Effects of Incretin‐Based Therapies, Diet and Exercise Interventions, and Bariatric Surgery on Fat‐Free Mass in Adults With Overweight or Obesity: A Systematic Review and Meta‐Analysis

ABSTRACT

Background

Maintaining fat‐free mass (FFM) during weight loss is important to maintain a healthy body composition. This review examined changes in FFM during weight loss induced by diet and exercise interventions, incretin‐based therapies, and bariatric surgery, and assessed whether the relative loss of FFM differs by weight‐loss modality.

Methods

A systematic search of PubMed, Cochrane Library, and Embase was conducted for randomised controlled trials published from 2015 to October 21st, 2025. Eligible studies included adults with overweight or obesity (with or without type 2 diabetes), achieving ≥ 10% weight loss of total body weight through diet and exercise, incretin‐based, or surgical interventions. The studies were required to report measurements of either FFM using dual‐energy X‐ray absorptiometry or muscle mass using computed tomography or magnetic resonance imaging.

Results

Twelve diet and exercise, five incretin‐based therapy, and four bariatric surgical studies (1334 participants) were included. The pooled changes in FFM were −1.8 kg (95% CI: −2.6, −1.0) for diet and exercise interventions, −4.8 kg (95% CI: −5.6, −3.9) for incretin‐based therapies versus placebo, and −9.1 kg (95% CI: −12.3, −6.0) for bariatric surgery. The proportion of total weight loss attributable to FFM was 14.9% for diet with/without exercise interventions (22.3% for diet without exercise and 7.7% for diet with exercise), 33.3% for incretin‐based therapies, and 34.2% for surgical interventions.

Conclusion

Among the included studies with ≥ 10% weight loss in individuals with overweight or obesity, diet and exercise interventions were associated with the smallest reductions in FFM, whereas incretin‐based therapies and bariatric surgery showed substantially greater losses. Given the importance of FFM, strategies to preserve FFM, particularly exercise, should be included in all weight‐loss approaches.

1. Introduction

Obesity is an escalating public health issue, with around one in eight individuals affected worldwide in 2022 [1]. Overweight and obesity are associated with numerous health complications, including cardiovascular diseases [2], type 2 diabetes [3], cancers [4], and digestive diseases [5]. Consequently, the demand for effective treatments of obesity has increased.

Traditionally, weight reduction relied on behavioural and diet and exercise interventions [6], and for eligible patients, bariatric surgery [7]. However, diet and exercise interventions require sustained adherence and substantial behavioural changes, while bariatric surgery remains highly invasive and irreversible. Recent advances in incretin‐based pharmacotherapies have changed obesity management [6]. These treatments can induce weight losses approaching those achieved through bariatric surgery [8], yet with considerably lower invasiveness.

Despite their efficacy, concerns have emerged regarding the potential loss of fat‐free mass (FFM) associated with these pharmacological interventions [8, 9, 10]. FFM, at the molecular level, comprises all nonfat body components [11]. Preserving FFM during weight loss is critical, as skeletal muscle and organ tissues exhibit a higher metabolic rate than fat mass [8], supporting metabolism and enabling sustained weight loss. Loss of FFM, particularly muscle and bone, can impair physical function and metabolic health [6], and muscle loss increases the risk of frailty [12], insulin resistance, and weight regain due to reduced metabolic rate, especially in elderly people [13], while bone loss can elevate the risk of osteoporosis and fractures, particularly in postmenopausal women [14].

The extent of FFM reduction with incretin‐based therapies raises questions regarding long‐term metabolic implications. Several reviews and meta‐analyses have examined FFM loss during incretin‐based therapies [8, 15, 16, 17, 18]. However, no studies have investigated differences in FFM loss with incretin‐based therapies, diet and exercise interventions, or bariatric surgery.

Therefore, this systematic review and meta‐analysis evaluate whether FFM reductions observed with incretin‐based therapies are proportional to those seen with bariatric surgery or diet and exercise interventions, when undergoing weight loss of at least 10% of baseline body weight. Although more than 5% weight loss is typically considered clinically meaningful [19], newer incretin‐based agents achieve larger reductions [20, 21], making it particularly relevant to examine how larger weight losses influence FFM. Understanding whether FFM loss differs across interventions is crucial for evaluating the safety and metabolic sustainability of emerging obesity treatments.

2. Methods

The present systematic review and meta‐analysis followed the Cochrane Handbook for Systematic Reviews of Interventions and adhered to the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines (Supporting Information) [22, 23].

2.1. Data Sources and Search Strategies

The search was conducted in PubMed, Cochrane Library, and Embase, covering the period 2015 to October 21^st 2025. Search strings were built around predefined interventions, outcomes, measurement methods, and study design, using Boolean operators (AND, OR). Terms covered diet and exercise interventions, incretin‐based pharmacotherapies, and bariatric surgery, as well as terms for FFM, imaging methods, magnetic resonance imaging (MRI), computed tomography (CT), and dual‐energy X‐ray absorptiometry (DXA), and the study design randomised controlled trials (RCTs). Searches were limited to terms appearing in titles or abstracts. Search strings for each database are provided in Supporting Information. Additional manual searches were conducted in PubMed.

2.2. Study Selection

Study screening and selection were conducted using Covidence [24]. Initially, all titles and abstracts identified through the search were screened, and studies meeting the inclusion criteria were retrieved for full‐text assessment. The inclusion criteria for this systematic review were, (i) RCTs or substudies to RCTs (ii) adult participants (≥ 18 years) with overweight or obesity, with or without type 2 diabetes, (iii) one of the following interventions: treatment with incretin‐based pharmacotherapies, diet and exercise interventions, or bariatric surgery, (iv) assessment of muscle mass or FFM by MRI, CT or DXA, and (v) achievement of ≥ 10% total body weight loss in all three intervention modalities to ensure comparability. Studies including individuals with type 1 diabetes were excluded. RCTs were chosen as an inclusion criterion, as this study design represents the highest level of evidence and randomisation minimises confounding and selection bias [25]. Restricting inclusion to a single study design also improves comparability across studies by ensuring a consistent methodological standard. RCTs comparing different active treatments without a control group were also included, as control groups are often not feasible in surgical interventions due to ethical and practical constraints. Two reviewers (L.B.C. and I.Ø.C.) independently screened records and selected studies according to the predefined criteria. Any discrepancies were resolved through discussion or by consulting a third reviewer (L.B.M.).

2.3. Data Extraction and Outcomes

This review compared three weight‐loss approaches in terms of changes in FFM and total body weight, and examined their relationship. Primary endpoints were changes in FFM or muscle mass and total body weight (kg). Secondary endpoints were change in FM (kg) and percentage weight change. Muscle strength was an exploratory endpoint but was not reported, as it was assessed only in a few diet and exercise studies.

Data extracted from the 21 articles included study and participant characteristics (Tables S1–S3). For diet and exercise and surgical studies, mean ± standard deviation (SD) was collected for each arm when available. For pharmacological studies where SDs were missing, estimated treatment differences (ETDs) and 95% confidence interval (95% CI) were extracted.

In 16 of the included studies, the term “lean mass” was used and not clearly defined, making it unclear which tissue components were included. More broadly, the term “lean mass” lacks a consistent definition regarding which body components are included [11]. The terminology used (e.g., lean mass, lean body mass, lean soft tissue mass) is specified for each study in Table 1. For comparability, these measures were treated as equivalent, as they represent closely related measures and any differences are likely to be minor; however, variation in definitions and terminology may still lead to differences in the components assessed across studies.

TABLE 1.

Body weight and body composition outcomes in the included studies.

Study	Study arm	n	Weight change (kg)	Weight change (%)	FFM change (kg)	FM change (kg)
(A) Diet and exercise studies
Anyiam 2024	VLDC	10	−13.1 (6.0)	−11.3 (5.7)	−3.8 (2.5) a	−9.0 (4.1)
Ard 2019	OPTIFAST, week 26	135	−13.2 (0.6) b	−12.4 (0.6)	−2.3 (2.8) c	−11.3 (7.5)
Beavers 2025	WL + RT, 12 months	47	−10.4 (6.4)	−11.3 (7.1) d	−0.9 (1.9) c	−10.0 (5.7)
Benito 2015	E‐group	26	−10.6 (0.8)	−11.3 (0.9) d	0.4 (1.3) c	−9.0c
Benito 2015	SE‐group	24	−9.9 (0.9)	−10.1 (0.9) d	−0.1 (1.7) c	−8.6 e
Benito 2020	SE‐group	22	−8.5 (2.8)	−10.6 (3.5) d	0.9 (1.4) c	−8.0 e
Brennan 2021	WLEX group	20	−10.6 (4.9)	−11.0 (5.0)	−0.9 (1.9) c	−8.6 (4.3)
Haywood 2017	Ex/VLCD	41	−10.4 (5.4) b	−10.0 (5.2)	−2.2 (2.2) b , c	−6.8 (4.8) b
Magkos 2021	P52	45	−13.3 (3.4)	−13.2 (2.7)	−3.6 (1.5) b , c	−9.7 (2.7)
Magkos 2021	P77	39	−11.5 (3.1)	−11.6 (2.5)	−3.1 (1.4) b , c	−8.4 (2.5)
Marin‐Alejandre 2021	AHA group, 6 months	37	−10.2 e	−10.8 e	−1.2 c , e	−9.1 e
Normandin 2018	Diet	15	−11.2 (4.4)	−11.9 (4.7) d	−2.3 (1.9) c	−7.9 (3.2)
Normandin 2018	Diet + vest	18	−11.0 (6.3)	−10.9 (6.3) d	−2.9 (1.6) c	−7.6 (4.5)
Seimon 2019	Severe group, 6 months	50	−17.8 (6.7)	−19.9 (7.4)	−3.0 (3.3) c	−12.2 (5.6)
Stentz 2016	HC diet	12	NA	−11.3 (6.2)	NA	NA
(B) Pharmacological studies
Coskun 2025	8 mg retatrutide	28	−16.8 (10.1)	−17.1 (9.5)	−6.5 (3.7) c	−10.9 (5.3)
Coskun 2025	8 mg retatrutide vs. placebo	28/25	−14.7 (7.0)	−14.7 (6.9)	−5.9 (3.0) c	−9.2 (4.0)
Coskun 2025	12 mg retatrutide	24	−17.5 (13.7)	−16.0 (8.8)	−6.4 (4.4) c	−10.5 (8.8)
Coskun 2025	12 mg retatrutide vs. placebo	24/25	−15.4 (9.7)	−13.6 (6.5)	−5.8 (3.5) c	−8.8 (6.6)
Garvey 2025	CagriSema 2.4 mg of each	154	−24.2	−21.8	−7.0 f	−14.1
Garvey 2025	CagriSema, 2.4 mg of each vs. placebo	154/55	−19.7 (10.4)	−17.9 (9.1)	−4.9 (4.5) f	−11.7 (8.0)
Garvey 2025	Semaglutide 2.4 mg	21	−18.7	−17.0	−4.6 f	−12.4
Look 2025	Pooled tirzepatide (5/10/15 mg)	124	−21.6 b	−21.3	−5.6 (0.3) c	−15.9
Look 2025	Pooled tirzepatide vs. placebo	124/36	−16.2 (9.3) b	−16.2 (9.3)	−4.4 (3.2) c	−12.3 (7.4)
Wilding 2021	Semaglutide 2.4 mg	95	−15.3 g	−14.9 g	−5.3 a	−8.4
Wilding 2021	Semaglutide 2.4 mg vs. placebo	95/45	−12.7 (10.7) g	−12.4 (9.9) g	−3.4 (3.7) a	−7.0 (7.9)
(C) Surgical studies
Hofsø 2020	Roux‐en‐Y gastric bypass	44	−34.5 (9.2)	−29.1 (7.8) d	−10.2 (3.3) c	−23.0 (6.7)
Hofsø 2020	Sleeve gastrectomy	48	−28.7 (12.9)	−23.6 (10.6) d	−9.8 (3.6) c	−18.2 (9.3)
Maghrabi 2015	Roux‐en‐Y gastric bypass, 24 months	18	−26.5 (10.6) b	−25.2 (10.1)	−6.8 (2.5) b , c	NA
Maghrabi 2015	Sleeve gastrectomy, 24 months	19	−22.5 (7.4) b	−22.5 (7.4)	−6.5 (2.2) b , c	NA
Schneider 2016	Laparoscopic sleeve gastrectomy	23	−32.1 (12.7)	−26.7 (10.6) d	−10.5 (3.7) c , h	−17.7
Schneider 2016	Laparoscopic Roux‐en‐Y gastric bypass	19	−39.0 (9.7)	−31.0 (7.7) d	−17.1 (4.2) c , h	−20.9
Varma 2019	Roux‐en‐Y gastric bypass	4	−9.5 (0.2)	−10.0 (0.2)	−3.4 (1.0) b , c	−5.8 (1.1) b

Abbreviations: AHA = American heart association dietary intervention, CagriSema = cagrilintide + semaglutide, E‐group = endurance training, Ex = exercise, NA = not available, P52 = VLCD with 52 g protein, P77 = VLCD with 77 g protein, SE‐group = strength and endurance training, Severe group = severe energy restriction, VLCD = very low‐calorie diet, vs. = versus, WL + RT = weight loss + resistance training, WLEX = weight loss + exercise.

^a Value was measured and reported as “lean body mass” in the study, which per definition includes all non‐fat molecules in the body, meaning including bone mass [11].

^b Mean and SD estimated from mean percentage change and SD and mean baseline weight (kg).

^c Value was measured and reported as “lean mass” in the study without further specification of the tissues included in the measurement.

^d Mean and SD estimated from mean weight change (kg) and SD and mean baseline weight (kg).

^e Mean estimated by calculating the difference between baseline weight (kg) and final weight (kg).

^f Value was measured and reported as “lean soft‐tissue mass” in the study, which per definition includes all non‐fat, non‐bone mineral molecules in the body, meaning excluding bone mass [11].

^g Values from the main study population and not the DXA‐substudy population.

^h SD values are imputed from two similar trials [26, 27] corresponding to Cochrane Handbook section 6.5.2.7 [23].

When results were presented as means with CIs, SEMs, or ranges, SDs were estimated using standard formulas. For medians with ranges or interquartile range (IQR), means and SDs were similarly approximated. In pharmacological trials reporting only 95% CIs for ETDs, SDs were derived by converting CIs to standard error (SE) and then to SD following Cochrane guidelines (Section 6.5.2.3) [23]. All equations, including calculations between percentages and kilograms, are described in Supporting Information.

When baseline DXA‐subgroup data were missing, information from the main trial was used and marked accordingly. Additionally, authors from included studies were contacted to request missing data.

2.4. Quality Assessment

Quality assessment was performed using the Cochrane Risk of Bias tool (RoB2) [28]. Two reviewers (L.B.C. and I.Ø.C) independently assessed each study according to the RoB2 framework, and discrepancies were resolved through discussion. Studies were categorised as intention‐to‐treat (ITT) or per‐protocol (PP). Each PP study used its own definition of what qualified a participant as per protocol, which this review adhered to when assessing bias. RoB2 evaluates five domains of potential bias: (1) the randomisation process, (2) deviations from the intended interventions, (3) missing outcome data, (4) measurement of the outcome, and (5) selection of the reported result. Each domain was assessed as indicating “low,” “some concerns,” or “high” risk of bias in accordance with Cochranes guidance [23]. Trials were classified as overall low risk of bias only if all five domains were rated low.

2.5. Statistical Analysis

Statistical analyses were conducted using R version 4.4.2 using the packages metafor, dplyr, and knitr [29]. Within‐group changes in FFM or muscle mass, FM and total weight from baseline to post weight loss were extracted, as most diet and exercise and surgical studies lacked control groups and instead compared different interventions. Pharmacological studies were analysed relative to placebo by calculating the comparison effective sample size: $n_{\mathit{eff}}={\left(\frac{1}{n_a}+\frac{1}{n_p}\right)}^{-1},$with n _a as the active treatment sample size, and n _p as the placebo sample size. As muscle mass was only reported in one study, a comparison was not feasible. Pooled mean changes in FFM and corresponding 95% CI were estimated using a random‐effects model, chosen because true intervention effects were expected to vary across studies due to methodological and clinical heterogeneity. Statistical heterogeneity was quantified using the $I^2$ statistic and interpreted per the Cochrane Handbook (Section 10.10.2) [23]. The significance level was set at 0.05. Fat‐free mass ratio (FFM‐ratio) was calculated for each study arm as $\frac{∆\mathit{FFM}}{\left(∆\mathit{FFM}+∆\mathit{FM}\right)}$, and visualised in boxplots (with medians, means, and IQR) using ggplot2, dplyr, tidyr, and ggdist in R [29]. For the two intervention arms in Maghrabi 2015, the FFM‐ratio was calculated using ∆Weight rather than $\left(∆\mathit{FFM}+∆\mathit{FM}\right)$, as FM data were not reported [30].

3. Results

3.1. Study Selection

Figure 1 illustrates the study selection process. A total of 159 articles were assessed in full text, of which 141 were excluded as conference abstracts, trial registrations, or otherwise not meeting the inclusion criteria. Three additional eligible studies were identified through manual search. Ultimately, 21 studies were included.

FIGURE 1.

Preferred reporting items for systematic reviews and meta‐analyses (PRISMA), flowchart of the study selection process [22].

3.2. Study Characteristics

The characteristics of the studies are presented in Tables S1 and S2. Twelve diet and exercise studies, five pharmacological studies, and four surgical studies were included.

The twelve diet and exercise studies differed substantially in intervention content and duration. Most had between two and four parallel intervention arms and no control group. Seven studies included diet‐only arms (Anyiam 2024, Ard 2019, Magkos 2021, Marin‐Alejandre 2021, Normandin 2018, Seimon 2019, Stentz 2016) [31, 32, 33, 34, 35, 36, 37]. Five studies had study arms included that examined combined diet and exercise interventions (Beavers 2025, Benito 2015, Benito 2020, Brennan 2021, Haywood 2017) [38, 39, 40, 41, 42]. Normandin 2018 had both a study arm with diet‐only and one with both diet and weighted vest included [35]. Exercise interventions included resistance training (Beavers 2025) [38], endurance (E‐group), and combined strength‐endurance (SE‐group) programs (Benito 2015 and 2020) [39, 40], aerobic and endurance exercises (Haywood 2017) [42], and combined aerobic and resistance training (Brennan 2021) [41]. Concomitant components included dietary counselling, diet diaries, written guidance, and behavioural sessions. Study duration ranged from eight weeks (Magkos 2021) [33] to two years (Marin‐Alejandre 2021) [34]. Blinding was rarely feasible, with Beavers 2025 being the only study with blinding [38].

The five included pharmacological studies had more in common in terms of design (see Tables S1 and S2). Three were substudies of larger RCTs (Look 2025, Coskun 2025, Sattar 2025) [21, 43, 44]. As for Wilding 2021 and Garvey 2025, the main studies were included, and data from the DXA‐subgroups were extracted [20, 45]. Garvey 2025 (REDEFINE‐1), Look 2025 (SURMOUNT‐1), and Wilding 2021 (STEP‐1) were double‐blind, placebo‐controlled phase 3 trials [20, 21, 45]. Coskun 2025 was a phase 2 double‐blind, placebo‐controlled randomised trial [43], and Sattar 2025 was a post hoc, exploratory analysis of an open‐label phase 3 trial (SURPASS‐3) [44]. The studies investigated tirzepatide (glucagon‐like peptide‐1 (GLP‐1) and glucose‐dependent insulinotropic polypeptide (GIP) receptor agonist) [21, 44], retatrutide (GLP‐1, GIP, and glucagon (GCG) receptor agonist) [43], semaglutide (GLP‐1 receptor agonist) [20], or cagrilintide (amylin analogue) combined with semaglutide [45]. Concomitant components included calorie restrictions, dietary counselling, and behavioural counselling. Treatment duration ranged from 36 weeks (Coskun 2025) to 72 weeks (Look 2025) [20, 21, 43, 44, 45].

Surgical interventions included four Roux‐en‐Y gastric bypass (RYGB) and three sleeve gastrectomy (SG) arms [30, 46, 47]. Designs varied, but all procedures were performed laparoscopically (see Tables S1 and S2). Hofsø 2020, a randomised triple‐blind trial, compared SG and RYGB [26], Maghrabi 2015 a randomised controlled substudy comparing RYGB, SG, and intensive pharmacological therapy [30], Schneider 2016 a prospective randomised trial comparing laparoscopic SG and RYGB [46], and Varma 2019, a small randomised pilot trial comparing pharmacological therapy, adjustable gastric banding, and RYGB [47]. Hofsø 2020 was the only surgical trial incorporating blinding [26]. Concomitant peri‐ or postoperative components included micronutrient supplementation (Hofsø 2020, Maghrabi 2015) [26, 30], medical prophylaxis (Schneider 2016) [46], and dietitian follow‐up (Varma 2019) [47]. Duration ranged from approximately 9 (Varma 2019) to 24 months (Maghrabi 2015) [30, 47].

3.3. Participants

Baseline characteristics of the participants are presented in Table S3. Among the twelve diet and exercise studies, fifteen study arms with a total of 577 participants were included. Mean age ranged from 37.4 to 70.3 years, baseline body weight from 80.2 to 121.6 kg, FFM from 38.2 to 65.6 kg, FM from 29.2 to 52.3 kg, and BMI from 27.7 to 40.0 kg/m².

In the five pharmacological studies, seven active treatment arms and four placebo groups were included, with a total of 582 participants. Mean age ranged from 45.5 to 58.2 years, baseline body weight from 91.3 to 107.2 kg, FFM from 50.4 to 60.7 kg, FM from 38.3 to 49.7 kg, and BMI from 33.7 to 39.7 kg/m². Sattar 2025 reported muscle volume rather than FFM or FM due to MRI methodology [44].

The four bariatric surgery studies had seven study arms included, with a total of 175 participants. Mean age ranged from 40.3 to 54.3 years, baseline body weight from 94.9 to 125.8 kg, FFM from 46.9 to 67.9 kg, FM from 43.3 to 56.4 kg, and BMI from 35.0 to 44.4 kg/m².

Overall, 1334 participants across 33 study arms (including placebo) were included. Participants were primarily female, with mean ages ranging from 37.4 to 70.3 years, body weight from 80 to 125.8 kg, FFM from 38.2 to 67.9 kg, FM from 29.2 to 56.4 kg, and BMI from 27.7 to 44.4 kg/m².

3.4. Quality Assessment

Figure 2 presents the risk of bias for per‐protocol (PP) studies, and Figure 3 for intention‐to‐treat (ITT) studies. ITT data were used when available; otherwise, PP results were included in the meta‐analysis, and the risk‐of‐bias assessment was performed accordingly. Both analyses were conducted using the RoB2 tool and included ten ITT and eleven PP studies [28]. Across both groups, most studies demonstrated a low risk of bias in D1 and D4. High risk of bias was mainly seen in D3 and D5 for both groups. Bias in D3 largely stemmed from incomplete follow‐up and missing outcome data. In D5, bias mainly arose from missing protocols or prespecified statistical analyses, increasing selective reporting risk. Several ITT trials also showed bias in deviations from intended interventions (D2), often due to unbalanced deviations between groups.

FIGURE 2.

Risk of bias assessment of the per‐protocol studies. Among PP studies, three showed high risk of bias (Look 2025, Stentz 2016, Varma 2019), while the remaining demonstrated low risk.

FIGURE 3.

Risk of bias assessment of the intention‐to‐treat studies. Most ITT studies demonstrated low risk of bias, with isolated exceptions (Sattar 2025, Marin‐Alejandre 2021).

Overall, these findings suggest a moderate‐quality evidence base. Most studies followed appropriate methodological standards, though some remained vulnerable to bias from missing data and selective reporting.

3.5. Results of Included Studies

All studies included achieved ≥ 10% weight loss. Weight changes (kg and %) and changes in FFM and FM from baseline to follow‐up for each study arm are presented in Table 1. For pharmacological studies, values are reported both for active treatment arms alone and as active and placebo comparisons (Table 1B).

Due to incomplete variance reporting (SD, SEM, and 95% CI) in several pharmacological studies, a direct comparison of all three intervention classes was not feasible. In the pharmacological trials, variance was typically available for active and placebo comparisons, whereas diet and exercise and surgical studies did often not have a placebo or control arm. Consequently, a direct statistical comparison across interventions was not performed, as the underlying analytic structures differed. Instead, a descriptive quantitative synthesis of the FFM‐ratio was conducted, visualised using boxplots.

3.6. Change in Fat‐Free Mass in Diet and Exercise Studies

The heterogeneity was 95.7%, meaning considerable heterogeneity [23]. Marin‐Alejandre 2021 and Stentz 2016 were not included in this analysis, as SDs were missing [34, 37]. Based on the random‐effects model, the pooled mean reduction in FFM across the ten diet and exercise study arms was −1.80 kg (95% CI: −2.64, −0.96). Results are presented in Figure 4A.

FIGURE 4.

(A) Heterogeneity: ${\mathit{Tau}}^2$= 2.254; ${\mathit{Chi}}^2$= 317.92; df = 12 (p < 0.001); $I^2$= 95.7%. Test for overall effect: Z = −4.18 (p < 0.001). (B) Heterogeneity: ${\mathit{Tau}}^2$= 0.443; ${\mathit{Chi}}^2$= 7.19; df = 4 (p = 0.126); $I^2$ = 44.8%. Test for overall effect: Z = −10.61 (p < 0.001). $n_{\mathit{eff}}$was calculated using the equation $n_{\mathit{eff}}={\left(\frac{1}{n_a}+\frac{1}{n_p}\right)}^{-1}$, where n _a  = number of participants in the treatment group, and n _p  = number of participants in the placebo group. (C) Heterogeneity: ${\mathit{Tau}}^2$= 17.832; ${\mathit{Chi}}^2$= 228.72; df = 6 (p < 0.001); $I^2$= 98.1%. Test for overall effect: Z = −5.66 (p < 0.001). Panel (A): Forest plots showing the pooled effects of the diet and exercise study arms on fat‐free mass change in kilograms, estimated using a random‐effects model. Data are presented as mean difference and 95% CI, along with mean, SD, and number of participants (n) for each study arm, and weight (%) contributions of each study to the aggregated FFM change. The analysis was conducted in R version 4.4.2. Panel (B): Forest plots showing the pooled effects of the pharmacological study arms compared with their respective placebo arms on fat‐free mass change in kilograms, estimated using a random‐effects model. Data are presented as mean difference and 95% CI, along with mean, SD, and the effective sample size for each two‐group comparison (n_eff), and weight (%) contributions of each study to the aggregated FFM change. The analysis was conducted in R version 4.4.2. Panel (C): Forest plots showing the pooled effects of the diet and exercise study arms on fat‐free mass change in kilograms, estimated using a random‐effects model. Data are presented as mean difference and 95% CI, along with mean, SD, and number of participants (n) for each study arm, and weight (%) contributions of each study to the aggregated FFM change. The analysis was conducted in R version 4.4.2. Abbreviations: 12 M = 12 months, 6 M = 6 months, CagriSema = cagrilintide + semaglutide, Ex = exercise, LRYGB = laparoscopic Roux‐en‐Y gastric bypass, LSG = laparoscopic sleeve gastrectomy, P52 = VLCD with 52 g protein, P77 = VLCD with 77 g protein, Reta = retatrutide, RYGB = Roux‐en‐Y gastric bypass, Sema = semaglutide, Severe = severe energy restriction, SG = sleeve gastrectomy, TZP = tirzepatide, VLCD = very low‐calorie diet, W26 = week 26, WL + RT = weight loss + resistance training, WLEX = weight loss + exercise.

3.7. Change in Fat‐Free Mass in Pharmacological Studies Versus Placebo

Moderate statistical heterogeneity was observed across the pharmacological studies (I ²  = 44.8%) [23]. Sattar 2025 assessed muscle volume in litres, which could not directly be compared with loss of FFM in kilograms, and was therefore not included in the analysis [44]. Based on the random‐effects model, the pooled mean difference in FFM loss between active treatment and placebo across the five included study arms was −4.75 kg (95% CI: −5.62, −3.87). Results are shown in Figure 4B.

3.8. Change in Fat‐Free Mass in Surgical Studies

Considerable heterogeneity was observed among the surgical intervention studies (I ²  = 98.1%) [23]. Based on the random‐effects model, the pooled mean reduction in FFM across the seven surgical study arms was −9.14 kg (95% CI: −12.31, −5.98). Results are shown in Figure 4C.

3.9. Fat‐Free Mass Ratio Across Interventions

FFM‐ratio was calculated for all study arms, except two. Sattar 2025 reported outcomes in MRI‐derived muscle volume rather than FFM; Stentz 2016 did not report changes in FM and FFM [37, 44]. Therefore, these studies were not included in the analysis. For the remaining studies, FFM‐ratio values were grouped into the three intervention categories. Boxplots illustrating the distribution of FFM‐ratio values are shown in Figure 5A, with subgroups, IQR, minimum and maximum values, median, mean, and SD presented. Mean FFM‐ratio was 0.149 (SD = 0.132) for diet and exercise, 0.333 (SD = 0.056) for pharmacological and 0.342 (SD = 0.064) for surgical interventions. The mean FFM‐ratio for diet without exercise interventions was 0.223 (SD = 0.064) and for diet with exercise interventions the FFM‐ratio was 0.077 (SD = 0.147). Individual study‐level FFM‐ratio values across intervention categories are displayed in Figure 5B, illustrating notable between study variability. The endurance‐group (E‐group) in Benito 2015 and strength and endurance‐group (SE‐group) in Benito 2020 had an increase in FFM rather than loss, which is why these study arms have negative FFM‐ratio values in Figure 5 [39, 40].

FIGURE 5.

(A) Fat‐free mass ratio (FFM‐ratio) categorised into the three intervention classes: Diet and exercise, pharmacological, and surgical. They are illustrated using boxplots, showing the median, interquartile range, minimum, maximum, mean (red points) and individual study arm (depicted as points). Diet and exercise studies incorporating exercise as part of the intervention are indicated by green rather than black points. Data are presented as number of subgroups, minimum (min), first quartile (Q1), median, third quartile (Q3), maximum (max), mean and SD. (B) Fat‐free mass ratio (FFM‐ratio) across the three intervention categories: Diet and exercise, pharmacological, and surgical. Each study arm and its corresponding sample size (n) are shown on the left. Studies are ranked from top to bottom according to FFM‐ratio, with higher ratios (indicating greater proportional FFM loss relative to total weight loss) positioned at the top. The analysis was conducted in R version 4.4.2.

4. Discussion

To our knowledge, this is the first systematic review investigating differences in FFM changes during ≥ 10% weight reduction with either diet and exercise interventions, incretin‐based pharmacotherapies, or bariatric surgery. Diet and exercise interventions showed the smallest FFM reductions, whereas pharmacological and surgical interventions demonstrated larger losses. Pooled mean FFM changes were −1.80 kg for diet and exercise interventions, −4.75 kg for incretin‐based therapies versus placebo, and −9.14 kg for bariatric surgery. FFM loss accounted for 14.9% (22.3% without exercise and 7.7% with exercise), 33.3%, and 34.2% of total weight loss, respectively, illustrating that the larger FFM loss with incretin‐based therapy or bariatric surgery is not solely explained by larger weight losses. Under standardized conditions, a FFM loss of ≤ 25% is recommended during weight loss [8, 11, 15]. Both pharmacological and surgical interventions exceeded this threshold. The high amount of FFM loss observed in bariatric surgery may partly be explained by the differences in baseline body composition, as participants undergoing surgery generally had higher baseline BMI, body weight, and consequently greater baseline FFM to support the higher body weight. The greater reductions in FFM observed after surgery may therefore reflect both the greater total weight loss and the higher baseline FFM in these participants. In this context, the similar proportion of FFM loss relative to total weight loss observed for pharmacological and surgical interventions may suggest that the composition of weight loss follows broadly comparable patterns across these modalities despite differences in absolute weight reduction. While these findings suggest similar patterns in the composition of weight loss, the clinical relevance of FFM loss, and the importance of its preservation, depends on what FFM represents.

Preservation of FFM is important, but FFM must be distinguished from skeletal muscle mass, as it is roughly twice the quantity of true muscle mass [11]. Therefore, FFM loss should not be assumed to fully represent muscle loss [11], which CT or MRI better quantify [44]. One MRI substudy (Sattar 2025) was included. According to the study's authors, reductions in muscle volume appeared proportional to weight loss [44]. Although not equivalent to muscle mass, FFM encompasses metabolically active tissues essential for muscle function and metabolic health, and a higher FFM percentage with lower FM percentage is associated with better metabolic outcomes and lower mortality, making FFM preservation a meaningful target [48, 49].

Muscle strength is also clinically relevant but could not be compared in this review because it was inconsistently measured. Strength predicts mortality more strongly than muscle mass and depends on muscle quality and muscle size [6].

Earlier reviews reported lower FFM losses than was found in our analyses: Karakasis et al. (2024) found FFM accounted for $~25\%$ of weight loss [16], and Jiao et al. (2024) found a −1.02 kg FFM change [18], far lower than the −4.75 kg observed in our analysis, but both examined far smaller weight losses and less potent agents such as liraglutide and exenatide. The larger FFM changes observed here likely reflect greater weight reductions and inclusion of newer incretin‐based therapies. These findings raise the question of whether this higher FFM loss is clinically concerning.

For incretin‐based pharmacotherapies, evidence shows that although FFM decreases, they may enhance skeletal muscle function [16]. In Sattar (2025), tirzepatide reduced muscle fat infiltration and muscle volume [44]. Lower infiltration indicates better muscle quality, which is more strongly associated with favourable outcomes than muscle quantity [50, 51, 52], and both muscle quality and myosteatosis predict mortality [53]. Strength correlates closely with MRI‐based muscle measurements, with low muscle volume and increased fat infiltration predicting lower strength [54, 55]. GLP‐1RAs also appear to improve strength per BMI unit and reduce muscle fat infiltration [56], indicating that FFM declines should be interpreted alongside potential improvements in muscle quantity and quality [8].

Regarding bariatric surgery, although FFM loss occurs, Bolte et al. report improvements in physical ability and quality of life, including increased relative muscle strength [6]. Because strength predicts mortality [6], these functional gains help contextualise FFM reductions, as weight loss still improves functional capacity and insulin sensitivity in individuals with overweight and obesity [6].

Finally, the one‐quarter FFM rule has been criticised by a review [57]. They concluded that the rule is at best an approximation as it is influenced by factors such as diet composition, sex, baseline adiposity, and physical activity level [57].

The question then becomes how to lose weight while preserving FFM and how loss of FFM influences weight loss maintenance. Incretin‐based pharmacotherapy and bariatric surgery achieve large weight losses and are often accompanied by substantial improvements in glycated hemoglobin, blood pressure, and lipid profiles [58, 59]. However, a rapid and pronounced energy deficit can activate adaptive mechanisms whereby the body utilizes not only fat but also muscle as a fuel source [60]. This can result in a greater proportion of FFM loss compared to exercise and diet alone, which may increase the risk of sarcopenia and reduced physical capacity [57] and an increased risk of weight regain over time [61].

Combined interventions with exercise may be more optimal in preserving FFM. Lundgren et al. (2021) found that combining liraglutide (GLP‐1RA) with exercise in weight‐loss maintenance following a low‐caloric diet, did not reduce FFM, even when weight loss continued [62]. Jensen et al. (2024) investigated the post‐weight‐loss maintenance phase in the same population and found that pharmacotherapy alone led to substantial weight regain after treatment termination, resulting in a higher body‐fat percentage than before liraglutide treatment [61]. In contrast, the exercise‐only and combined groups regained less weight and had lower body‐fat percentage one year after treatment termination [61]. Thus, exercise appears essential to FFM preservation. Similarly, another review found that weight regain and reversal of cardiometabolic health markers occurred faster after termination of pharmacological treatment than behavioural weight management programs [63], highlighting the importance of the integration of exercise. Both endurance and resistance training preserve muscle mass [64], and resistance training further improves strength and body composition independent of weight loss [6, 8, 11, 64]. This aligns with the findings of the present review, where interventions combining diet with exercise showed the lowest proportion of FFM lost (7.7% of total weight loss) and in some of the included studies, interventions combining diet and exercise (e.g., Benito 2015, Benito 2020) showed an increase in FFM rather than a loss during weight loss.

Protein intake represents another important factor worth considering. Higher protein intake during weight loss may preserve FFM more effectively than lower protein diets [8]. Calorie restriction reduces net protein utilization, thereby increasing the amount of dietary protein required to maintain protein balance and preserve FFM [65]. Consuming more than the recommended dietary allowance for protein (0.8 g/kg) has been shown to reduce the loss of FFM during calorie‐restricted and surgical weight loss, and similar effects may occur during incretin‐based weight loss [66, 67, 68]. Protein intake may therefore be an important factor in preserving FFM across different weight loss interventions [69]. Combining resistance training with protein supplementation could be an ideal approach to preserving FFM [8].

Bone health is another consideration. Bariatric surgery reduces bone mineral density, whereas incretin‐based therapies do not appear to adversely affect BMD, according to meta‐analytic evidence [70]. However, at clinically relevant bone specific sites, such as hip and spine, liraglutide reduced BMD, which was not observed when combined with exercise [71].

Study duration varied both across intervention types and within each intervention category. Bariatric surgery differs fundamentally from diet and exercise and pharmacological interventions, as it is a single procedure followed by a dynamic weight trajectory. Most weight loss after bariatric surgery occurs within the first year postoperatively, after which weight regain may occur, particularly in the absence of sustained diet and exercise changes [72]. In contrast, diet and exercise and pharmacological interventions require continuous treatment throughout the trial period. These differences in intervention dynamics and follow‐up duration may influence both the magnitude and composition of weight loss observed across studies. Furthermore, weight regain is a relevant consideration across all intervention types, as weight regain most often occurs when the treatment stops or is less intensive.

The key strength in this review is that it provides a clinically interpretable comparison of FFM changes across diet and exercise, pharmacological, and bariatric surgery interventions. These modalities are typically evaluated separately, and comparisons across all three approaches are lacking. By applying a clinically meaningful ≥ 10% weight‐loss threshold, the review evaluates FFM changes under comparable conditions of substantial weight reduction, enabling interpretation of the composition of weight loss rather than weight change alone.

Additional strengths include the use of objective body composition assessments (DXA and CT/MRI) and a consistent randomised‐trial evidence standard across modalities, which strengthens internal validity and comparability. Methodological transparency is further enhanced by explicitly reporting how heterogeneous “lean mass” terminology was handled in synthesis. Together, these features strengthen reproducibility and improve the clinical applicability of the findings when counseling patients about expected changes in body composition during substantial weight loss.

This review has several limitations. Considerable heterogeneity was present across interventions, particularly among surgical and diet and exercise studies, limiting comparability. At the same time, this heterogeneity reflects important differences in study designs, populations, and outcome measures, and highlights a need for more standardised approaches in future research. Pharmacological therapy data originated from substudies, often with smaller sample sizes, which may reduce statistical power. Surgical studies also included relatively few participants. These smaller sample sizes may in part reflect the practical and cost‐related challenges associated with body composition assessments such as DXA, as well as the logistical challenges of conducting such interventions.

Missing or inconsistently reported variance measures required assumptions that may have influenced FFM‐ratio estimates. Finally, eleven studies used per‐protocol analyses, which preserve randomisation less effectively than ITT analyses.

5. Conclusion

This systematic review finds that among studies achieving ≥ 10% weight loss, diet with exercise interventions is associated with the smallest reductions of FFM, while incretin‐based therapies and bariatric surgery showed substantially greater FFM losses, indicating clear differences across weight loss modalities. Our findings highlight the critical role of integrating exercise in preserving FFM during weight loss irrespective of the weight loss method and support the integration of exercise as an essential component of all weight‐loss strategies. This focus becomes even more timely as incretin‐based therapies now enable very large weight losses with proportionally larger reductions in FFM during weight loss compared with diet with exercise interventions.

Future research investigating the effect of weight loss on FFM according to weight‐loss approach could directly compare approaches in an RCT, ideally using MRI or CT to assess muscle quantity and quality. Incorporating measures of muscle strength and function would provide valuable insight into how FFM loss affects skeletal muscle mass.

Author Contributions

L.B.‐C. and I.Ø.C. contributed equally and are co‐first authors of the manuscript. R.M.S. initiated the study. L.B.‐C., I.Ø.C., L.B.M., R.M.S., S.B.K.J., and S.S.T. contributed to the design. L.B.‐C. and I.Ø.C. conducted data collection. L.B.‐C. and I.Ø.C. performed the analysis. All authors assisted in writing the manuscript.

Funding

L.B.M. is supported by the Danish Cardiovascular Academy (NNF20SA0067242) and The Danish Heart Foundation (PhD2023009‐HF and PD2‐2024005‐DCA), and R.M.S. is supported by the Danish Diabetes and Endocrine Academy (NNF22SA0079901).

Conflicts of Interest

Laura Busk‐Cirera holds Novo Nordisk A/S stocks. Signe Sørensen Torekov research grants, lecture fee, advisory board meeting, Novo Nordisk, lecture fee, Merck, and Ferring. The other authors declare no conflicts of interest.

Supporting information

Table S1: Baseline characteristics of the studies included.

Table S2: Baseline characteristics of the studies included.

Table S3: Baseline characteristics of participants.

Busk‐Cirera L., Carlsen I. Ø., Madsen L. B., et al., “Effects of Incretin‐Based Therapies, Diet and Exercise Interventions, and Bariatric Surgery on Fat‐Free Mass in Adults With Overweight or Obesity: A Systematic Review and Meta‐Analysis,” Diabetes, Obesity and Metabolism 28, no. 9 (2026): 7737–7751, 10.1111/dom.71006.

Data Availability Statement

Data used in this systematic review available in articles found from searches in public databases (pubmed, Embase, Cochrane Library).