Effects of exercise training on obesity‐related parameters in people with intellectual disabilities: systematic review and meta‐analysis

Abstract Background Efforts to synthesise existing knowledge concerning the effects of exercise interventions on obesity (i.e. changes in body weight and composition) have been made, but scientific evidence in this matter is still limited. This systematic review and meta‐analysis aims to identify and critically analyse the best available evidence regarding the use of physical exercise as a strategy to attenuate obesity through its effects on adiposity‐related anthropometric parameters in people with intellectual disability (ID). Methods Following the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses guidelines, a literature search was performed using PubMed, Scopus, SPORTDiscus, CINAHL and the Cochrane Library through specific keywords up to July 2020. The search adhered to the population, intervention, comparison and outcome strategy. Randomised controlled trials addressing the effects of the exercise intervention on adiposity‐related anthropometric parameters (body mass index, waist circumference, waist–hip ratio, fat percentage or body weight) in children, adolescents and adults with ID were included. The methodological quality of the studies found was evaluated through the PEDro scale. Results A total of nine investigations with children and/or adolescents (10–19 years) and 10 investigations with adults (18–70 years) were selected, mostly experiencing mild and moderate ID. Methodological quality was fair in 13 of these publications, good in five and excellent in one. Seventeen trials reported comparable baseline and post‐intervention data for the intervention and control groups and were included in the meta‐analysis. In nine studies, the intervention group performed a cardiovascular training programme. Five papers described a combined training programme. Two trials executed whole‐body vibration training programmes, and one publication proposed balance training as the primary intervention. According to the meta‐analysis results, the reviewed studies proposed exercise modalities that, in comparison with the activities performed by the participants' in the respective control groups, did not have a greater impact on the variables assessed. Conclusions While physical exercise can contribute to adiposity‐related anthropometric parameters in people with mild and moderate ID, these findings show that exercise alone is not sufficient to manage obesity in this population. Multicomponent interventions appear to be the best choice when they incorporate dietary deficit, physical activity increase and behaviour change strategies. Finding the most effective modality of physical exercise can only aid weight loss interventions. Future research would benefit from comparing the effects of different exercise modalities within the framework of a multicomponent weight management intervention.

intervention, comparison and outcome strategy. Randomised controlled trials addressing the effects of the exercise intervention on adiposity-related anthropometric parameters (body mass index, waist circumference, waist-hip ratio, fat percentage or body weight) in children, adolescents and adults with ID were included. The methodological quality of the studies found was evaluated through the PEDro scale. Results A total of nine investigations with children and/or adolescents (10-19 years) and 10 investigations with adults (18-70 years) were selected, mostly experiencing mild and moderate ID. Methodological quality was fair in 13 of these publications, good in five and excellent in one. Seventeen trials reported comparable baseline and post-intervention data for the intervention and control groups and were included in the meta-analysis. In nine studies, the intervention group performed a cardiovascular training programme. Five papers described a combined training programme. Two trials executed whole-body vibration training programmes, and one publication proposed balance training as the primary intervention. According to the meta-analysis results, the

Introduction
Overweight and obesity, defined as a body mass index (BMI) over 25 or 30 kg/m 2 , respectively, are an excessive fat accumulation (i.e. adiposity) that is associated with poor health outcomes, including an increased risk of incidence and mortality from cancers (Parra-Soto et al. 2021) and cardiovascular diseases (Dwivedi et al. 2020). Because of the important health burden and the economic costs to the healthcare systems worldwide derived from obesity, it is a health condition that needs treatment and, more importantly, prevention (Meldrum et al. 2017).
Excessive adiposity is largely due to a chronic imbalance between energy intake and energy expenditure favouring the former. Strategies directed at reducing its prevalence should have as a priority to create a negative energy balance. To this aim, exercise promotion plays a key role (Petridou et al. 2019). Indeed, exercise can help modify several adiposity-related anthropometric parameters linked to obesity (i.e. BMI, fat mass and waist circumference) associated with morbidity and mortality (Huxley et al. 2010). Internationally recognised guidelines support increasing physical activity (PA) as an integral part of a lifestyle intervention for weight loss, but that this must be combined with a dietary change that produces a calorie deficit and with behaviour change strategies (Scottish Intercollegiate Guidelines Network 2010; National Institute for Health and Care Excellence 2014a,b).
Consequently, there are currently different systematic reviews and meta-analyses to provide scientific evidence about the benefits of exercise on obese people and the training modalities that are more suitable for them. These pieces of research are generally focused on children (García-Hermoso et al. 2018), adults (Kim et al. 2019) and elderly persons (Martínez-Amat et al. 2018) with obesity or in individuals with co-morbidities related to this condition (Ostman et al. 2017). However, research of this kind around people with intellectual disability (ID) has been sparse. This is a population often neglected, despite being at high risk of developing obesity and associated chronic health conditions (Hsieh et al. 2014).
Compared with the general population, the prevalence of obesity is much higher in persons with ID, in both adults (Doody & Doody 2012) and youth (Hinckson et al. 2013). Obesity leads to a significant contribution to the reduced life expectancy and increased health needs in these individuals . Therefore, there is a need for actions to reduce its prevalence. Obesity has multifactorial causes and decreased resting energy expenditure, and low PA levels have been shown to be determinant factors in those with ID (Bertapelli et al. 2016). Accordingly, physical exercise performance takes a prominent role among the lifestyle modifiable factors that can contribute to reducing obesity in this population (Spanos et al. 2013).
Efforts to synthesise existing knowledge concerning the effects of exercise interventions on obesity in people with ID have been made (Conrad & Knowlden 2020). Hamilton et al. (2007) reviewed the scientific evidence on the effectiveness of interventions for obesity in ID and indicated that PA interventions proved to be effective in the short term and should be considered an essential component of a weight loss programme. However, they specifically focused on adults with ID. In addition, most of the studies found had significant methodological problems that prevented them from extracting solid conclusions. Similarly, Maïano et al. (2014) reviewed the effectiveness of lifestyle interventions targeting changes in body weight and found that PA programmes were somewhat effective in provoking changes in body weight and composition. The authors focused on young men with ID, and they also observed serious limitations in the reviewed studies, such as small sample sizes or methodological flaws that led to inconclusive findings. Another systematic review and meta-analysis of randomised controlled trials (RCTs) by Harris et al. (2015) indicated that PA interventions did not significantly change body weight or BMI on young adults with ID, and its effects on other measures of body composition were inconsistent. The generalisability of their findings was limited because of the heterogeneity of the sample population in age ranges and the fact that no comparison on the effects of different PA programmes could be performed. Indeed, the authors highlighted that published RCTs were inadequate to form firm conclusions. More recently, the systematic review and meta-analysis by Kapsal et al. (2019) showed that PA generally had a positive impact on physical and psychological outcomes, but not in the case of BMI. However, certain methodological aspects such as excluding adults from their analysis and an insufficient number of RCTs that led to a considerable risk of bias limited the strength of the obtained findings. Finally, Casey & Rasmussen (2013) did review the literature on the effects of exercise on both young and adults with ID. They found that although exercise interventions appear effective at maintaining fat levels, they seemed to be particularly ineffective at reducing body fat percentage. These results were limited because the authors conducted a scoping review; thus, there was a lack of uniformity in study design and measurement, and no quality appraisal was performed. Besides, body fat was the only adiposity-related outcome analysed.
Another important aspect to highlight is that BMI is an indirect marker of adiposity that does not consider body composition (i.e. fat and lean mass). It has been reported that anthropometric proxies for central obesity, such as waist circumference, are better predictors of health risks (Janssen et al. 2004;Schneider et al. 2010). Furthermore, it has been suggested a higher risk of mortality for high waist circumference compared with high BMI (Yusuf et al. 2005;Pischon et al. 2008;Staiano et al. 2012). The body fat percentage, a direct measure of body fat, is associated with coronary heart disease (Lavie et al. 2009), a leading cause of death in developed countries (World Health Organization 2020), and reflects better the serum lipid profile and related metabolic alterations than the anthropometric surrogates (Nagaya et al. 1999). In spite of this, no previous systematic reviews have performed a meta-analysis comparing the effects of different types of physical exercise on these different adiposity-related parameters in people with ID.
Altogether, it seems that no systematic review and meta-analysis of RCTs on the effects of PA when performed as a sole intervention on obesity in both children and adults with ID and that also examines the impact of different exercise modalities has been published so far. Therefore, this review aims to answer the following research questions: • Is physical exercise an accurate strategy to attenuate obesity through its effects on adiposity-related anthropometric parameters in people with ID? • What is the methodological quality of the studies that have addressed this topic? • Is there any physical exercise modality that should be specifically prescribed for this aim in this population?

Methods
The systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (Moher et al. 2009). The review protocol was registered at the PROSPERO database (registration number CRD42020162080).

Search strategy
A systematic search of randomised controlled trials (RCT) that examined the effects of different PA interventions on adiposity-related anthropometric parameters in people with ID until July 2019 was performed. Articles from PubMed, Scopus, SPORTDiscus, CINAHL and the Cochrane Library were identified. The search adhered to the population, intervention, comparison and outcome strategy. Only terms regarding the population and the intervention, in a combination of standardised MeSH and free-text terms, according to recommendations from the Cochrane Handbook for Systematic Reviews of Interventions (Higgins & Green 2008), were used. Consequently, the search syntax was developed as follows: (exercise OR 'physical activity' OR 'physical exercise') AND ('intellectual disability' OR 'mental retardation' OR idiocy).

Eligibility criteria and study selection
The inclusion criteria were as follows: (1) RCT based on exercise or PA programmes, (2) sample entirely made up of individuals with ID and (3) studies analysing the effects of the intervention on adiposity-related anthropometric parameters (BMI, waist circumference, waist-hip ratio, fat percentage or body weight). Investigations that carried out interventions combined with other types of non-exercise therapies or that were merely based on the promotion of PA, as well as research not published in English, French, Portuguese or Spanish in peer-reviewed journals, were excluded. Titles and abstracts of search results for key criteria were screened, and then full-text potentially relevant articles for inclusion were assessed. Also, the reference lists of relevant studies for any additional suitable publications were screened. Two authors (M. A. S.-L. and J. S.-B.) independently assessed eligibility, with discrepancies resolved through discussion with a third author (C. A. P.).

Data extraction
One researcher (J. S.-B.) extracted information using a predeveloped form comprising participants' characteristics, exercise programmes, outcomes and measurement tools, and main findings from the original works, and two investigators (M. A. S.-L. and C. A. P.) confirmed the accuracy of data extraction. The data extraction procedure was not blind, as the names of the authors of the selected studies and the title of the journals in which they were published were identifiable. Missing data were obtained by contacting the study authors whenever possible.

Methodological quality assessment
The information on the quality of RCTs was retrieved directly from PEDro, the Physiotherapy Evidence Database. If a trial was not included in PEDro, two authors (M. A. S.-L. and J. S.-B.) appraised its quality using the PEDro scale. A third author (C. A. P.) provided advice to reach a consensus in case of disagreement. The PEDro scale rates internal study validity and the presence of replicable statistical information on a scale from 0 to 10. The suggested cut-off points to categorise studies by quality were excellent (9-10), good (6-8), fair (4-5) and poor (≤3) (Silverman et al. 2012).

Data analysis
A meta-analysis to measure post-intervention changes between exercise intervention groups and control groups unassigned to any exercise programmes was used. The standardised mean differences (SMDs) and their 95% confidence intervals (CIs) were calculated to assess the change in each outcome. The SMD was computed using intervention and control group sample sizes, baseline and post-intervention mean, and standard deviation (SD) for each selected outcome measure. To obtain the change from baseline and post-intervention SD, the following formula was used: SD change = √(SD 2 baseline + SD 2 post-intervention ) À (2 × Corr × SD 2 baseline × SD 2 post-intervention ), where Corr = 0.5. The value for Corr was imputed on the assumption of moderate correlation between baseline and post-intervention measurements (Gates et al. 2013;Northey et al. 2018). Multiple publications from the same trial were identified in order to avoid double counting the same sample of participants (Senn 2009).
Both fixed-effect and random-effect models were applied to determine the pooled effects, considering the first if I 2 heterogeneity was <30% or the second if higher (DerSimonian & Laird 1986). Forest plots displaying SMDs and 95% CIs to compare the effects between the intervention and control groups were utilised. The SMDs were considered as significant when their 95% CIs excluded zero, while pooled SMD values were evaluated as small (≤0.2), medium (0.2-0.8) or large effects (≥0.8) (Cohen 1992).
All statistical analyses were performed using STATA (version 13; StataCorp LLC, College Station, TX). Figure 1 provides a full depiction of the systematic review process. A total of 4015 records from the database search were obtained. After excluding duplicates, 2704 records were identified. Titles and abstracts were screened, and 248 studies for full-text assessment were retrieved. Finally, 19 RCTs met the full inclusion criteria and were included in the systematic review. Of the total, 17 trials reported comparable baseline and post-intervention data for both the intervention and control groups and were included in the meta-analysis. From these, three publications were derived from the same sample (Ordoñez et al. 2013;Ordonez et al. 2014;Rosety-Rodriguez et al. 2014).

Design and samples
A total of nine studies included children and/or adolescents (10-19 years), and 10 investigations included adults (18-70 years).
Five papers described a combined training programme. Rimmer et al. (2004)  Two studies executed whole-body vibration training programmes (González-Agüero et al. 2013;Matute-Llorente et al. 2015), and one investigation proposed balance training as the primary intervention (Mikolajczyk & Jankowicz-Szymanska 2015). These three trials were included in the meta-analysis as 'other interventions'.
Regarding exercise intensity prescription, four publications used the participant's maximum heart rate (HR max ), ranging from 50% to 70% HR max (Ordoñez et al. 2013;Ordonez et al. 2014;Rosety-Rodriguez et al. 2014;Kim 2017); two papers prescribed exercise intensity based on maximum oxygen consumption (V O₂ max ), ranging from 50% to 80% V O₂ max (Rimmer et al. 2004;Boer & Moss 2016); and one RCT opted for the ventilatory anaerobic threshold increasing from 90% up to 110% (Calders et al. 2011). The one-repetition maximum was used to prescribe the intensity of the strength training programme in two studies (Rimmer et al. 2004;Calders et al. 2011).

Clinical significance of the results
A total of 14 studies reported body weight measurements. Of those, only three studies (Lin & Wuang 2012;Boer & Moss 2016;Kim 2017) reported an average weight loss superior to À3% (ranging between 3.2% and 6.6%), which is considered of clinical relevance (National Institute for Health and Care Excellence 2014b).

Sensitivity analysis
The results of the sensitivity analysis, including only aerobic exercise programmes in children or adolescents and adults, exhibited significant improvements for BMI when the analysis was not stratified by age groups (Fig. S1). A trend towards improvement was generally found for all variables (weight, BMI, waist circumference and body fat, Fig. S2).

Discussion
This systematic review aimed at summarising, critically evaluating and integrating the available scientific evidence regarding the impact of exercise on a number of obesity-related parameters related in children, adolescents and adults with ID. A considerable number of RCTs ranging from fair to excellent methodological quality were found, and the meta-analysis included 85% of these trials. Thus, the information provided here can help guide best practices among those therapists and researchers interested in the possibilities and potential impact of prescribing exercise as a weight loss strategy for people with ID.
According to the meta-analysis results, the reviewed studies proposed exercise modalities that, in comparison with the activities performed by the participants in the respective control groups, did not have a greater impact on the variables assessed. Given that the training programmes were generally well designed, and exercise adherence was usually high, our results would imply that exercise alone does not lead to significant anthropometric and body weight parameter changes. This finding is in line with previous observations indicating that exercise often does not produce significant morphological 435 Figure 3. Meta-analysis for the body mass index by age and type of exercise programme. CG, control group; CI, confidence interval; IG, intervention group; SMD, standardised mean difference. [Colour figure can be viewed at wileyonlinelibrary.com] changes in people with ID Spanos et al. 2013;Harris et al. 2015;Martínez-Aldao et al. 2019). Moreover, these results reinforce the idea that interventions targeting obesity should focus on reducing energy intake rather than solely on exerciseinduced energy expenditure (Westerterp 2019).
Nevertheless, it is worth mentioning that the sensitivity analysis showed a significant trend revealing that aerobic exercise might contribute to significant changes in obesity-related parameters. Indeed, when all the investigations that utilised this exercise modality were analysed together (without considering the participants' age), the results were statistically significant. Thus, it is plausible to think that if the reviewed trials had a greater sample size, the meta-analysis would have detected a significant impact of aerobic exercise on anthropometric and body composition parameters. Indeed, this exercise modality proves to be central for body weight management among overweight and obese adults (Ismail et al. 2012).
The idea of which exercise modality could have a greater influence on adiposity-related anthropometric parameters cannot be elaborated further, because only three investigations comparing different exercise modalities were found. From the obtained data, it seems that aerobic interval training might have certain advantages over continuous aerobic training, while combined endurance and muscular training do not exhibit a greater effect than aerobic exercise alone. This is supported by similar findings from previous works evaluating other populations (Monteiro et al. 2015;Wewege et al. 2017).
In this piece of research, publications including both adult and young populations mostly with mild to moderate ID were reviewed. Although ID causes were not always reported, many investigations focused on young people with DS were found. This is a fact that deserves some appreciation because this group of individuals presents a high obesity prevalence (O'Shea et al. 2018); but yet, based on previous findings (Bertapelli et al. 2016) and also to our results, exercise-based programmes appear to be insufficient to achieve weight or fat loss.
Recent evidence has identified the difficulty of procuring an effective strategy to reduce obesity among people with ID. For instance, Harris et al. (2018) reported that current multicomponent weight management interventions were not more effective than no treatment in this population. The authors suggested that this lack of effect was due to the interventions did not adhere to clinical recommendations regarding diet, exercise and behaviour change techniques. However, they did show that multicomponent interventions that specifically included an energy-deficient diet were effective. Similarly, Ptomey et al. (2018) demonstrated that a well-controlled and designed intervention combining diet, PA and counselling led to significant weight changes in obese people with ID. Altogether, these results suggest that multicomponent interventions are a more effective approach than prescribing exercise alone for obesity management in people with ID.
The main strength of this review lies in its ability to build upon currently existing revisions and meta-analysis of the RCTs regarding the impacts of exercise on anthropometric and body composition parameters in people with ID, in order to reach more robust conclusions. Nevertheless, some methodological weaknesses must be recognised. First, various reviewed papers did not inform the aetiology and severity of ID. Also, the low number of studies included in the meta-analyses did not allow a stratified analysis by ID aetiology. Second, most of the studies had samples wholly made up of DS participants. Although this finding was somehow expected, because people with DS are one of the most at-risk groups within the ID population for weight gain and obesity, it shows the need for further studies including different diagnoses within the ID population. Third, the different procedures to measure adiposity across the included studies limit the quality of the metaanalysis. Fourthly, investigations with two exercise-based programmes were not included in the meta-analysis. Hence, proper advice about the benefits of a specific exercise modality compared with other exercise options could not be given. Fifthly, because of the reduced number of studies included per meta-analysis, a moderator analysis using 437 Figure 5. Meta-analysis for the waist circumference by age and type of exercise programme. CG, control group; CI, confidence interval; IG, intervention group; SMD, standardised mean difference. [Colour figure can be viewed at wileyonlinelibrary.com] meta-regression could not be performed, which would reduce the risk of type I error (Gordon et al. 2018). Sixthly, the analysed evidence might be incomplete because of language restrictions and because the grey literature was not reviewed. Finally, another limitation is the potential selective publication in the scientific literature (publication bias).

Conclusions
While physical exercise can contribute to adiposity-related anthropometric parameters in people with mild and moderate ID, these findings show that exercise alone is not sufficient to manage obesity in this population. Multicomponent interventions appear to be the best choice when they incorporate dietary deficit, PA increase and behaviour change strategies. Finding the most effective modality of physical exercise can only aid weight loss interventions. There is a need for future research aimed at comparing the impacts of different exercise modalities within the framework of a multicomponent intervention.