9 10 12 14 Research Unit16 - CORE · This is a post-print version of the following article: Brown,...

This is a post-print version of the following article: Brown, Helen Elizabeth, Pearson, Natalie, Braithwaite, Rock E., Brown, Wendy J. and Biddle, Stuart J. H. (2013) Physical activity interventions and depression in children and adolescents: A systematic review and meta-analysis. Sports Medicine, 43 3: 195-206.

Physical activity interventions and depression in children and adolescents: a systematic review and 1

meta-analysis 2

Running title: PA interventions and depression 3 4 5 6 7 Authors: Helen E. Brown1, Natalie Pearson2, 4, Rock E. Braithwaite3, Wendy J. Brown1 and Stuart J. 8

H. Biddle1, 2, 4 9

1School of Human Movement Studies, The University of Queensland, Brisbane, QLD, Australia. 10

2School of Sport, Exercise & Health Sciences, Loughborough University, Loughborough, 11

Leicestershire LE11 3TU, UK. 12

3Department of Kinesiology and Recreation Administration, Humboldt State University, Arcata, CA, 13

United States. 14

4 The NIHR Leicester-Loughborough Diet, Lifestyle and Physical Activity Biomedical 15 Research Unit 16 17 18 19 Article type: Systematic review/meta-analysis 20 21 Keywords (not included in title): mental health, youth, exercise, programs 22 23 Abstract word count: 285 words 24

Manuscript word count: 2843 words (inclusive of all pages except the abstract and title page) 25

26

27

28


Date of manuscript submission: 1/3/12 29

30

31

32

Abstract 33

Context: Evidence suggests chronic physical activity (PA) participation may be both protective 34

against the onset of and beneficial for reducing depressive symptoms. 35

Objective: To assess the impact of PA interventions on depression in children and adolescents using 36

meta-analysis. 37

Data sources: Published English language studies were located from manual and computerised 38

searches of the following databases: PsycInfo, The Cochrane Database of Systematic Reviews and 39

The Cochrane Central Register of Controlled Trials, Trials Register of Promoting Health Interventions 40

(TRoPHI; EPPI Centre), Web of Science and MEDLINE. 41

Study selection: Studies meeting inclusion criteria (1) reported on interventions to promote or 42

increase PA; (2) included children aged 5-11 years and/or adolescents aged 12-19 years; (3) reported 43

on results using a quantitative measure of depression; (4) included a non-physical control or 44

comparison group; (5) were published in peer-reviewed journals written in English, up to and 45

including May 2011. 46

Data extraction: Studies were coded for methodological, participant, and study characteristics. 47

Comprehensive Meta-Analysis version-2 software was used to compute effect sizes; with sub-group 48


analyses to identify moderating characteristics. Study quality was assessed using the Delphi 49

technique. 50

Results: Nine studies were included (N= 581); most were school-based RCTs randomized by 51

individual. Studies used a variety of measurement tools to assess depressive symptoms. The 52

summary treatment effect was small but significant (Hedges’ g=-0.26, SE=0.09, 95% C.I = -0.43, -0.08, 53

p=0.004). Sub-group analyses showed that methodological (e.g. studies with both education and PA 54

intervention; those with a higher quality score; and less than three months in duration) and 55

participant characteristics (e.g. single gender studies; those targeting overweight or obese groups) 56

contributed most to the reduction in depression. 57

Conclusions: There was a small significant overall effect for PA on depression. More outcome-58

focused, high quality trials are required to effectively inform the implementation of programs to 59

reduce depressive symptoms in children and adolescents. 60

61

62

63

64

65

66

67


68

69

70

71

72

73

74

75

76

77

Introduction 78

Current epidemiological data suggest a global prevalence of child and adolescent mental health 79

disorders of approximately 20%(1) of which depression is the most frequently diagnosed(2). 80

Depression during youth occurs frequently, is chronic and recurrent, and is associated with 81

morbidity and mortality(3). Depression in children and adolescents can also result in substantial 82

impairment in social functioning (such as relationships with peers and family members)(4), cognitive 83

development and scholastic achievement(5) and associated suicide is reported to be the 3rd leading 84

cause of death among adolescents(6). 85


Psychotherapy and pharmacological intervention are suggested as treatment options for depression 86

in children and adolescents(3) but are often reported to be ineffective(7). Cognitive behavioural 87

therapy can be of use but may be inaccessible and expensive(8). There is, however, growing interest 88

in the use of physical activity (PA) interventions for the management of depression. Building on 89

reviews that cite positive outcomes for the prevention and treatment of depression in adult 90

populations(9-14), emerging evidence suggests similar outcomes may be attainable in children and 91

adolescents. 92

A recent analysis(15) of the evidence from two systematic reviews of the effect of exercise on 93

depression in this age group indicates encouraging results for the efficacy of PA interventions(7, 16). 94

Statistically significant differences in depression scores for exercise conditions compared with no 95

intervention are reported in a Cochrane Review by Larun and colleagues(7). Given the date of this 96

review, it is timely to update the evidence. The purpose of this study, therefore, was to determine 97

the overall efficacy of PA interventions on depression in children and adolescents by conducting a 98

meta-analysis of available evidence published up to May 2011 and including both clinical and non-99

clinical populations. Both prevention (with non- clinical samples) and management (clinical) studies 100

were included. 101

Methods 102

Search strategies and inclusion criteria 103

A broad literature search strategy was developed using keywords from five categories: population, 104

study design, PA behaviour, mental health outcome and intervention type. Key terms (see Figure 1 105

for full search details) from each category were combined to locate all relevant literature using 106

PsycInfo, The Cochrane Database of Systematic Reviews and The Cochrane Central Register of 107


Controlled Trials, Trials Register of Promoting Health Interventions (TRoPHI; EPPI Centre), Web of 108

Science and MEDLINE. Relevant articles were selected by screening the titles and reviewing abstracts 109

and, when abstracts were not available or did not provide sufficient data, the full text article was 110

retrieved and screened to determine whether inclusion criteria were met. In addition, reference lists 111

of primary studies, review papers, and identified articles were screened for titles that included key 112

terms. Only studies that (1) reported measures on interventions to promote or increase PA; (2) 113

included children and/or adolescents aged 5-19 years; (3) reported on results from any quantitative 114

measure of depression; (4) included a non-physical control or comparison group; (5) were published 115

in peer-reviewed journals written in English; and (6) up to and including May 2011 were included. 116

Data extraction 117

Information about study location and design, randomisation, setting, participants, target population, 118

depression assessment, intervention, and effect sizes and their variation was extracted from each 119

paper by two researchers independently. Study authors were emailed when missing information 120

prevented data extraction. If author(s) did not respond to requests within one month, either the 121

study or the relevant outcome was excluded from the analysis. Differences in facts (committed by 122

transposing information during the coding process) and interpretation (coding differences based on 123

interpretation of the available information) were discussed and resolved by the researchers. 124

Quality assessment 125

Study quality was assessed using the Delphi list(17). Studies were given scores (0 or 1) based on the 126

method of randomisation, treatment allocation, similarity between groups at baseline, reporting of 127

eligibility criteria, blinding of outcome assessors and participants, whether point estimates and 128

measures of variability were presented for the primary outcome measures, and whether intention-129


to-treat analysis was used. Each included paper was given a score out of eight with higher scores 130

meaning higher study quality. 131

Effect Size Calculations 132

Comprehensive Meta-Analysis (CMA) version-2 software was used to compute all effect sizes in the 133

current study(18). A random effects model using Hedges’ g as the effect size index was selected to 134

measure differences in depression between PA experimental and comparison groups(19). The 135

statistical assumption supporting a random effects model suggests that there will be within study 136

error (sampling error) and between study variance. Standardised mean differences were adjusted by 137

the inverse weight of the variance to prevent inflation of study weights, providing more accurate 138

estimates of effect size. The meta-analytic literature has found that Hedges’ g prevents 139

overestimation of an effect size value when sample sizes are fewer than 20 studies(20, 21). The 140

standard formula for Hedges’ g used to correct for bias in small samples was: 141

𝑔 = 𝑑 �1 − 34𝑁−9

� Where, 142

𝑑 = 𝑀1−𝑀2𝑆

143

In these equations, d represents Cohen’s (1969) formula for power(22); M is the sample mean, S the 144

within group sample standard deviation, g is the effect size obtained from Hedges (1981) correction 145

formula(23), using the total sample size N in the denominator. 146

147

Descriptive measures such as means, standard deviations, and sample sizes were used by CMA to 148

calculate estimates of effect size. When descriptive data were insufficient then CMA provided 149

options for effect sizes to be calculated from combinations of available data such as F, t, r, and/or p-150

values. When performing the analysis in CMA each study contributed one effect size calculation to 151


the overall analysis. When a study contained more than one measurement of depression, CMA 152

provided an option that averaged outcomes to calculate an overall summary treatment effect. 153

Coding 154

Studies were coded separately by two authors on three primary classifications (methodological 155

characteristics, participant characteristics, and study characteristics), using processes that develop 156

and refine coding sheets(24) to enable subgroup analysis. Methodological characteristics were 157

coded according to research design (randomized control trial OR other design), intervention duration 158

(< 3 months OR > 3 months), intervention frequency (< 3 days/week OR > 4 days/week), intervention 159

delivery (Physical Activity, Education, OR Both), level of randomisation (school, class, OR individual), 160

and study quality (Delphi score < 4 OR score > 5). Participant characteristics included sample size (N), 161

participant age (mean age < 13 OR > 13), gender (males OR both males and females), and target 162

population (Overweight, At-risk, OR School-based). Study characteristics were categorised by 163

country (United Kingdom OR United States), study year (< 1996 OR > 1996), and measures 164

(Depression only OR Multiple measures). 165

Heterogeneity of Variance 166

The three statistics used to assess homogeneity of variance included the QTotal (QT) value which is 167

based on a χ-square (χ2) distribution, tau-square (τ2) value, and I-square (I2) value. All three 168

statistics (QT, τ2, and I2) were used to interpret heterogeneity of variance. When the QT statistic is 169

significant then a procedure is used to conduct subgroup (moderator) analyses by 170

compartmentalizing variance into QBetween (QB) and QWithin (QW) values, with significant QB 171

values (p < .05) needing a statistical technique (T-test or ANOVA) to determine group 172

differences(21). The τ2 statistic provides an estimate of total variance between studies with larger 173


values reflecting the proportion of variance that can be attributed to real differences between 174

studies in a random effects model. When the number of studies per subgroup is small (k < 5) τ2 can 175

be imprecise, therefore, a pooled estimate of variance was used for all calculations(25). The I2 176

statistic is the ratio of excess dispersion to total dispersion and can be interpreted as the overlap of 177

confidence intervals explaining low (25%), moderate (50%), and high (75%) values of the total 178

variance attributed to covariates(26). Larger values of I2 require techniques (i.e., moderator analysis 179

or meta-regression) to provide explanations(25, 26). Research suggests that smaller samples sizes 180

increase the likelihood that assumptions will be violated when using a random effects model, as 181

error can be overestimated(21). A conservative alpha level (α < .01) was established to prevent type 182

I errors when interpreting results from the moderator analyses. 183

Outlier analysis and publication bias 184

Outliers were identified by analyzing relative residual values (Z < or > + 1.96) and if present were 185

analyzed by using a “one study removed” technique that is available with the CMA version-2 186

software. The criterion for outlier inclusion was a large residual value that did not influence 187

significant (p < .01) effect sizes (Hedges g) and remained within the 95% confidence interval. 188

Publication bias was analyzed through visual inspection of a funnel plot, a Fail Safe N calculation(27) 189

and a “Trim and Fill” procedure(28, 29). Funnel plots provide a visual representation of studies 190

according to standard error (y-axis) and effect size (x-axis) with symmetrical distributions being 191

indicative of a lack of publication bias. Fail Safe N calculations are based on the number of studies 192

needed to nullify significant effects(27). The “Trim and Fill” procedure is an iterative statistical 193

process that adds/removes studies to balance an asymmetrical funnel plot and provide an unbiased 194

estimate of effect size(28, 29). 195


Results 196

The searches yielded 388 titles of potentially relevant articles; of which nine studies met the 197

inclusion criteria(30-38). These studies comprised nine independent samples including 581 children 198

and/or adolescents. The literature search strategy is shown in Figure 1, which also shows the 199

primary reasons for exclusion of studies at each stage of the process. 200

FIGURE 1 NEAR HERE. 201

Methodological characteristics 202

Of the nine studies reviewed, five were randomised controlled trials (RCT)(30-32, 34, 37), two were 203

controlled trials (CT)(35, 38), one was a cluster randomised controlled trial(33) and one was a quasi-204

experimental study(36) (see Table 1). Of these, four were randomised at the individual level(30-32, 205

37), three at the school level(34, 35, 38) and two at the class level(33, 36). Analysis for all studies was 206

conducted at the individual level. Data used to compute effects sizes by CMA included independent 207

group means and standard deviations for six papers (30, 32, 34, 35, 37, 38) and combinations of 208

mean difference scores, t-values, and p-values in the three remaining studies (31, 33, 36). 209

Intervention periods ranged from nine(33) to 40 weeks(36) and sessions lasted between 20(32) and 210

90(31) minutes. Five studies held sessions on three days each week(30, 31, 36-38). The remaining 211

studies held sessions on either two days(35), five days(32), four days(34) or 2-3days(33). 212

Intervention mode varied across studies. Aerobic exercise, included in six interventions, was the 213

most common activity(30-32, 35, 37, 38). One intervention comprised health education(33), one was 214

sport and physical education lessons(36), and one was yoga and mindfulness training(34). Three 215

interventions were delivered by research staff(30, 32, 35), three by trained counsellors(31, 34, 38) 216


and two by physical education staff(36, 37). The remaining study did not provide delivery 217

information(33). 218

TABLE 1 NEAR HERE. 219

Participant characteristics 220

Studies included between 19(33) and 207(32) participants (median sample 81 participants), and 221

were conducted with both male and female participants. Two studies included only boys(31, 37) 222

with the remaining studies using co-educational samples(30, 32, 33, 35, 36, 38). Participants were 223

aged between 8 and 19 years. 224

The majority of studies targeted at-risk groups for depression. Two studies targeted clinically 225

overweight young people (Body Mass Index Centile > 85th or 98th national average)(30, 37). Two 226

studies included only criminally institutionalised youth offenders(31, 37). One study targeted those 227

with low socio-economic status(36) and one those of Hispanic origin(33). The remaining three 228

studies targeted the general population(34, 35, 38). Four interventions were conducted as part of 229

the physical education curriculum(33-36), two in juvenile detention centres(31, 37) and two as part 230

of larger interventions to reduce obesity(30) and improve metabolism(32). The remaining 231

intervention was conducted in an after-school program(38). 232

Study characteristics 233

The majority of included interventions were conducted in the USA(31-34, 37, 38), with remaining 234

interventions conducted in the UK(30, 35) and Chile(36). A variety of assessment tools was used to 235

measure depressive symptoms. Three studies measured depression using the Beck Depression 236

Inventory (BDI)(31, 33, 37). Remaining studies used the Children’s Depression Inventory (CDI)(30), 237


the Profile of Mood States (POMS)(38), the Hospital Anxiety and Depression Scale (HADS)(36), the 238

Short Mood and Feelings Questionnaire (SMFQ-SF)(34), the Multiple Affect Adjective Check List 239

(MAACL)(35) or the Reynolds Child Depression Scale (RCDS)(32). 240

Quality assessment 241

There was good initial agreement (89%) between the two independent reviewers on the 242

methodological quality of included studies. Quality ratings for each study are shown in Table 2. Two 243

studies scored 7 out of the maximum score of 8; neither blinded patients to the intervention(30, 32). 244

A further study did not include an intention-to-treat analysis and did not blind participants and 245

therefore scored 6 out of 8(37). The remaining studies received lower quality scores ranging from 246

5(33) to 2(35, 38). 247

INSERT TABLE 2 NEAR HERE. 248

Meta-analysis 249

Individual and summary effect size statistics are shown in Figure 2. When interpreting standardized 250

mean differences, Cohen’s criteria(22) were used to determine small, r = .10; medium, r = .30; large, 251

r = .50 effect sizes. Negative effect sizes were indicative of PA treatment groups having decreased 252

depression scores when compared with control or comparison groups. 253

INSERT FIGURE 2 NEAR HERE. 254

The PA intervention summary treatment effect for depression was small but significant (Hedges’ g = 255

-0.26, SE = 0.09, 95% C.I. = -0.43, -0.08, p = 0.004) representing about one quarter of a standard 256

deviation’s decrease in depression for experimental groups. Analysis revealed a significant 257

heterogeneous distribution (QT = 19.84, p = .002, I2 = 59.68) indicating the need to conduct 258

subgroup (moderator) analyses using the methodological, participant, and study characteristics 259


coded for each study. No outliers were identified as relative residuals (Z-values) ranged from 0.69 to 260

-1.77, therefore, a sensitivity analysis was not conducted. The potential for publication bias was 261

marginal as there was a symmetrical funnel plot, no studies were added when the Trim and Fill 262

procedure was performed, and a moderate Fail Safe N calculation was needed to nullify a significant 263

effect (α < 0.01). 264

The review of homogeneity statistics found a moderate (I2 = 59.68) significant (QT = 19.84) 265

heterogeneous distribution that would explain a portion of the variance using subgroup analyses 266

(see Tables 3 and 4). There were no significant (p < .01) differences between categories coded, 267

however, trends were present with many subgroups. Borenstein and colleagues(25) recommend 268

caution when interpreting subgroup analyses with fewer than five studies as estimates of effect size 269

may be imprecise. We chose to report these findings only to provide guidance for future research by 270

providing conservative interpretations of findings(25). 271

INSERT TABLES 3 AND 4 NEAR HERE 272

Trends from the moderator results suggest that methodological characteristics were important. The 273

largest effects on depression were for higher quality, short (less than three months in duration), 274

randomised controlled trials (randomised by individual) that included both education and physical 275

activities. Participant characteristics that contributed most to the reduction of depression included 276

single gender studies, studies that targeted overweight/obese youth or adolescence, and studies 277

that were conducted in the United Kingdom. The only study characteristic that demonstrated a 278

trend was interventions with a single measure of depression. All the PA intervention moderators 279

demonstrated lower between study variance (τ2) and explained moderate to large portions of 280

subgroup variance (I2). 281


Discussion 282

This meta-analysis updated the extant literature on the impact of PA interventions on depression in 283

children and adolescents. The results indicated a small but significant summary treatment effect. 284

This is in contrast with the findings of Larun and colleagues(7) who conducted a systematic review of 285

five exercise interventions on depression in children and adolescents (up to 20 years old), and found 286

a significant moderate effect (standard mean difference effect size = −0.66; CI, −1.25, −0.08). They 287

noted however that the number of included trials was low, that the trials were of low quality and 288

highly varied in respect of methodological characteristics, such as sampling and measurement. We 289

included results from more studies (nine, compared with five), including some considered to be of 290

higher quality. As the results from the study by Norris and colleagues (which used flexibility activities 291

and lower-intensity aerobic exercise)(35) contradicted those from all the other included studies, we 292

considered the possibility that it’s inclusion may have contributed to our lower effect size. Sensitivity 293

analysis found, however, that omission of this study had a marginal effect on the overall findings. 294

295

The results of individual studies in our meta-analysis were variable, as indicated by the moderate, 296

significant heterogeneity statistics. This might be explained by our inclusion of both preventive and 297

treatment trials, with a wide range of baseline depression levels, and with greater effects in the 298

clinical trials. However, given the relative paucity of literature with children and adolescents, and the 299

importance of mental health in this age group, it is sensible, in our view, not to be too restrictive 300

with exclusion criteria at this point. 301

302


The findings also indicated that studies that focused on depression (i.e. did not measure additional 303

outcomes) reported a greater treatment effect(30-33, 37). This suggests a possible need to test 304

outcome-specific interventions, rather than generic ‘fix-all’ treatment approaches, when working 305

with depression in children and adolescents. Moreover, young people referred to physical activity 306

programmes for specific health reasons, for example as part of a study examining obesity(30) or 307

metabolic outcomes(32), may have different motivations and expectation, which may affect 308

psychological outcomes. 309

Our moderator analyses also demonstrated that studies with higher quality scores indicated greater 310

treatment effects. These studies(30, 34) were distinct in that they concealed treatment allocation, 311

blinded the outcome assessor and performed intention-to-treat analysis. Further experimental work 312

should ensure high methodological quality using these criteria. 313

The majority of studies targeted at-risk groups for depression including criminally institutionalised 314

male youth offenders(31, 37), those with low socio-economic status(36) and those of Hispanic 315

origin(33). Two studies targeted young people with Body Mass Index (BMI) centile > 85th or 98th 316

national average and reported a greater treatment effect of physical activity in clinically overweight 317

participants(30, 32). Previous evidence also supports the premise that elevated BMI in children and 318

adolescents is associated with depression(39, 40). Considering that PA is recognised as an effective 319

strategy for weight loss(41, 42), such interventions may have dual benefit when working with 320

overweight children and adolescents. 321

We are not able to do more than speculate on the likely mechanisms of why depression is reduced 322

after physical activity. Plausible reasons include neuro-biological mechanisms, whereby 323

neurotransmitters released during activity may mediate changes in depressive symptoms and mood, 324


and psychosocial mechanisms, which result in improved mastery and elevations in physical self-325

worth, the latter being implicated in improved self-esteem(15, 43). However, there are no definitive 326

explanations as to which may be more important, and further work in this area is required. Coupled 327

with this is the need to investigate short and long term effects of physical activity, in the context of 328

possible spontaneous remission from depression with the passage of time. 329

Conclusion 330

This meta-analysis assessed the impact of PA interventions on depression in children and 331

adolescents. As only nine studies were included, caution is required in interpretation of the results. 332

However, study quality was higher than in previous reviews, and the small but significant treatment 333

effect suggests that physical activity may play a role in the prevention and treatment of depression 334

in young people. Further high quality outcome-focussed studies are now needed to confirm this 335

finding, and to inform the implementation of programs to effectively reduce depressive symptoms in 336

children and adolescents. 337

338

Funding sources: This work was supported by an Alf Howard International Travel Scholarship (School 339

of Human Movement Studies, University of Queensland) that enabled international collaborative 340

work to be undertaken. HEB was partially supported by an Australian NHMRC Program Grant (# 341

301110). 342

References 343

1. W.H.O. Atlas: child and adolescent mental health resources: global concerns, implications for 344 the future. 2005;ISBN 92 4 156304 4. 345


2. Satcher DS. Executive summary: a report of the Surgeon General on mental health. Public 346 Health Rep. 2000;115(1):89-101. Epub 2000/09/01. 347 3. Neal D R. Treatment of depression in children and adolescents. Lancet. 2005;366(9489):933-348 40. 349 4. Guberman C, Manassis K. Symptomatology and family functioning in children and 350 adolescents with comorbid anxiety and depression. J Can Acad Child Adolesc Psychiatry. 351 2011;20(3):186-95. Epub 2011/08/02. 352 5. Lundy SM, Silva GE, Kaemingk KL, Goodwin JL, Quan SF. Cognitive Functioning and Academic 353 Performance in Elementary School Children with Anxious/Depressed and Withdrawn Symptoms. 354 Open Pediatr Med Journal. 2010;4:1-9. Epub 2010/07/29. 355 6. W.H.O. The World Health Report. ISBN 92 4 156201 3. 2001. 356 7. Larun L, Nordheim Lena V, Ekeland E, Hagen Kåre B, Heian F. Exercise in prevention and 357 treatment of anxiety and depression among children and young people. Cochrane Database Syst Rev 358 [Internet]. 2006; (3). Available from: 359 http://www.mrw.interscience.wiley.com/cochrane/clsysrev/articles/CD004691/frame.html. 360 8. Biddle SJH, Fox, K.R. and Boutcher, SH. . Physical activity and psychological well-being. 361 London; Routledge 2000. 362 9. Martinsen EW. Physical activity in the prevention and treatment of anxiety and depression. 363 Nord J Psychiatry. 2008;62 Suppl 47:25-9. Epub 2008/09/09. 364 10. Fox KR. The influence of physical activity on mental well-being. Public Health Nutr. 365 1999;2(3A):411-8. Epub 1999/12/28. 366 11. Landers DM. The influence of exercise on mental health. Research Digest (President's 367 Council on Physical Fitness and Sports (US). 1997;Series 2 (no. 12). 368 12. Dunn AL, Trivedi MH, Kampert JB, Clark CG, Chambliss HO. Exercise treatment for 369 depression: Efficacy and dose response. Am J Prev Med. 2005;28(1):1-8. 370 13. Dunn AL, Trivedi MH, O'Neal HA. Physical activity dose-response effects on outcomes of 371 depression and anxiety. Med Sci Sport Exer. 2001;33(6):S587-S97. 372 14. Strohle A. Physical activity, exercise, depression and anxiety disorders. J Neural Transm. 373 2009;116(6):777-84. 374 15. Biddle SJH, Asare M. Physical activity and mental health in children and adolescents: a 375 review of reviews. Brit J Sport Med. 2011;45(11):886-95. 376 16. Calfas KJ. Effects of physical activity on psychological variables in adolescents. Pediatr Exerc 377 Sci. 1994;6:406-23. 378 17. Verhagen AP, de Vet HC, de Bie RA, Kessels AG, Boers M, Bouter LM, et al. The Delphi list: a 379 criteria list for quality assessment of randomized clinical trials for conducting systematic reviews 380 developed by Delphi consensus. J Clin Epidemiol. 1998;51(12):1235-41. Epub 1999/03/23. 381 18. Borenstein M, Hedges, L., Higgins, J., & Rothstein, H. . Comprehensive meta-analysis: A 382 computer program for research synthesis Englewood, NJ: Biostat. 2005. 383 19. Borenstein M, Hedges, L.V., Higgins, J.P.T., & Rothstein, H.R. . A basic introduction to fixed 384 and random effects models for meta-analysis. . Research Synthesis Methods. 2010;1:97-111. 385 20. Field AP. Meta-analysis of correlation coefficients: a Monte Carlo comparison of fixed- and 386 random-effects methods. Psychol Methods. 2001;6(2):161-80. Epub 2001/06/20. 387 21. Hedges LVO, I. Statistical methods for meta-analysis. New York: Academic Press. 1985. 388 22. Cohen J. Statistical power analysis for the behavioral sciences Hillsdale, NJ: Erlbaum. 389 1988;2nd ed. 390

http://www.mrw.interscience.wiley.com/cochrane/clsysrev/articles/CD004691/frame.html


23. Hedges LV. Distribution Theory for Glass's Estimator of Effect Size and Related Estimators. 391 Journal of Educational Statistics. 1981;6(2):107-28. 392 24. Brown SA, Upchurch SL, Acton GJ. A framework for developing a coding scheme for meta-393 analysis. Western J Nurs Res. 2003;25(2):205-22. Epub 2003/04/02. 394 25. Borenstein M. Introduction to Meta-analysis. Chichester, UK: Wiley. 2009. 395 26. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. 396 BMJ. 2003;327(7414):557-60. 397 27. Rosenthal R. The file drawer problem and tolerance for null results. Psychological Bulletin. 398 1979;86(3):638-41. 399 28. Duval S, Tweedie R. A nonparametric "trim and fill" method of accounting for publication 400 bias in meta-analysis. J Am Stat Assoc. 2000;95(449):89-98. 401 29. Duval S, Tweedie R. Trim and fill: A simple funnel-plot-based method of testing and adjusting 402 for publication bias in meta-analysis. Biometrics. 2000;56(2):455-63. 403 30. Daley AJ, Copeland RJ, Wright NP, Roalfe A, Wales JKH. Exercise therapy as a treatment for 404 psychopathologic conditions in obese and morbidly obese adolescents: A randomized, controlled 405 trial. Pediatrics. 2006;118(5):2126-34. 406 31. Hilyer JC, Wilson DG, Dillon C, Caro L, Jenkins C, Spencer WA, et al. Physical-Fitness Training 407 and Counseling as Treatment for Youthful Offenders. J Couns Psychol. 1982;29(3):292-303. 408 32. Petty KH, Davis CL, Tkacz J, Young-Hyman D, Waller JL. Exercise effects on depressive 409 symptoms and self-worth in overweight children: a randomized controlled trial. J Pediatr Psychol. 410 2009;34(9):929-39. Epub 2009/02/19. 411 33. Melnyk BM, Jacobson D, Kelly S, O'Haver J, Small L, Mays MZ. Improving the Mental Health, 412 Healthy Lifestyle Choices, and Physical Health of Hispanic Adolescents: A Randomized Controlled 413 Pilot Study. J School Health. 2009;79(12):575-84. 414 34. Mendelson T, Greenberg MT, Dariotis JK, Gould LF, Rhoades BL, Leaf PJ. Feasibility and 415 Preliminary Outcomes of a School-Based Mindfulness Intervention for Urban Youth. J Abnorm Child 416 Psych. 2010;38(7):985-94. 417 35. Norris R, Carroll D, Cochrane R. The effects of physical activity and exercise training on 418 psychological stress and well-being in an adolescent population. J Psychosom Res. 1992;36(1):55-65. 419 36. Bonhauser M, Fernandez G, Puschel K, Yanez F, Montero J, Thompson B, et al. Improving 420 physical fitness and emotional well-being in adolescents of low socioeconomic status in Chile: results 421 of a school-based controlled trial. Health Promot Int. 2005;20(2):113-22. Epub 2005/03/25. 422 37. MacMahon JR, Gross RT. Physical and psychological effects of aerobic exercise in delinquent 423 adolescent males. Am J Dis Child. 1988;142(12):1361-6. Epub 1988/12/01. 424 38. Annesi JJ. Correlations of depression and total mood disturbance with physical activity and 425 self-concept in preadolescents enrolled in an after-school exercise program. Psychol Rep. 2005;96(3 426 Pt 2):891-8. Epub 2005/09/22. 427 39. Goodman E, Whitaker RC. A Prospective Study of the Role of Depression in the Development 428 and Persistence of Adolescent Obesity. Pediatrics. 2002;110(3):497-504. 429 40. Lumeng JC, Gannon K, Cabral HJ, Frank DA, Zuckerman B. Association Between Clinically 430 Meaningful Behavior Problems and Overweight in Children. Pediatrics. 2003;112(5):1138-45. 431 41. Strong WB, Malina RM, Blimkie CJ, Daniels SR, Dishman RK, Gutin B, et al. Evidence based 432 physical activity for school-age youth. J Pediatr. 2005;146(6):732-7. Epub 2005/06/24. 433 42. Washington RL. Does increasing physical activity really make a difference? Show me the 434 evidence. Pediatrics. 2005;146(6):A2. 435


43. Fox KR. The effects of exercise on self-perceptions and self-esteem. In: Biddle SJH, editor. 436 Physical activity and psychological well-being. London: Routledge.; 2000. p. 88-117. 437

438

439

440


441

Study (authors, country,

study design)

Level of randomisation;

number of clusters Setting Participants (number; age;

% male)a Target population Assessment period; depression measure

Intervention: description; delivery; duration

Annesi[1] (2005; USA; QE)

After-school program

N=2

After-school program n=49; 10.5 ± 0.9yrs; 46% None 0 and 12 weeks; POMS-

SF

45 min aerobic, resistance and stretching exercise 3 day/week; delivered by trained counsellors; 12 weeks

Bonhauser et al.[2] (2005; Chile; CRT)

Class enrolment

N=4 Public school

n=198; 15.53 ± 0.9yrs; 54% additional education and 43% usual physical education

Low socio-economic status 0 and 40 weeks; HADS

90 min additional sport and physical education classes 3 day/week vs. usual physical education (comparison); delivered by physical education staff; 40 weeks

Daley et al.[3] (2009; UK; RCT) Individual

Larger study population (examining the effect of exercise on obesity)

n=81; 13.1yrs; 44.4% BMI >98th percentile 0, 8, 14 and 28 weeks; CDI

30 min aerobic exercise 3 day/week; delivered by trained research staff; 28 weeks

Table 1: descriptive characteristics of studies meeting inclusion criteria


Hilyer et al.[4] (1982; USA; RCT) Individual

State industrial school for youth offenders

n=30; 15.5-18.6yrs; 100% Labelled ‘criminally institutionalised’ 0 and 20 weeks; BDI

90 min ‘physical fitness’ program 3 day/week; delivered by counsellors; 20 weeks

MacMahon et al.[5] (1988; USA; RT)

Individual Juvenile detention centre

n=98; 16.3yrs (14-18.25); 100%

Labelled ‘criminally institutionalised’ 0 and 12 weeks; BDI

40 min vigorous exercise 3 day/week (intervention) vs. 40 min ‘light’ exercise 3 day/week (comparison); delivered by physical education staff; 12 weeks

Melnyk et al.[6] (2009; USA; CRCT)

Class enrolment

N=2 Public school n=19; 15.5 ± 0.63; 68% Majority Hispanic 0 and 9 weeks; BDI

50 min health education (promoting physical activity) 2-3 day/week; delivery unknown; 9 weeks

Mendelson et al.[7] (2010; USA; CRCT)

School enrolment

N=4 Public school n=97; age range: 9.7-10.6;

39.2% None 0 and 12 weeks; SMFQ-C

45 min yoga-based physical activity, breathing techniques and guided mindfulness practices 4 day/week; delivered by HLF instructors; 12 weeks


Norris et al.[8] (1992; UK; CRCT)

School enrolment

N=2 Public school n=60; mean age: 16.6yrs;

52% None 0 and 10 weeks; Multiple Affect Adjective Check List

30 min exercise sessions (high- vs. moderate-intensity vs. flexibility) 2 day/week; delivered by trained research staff; 10 weeks

Petty et al.[9] (2009; USA; RCT) Individual

Larger study population (examining the effect of exercise on metabolism)

n=207; age range:: 9.3-9.4yrs; 36% black and 52% white

BMI >85th percentile 0 and 14 weeks; RCDS

20 min/day or 40 min/day aerobic exercise program; delivered by trained research staff; 13 ± 1.6 weeks

Note. Study (authors, country, study design): USA = United States of America, UK = United Kingdom; QE = quasi-experimental study design, CRT = cluster-randomised trial, RCT = 442 randomised controlled trial, RT = randomised trial, CRCT = cluster-randomised controlled trial. Level of randomisation; number of clusters: N = number of clusters. Participants (number; 443 age; %male): n = number of participants, yrs = years of age. Target population: BMI = Body Mass Index. Assessment period; depression measure: POMS-SF = Profile of Mood States–Short 444 Form, HADS: Hospital Anxiety and Depression Scale, CDI = Children’s Depression Inventory, BDI = Beck Depression Inventory, SMFQ-C = Short Moods and Feelings Questionnaire-Children, 445 RCDS: Reynolds Child Depression Scale. Intervention: description; delivery; duration: min = minutes, HLF = Holistic Life Foundation. 446 447 aMeans and standard deviations and/or age ranges are provided where appropriate. 448

449


450

451

Annesi 2005

Bonhauser et al 2005

Daley et al 2009

Hilyer et al 1982

MacMahon et al. 1988

Melnyk et al. 2009

Mendelson et al. 2010

Norris et al. 1992

Petty et al. 2009

Was a method of randomisation performed? N Y Y Y Y Y Y N Y

Was the treatment allocation concealed? N N Y Y Y N N N Y

Were the groups similar at baseline regarding the most important prognostic indicators? Y Y Y Y Y Y Y N Y

Were the eligibility criteria specified? N N Y N Y Y N N Y

Was the outcome assessor blinded? N N Y N Y N N N Y

Was the patient blinded? N N N N N Y N Y N

Were point estimates and measures of variability presented for the primary outcome measures? Y Y Y Y Y Y Y Y Y

Did the analysis include an intention-to-treat analysis? n Y y N N N N N Y

Total Delphi score / 8 2 4 7 4 6 5 3 2 7

Table 2: quality assessment of studies meeting inclusion criteria


452

453

454

Table 3: coding information for studies meeting inclusion criteria 455

Methodological Characteristics Participant Characteristics Study Characteristics

Study Design Duration Frequency Delivery Randomization Quality Age Gender Target Country Year Measure

Annesi, 2005 O <3 1 PA S L A B S US A M

Bonhauser et al., 2005 O >3 1 PA C L A B S US A M

Daley et al., 2006 RCT <3 1 PA I H A B O UK A D

Hilyer et al., 1982 RCT >3 1 B I L A M A US B D

McMahon & Gross, 1988 RCT >3 1 PA I H A M A US B D

Melnyk et al., 2009 O >3 1 E S H A B S US A D

Mendelson et al., 2010 RCT <3 2 B C L Y B S US A M

Norris et al., 1992 O <3 1 PA C L A B S UK B M

Petty et al., 2009 RCT >3 2 PA I H Y B O US A D

Note. Design (Research Design): O = other design, RCT = randomized control trial. Duration (Research Duration): < 3 = less than 3 months, > 3= greater than or equal to 3 months. 456 Frequency (Intervention Frequency): 1 = less or equal to 3 days/week, 2 = greater than 4 days per week. Delivery (Intervention Delivery): PA = physical activity, E = education, B = both PA 457 and education. Randomization (Randomization Level): S = School, C = class, I = individual. Quality (Study Quality): L = Delphi score < 4, H = Delphi score > 5.Age: A = adolescent mean age > 458 13 years, Y = youth mean age < 13 years. Gender: M = males only, B = Both males and females. Target (Intervention Target): S = school based, O = obese/overweight, A = at-risk/juvenile 459


delinquent. Country: UK = United Kingdom, US = United States. Year: A = 1996 and after, B = before 1996. Measure: M = multiple measures taken during intervention, D = Depression only 460 measures taken. 461 462 463 464 465 466 467 468 469 470

Effect Size Statistics Null Test Heterogeneity Statistics Publication Bias k g SE s2 95% C.I. Z Q τ2 I2 Fail Safe N

Random Effects Model a 9 -0.257 0.090 0.008 (-0.433, 0.080) -2.853 19.84* 0.041 59.68 22 Methodological Characteristics b

Research Design 1.498b Randomized Control Trial 5 -0.345 0.110 0.012 (-0.561,- 0.129) -3.138 0.049 66.37

Other Design 4 -0.135 0.132 0.017 (-0.393, 0.123) -1.026 0.001 2.97 Intervention Duration 1.097b

< 3 Months 6 -0.329 0.108 0.012 (-0.541, -0.118) -3.057 0.066 65.42 > 3 Months 3 -0.149 0.135 0.018 (-0.413, 0.115) -1.106 0.000 0.000

Intervention Frequency 0.368b < 3 Days/Week 7 -.288 0.107 0.011 (-0.497,- 0.078) -2.694 0.058 66.01 > 4 Days/Week 2 -.156 0.190 0.036 (-0.528, 0.216) -0.820 0.000 0.000

Intervention Delivery 2.559b PA 6 -0.198 0.096 0.009 (-0.386, -0.009) -2.058 0.007 18.81

Education 1 -0.103 0.277 0.077 (-0.646, 0.441) -0.369 0.000 0.000 Both PA & E 2 -0.458 0.147 0.022 (-0.746, -0.170) -3.118 0.083 77.95

Level of Randomization 3.196b School 2 -0.275 0.197 0.039 (-0.661, 0.110) -1.399 0.004 6.88 Class 3 -0.061 0.142 0.020 (-0.611, -0.161) -0.429 0.000 0.000

Individual 4 -0.386 0.115 0.013 (-0.433, 0.080) -3.365 0.054 70.95 Study Quality 2.259b

Low-Score < 4 5 -0.145 0.111 0.012 (-0.363, 0.074) -1.297 0.000 0.000 High-Score > 5 4 -0.386 0.116 0.013 (-0.614, -0.159) -3.328 0.059 71.17

Participant Characteristics b Age 0.368b

Table 4: sub-group analysis of included studies Effect Sizes (95% Confidence Limits) and Between-Group Tests of Heterogeneity (Q), Assessment of the Amount of Variability Caused by Random Error (I2), and Tests of Moderators for Studies Included in the Meta-analysis


< 13 years 2 -0.156 0.190 0.036 (-0.528, 0.216) -0.820 0.000 0.000 > 13 Years 7 -0.288 0.107 0.011 (-0.497, -0.078) -2.694 0.058 66.01

Gender 0.234b Males Only 2 -0.351 0.218 0.048 (-0.778, 0.077) -1.607 0.042 46.82

Males & Females 7 -0.233 0.106 0.011 (-0.440, -0.026) -2.209 0.049 66.50 Population Target 1.466b

Obese/Overweight 2 -0.407 0.162 0.026 (-0.725, -0.089) -2.507 87.60 0.090 At-Risk 3 -0.249 0.158 0.025 (-0.559, 0.062) -1.571 32.39 0.017 School 4 -0.143 0.146 0.021 (-0.454, -0.064) -0.986 2.18 0.001

Country 0.493b UK 2 -0.358 0.165 0.027 (-0.682, -0.034) -2.168 0.219 90.02 US 7 -0.223 0.099 0.010 (-0.417, -0.029) -2.251 0.000 0.000

Study Characteristics b Year 0.152b

Before 1996 3 -0.200 0.172 0.029 (-0.536, 0.136) -1.165 0.043 64.59 1996 and After 6 -0.279 0.110 0.012 (-0.495, -0.063) -2.535 0.054 55.48

Measures 1.433b Depression Only 5 -0.345 0.111 0.012 (-0.563, -0.127) -3.107 0.050 66.38

Multiple Measures 4 -0.140 0.130 0.017 (-0.310, 0.114) -1.081 0.001 2.52 Note:. PA = physical activity, E = education. 471 k = number of effect sizes. g = effect size (Hedges g). SE = standard error. S2 = variance. 95% C. I. = confidence intervals (lower limit, upper limit). Z = test of null hypothesis. 472 τ2 = between study variance in random effects model. I2 = total variance explained by moderator. * indicates p < .01. a = Total Q-value used to determine heterogeneity. b = Between 473 Q-value used to determine significance (α < 0.01). 474 475

476

477

478

479

9 10 12 14 Research Unit16 - CORE · This is a post-print version of the following article: Brown,...

Documents

Transcript of 9 10 12 14 Research Unit16 - CORE · This is a post-print version of the following article: Brown,...