mechanism of action is sufficient to sustain a calorie deficit that overrides this concurrent reduction in REE, effectively maintaining weight loss.

Notably, reduction in REE may also be relevant when medication is discontinued as recent evidence indicates that cessation of OMM is followed by clinically meaningful weight regain [35]. In a recent systematic review and meta-analysis, weight regain after stopping weight management medications was consistently observed across drug classes, reinforcing the concept that obesity treatment effects often attenuate when active therapy is withdrawn. Specifically, authors reported that weight regain after treatment cessation was faster with newer, more effective incretin mimetics (including semaglutide), reaching an estimated rate of 0.8 kg per month [35]. These observations, together with trial extension data showing substantial regain after semaglutide withdrawal, support the hypothesis that treatment-associated reductions in energy expenditure could contribute – alongside behavioral and neuroendocrine mechanisms – to post-discontinuation weight regain, and highlight the value of assessing REE by indirect calorimetry in real-world obesity care. Furthermore, this is clinically important and underscores the need to view obesity as a chronic disease requiring sustained, long-term management, rather than short-term medication courses. In this context, medication withdrawal may promote weight regain, particularly if it occurs against a background of reduced REE and AT.

Preserving function while reducing adiposity is crucial to obesity care, particularly in severe obesity where deconditioning and mobility limitations are common. Handgrip strength is increasingly viewed as a simple marker of muscle function and overall health risk, supporting its inclusion alongside BC endpoints. In our cohort, absolute dominant handgrip strength did not significantly change in either group and did not differ between groups after adjustment, suggesting that clinically meaningful strength loss did not accompany semaglutide-associated weight reduction. Moreover, relative handgrip strength improved modestly in the semaglutide group and remained significantly different after adjustment versus lifestyle group, while recognizing that concurrent reductions in BIA-derived lean compartments highlight the need for ongoing surveillance of muscle health and function during long-term OMM. This functional preservation is highly coherent with our reference-based structural data, confirming that the decline in crude SMM did not impair physical performance because tissue loss represented a proportional load realignment (SMI-SDS) rather than qualitative degradation. This is clinically relevant because relative strength integrates neuromuscular performance with body mass, and prior work suggests relative handgrip strength can mitigate functional dependence associated with excess adiposity. Our findings therefore support a treatment profile in which fat loss and central adiposity improvements occur without compromising – and potentially improving – strength-to-weight balance, which may translate into better physical capability in daily activities [36, 37].

Limitations of the Study

This study should be interpreted in light of several limitations. First, the retrospective, single-center observational design does not allow for definitive causal inference, and there is a relevant risk of confounding by indication. Because treatment allocation reflected routine clinical decision-making rather than randomization, it was inherently influenced by patient preference, socioeconomic factors (such as medication affordability or access), and clinician judgment. Although our statistical models adjusted for baseline age, gender, and body weight, we acknowledge that we could not account for critical unmeasured confounders. Patients opting for semaglutide may differ from the lifestyle educational group in terms of intrinsic motivation, eating behaviors, previous weight-loss treatment history, or specific comorbidities. Consequently, selection bias cannot be entirely excluded, and our findings must be interpreted strictly as real-world clinical associations rather than isolated pharmacological effects. Nevertheless, this real-world approach is highly advantageous for identifying clinical trends and providing a scalable framework for other tertiary centers to organize the integrated treatment and longitudinal monitoring of patients with obesity. These data serve as a vital precursor highlighting the need for future prospective, randomized, multicenter investigations. Second, because this was a single-center, retrospective study conducted in a tertiary obesity unit, and included adults with obesity without diabetes undergoing routine BIA and indirect calorimetry, our findings may be most generalizable to similar specialized clinical settings. Moreover, only a small subset of the 1,447 patients initially assessed fulfilled our inclusion criteria (Fig. 1), largely because eligibility required complete baseline (V1) and 12-month (V2) morphofunctional assessments, including indirect calorimetry and BIA, within the study window. The implementation of these assessments at our center and the progressive national availability of semaglutide during the study period meant that many patients could be excluded for technical and organizational reasons (e.g., absence of a V2 assessment with full measurements), in addition to any patient-level adherence issues. Generalizability to broader populations (e.g., primary care, different health systems, or patients with diabetes) and to centers with different prescribing patterns and follow-up intensity should be confirmed in larger, multicenter cohorts. Third, treatment exposure and follow-up assessments were derived from routine clinical records. Furthermore, while all patients adhered to the mandatory eight-session structured lifestyle educational program as a condition of their clinical follow-up, we lack specific quantitative or psychometric measures to evaluate how this education translated into individual behavioral changes. This absence of granular behavioral data represents a potential unmeasured confounding factor inherent to the retrospective design of the study. Additionally, the specific factors influencing the choice of treatment strategy (lifestyle group or semaglutide group) were not formally documented, which may introduce selection bias despite the use of a lifestyle-only comparison group managed under the same institutional protocols. Fourth, treatment exposure and follow-up assessments were derived from routine clinical records; although we applied an a priori ≥70% semaglutide exposure threshold to strengthen exposure classification, adherence could not be verified through patient-reported administration logs or home diaries. Fifth, missing data were handled using an available-case approach, which maximizes data use but may introduce bias if data are not missing at random. Sixth, while our models accounted for age and gender as covariates, this study was not specifically powered to explore differential treatment responses through stratified subgroup analyses or specific dose-response relationships. Because stratifying our cohort by formulation (oral versus subcutaneous) or dose intensity would result in subgroups too small for robust statistical inference, future adequately powered research is needed to clarify how these different dosing regimens influence individual responses to GLP-1 receptor agonists. Finally, BC was assessed by BIA, a feasible clinical method but one that may be influenced by hydration status and prediction-equation assumptions in severe obesity. As a result, our conclusions regarding the proportional contribution of SMM and FFM to weight loss should be interpreted with caution and considered descriptive, real-world associations rather than definitive compartment-level inferences.

Conclusion

Our results demonstrate that a 12-month treatment with semaglutide combined with a lifestyle educational program leads to a significant reduction in total adiposity and improvements in central anthropometric indices in adults with obesity without diabetes. In line with our primary objective, this therapeutic strategy achieved greater changes in body weight, BMI, and descriptive BIA-derived adiposity compartments compared to lifestyle educational intervention alone. Crucially, when evaluating these changes against a continuous reference population, the absolute decrease in SMM observed in the semaglutide group represents a structural realignment to a lower total body weight – demonstrated by a significant improvement in the SMI-SDS towards the reference average – rather than a detrimental muscle wasting process. Regarding our physiological hypotheses, absolute functional strength was maintained in both treatment cohorts, while relative handgrip strength improved selectively in the semaglutide group. Furthermore, although measured REE declined in both groups, the magnitude of the longitudinal AT did not differ significantly between groups, indicating that semaglutide does not exacerbate the metabolic counter-regulation associated with weight reduction. These real-world findings reinforce the clinical utility of comprehensive multi-modal monitoring. However, given the inherent mathematical assumptions of trunk-level bioimpedance estimates, regional outputs such as VFAs should be interpreted strictly as indicative trends alongside standardized anthropometric measures like waist circumference. Ultimately, these data underscore that managing obesity as a chronic condition requires the application of dynamic, standardized metrics to accurately capture and evaluate musculoskeletal status during treatment with OMM.

Acknowledgments

This work has been carried out by Francesca Filippi-Arriaga within the framework of the Doctoral Program in Medicine of the Universitat Autònoma de Barcelona. The authors utilized AI-assisted technologies for grammar and linguistic refinement and preliminary data validation during the preparation of this manuscript. The final analysis and interpretation of the data were performed solely by the authors, who maintain full responsibility for the content and integrity of the work.

Statement of Ethics

The study protocol was approved by the Clinical Research Ethics Committee (CEIm) of Hospital Universitari Vall d’Hebron (Barcelona, Spain) (CEIm code: EOM(AG)026/2025(6498). The CEIm issued a favorable opinion and approved the proposed waiver of informed consent according to the design and implications of the study.

Conflict of Interest Statement

A.C. reports speaking fees from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Novo Nordisk, Sanofi, and Menarini and research grants from Eli Lilly, Novo Nordisk, and Menarini. A.C. is a member of a Boehringer Ingelheim Data Monitoring Committee. E.F. reports speaking fees from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Novo Nordisk, and Sanofi. P.C. reports speaking fees from AstraZeneca, Boehringer Ingelheim, Eli Lilly, and Novo Nordisk. V.F. reports speaking fees from Eli Lilly and Novo Nordisk. M.C. reports speaking fees from Eli Lilly and Novo Nordisk. F.F.-A., A.P., A.J., E.V., and C.O.-V. declare no conflicts of interest.

Funding Sources

The authors declare that no financial support was received for the research, authorship, and/or publication of this article.

Author Contributions

F.F.-A. and A.C.: conceptualization, methodology, formal analysis, investigation, data curation and original draft preparation, and writing – review and editing. M.C., A.P., A.J., E.V., P.C., V.F., C.O.-V., and E.F.: methodology, investigation, and results review. M.C., E.F., and A.C.: supervision. All authors have read and agreed to the published version of the manuscript.

Funding Statement

The authors declare that no financial support was received for the research, authorship, and/or publication of this article.

Data Availability Statement

The data that support the findings of this study are not publicly available due to privacy reasons but are available from the corresponding author upon request.