Effects of exercise training programmes on fasting gastrointestinal appetite hormones in adults with overweight and obesity: A systematic review and meta-analysis

,


Introduction
The World Health Organization (WHO) documented that more than 1.9 billion adults aged 18 years and over are living with overweight or obesity (BMI ≥25 kg/m 2 ; WHO, 2021). Body mass gain results from repeated episodes of increased energy intake and/or reduced accumulation of physical activity and exercise (Blundell et al., 2015;Hall et al., 2012;Spiegelman & Flier, 2001). Thus, participation in exercise programs has been advocated as a non-pharmacological and non-surgical approach to obesity prevention and treatment (Dorling et al., 2018). However, the effectiveness of exercise-induced body mass loss in the absence of caloric restriction is variable (Gibbons et al., 2017;Jackson et al., 2018;King et al., 2008;Manthou et al., 2010) and is attenuated by compensatory changes in appetite and energy intake (Blundell et al., 2015;King et al., 2008). Therefore, understanding how appetite regulatory mechanisms are altered by exercise in overweight and obese individuals is of great interest.
The regulation of appetite is controlled by a coordinated and complex system of central neurons and peripheral appetite hormones (Augustine et al., 2018). It is well established that gastrointestinal hormones such as anorexigenic peptide YY (PYY), glucagon-like peptide 1 (GLP-1), cholecystokinin (CCK) and orexigenic ghrelin are acutely altered in response to feeding (Suzuki et al., 2011). Despite the lack of a definitive consensus, several studies suggest that an acute bout of exercise suppresses levels of acylated ghrelin while simultaneously increasing PYY and GLP-1, and this has the potential to modify food and drink intake following exercise (Stensel., 2010;King et al., 2013, Schubert et al., 2014Douglas et al., 2016;Ouerghi et al., 2021). Existing evidence also suggests that changes in gastrointestinal appetite hormone concentrations depend on the mode, intensity, and duration of acute exercise sessions, along with the body weight and nutritional status of investigated participants (Dorling et al., 2018;Zouhal et al., 2019).
The evidence regarding the effects of exercise training on gastrointestinal hormones has not been extensively synthesised. Previous narrative reviews suggest that chronic exercise has no influence on ghrelin concentrations independently of body mass loss and leads to increased fasting concentrations of PYY and GLP-1 Martins et al., 2008;Stensel, 2010). In addition, a narrative review suggested that exercise training increases fasting hunger and postprandial satiety and improves the coupling between exergy expenditure and energy intake exergy expenditure in response to food intake (Dorling et al., 2018). These reviews, however, did not adhere to a rigorous systematic methodology to assess study selection, data extraction, and study quality. A more recent systematic review summarising evidence on total and acylated ghrelin responses to chronic exercise reports that data are inconsistent, with studies showing increased, decreased, or unchanged circulating concentrations of total ghrelin and acylated ghrelin (Ouerghi et al., 2021). The inconsistency of the data might be related to large differences between studies in terms of participant characteristics, particularly body mass status (Ouerghi et al., 2021). Adults living with overweight and obesity have altered concentrations of gastrointestinal hormones in response to exercise (Lean & Malkova, 2016;Zouhal et al., 2019), so research is needed with a specific focus on this population. Ouerghi et al. (2021) also included studies that did not apply a randomised control trial design and hence did not provide the most accurate estimate of the exercise training response. Therefore, this systematic review and meta-analysis aimed to fill the evidence-based gap by quantifying the effect of exercise training programs (duration ≥ four weeks) from randomised controlled studies on the fasting gastrointestinal appetite hormones of adults living with overweight and obesity.

Literature search
A systematic search strategy based on keywords related to exercise interventions and gastrointestinal appetite hormones was developed to identify relevant papers using MEDLINE, EMBASE, Cochrane Library, Web of Science, and Scopus databases (Supplementary material 1). Protocols and ongoing trials were identified through the International Standard Randomized Controlled Trial Number (ISRCTN) trials registry and ClinicalTrials.gov. The search was conducted from database inception up to November 25th, 2021. In addition, hand searches of reference lists of key reviews and retrieved full-text articles were conducted. Citation screening through Google Scholar was also performed for included full-text articles as a final check.

Eligibility criteria
Criteria for consideration in this systematic review included adults (aged ≥18 years) with overweight and obesity (BMI ≥25 kg/m 2 ), an exercise intervention (≥4 weeks) that investigated the impact on at least one gastrointestinal appetite hormones (total ghrelin, acylated ghrelin, PYY, CCK, and GLP-1). The inclusion of exercise interventions with a minimum exercise duration of 4 weeks was based on the evidence that this duration achieves measurable body weight and body fat change (Jackson et al., 2018;Manthou et al., 2010). For studies that included adolescents and adults, at least 50% of the sample were required to be aged ≥18 years; further, at least 50% of participants in selected studies had to have a BMI ≥25 kg/m 2. Studies were excluded if they were not published in English and did not apply a randomised control trial design since randomized controlled trials are considered the most reliable method for determining the effectiveness of interventions (Akobeng, 2005). The primary outcome of this review was to examine the differences in fasting gastrointestinal hormones between the exercise and control groups following interventions. The change in body mass and BMI after exercise training programs was also considered.

Selection process
The articles identified were electronically imported into Covidence software for screening (www.covidence.org) and duplicate studies were automatically removed. The screening of the title and abstracts was conducted by two independent reviewers (SA & TA). Potentially relevant articles were sourced, and the full-text articles were screened against the criteria above (TA & DM). Any disagreements on study selection were discussed and resolved by a consensus decision. Reliability between reviewers for the title and abstract screening and full-text screening were calculated in SPSS (version 27; SPSS IBM, New York, NY, USA) using Cohen's kappa scores, demonstrating almost perfect agreement and moderate agreement (ĸ = 0.99 and 0.65, for the title and abstract and full-text articles screening, respectively; Landis & Koch, 1977).

Data extraction
Two reviewers (TA & SW) independently conducted the data extraction from a pre-specified data extraction template using Microsoft Excel adapted from reviews McGarty et al., 2018). The results were compared, and a consensus was agreed upon. Data extraction included study characteristics (e.g., the location and aim of the study), intervention components (e.g., frequency, intensity, type of exercise sessions, and the duration of the whole intervention), outcome measures (e.g., total ghrelin, acylated ghrelin, PYY, GLP-1, CCK, body mass and BMI), and results (e.g., post-intervention and control mean, standard deviation, and sample size).

Risk of bias
The Cochrane Collaborations tool version 2 (Higgins et al. 2011) was used to judge the risk of bias in included studies. The studies were judged as low, with some concerns or high risk of bias according to the following six domains: (1) the randomisation process; (2) deviations from the intended interventions (effect of assignment to intervention); (3) deviations from the intended interventions (effect of adhering to intervention); (4) missing outcome data; (5) measurement of the outcome; (6) selection of the reported result. Two researchers (TA & LH) independently conducted the risk of bias, and disagreement was resolved through discussion with a third reviewer (AM).

Data synthesis
Meta-analyses were conducted using Comprehensive Meta-Analysis (Version 3.0 for Windows: Biostat, Englewood, Colorado, USA). Effect sizes for each outcome were calculated as the standardized difference in means, expressed as Cohen's d (Cohen, 1988). Effects sizes were calculated based on imputed data (post-intervention and control mean, standard deviation, and sample size) and interpreted as small (d = 0.20), moderate (d = 0.50) or large (d = 0.80; Cohen, 1988). A random-effects model was used to compute effect sizes to account for true variation between study effects, as well as a random error within a study (Der-Simonian & Laird, 1986). Heterogeneity was assessed using Cochrane's Q statistic, with a significance level of P < 0.05 indicating evidence of statistical heterogeneity. The I 2 statistic was used to quantify the degree of heterogeneity, with I 2 ≥ 50% indicating substantial heterogeneity (Higgins et al., 2003). Due to the small number of studies included in this systematic review, it was not possible to conduct any subgroup or moderator analysis to explore heterogeneity between studies or explore the relationship between changes in gastrointestinal hormones and changes in body weight/BMI (Higgins & Green, 2011). The limited number of studies also meant that there was not adequate power of reliability of tests to detect the presence of publication bias (Higgins & Green, 2011).

Study selection process
The electronic database searches identified a total of 13,204 records. Seven further studies were identified from hand searching of bibliographies of included studies. Duplicate studies were removed, and initial title and abstract screening were conducted for 9325 studies. Sixty-eight studies underwent full-text screening, and nine studies met the eligibility criteria (Fig. 1). The full list of excluded studies and the reasons for exclusion are presented in the supporting information (Supplementary Materials). The nine eligible studies are descriptively summarized in Table 1. Three studies did not provide sufficient data for appetite hormones and therefore were not included in the meta-analysis (Foster-- Schubert et al., 2005;Mason et al., 2015;Shakiba et al., 2019). The outcome data from two studies were not reported in suitable units for the meta-analysis (Foster- Schubert et al., 2005;Mason et al., 2015), and one study did not report data for the control group to allow comparison with the intervention group (Shakiba et al., 2019).
A total of 707 participants (259 men and 448 women) were recruited across the nine studies, with the number of participants included in the studies ranging from 26 to 186. Participants were physically inactive adults with a mean age range of 28-69 years and a mean BMI between 27.1 and 31.8 kg/m 2 . Three studies recruited adults classified as living with overweight (Ahmadi et al., 2019;Rosenkilde et al., 2013;Sim et al., 2015), five recruited adults living with overweight and/or obesity (Foster-Schubert et al., 2005;Guelfi et al., 2013;Mason et al., 2015;Quist et al., 2019;Shakiba et al., 2019), and one study recruited only adults living with obesity (Kang et al., 2018). Five studies recruited male participants (Guelfi et al., 2013;Rosenkilde et al., 2013;Sim et al., 2015;Ahmadi et al., 2019;Shakiba et al., 2019), three only female participants (Foster-Schubert et al., 2005;Kang et al., 2018;Mason et al., 2015), and one study included both males and females (Quist et al., 2019). Participants were asked not to change their habitual diets during exercise training in all studies.

Adherence to the interventions
Adherence to training sessions was reported by six studies (Foster-- Schubert et al., 2005;Guelfi et al., 2013;Mason et al., 2015;Quist et al., 2019;Rosenkilde et al., 2013;Sim et al., 2015). Adherence to the exercise intervention sessions was high, ranging from 82 to 99%. In the study, conducted by Foster-Schubert et al. (2005), there was similar adherence to supervised and home-based exercise sessions, which accounted for 82% and 87%, respectively. In the study by Guelfi et al. (2013) adherence at 92% did not differ between endurance and resistance training groups. The intensity of exercise sessions also had no effect on training adherence in the study by Sim et al. (2015), in which the high-intensity intermittent exercise group and the moderate-intensity continuous group achieved adherence of 98% and 97%, respectively. Quist et al. (2019) found that women and men similarly adhered to the intervention sessions. For men and women, adherence to cycle training was 91% and 100%, respectively. The adherence was similar for moderate-intensity exercises (men: 100%, women: 99%) and vigorous-intensity exercises (men: 95%, women: 90%).

Attrition
Five studies presented data for participants who completed the study and did not report any participants dropping out of the study (Ahmadi et al., 2019;Guelfi et al., 2013;Kang et al., 2018;Shakiba et al., 2019;Sim et al., 2015). Four studies reported participants who dropped out from the interventions (Foster- Schubert et al., 2005;Mason et al., 2015;Quist et al., 2019;Rosenkilde et al., 2013), with attrition rates ranging from 2.9% to 7.5%. In the study by Foster- Schubert et al. (2005), four participants (4.5%) dropped out of the exercise and one (1.2%) of the control group. In the study by Quist et al. (2019), the interventions were completed by 87 out of 130 participants, resulting in an overall attrition rate of 33% (11% control group; 46% bike group; 23% moderate-intensity exercise group, and 24% vigorous-intensity exercise group). Only one study reported reasons for dropping out from the exercise intervention and identified that a participant in the high-intensity exercise group dropped out due to the recurrence of a previously experienced injury (Rosenkilde et al., 2013).

Assessment of risk of bias
Eight studies were judged to be at high risk of bias, with one study considered some concern of risk of bias (Mason et al., 2015). This was due mainly to a lack of reporting across domains. The risk of bias for each domain is presented in the supporting information (Table S1, Supplementary material). Five studies were assessed as low risk of bias for the randomisation process as they reported appropriate methods for sequence generation and allocation concealment; for example, computerised program, block randomisation, and random number generator software (Foster-Schubert et al., 2005;Mason et al., 2015;Quist et al., 2019;Rosenkilde et al., 2013;Sim et al., 2015). Four studies were judged as some concern because although they reported the participants were randomised, they did not provide any details of the methodologies used (Ahmadi et al., 2019;Guelfi et al., 2013;Kang et al., 2018;Quist et al., 2019).
The effect of the intervention assignment was judged as some concern. This was mainly because of the nature of the physical activity interventions making blinding participants and researchers challenging. Information was also not reported regarding deviation arising because of the trial context. Only two studies specified adhering to intention-totreat analysis (Quist et al., 2019;Rosenkilde et al., 2013). The risk of bias in relation to the effect of adherence to the intervention was high, again due to the lack of blinding of participants and personnel. There was also a lack of information on protocol interventions balanced across intervention groups and inappropriate analysis to estimate the effect of adherence.
The risk of bias for missing outcomes was judged as low for most of the studies and as high for only one study (Quist et al., 2019), which had a high number of dropouts (>20%) and, therefore, missing data on the outcome measures. It was assumed that outcome assessments were not blinded. Selection bias was assessed as low risk of bias in four studies as they reported having a published protocol with a pre-specified statistical analysis plan (Foster- Schubert et al., 2005;Mason et al., 2015;Quist et al., 2019;Rosenkilde et al., 2013). Five studies were assessed as some concern as it was not reported whether they followed a pre-specified plan (Mason et al., 2015;Guelfi et al., 2013;Kang et al., 2018;Ahmadi et al., 2019;Quist et al., 2019).

Power and sample size
Only one study reported sample size calculation for appetite hormone outcomes (Shakiba et al., 2019). Four studies conducted sample size calculations for primary outcomes; however, these were not related to appetite hormones (Foster-Schubert et al., 2005;Rosenkilde et al., 2013;Mason et al., 2015;Quist et al., 2019). Four studies did not report sample size calculation; therefore, it is uncertain whether these studies were adequately powered to detect a significant difference in outcome measures.
Two studies did not provide sufficient data to be included in the meta-analysis (Foster-Schubert et al., 2005;Mason et al., 2015). According to one of these studies, a significant (P = 0.03) increase in fasting total ghrelin concentration was found following endurance exercise training at 75% HR max for 12 months (Foster-Schubert et al., 2005), whereas no change (P = 0.53) in fasting total ghrelin was found in another study that applied 12 months of exercise intervention (Mason et al., 2015).
One study did not provide sufficient data to be included in the metaanalysis for acylated ghrelin. This study reported a significant decrease of acylated ghrelin concentration in the resistance (P < 0.001), a combined (resistance and endurance) (P < 0.001) and endurance training (P < 0.05) groups after 12 weeks in men with overweight (Shakiba et al., 2019, Table 1).

PYY
PYY levels were reported in four studies. Three out of four studies (seven intervention groups) were included in the meta-analysis. A pooled effect size (Fig. 4) indicated that there was no significant difference in fasting PYY concentration between the exercise and control interventions (d = − 0.16, 95% CI: − 0.62 to 0.31, P = 0.51). There was substantial heterogeneity between studies [Q-statistic = 8.30, df = 6, (P = 0.22); I 2 = 27.68%].
Two studies were not included in the meta-analysis due to insufficient data (Rosenkilde et al., 2013;Shakiba et al., 2019). In one of these studies, the fasting concentration of PYY was significantly increased due to 12 weeks of resistance (P = 0.03), combined (endurance and resistance) (P = 0.04), and endurance (P = 0.04) training (Shakiba et al., 2019). Another study found that an increase in the fasting PYY concentration tended (P = 0.07) to be higher in the high-dose endurance exercise training group compared to the control group (Rosenkilde et al., 2013).

GLP-1
Two studies measured the effects of exercise training on GLP-1 (Quist et al., 2019;Shakiba et al., 2019). One study investigated the impact of the 6-month endurance training programme with the intensity of exercise sessions corresponding to 70% of V O 2 peak and it found that after six months, fasting GLP-1 concentration in the exercise group was significantly higher than in the control group (P = 0.04) (Quist et al., 2019). In the study by Shakiba et al. (2019), fasting GLP-1 concentration was not affected by exercise regimens (resistance, endurance, and combined training) after 12 weeks (P > 0.05).

CCK
Only one study investigated the effect of exercise on fasting plasma level of CCK and found that after six months of exercise intervention, fasting CCK concentration did not change (P ≥ 0.5) in any endurance exercise training group (moderate intensity leisure, vigorous-intensity leisure or active commuting by bike) compared with the control group (Quist et al., 2019).

Body mass and BMI
Overall, based on results from seven selected studies, the metaanalysis showed a significant reduction in body mass (d: − 0.22, 95% CI: − 0.42 to − 0.03, P = 0.03). A BMI analysis was conducted based on eight studies and found a significant reduction (d: − 0.31, 95% CI: − 0.50 to − 0.12, P = 0.001) after exercise interventions in comparison to Fig. 2. Forest plot of randomised controlled trials comparing the impact on ghrelin of interventions involving exercise training with non-exercise control group comparators. Effects of trials are presented as means and SDs at the end of the interventions and SMD (95% CI). a, endurance training; b, combined (aerobic and resistance) training; c, endurance training (moderate intensity); d, endurance training (high intensity); DF, degrees of freedom; SMD, standardised mean difference; CI, confidence interval; Q, heterogeneity statistic to test homogeneity; I 2 , index of heterogeneity beyond within-study sampling error. In the study of Ahmadi et al. (2019) ghrelin concentration was measured in ng/ml. control (Figs. 5 and 6). In both cases, there was no evidence of statistical heterogeneity between studies for body mass [Q-statistic = 6.16, df = 13, (P = 0.94); I2 = 0%] and BMI [Q-statistics = 10.21, df = 14(P = 0.47); I 2 = 0%]. Two studies had insufficient results to be included in the meta-analysis (Foster-Schubert et al., 2005;Ahmadi et al., 2019). These studies reported significantly reduced body mass and BMI due to 12 months of endurance training (Foster-Schubert et al., 2005) and significantly reduced BMI due to 8 weeks of endurance training (Ahmadi et al., 2019).

Correlation between body mass and appetite hormone changes
Three studies analysed the correlation between body mass and appetite hormones (Foster-Schubert et al., 2005;Mason et al., 2015;Shakiba et al., 2019) and found that weight loss (P < 0.05) and reduction in BMI (P < 0.05) that occurred with exercise was positively associated with an increase in plasma concentration of total ghrelin (P < 0.05) (Foster-Schubert et al., 2005). Increased ghrelin concentrations were also linked to more advanced body weight loss (P < 0.0001), and BMI reduction (P < 0.0001) in the study by Mason et al. (2015). In the study of Shakiba et al. (2019) body mass loss and BMI reduction were positively correlated with a reduction in acylated ghrelin (P = 0.003, P = 0.009, respectively) and negatively correlated with an increase in plasma PYY concentration (P = 0.003, P = 0.03, respectively).

Discussion
To the best of our knowledge, this is the first systematic review with meta-analysis to investigate the effect of exercise training programs with Fig. 3. Forest plot of randomised controlled trials comparing the impact on acylated ghrelin of interventions involving exercise training with non-exercise control group comparators. Effects of trials are presented as means and SDs at the end of the interventions and SMD (95% CI). a, endurance training; b, resistance training program; c, endurance training program (moderate intensity leisure); d, endurance training (vigorous-intensity leisure); e, endurance training (active commuting by bike); f, endurance training (high intensity intermitted exercise training); g, endurance training (continuous moderate intensity training); DF, degrees of freedom; SMD, standardised mean difference; CI, confidence interval; Q, heterogeneity statistic to test homogeneity; I2, index of heterogeneity beyond within-study sampling error. Fig. 4. Forest plot of randomised controlled trials comparing the impact on PYY of interventions involving exercise training with non-exercise control group comparators. a, endurance training; b, resistance training; c, endurance training (moderate intensity leisure); d, endurance training (vigorous-intensity leisure); e, endurance training (active commuting by bike); f, endurance training (continuous moderate-intensity training); g, endurance training (high intensity intermitted exercise training); DF, degrees of freedom; SMD, standardised mean difference; CI, confidence interval; Q, heterogeneity statistic to test homogeneity; I2, index of heterogeneity beyond within-study sampling error. a duration of at least four weeks on fasting concentrations of gastrointestinal appetite hormones in healthy adults with overweight and obesity. The principal findings of this systematic review and metaanalysis were that exercise training programmes had no impact on fasting concentrations of gastrointestinal appetite hormones such as total ghrelin, acylated ghrelin and PYY in comparison to the control group. We also found that exercise training decreased body mass and BMI in comparison to the control group, though the relation between changes in body mass or BMI and fasting appetite hormones could not be fully explored due to the small number of studies included in this review. The results obtained should be interpreted with caution since our review identified only a few studies, and these were underpowered and carried a high risk of bias. Thus, this systematic review highlights the need for more well-reported RCTs.
Although our systematic review obtained only a few studies in individuals living with overweight and obesity, many studies examining the influence of acute and chronic exercise on appetite regulation have assessed ghrelin, given its relatively unique status as an identified orexigenic hormone (Dorling et al., 2018). Our results indicate that exercise training did not alter fasting concentrations of total ghrelin or acylated ghrelin in individuals with overweight and obesity. The acylated ghrelin results support a recent review showing that, despite some variability, exercise training does not influence acylated ghrelin in most settings (Ouerghi et al., 2021). Nonetheless, in contrast to our results, the same review also demonstrated that exercise increased total ghrelin compared to the control (Ouerghi et al., 2021). It is possible that differences in the populations studied are the reason for these variations. Indeed, the systematic review of Ouerghi et al. (2021) included studies which involved participants with diverse characteristics (e.g., sex, age, body mass and body composition, physical fitness, and health status) while participants of the studies considered in our systematic review were physically inactive adults of ≥18 years and a BMI ranging between 27.1 and 31.8 kg/m 2 . In addition, the systematic review of Ouerghi et al. (2021) applied physical exercise intervention alone or combined with other interventions.
Further to ghrelin, numerous anorexigenic hormones have been implicated in appetite regulation in humans (Karra & Batterham., 2010). In this regard, our meta-analysis found no overall effect on fasting concentrations of PYY, with very low substantial heterogeneity between studies that included seven exercise interventions. Though we could not perform a meta-analysis on other anorexigenic hormones, one study that examined CCK concentrations in response to exercise training in individuals living with overweight and obesity found no changes (Quist et al., 2019), and the studies that assessed GLP-1 yielded mixed findings, with one study finding an increase (Quist et al., 2019) and one study finding no change with exercise (Shakiba et al., 2019). Thus, the effect of exercise training on these gastrointestinal satiety hormones clearly requires further investigation in individuals living with overweight and obesity.
Evidence from single-arm exercise intervention studies indicates that training reduces the postprandial suppression of acylated ghrelin and exerts little influence on postprandial PYY, GLP-1 and CCK concentrations (Martins et al., 2010(Martins et al., , 2013); yet there are relatively few randomised controlled trials applying exercise training that have assessed gastrointestinal hormones in the postprandial state (Dorling et al., 2018). As a result of this and evidence indicating that exercise training Fig. 5. Forest plot of randomised controlled trials comparing the impact on BMI of interventions involving exercise training with non-exercise control group comparators. a, endurance training; b, endurance training(cycling); c, resistance training; d, combined training; e, endurance training (moderate to vigorous-intensity exercise); f, endurance training (moderate intensity leisure); g, endurance training (vigorous-intensity leisure); h, endurance (active commuting by bike); i, endurance training (moderate-intensity exercise); j, endurance training (high-intensity exercise); k, resistance training; l, endurance training (running); m, combined training; n, endurance training (continuous moderate-intensity training); o, endurance training (high intensity intermitted exercise training); DF = degrees of freedom; SMD = standardised mean difference; CI = confidence interval; Q = heterogeneity statistic to test homogeneity; I2 = index of heterogeneity beyond withinstudy sampling error. enhances the coupling between energy intake and exergy expenditure after food consumption (Blundell et al., 2015), additional controlled studies are needed to test how postprandial concentrations of gastrointestinal hormones are influenced by exercise training in individuals with overweight and obesity. Regardless, our systematic review suggests that any compensatory increase in energy intake due to exercise training is unlikely to be related to fasting gastrointestinal appetite hormone changes. At the same time, it is well established that during diet-induced weight loss, these hormones change the direction to favour increased appetite, energy intake and weight regain (Lean & Malkova, 2016;Thom et al., 2020). Thus, findings from this systematic review, combined with the evidence from body weight loss studies applying caloric restrictions, suggest that in individuals living with overweight and obesity sustainability of the achieved negative energy balance might be less challenging during exercise training than applications of caloric restriction.
It is possible that the collective tendency for studies to show nonsignificant average changes in appetite-related hormones is due to the small exercise-induced body mass loss observed in individuals with overweight and obesity. In line with most studies (Swift et al., 2014), we observed a small decrease in body mass and BMI in response to exercise, and while we could not explore the relationship between appetite-related hormones and body mass, previous studies have indicated that body mass and adiposity significantly influence appetite-related hormones (Lean & Malkova., 2016). Some studies included in this systematic review showed that greater exercise-induced body mass loss increased total ghrelin (Ahmadi et al., 2019;Foster--Schubert et al., 2005;Kang et al., 2018;Mason et al., 2015), though Shakiba et al. (2019) revealed that body mass was positively associated with circulating acylated ghrelin. Thus, further confirmatory work needs to test if greater body mass loss with exercise increases gastrointestinal hormone concentrations.
The finding of no impact of exercise training on mean changes of gastrointestinal appetite hormones and energy intake found in our systematic review might mask the presence of inter-individual variability. Heterogeneity in acylated ghrelin and PYY levels in response to acute exercise bouts has been reported (Goltz et al., 2019) and inter-individual variability in subjective appetite, energy intake and weight change with exercise training is well established (King et al., 2008). In addition, failure to find changes in gastrointestinal appetite hormones may be related to variations in methodological procedures, including differences in food intake during days leading to blood collection (Beaulieu et al., 2017), participant acclimatization to blood sampling procedures and differences in blood collection and processing methods (Chandarana et al., 2009). Given these methodological factors have a significant influence on fasting gastrointestinal hormone concentrations (Chandarana et al., 2009), standardized methodologies in the field should be devised and followed by exercise training studies in individuals with overweight and obesity to facilitate between-study comparisons (Dorling et al., 2018). Efforts could additionally be made to enhance the robustness of study designs by blinding researchers during sample processing and analysis where possible.
A key strength of this systematic review is that a comprehensive literature search strategy was used to identify all relevant research. This systematic review applied rigorous methodology in accordance with PRISMA guidelines (Page et al., 2021), including double screening of studies, data extraction, and assessment of the risk of bias. Further, the inclusion of data from randomised controlled trials provided a true estimate of intervention effects, and we investigated the impact of exercise Fig. 6. Forest plot of randomised controlled trials comparing the impact on the weight of interventions involving exercise training with non-exercise control group comparators. a, endurance training (cycling); b, resistance training; c, combined training; d, endurance training (moderate to vigorous intensity); e, endurance training (moderate intensity leisure); f, endurance training (vigorous-intensity leisure); g, endurance training (active commuting by bike); h, endurance training (moderate-intensity exercise); i, endurance training (high-intensity exercise); j, resistance training; k, endurance training; l, combined training; m, endurance training (continuous moderate-intensity training); n, endurance training (high intensity intermitted exercise training); DF = degrees of freedom; SMD = standardised mean difference; CI = confidence interval; Q = heterogeneity statistic to test homogeneity; I2 = index of heterogeneity beyond within-study sampling error. on a variety of appetite hormones in this review, rather than just focusing on one hormone. However, only a small number of studies (n = 9) met the inclusion criteria and the number of studies and interventions measuring appetite hormones other than acylated ghrelin was low. Therefore, neither a moderator nor a subgroup (e.g., sex, habitual physical activity, adiposity level, habitual diet) analysis could be conducted to examine the current study's data or assess the heterogeneity between effect estimates between the studies. Such analyses are important because inter-individual variations in gastrointestinal appetite hormones are evident and exploratory analyses have indicated that there are predictors of energy intake and body mass change during exercise training (Dorling et al., 2021;Hӧ;chsmann et al., 2020). Similarly, the numbers of participants in most studies and interventions were small, which suggests that these studies might be underpowered primarily to detect a significant difference between the groups. The restriction to the English language may be considered a limitation in this review.
In conclusion, this systematic review and meta-analysis indicates that exercise training programmes in individuals living with overweight and obesity have no impact on fasting concentrations of total and acylated ghrelin, PYY, GLP-1 and CCK, which suggests that any upregulation of appetite and energy intake, which is common during participation in exercise training, would be related to factors different from changes in fasting concentrations of gastrointestinal appetite hormones. While the low average body mass loss induced by exercise training may play a role in these findings, there were limited interventions of a high standard as well as inter-individual variability. As a result, further studies that test the influence of exercise training on appetite-related hormones, including responses in the postprandial state, are needed in individuals living with overweight and obesity.