N U T RI TI O N RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40

Available online at www.sciencedirect.com

ScienceDirect
www.nrjournal.com

Original Research

A meta-analysis of school-based obesity prevention

programs demonstrates limited efficacy of
decreasing childhood obesity

Ling-Shen Hung a , Diane K. Tidwell b,⁎, Michael E. Hall c , Michelle L. Lee d ,

Chiquita A. Briley e , Barry P. Hunt b
a
National Health Research Institutes of Taiwan, 35 Keyan Road, Zhunan Miaoli County 35053 Taiwan
b
Mississippi State University, Box 9805, Food Science, Nutrition and Health Promotion Department, Mississippi State, MS 39762 USA
c
Florida Atlantic University, Department of Exercise Science and Health Promotion, 777 Glades Road, Boca Raton, FL 33431 USA
d
East Tennessee State University, P.O. Box 70690, Department of Allied Health Sciences, Johnson City, TN 37614 USA
e
Tennessee State University, Department of Family and Consumer Sciences, 3500 John A. Merritt Blvd, Nashville, TN 37209 USA

ARTI CLE I NFO A BS TRACT

Article history: Childhood obesity is a global concern. The objectives of this meta-analytical study were to
Received 16 August 2014 evaluate the effectiveness of school-based childhood obesity prevention programs, and to
Revised 7 January 2015 examine program components (moderators). The methods included searching databases
Accepted 13 January 2015 (PubMed, Google Scholar, and the university's EBSCOhost Web service) as well as
handsearching reference lists of articles published in English. Selection criteria for studies
Keywords: to be included in the meta-analysis were limited to studies that reported body mass index
Meta-analysis (BMI) or skinfold thickness as outcome measures and were school-based obesity prevention
Moderators interventions; cross-sectional design studies were excluded. We hypothesized the meta-
Childhood obesity analysis would yield a summary effect size of magnitude which would indicate that school-
School-based programs based interventions have been effective in improving children's BMI or skinfold thickness
values. A total of 26 114 children from 27 school-based childhood obesity prevention
programs provided 54 effect sizes. A random-effects model calculated a small summary
effect size of 0.039 (95% confidence interval −0.013 to 0.092). Heterogeneity among studies
was observed which disappeared after pooling studies that used a randomized controlled
trial design with one program moderator (physical activity or nutrition). We failed to accept
our hypothesis and concluded that overall, school-based interventions have not been
effective for improving body mass index or skinfold thickness to curb childhood obesity;
however, randomized controlled trials that focused on physical activity or nutrition
appeared to produce promising results.
© 2015 Elsevier Inc. All rights reserved.

Abbreviations: BMI, body mass index; CI, confidence interval; k, number of studies (number of effect sizes in a meta-analysis); n,
number of participants in the studies; RCT, randomized controlled trial; SE, standard error; SMD, standardized mean difference.
⁎ Corresponding author. Box 9805, FSNHP, Mississippi State, MS 39762 USA. Tel.: +1 662 325 0239 (Office), +1 662 325 3200 (Department).
E-mail addresses: lhung216@gmail.com (L.-S. Hung), d.tidwell@msstate.edu (D.K. Tidwell), mhall61@fau.edu (M.E. Hall),
LEEML2@mail.etsu.edu (M.L. Lee), cbriley1@Tnstate.edu (C.A. Briley), bhunt@fsnhp.msstate.edu (B.P. Hunt).

http://dx.doi.org/10.1016/j.nutres.2015.01.002
0271-5317/© 2015 Elsevier Inc. All rights reserved.
230 N U TR I TION RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40
1. Introduction
Childhood obesity has increased significantly worldwide,
which has serious consequences on children's health [1] and
health care costs [2,3]. Among US children and adolescents
aged 2 to 19 years, 16.9% were obese and 31.8% were either
overweight or obese in 2009 to 2010 [4]. The number of
children experiencing weight problems is alarming because it
places them at risk for physical and psychosocial problems
[5]. Ebbeling et al [6] reported that obese children experienced
complications related to cardiovascular, endocrine, musculo-
skeletal, and gastrointestinal functions, and also pulmonary
complications such as sleep apnea, asthma, and exercise
intolerance. Dietz [7] reported the most widespread conse-
quences of childhood obesity were psychosocial and obese
children become targets of systematic discrimination and are
more likely to have low self-esteem. Ebbeling et al [6] also
discussed psychosocial consequences of childhood obesity
including sadness, loneliness, and a negative self-image.
There are many causes of childhood obesity, including
genetic, biological, and environmental factors [6]. Although
genetic and biological factors are important, they cannot fully
explain the current global childhood obesity trends. Environ-
mental and societal factors associated with food intake and
physical activity should be the primary focus for understand-
ing the macro-level impacts on obesity [6,8,9]. Schools are
considered a primary setting for implementing childhood
obesity programs because of school meal programs and
physical activity education [8–10]. Research has indicated
that students attending schools with à la carte programs (i.e.,
snack bars, vending machines, kiosks with individual food
and beverage sales) consumed less fruits and vegetables, and
had higher intakes of calories from total and saturated fat
compared to students who were not exposed to à la carte
programs [11]. Also, the availability of snack vending ma-
chines in schools was associated with a higher body mass
index (BMI) z score [12]. Schools that offered nutrition
education programs consistent with the Dietary Guidelines
for Americans [13] had lower rates of obesity [14,15]. A health
education curriculum that highlights the importance of
nutrition and physical activity can help students adopt and
maintain healthy lifestyles regarding eating and physical
activity [16].
School-based programs include structured interventions
delivered in schools by teachers, or other professionals, that
focus on lifestyle factors for obesity prevention, such as
nutrition education classes or online resources available for
students and parents, and physical activity education and
activities that promote movement and exercise. The school-
based programs in the current study referred to studies
implemented in schools (from elementary to high schools)
to improve students' knowledge and lifestyle behaviors for
healthy weight management. Some studies included out-of-
school or at-home activities; however, the foci of the
interventions were implemented at school. Hence, those
studies were included in our study.
There are many individual school-based childhood obesity
studies in the literature; however, there are a limited number
of meta-analytical studies that examine the pooled effective-
ness of studies based on BMI or skinfold thickness outcomes.
Six school-based meta-analyses with BMI or skinfold thick-
ness values as outcomes were identified [17–21]. These meta-
analyses [17–21] present conflicting results. Gonzalez-Suarez
et al [17] reported their meta-analysis showed convincing
evidence that school-based interventions were effective, at
least short-term, in reducing the prevalence of childhood
obesity. In addition, Sobal-Goldberg et al [18] reported more
recent school-based obesity prevention programs with a ran-
domized controlled trial (RCT) design were at least mildly
effective in reducing BMI in children. However, Harris et al [19]
and Kanekar and Sharma [20] reported school-based interven-
tions did not improve BMI. Katz et al. [21] concluded that school-
based interventions with both nutrition and physical activity
interventions were effective and Lavelle et al [22] summarized,
“There is growing evidence that school-based interventions that
contain a physical activity component may be effective in
helping to reduce BMI in children.” Therefore, the objectives of
this study were to evaluate the effectiveness of school-based
childhood obesity prevention programs which reported BMI or
skinfold thickness as outcome measures, and to evaluate
program, or intervention, components (moderators) that may
influence the effectiveness of the program outcomes. We
hypothesized that the meta-analysis would yield a summary
effect size of magnitude which would indicate that school-based
interventions have been effective in improving children's BMI or
skinfold thickness values. The findings of this study could
advance human nutrition by assisting in designing effective
obesity prevention programs for children in school settings.
2. Methods and materials
2.1. Search strategy, inclusion criteria, and selecting
studies for the meta-analysis
Generally accepted guidelines for conducting and reporting a
meta-analysis were used [23]. An electronic database search was
performed to locate school-based childhood obesity prevention
programs published in peer-reviewed English-language
journals. The key words of childhood obesity, school-based,
intervention, physical activity, physical education, and seden-
tary behavior were combined to locate potential studies. Studies
were retrieved from PubMed, Google Scholar, and the
university's comprehensive EBSCOhost Web Service which
included MEDLINE, PsycINFO, ERIC, Agricola, Academic Search
Premier and Academic Search Complete databases, and also
reference lists from childhood obesity systematic literature
reviews. The following inclusion criteria were used for selecting
studies for the meta-analysis: (1) programs must be school-
based; (2) age range of boys or girls must be within 6 to 18 years
old; (3) outcome measurements must be based on BMI or
skinfold thickness values; (4) studies used research designs
such as a RCT, cohort, or interventions with control groups; and
(5) pre-post intervention statistics such as means, standard
deviations (SD), standard errors (SE), sample sizes, and P values
were reported that would enable the calculation of effect sizes.
The year of publication was not a criterion.
After an initial review of the abstracts, the identified
school-based programs that might be included in the meta-
analysis were subjected to a rigorous evaluation for study
 N U T RI TI O N RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40
592 potential relevant studies retrieved from
electronic database search
83 studies evaluated
16 studies added
through manual search
and evaluated
27 studies included in meta-analysis
Fig. 1 – Flowchart of the process for selecting studies for the meta-analysis.
content and data and statistical information. A protocol was
established a priori for data abstraction that included
recording the first author and year of publication, country
where the intervention was implemented, research design,
intervention components (moderators), participants' charac-
teristics, and outcome measurements with statistical infor-
mation for BMI or skinfold thickness values.
2.2. Statistical analyses
2.2.1. Effect size, heterogeneity, and publication bias
The primary outcome measurement for calculating the effect
size, which was a standardized mean difference (SMD), was
mean change of BMI or skinfold thickness. Transforming the
BMI or skinfold thickness values into a SMD scale allowed
data to be pooled and compared. A random effects model with
95% confidence interval (CI) was used to compute a weighted
summary effect size for all studies and to test for heteroge-
neity [24]. The random effects model was used due to the
expected variability among studies. Effect sizes were reported
as SMDs and defined as small (0.20), medium (0.50), and large
(0.80 or greater) [25].
To determine heterogeneity and its magnitude, the values
for Q, P, and I2 were calculated. When P is less than 0.05 for the
chi-squared Q statistic, the null hypothesis for heterogeneity
is rejected and consequently, there is variance among the
studies that is greater than expected; variance is beyond
sampling error or chance [26]. The I2 statistic determines the
magnitude of heterogeneity, if I2 is greater than 50%, the
meta-analysis is considered heterogeneous [24]. Huedo-
Medina et al [27] summarized heterogeneous magnitude as
231
509 studies excluded after
reviewing abstracts
72 studies did not fulfill all the
inclusion criteria and were
excluded
small when I2 is equal to 25%, 50% as medium, and 75%
as large.
Publication bias in meta-analytical research is a problem
that may occur due to the tendency of sometimes publishing
studies which show significant or positive results and not
publishing studies with null or negative results. This can limit
the true findings of a meta-analysis. Publication bias can be
determined by visual examination of the funnel plot. A
skewed funnel plot indicates a biased meta-analysis. The
robustness of the publication bias can be further analyzed by
conducting Classic Fail-Safe N and Duval and Tweedie's Trim
and Fill analyses. The Classic Fail-Safe N test determines the
number of missing null studies (studies that would have an
effect size of zero) required to add into the meta-analysis to
achieve an unbiased result, which is indicated by a nonsig-
nificant P value. Duval and Tweedie's Trim and Fill analysis
addresses the best estimate of the unbiased effect size by
determining the number of studies that should be trimmed
(removed or altered) and filled (imputed studies), which
produces a shift in the funnel plot to correct the variance.
The visual presentation of the effect size between the
observed and imputed studies determines how small or
large a shift is required to correct for bias. According to
Borenstein et al [26], “If this shift is trivial, then one can have
more confidence that the reported effect is valid” (p. 286).
2.2.2. Subgroup analyses and moderators
When Q is significant and the I2 value is large, variation exists
among the studies and subgroup analysis is warranted which
may explain the heterogeneity [26]. The moderators that were
planned for investigation of heterogeneity using subgroup
 232
N U TR I TION RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40
Table 1 – Characteristics and components of school-based studies included in the meta-analysis
First Country Research Components of school-based obesity prevention programs
author design
[reference] No. of Mean age Program Nutrition Physical Parental Specialist Theory Outcome measurement,
children category duration activity involvement involvement based BMI or skinfold thickness
of children (y) (e.g., nurse, and positive (+) or negative
(y) nutritionist) (−) program effect

Alexandrov Moscow, USSR RCT 766 ≥10 ≥1 Yes Yes Yes No No BMI (+)
[23]
Angelico Italy Cohort 150 <10 ≥1 Yes Yes Yes No No BMI (−)
[24]
Bayne- New York, USA RCT 442 ≥10 ≥1 Yes Yes No No No BMI (+)
Smith [25]
Bush [26] DC, USA RCT 431 ≥10 ≥1 Yes Yes Yes Yes Yes Skinfolds (−)
Caballero Arizona, New Mexico, RCT 1409 <10 ≥1 Yes Yes Yes No Yes BMI (+)
[27] South Dakota, USA
Carrel [28] Wisconsin, USA RCT 50 ≥10 <1 Yes Yes No No No BMI (−)
Donnelly Nebraska, USA RCT 108 <10 ≥1 Yes Yes No Yes No BMI (+)
[29]
Hansen [30] Denmark North Mix of case control 132 ≥10 <1 No Yes No No No BMI (−)
and RCT
Harrell [31] Carolina, USA RCT 1274 <10 <1 Yes Yes No Yes No BMI (−)
Harrell [32] Mississippi, USA Intervention with 186 ≥10 <1 Yes Yes Yes Yes No BMI (−)
control group
James [33] England RCT 574 <10 ≥1 Yes No No No No BMI (+)
James [34] England RCT 434 ≥10 ≥1 Yes No No No No BMI (+)
Kain [35] Chile RCT 3086 <10 <1 Yes Yes Yes Yes No BMI Boys (+)
BMI Girls (−)
Killen [36] California, USA RCT 1130 ≥10 <1 Yes Yes No No Yes BMI (+)
Kriemler Switzerland RCT 502 <10 <1 No Yes No No No BMI (+)
[37]
Luepker [38] California, Louisiana, RCT 3959 <10 ≥1 Yes Yes Yes No Yes BMI (−)
Minnesota, Texas, USA
Manios [39] Greece RCT 393 <10 ≥1 Yes Yes Yes No Yes BMI (+)
Manios [40] Greece RCT 641 ≥10 ≥1 Yes Yes Yes Yes Yes BMI (+)
Nader [41] USA RCT 3660 <10 ≥1 Yes Yes Yes No Yes BMI (+)
Pangrazi Arizona, USA RCT 599 <10 <1 No Yes No No No BMI (+)
[42]
Puska [43] North Karelia, Finland RCT 851 <10 ≥1 Yes Yes Yes Yes No BMI (+)
Ritenbaugh New Mexico, USA Intervention with 102 ≥10 ≥1 Yes Yes No Yes No BMI (−)
[44] control group
Robinson California, USA RCT 192 ≥10 <1 No No Yes No Yes BMI (+)
[45]
Sallis [46] California, USA RCT 955 <10 ≥1 No Yes No No No Skinfolds: Boys (+) for both groups

N U T RI TI O N RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40
Girls (+) for teacher-led
Girls (−) for specialist-led
Trevino [47] Texas, USA RCT 387 <10 <1 Yes Yes Yes Yes Yes BMI Boys (−)
BMI Girls (+)
Vandongen Australia RCT 971 <10 <1 Yes Yes Yes Yes No BMI (+) for 6 groups and (−) for 4
[48] groups
Wardle [49] London, England Cohort 2730 <10 ≥1 No Yes No No No BMI (+)

233
 234
N U TR I TION RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40
Fig. 2 – Forest plot of the meta-analysis. Each study is identified by first author and reference. The individual effect sizes are identified as “Std diff in means” with the standard
error, variance, lower and upper limits (confidence intervals), Z value, and P values provided for each study. The overall summary effect size of the meta-analysis (.039) is noted
on the bottom line.
 N U T RI TI O N RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40 235

Table 2 – Results of the overall summary effect size and the subgroup moderators
k (n) a Effect size SE Z value (2-tailed Q (df) c P (between
(95% CI) b test of null) subgroup effects) d

Summary effect size 54 (26 114) 0.039 (−0.013 to 0.092) 0.027 1.458 167.77*** (53)
Program duration 0.17 (1) .680
<1 year 28 (8509) 0.038 (−0.004 to 0.080) 0.021 1.776
≥1 year 26 (17 605 0.017 (−0.076 to 0.109) 0.047 0.357
Age 0 (1) .989
<10 years old 22 (14 229) 0.036 (−0.095 to 0.167) 0.067 0.536
≥10 years old 32 (11 885) 0.035 (0.003 to 0.067) 0.017 2.111*
Nutrition component 2.56 (1) .110
No 17 (5110) 0.088 (0.024 to 0.151) 0.033 2.690**
Yes 37 (21 004) 0.013 (−0.053 to 0.079) 0.034 0.379
Physical activity component - - - - -
No 3 (1200) 0.029 (−0.026 to 0.084) 0.028 1.038
Yes 51 (24 914)
Parental involvement 0.58 (1) .447
No 20 (9032) 0.062 (0.015 to 0.109) 0.024 2.577*
Yes 34 (17 082) 0.028 (−0.045 to 0.102) 0.037 0.755
Specialist involvement 0.27 (1) .605
No 29 (18 077) 0.052 (−0.034 to 0.137) 0.044 1.185
Yes 25 (8037) 0.026 (−0.017 to 0.069) 0.022 1.204
Theory-based program 0.95 (1) .329
No 43 (15 321) 0.013 (−0.062 to 0.088) 0.038 0.333
Yes 11 (10 793) 0.059 (0.006-0.111) 0.027 2.178*
Research Design 6.50* (1) .011*
RCT 45 (22 814) 0.050 (0.024-0.076) 0.014 3.699***
Other 9 (3300) −0.378 (−0.707 to −0.050) 0.167 −2.259*
No. of program components/moderators 5.85* (1) .016*
One 18 (5926) 0.112 (0.054-0.170) 0.030 3.792***
Two or more 36 (20 188) 0.003 (−0.064 to 0.070) 0.034 0.082

k, number of studies or effect sizes if a study has more than one effect size such as results reported separately for boys and girls.
n, number of children represented in the studies.
*P < .05; **P < .01; ***P < .001.
a
Analyses were not conducted with subgroups containing less than 5 comparisons.
b
Effect sizes are reported as weighted standardized mean differences with 95% confidence intervals.
c
Significant Q values indicate heterogeneity, df is degrees of freedom. The summary Q is the overall heterogeneity for the study and the other
Q values are the between-group results, which tested heterogeneity between the two categorical subgroup moderators.
d
A significant P value indicates a difference between the two categorical moderator subgroup effect sizes.

analysis were (1) participants' age (less than 10 years old
versus 10 years or older), (2) program duration (less than
1 year versus 1 year or longer), (3) a nutrition component
versus not having a nutrition component in the intervention,
(4) a physical activity component versus no physical activity
component, (5) parental involvement versus no parental
involvement, (6) specialists such as the involvement of
nutritionists, nurses, and physicians versus no involvement
of specialists, (7) research design (RCTs versus non-RCT
studies), and (8) theory based versus programs that were not
designed based on a social science or health promotion
theory. The mixed effects model was used for subgroup
analysis as recommended by Borenstein et al [26]. Compre-
hensive Meta-Analysis software (Version 3, 2013, Biostat, Inc,
Englewood, NJ, USA) was used for all statistical analyses.
performed based on the inclusion criteria which excluded 581
studies. Twenty-seven studies [28–54] were included in the
meta-analysis with publication dates ranging from 1982 to
2010. The meta-analysis was completed in 2010. Among the
27 included studies [28–54], six [30,36,43,46,51,53] required
additional hand calculations for effect sizes to be computed,
such as calculating pre-post correlations, SD, or SE with the
use of basic statistical equations using the information
reported in the studies.
A total of 26 114 children represented in 27 school-based
childhood obesity prevention programs resulted in 54 effect
sizes. Some studies yielded more than one effect size because
they reported male and female BMI changes separately or
measured program effect at different time points. Characteris-
tics of the studies including program components and outcome
measurement information are summarized in Table 1. The
effect sizes are demonstrated in a forest plot (Fig. 2).

3. Results
The electronic database search generated 592 potential
relevant studies and a manual search of reference lists
added an additional 16 studies (Fig. 1). An evaluation was
3.1. Effect size and heterogeneity
The weighted summary effect size for all 54 effects was small
(d = 0.039, 95% CI −0.013 to 0.092, P = .145, Fig. 2 denotes the
236 N U TR I TION RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40
Funnel Plot of Standard Error by Std diff in means
0
1
Standard Error
-4 -3 -2
Fig. 3 – Funnel plot of the meta-analysis. The hollow circles represent the studies in the meta-analysis and the dark circles
represent the imputed studies that would correct the publication bias.
effect size as “Std diff in means”). This indicated there was not
strong evidence that school-based programs have been
effective in obesity prevention. Thus, we rejected our hypoth-
esis and determined that overall, school-based obesity
prevention programs have not been effective in improving
BMI or skinfold thickness parameters. The individual effect
sizes ranged from −3.705 (least effective) to 0.368 (most
effective). Most effect sizes for the programs were small with
values less than 0.20; only three studies [44, 47 included two
of the girls PLAY groups] had interventions that produced
effect sizes greater than 0.30 (Fig. 2). This meta-analysis was
heterogeneous (Q(df) = 167.77(53), P < .001) with an I2 value of
68.41%, which implied that variation existed among the
interventions and subgroup analysis investigating the inter-
vention components (moderators) should be conducted [26].
3.2. Subgroup analysis and heterogeneity
Results from the subgroup analysis determined that research
design had a significant between-group effect (P = .011,
Table 2). Studies using a RCT design were more effective in
improving BMI or skinfold thickness values in children
compared to non-RCT interventions. The research design
moderator was investigated further and separate meta-
analyses were conducted using the RCTs and non-RCT
studies. The meta-analysis with the RCTs (k = 45) yielded a
small summary effect size of 0.050 (0.014 SE, P < .001);
however, heterogeneity was deemed nonexistent as Q(df)
was 39.72(44) (P = 0.656) with an I2 value of 0%. The RCT
studies had the least amount of variability and the narrowest
range for the 95% CI, which did not include absolute zero, and
a significant Z value of 3.699 (P < .001) further indicated that
the effect size was not zero (Table 2). In comparison, the
meta-analysis of the non-RCT research designs (interventions
with control groups and cohorts, k = 9) resulted in a negative
effect size of −0.378 (0.167 SE, P = 0.024) with extreme
heterogeneity (Q(df) = 125.21(8), P < .001) of very high
-1 0 1 2 3 4
Std diff in means
magnitude with an I2 value of 93.61%. The non-RCT group
had the largest 95% CI range (Table 2), which also did not
include absolute zero but the values were negative indicating
they were the least effective programs.
A moderator was added post hoc as the literature indicated
that having only one program component (physical activity or
dietary intervention) may be more effective in childhood
obesity prevention programs, although evidence was insuffi-
cient in school-based studies that a combination of diet and
physical activity was more effective at preventing obesity or
overweight [55]. This led us to examine the number of
program components as a moderator and the between-
group effect was significant (P = .016, Table 2). A meta-
analysis with interventions that contained only one program
component (k = 18) yielded a summary effect size of 0.112
(0.030 SE, Z = 3.792, P < .001) and heterogeneity was
nonexistent (Q(df) = 10.88(17), P = 0.863) with an I2 value of
0%. Conversely, a meta-analysis of interventions with two or
more program components (k = 36) resulted in a very small
effect size of 0.003 (0.034 SE, P = 0.935) with extreme
heterogeneity (Q(df) = 150.07(35), P < .001) of high magnitude
with an I2 value of 76.68%. Programs that contained only one
moderator (physical activity or nutrition) and a RCT design
produced an effect size of 0.168 (95% CI = 0.085-0.252) with no
heterogeneity (Q(df) = 6.08(12), P = 0.912, I2 = 0%). After
additional investigation with all the moderators, it was
observed that interventions with two or more components
implemented with a non-RCT research design produced the
highest heterogeneity with an I2 value of 96.72%, which
indicated that it strongly contributed to overall heterogeneity
of the meta-analysis.
The other between-group moderator components did not
have significantly different effect sizes (Table 2). Neither
categorical groups of interventions with a program duration of
less than one year or programs implemented for one year or
longer were effective in overall improvement of children's BMI
or skinfold thickness status. The between-group analysis for
 N U T RI TI O N RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40
program duration had variability between the 2 SEs; however,
the effect sizes were similar and therefore, the between-group
analysis for program duration was not significant (P = .680).
Similarly, the effect sizes were not significantly different when
comparing interventions that targeted children less than
10 years old vs interventions that targeted children 10 years or
older (P = .989), although there was less variability with the older
children as shown by a smaller SE (Table 2). Programs that
included a nutrition intervention component did not produce a
significantly different effect size than interventions without a
nutrition component (P = .110).
Subgroup analysis for studies with a physical activity
component (k = 51) and those without a physical activity
component (k = 3) was not conducted due to the inadequate
number of studies without physical activity. The Agency for
Healthcare Research and Quality [56] recommended that
subgroups should have a minimum of four studies for
categorical subgroup analysis. Other subgroup analyses were
not significant such as the between-group effect for parental
involvement versus no parental involvement in the interven-
tions (P = .447), and programs with specialists such as
nutritionists, nurses, and physicians that assisted in the
school-based programs did not have a greater effect com-
pared to programs without specialists (P = .605). Lastly,
interventions that were based on a social science or health
promotion theory did not have a larger effect size compared
to non-theory based programs (P = .329, Table 2).
3.3. Publication bias, Classic Fail-Safe N, and Duval and
Tweedie's Trim and Fill analysis
An asymmetrical funnel plot was produced by this meta-
analysis (the hollow circles in Fig. 3), which indicated
publication bias [26]. Most studies included in the meta-
analysis were large studies and therefore, plotted on top of
the funnel. Smaller studies were lacking in this meta-
analysis. Results of the Classic Fail-Safe N test (P = .047)
demonstrated that 2 small null studies were missing from the
meta-analysis to bring the P value to greater than α (.05). The
Duval and Tweedie's Trim and Fill analysis indicated that 15
studies should be trimmed from the left side of the funnel plot
(Fig. 3). After trimming those studies, the adjusted effect size
would be .008, which is the best estimate of the unbiased
effect size. As presented in Fig. 3, a small shift of the summary
effect size is observed on the bottom of the graphic between
the present study (represented by hollow diamond) and the
estimated unbiased effect size (dark diamond) and therefore,
the impact of publication bias for the present study may be
considered trivial [26].
4. Discussion
This meta-analysis observed a small summary effect size of
0.039 for the 27 studies with 54 effects, which indicated that
school-based interventions have not been effective in im-
proving BMI or skinfold thickness values to help curb
childhood obesity. Therefore we rejected our hypothesis and
determined that overall, school-based interventions have not
been effective for improving BMI or skinfold thickness to curb
237
childhood obesity. This finding was consistent with other
school-based meta-analyses conducted by Harris et al [19]
which included 18 studies and Kanekar and Sharma's meta-
analysis of five studies [20]. Conversely, Katz et al [21]
concluded from their meta-analysis of eight studies that
school-based interventions combining nutrition and physical
activity interventions were effective at achieving weight
reduction. The largest meta-analysis of 43 school-based
studies with 60 effect sizes reported a small summary effect
size of -0.17 [22]. A meta-analysis of school-based obesity
prevention RCTs observed that more recent, longer duration,
and comprehensive interventions which included parental
support were at least mildly effective in reducing BMI [18]. Our
results indicated that school-based intervention RCTs with a
physical activity or nutrition component were more effective
in improving BMI or skinfold measurement values compared
to interventions with multi-components and a non-RCT
research design.
The intervention in the present meta-analysis with the
greatest effect size of 0.368 and concomitant improvement in
BMI was the girls PLAY and physical education intervention
conducted by Pangrazi et al [47], which was a RCT and
included children less than 10 years old and focused only on
physical activity with a program duration of less than one
year. The least effective intervention with the smallest effect
size of -3.705 was computed for the study conducted by
Angelico et al [29]. This study was a cohort research design
and also included children less than 10 years old but had a
program duration of more than one year and included
components of nutrition, physical activity, and parental
involvement [29].
Subgroup analyses indicated duration of the intervention,
children's age, parental or specialists' involvement, or a
nutrition component did not yield significant between-group
effects. Other meta-analyses have reported conflicting re-
sults. Harris et al [19] reported programs with a shorter
duration (less than one year) produced a greater effect size
regarding improvement of students' BMI compared to pro-
grams with a longer duration. However, Kamath et al [57]
demonstrated that program effect was greater for interven-
tions implemented for more than six months compared to
studies with durations of less than 6 months in a meta-
analysis of various types of interventions that included both
school-based and non-school-based interventions. Overall it
is difficult to determine optimal program duration for
improving students' BMI due to different time periods used
for interventions. It would be beneficial if future studies could
determine optimal program durations, especially for different
children's ages.
Although the present meta-analysis indicated that
children's age did not have an impact on program outcome,
Kamath et al [57] reported trials with children had a larger
reduction in sedentary activity than trials with adolescents.
However, the meta-analysis conducted by Stice et al [58]
suggested that older children were better in understanding
the concepts delivered in their intervention and a larger
effect size would be detected for programs that targeted
older children.
Unexpectedly, overall results of our meta-analysis indicat-
ed that studies promoting a healthy diet did not yield a
238 N U TR I TION RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40
greater effect size than programs not targeting dietary
improvement. The influence of physical activity could not be
determined due to an inadequate number of school-based
studies without a physical activity component. Findings
associated with dietary and physical activity components
were demonstrated in a meta-analysis by Stice et al [58] with
various types of school-based and non-school-based inter-
ventions; they suggested the results could be due to compen-
sation. Students might have positive behavior changes
following the interventions regarding nutrition and physical
activity at school, but they altered their behaviors outside
school toward unhealthy diet and physical inactivity behav-
iors, and consequently, the effects of the program were
negated [58]. In addition, perhaps the amount of physical
activity in interventions and the adherence of participants in
the intervention groups were not sufficient to achieve
significant improvements in BMI values [19].
Although our study demonstrated that overall, a nutrition
component in the interventions did not yield a significant
impact on improving children's weight status, it should not be
abandoned from childhood obesity prevention programs.
Maintaining a healthy body weight requires energy balance.
Factors related to healthy nutrition behaviors should be
involved in obesity prevention [6,9,10]. It may be more
important for future meta-analyses of childhood obesity
research to investigate factors that impede or assist healthy
or unhealthy diet and physical activity behavioral changes in
school settings. For instance, Kamath et al [57] analyzed the
effect size of different factors that increased or decreased
participants' healthy or unhealthy diets and physical activi-
ties. Their study suggested that programs involving reinforce-
ment yielded a greater program effect for increasing healthy
dietary behaviors compared to programs that did not involve
reinforcements, such as incentives and rewards [57]. Kamath
et al [57] advocated that interventions should include
reinforcement and multiple cognitive components, such as
self-efficacy and decision making, to increase children's
physical activity. Programs that were implemented for a
longer duration (more than six months), targeted children
versus adolescents, and involved multiple cognitive compo-
nents versus one or no cognitive components, had a greater
program effect regarding decreasing participants' sedentary
activity [57].
Consistent with our study, Stice et al [58] also observed
that parental involvement did not produce a greater program
effect compared to programs without parental involvement.
This may suggest that parental participation in the program
activities was insufficient. Ebbeling et al [6] iterated that
parental support and healthy family eating habits could lower
children's risk for developing obesity. If parents take an active
role in promoting and monitoring their children's lifestyles
regarding healthy eating and physical activity, it may help
children maintain healthy body weights. However, the
influence of the home environment on childhood obesity
prevention is not fully understood and requires more in-
depth research [58].
Our subgroup analysis implied that specialist involvement
was not an important program component in childhood
weight control interventions. The same finding was detected
in the meta-analysis conducted by Stice et al [58]. The
specialists may not have been involved at the level required
to produce a treatment effect, such as a nutritionist delivering
multiple interactive nutrition sessions versus only being able
to provide one or two limited nutrition sessions to the
children. Overall, most of the studies used teachers to deliver
the programs to the students.
Programs based on the social cognitive theory or social
learning theory did not provide more effective constructs at
changing behaviors. Jeffery [59] stated that successful weight
loss programs did not derive from cognitive decision-making
models, and suggested that applied theories such as classical
behavior theory, communications theory, and learning theory
may be more helpful than the social cognitive theories.
Indeed, Davison and Birch [60] propose that a multifaceted
theoretical system is needed such as the ecological systems
theory since childhood obesity involves a complex, interre-
lated set of factors.
The present meta-analysis had publication bias due to
the absence of smaller studies as indicated by the funnel
plot, although only a small shift of the summary effect size
was detected in the plot [26]. Additional studies that
reported BMI or skinfold thickness outcome measures
could not be located in the literature. In contrast, a separate
meta-analysis conducted using only the studies with a RCT
design and one moderator produced more robust results
with no evidence of publication bias. It is well known that
RCTs are equated with strong evidence and high quality
research. The use of one or more components such as
physical activity, nutrition, or parental involvement in an
intervention for obesity prevention in children warrants
additional research. When considering the overall results
produced by this meta-analysis and the small effect size that
was observed, we concluded that school-based interven-
tions have not been effective in combatting childhood
obesity. However, school-based RCTs that focused only on
physical activity or nutrition appeared to produce promising
results for impacting childhood obesity.
The majority of the school-based obesity prevention
programs included programs with multi-components and
the variables of the interventions were implemented using
various methods. Therefore, it became difficult to determine
why some variables produced effective program outcomes
while others did not. When comparing our study with other
meta-analytic childhood obesity prevention research, we
discovered inconsistencies in the literature and different
comparison methods used for moderators, such as different
cut-off points for program duration and participants' ages
regarding younger versus older children. This made it
difficult for determining the best age range of children and
duration of the program for effective pediatric obesity
prevention interventions. Also, other moderators may be
important components in interventions which were not
investigated in this meta-analysis, such as television and
computer screen viewing, food and micronutrient intakes,
and family dynamics. As with all meta-analyses, the
conclusions that can be drawn are limited by the robustness
of the studies used for the analyses. It should also be noted
that meta-analyses are observational studies and not
experimental studies; therefore, cause and effect cannot be
determined.
 N U T RI TI O N RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40
Acknowledgment
The authors thank Wes Schilling for sharing his statistical
expertise. This research was funded by the Department of
Food Science, Nutrition and Health Promotion and the
Mississippi Agricultural and Forestry Experiment Station.
Approved for publication as Journal Article No. J-12388 of the
Mississippi Agricultural and Forestry Experiment Station,
Mississippi State University.
REFERENCES
[1] Han JC, Lawlor DA, Kimm SYS. Childhood obesity. Lancet
2010;375:1737–48.
[2] Kraak VI, Story M. A public health perspective on healthy
lifestyles and public-private partnerships for global
childhood obesity prevention. J Am Diet Assoc 2010;110(2):
192–200.
[3] Finkelstein EA, Ruhm CJ, Kosa KM. Economic causes and
consequences of obesity. Annu Rev Public Health 2005;26:
239–57.
[4] Ogden CL, Carroll MD, Kit BK, Flegal KM. Prevalence of obesity
and trends in body mass index among US children and
adolescents, 1999-2010. JAMA 2012;307(5):483–90.
[5] Daniels SR. Complications of obesity in children and
adolescents. Int J Obes 2009;33:S60–5.
[6] Ebbeling CB, Pawlak DB, Ludwig DS. Childhood obesity:
public-health crisis, common sense cure. Lancet 2002;360:
473–82.
[7] Dietz WH. Health consequences of obesity in youth:
childhood predictors of adult disease. Pediatrics 1998;101:
518–25.
[8] Sallis JF, Glanz K. Physical activity and food environments:
solutions to the obesity epidemic. Milbank Q 2009;87(1):
123–54.
[9] Koplan JP, Liverman CT, Kraak VI. Preventing childhood
obesity: health in the balance: executive summary. J Am Diet
Assoc 2005;15(1):131–8.
[10] Story M, Nanney MS, Schwartz MB. School and obesity
prevention: creating school environments and policies to
promote healthy eating and physical activity. Milbank Q
2009;87(1):71–100.
[11] Kubik MY, Lytle LA, Hannan PJ, Perry CL, Story M. The
association of the school food environment with dietary
behaviors of young adolescents. Am J Public Health 2003;93
(7):1168–73.
[12] Fox MK, Dodd AH, Wilson A, Gleason PM. Association
between school food environment and practices and body
mass index of US public school children. J Am Diet Assoc
2009;109(Suppl. 1):S108–17.
[13] US Department of Agriculture, US Department of Health and
Human Services. Dietary Guidelines for Americans, 2010. 7th
ed. Washington, DC: US Government Printing Office; 2010.
[14] Veugelers PJ, Fitzgerald AL. Effectiveness of school programs
in preventing childhood obesity: a multilevel comparison.
Am J Public Health 2005;95(3):432–5.
[15] Foster GD, Sherman S, Borradaile KE, Grundy KM, Vander
Veur SS, et al. A policy-based school intervention to prevent
overweight and obesity. Pediatrics 2008;121(4):794–802.
[16] Story M, Kaphingst KM, French S. The role of schools in
obesity prevention. Future Child 2006;16(1):109–42.
[17] Gonzalez-Suarez C, Worley A, Grimmer-Somers K, Dones V.
School-based interventions on childhood obesity: a
meta-analysis. Am J Prev Med 2009;37(5):418–27.
239
[18] Sobal-Goldberg S, Rabinowitz J, Gross R. School-based obesity
prevention programs: a meta-analysis of randomized con-
trolled trials. Obesity 2013;21:2422–8.
[19] Harris KC, Kuramoto LK, Schulzer M, Retallack JE. Effect of
school-based physical activity interventions on body mass
index in children: a meta-analysis. CMAJ 2009;180(7):
719–26.
[20] Kanekar A, Sharma M. Meta-analysis of school-based child-
hood obesity interventions in the U.K. and U.S. Int Q
Community Health Educ 2009;29(3):241–56.
[21] Katz DL, O'Connell M, Njike VY, Yeh M-C, Nawaz H. Strategies
for the prevention and control of obesity in the school
setting: systematic review and meta-analysis. Int J Obes 2008;
32:1780–9.
[22] Lavelle HV, Mackay DF, Pell JP. Systematic review and meta-
analysis of school-based interventions to reduce body mass
index. J Public Health 2012;34(3):360–9.
[23] Higgins JPT, Green S, editors. Cochrane Handbook for
Systematic Reviews of Interventions Version 5.1.0The
Cochrane Collaboration; 2011 [Available from: http://www.
cochrane-handbook.org].
[24] Borenstein M, Higgins JPT. Meta-analysis and subgroups. Prev
Sci 2013;14:134–43.
[25] Cohen J. A power primer. Psychol Bull 1992;112:155–9.
[26] Borenstein M, Hedges LV, Higgins JPT, Rothstein HR. Intro-
duction to meta-analysis. West Sussex, England: John Wiley
& Sons Ltd.; 2009.
[27] Huedo-Medina TB, Sánchez-Meca J, Marín-Martínez F, Botella
J. Assessing heterogeneity in meta-analysis: Q statistic or I2
index? Psychol Methods 2006;11(2):193–206.
[28] Alexandrov AA, Maslennikova GY, Kulikov SM, Propirnij GA,
Perova NV. Primary prevention of cardiovascular disease:
3-year intervention results in boys of 12 years of age. Prev
Med 1992;21:53–62.
[29] Angelico F, Del Ben M, Fabiani L, Lentini P, Pannozzo F,
Urbinati GC. Management of childhood obesity through a
school-based programme of general health and nutrition
education. Public Health 1991;105:393–8.
[30] Bayne-Smith M, Fardy PS, Azzollini A, Magel J, Schmitz KH,
Agin D. Improvements in heart health behaviors and
reduction in coronary artery disease risk factors in urban
teenaged girls through a school-based intervention: the
PATH program. Am J Public Health 2004;94(9):1538–43.
[31] Bush PJ, Zuckerman AE, Theiss PK, Taggart VS, Horowitz C,
Sheridan MJ, et al. Cardiovascular risk factor prevention in
black schoolchildren: Two-year results of the “know your
body” program. Am J Epidemiol 1989;129(3):466–82.
[32] Caballero B, Clay T, Davis SM, Ethelbah B, Rock BH, Lohman
T, et al. Pathways: a school-based, randomized controlled
trial for the prevention of obesity in American Indian
schoolchildren. Am J Clin Nutr 2003;78:1030–8.
[33] Carrel AL, Clark RR, Peterson SE, Nemeth BA, Sullivan J, Allen
DB. Improvement of fitness, body composition, and insulin
sensitivity in overweight children in a school-based exercise
program. Arch Pediatr Adolesc Med 2005;159:963–8.
[34] Donnelly JE, Jacobsen DJ, Whatley JE, Hill JO, Swift LL,
Cherrington A, et al. Nutrition and physical activity program
to attenuate obesity and promote physical and metabolic
fitness in elementary school children. Obes Res 1996;4(3):
229–43.
[35] Hansen HS, Froberg K, Hyldebrandt N, Nielsen JR. A con-
trolled study of eight months of physical training and
reduction of blood pressure in children: the Odense school-
child study. Br Med J 1991;303:682–5.
[36] Harrell JS, McMurray RG, Bangdiwala SI, Frauman AC, Gansky
SA, Bradley CB. Effects of a school-based intervention to
reduce cardiovascular disease risk factors in elementary-
school children: the cardiovascular health in children (CHIC)
study. J Pediatr 1996;128(6):797–805.
240 N U TR I TION RE S E ARCH 3 5 ( 2 0 15 ) 2 2 9–2 40
[37] Harrell TK, Davy BM, Stewart JL, King DS. Effectiveness of
a school-based intervention to increase health knowledge
of cardiovascular disease risk factors among rural Mis-
sissippi middle school children. South Med J 2005;98(12):
1173–80.
[38] James J, Thomas P, Cavan D, Kerr D. Preventing childhood
obesity by reducing consumption of carbonated drinks:
cluster randomised controlled trial. Br Med J 2004;328:1237–9.
http://dx.doi.org/10.1136/bmj.38077.458438.EE.
[39] James J, Thomas P, Kerr D. Preventing childhood obesity: two
year follow-up results from the Christchurch obesity pre-
vention programme in schools (CHOPPS). Br Med J 2007;335:
762. http://dx.doi.org/10.1136/bmj.39342.571806.55.
[40] Kain J, Uauy R, Albala C, Vio F, Cerda R, Leyton B. School-
based obesity prevention in Chilean primary school children:
methodology and evaluation of a controlled study. Int J Obes
2004;28:483–93.
[41] Killen JD, Telch MJ, Robinson TN, Maccoby N, Taylor B,
Farquhar JW. Cardiovascular disease risk reduction for tenth
graders: a multiple-factor school-based approach. JAMA 1988;
260(12):1728–33.
[42] Kriemler S, Zahner L, Schindler C, Meyer U, Hartmann T,
Hebestreit H, et al. Effect of school based physical activity
programme (KISS) on fitness and adiposity in primary
schoolchildren: cluster randomized controlled trial. Br Med J
2010;340:c785. http://dx.doi.org/10.1136/bmj.c785.
[43] Luepker RV, Perry CL, McKinlay SM, Nader PR, Parcel GS,
Stone EJ, et al. Outcomes of a field trial to improve children's
dietary patterns and physical activity: the Child and Adoles-
cent Trial for Cardiovascular Health (CATCH). JAMA 1996;275
(10):768–76.
[44] Manios Y, Kafatos A, Mamalakis G. The effects of a health
education intervention initiated at first grade over a 3 year
period: physical activity and fitness indices. Health Educ Res
1998;13(4):593–606.
[45] Manios Y, Moschandreas J, Hatzis C, Kafatos A. Health and
nutrition education in primary schools of Crete: changes in
chronic disease risk factors following a 6-year intervention
programme. Br J Nutr 2002;88:315–24.
[46] Nader PR, Stone EJ, Lytle LA, Perry CL, Osganian SK, Kelder S,
et al. Three-year maintenance of improved diet and physical
activity: The CATCH Cohort. Arch Pediatr Adolesc Med 1999;
153(7):695–704.
[47] Pangrazi RP, Beighle A, Vehige T, Vack C. Impact of promoting
lifestyle activity for youth (PLAY) on children's physical
activity. J Sch Health 2003;73(8):317–21.
[48] Puska P, Vartiainen E, Pallonen U, Salonen JT, Pöyhiä P, Koskela
K, et al. The North Karelia youth project: evaluation of two
years of intervention on health behavior and CVD risk factors
among 13- to 15-year old children. Prev Med 1982;11:550–70.
[49] Ritenbaugh C, Teufel-Shone NI, Aickin MG, Joe JR, Poirier S,
Dillingham DC, et al. A lifestyle intervention improves
plasma insulin levels among Native American high school
youth. Prev Med 2003;36:309–19.
[50] Robinson TN. Reducing children's television viewing to
prevent obesity: a randomized controlled trial. JAMA 1999;282
(16):1561–7.
[51] Sallis JF, McKenzie TL, Alcaraz JE, Kolody B, Faucette N, Hovell
MF. The effects of a 2-year physical education program
(SPARK) on physical activity and fitness in elementary school
students. Am J Public Health 1997;87(8):1328–34.
[52] Trevino RP, Hernandez AE, Yin Z, Garcia OA, Hernandez I.
Effect of the Bienestar health program on physical fitness in
low-income Mexican American children. Hisp J Behav Sci
2005;27(1):120–32.
[53] Vandongen R, Jenner DA, Thompson C, Taggart AC, Spickett
EE, Burke V, et al. A controlled evaluation of a fitness and
nutrition intervention program on cardiovascular health in
10- to 12-year-old children. Prev Med 1995;24:9–22.
[54] Wardle J, Brodersen NH, Boniface D. School-based physical
activity and changes in adiposity. Int J Obes 2007;31:1464–8.
[55] Wang Y, Wu Y, Wilson RF, Bleich S, Cheskin L, Weston C,
et al. Childhood obesity prevention programs: comparative
effectiveness review and meta-analysis. Comparative effec-
tiveness review No. 115. (Prepared by the Johns Hopkins
University Evidence-based Practice Center under contract No.
290-2007-10061-I.) AHRQ Publication No. 13-EHC081-EF.
Rockville, MD: Agency for Healthcare Research and Quality;
2013 [Available from: www.effectivehealthcare.ahrq.gov/
reports/final.cfm].
[56] Agency for Healthcare Research and Quality. Methods guide
for effectiveness and comparative effectiveness reviews.
AHRQ Publication No. 10(14)-EHC063-EF. Rockville, MD:
Agency for Healthcare Research and Quality; 2014 [Available
from: http://www.effectivehealthcare.ahrq.gov/ehc/
products/60/318/CER-Methods-Guide-140109.pdf].
[57] Kamath CC, Vickers KS, Ehrlich A, McGovern L, Johnson J,
Singhal V, et al. Behavioral interventions to prevent childhood
obesity: a systematic review and metaanalyses of randomized
trials. J Clin Endocrinol Metab 2008;93(12):4606–15.
[58] Stice E, Shaw H, Marti CN. A meta-analytic review of obesity
prevention programs for children and adolescents: The
skinny on interventions that work. Psychol Bull 2006;132(5):
667–91.
[59] Jeffrey RW. How can health behavior theory be made more
useful for intervention research? Int J Behav Nutr Phys Act
2004;1:10. http://dx.doi.org/10.1186/1479-5868-1-10.
[60] Davison KK, Birch LL. Childhood overweight: a contextual
model and recommendations for future research. Obes Res
2001;2(3):159–71.