The effects of exercise interventions on weight regain after weight loss: a systematic review and meta-analysis Authors: Jiachen Wang, Yun Chen, Kai Xu, Jiansong Dai Published: 2026 ABSTRACT AbstractTo systematically evaluate the effects of exercise during the post–weight-loss maintenance phase on weight regain. This study was registered in PROSPERO (CRD420251240526) and reported in accordance with the PRISMA statement. We searched PubMed, Embase, Web of Science, the Cochrane Library, and CNKI from database inception to October 19, 2025. Randomized controlled trials investigating exercise interventions during the weight-maintenance phase after successful weight loss in individuals with overweight/obesity were included. Risk of bias was assessed using the revised Cochrane Risk of Bias tool for randomized trials, RoB 2.0. Eleven trials involving 568 participants were included. Meta-analysis showed that the exercise group experienced significantly less weight regain than the control group (MD = − 2.81 kg, 95% CI: −5.12 to − 0.51; I² = 0%). Evidence for fat-mass reduction remained inconclusive because the primary random-effects analysis was non-significant and substantial heterogeneity was present (MD = − 3.39 kg, 95% CI: −7.24 to 0.46; I² = 69%). Exercise confer a small but consistent benefit for weight maintenance after weight loss; however, evidence for improvements in fat mass remains inconclusive. Further trials with longer follow-up, standardized exercise prescriptions, objective adherence monitoring, and harmonized body-composition assessment are warranted.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-57804-8.Subject terms: Diseases, Health care, Medical research Introduction Obesity has emerged as a major public health challenge worldwide1. Pharmacotherapy, energy-restricted diets, and bariatric surgery provide substantial short-term weight reduction1–6. However, post-treatment weight regain is common and rapid. For instance, individuals may regain more than 60% of their initial loss within one year of discontinuing GLP-1 receptor agonists2,7, while one-third to two-thirds of diet-induced weight loss is typically regained over a similar period3, a lifestyle-driven trajectory significantly influenced by behavioral sustainability and the frequency of nutritional follow-up8. Even after bariatric surgery, approximately 30% to 50% of patients experience significant weight regain within five years5,6,9. This recidivism is a complex, multi-factorial phenomenon. It links closely to pre-surgery cortisol levels10. Post-operative mental health and eating-related comorbidities also drive this relapse11. This phenomenon is driven by complex physiological and behavioral adaptations. Weight loss triggers compensatory responses, including alterations in appetite-regulating hormones such as ghrelin and GLP-12,7, as well as a persistent reduction in resting energy expenditure3. These adaptations create a metabolic environment conducive to energy storage, often outlasting the active weight-loss phase. Exercise is a low-cost, non-invasive countermeasure. It directly counteracts these mechanisms. Training increases energy expenditure, preserves fat-free mass, and improves insulin sensitivity. However, existing research evaluates exercise within isolated weight-loss pathways, leading to a lack of integrated evidence across different treatment modalities. Therefore, this systematic review and meta-analysis aims to quantify the effect of exercise interventions on attenuating weight regain. In this study, weight regain is defined as the net change in body weight at the end of the maintenance phase relative to the initial, pre-weight-loss baseline. Methods This study was carried out based on the guidelines of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement. The study protocol was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO) under registration number: CRD420251240526. Search strategy Two investigators systematically searched PubMed, Embase, Web of Science, the Cochrane Library, and China National Knowledge Infrastructure (CNKI) using a combination of Medical Subject Headings and free-text terms to identify studies meeting the inclusion criteria. The search strategy was designed by all authors and implemented by the first author. The search covered the period from database inception to October 19, 2025. A comprehensive search strategy was developed using controlled vocabulary and free-text terms related to exercise, weight loss, weight maintenance, weight regain, overweight/obesity, and randomized controlled trials. The strategy was adapted for each database according to its indexing system and search syntax. The complete database-specific search strategies are provided in supplementary materials. Inclusion and exclusion criteria for studies In our meta-analysis, eligible publications were included based on the following criteria:1) participants with overweight or obesity (BMI ≥ 25 kg/m²); 2) participants had achieved weight loss prior to entering the weight-maintenance phase through energy-restricted diets, pharmacotherapy, surgery, or other approaches; 3) the maintenance-phase intervention consisted of an exercise intervention; 4) at least one of the following outcomes was reported after the intervention: anthropometric measures, body composition measures, or blood-based indicators; 5) randomized controlled trial design; and 6) manuscripts published in the English language. The exclusion criteria were: (1) non-randomized studies; (2) conference abstracts, commentaries, reviews, or animal experiments; (3) non-English publications; (4) absence of a complete study protocol; and (5) duplicate publications. Study selection and data extraction After the initial de-duplication of the retrieved records in EndNote X9, two reviewers independently screened titles, abstracts, and full texts. They extracted data based on predefined criteria. Reviewers cross-checking all entries to ensure consistency. Any disagreements were resolved through adjudication by a third reviewer. This study strictly followed the Cochrane Handbook for Systematic Reviews of Interventions and was conducted and reported in accordance with the PRISMA guidelines for systematic reviews and meta-analyses. Risk of bias assessment Risk of bias was assessed independently by two reviewers using the revised Cochrane Risk of Bias tool for randomized trials, RoB 2.0. Disagreements were resolved through discussion or consultation with a third reviewer. The assessment was conducted according to the five RoB 2.0 domains: (1) bias arising from the randomization process; (2) bias due to deviations from intended interventions; (3) bias due to missing outcome data; (4) bias in measurement of the outcome; and (5) bias in selection of the reported result. Overall risk of bias was judged as “low risk of bias,” “some concerns,” or “high risk of bias” according to the RoB 2.0 algorithm. For exercise interventions, lack of blinding of participants and intervention providers was not automatically judged as high risk of bias; instead, we evaluated whether knowledge of the assigned intervention led to deviations from intended interventions that were likely to affect the outcomes. Statistical analysis Data were analyzed using RevMan 5.4. Effect estimates were expressed as risk ratios (RRs) or mean differences (MDs) with corresponding 95% confidence intervals (CIs). Clinical and methodological heterogeneities were anticipated across trials. Variations included participant traits, weight-loss methods, exercise modalities, and durations. Therefore, a random-effects model served as the primary analytic model. Furthermore, fixed-effect model analyses were performed as sensitivity analyses to evaluate the robustness of the findings. P < 0.05 was considered statistically significant. If a study included multiple exercise arms, only the single arm most consistent with the research objectives was selected for analysis to avoid unit-of-analysis errors. Specifically, we implemented a pre-specified hierarchical strategy to ensure transparency in this selection: first, immediate exercise interventions after weight loss were prioritized over delayed starts to address the critical window of physiological adaptation ; second, resistance training (RT) was selected given its established role in preserving fat-free mass and resting metabolic rate ; finally, when multiple dosages were available, we prioritized more conservative exercise volumes to evaluate the effectiveness of sustainable, real-world prescriptions. Outcomes The primary outcomes were the amount of weight regain and the change in fat mass (both standardized to kilograms), assessed at the last follow-up. Weight regain was defined as the difference between body weight at the last follow-up during the maintenance phase (Wt) and the initial baseline weight recorded prior to any weight-loss intervention (W0): ΔW = Wt − W0. Similarly, the change in fat mass was calculated as ΔFM = FMt − FM0. A positive ΔW (or ΔFM) indicates an increase in body weight (or fat mass) during the maintenance phase (i.e., regain), whereas a negative value indicates further reduction or maintenance at a lower level during follow-up.The absolute changes in body weight and fat mass in kilograms were selected as the primary outcomes. Kilograms were the most frequently reported or derivable metric across trials. This choice provides direct clinical interpretability. However, we acknowledge that these absolute kilogram-based metrics may not fully account for variations in baseline body size, initial weight-loss magnitude, or post-weight-loss starting status across different studies.Additionally, adherence was evaluated as a supplementary outcome to assess the feasibility and behavioral compliance of the long-term exercise interventions. Adherence data were systematically extracted from the attendance logs or protocol completion rates reported in the primary literature. An adherence event was defined as the number of participants reported as adherent to the intervention or as having completed the prescribed protocol/final follow-up in each study arm. The denominator was the number of participants assigned to the corresponding group. Because adherence definitions varied across trials, we preferentially extracted protocol completion or attendance-based adherence data when available. Results Study search results A comprehensive search of PubMed, Embase, the Cochrane Library, Web of Science, and CNKI was conducted up to October 19, 2025. A total of 1,529 records were identified, of which 441 duplicates were removed. After screening titles and abstracts, 1,052 records were excluded. Full texts of the remaining 36 articles were reviewed in detail; 25 were excluded due to ineligible interventions, inability to obtain sufficient details of the intervention protocols, or unavailable outcome data. Ultimately, 11 studies were included in the present analysis. The study selection process is presented in the PRISMA flow diagram (Fig. 1). Flow diagram of study selection. Risk of bias of the included literature The risk of bias of the included randomized controlled trials was assessed using the revised Cochrane Risk of Bias tool for randomized trials, RoB 2.0, at the outcome level. The assessment was conducted according to the five RoB 2.0 domains: bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome, and bias in selection of the reported result. Overall judgments were categorized as low risk of bias, some concerns, or high risk of bias according to the RoB 2.0 framework. (Figures 2 and 3). Overall summary of risk of bias for the included studies. Risk-of-bias assessment for the included studies. Characteristics of the included literature All 11 included studies12–22 reported the effects of exercise interventions on post–weight-loss weight regain in individuals with overweight/obesity. All studies were randomized controlled trials published in English, with publication dates ranging from 1996 to 2023. A total of 568 participants were included in the analysis, comprising 216 in the control groups and 352 in the exercise groups. The mean age of participants varied widely across studies, ranging from 38.6 to 70 years. Sex composition also differed substantially: three studies enrolled men only (n = 3) and three enrolled women only (n = 3); among the remaining mixed-sex trials, women generally predominated (e.g., Louisa Y Herring, 2017). The initial weight-loss approaches included bariatric surgery (n = 4) and (very) low-calorie diets (n = 7). During the weight-maintenance phase, interventions consisted of resistance training and/or aerobic exercise (e.g., cycle ergometry, walking, and deep-water running). Intervention duration ranged from 12 weeks (Louisa Y Herring, 2017) to 53 weeks (Lejeune, 2003). Overall, adherence was generally good, with most studies reporting high completion rates. Regarding outcomes, the included trials primarily assessed anthropometric measures (e.g., height and body weight) and body composition, often complemented by imaging-based evaluations to quantify changes in body composition (Table 1). Table 1Characteristics of Included Studies.StudyYearSample size(n)Mean age(years)Sex(M/F)Adherence(completed/assigned)Intervention durationControl groupExercise groupInterventionWeight(kg)Fat mass(kg)Waist(cm)InterventionWeight(kg)Fat mass(kg)Waist(cm)Kaikkonen1220238046.121/5980/12012 monthsWeight-loss counseling; usual lifestyle-6.7–-4.9Resistance training-5.8–-8.7Bellicha 1320225443.5Women only54/766 monthsAdditional protein supplementation-34.0-21.0–Additional protein supplementation + resistance training-34.0-23.8–Lundgren142021166Placebo: 43; Exercise: 43Placebo: 18/31; Exercise: 17/31166/19552 weeksPlacebo + usual physical activity+ 6.1––Placebo + aerobic exercise+ 2.0––Hintze15201854Not reportedWomen onlyExercise: 25/36; Control: 29/3412 monthsNo exercise intervention-4.2-3.1–Resistance training-3.0-2.7–Herring16201721Exercise: 52.4; Control: 44.3Exercise: 1/11; Control: 1/11Exercise: 11/12; Control: 10/1212 weeksRoutine follow-up care+ 2.9+ 2.1–Resistance training-2.7-1.9–Coleman17201643Exercise: 52.0; Control: 46.6Exercise: 4/21; Control: 4/22Exercise: 20/26; Control: 23/256 monthsUsual care+ 1.2––Functional strength, flexibility, and aerobic training-0.2––Lejeune1820032939.0Men onlyDiet: 15/20; Diet+exercise: 14/2053 weeksDiet restriction+ 7.1+ 5.3–Diet restriction + aerobic exercise+ 8.9+ 6.0–van Aggel-Leijssen19200229Control: 38.6; Exercise: 39.3Men onlyControl: 15/17; Exercise: 14/2040 weeksDiet restriction+ 7.1+ 5.3–Aerobic exercise+ 8.9+ 6.0–Borg2020026842.6Men only68/906 monthsNo exercise intervention000Resistance training-0.5-1.2-1.4Fogelholm2120007440.0Women only74/8240 weeksNo exercise intervention+ 9.7+ 8.4–Walking (aerobic)+ 5.9+ 8.5–Ballor22199618Not reportedAerobic: 4/5; Resistance: 4/518/1812 weeksNone–––Aerobic: walking; Resistance: resistance trainingAerobic: -2.5; Resistance: +0.4Aerobic: -1.8; Resistance: -1.2– Characteristics of Included Studies. Meta-analysis Body weight Heterogeneity testing was performed for the nine included studies[13–17,19−21]. Given the anticipated clinical and methodological heterogeneity across the trials, a random-effects model was used as the primary analytic model. No statistical heterogeneity was observed (Chi2 = 3.95, df = 8, P = 0.86; I2 = 0%; Tau2 = 0.00). The pooled results indicated that, compared with the control group, the exercise intervention significantly reduced weight regain (MD = − 2.81 kg, 95% CI: −5.12 to − 0.51; Z = 2.39, P = 0.02) (Fig. 4). Forest plot of the meta-analysis comparing body weight between the exercise and control groups (Random-effects model). Additionally, a sensitivity analysis using a fixed-effect model yielded the identical pooled estimate(MD = − 2.81 kg, 95% CI: −5.12 to − 0.51; P = 0.02), demonstrating that the findings are robust to the choice of different statistical models(Fig. 5). Forest plot of the meta-analysis comparing body weight between the exercise and control groups (Sensitivity analysis: Fixed-effect model). Fat mass For fat mass, six studies13,15,16,19–21 were included in the meta-analysis.The random-effects model showed that the exercise group had a greater reduction in fat mass than the control group, but the between-group difference did not reach statistical significance (MD =−3.39 kg, 95% CI:−7.24 to 0.46; P = 0.08). Substantial heterogeneity was observed across studies (Tau² = 15.02; Chi² = 15.97, df = 5, P = 0.007; I² = 69%) (Fig. 6). Forest plot of the meta-analysis comparing fat mass between the exercise and control groups (Random-effects model). A fixed-effect sensitivity analysis showed a statistically significant effect in favor of exercise (MD = − 3.48 kg, 95% CI: −5.52 to − 1.45; P = 0.0008; I² = 69%) (Fig. 7). However, because substantial between-study heterogeneity was present, the random-effects model was retained as the primary analysis for interpretation. Forest plot of the meta-analysis comparing fat mass between the exercise and control groups (Sensitivity analysis: Fixed-effect model). Adherence analysis The forest plot showed that, across the nine included studies13–17,19–21, the individual effect estimates (risk ratios, RRs) and their 95% CIs all crossed 1.0. Thus, no single trial detected significant adherence differences between groups. In the pooled analysis, the overall RR was 0.94 (95% CI: 0.85 to 1.03). This interval included 1.0 (Z = 1.26, P = 0.21). the combined results further confirmed that adherence did not differ significantly between groups at the aggregate level. In this study, RR was calculated as the proportion of adherent participants in the exercise group divided by the proportion of adherent participants in the control group.Heterogeneity assessment indicated moderate between-study heterogeneity (Chi² = 13.87, df = 8, P = 0.09; I² = 42%). Funnel plot inspection suggested an overall symmetrical inverted funnel shape for the nine studies; however, slight asymmetry was observed in the lower part of the plot, implying potential bias. However, because the number of trials included in this analysis was fewer than 10 (N = 9), the power of the funnel plot to reliably detect publication bias is inherently limited. Consequently, these funnel plot findings must be interpreted with extreme caution and should be considered purely as an exploratory conclusion rather than definitive evidence of publication bias (Figs. 8 and 9). Forest plot of the meta-analysis comparing adherence between the exercise and control groups. Funnel plot of the meta-analysis comparing adherence between the exercise and control groups. Discussion This study synthesized evidence from 11 trials including a total of 568 participants to systematically evaluate the effects of exercise interventions on post–weight-loss weight regain. Compared with the control group, the exercise group exhibited a smaller amount of weight regain at the last follow-up, and the between-group difference reached statistical significance (MD = − 2.81 kg, 95% CI: −5.12 to − 0.51; I² = 0%; P = 0.02). In contrast, the between-group difference in fat mass change (ΔFM) was not statistically significant (MD = − 3.39 kg), with the confidence interval crossing zero (95% CI: −7.24 to 0.46; P = 0.08) and substantial heterogeneity observed (I² = 69%). Although the fixed-effect sensitivity analysis showed a significant reduction in fat mass, the random-effects model was considered more appropriate because substantial clinical and statistical heterogeneity was present. Therefore, the evidence for an effect of exercise on fat mass remains inconclusive. Because only six studies were included in the fat-mass analysis, prediction intervals were not estimated, and the heterogeneous result should therefore be interpreted cautiously. Collectively, these findings suggest that, during the weight-maintenance phase, exercise can help attenuate overall weight regain, whereas its effect on fat mass remains uncertain. This uncertainty may be attributable to differences across included studies in intervention modality, intensity and duration, participant characteristics, and outcome assessment methods. Our findings are broadly consistent with previous evidence supporting the role of exercise in long-term weight management. For example, Marc-Hernández et al. reported that exercise training after sleeve gastrectomy helped reduce weight regain and improve cardiometabolic outcomes23. However, previous reviews have also shown that the effect of exercise on body weight is often modest and may vary according to population, exercise prescription, and weight-loss pathway24,25. The present study extends this evidence by focusing specifically on the post–weight-loss maintenance phase and by combining trials conducted after different initial weight-loss approaches, including diet-induced weight loss and bariatric surgery. Exercise modality may partly explain the observed differences across trials. Aerobic exercise and walking were associated with smaller weight regain in several studies. For example, Lundgren et al. reported lower weight regain in the aerobic exercise group than in the control group14, and Fogelholm et al. found that walking attenuated weight regain after a very-low-energy diet21. In contrast, resistance training appeared to have a more consistent influence on body composition than on body weight. In the trials by Borg et al. and Ballor et al., resistance training produced relatively small changes in total body weight but was associated with reductions in fat mass20,22. Kaikkonen et al. also reported greater waist-circumference reduction in the resistance-training group than in controls12. These findings suggest that body weight alone may underestimate the benefits of resistance-based interventions, particularly when changes in fat-free mass accompany changes in fat mass. Intervention duration is another important source of variation. Short-term interventions, such as those in Herring et al. and Ballor et al., showed favorable changes in body weight or fat mass but provided limited information on long-term maintenance16,21. Medium- and long-term trials showed less consistent results. Coleman et al. and Lundgren et al. suggested that exercise attenuated weight regain14,17, whereas Hintze et al. found no significant advantage of a one-year resistance-training program for body composition or energy expenditure15. These inconsistent findings may reflect differences in exercise adherence, supervision, intervention intensity, and concurrent dietary or behavioral support. The non-significant pooled effect on fat mass should be interpreted with caution. Fat-mass outcomes were more heterogeneous than body-weight outcomes, which may be explained by differences in exercise mode, dose, supervision, dietary control, and body-composition assessment methods. Previous trials and reviews have shown that aerobic exercise, resistance training, and combined exercise can produce different effects on total fat mass, visceral adiposity, and subcutaneous fat26–30. In addition, exercise-induced energy expenditure may be partly offset by compensatory increases in energy intake or reductions in non-exercise activity thermogenesis, which can attenuate changes in fat mass31,32. Several physiological mechanisms support the use of exercise during the weight-maintenance phase. Weight loss is accompanied by hormonal adaptations, increased appetite, and persistent reductions in energy expenditure, all of which promote weight regain33–35. Sleep and stress may further influence this process36. Exercise can counter these adaptations by increasing total energy expenditure, preserving fat-free mass, attenuating reductions in resting metabolic rate, improving insulin sensitivity, and enhancing fat oxidation31,32,34,37,38. Exercise may also improve appetite regulation, eating behavior, and self-efficacy, which are important for long-term adherence to weight-control behaviors16,21. These findings have practical implications. Exercise should be considered an important component of post–weight-loss maintenance, but it should not be evaluated solely by changes in body weight. In clinical practice, aerobic exercise may be particularly useful for limiting weight regain, whereas resistance training may provide additional benefits for preserving or improving body composition. Combining exercise with dietary strategies, such as higher-protein or low–glycemic index diets, may further support weight maintenance39,40. Recent evidence also suggests that combining exercise with pharmacotherapy may produce greater weight-maintenance benefits than either strategy alone41. This review has several limitations. First, although absolute changes in body weight and fat mass are clinically intuitive, they may not fully account for differences in baseline body size, initial weight-loss magnitude, or post–weight-loss starting status across studies. Because many trials did not consistently report the percentage of lost weight regained or standardized post–weight-loss baseline values, relative metrics could not be reliably synthesized. Second, the fat-mass analysis showed substantial heterogeneity, likely reflecting differences in exercise prescription, supervision, dietary co-interventions, and body-composition measurement methods. Third, long-term exercise interventions are vulnerable to declining adherence, which may reduce the delivered exercise dose and underestimate the true effect of exercise. Fourth, only English-language full-text articles were included. Although CNKI was searched to identify potentially relevant English-language records indexed in Chinese databases, the exclusion of non-English studies may have introduced language bias. Future trials should use longer follow-up periods, standardized and clearly quantified exercise prescriptions, objective adherence monitoring, and harmonized body-composition assessment methods. They should also measure dietary intake, non-exercise physical activity, and cardiorespiratory fitness to better clarify how exercise modality and dose influence weight regain and body composition after weight loss. Conclusion This study conducted a systematic review and meta-analysis to evaluate the effects of exercise interventions during the post–weight-loss maintenance phase on outcomes related to weight regain. Based on the available evidence, exercise interventions produced a small but statistically significant benefit for weight maintenance compared with control conditions (MD = − 2.81 kg), with low heterogeneity across studies (I² = 0%), suggesting that this finding was relatively consistent within the included evidence base. In contrast, although fat mass showed a decreasing trend (MD = − 3.39 kg), the pooled effect did not reach statistical significance and substantial between-study heterogeneity was observed (I² = 69%), indicating that definitive evidence for an effect of exercise on fat mass remains lacking. These results suggest that exercise during the maintenance phase after weight loss can help attenuate weight regain, but such benefits may not necessarily be accompanied by substantial reductions in fat mass. Future research requires longer follow-up periods. Exercise prescriptions must be standardized and quantifiable. Adherence monitoring must be objective. Researchers must harmonize body-composition assessment methods and conditions. Studies should also track dietary intake and non-exercise activity changes. These steps will clarify how exercise modalities and doses affect weight regain and body composition. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Author contributions J.C.W. and Y.C. conceived the study. J.C.W. and Y.C. designed the search strategy. J.C.W. and Y.C. performed study screening and data extraction. J.C.W. conducted the meta-analysis. J.C.W. and Y.C. assessed risk of bias. J.C.W. drafted the manuscript. All authors critically revised the manuscript and approved the final version. Funding This research was funded by the Jiangsu Provincial Undergraduate Innovation Training Program. Data availability All data generated or analysed during this study are included in this published article and its Supplementary Information files. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement All data generated or analysed during this study are included in this published article and its Supplementary Information files.