The effect of resistance training interventions on weight status in youth: a meta-analysis

Background There has been a rise in research into obesity prevention and treatment programmes in youth, including the effectiveness of resistance-based exercise. The purpose of this meta-analysis was to examine the effect of resistance training interventions on weight status in youth. Methods Meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines and was registered on PROSPERO (registration number CRD42016038365). Eligible studies were from English language peer-reviewed published articles. Searches were conducted in seven databases between May 2016 and June 2017. Studies were included that examined the effect of resistance training on weight status in youth, with participants of school age (5–18 years). Results There were 24 complete sets of data from 18 controlled trials (CTs) which explored 8 outcomes related to weight status. Significant, small effect sizes were identified for body fat% (Hedges’ g = 0.215, 95% CI 0.059 to 0.371, P = 0.007) and skinfolds (Hedges’ g = 0.274, 95% CI 0.066 to 0.483, P = 0.01). Effect sizes were not significant for: body mass (Hedges’ g = 0.043, 95% CI − 0.103 to 0.189, P = 0.564), body mass index (Hedges’ g = 0.024, 95% CI − 0.205 to 0.253, P = 0.838), fat-free mass (Hedges’ g = 0.073, 95% CI − 0.169 to 0.316, P = 0.554), fat mass (Hedges’ g = 0.180, 95% CI − 0.090 to 0.451, P = 0.192), lean mass (Hedges’ g = 0.089, 95% CI − 0.122 to 0.301, P = 0.408) or waist circumference (Hedges’ g = 0.209, 95% CI − 0.075 to 0.494, P = 0.149). Conclusions The results of this meta-analysis suggest that an isolated resistance training intervention may have an effect on weight status in youth. Overall, more quality research should be undertaken to investigate the impact of resistance training in youth as it could have a role to play in the treatment and prevention of obesity.

Physical activity guidelines and position statements emphasise the importance of 'activity to strengthen muscle and bone' and research suggests that resistance training might have an impact on weight status in youth.
This meta-analysis found that resistance training has a positive effect on body fat percentage and skinfolds in youth. Further research is required to investigate the role resistance training may play in the treatment and prevention of obesity.

Background
Obesity is a worldwide concern. In 2016, more than 1.9 billion adults were classified as overweight, 13% were obese and 41 million children aged under 5 were overweight or obese [1]. Childhood obesity is a critical public health threat as the prevalence of obesity amongst youth continues to increase worldwide, and there is the risk of developing obesity-related diseases at an increasingly younger age [2]. Prevention and treatment programmes suitable for youth have been developed for which physical activity is an integral component [3].
The current guidelines for children aged 5-18 recommend 60 min of daily physical activity (PA), and minimising the time spent sitting each day. They also recommend activity that strengthens muscle and bone, at least 3 days a week [3,4]. However, despite these guidelines, one of the more recent global surveillance studies, the Health Behaviour in School-aged Children survey [5], reported that across Europe, less than 50% of young people were meeting the current PA guidelines. Additionally, the survey demonstrated a decline in PA levels with age; 25% of 11 years olds met the recommendations, compared to just 16% of 15 year olds [5] which indicates that as children advance to adolescence, sedentary behaviour becomes more common.
A growing volume of studies have now been published that seek to examine the effectiveness of PA interventions designed to combat these low levels [6]. However, in a systematic review of 57 studies that investigated PA interventions in children and adolescents, very few studies that were included examined interventions which addressed compliance with the muscle and bone component of the guidelines [6].
The benefits of resistance training (RT) in youth are well documented, and key organisations (NSCA, UKSCA and BASES) have developed position statements in support of this [7][8][9]. A benefit identified in these position statements is the positive effect of RT on weight status, although the evidence to support these statements is not strong. In fact, the majority of research investigates multi-component interventions with a focus on overweight/obese youths [10][11][12].
A systematic review in 2011 identified that very few randomised control trials have examined the effects of RT alone on body composition in overweight or obese adolescents [10]. From just seven studies included in this review that focused on RT alone, the authors reported inconclusive results and the recommendation was for more studies to be conducted that concentrate solely on RT interventions [10]. Also examining the effect of RT alone on overweight or obese children, a further review found only six studies met the inclusion criteria [11]. Three out of the six studies showed a significant decrease in percentage fat and a significant increase in fat free mass, although none of the studies found a decrease in total fat mass. Additionally, all studies reported an increase in body weight. Again, the findings were inconclusive, but they identified there was a lack of randomised control trials (only two studies) and the studies had low sample sizes [11]. In a systematic review and meta-analysis, it was found that interventions that included a RT component had a very small to small effect on body composition in overweight and obese youth [12]. Studies that included interventions that were RT alone showed similar effects to those studies that included resistance plus an aerobic or dietary component. However, across studies, there was large variation in intervention content and assessment methods [12].
The studies included in these reviews involved only overweight and obese participants and therefore it could be argued that they are focused on the treatment of obesity rather than prevention. However, it has been suggested that early prevention holds better than treatment with 'primordial prevention' being the first of three levels of obesity prevention [13]. This emphasises the prevention of risk factors to maintain a healthy weight throughout childhood and into the teens [13].
There has been one systematic review to date with emphasis on the impact of RT on the weight status of both healthy weight and overweight/obese youth [14]. This review included studies using a variety of assessment tools and also studies that combined aerobic training and dietary interventions, however, there were still only 12 studies included in total. Investigating the impact of RT on body composition, they identified that due to limitations in methodologies it was difficult to reach conclusions about the isolated role of RT [14].
Since the focus of the published reviews has been on obese and overweight youth, or on multicomponent interventions, the isolated role of RT on weight status in youth is not currently known. Further to this, the only review to focus on both healthy weight and overweight/obese youth to date is a decade old. Therefore, the purpose of this review was to systematically examine the impact of resistance training interventions on weight status in youth.

Methods
The search strategy and inclusion criteria were specified and documented in advance on PROSPERO (number CRD42016038365). The conduct and reporting of this review adhered to the guidelines outlined in the PRISMA statement [15]. The PRISMA flow diagram detailing the systematic search and included studies is shown in Fig. 1.

Search strategy
Electronic literature databases were searched from the year of their inception up to and including June 2017. These were PubMed, MEDLINE, ERIC, PsycINFO, Embase, Sport Discus and Scopus. Relevant references from published literature were followed up and included when they met the inclusion criteria and literature not identified in the electronic searches was sourced. ResearchGate was used to identify research papers written by key researchers in the field. Additionally, these researchers were contacted regarding any literature not yet published and the authors of this review searched their personal libraries.
The search terms were related to weight status, youths and RT (Table 1). The Boolean operator "AND" was used between categories, and the phrase "OR" was used within categories. The search strategy was adapted for each database, and searches were logged.
The screening and data extraction process was undertaken by HC and AD. Titles of potentially relevant articles were retrieved using the search strategy, duplicates then removed, and then the titles and abstracts were screened by HC. Ten percent (n = 418) of the titles and abstracts were also screened by AD. The inter-rater reliability for the two authors was found to be Kappa = 0.897 suggesting a strong level of agreement [16]. Full-text copies were obtained for potentially eligible articles and assessed for inclusion by HC and AD. During the review of full-text articles, a majority decision was taken in consultation with the other authors when disagreements regarding inclusion/exclusion occurred.

The inclusion/exclusion criteria
Studies with participants of school age between 5 and 18 years were included. No studies were included where the subject group were identified as having a pathological condition or disability which affects movement, (e.g. cerebral palsy/dyspraxia) and no studies were included where the subject group was identified as having a behavioural or neuropsychological condition (e.g. autism or ADHD).
To allow an isolated review of strength training, studies included RT methods but were excluded if they contained plyometric, vibration or neuromuscular training, or training specifically for rehabilitation purposes. Additionally, studies were excluded if the warm up and cool down activities were more than 10 min duration each to ensure there was not a significant aerobic component in the intervention. Studies targeting physical activity or diet were not included unless both the control group and intervention group received the same dietary  and physical activity intervention, but the intervention group received a different RT input. There was no restriction on location (e.g. school based or sports centre) or timing (e.g. during or after school).
Although studies were included that used a control group and also those that did not, for the purpose of this paper, the analysis focused on studies that included a control group, and therefore are referred to as controlled trials (CTs). Objective measures included a measure of weight status through body mass, body mass index (BMI), skinfolds, DEXA scans, BodPod, MRI scans, waist circumference, hip circumference or waist to hip ratio.
It is important to note that studies were only included when weight status was mentioned as an aim of the study in the abstract and not just as a demographic characteristic of the participants.

Data extraction
Data were extracted using an electronic form by HC and included study characteristics (e.g. country, year); participant characteristics (e.g., age, sex); intervention components (e.g. setting, content); and changes in the outcomes (e.g. change in weight status). The outcome data were extracted in the form of mean, standard deviation and sample size. If the outcome data was not visible within the paper, authors were contacted and asked to provide this data within a set time frame. To check reliability, AD carried out data extraction on the first 10% of the included studies, which were in alphabetical order of the first author. Following this, any disagreements were resolved through discussion with all authors.

Methodological quality and risk of bias assessment
The "Quality Assessment Tool for Quantitative Studies" developed by the Effective Public Health Practice Project in 1998 [17] was used to assess the quality and risk of bias of the included studies. The results of the assessment lead to an overall methodological rating of strong, moderate or weak in eight sections: selection bias, study design, confounders, blinding, data, collection methods, withdrawals and dropouts, intervention integrity and analysis. The assessment tool has been found to be valid and reliable [18]. To check reliability, AD carried out this assessment on 10% of the included studies and any disagreements were resolved through discussion between the two authors. Overall, the data extraction and risk of bias accuracy of one author was deemed to be acceptable.

Meta-analysis
Random effects meta-analyses were conducted with the Comprehensive Meta-analysis software (version 2.2.064). Hedges' g with randomised effects and 95% CIs were calculated for trials with sufficient data. The magnitude of Hedges' g was interpreted using Cohen's (1988) [19] convention as small (0.2), medium (0.5) and large (0.8). A significance level of p ≤ 0.05 was applied. Heterogeneity was assessed using the I 2 statistic. For interpretation, I 2 values of 25, 50 and 75 were considered to indicate low, moderate and high heterogeneity, respectively [20]. Publication bias was assessed by calculating Egger bias statistics [21] and Rosenthal's fail-safe N [22]. Corresponding funnel plots were created.
A moderator analysis was conducted to determine whether the intervention effects on the outcomes differed by sex of participants (males or females), sex of training group (i.e. the training group was designed for either males, females or mixed sex), weight status (healthy weight, overweight, obese or mixed weight status), age (< 12 or > 12 years, based on primary and secondary school age split), pubertal stage (<Tanner stage 2 or >Tanner stage 2, based on pre-pubertal and post-pubertal stages), location (school during P.E, school during free time or community), type of control (no resistance training, nutrition input only, normal activity, wait list) and quality of study (weak, moderate or strong). Additional moderator analyses were planned for ethnicity and supervised compared to self-regulated sessions. However, there was insufficient data to allow these analyses. Although data was also extracted for frequency and duration of interventions, a moderator analysis was not conducted on this data due to the inappropriateness of separating their independent and combined impact on training outcomes.
It is important to note that for outcomes where a decrease in score was a positive intervention effect (e.g. BMI) and an increase in score was a positive effect (e.g. fat free mass), this was accounted for in the analysis.

Results
Out of an initial 6290 studies identified through database searches, 18 studies met the inclusion criteria and included sufficient data for the meta-analysis. Twenty four data sets were included in the meta-analysis (some studies had more than one intervention group).

Study characteristics
The surveyed studies were conducted in eight different countries (USA, Australia, New Zealand, Hong Kong, Brazil, Tunisia, Austria and Japan). In total, there were 554 participants in the experimental groups (sample sizes ranged from 8 to 78 participants) and 599 participants in the control groups (sample sizes ranging from 5 to 129 participants).
The age of all participants ranged from 8 to 16 years. Eight outcomes related to weight status were included in the analysis: body mass (kg), BMI (kg/m 2 ), body fat (%), fat-free mass (kg), fat mass (kg), lean mass (kg), skinfolds (mm) and waist circumference (cm). The average attendance for the studies that reported it was 88%. The study details included in the analysis can be found in Table 2.
Based on the thresholds categorised [20], moderate to high heterogeneity was identified for BMI and waist circumference. For all other outcomes, heterogeneity was low.

Publication bias
To identify possible publication bias, effect sizes were plotted against standard errors to generate funnel plots as illustrated in Fig. 3. There was no indication of publication bias with no statistically significant result from Egger's test [21]. Rosenthal's fail-safe N [22] found that 241 additional studies would be needed for the cumulative effect to be non-significant. Therefore, it can be concluded that the meta-analysis provides a satisfactory representation of the effect of a RT intervention on weight status.

Quality appraisal
Through the quality appraisal process, 44% of the included studies were classified as 'strong' , 50% were classified as 'moderate' and 6% were classified as 'weak'.

Moderator analysis
The results (Table 3) suggest a moderation effect of weight status on BMI and waist circumference, with significant differences between healthy weight, mixed weight, overweight/obese and obese participants. The results suggest a moderation effect of quality score on waist circumference, with significant differences between strong and moderate studies.

Summary of evidence
The purpose of this meta-analysis was to examine the effect of resistance training (RT) interventions on weight status in youth. In summary, there was a small, statistically significant, effect of RT interventions on body fat % and skinfolds, but no overall significant effect on body mass, BMI, fat-free mass, fat mass, lean mass or waist circumference.
The UKSCA's [8] and NSCA's [7] position statements on youth RT both suggest that RT may have a positive impact on body fat and the significant findings of this meta-analysis for body fat % and skinfolds support these statements.
While more studies are required to provide further understanding of the mechanism underlying a reduction in body fat due to a RT intervention, it has been reported that a possible cause could be due to an increase in skeletal muscle mass and resulting increase in basal metabolic rate [23], in particular, this has been noted in adolescents [8]. However, as there were no overall significant effect sizes evident for lean mass or fat free mass in this analysis, this suggests that these changes could possibly be due to increases in overall energy expenditure that may have occurred by simply taking part in an active intervention, rather than an increase in metabolically active lean tissue. It is important to note, however, that with data from both pre-pubertal and post-pubertal participants included in the analysis, this may have had an impact on the results.
Although all of the studies that measured fat mass were in favour of the intervention, this result was not significant and the effect size was trivial. It is worth considering why the effects were not consistent across different measures of body fat. Although body scans have been used in the included studies to measure both body fat % and fat mass, studies that have used skinfolds to measure body fat % have also been included in the analysis. DEXA scans measure both subcutaneous and visceral fat; however, skinfolds measure subcutaneous fat only. As there have been variable findings with regards to the impact of training on different locations of fat [24][25][26][27][28], this could explain why there were significant findings for body fat % and skinfolds, but not for fat mass. This emphasises the importance of validity when selecting measurement tools and methods used to assess intervention outcomes indicative of body fat. An additional factor to consider in this meta-analysis is that there was more than double the number of studies that investigated body fat % as an outcome in comparison to fat mass, providing greater statistical power to the findings for body fat %.
Resistance training did not demonstrate a significant effect on body mass or BMI; however, due to growth,  these results should be interpreted with caution. Since the studies included in the analysis assessed a combination of healthy weight, overweight and obese participants, changes in these measures over time are likely to be variable and this is explored further in the moderator analysis discussed below. For the purpose of this analysis, a decrease in body mass and BMI was analysed as a favourable change. However, with the interventions being resistance based, a subsequent increase in lean mass/ fat-free mass (and therefore body mass) was possible, and this may have obscured the findings with an overall positive trend for RT impacting on lean mass and fat free mass in this analysis. It has been identified that there is mixed evidence with regards to whether youth may experience increases in muscle mass following RT, most likely due to inadequate levels of circulating testosterone [7] although it has been suggested that resistance training may develop lean body mass in adolescents [8]. This will have been exacerbated by the inclusion of youth from 8 to 16 years of age of varying pubertal status and by varying intervention duration. It has also been suggested that periods of training in excess of 10 weeks are required for increases in lean muscle mass to occur [10]. In this review, for lean mass, 67% of the data sets were from studies that included interventions that were > 10 weeks, and for fat-free mass, these interventions were > 10 weeks for only 50% of the studies. This suggests that the intervention duration for several of the studies may not have been long enough to invoke positive measurable changes. For the outcome of waist circumference, there was no significant effect size evident. With a combination of healthy weight, overweight and obese participants, it might be expected that those studies including overweight/obese participants would show a larger effect on waist circumference than the studies that included healthy weight participants, and this is further explained by the moderator analysis discussed below. There were two studies that included obese participants only; one finding a large effect of the intervention on waist circumference [29] and the other finding no significant effect [30]. In the 2004 study, the authors did identify that with only a 6-week intervention, this may not have been long enough to have a positive impact on the measured outcomes. Additionally, only six studies investigated waist circumference as an outcome, and this outcome therefore had less statistical power than some of the others.

Previous reviews
Overall, these findings are similar to previous reviews. In a meta-analysis published in 2013 [12], there was a significant effect of interventions including a RT component on body fat % in overweight or obese youths. They also reported no significant effect sizes for body mass, BMI waist circumference, fat mass or lean mass. Out of the 40 studies included, only 9 studies were RT-only studies, and out of these 9, 6 were CTs. However, although interventions were included that also incorporated an aerobic and/or dietary component, similar effects were found for studies that included interventions that were RT alone when authors performed a sub-analysis. There were only three studies included that incorporated RT only and therefore interpreting the data should be undertaken with caution.
In a further systematic review [11], only six studies were included that investigated RT only interventions in overweight/obese youth. Three out of the six studies showed a significant decrease in percentage fat and a significant increase in fat-free mass, although none of the studies found a decrease in total fat mass, which is in support of the findings of this review. Four studies reported significant changes in body composition, with an increase in fat-free mass and BMI and additionally, all studies reported an increase in body weight. This is conflicting with the current review, although their review  only included overweight and obese youth. Unlike the studies in the current analysis, all studies in their review included RT of moderate to submaximal intensity during treatment, rather than high intensity work and while acknowledging the limitations of using percentage of 1RM to prescribe intensity, higher intensity work (however calculated) will provide a greater stimulus for overload than low-medium intensity work. This may have had an impact on their overall findings with regards to weight status, as high intensity work has been reported as a key component to elicit changes in body composition [10].
Supporting the variable findings with regards to the impact of RT on obese adolescents, inconclusive results were found from just seven studies included that focused on RT alone [10].
In the only systematic review to date that included both healthy weight and overweight/obese participants, 12 studies were included [14]. It was reported that for CTs, no studies found a significant change in BMI body fat, two studies reported a change in lean body mass and one study that reported a change in waist circumference. In their review, there were only five RCTs and three of these included a dietary and/or aerobic component. The interventions were also different to the current review with all but two of the RT interventions being circuit based.
Overall, with just one of the reviews described above including a meta-analysis, it makes comparisons with the current meta-analysis somewhat challenging.

Moderator analysis
To investigate the findings further, a moderator analysis was completed on all outcomes to identify if any effects could be explained by specific moderator variables. It was found that weight status was a moderator for BMI and waist circumference, and these outcomes showed moderate and high heterogeneity respectively and therefore the variance between studies could be explained by weight status. All other outcomes did not display significant heterogeneity, and no significant findings were apparent from the moderator analysis (Table 2).
There was a significant effect of the intervention on the BMI and waist circumference of obese participants ( Table 2) but not on other weight categories, indicating that RT could be an effective intervention on these outcomes in obese individuals. It would seem plausible that BMI and waist circumference outcomes would vary significantly across studies due to the inclusion of both healthy weight and overweight/obese participants. It has been reported that obese youth are more sedentary than their healthy weight peers [31] and require more energy to move [32]. Therefore, an increase in physical activity might have a larger relative increase in energy expenditure reflected in reduced BMI and waist circumference.
It should be noted, however, that for waist circumference, there were only six studies included in the analysis. Three were mixed weight with one study each for obese, overweight and healthy weight participants. This small number of studies may explain the high heterogeneity, and therefore interpreting the results should be undertaken cautiously. Additionally, a longer term follow up study would be beneficial to investigate resistance training as an obesity prevention method.
These findings conflict with previous findings [12] that suggested that there was a very small moderation effect of age and sex on various weight status outcomes. It was reported that for youth 12 years or older, there was an intervention effect on body mass, BMI, waist circumference, body fat %, fat mass and lean mass, and for males, there was an intervention effect on body mass, BMI, body fat % and fat mass. However, these were small influences on intervention effects, and their analysis included studies that incorporated an aerobic or dietary component which is different to the analysis in the current review and therefore it is difficult to make conclusions regarding RT alone based on their findings.

Strengths and limitations
There were a number of strengths to this review. There should be strong confidence in the main findings given the rigorous review process. A strict inclusion/exclusion criteria resulted in an analysis of 24 data sets that examined the effects of RT on weight status in 554 youths from 8 countries.
This is also the first meta-analysis to include healthy weight and overweight/obese participants taking part in RT only interventions, which is important to identify the impact of RT not only as a treatment for obesity but also as a prevention.
There was high compliance reported in the included studies. For the studies who reported it, compliance was 88%. As well as a strength of the current meta-analysis, high compliance adds substance to the potential for RT as a viable mode of intervention to improve weight status.
There are however limitations that need to be considered when interpreting the results. There was large variability within the study interventions with regards to participant numbers (ranging from 5 to 129 participants), frequency, duration and programme content. The frequency ranged from 2 to 6 times a week and duration ranged from 8 to 20 weeks. Programmes also involved a mixture of sets and reps with a range of intensities and some being circuit based. The forest plots also indicate large variation with the individual studies' results.
For certain outcomes there were a variety of different methods of measurement. For example, body fat % was measured by DEXA, BodPod, bioelectrical impedance scales, skinfolds and MRI scanning. While reporting standardised mean differences allowed us to pool the data for the purposes of the meta-analysis, differences between measurement tools should be acknowledged.
A limitation of the moderator analysis was not all of the studies reported data to enable a thorough investigation, so limited conclusions can be made based on this additional, incomplete level of analysis.
Finally, there was a mixture of quality of the studies included, with only 44% of the studies classified as 'strong'. When moderating for quality of studies, it is difficult to make conclusions despite there being significant findings for waist circumference, as there was only one 'strong' study for this outcome. This had a large effect size in comparison to the 'moderate' and 'weak' studies.

Conclusions
The literature suggests that RT may have a positive effect on weight status in youth, although the effects have not been clearly established and it is clear that more quality research is required to investigate this further. This meta-analysis provides an overview of the current research evidence and an insight into the potential benefits of such interventions. It adds to previous systematic reviews by including both healthy weight and overweight/obese youths and including RT only interventions which explores not only the treatment of obesity, but also prevention which is a vital component in combatting rising levels of obesity in children.
Overall, this meta-analysis found a small, statistically significant effect of RT interventions on body fat % and skinfolds, but no overall significant effect on body mass, BMI, fat free mass, fat mass, lean mass or waist circumference. While we can conclude that RT interventions have a small positive impact on some indicators of weight status, it is noted that this reflects only a small body of published work.
With RT interventions offering potential benefits for youth with regards to weight status, it is imperative that more robust and quality studies should be conducted to further, and unequivocally, investigate the role RT may play in the treatment and prevention of obesity. Based on the findings of this meta-analysis, and in support of the conclusions of previous reviews, future studies should be designed as randomised controlled trials with large samples and include a treatment group with an isolated RT intervention. There should be careful consideration into appropriate intervention content and assessment methods. If validated, this type of intervention, as recommended by the UK and WHO physical activity guidelines, could ultimately have a positive impact on the prevalence of common obesity-related diseases, such as type 2 diabetes, cardiovascular disease and cancer, and improve the health of individuals not only during childhood but as they progress through life.