Randomized controlled trials (RCTs) for inclusion in a meta-analysis of effectiveness of dietetic interventions in child and adolescent obesity.
Heterogeneity among 17 randomized controlled trials of interventions for child obesity highlighting the variability of study components and follow-up periods. BT indicates behavioral therapy; CBT, cognitive behavioral therapy; Ctl, control; DA, dietary advice; PA, physical activity; PI, postintervention; SB, sedentary behavior. * indicates lifestyle education included behavioral therapy plus dietary advice plus physical activity.
Meta-analysis of 7 randomized controlled trials of interventions for child obesity where a “true” control arm was used and results were reported postintervention. CI indicates confidence interval; SMD, standardized mean difference.
Meta-analysis of 3 randomized controlled trials of interventions for child obesity where a “true” control arm was used and results were reported after a period of follow-up. CI indicates confidence interval; SMD, standardized mean difference.
Collins CE, Warren J, Neve M, McCoy P, Stokes BJ. Measuring Effectiveness of Dietetic Interventions in Child ObesityA Systematic Review of Randomized Trials. Arch Pediatr Adolesc Med. 2006;160(9):906-922. doi:10.1001/archpedi.160.9.906
To assess the effectiveness of dietetic treatment for obese children and to report details of dietary interventions.
English-language articles from 1975 to 2003 available from health and medical databases.
Randomized controlled trials with subjects younger than 18 years of age that included a dietary intervention in isolation or in combination with lifestyle modifications and/or psychological therapies. One person searched the databases; 2 people independently critically appraised the articles for methodological quality and then extracted data using standardized tools.
Thirty-seven randomized controlled trials met the inclusion criteria; 17 contained sufficient information for a Forest plot of the standardized effects. Eight studies had a true control and were included in a meta-analysis. The random effects model was reported if the Q noncombinability χ2 statistic was significant at the 10% level because it has low power as a strict test of homogeneity.
The 2 strongly qualified meta-analyses suggest that interventions that include a dietary treatment do achieve relative weight loss. Details of the dietary intervention or participant food intake are rarely described.
It is not possible to evaluate the effectiveness of dietary treatment for childhood obesity because of the lack of high-quality studies and the heterogeneity of designs, treatment combinations, outcome measures, and follow-up. There is an urgent need to improve the quality of studies in this area because childhood obesity poses major health risks for populations, yet there is limited evidence on which to base treatment strategies.
Obesity in children is increasing rapidly around the world. The associated major adverse health consequences have been systematically reviewed1 and the psychological consequences and social marginalization reported.2
Prevention is clearly the ideal solution to the problem. However, for those already overweight or obese, or at an increased risk of becoming so, effective interventions will offer the only chance of reducing the probability of progression to adult obesity, its associated comorbidities, and the increased risk of premature mortality.3 Essential elements of interventions are likely to include diet and food intake modification,4 and these have been identified for adult interventions. The efficacy of such interventions in the pediatric population is not known. An examination of successful interventions along with the key dietary changes achieved by participants would inform the optimal dietary treatment for broad dissemination. Therefore, the objective of this review was to identify and present the best available evidence on the effectiveness of dietetic treatment and management for overweight or obese children and adolescents. A secondary goal was to specifically describe the elements of successful dietary treatments.
We performed a systematic review of the published English-language literature from 1975 to 2003. These limits were set to contain costs and maintain timelines. Because the prevalence of child obesity has increased predominantly since the late 1970s,5 it was considered likely that few relevant studies would have been published before then. An additional author search was carried out in 2004 for the most up-to-date results of specific studies identified in the original search.
Electronic databases were searched (CINAHL [Cumulative Index to Nursing and Allied Health Literature], MEDLINE, PREMEDLINE, DARE [Database of Abstracts of Review of Effectiveness], Cochrane, EMBASE [Excerpta Medica Database], Austrom, Current Concepts, and Dissertation Abstracts). In addition, we searched nutrition and dietetics journals by hand (the Journal of the Dietitians Association of Australia, the International Journal of Obesity, and the Journal of Human Nutrition and Dietetics) and government reports from the United States, United Kingdom, and Australia along with the reference lists and bibliographies of retrieved articles. The MeSH (Medical Subject Headings of the National Library of Medicine) keyword search terms were dietetic, paediatric (pediatric), child, adolescent, family, parent, school, overweight, obesity, intervention, weight control or weight management or weight loss, and healthy weight.
The current report is limited to the randomized controlled trials (RCTs) identified for a more extensive review undertaken for the Joanna Briggs Institute (JBI), Adelaide, Australia. Any trial that evaluated the effectiveness of nutrition or dietary interventions for treating overweight or obesity in children and adolescents was included for the JBI review.
Participants aged younger than 18 years who were defined as overweight or obese were included. We also included programs that were directed exclusively at parents of overweight or obese children and adolescents. Participants were free-living outpatients; inpatients in clinical obesity units; or students attending community programs, one-off programs, camps, or schools. Studies that reported at least one of the following primary outcomes were included: percentile of body mass index (BMI), calculated as weight in kilograms divided by height in meters squared; BMI z score; percentage of children overweight for age; waist measurement; skin folds; percentage of body fat; or percentage of lean body mass.
We considered interventions aimed at the treatment of elevated body weight delivered by any health professional or nutrition scientist. Initially, the aim was to review only studies that compared dietary modification with a standard control (ie, no treatment). Because of the limited number of such studies, we also reviewed interventions that included dietary modification both as part of an intervention and/or control with various combinations of lifestyle modifications (physical activity and/or sedentary behavior modification) and psychological components (cognitive and/or behavioral therapy).
Studies that did not report a weight-related variable as a primary outcome were excluded from the review. Those studies that were assessed as being of poor methodological quality using the JBI tools6 or that did not meet the inclusion criteria were excluded from the review.
Articles were retrieved if the information contained in the title, abstract, and descriptor/MeSH headings appeared to meet the inclusion criteria. Working independently, 2 reviewers critically appraised the articles for methodological quality using the JBI critical appraisal tools.6 Particular attention was paid to sources of bias in the studies.7 If disagreement between reviewers occurred in any of these steps, a third person repeated the procedure to achieve consensus.
We used Review Manager (RevMan 4.2, The Cochrane Collaboration, Oxford, England) to manage the systematic review. Interventions and study design defined comparability of the studies. Heterogeneity was assessed using the standard Q statistic for heterogeneity and visual interpretation of graphs. Significant heterogeneity was assumed when P<.10. All outcome data reported were continuous, but because studies variously reported changes in percentage overweight, BMI z score, percentage body fat, or body weight, standardized mean differences were used in the analyses. The standardized mean difference has the important property that its value does not depend on the measurement scale. This enabled comparisons of differences in mean weight loss when expressed in percentage body fat, weight for height, or another outcome measure. Studies were excluded from the analyses if they did not report numbers of participants or standard deviations.
There were 1310 articles identified by the search strategy (Figure 1). After excluding studies due to duplication and assessing titles and abstracts to meet inclusion criteria, we retrieved 329 full text articles for reviewing. One hundred twenty-five articles met the inclusion criteria for the main review and details will be reported through JBI publications. After we excluded the nonrandomized trials, there were 49 articles arising from 37 individual studies because some reported the results of the same study more than once. These 37 RCTs are reported here.
The total number of participants across all studies was 2262. The focus of the interventions was diet with various combinations of modifications to physical activity and/or sedentary behavior or behavioral and/or cognitive therapy. Of the 37 studies, 18 had two intervention arms,8- 25 11 had three,26- 35 6 had four,36- 41 and 1 had five arms (Table 1).43 Thirty-two studies8- 14,16- 21,23,24,26- 29,31- 40,42,43 included physical activity as a component of at least 1 of the study arms while 14 studies8,9,11,13,16,18,20,26,27,31,32,34,35,42 included behavioral therapy as a component of at least 1 of the study arms. Six studies8,19,29,31,33,43 included cognitive behavioral therapy specifically as a component of at least 1 of the study arms. Six studies12,14,15,25,28,39 included decreasing sedentary behavior as a component of at least 1 of the study arms. The length of intervention in each RCT ranged from 6 weeks to 18 months with the majority (n = 17) between 3 months and 6 months.14- 20,26- 29,34,35,37- 39,41 Nine studies8- 10,21- 23,31,33,40 described interventions that were run for up to 3 months, and 7 studies12,13,25,30,32,35,43 had interventions that ran for longer than 6 months. The length of intervention was unclear or varied in 3 studies.11,24,36 The length of follow-up was very diverse among the studies. The shortest was 1 month38 while the longest was 10 years.44 Nine studies17,18,22,23,26,32,37,38,42 included no follow-up of participants postintervention. Table 1 also reports the retention rates at the longest point of follow-up except for the 5 studies with a follow-up exceeding 2 years where interim rates are given.24,25,35,43,45 Studies undertaking intention-to-treat analyses are also highlighted in Table 1. The majority of studies were conducted in the United States (n = 29), 5 were in different European countries,21,24,25,30,43 2 were in Australia,8,17 and 1 was in Hong Kong.23
Because of the various combinations of treatments that included a dietary intervention and the variable composition of the control groups, a meta-analysis to show the effectiveness of dietetic interventions per se was not possible. Results of the main weight outcomes for each of the 37 RCTs are presented in Table 2 with the significance indicated. A Forest plot of the standardized effects for the studies that contained sufficient information (n = 17) highlights the heterogeneity of the interventions (Figure 2). It was not appropriate to report a meta-analysis result because only a minority of studies had an adequate control group and the treatments reported were highly diverse. However, a meta-analysis was undertaken on the subset of studies (n = 8) that included both a dietary intervention component and an adequate control group (Figure 3). In these studies, diet was a component of the interventions and the control group either received no intervention, was placed on a waiting list, or received usual care. Caution must be exercised in the interpretation of the analysis because diet was only a component of the interventions, making its contribution difficult to evaluate. In addition, the number of studies is limited. The meta-analysis suggests that interventions that contain a dietary component are effective in achieving relative weight loss in overweight/obese children and adolescents (pooled standardized mean difference, −1.82; 95% confidence interval, −2.40 to −1.23) (Figure 3). The 2 studies with the greatest standardized effect were those of Rochini et al38 and Becque et al26 (Figure 2). Becque et al reported that in adolescents, a reduction in the percentage body fat of approximately 3% was achieved after 20 weeks of treatment with diet and behavior change with or without exercise.26 Postintervention, participants in the study by Rochini et al of obese adolescents significantly decreased their percentage body fat by 3% to 6%38 (Table 2). Only 4 studies had follow-up data, and a meta-analysis of the outcomes of these results suggests a diminishing effect of the intervention over time (pooled standardized mean difference, −0.64; 95% confidence interval, −0.89 to −0.39) (Figure 4).
A summary of the dietary interventions and the studies that measured and reported dietary compliance and outcomes is given in Table 3. Sixteen studies8- 16,22,28,29,35,36,39,40 used the Stoplight/Traffic Light diet (or a variation of it) as their dietary intervention. Five studies18,26,36- 38 included food or calorie exchange programs while only 1 study22 compared 2 different dietary interventions (low fat vs low carbohydrate). Compliance with the dietary intervention was reported in only 3 studies13,22,28 although reported as having been measured by food diaries or records in 23 studies. An attempt to describe the dietary intake of those participating in or completing an intervention was made in most studies, but the level of detail was of variable quality. A dietitian was reported to be involved in the formulation and/or delivery of the dietary intervention in 13 primary studies9,16- 19,21- 23,25,26,30,34,35 (Table 3).
All 37 RCTs had at least 1 methodological weakness using the JBI critical appraisal tools.6 Although all studies reported randomization of participants to study groups, the method of randomization was described in a minority of studies. Allocation concealment was unclear in 24 studies, adequate in 11 studies,9,10,13,18,23,25,27,31,34,39- 41 and not used in 2 studies. The number of participants in the studies varied but was generally low; 16 had fewer than 50 participants and only 2 studies had sample sizes greater than 150.24,43 The nature of dietary and lifestyle interventions prevents the blinding of participants. Only 1 study mentions assessor blinding and this was only for some of the assessments.23 Outcomes were measured using varying scales among the studies but within each study were reported as being measured in the same manner for all participants. The retention rates were greater than 80% for 17 of 37 RCTs; the lowest rate was 63% at 1-year follow-up after a long intervention period of 14 to 18 months.27 The retention rate was unclear in 1 study and difficult to ascertain at various stages of the intervention and/or follow-up in several studies.
The initial aim of this review was to assess the effectiveness of dietary interventions in the management of overweight and obesity in children and adolescents. Only a very limited number of studies (n = 7)17,21- 23,30,35,41 tested a dietary intervention as the sole component of treatment and compared this with a nonintervention control group or a treatment group with a different approach or intensity. Therefore, we evaluated studies that comprised dietary treatments in combination with other treatments. This resulted in a diverse set of studies that made it impossible to report the effectiveness of dietary interventions in childhood/adolescent overweight and obesity. Although the Forest plot and the results in Table 2 indicate significant reductions in weight in many studies, the study quality, range of outcome measures, and short time span of most studies limit the conclusions that can be drawn. This review further highlights the paucity of quality research in this area that has recently been acknowledged in a Cochrane Review.49 The analyses undertaken suggest that interventions that include a dietary treatment can be effective in this group, but many caveats apply, including most importantly the limited number of quality studies undertaken to date and inadequate long-term follow-up data.
Many studies inadequately described the details of the dietary intervention itself, making replication almost impossible without assistance from study authors. The Stoplight/Traffic Light diet50 was the most commonly reported dietary intervention strategy. It is calorie controlled and categorizes food into different colored groups, red (very limited), green (eat freely), and yellow (limited). It then prescribes a recommended number of daily/weekly servings of each of these food groups. Although the Stoplight diet as described originally by Epstein and Squires51 is frequently used as the dietary prescription in studies conducted in the United States, these studies do predominantly arise from a single research group and the applicability to other settings has not been established. From Table 3, the evolution within the Stoplight/Traffic Light diet is apparent, both in terms of the gradual increase in the number of red foods allowed per week, the reduced focus on tight calorie restriction, a redefinition of the food groups to align more closely to the food pyramid, and the integration of behavioral therapy.
Given that the rationale and objectives of the dietary interventions were rarely described, it was not surprising that the quality of most of the studies was poor in terms of assessing changes in dietary intake in response to the intervention at either the individual or group level. Many studies did not report dietary outcomes, and it is unclear whether food intake data was not collected or was collected but not used because it was of poor quality or because the focus in reporting was only on changes in the primary outcome of weight status.
A challenge for researchers will be in estimating and validating changes in food intake, for example, through the use of dietary biomarkers. The inherent difficulties of dietary assessment in children are well recognized52 and researchers have recently made attempts to validate dietary changes within their studies.22,38
In adult obesity, long-term studies indicate that following a low-fat diet in conjunction with modest daily energy intake restriction (500-1000 kcal/d) results in approximately 5% weight loss after 2 years of follow-up.53 For children, no such evidence is available. Only 3 studies in the meta-analysis had a follow-up of longer than a year. To date, the changes achieved in the dietary intakes of children and adolescents who have successfully lost weight, even in the short-term, have not been described qualitatively or quantitatively. This presents additional challenges when designing dietary interventions for child obesity trials. An attempt to address this gap has been made recently through the use of dietary modeling of food intakes from a reference population to achieve a reduction in energy density.54
In the absence of consensus of expert opinion, a rational approach is to base interventions on observational and epidemiological data. Targeting an increase in fruit and vegetable consumption may be more effective than focusing on a reduction in high-energy, low-nutrient-dense foods.55 Beverage intake may also be a particular problem in children, and consumption of soft drinks has been linked to obesity.48 Behaviors that could be targeted in children to achieve energy reduction and an increase in nutrient density have recently been described.4
There are practical and ethical considerations about conducting a trial with a no-intervention control. However, the following methodological considerations are important when addressing the significant gaps in the evidence and weaknesses in study design on this serious public health issue. A number of trials in this review have no clear reference arm with the treatment (ie, dietary modification) under test often present in both the treatment and the control arms. This makes it difficult for clinicians to extract meaningful results and strategies applicable in general patient settings.
The variety of outcome measures in the trials included body weight, percentage overweight, BMI, and BMI z score with some investigators using a number of different measures. To make direct comparisons of effectiveness, we recommend that BMI z score be reported in all intervention studies in child obesity as a clinically meaningful outcome measure along with the proportion of subjects whose final (follow-up) BMI z score is in the healthy weight range for the appropriate population demographic.
The small arm sizes in many of the trials in this study raise statistical concerns. It would seem that a standard power calculation (using readily available freeware such as PS56 available from http://biostat.mc.vanderbilt.edu/twiki/bin/view/Main/PowerSampleSize) is rarely performed. This is important because the large standard deviations commonly reported indicate that sample sizes well in excess of those commonly used would be required to achieve 80% power. We therefore recommend that studies be powered (ie, employ sample sizes) to have an 80% chance of rejecting a null hypothesis or no-intervention effect if the true intervention effect is more than a specified clinically important value.
Within the 37 trials reported, the age ranges in the subgroups varied from 3 to 12 years to 12 to 18 years with numerous prepubescent and postpubescent subgroup age ranges. One of the authors (B.J.S.) has previously noticed broadly in medical research this use of incompatible age ranges across studies, often by the same researchers. Such inconsistency makes valid meta-analysis difficult to achieve because the assumption then has to be made that age has no effect on weight outcomes when we know that body composition does indeed vary with age and pubertal status. We recommend that pediatric weight loss researchers seek consensus on age group targets for children and adolescents younger than 20 years so that later meta-analyses can be facilitated.
Most of the trials reported results immediately postintervention and at 1 or more follow-up times. There is wide variation in both intervention lengths and the follow-up intervals and duration. Some trials employ follow-up durations as short as 3, 6, or 8 months. We suggest that the success of a weight loss intervention designed ultimately to produce lasting improvements in the health states of obese patients cannot be established clinically, as opposed to statistically, in as short a time frame as 8 months. More consistent intervention and follow-up durations across the field would facilitate informal as well as more formal comparisons of trial results via meta-analysis. We therefore recommend that professional bodies lobby major funding agencies to raise awareness of these issues and advocate for priority funding of large intervention trials with adequate follow-up periods.
The fact that only 2 strongly qualified meta-analyses could be presented in this article is in itself an indication of the significant heterogeneity of interventions for weight management in children and adolescents. Consistency of patient attributes, treatment types, and trial protocols is essential if the meta-analytic calculations are to yield results that can be relied on and incorporated into the knowledge base of the field.
Many trials in weight loss research are small in comparison with other areas, such as cancer treatment research. This may be because funding is generally limited, because it is difficult to recruit into lifestyle interventions particularly for child obesity due to its multifactorial etiology, or because the research is opportunistic.
Although there is insufficient evidence to determine the effectiveness of dietary interventions in treating excessive weight gain in children and adolescents, this systematic review does indicate that interventions that include dietary modification, both as part of an intervention or control and in various treatment combinations, are effective in gaining improvements in weight-related outcomes. This effect appeared to diminish with time. It is difficult to make evidence-based dietary recommendations for child weight management because details of the dietary interventions or the changes made by those who are successful are lacking. The evidence base for pediatric weight management research would increase if researchers considered carefully the methodological quality of their trials, including sample size, at the earliest planning stage. This article has highlighted current gaps in knowledge and limitations in current study designs. Researchers in clinical, public health, and research settings are encouraged to fully describe the rationale and aims of their dietary interventions, evaluate whether subjects adhere to the dietary recommendations, and report dietary outcomes to address the urgent need to develop effective dietary interventions for both childhood obesity treatment and prevention.
Correspondence: Clare E. Collins, PhD, BSc, Dip(Nutr&Diet), Dip(ClinEpi), APD, School of Health Sciences, Mail Box 38, Hunter Building, University of Newcastle, Callaghan, NSW 2308, Australia (firstname.lastname@example.org).
Accepted for Publication: February 16, 2006.
Author Contributions:Study concept and design: Collins, Warren, Neve, McCoy, and Stokes. Acquisition of data: Collins, Warren, Neve, and McCoy. Analysis and interpretation of data: Collins, Warren, Neve, McCoy, and Stokes. Drafting of the manuscript: Collins, Warren, McCoy, and Stokes. Critical revision of the manuscript for important intellectual content: Collins, Warren, Neve, McCoy, and Stokes. Statistical analysis: Neve and Stokes. Administrative, technical, and material support: Collins, Warren, Neve, McCoy, and Stokes.
Funding/Support: This study was supported by a grant from the Joanna Briggs Institute, Adelaide, Australia.
Acknowledgment: We thank Susan Day, AUDIS, for assistance with search and retrieval strategy; Paul Carless, MMedSc(ClinEpi), for methodological and statistical advice; and Tracy Burrows, BHSc(N&D), and Rachel Sutherland, MPH, for identification and retrieval of references and for creating Figure 1 (T.B.).