UNDER EMBARGO UNTIL NOVEMBER 7, 2014, 12:01 AM ET Podcast available online at www.jneb.org
Research Article
Influence of Behavioral Theory on Fruit and Vegetable Intervention Effectiveness Among Children: A Meta-Analysis Cassandra S. Diep, PhD1,2; Tzu-An Chen, PhD1; Vanessa F. Davies, MSc1,3; Janice C. Baranowski, MPH, RD1; Tom Baranowski, PhD1 ABSTRACT Objective: To test the hypotheses that interventions clearly based on theory, multiple theories, or a formal intervention planning process will be more effective in changing fruit and vegetable consumption among children than interventions with no behavioral theoretical foundation. Design: Systematic review and meta-analysis. Setting: Identification of articles in PubMed, PsycInfo, Medline, Cochrane Collaborative database, and existing literature reviews and meta-analyses. Participants: Children aged 2–18 years. Interventions: Change in fruit and/or vegetable consumption in dietary change interventions. Methods: Meta-analysis, meta-regression analysis, and summary reporting for articles. Conclusions and Implications: Predicating an intervention on behavioral theory had a small to moderate enhancement (P < .001) of outcome effectiveness. Differences in mean Hedges’ g effect sizes between theory and non-theory interventions were 0.232 for fruit, 0.043 for vegetables, and 0.333 for fruit and vegetables combined. There was mixed support, however, for enhanced dietary change with multiple theories or a formal planning process. After controlling for study quality, theory use was related only to vegetable consumption (b ¼ 0.373; P < .001). More research is needed on theory’s influences on dietary behaviors to guide future interventions among children. More research is also needed to identify what may be effective practical- or experience-based procedures that complement theory, to incorporate into interventions. Key Words: meta-analysis, theory, dietary change, children (J Nutr Educ Behav. 2014;46:506-546.) Accepted May 19, 2014.
INTRODUCTION Higher levels of fruit and vegetable (FV) intake protect against hypertension, heart disease, stroke, and other chronic diseases1 and may be a strategy for obesity prevention among children and adults.2–7 Many child disease prevention interventions have targeted FV intake because of their low energy density, high fiber content, and abundance of phytochemicals.1 Unfortunately, FV interventions have been minimally
1
effective and optimal design components remain elusive.8,9 One important component of intervention design may be theory. Using theory may increase the effectiveness of a behavioral change intervention by providing information on which variables influence a particular behavior.10–13 Theory should embody what the behavioral sciences have learned about behavior and its change, identify key constructs demonstrated to predict behavior, and identify procedures to
US Department of Agriculture/Agricultural Research Service Children’s Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, TX 2 Department of Family and Community Medicine, Baylor College of Medicine, Houston, TX 3 Department of Public Health, Federal University of Santa Catarina, Florian opolis, Brazil Address for correspondence: Cassandra S. Diep, PhD, Baylor College of Medicine, USDA/ ARS Children’s Nutrition Research Center, 1100 Bates St, Houston, TX 77030; Phone: (713) 798-0387; Fax: (713) 798-7098; E-mail:
[email protected] Ó2014 SOCIETY FOR NUTRITION EDUCATION AND BEHAVIOR http://dx.doi.org/10.1016/j.jneb.2014.05.012
506
change these constructs to lead to behavior change.14 Given this, an early review of the dietary change intervention literature made the then bold statement that simply promoting knowledge is insufficient: Interventions based on behavioral theory are more likely to attain dietary behavior change.15 However, theories have not always been highly predictive of the targeted behavior.14,16 A meta-analysis of adult dietary change studies17 has revealed no association between theory use and intervention effectiveness. Reviews and meta-analyses of behavioral interventions for other health behaviors also reveal inconsistent results about the effectiveness of theory use in intervention design. Three reviews of behavioral interventions for human immunodeficiency virus risk behaviors18 report evidence supporting increased effectiveness when interventions are predicated on theory,19–21 2 reported possible supporting evidence,22,23 1 reported no evidence,24 and 1 reported
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 contradictory evidence.25 The latter 2 reviews include only studies among Hispanics/Latinos in the US and Latin Americans in the western hemisphere, and find extensive moderating effects on human immunodeficiency virus–related behavior change outcomes.24,25 Similar inconsistencies are found in reviews of interventions for mammography26 as well as for Internet methods27 and tailored print messaging28 targeting various health behaviors (eg, physical activity, dietary behavior, alcohol consumption, smoking cessation). Examining the effectiveness of theory in dietary change studies in children may provide different or additional insights. An important issue when evaluating theory in interventions is what constitutes a behavioral theory. Traditional behavioral theories that specify expected relationships among theoretically defined variables (eg, Social Cognitive Theory, Theory of Planned Behavior29) clearly qualify, but other theories may be more tenuous. For example, Social Ecological Theory30 simply states that aspects of the environment should be related to behavior without specification of those aspects. In a related manner, it is not always clear which theory is most applicable in a specific population (eg, gender, age, ethnic, or socioeconomic status [SES] groups) for a specific behavior. Another issue related to theory is the inconsistency in theory implementation in interventions. At one time, it was thought interventions needed to be consistent with a single theory,31 but more recently investigators have recognized the complexity of behavior change and may use multiple theories to guide an intervention.32 Alternatively, some investigators simply name theories without explaining how they used them; others provide minimal reports, whereas others employ formal intervention planning procedures such as ‘‘intervention mapping’’33 or a logic model34 to inform how to apply findings in the literature to the intervention. In the absence of empirically validated effective intervention procedures, common sense must be used to design and evaluate theory-based interventions.35 Thus, reviews and meta-analyses of theory's impact on
intervention effectiveness in children are needed to clarify the issues related to number and type of theories, implementation strategy, and study quality. This systematic review and metaanalysis tested the hypothesis: Interventions clearly based on theory were more effective in changing FV intake among children than interventions with no behavioral theoretical foundation. In addition, this study tested whether effectiveness varied by type and number of theories, having a formal intervention planning process (eg, intervention mapping33 or qualitative formative research), or study quality.
METHODS Search The first author, who has expertise in the content area and has conducted other systematic reviews, performed all searches. The authors searched for theory-based, peer-reviewed studies in PubMed, PsycInfo (Ovid), Medline (Ebsco), and the Cochrane Collaborative database without a limit on year of publication. Search terms included combinations of the following: ‘‘dietary change,’’ ‘‘obesity prevention,’’ ‘‘children,’’ ‘‘adolescents,’’ and ‘‘theory.’’ The authors also searched published literature reviews and meta-analyses on FV interventions among children and adolescents,36– 49 and searched references of all studies for further relevant studies. To find articles without theory, searches occurred in the published literature reviews and meta-analyses. In the previous searches for theorybased studies, the earliest study found by the authors was published in 1989, so they only looked for non-theory articles published in or after 1989 for comparison. Figure 1 illustrates the literature search selection process. The exported files from the literature search were uploaded and searched for duplicates in RefWorksCOS (ProQuest, LLC, Bethesda, MD, 2008) and then imported into Microsoft Excel for reviewing purposes. When an export was not possible, the authors manually entered the data into RefWorks and Excel. Files were copied and saved for the record
Diep et al 507 and searches were recorded using PRISMA guidelines.50
Study Selection Two stages identified relevant studies for inclusion. The first author screened reference titles and abstracts identified by the search strategies mentioned previously for inclusionary and exclusionary criteria. Studies had to have (1) at least 1 control group for comparison (eg, a no-intervention control group, a control group that received a similar intervention not based on theory, or both); (2) targeted dietary change (specifically, fruit [F], vegetable [V], or FV consumption); (3) reported means and standard deviations of F, V, and/or FV consumption in each group at baseline and after the intervention; (4) targeted children or adolescents (aged 2–18 years) for dietary change; and (5) employed behavior change procedures (eg, goal-setting, recipe preparation, or modified school meals). Exclusionary criteria were: (1) articles not in English; (2) non-research articles; and (3) obesity treatment studies (children were selected to already be obese). The authors obtained full texts of all references included in the first stage. The study was the unit of analysis. If more than 1 publication appeared in a study, they were combined to characterize the study. Some studies had more than 1 F or V target, in which case the authors attempted to identify any F, V, or FV combined targets and reviewed each.
Data Extraction The authors piloted a pre-coded data extraction form to ensure it captured all relevant information and then applied this form to all included studies. The information extracted from each study included program details (eg, name, focus, duration), theoretical basis, sample size and characteristics (eg, average age, race/ ethnicity), conclusions, baseline and postintervention diet measures (eg, FV consumption), baseline and postintervention weight status (eg, mean body mass index, percent overweight/obese), and moderating variables. Theoretical basis was judged from descriptions of the interventions;
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
508 Diep et al
236 articles identified through literature searches
207 articles identified through reference searches and published literature reviews
241 articles eliminated after primary screening of abstracts: Was not in English (n = 1) Was not a research article (n = 34) Did not have a control group for comparison (n = 89) Did not target dietary change (n = 62) Did not report necessary means and SDs (n = 13) Did not target children or adolescents (n = 29) Was an obesity treatment study (n = 13)
202 articles remained for secondary screening of full text 160 articles eliminated after secondary screening of full text: Was not a research article (n = 1) Did not have a control group for comparison (n = 21) Did not target dietary change (n = 12) Did not report necessary means and SDs (n = 114) Did not target children or adolescents (n = 8) Was an obesity treatment study (n = 4)
42 articles remained for final review 13 articles eliminated for not reporting fruit and/or vegetable results
29 articles remained in literature review and meta-analysis
Figure 1. Literature search and meta-analysis selection process for retrieving articles on behavioral theory use in dietary change interventions in children.
the authors searched for explicit and implicit statements of theory or constructs. Tables 1 and 2 list all variables. Following Cochrane51 and PRISMA52 guidelines, the authors also assessed risk of bias using 6 items: whether (1) the study was adequately random-
ized (ie, random sequence generation); (2) the allocation sequence was concealed from those involved in participant enrollment and assignment; (3) patients and/or data collectors were blinded; (4) incomplete outcome data were adequately ad-
dressed; (5) the study was free from selective outcome reporting; and (6) there were no other possible sources of bias (eg, self-reported methods for diet assessment, low statistical power, and not controlling for clustering when appropriate). A study received 2 points for the presence of each risk of bias, 1 for unclear presence or moderate risk, and 0 for no presence. Points were summed and divided by the total possible points, which varied across studies from 12 to 18. Studies with scores < 50% received a low risk of bias (ie, high quality) classification; those $ 50% had a high risk of bias (ie, low quality). Table 3 includes risk of bias variables and coding. Two abstractors (ie, CD and VD, both with training in research methods, statistics, and the content area) independently abstracted all articles. Inter-abstractor agreement was 88.2%. Both abstractors discussed and resolved all differences. TB and JB oversaw and approved all results.
Statistical Analysis To represent the impact of the behavioral intervention on dietary change, effect sizes (Hedges' g) were calculated, defined as the standardized mean difference (the difference between treatment vs control means divided by their pooled standard deviation) with Hedges' correction for small sample size.53 Analyses included only studies that explicitly measured F, V, or FV combined consumption (there were insufficient numbers of studies targeting other dietary outcomes to conduct a meta-analysis with them). If a study reported outcomes from > 1 intervention or intervention group for F, V, and/or FV, the mean of the effect sizes was calculated so that each study had only 1 F, V, and/or FV combined effect size. For studies that reported > 1 follow-up measure, effect size calculations used data from the time point closest after intervention, when an effect was most likely to be detected. Based on the Q statistic, I2,54 and visual inspection of forest plots, heterogeneity (or systematic differences between studies) was likely. The authors examined potential outliers55 and undertook sensitivity analysis to explore the effect of removing studies
None
Gimme 5 on eating more FJV: RCT in schools and grocery
Baranowski (2000)62
None
Qualitative CHILDREN study on formative overcoming barriers in research accessing PA areas, increasing FV availability, and increasing parental support: RCT in school and home for 12 mo (2005–2006). Activities included discussion, active learning, cues, modeling, guided practice, enactment, problem solving, goal setting, self-reevaluation, environmental reevaluation, arguments, direct experience, mobilized social support, educational material, parental meetings, and FV bazaars delivered by schoolteachers (trained by research team) and parent involvement at home
Program Details Intervention to increase FV consumption: RCT in schools for 9 mo (1999– 2000). Activities included FV provision in schools, marketing newsletters, and curriculum materials that seemed to be delivered by teachers.
Formal Planning Process
Angelopoulos (2009)61
First Author (Publication Year) Anderson (2005)60
SCT
TPB and ecological model
TPB
Theory
Sample Size and Characteristics 511 intervention and 464 control children aged 6–7 and 10–11 y in Scotland; 65 control (mean age, 8.6 2.23 y; 52% female) and 64 intervention (mean age, 8.4 2.21 y; 56% female) completed food diaries
FV-Related Conclusions Whole-school approach to increasing FV intake had modest but significant effect on F intake
(continued)
Intervention based on 1,732 US fourth- and FV consumption in SCT and targeted/ fifth-graders control group
Daily F consumption 646 fifth-graders in Focus groups to increased in IG and Greece: 321 develop TPB decreased in CG. intervention questionnaires and No other significant children (mean age, identify differences in FV 10.25 0.44 y; determinants of consumption 57.3% female) and children’s observed. 325 control behaviors. Results children (mean age, of TPB 10.29 0.44 y; questionnaires 54.2% female) applied to F, V, dairy, and sweets consumption components. Intervention pursued changes in TPB constructs to change child’s environment (reflects ecological models).
Theoretical Basis Description Assessments examined beliefs, attitudes, and knowledge. TPB not used to design intervention.
Table 1. Summary of Intervention Details, Theoretical Basis, Sample Size and Characteristics, and Conclusions
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 509
Birnbaum (2002)63
First Author (Publication Year)
Table 1. Continued
Systematic program Teens Eating for Energy and planning approach Nutrition at School to promote health behavior change (diet): RCT in schools for 1 school y (1998–1999). Activities included curriculum sessions, activities, and intervention-related messages for parents or guardians delivered by peer leaders and teachers.
stores for 6 wk (January to February, 1995 and January to February, 1996). Activities included role playing, snack and meal preparation, goal setting, and problem solving delivered by teachers and produce managers.
Program Details
Formal Planning Process
SCT and TPB
Theory
displayed substantial decline from y 1 to y 2, with either no decline or mitigated decline in treatment group
FV-Related Conclusions
Peer leaders plus 1,755 control US curriculum plus seventh-graders environment (48.7% female, groups reported 70.8% white, largest increases in 11.7% African FV consumption. American, 3.7% Students exposed API, 2.2% to curriculum plus Hispanic/Latino, environment 6.1% multiracial, interventions also 1.5% Native improved. American, 4.1% Students exposed other); 226 peer to environment leaders plus interventions curriculum plus showed declining environment FV intake. Control (53.5% female, students’ choices 72.6% white, 5.8% remained stable. African American, 9.3% API, 1.8% Hispanic/Latino, 5.8% multiracial, 0.9% Native American, 4.0% other), 677 curriculum plus environment (49.3% female,
completed 7-d food record in y 1, 1,864 in y 2, and 1,946 in y 3. Final cohort sample (n ¼ 1,172) was 15.3% African American and 84.7% EuroAmerican, almost equally split by gender.
measured knowledge, preferences, outcome expectations, selfefficacy, social norms, asking behaviors, and availability– accessibility Intervention (with 3 different exposure groups) based on SCT, TPB, and systematic program planning approach. Curriculum sessions informed by SCT and developed using systematic program planning approach; assessments based on TPB.
Sample Size and Characteristics
Theoretical Basis Description
510 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
Comics for Health for None prevention, assessment, and
Branscum (2013)65
None
Challenge! for health promotion/obesity prevention: RCT in adolescents’ homes and community sites (eg, convenience stores, parks) with 12 sessions (duration unclear but research assistants enrolled adolescents and caregivers between 2002 and 2004). Activities included mentorship (role modeling and support), participatory learning, goal setting, and making healthy snacks and recipes delivered by trained, college-enrolled (or recently graduated) black mentors.
Black (2010)64
Life Course Theory and SCT
SCT and MI
Intervention based on 34 control children Life Course Theory (43% female, 82%
(continued)
Both interventions found significant,
In most dietary 235 black Intervention categories adolescents aged incorporated (including FJ and 11–16 y in US: 114 principles of SCT V), intervention control (mean age, and MI to help adolescents had 13.3 1.0 y; adolescents greater decreases 47.4% female, identify personal in average daily 96.5% nonchallenges and consumption than Hispanic black) and goals related to diet controls. No 121 intervention and PA. Mentors intervention effects (mean age, 13.3 trained in on body mass 1.0 y; 51.2% motivational index category or female, 97.5% noninterviewing. body composition. Hispanic black). 56% living below federal poverty level. There were 93 control and 91 intervention at postintervention and 90 control and 89 intervention at delayed follow-up.
70.1% white, 8.6% African American, 7.4% API, 2.4% Hispanic/Latino, 5.0% multiracial, 2.5% Native American, 4.0% other); and 845 environment (49.6% female, 62.1% white, 10.2% African American, 12.3% API, 4.0% Hispanic/Latino, 5.1% multiracial, 1.3% Native American, 5.0% other)
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 511
Eat 5 for increasing FV consumption: RCT for 1 mo in junior Girl Scout troops. Activities included classes,
Cullen (1997)67 None
None Web-Based Active Balance Childhood to enhance adolescents’ self-efficacy and facilitate their understanding and use of problem-solving skills related to nutrition, PA, and coping: RCT for 8 wk (data collected from 2007 to 2009). Activities included curriculum, interactive dietary preparation software, and goal setting delivered by research team over Web in home, library, or community center.
treatment of child and adolescent overweight and obesity (screen time, water and sugar-free drinks, PA, FV): group RCT with 4 lessons (each 30 min) in YMCA after-school programs. Activities included instructor modeling and practice, positive role models, role playing positive and vicarious reinforcement, goal-setting activities, storytelling, and character development delivered by corresponding author.
Program Details
Chen (2011)66
First Author (Publication Year)
Table 1. Continued Formal Planning Process
SCT
TTM and SCT
Theory white, 6% African American, and 12% Asian) and 37 intervention children (53% female, 73% white, 14% African American, and 5% Asian) aged 8–11 y in US
and targeted elementary-aged children (timing) in schools (environment). Intervention operationalized, targeted, and measured SCT constructs (including selfefficacy, expectations, and self-control).
yet modest effects for FV consumption
FV-Related Conclusions
SCT used to guide selection of behavioral determinants and
22 US Junior Girl Scout troops with about 300 girls aged 9–12 y (75%
At posttest, girls in intervention condition troops reported
Significantly more Intervention designed 54 Chinese adolescents in IG Americans aged to be individually than in CG 12–15 y (mean age, tailored to decreased their 12.52 3.15 y) participants’ waist-to-hip ratio and their families: behavioral stage. while also 27 control (52% Family component increasing FV female) and 27 also targeted intake intervention (41% participants’ female). About 40% behavior and of families had an personal factors. annual household income < $40,000. There were 25 control and 26 intervention at time 1 and 24 control and 26 intervention at times 2 and 3.
Sample Size and Characteristics
Theoretical Basis Description
512 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
LA Sprouts for gardening and None nutrition - non-randomized intervention pilot study in Milagro Allegro Community Garden for 12 wks in 2010. Activities included educational and cookery lessons aimed to children, and classes aimed to parents delivered by study staff member or graduate student trained in nutrition (supervised by RD).
None Action Schools! BC for increasing FV consumption; enhancing knowledge, attitudes, and perceptions regarding FV; and increasing willingness to try new FV: RCT in school for 12 wk (2005–2006). Activities included classroom and tasting activities by teachers.
Social marketing Gimme 5 for increasing FV method of focus consumption among fourthgroups and fifth-graders: RCT in elementary health class for 6 wk (October to November, 1991). Activities included recipe preparation, taste testing, goal setting, self-
Davis (2011)68
Day (2008)69
Domel (1993)70
FV preparations and tastings, and parent information sheets by principal investigator and troop leaders.
SCT
None
None
Social marketing method of focus groups identified ways to promote behavior. ‘‘Curriculum development was based on the
None
None
intervention methods (eg, intervention targeted knowledge, skills, self-efficacy)
significantly greater FV intake compared with control condition troops, but increased intake not sustained
(continued)
No differences in total 301 US fourth- and fifth-graders FV daily servings, greater increase in (51.2% female, just F daily servings in over half African intervention American and the compared with rest predominantly control, and small Anglo American): increase in V daily 106 control (48%
198 control Canadian Significant differences found between youth (mean age, conditions over 9.9 0.57 y; 57% time. F servings, FV female) and 246 servings, and FV intervention (mean variety increased in age, 10.1 0.64 y; intervention 49% female) schools.
Compared with CG, 104 Latino fourthLA Sprouts and fifth-grade participants students in Los increased dietary Angeles, CA: 70 fiber intake. No control (mean age, significant changes 9.9 0.7 y; 41.4% in F or V specifically. female, 4% white, 93% Latino, 3% Asian) and 34 intervention (mean age, 9.7 0.7 y; 61.8% female, 97% Latino, 3% Asian)
Caucasian, 11% Hispanic, 3% African American, 11% other ethnic groups). At baseline, 133 control and 126 intervention with diet measures; at postintervention, 109 control and 101 intervention.
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 513
Dzewaltowski (2009)71
First Author (Publication Year)
Table 1. Continued
Healthy Youth Places for building middle school environments that promote FV and PA: schoolrandomized trial in classroom, school lunch, and after-school program. Activities included youth-led school advocacy groups (‘‘change teams’’) and curriculum by expert staff.
monitoring, problem solving, fun activities (raps, games, role playing, mock newspaper columns, comic strips, rhymes), and weekly newsletter by teachers.
Program Details
None
Formal Planning Process
SCT
Theory
servings in control vs no change in intervention
eligible for free or reduced-price lunch) and 195 intervention (63% eligible for free or reduced-price lunch)
concept of reciprocal determinism within social cognitive theory with focus groups providing details on theoretical constructs.’’ Intervention appeared to target these factors.
1,204 US sixth-grade Intervention and Intervention model control schools did control students designed to not change (mean age, 12.40 influence direct differently over time 0.43 y; 47.2% personal and proxy on FV combined, F, boys, 36.8% free agency (2 modes of or V and reduced lunch, individual human 78.7% white, 9.0% agency targeted by black, 4.2% SCT). Intervention Hispanic, 2.5% model targeted American Indian, project, school, 1.4% Asian) and and place levels. 1,007 intervention Training sessions (mean age, 12.36 for school staff 0.40 y; 46% emphasized boys, 29.8% free theory-based and reduced lunch, principles of 77.7% white; behavior change, 10.0% black, and several 4.67% Hispanic, constructs (eg, self1% American efficacy, group Indian, 3.5% norm) assessed Asian). Study through survey. cohort included 815 control and 767 intervention students.
FV-Related Conclusions
Sample Size and Characteristics
Theoretical Basis Description
514 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
Switch to promote being active None for $ 60 min/d, limiting total screen time to # 2 h/d, and eating $ 5 FV/d: RCT in communities, schools, and homes from 2005 to 2006. Activities included paid advertising, unpaid media, curricula, materials, and resources by teachers and research team.
None Intervention to test whether culturally adapted nutrition curriculum affects FV consumption in American Indian third-graders: intervention pilot study for 6 wk in school. Activities included nutrition lessons by registered dietitian.
Intervention to increase children’s F intake: nonrandomized controlled trial (intervention pilot study)
Gentile (2009)73
Govula (2007)74
Gribble (2003)75
None
None Intervention to increase FV consumption of upperelementary school students and change home nutritional environment: quasiexperimental study design for 6 wk after school. Activities included nutrition education sessions, literacy and health communication sessions, and developing media campaign.
Evans (2006)72
SLT
None
Social ecological framework
SCT
Curriculum covered topics related to knowledge, skills, goal-directed
None
Socio ecological framework guided development of Switch (integrated programming at community, school, and family levels)
Intervention targeted home environment and certain personal factors (eg, self-efficacy, motivation) to change behavior. Child and parent psychosocial constructs measured through scales and questions.
17 US child–parent pairs in control (for child, mean age, 11.6 0.5 y;
(continued)
Significant increase in F intake by children in experimental vs control group
Change in intake from 21 control and 12 baseline to intervention completion American Indian between groups students aged 8– was significantly 11 y (45.4% female) different for total FV and V (higher for intervention than control)
1,323 third- through Switch program yielded small to fifth-grade children modest treatment and their parents effects for (47% male, 90% promoting white) from 10 children’s FV schools in 2 US consumption, states. At baseline, particularly 6 mo 674 control (mean postintervention age, 9.6 0.9 y) and 685 intervention (mean age, 9.6 0.9 y).
Intervention not 18 US intervention effective in (66.6% female, changing FV 50.0% black, consumption in 33.3% white, children 27.7% with total household income < $20,000) and 21 control (42.8% female, 23.8% black, 57.1% white, 9.6% with total household income < $20,000) fourthand fifth-graders. Fifteen control and 13 intervention included in analyses.
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 515
None
Be Active Eat Well for preventing childhood obesity: longitudinal study from 2003 to 2006
Intervention to promote FV intake: RCT in school. Activities included internet-
Johnson (2012)77
Mangunkusumo (2007)78
None
Intervention on PA and healthy None eating (increasing F, reducing soft drink, increasing water, and reducing fat intake): RCT for 2 y (2003–2005) in school and home. Activities included computer-tailored intervention, information through posters and folders, and work group meetings by research and school staff.
for 10 wk in food science lab at University of Cincinnati. Activities included instruction, pen-and-paper activities, interactive discussions, case studies, brainstorming activities, and games by nutrition graduate research assistant.
Program Details
Haerens (2006)76
First Author (Publication Year)
Table 1. Continued Formal Planning Process
Behavioral Change Theory
Ecological theory
TPB and TTM
Theory 82.3% female, 94.1% white, 5.9% other) and 9 child– parent pairs in intervention (for child, mean age, 10.8 0.6 y; 33.3% female, 55.5% white, 44.5% other)
behaviors, reinforcement, rewarding/ punishing, etc. Constructs also measured as variables. Intervention appeared to target SLT factors.
FV-Related Conclusions
Intervention informed by Behavioral Change Theory
486 children aged 9– 12 y in The Netherlands: 223
No significant effects found on intake
1,812 children aged Appeared to be Intervention greater increases in 4–12 y in Australia: recognized multiple F and V 977 control (mean hierarchies of consumption in IG age, 8.19 2.15 y; influence including than comparison 49.2% female) and individual, 835 intervention household, and (mean age, 8.16 environment 2.25 y; 53.7% (collected data on female) multiple hierarchies)
No positive 2,991 seventh- and Questionnaires on intervention effects eighth-graders in psychosocial on F consumption West Flanders/ determinants used Belgium (mean to tailor feedback age, 13.1 0.8 y; and advice for F 36.6% female, intake or PA based 32.6% higher SES): on TPB. Also, TTM 759 control, 1,226 ‘‘used for matching intervention with content and parental approach of this involvement, 1,006 feedback to the intervention alone; stages of 2,840 at baseline changes.’’ and 2,287 at 2-y postintervention
Sample Size and Characteristics
Theoretical Basis Description
516 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
None Go For Health to influence students’ diet and PA (reduce sodium, total fat, saturated fat; increase PA; and develop behavioral capability, self-efficacy, and positive expectancies): controlled trial (unclear whether randomized or quasi-experimental) from 1985 to 1987 in school. Activities included modeling, self-monitoring of behavior, contracting to try new behaviors, skill development, verbal praise, material rewards, cues and reinforcing messages, modified school meals and PE, new recipes, food protocol, training for food service workers, and lessons in health education by teacher and staff.
None School fruit and vegetable scheme to provide free FV for schoolchildren to improve daily FV consumption: intervention trial for 10 mo (2004) in school. Activities included free FV and
Parcel (1989)79
Ransley (2007)80
tailored nutrition advice, brief dietary counseling, and oral and written feedback about nutrition by research team and nurse.
control (mean age, 10.3 0.5 y; 51.1% female, 84.6% Dutch background) and 263 intervention (mean age, 10.3 0.5 y; 54.4% female, 88.6% intervention)
None
None
(continued)
School fruit and 3,703 children aged vegetable scheme 4–6 y in England: promoted increase 2,143 control in adjusted F intake (mean age, 72.4 after 3 mo but 10.3 mo; 51% effect waned over female, 20% with time free school meals eligibility) and 2,681
No overall effect At baseline, 194 SLT and Organization Intervention to found for control and 213 change model develop behavioral percentage of F or intervention US capability, V selected in third- and fourthexpectations, and relation to all foods graders (62.3% self-efficacy for selected Anglo, 20.9% healthful eating and Mexican, 14.8% PA, and also black, 2% Asian influence American and expectancies. American Indian). A Methods included total of 147 control modeling, selfand 205 monitoring, cues, intervention at rewards, etc, and posttest 1; 159 allowed students to control and 238 practice healthful intervention at behaviors. SLT posttest 2. constructs measured. Implementation based on organizational change model (ie, sequence of 4 phases).
(explored effects on potential determinants of intake and selfreported intake as initial steps towards behavioral change). In addition, selfefficacy targeted during dietary counseling.
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 517
Formal Planning Process
None High 5 (under 5-A-Day for Better Health Initiative) to increase FV consumption among fourth graders: randomized experimental design for 7 wk (1994–1995) in school and home. Activities included behavioral curricula in fourth and fifth grades, parental involvement/education, school food service changes, and industry involvement and
Reynolds (2004)83
Intervention mapping protocol
Intervention to make FV consumption habitual: RCT for 8 mo (2004–2005) in school. Activities included daily free FV, classroom curriculum, and parental involvement delivered by schoolteacher.
educational materials and activities.
Program Details
Reinaerts (2007)81 Reinaerts (2008)82
First Author (Publication Year)
Table 1. Continued
SCT
None
Theory intervention (mean age, 72.1 10.5 mo; 49% female, 18% with free school meals eligibility) at baseline; 1,648 control and 2,045 at postintervention
Sample Size and Characteristics
FV-Related Conclusions
SCT used to design both interventions and select evaluation measures. Each potential mediator derived from SCT constructs, targeted in intervention activities, and measured as part of evaluation. For
1,584 fourth-graders Both Alabama and Minnesota in Alabama, but programs unclear distribution increased FV intake in control (mean between y 1 and 2 age, 8.71 0.57 y; 49.56% female, 78.88% white, 21.12% non-white) and intervention (mean age, 8.68 0.59 y; 50.57% female, 76.26%
Similar effects for free Both distribution and 436 parents of distribution children aged 4–12 multi-component program and y in The programs multicomponent Netherlands with developed program in repeated measures according to increasing (mean age, 8 2.2 intervention children’s F y; 53% female, mapping protocol, consumption over 82% Dutch origin): a stepwise time. Distribution 227 control, 85 in approach to ensure program also distribution systematic increased program, and 124 evidence-based children’s V in multicomponent and theory-driven consumption over program with diet development and time. measures implementation of health-promoting interventions
Theoretical Basis Description
518 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
Rosenkranz (2010)84
None Healthier Troops in Supplemental Assistance Program for preventing obesity by modifying Girl Scout troop meeting environments and by empowering girls to improve quantity and/or quality of family meals in home environments: group RCT for about 4 mo (October, 2007 to April, 2008). Activities included interactive educational curriculum, troop meeting policies, and badge assignments completed at home by troop leaders.
5-A-Day Power Plus (under 5- None A-Day for Better Health Initiative) to increase FV consumption among fourthgraders: randomized experimental design for 8 wk from 1995 to 1996 in school and home. Activities included behavioral curricula in fourth and fifth grades, parental involvement/education, school food service changes, and industry involvement and support by teachers.
support by curriculum coordinators.
SCT
SCT
Intervention targeted role modeling, skill building, enhancement of self-efficacy and proxy efficacy, and reinforcement of behavior
example, activities built skills, selfefficacy, outcome expectancies.
(continued)
42 control (mean age, Significant differences by condition within 10.5 1.3 y; time 1 and time 2 75.0% nonfor FV consumption Hispanic but no significant Caucasian, 25.0% differences in racial/ethnic change from time 1 minority, 35.0% to time 2 by lower SES) and 34 condition intervention (mean age, 10.6 1.1 y; 79.4% nonHispanic Caucasian, 20.6% racial/ethnic minority, 28.1% lower SES) US Girl Scout troops aged 9–13 y; 39 control and 33 intervention in analyses
522 fourth-graders in Minnesota, but unclear distribution in control (mean age, 10.00 0.38 y; 52.06% female, 54.68% white, 45.32% non-white) and intervention (mean age, 10.05 0.41 y, 50.59% female, 45.88% white, 54.12% nonwhite). More than 60% of students received free or reduced-cost lunches.
white, 23.74% nonwhite)
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 519
None Schoolgruiten Project to increase availability, accessibility and exposure to FV and knowledge and skills: quasi-experimental study in 2003 in school. Activities included free piece of F or V and lessons to increase knowledge and skills by teachers.
None 5-A-Day Achievement Badge Program for FJ and low-fat V consumption: 2-condition (treatment, active-attentionplacebo-control) group randomized trial for 9 wk from 2003 to 2004 in troops and over Internet. Activities
Thompson (2009)87
Program Details Eating smart, Eating for me to None encourage healthy eating, exercise, and body image, while discouraging calorierestrictive dieting, exercising for weight loss, and development of body dissatisfactory: school intervention trial with lessons by classroom teachers and newsletters to parents
Tak (2007)86
First Author (Publication Year) Smolak (1998)85
Table 1. Continued Formal Planning Process
SCT
None
None
Theory
Intervention taught functional knowledge and skills to enhance self-efficacy; made outcome expectancy– related comments
None
Theoretical Basis Description None
473 Boy Scouts aged Boy Scout troop plus Internet 10–14 y in intervention Houston, TX: 228 promoting FJ and control and 224 low-fat V intervention at consumption baseline, 234 resulted in shortcontrol and 235 term changes in FJ intervention at
565 children of Dutch Schoolgruiten Project had significant ethnicity and 388 effect on F intake of children of nonchildren of Dutch Western ethnicity ethnicity and on V (mean age, 9.9 y) intake of children of and their parents in non-Western The Netherlands: ethnicity 333 control children of Dutch ethnicity, 120 control children on non-Western ethnicity, 232 intervention children of Dutch ethnicity, 268 intervention children of nonWestern ethnicity
Sample Size and FV-Related Characteristics Conclusions Behavior, including 222 white public eating patterns, not school fifth-graders changed by (54.5% female) in participation in US: 55 in control curriculum and 167 in intervention
520 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
None Intervention to increase FV intake and PA in African American adolescents: 3group randomized design for 12 wk in school (after-school program). Activities included behavioral records, diet and PA checklist, reinforcement (reward) plan, environmental cues, social support, positive self-talk, and strategic selfpresentation videotape session. None (see above)
SCT
consumption among US Boy Scouts
Intervention based on 53 African American SCT plus MI and SCTonly groups adolescents aged SCT principles (eg, showed significant 11-15 y: 16 control education, increase in FV (0% female), 17 behavioral skills intake compared (52.9% female) in training, and with control group SCT plus MI, and feedback). For MI, 20 (65.0% female) self-concept and in SCT only; 11 motivation control, 14 SCT incorporated. SCT plus MI, and 14 and MI theoretical SCT only with post measures were measures measured.
post-1, and 219 control and 213 intervention at post-2
SCT and MI
to enhance motivation; taught recipe preparation to enhance preference, knowledge, and home availability; which would enhance consumption. Constructs were components of program and measured.
API indicates Asian and Pacific Islander; CG, control group; F, fruit; FJ, fruit and juice; FJV, fruit, juice, and vegetables; FV, fruit and vegetables; IG, intervention group; MI, motivational interviewing; PA, physical activity; RCT, randomized controlled trial; SCT, Social Cognitive Theory; SES, socioeconomic status; SLT, Social Learning Theory; TPB, Theory of Planned Behavior; TTM, Transtheoretical Model–Stages of Change; V, vegetables.
Wilson (2002)88
included in-troop sessions, recipe preparation and tasting, goal setting, selfmonitoring, and problem solving by troop leaders.
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 521
Mean daily servings PL + C + E mean daily servings of of FV: 4.76 0.04 FV: 4.88 0.06 C + E mean daily servings of FV: 4.51 0.04
Birnbaum (2002)63
Not provided
Not provided (Least-squares mean values [SE]) FV: 2.3 (0.1) V: 1.1 (0.1) FJ: 1.2 (0.1) FV weekday lunch: 0.9 (0.1) FV all other times: 1.6 (0.1)
(Least-squares mean values [SE]) FV: 2.4 (0.1) V: 1.2 (0.1) FJ: 1.2 (0.1) FV weekday lunch: 1.0 (0.1) FV all other times: 1.7 (0.1)
Baranowski (2000)62
Not provided
Not provided (Least-squares mean values [SE]) Y 2 FV: 2.3 (0.1) V: 1.2 (0.1) FJ: 1.1 (0.1) FV weekday lunch: 1.0 (0.1) FV all other times: 1.6 (0.1) Y 3 FV: 2.3 (0.1) V: 1.1 (0.1) FJ: 1.1 (0.1) FV wkday lunch: 0.9 (0.1) FV all other times: 1.6 (0.1) Mean daily servings PL + C + E mean daily servings of of FV: 4.80 FV: 5.80 0.05 0.03 C + E mean daily servings of FV: 4.95 0.04 E mean daily
(Least-squares mean values [SE]) Y 2 FV: 2.1 (0.1) V: 1.0 (0.1) FJ: 1.1 (0.1) FV weekday lunch: 0.8 (0.1) FV all other times: 1.5 (0.1) Y 3 FV: 2.1 (0.1) V: 1.1 (0.1) FJ: 1.0 (0.1) FV wkday lunch: 0.8 (0.1) FV all other times: 1.5 (0.1)
Adjusted for race and parents’ education
Neither gender nor ethnic group moderated effect
Control mean BMI: Adjustments in analyses made for 20.2 3.2 MVPA level, BMI, Control mean BMI zgender, and score: 0.67 0.8 school region, Intervention mean when appropriate BMI: 19.2 2.9 Intervention mean BMI z-score: 0.41 0.9
(exchanges/d) F: 1.5 1.8 V: 1.0 1.4 Dairy: 2.5 1.1 Fats and oils: 6.4 3.8 Meat: 3.6 2.9 Grains: 5.7 2.8 Sweets and beverages: 1.7 2.2
(exchanges/d) F: 1.1 1.6 V: 1.2 1.2 Dairy: 3.0 1.9 Fats and oils: 9.4 5.0 Meat: 4.8 3.2 Grains: 7.7 3.3 Sweets and beverages: 2.8 3.2
Control mean BMI: 20.1 3.4 Control mean BMI zscore: 0.83 0.9 Intervention mean BMI: 20.3 3.6 Intervention mean BMI z-score: 0.87 0.9
(exchanges/d) F: 1.1 1.2 V: 1.2 1.1 Dairy: 2.7 1.3 Fats and oils: 8.0 4.7 Meat: 4.4 2.5 Grains: 6.6 2.7 Sweets and beverages: 2.5 2.2
(exchanges/d) F: 1.3 1.5 V: 1.1 1.0 Dairy: 2.8 1.4 Fats and oils: 8.7 5.1 Meat: 5.5 3.3 Grains: 7.4 3.1 Sweets and beverages: 2.6 2.7
Angelopoulos (2009)61
Not provided
Possible Moderating Variables
Not provided
Postintervention Weight Status
F: 183 17.0 V: 52 48.6 FV : 235 151.2
Postintervention Group Diet Measures
F: 107 14.2 V: 55 42.3 FV: 163 109.6
Postintervention Control Group Diet Measures
Not provided
Baseline Weight Status
F: 133 11.9 V: 69 41.1 FV: 202 101.9
F: 100 11.7 V: 70 58.1 FV: 170 109.6
Baseline Intervention Group Diet Measures
Anderson (2005)60
First Author (Publication Year)
Baseline Control Group Diet Measures
Table 2. Summary of Baseline and Postintervention Diet, Weight Status, and Moderating Variables for Each Intervention
522 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
(Servings/d) FJ: 1.3 1.3 V: 2.1 2.2 Snacks/desserts: 4.9 4.1 Milk: 1.0 1.2 Nondiet soda: 0.8 0.7 Fried foods: 0.4 0.3
(Servings or glasses) FV: 3.35 3.17 Sugar-free beverage: 4.03 2.76 SSB: 0.94 1.17
(Servings/d) FV: 2.28 0.61
Black (2010)64
Branscum (2013)65
Chen (2011)66
Control mean z-BMI: 0.59 1.1 (31.6% overweight/ obese) Intervention mean z-BMI: 0.76 1.2 (44.6% overweight/ obese)
(Servings/d) FV: 2.19 0.48
Control mean BMI: 20.25 3.21 (33.3% overweight/ obese) Intervention mean
(Servings or glasses) Control mean BMI percentile: FV: 3.41 2.68 55.52 27.96 Sugar-free beverage: 3.35 Intervention mean BMI percentile: 2.58 60.15 26.39 SSB: 1.36 2.00
(Servings/d) FJ: 1.6 1.3 V: 2.4 2.0 Snacks/desserts: 5.7 4.7 Milk: 1.0 1.2 Nondiet soda: 0.8 0.7 Fried foods: 0.8 0.4
E mean daily servings of FV: 4.76 0.03
(Servings/d) Time 1 FV: 2.14 0.66 Time 2 FV: 2.11 0.55
(Servings or glasses) Posttest FV: 4.15 2.97 Sugar-free beverage: 5.24 3.25 SSB: 1.00 1.54 Follow-up FV: 4.91 3.18 Sugar-free beverage: 6.06 3.71 SSB: 1.09 1.94
(Servings/d) Postintervention FJ: 1.4 1.4 V: 1.9 2.3 Snacks/desserts: 5.4 5.2 Milk: 0.8 1.0 Nondiet soda: 0.7 0.7 Fried foods: 0.7 0.5 Delayed follow-up FJ: 1.1 1.2 V: 2.0 2.4 Snacks/desserts: 4.3 4.0 Milk: 0.9 1.1 Nondiet soda: 0.7 0.7 Fried foods: 0.8 0.5
Control mean z-BMI BMI not moderator for diet PI: 0.61 1.14 (34.41% overweight/ obese) Control mean z-BMI Delayed followup: 0.65 1.14 (42.70% overweight/ obese) Intervention mean zBMI PI: 0.73 1.16 (39.56% overweight/ obese) Intervention mean zBMI DFU: 0.77 1.14 (39.33% overweight/ obese)
(Servings/d) Time 1 FV: 2.36 0.64 Time 2 FV: 2.41 0.64
Control Time 3 mean BMI: 20.21 3.13 Intervention Time 3 mean BMI: 20.76 3.08
(continued)
Not provided
Not provided (Servings or glasses) Control FU mean BMI percentile: Posttest FV: 57.26 27.84 4.68 3.08 Intervention FollowSugar-free up mean BMI beverage: 4.62 percentile: 59.23 3.17 26.31 SSB: 0.89 0.98 Follow-up FV: 4.62 3.33 Sugar-free beverage: 4.21 2.62 SSB: 0.89 0.95
(Servings/d) Postintervention FJ: 1.2 1.2 V: 2.0 2.0 Snacks/desserts: 4.0 3.5 Milk: 1.0 1.1 Nondiet soda: 0.6 0.7 Fried foods: 0.6 0.4 Delayed follow-up FJ: 1.1 0.9 V: 1.9 1.9 Snacks/desserts: 3.6 3.7 Milk: 0.7 1.0 Nondiet soda: 0.5 0.6 Fried foods: 0.6 0.4
servings of FV: 4.44 0.04
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 523
Baseline Intervention Group Diet Measures BMI: 20.79 3.12 (37% overweight/ obese)
Baseline Weight Status
Meat (servings/d): 2.0 1.7 Dairy (servings/d): 2.1 1.1 V (servings/d): 1.9 1.3 F (servings/d): 4.1 0.9 Whole grains (oz/d): 0.7 0.7
Servings of F: 3.36 3.10 Servings of V: 2.08 2.11 Servings of FV: 5.44 4.26 Variety of FV: 3.93 2.54
(daily servings) Total FV: 2.05 1.04 F: .42 .42 J: .39 .44 V: 1.16 .65 Legumes: .09 .14
(Cohort leastsquares mean
Day (2008)69
Domel (1993)70
Dzewaltowski (2009)71
Not provided
(Cohort leastsquares mean
Control mean BMI: 19.18 3.78
Not provided (daily servings) Total FV: 2.66 1.25 F: .60 .53 J: .34 .47 V: 1.55 .75 Legumes: .17 .21
Servings of F: 2.31 2.75 Servings of V: 1.94 1.94 Servings of FV: 4.24 3.69 Variety of FV: 3.33 2.15
Control mean BMI Meat (servings/d): percentile: 2.1 2.4 80.1 24.4 (61% Dairy (servings/d): overweight/ 2.1 1.3 obese) V (servings/d): Intervention mean 1.6 1.0 BMI percentile: F (servings/d): 74.5 80.1 (53% 4.0 0.7 overweight/ Whole grains (oz/d): obese) 0.8 0.9
Mean daily servings Mean daily servings Not provided of FV: 2.20 (1.96) of FV: 3.02 (2.21)
Baseline Control Group Diet Measures
Davis (2011)68
Cullen (1997)67
First Author (Publication Year)
Table 2. Continued
Time 3 FV: 2.63 0.71
Postintervention Group Diet Measures Postintervention Weight Status
(Cohort leastsquares mean
(daily servings) Total FV: 2.35 1.32 F: .48 .54 J: .46 .56 V: 1.36 .67 Legumes: .06 .10
Servings of F: 2.68 2.66 Servings of V: 1.97 1.84 Servings of FV: 4.65 3.46 Variety of FV: 3.83 2.37
Meat (servings/d): 2.5 3.4 Dairy (servings/d): 1.7 1.0 V (servings/d): 1.3 1.0 F (servings/d): 4.2 0.8 Whole grains (oz/d): 0.6 0.6
Not provided
(Cohort leastsquares mean
Control mean BMI: 18.93 3.72
Not provided (daily servings) Total FV: 2.95 1.21 F: .90 .58 J: .38 .52 V: 1.55 .70 Legumes: .12 .16
Servings of F: 2.55 2.72 Servings of V: 1.87 1.90 Servings of FV: 4.42 3.45 Variety of FV: 3.80 2.56
Control mean BMI Meat (servings/d): percentile: 2.8 2.5 80.0 24.2 Dairy (servings/d): Intervention mean 1.7 1.2 BMI percentile: V (servings/d): 74.9 25.7 1.6 1.0 F (servings/d): 3.9 0.8 Whole grains (oz/d): 0.9 0.7
(Mean daily servings (Mean daily servings Not provided of FV) of FV) Posttest: 2.06 (1.71) Posttest: 3.39 (1.93) 3-mo follow-up: 3-mo follow-up: 2.32 (1.81) 2.89 (1.60)
Time 3 FV: 2.34 0.66
Postintervention Control Group Diet Measures
Analyses controlled for gender, race,
Not provided
Unclear
Not provided
Adjusted for grade level and pretest FV intake
Possible Moderating Variables
524 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
(Half serving portions) F: 1.0 1.0
(Pieces/wk or glasses/d): Boys F: 6.4 6.0 Soft drinks: 2.5 2.3 Water: 3.7 2.5 Girls F: 6.7 5.8 Soft drinks: 2.4 2.1 Water: 3.6 2.5
Gribble (2003)75
Haerens (2006)76 (Pieces/wk or glasses/d): Boys F: 4.9 5.1 Soft drinks: 3.4 2.5 Water: 3.2 2.7 Girls F: 5.4 5.0 Soft drinks: 2.8 2.2 Water: 3.5 2.6
Not provided
Not provided
Not provided
Total FV: 3.4 0.9 Total F: 2.4 0.8 Total V: 1.0 0.2
Total FV: 5.9 0.7 Total F: 3.4 0.5 Total V: 2.5 0.4
Govula (2007)74
(Half serving portions) F: 1.8 2.3
Control mean BMI: 18.5 3.5 (27.9% overweight/ obese) Intervention mean BMI: 18.4 3.3 (26.5% overweight/ obese)
(Servings/d) Child FV: 4.9 3.2
(Servings/d) Child FV: 4.1 2.9
Not provided
Gentile (2009)73
(In servings) FV: 1.94 1.01
Intervention mean BMI: 19.15 3.98
(In servings) FV: 2.06 1.83
[SE]) FV combined: 3.11 (0.20) F: 1.90 (0.11) V: 1.21 (0.10)
Evans (2006)72
[SE]) FV combined: 3.07 (0.19) F: 1.80 (0.10) V: 1.26 (0.10)
(Pieces/wk or glasses/d): 1-y post boys F: 5.8 5.3 Soft drinks: 2.7 2.5 Water: 4.1 2.8 1-y post girls F: 6.2 5.2 Soft drinks:
(Half serving portions) F: 0.8 0.6
(Pieces/wk or glasses/d): 1-y post boys F: 4.6 4.9 Soft drinks: 3.6 2.7 Water: 3.6 2.8 1-y post girls F: 5.7 5.1 Soft drinks:
(Half serving portions) F: 2.6 2.3
Not provided
Not provided
(continued)
Gender (clear gender differences, so all analyses applied in boys and girls separately); analyses also adjusted for age and SES
Not provided
Not provided
Not provided
Total FV: 4.9 0.5 Total F: 2.7 0.4 Total V: 2.2 0.2
Total FV: 4.7 0.5 Total F: 3.2 0.5 Total V: 1.5 0.2
Not provided
Control immediate Sex, level of family involvement, mean BMI: weight status 19.0 0.3 Intervention immediate mean BMI: 19.0 0.2 Control 6 mo mean BMI: 19.5 0.1 Intervention 6 mo mean BMI: 19.4 0.1
Not provided
BMI, and SES; possible effect modifications owing to gender, race, free and reduced-lunch eligibility, and BMI tested
(Servings/d [mean SE]) FV (immediate): 4.4 (0.2) FV (6 mo: 4.1 (0.2)
(In servings) FV: 1.68 1.13
Intervention mean [SE]) BMI: 19.00 Postintervention y 1 3.64 FV combined: 2.91 (0.20) F: 1.76 (0.11) V: 1.14 (0.11) Postintervention y 2 FV combined: 2.84 (0.20) F: 1.68 (0.11) V: 1.16 (0.10)
(Servings/d [mean SE]) FV (immediate): 4.2 (0.1) FV (6 mo): 4.0 (0.1)
(In servings) FV: 2.00 1.10
[SE]) Postintervention y 1 FV combined: 2.89 (0.19) F: 1.68 (0.11) V: 1.20 (0.10) Postintervention y 2 FV combined: 2.76 (0.19) F: 1.58 (0.10) V: 1.18 (0.10)
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 525
(Average daily use of (Average daily use of Not provided table salt or FV table salt or FV as % of total as % of total foods) foods) Salt use: Salt use: 0.30 0.42 0.35 0.48 FV: 21.38 8.49 FV: 22.06 7.98
Not provided
Parcel (1989)79
F frequency previous wk (servings/d): 1.1 (0.8) 24-h recall V (g): 76.9 (66.7)
F frequency previous wk (servings/d): 1.2 (0.7) 24-h recall V (g): 77.9 (64.6)
Control mean BMI: 17.8 (2.81) Intervention mean BMI: 17.8 (2.90)
Baseline Weight Status
Mangunkusumo (2007)78
(Servings) F: 1.90 (1.30) V: 1.83 (1.38) Packaged snacks: 0.56 (0.72) Fast food: 0.17 (0.44) Sweet drink: 2.07 (1.60)
Baseline Intervention Group Diet Measures
(Servings) F: 1.96 (1.32) V: 1.74 (1.44) Packaged snacks: 0.63 (0.78) Fast food: 0.20 (0.45) Sweet drink: 1.79 (1.64)
Baseline Control Group Diet Measures
Johnson (2012)77
First Author (Publication Year)
Table 2. Continued
(Average daily use of table salt or FV as % of total foods) Posttest 1 salt use: 0.29 0.41 FV: 23.92 7.76 Posttest 2 salt use: 0.18 0.30 FV: 22.15 8.50
F frequency previous wk (servings/d): 1.2 (0.7) 24-h recall V (g): 74.9 (69.7)
Not provided
Control mean BMI: 19.2 (3.61) Intervention mean BMI: 19.7 (3.88)
Postintervention Weight Status
(Average daily use of Not provided table salt or FV as % of total foods) Posttest 1 salt use: 0.19 0.33 FV: 23.07 8.18 Posttest 2 salt use: 0.18 0.32 FV: 23.26 8.37
F frequency previous wk (servings/d): 1.1 (0.7) 24-h recall V (g): 76.9 (68.7)
(Servings) F: 2.20 (1.44) V: 2.03 (1.53) Packaged snacks: 0.69 (0.82) Fast food: 0.16 (0.41) Sweet drink: 1.52 (1.52)
2.6 2.2 Water: 3.7 2.6 2-y post boys F: 4.7 5.1 Soft drinks: 3.5 2.8 Water: 3.7 3.0 2-y post girls F: 5.8 5.6 Soft drinks: 2.3 2.1 Water: 3.8 2.8
2.3 2.2 Water: 3.7 2.7 2-y post boys F: 6.4 5.4 Soft drinks: 2.7 2.5 Water: 3.9 2.6 2-y post girls F: 6.6 5.7 Soft drinks: 2.1 2.1 Water: 4.0 2.8 (Servings) F: 2.01 (1.48) V: 1.80 (1.65) Packaged snacks: 0.75 (0.84) Fast food: 0.14 (0.38) Sweet drink: 1.51 (1.44)
Postintervention Group Diet Measures
Postintervention Control Group Diet Measures
Not provided
Analyses adjusted for region, gender, age, and baseline measure
Confounders were child gender, child age in y, and maternal education level and group
Possible Moderating Variables
526 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
(Servings/d) Alabama FV: 2.96 2.26
(Servings/d) Alabama FV: 2.91 2.37
Reynolds (2004)83
Distribution: 24-h FJV (times/d): 2.87 (1.19) F (portions/d): 1.25 (0.83) V snack (times/d): 0.32 (0.26) V (g/d): 43.6 (22.1)
Multicomponent: 24-h FJV (times/d): 2.47 (1.06) F (portions/d): 1.08 (0.72) V snack (times/d): 0.22 (0.22) V (g/d): 45.2 (25.2)
24-h FJV (times/d): 2.63 (1.25) F (portions/d): 1.12 (0.68) V snack (times/d): 0.28 (0.25) V (g/d): 47.7 (24.1)
Reinaerts (2007)81
(Portions) F: 1.9 (1.4) V: 1.5 (1.5) FV eaten daily: 3.4 (2.2)
Reinaerts (2008)82
(Portions) F: 1.8 (1.3) V: 1.6 (1.5) FV eaten daily: 3.4 (2.2)
Ransley (2007)80
Not provided
Not provided
Not provided
(Servings/d) Alabama FV: 2.74 2.12
24-h FJV (times/d): 2.88 (1.16) F (portions/d): 1.07 (0.65) V snack (times/d): 0.29 (0.24) V (g/d): 46.5 (25.2)
(Portions) At 3 mo F: 1.7 (1.4) V: 1.6 (1.6) FV eaten daily: 3.3 (2.3) At 7 mo F: 1.6 (1.3) V: 1.6 (1.5) FV eaten daily: 3.2 (2.1)
Not provided
(Servings/d) Alabama FV: 4.20 2.96
Multicomponent: Follow-up I 24-h FJV (times/d): 3.10 (1.12) F (portions/d): 1.27 (0.70) V snack (times/d): 0.31 (0.29) V (g/d): 43.6 (24.7) FU II 24-h FJV: 3.02 (1.29) F: 1.18 (0.75) V snack: 0.26 (0.24) V: 48.4 (24.0) Not provided
Not provided Distribution: Follow-up I 24-h FJV (times/d): 3.38 (1.18) F (portions/d): 1.41 (0.83) V snack (times/d): 0.38 (0.28) V (g/d): 49.0 (26.5) Follow-up II 24-h FJV: 3.22 (1.13) F: 1.25 (0.76) V snack: 0.40 (0.30) V: 49.6 (25.6)
(Portions) At 3 mo F: 1.9 (1.4) V: 1.5 (1.6) FV eaten daily: 3.4 (2.3) At 7 mo F: 1.6 (1.4) V: 1.5 (1.5) FV eaten daily: 3.1 (2.1)
(continued)
Possibly race or ethnicity but not tested
Not provided
Unclear
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 527
Minnesota FV: 4.91 2.93
Baseline Intervention Group Diet Measures Baseline Weight Status
Children of Dutch ethnicity: F (pieces/d): 1.60 (1.08) V (g/d): 97.2 (42.1) Children of nonWestern ethnicity: F (pieces/d): 2.13 (1.23) V (g/d): 120.4 (62.7)
Tak (2007)86
Thompson (2009)87 (Means [SE]) FJ: 2.3 (0.1) Low-fat V: 1.6 (0.0)
Total V consumption: Boys: 1.75 (1.94) Girls: 2.97 (3.72)
Not provided
(Means [SE]) FJ: 2.5 (0.1) Low-fat V: 2.0 (0.1)
(Mean BMI percentile) Control spring: 63.6 (3.6) Control fall: 64.5 (2.2) Intervention spring: 64.7 (3.0)
Not provided Children of Dutch ethnicity: F (pieces/d): 1.54 (1.04) V (g/d): 99.1 (47.6) Children of nonWestern ethnicity: F (pieces/d): 1.97 (1.14) V (g/d): 120.6 (66.3)
Total V consumption: Boys: 2.45 (2.80) Girls: 2.54 (2.57)
FV servings/d (scale FV servings/d (scale Control mean BMI percentile: 0–8): 5.0 2.0 0–8): 3.7 1.9 64.5 23.8 Eating with Eating with Intervention mean television scale television scale BMI percentile: (scale 0–4): (scale 0–4): 65.2 27.0 1.1 0.7 1.1 0.8 Days/wk SSB (scale Days/wk SSB (scale 0–7): 3.1 2.2 0–7): 2.6 2.4
Minnesota FV: 4.74 2.72
Baseline Control Group Diet Measures
Smolak (1998)85
Rosenkranz (2010)84
First Author (Publication Year)
Table 2. Continued
Minnesota FV: 5.25 2.93
Postintervention Group Diet Measures Postintervention Weight Status
(Means [SE]) Post-1 FJ: 2.9 (0.1) Post-1 low-fat V: 1.9 (0.0) Post-2 FJ: 3.0 (0.1) Post-2 low-fat V: 2.2 (0.0)
Children of Dutch ethnicity: F (pieces/d): 1.37 (0.89) V (g/d): 93.8 (38.2) Children of nonWestern ethnicity: F (pieces/d): 1.77 (1.08) V (g/d): 104.2 (50.6)
Total V consumption: Boys: 2.92 (2.43) Girls: 1.93 (2.10)
Not provided
(Means [SE]) Post-1 FJ: 3.5 (0.1) Post-1 low-fat V: 2.5 (0.1) Post-2 FJ: 2.8 (0.1) Post-2 low-fat V: 2.1 (0.1)
Adjusted for children’s age, gender, parental education level, region of residence, and baseline intake levels; effect modification by gender and educational level assessed
Not provided
Parenting style and family cohesion
Possible Moderating Variables
Analyses rerun with (Mean BMI demographic and percentile) anthropometric Control spring Postcharacteristics; 1: 63.7 (3.6) neither household Control fall post-1: education nor 64.6 (2.2) psychosocial Control spring postcharacteristics 2: 67.0 (3.6)
Not provided Children of Dutch ethnicity: F (pieces/d): 1.55 (0.93) V (g/d): 102.5 (42.2) Children of nonWestern ethnicity: F (pieces/d): 1.80 (1.01) V (g/d): 120.2 (64.5)
Total V consumption: Boys: 2.17 (2.35) Girls: 2.47 (2.37)
FV servings/d (scale Control mean BMI FV servings/d percentile: 0–8): 4.9 1.7 (scale 0–8): 62.2 23.2 Eating with 3.7 1.8 Intervention mean television scale Eating with BMI percentile: (scale 0–4): television scale 64.8 26.9 0.8 0.7 (scale 0–4): Days/wk SSB (scale 1.1 0.7 0–7): 2.3 2.4 Days/wk SSB (scale 0–7): 2.2 2.4
Minnesota FV: 4.66 2.76
Postintervention Control Group Diet Measures
528 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
FV (servings): 2.3 1.0 kcal: 2,806 975
SCT + MI FV (servings): 2.6 1.4 SCT + MI kcal: 2,655 529 SCT only FV (servings): 2.5 1.2 SCT only kcal: 2,751 816
Not provided
FV (servings): 3.3 2.1 kcal: 2,670 911
SCT + MI FV (servings): 5.7 2.2 SCT + MI kcal: 2,535 493 SCT only FV (servings): 4.8 2.4 SCT only kcal: 2,818 708
Not provided
Control fall post-2: 69.3 (2.3) Intervention spring post-1: 64.7 (3.0) Intervention fall post-1: 67.5 (2.3) Intervention spring post-2: 67.8 (3.0) Intervention fall post-2: 70.5 (2.3) Results remained unchanged after controlling for sex and parental marital status
BMI indicates body mass index; F, fruit; FJ, fruit and juice; FJV, fruit, juice, and vegetables; FV, fruit and vegetables; MI, motivational interviewing; MVPA, moderate to vigorous physical activity; PL + C + E, Peer leaders plus curriculum plus environment; SCT, Social Cognitive Theory; SES, socioeconomic status; SSB, sugar-sweetened beverage; V, vegetables.
Wilson (2002)88
Intervention fall: 67.5 (2.3)
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 529
Unclear
Method of random digits
Unclear
Unclear
Statistical Package for the Social Sciences randomization procedure
Angelopoulos (2009)61
Baranowski (2000)62
Birnbaum (2002)63
Black (2010)64
Random Sequence Generation
Anderson (2005)60
First Author (Publication Year)
Randomization occurred after recruitment and baseline assessments, so risk of
Random assignment done after all baseline measures taken, so risk of selection bias was low
Yes
Unclear
Unclear
Allocation Concealment
Unclear
Unclear
Unclear
Unclear
Unclear
Blinding
Table 3. Summary of Risk of Bias (or Study Quality) for Each Intervention
High attrition rate, so risk of attrition bias was high
Unclear
No statistically significant differences detected in those remaining in cohort vs those not, and no differential attrition detected, so risk of attrition bias was low
Unclear
Only 44% of participants completed food diaries, but unclear how many dropped out of program
Incomplete Outcome Data Other Sources of Bias
Appeared to report complete data
1. Method of diet assessment was selfreport through Youth Adolescent FFQ 2. Power analysis conducted; sample size provided power of at least 0.77 to detect moderate effect size for change in overweight/obese status
1. Method of diet assessment was selfreport through Behavioral Risk Factor Surveillance System 2. Power analyses and sample size determination reflected study design 3. Hierarchical structure of data was taken into account in analyses
1. Method of diet assessment was selfreport but as 7-d food record 2. Power analysis conducted beforehand 3. School was unit of random assignment and analysis. Analyses did not account for clustering of students within schools.
Appeared to report complete data
Appeared to report complete data
1. Method of diet assessment was selfreport but as 24-h dietary recall 2. Power analysis conducted beforehand 3. School was main sampling unit. Analyses did not account for clustering of students within schools, although interschool variation was taken into account.
1. Method of diet assessment was selfreport through 3-d food diary 2. Power analysis conducted beforehand 3. Analyses did not take into account clustering at school level
Appeared to report complete data
Appeared to report complete data
Selective Outcome Reporting
530 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
Unclear
Computergenerated random number assignment
Unclear
Branscum (2013)65
Chen (2011)66
Cullen (1997)67 Unclear
Randomization occurred after recruitment and baseline assessments, so risk of selection bias was low
Unclear
selection bias was low
Unclear
Unclear
Unclear
18.5% not present at follow-up and no significant differences in demographics or pretest FV intake for those completing and not completing
93% of participants completed baseline and all follow-up measures, and no significant differences were found between those who provided followup data and those who dropped out. In addition, mixed modeling allowed for missing data. Thus, there was low risk of attrition bias.
Appears no participants dropped out during study, so risk of attrition bias was low
Appeared to report complete data
Web program addressed adequate dietary intake and used food diaries from 3 d but authors reported only FV intake. There should be more complete outcome data reporting related to dietary intake, so risk of selective outcome reporting was high.
Appeared to report complete data
(continued)
1. Method of diet assessment was selfreport through FFQ 2. Sample size seemed adequate 3. Analyses used troop as unit of measure
1. Method of diet assessment was selfreport through 3-d food diary 2. Sample size was small but power analysis conducted beforehand to determine sample size
1. Method of diet assessment was selfreport through School Physical Activity and Nutrition questionnaire 2. Sample sizes were small (ie, 37 in theory-based group and 34 in knowledge-based group) 3. Two interventions compared, rather than no intervention serving as control 4. Measurement bias: ‘‘Some children complained about the instrument’s length, which could have introduced bias if they hurried through it, instead of taking the time to fully read and respond to the items.’’
3. Contamination bias: Intervention and control adolescents lived in same communities, raising possibility of contamination and possibly reducing ability to detect intervention effects
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 531
Not available (not randomized)
Unclear
Unclear
Unclear
Day (2008)69
Domel (1993)70
Dzewaltowski (2009)71
Random Sequence Generation
Davis (2011)68
First Author (Publication Year)
Table 3. Continued
Children assigned to intervention or control based on school assignment, so risk of
Unclear
Children assigned to intervention or control based on school, so risk of selection bias was high
Those who were not enrolled in LA’s BEST served as controls, so risk of selection bias was high
Allocation Concealment
Unclear
Unclear
Unclear (but does not seem possible)
Unclear (but does not seem possible)
Blinding
High attrition rates (with more dropouts in control school), so risk of attrition bias was high
9% of intervention students excluded from analyses for incomplete diaries owing to forgotten meals, attrition, or absences; 20% excluded in control school. Because attrition rates differed, risk of attrition bias was high.
Participation rates for eligible students did not differ between usual practice (66%) and intervention (64%) schools
Unclear
post measures, so risk of attrition bias was low
Incomplete Outcome Data
Appeared to report complete data
Appeared to report complete data
Appeared to report complete data
Appeared to report complete data
Selective Outcome Reporting
1. Method of diet assessment was selfreport through Youth/Adolescent FFQ 2. Sample seems adequate, but no power analysis performed 3. Schol was unit of random assignment and analysis. Analyses did not account for clustering of students within schools.
1. Method of diet assessment was food diaries validated with school lunch observations 2. The design included only control and 1 intervention, so statistical power of inferences was limited. However, design was appropriate for initial evaluation of new curriculum. 3. School was unit of random assignment and analysis. Analyses did not account for clustering of students within schools.
1. Method of diet assessment was selfreport through FFQs 2. Statistical power unclear 3. Analyses did not account for clustering of students within schools
1. Method of diet assessment was selfreport through FFQ screener 2. Sample size was relatively small 3. Analyses did not account for clustering of students 4. Groups were not randomized
Other Sources of Bias
532 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
Unclear
Unclear
Not available (not randomized)
Not available (not randomized)
Unclear
Evans (2006)72
Gentile (2009)73
Govula (2007)74
Gribble (2003)75
Haerens (2006)76 Unclear
Those who were unable to attend > 1 session served as controls, so risk of selection bias was high
Children assigned to intervention or control based on school, so risk of selection bias was high
Yes
Children assigned to intervention or control based on school assignment, so risk of selection bias was high
selection bias was high
Unclear
Unclear
Unclear
Families were not told whether they were in intervention or control group so risk of performance bias was low
Unclear
There were 704 dropouts, and those not participating at follow-up were significantly
Unclear
Attrition rate seemed to be 0%, so risk of attrition bias was low
77% of children and 48% of parents provided data at all 3 time points, so risk of attrition bias was high
Because of high attrition rate (27% to 28%), risk of attrition bias was high. However, children who dropped out were not significantly different in demographics from children who did not drop out.
Appeared to report complete data
Appeared to report complete data
Appeared to report complete data
Appeared to report complete data
Appeared to report complete data
(continued)
Method of diet assessment was self-report through FFQ
1. Method of diet assessment was selfreport through 3-d food record 2. Groups were not randomized
1. Method of diet assessment was selfreport through FFQ 2. Sample size was small 3. Analyses did not account for clustering of students within schools 4. Groups were not randomized
1. Method of diet assessment was selfreport (adapted from National Youth Risk Behavior Survey) 2. Low power to detect differences 3. Analyses did not account for clustering of students within schools 4. Selection bias: ‘‘The 2 school districts were approached due to the requirements of funding agencies.’’
1. Method of diet assessment was selfreport but as 24-h dietary recall 2. Sample size was small and study did not have enough power 3. Analyses did not account for clustering of students within schools 4. Unclear whether groups were randomized
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 533
Unclear
Unclear
Unclear
Mangunkusumo (2007)78
Parcel (1989)79
Random Sequence Generation
Johnson (2012)77
First Author (Publication Year)
Table 3. Continued
Unclear
Parents and children, blinded regarding which group they belonged to, received information about project, and parents were allowed several weeks to actively give written consent for participation, so risk of selection bias was low
Unclear
Allocation Concealment
Unclear
Parents and children were blinded regarding which group they belonged to, so risk of performance bias was low
Unclear
Blinding
Appeared to report complete data
Appeared to report complete data
97% completed pre and post measures, so risk of attrition bias was low
Control group seemed to lose participants
Appeared to report complete data
Selective Outcome Reporting
Data missing for individual level variables ranged from 12% to 20% and 11% to 30% for householdlevel variables. However, followup checks indicated data were missing at random.
older and consumed more soft drinks than those participating at follow-up. Thus, risk of attrition bias was high.
Incomplete Outcome Data
1. Method of diet assessment was selfreport using FFQs 2. Low power
1. Method of diet assessment was selfreport through FFQs 2. Power analysis conducted beforehand 3. Analyses accounted for multiple levels
1. Method of diet assessment was selfreport through telephone interview of parents 2. Sample size seemed adequate 3. Analyses accounted for nested structure of data using multilevel analyses
Other Sources of Bias
534 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
Not available (not randomized)
Unclear
Unclear
Random number generator
Not available (not randomized)
Ransley (2007)80
Reinaerts (2007)81 Reinaerts (2008)82
Reynolds (2004)83
Rosenkranz (2010)84
Smolak (1998)85 Not available (no allocation because all children in curriculum
Yes
Children assigned to intervention or control based on school assignment, so risk of selection bias was high
Unclear
Children were recruited after assignment of schools, so risk of selection bias was high
Unclear
Research assistants who collected data were blinded to condition of each troop, so risk of detection bias was low
Unclear
Unclear
Unclear
88% of children participated in both pretest and posttest, so risk of attrition bias was low
Dropouts were low (n ¼ 4, all from family relocation) and unrelated to troop assignment or study, so risk of attrition bias was low
Unclear
Multilevel logistic analyses of dropouts revealed some selective dropout, so risk of attrition bias was high
Unclear
between baseline and posttest 2, whereas intervention group gained participants, so risk of attrition bias is unclear
(continued)
1. Method of diet assessment was selfreport through survey 2. Sample size for controls appeared low 3. Analyses did not take into account clustering at class level 4. Groups were not randomized
1. Method of diet assessment was selfreport through questionnaire 2. Sample was small and study’s statistical power appears low for physical activity and FV 3. Analyses adjusted for clustered data structure
Appeared to report complete data
Appeared to report complete data
1. Method of diet assessment was selfreport but as 24-h dietary recall 2. Power estimates conducted beforehand, although sample sizes were small 3. Analyses did not account for clustering of students within schools
1. Method of diet assessment was selfreport through FFQ by parents 2. Sample size seemed adequate 3. Multilevel analyses used to account for nested sampling design 4. Schools were not randomly assigned to control groups but matched based on school size and ethnicity
1. Method of diet assessment was selfreport by parents 2. Power analysis conducted beforehand 3. Analyses took into account levels in model (including school) 4. Groups were not randomized
Appeared to report complete data
Appeared to report complete data
Appeared to report complete data
3. Authors used student as unit of analysis to test whether effect had occurred. Because school was unit of intervention, authors also tested effects using school as the unit of analysis. 4. Unclear whether groups were randomized
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 Diep et al 535
Coin toss
Unclear
Thompson (2009)87
Wilson (2002)88 Students were recruited before randomization, so risk of selection bias was low
Troops were randomized to condition after baseline assessment, so risk of selection bias was low
Children were assigned to intervention or control based on school, so risk of selection bias was high
classrooms participated)
Allocation Concealment
Unclear
Participants, data collectors, and interventionists were not blinded to condition, so risk of performance and detection biases were high
Unclear
Blinding
Percentage of participants who remained in program was higher for SCT + MI (83%) than for SCT-only (70%) or comparison groups (69%). Because of high and unequal attrition rates, risk of attrition bias was high.
Missing data percentage was small (0.4% to 1.9%). No significant group missing data status interactions, so risk of attrition bias was low.
There were 86% children and 70% parents at followup, and selective dropout was found for parents in control group for those residing in eastern region and for children who reported lower F intake at baseline. Thus, risk of selection bias was high.
Incomplete Outcome Data
Appeared to report complete data
Appeared to report complete data
Appeared to report complete data
Selective Outcome Reporting
1. Method of diet assessment was selfreport through 3-d dietary food record 2. Sample size was small and too low statistical power to detect differences 3. Analyses did not account for clustering of students within schools
1. Method of diet assessment was selfreport through FFQ 2. Power analysis conducted beforehand
1. Method of diet assessment was selfreport through FFQs 2. Power issues for parent-report data 3. Analysis took into account clustering within schools/classes 4. Groups were not randomized
Other Sources of Bias
F indicates fruit; FFQ, Food Frequency Questionnaire; FV, fruit and vegetables; MI, motivational interviewing; SCT, Social Cognitive Theory.
Not available (not randomized)
Random Sequence Generation
Tak (2007)86
First Author (Publication Year)
Table 3. Continued
536 Diep et al Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 with results classified as outliers.56 Heterogeneity is present when the Q statistic exceeds the critical value of the chi-square with k – 1 degrees of freedom. In a related manner, I2 describes the ‘‘. percentage of total variation across studies that is due to heterogeneity rather than chance.’’57 Higgins et al57 suggested that an I2 over 75% signals high heterogeneity and 50% to 75% signals moderate heterogeneity. The authors assessed possible publication bias by evaluating funnel plots of the trial mean differences for asymmetry, which can result from the nonpublication of small trials with negative results. Funnel plots of studies showed effect sizes on the horizontal axis and standard errors (or precision) on the vertical axis. Egger's test of asymmetry tested for publication bias.58 Subgroup analyses assessed differences in intervention effectiveness by theory use (yes vs no), study quality (high vs low), and formal planning process (yes vs no). In addition, a metaregression analysis assessed whether the number of theories increased intervention effectiveness. The authors used random effects modeling59 to take into account the amount of variance at the participant and study level and to generalize the findings to a larger population, of which the studies actually carried out are a sample.42,43 In addition, to control for study quality (which could account for differences/ discrepancies in outcomes), the authors conducted separate randomeffects meta-regressions for F, V, and FV combined, using a maximum likelihood estimation with KnappHartung adjustment. Study characteristic (ie, theory use, number of theories, or formal planning process) was the independent variable and study quality was the control variable. T-AC, a statistician, conducted all analyses using Comprehensive MetaAnalysis software version 2.0 (Biostat, Inc, Englewood, NJ, 2005) and version 3.0 (2014) for F, V, and FV combined consumption.
RESULTS Included Studies The authors screened 443 articles, 29 of which met inclusionary and exclusionary criteria for this literature review and meta-analysis60–88
(Figure 1). Three articles reported results from multiple interventions or intervention groups,63,83,88 so these interventions' results were averaged. Two articles reported on the same 2 interventions,81,82 so they were combined to characterize the study. In total, 29 articles were included in this literature review and metaanalysis. Approximately 29% of screened articles were excluded because they did not have means and standard deviations of diet in each group at baseline and postintervention available for abstraction, 25% did not have a control group for comparison, and 17% did not target dietary change. The remaining excluded studies did not report F and/or V outcomes, did not target children (aged 2–18 years), targeted already overweight/obese children, were not research articles, or were not in English. All 29 articles included in this review appeared in journals between 1989 and 2013, with a majority published in or after 2000 (86.2%). Approximately 65.5% occurred in the US; the remaining occurred in Australia, Belgium, Canada, England, Greece, Scotland, and The Netherlands. Table 1 provides complete study method details.
Risk of Bias It was difficult to measure risk of bias (ie, study quality) for numerous studies because of unavailable or unclear information, especially for random sequence generation, allocation concealment, and blinding. Based on the information provided, 72.4% of studies clearly randomized participants to intervention or control group(s), but only 5 (17.2%) reported random sequence generation by random digits, Statistical Package for the Social Sciences, coin toss, or random number generator, and 9 (31.0%) reported concealment of allocation sequence, mainly by randomizing children to groups after recruitment and baseline assessment. Three studies (10.3%) reported blinding of participants, the interventionist, and/or outcome assessors. Ten studies (34.5%) adequately addressed risk of incomplete outcome data, and 28 (96.6%) had a low risk of selective outcome reporting (ie, they
Diep et al 537 appeared to report on all measured outcomes). Other potential sources of bias included self-reported measures of diet, particularly food frequency questionnaires; limited statistical power; not accounting for clustering at group levels in analyses; not randomizing participants to intervention or control group(s); and other sources of contamination, measurement, or selection biases. Table 3 provides complete risk of bias details.
Theory Use and Formal Planning Process Table 1 details the extent to which interventions were based on theory or a formal planning process. In the 29 articles, there were a total of 33 interventions/intervention groups, some from the same article but using different theories. Eight interventions (24.2%) had no theoretical foundation, 15 (45.5%) employed 1 theory, and the remaining 10 (30.3%) employed 2 theories. Of the 25 interventions based on theory, Social Cognitive Theory was most common, used to develop 17 interventions (51.5%). Theory of Planned Behavior guided 6 interventions (18.2%). Transtheoretical Model, Social Learning Theory, Motivational Interviewing, Life Course Theory, Behavioral Change Theory, Organization Change Model, and ecological models guided 1 or a few interventions each. Seven interventions (21.1%) included some sort of formal planning process to inform intervention design and theory use. These formal planning processes included qualitative formative research, systematic program planning approach, social marketing method of focus groups, and intervention mapping.33
Meta-Analyses Results The authors separated meta-analyses by F, V, and FV results; not all 29 articles reported each.
Fruit. Figure 2A illustrates a forest plot showing F effect sizes with 95% confidence interval (CI) for each study. The effectiveness of interventions designed to increase F consumption varied by study, from 0.048 to 4.827, meaning the means of the
538 Diep et al
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
A
B
C
Figure 2. Forest plots for studies included in the (A) fruit, (B) vegetables, and (C) fruit and vegetables meta-analyses, by effect size. T indicates theory use; N, no theory use. Note: Study weight is indicated by the size of the marker in the Forest plot; total effect size is indicated by diamonds. Not all 29 studies were included in (A), (B), and (C) because not all reported fruit, vegetables, and fruit and vegetables results, respectively.
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 intervention and control groups differed by approximately 0.048 to 4.827 SDs. With all the studies (k ¼ 14) included, the overall mean of the distribution of effects indicated a moderate effect on F, g ¼ 0.543 (95% CI, 0.345–0.742) (P < .05). When analysis excluded an outlier with an extreme effect size (g ¼ 4.82760), the mean effect size weighted by sample size showed a significant small effect of intervention on F consumption, g ¼ 0.322 (95% CI, 0.186–0.458) (P < .05). Significant heterogeneity was detected across studies (Q ¼ 126.26; P < .001; I2 ¼ 90.50%), which suggested that moderating factors such as being based on theory or not may account for differences in outcomes.
Vegetables. With all studies (k ¼ 16) included, the overall mean of the distribution of effects indicated a large effect on V, g ¼ 0.755 (95% CI, 0.450–1.061) (P < .05). After deleting outliers (g ¼ 8.46287 and g ¼ 3.41574), the estimated overall mean effect size of interventions on V consumption dropped to g ¼ 0.174 (P ¼ .001), suggesting that interventions had a small but significantly greater impact on V consumption than control conditions (Figure 2B). Heterogeneity tests revealed that variance in effect sizes for V consumption (Q ¼ 80.27; P < .001; I2 ¼ 83.81%) differed significantly more than that explained by random sampling error alone.
Fruit and vegetables combined. With all studies (k ¼ 28) included, the overall
Diep et al 539
mean of the distribution of effects indicated a large effect on FV combined (g ¼ 1.002; 95% CI, 0.472–1.532; P < .05). After removing 2 studies with outlying effect sizes (g ¼ 8.46287 and g ¼ 6.32263), interventions had a moderate effect on FV consumption (g ¼ 0.47; 95% CI, 0.31–0.63; P < .001). Heterogeneity tests were significant (Q ¼ 592.95; P < .001; I2 ¼ 95.78%). All subsequent analyses omitted the outlying studies60,63,74,87 mentioned earlier; their characteristics are detailed in Tables 1–3.
Subgroup Analyses and MetaRegression To determine whether study quality, theory use, number of theories, or a formal planning process was associated with dietary change, separate moderator analyses were conducted for each outcome variable of interest: F (n ¼ 13), V (n ¼ 14), and FV combined (n ¼ 26) (Table 4). Separate meta-regressions also controlled for study quality (Table 5).
Study quality. Significant differences in Hedges' g were detected for F, V, and FV combined between high- and low-quality studies (P < .001) (Table 4). High-quality studies had significantly higher Hedges' g for F (g ¼ 0.4) and V (g ¼ 0.25) consumption than low-quality studies (F, g ¼ 0.212; V, g ¼ 0.09). However, the relationship was reversed for FV combined: lowquality studies showed significantly
higher effect size (g ¼ 0.529) than high-quality studies (g ¼ 0.437).
Theory use. The effect sizes of theorybased and non–theory based studies differed significantly (P < .001) (Table 4) across F, V, and FV combined consumption. The effect was stronger when the interventions were theorybased (g ¼ 0.389 for F, 0.181 for V, and 0.524 for FV) than non–theory based (g ¼ 0.157 for F, 0.138 for V, and 0.191 for FV). After controlling for study quality, theory use remained a significant predictor for V only (b ¼ 0.373; P < .001) (Table 5), for which the R2 was 0.87. Number of theories. Meta-regression analyses revealed no association between the number of theories and F, V, or FV combined consumption, with and without controlling for study quality (Tables 4 and 5).
Formal planning process. There were significant differences in F and FV combined consumption between studies with and without formal planning processes. For F, interventions with a formal planning process showed a higher effect size (g ¼ 0.435) than interventions without (g ¼ 0.284). For FV combined, however, interventions not incorporating a formal planning process resulted in a higher effect size (g ¼ 0.5) than interventions incorporating a formal planning process (g ¼ 0.284). There were no significant differences for V
Table 4. Subgroup Analyses of Intervention Effects by Theory Use, Study Quality, and Formal Planning Process F
Theory use (yes/no) Yes No Study quality (high/low) High Low Formal planning process (yes/no) Yes No
V
FV
Mean
SE
n
Mean
SE
n
Mean
SE
n
0.389* 0.157*
0.101* 0.058*
8* 5*
0.181* 0.138*
0.078* 0.052*
9* 5*
0.524* 0.191*
0.098* 0.076*
20* 6*
0.400* 0.212*
0.104* 0.057*
9* 4*
0.250* 0.090*
0.097 0.03
7* 7*
0.437* 0.529*
0.104* 0.153*
16* 10*
0.435* 0.284*
0.131* 0.081*
3* 10*
0.176 0.171
0.052 0.064
3 11
0.284* 0.500*
0.082* 0.093*
3* 23*
F indicates fruit; FV, fruit and vegetables; V, vegetables. *P < .001. Note: Not all 29 studies were included in each subgroup analysis because not all reported fruit, vegetables, and fruit and vegetables results, respectively.
540 Diep et al
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
Table 5. Meta-Regression Analyses of Intervention Effects by Theory Use, Number of Theories, and Formal Planning Process, While Controlling for Study Quality F b Theory use (reference ¼ no) Study quality (reference ¼ high) R2 Number of theories (reference ¼ no) Study quality (reference ¼ high)
V
FV
SE
P
b
SE
P
.364
0.173
.061
.373
0.082
< .001
.337
0.170
.076
–.407
0.079
< .001
.520
.870
SE
P
.194
b
0.249
.446
.028
0.217
.899
.090
.091
0.095
.362
.091
0.053
.115
.081
0.124
.520
.227
0.187
.253
–.237
0.091
.025
.145
0.203
.484
R2
.200
Formal planning process (reference ¼ no)
.275
0.195
.190
.067
0.104
.531
.227
0.283
.431
.252
0.176
.184
–.166
0.087
.082
.141
0.198
.485
Study quality (reference ¼ high) R2
.330
.430
.030
.260
.060
F indicates fruit; FV, fruit and vegetables; V, vegetables. Note: Not all 29 studies were included in each subgroup analysis because not all reported fruit, vegetables, and fruit and vegetables results, respectively. consumption. After controlling for study quality, there were no significant relationships between formal planning process and F, V, or FV consumption (Table 5).
Publication Bias Figure 3A–C illustrates funnel plots of studies for dietary intake of F, V, and FV combined. Hedges' g for each outcome of interest was plotted against its standard error. Although some studies fell outside the envelope, the P of Egger's test of asymmetry for F, V, and FV combined was 0.09, 0.26, and 0.20, respectively, which suggests that unpublished results were unlikely to have an effect.
DISCUSSION This study assessed whether theory involvement in child FV interventions made a difference. To the best of the authors' knowledge, this was the first systematic assessment of whether basing child FV interventions on behavioral theory enhanced its effectiveness. Interventions were effective in enacting dietary changes, even if temporarily. With outliers removed, interventions had small to moderate impacts on F, V, and FV combined consumption compared with control conditions. Significant heterogeneity across studies indicated that moderators (eg, theory use) were possible sources.
Without taking into account study quality, interventions based on a theoretical framework had significantly larger effect sizes for F, V, and FV combined consumption than interventions without a theoretical framework. Specifically, the differences in mean Hedges' g between theory and non–theory based interventions were 0.232 (small), 0.043 (small), and 0.333 (small to moderate) for F, V, and FV, respectively. A similar meta-analysis among adults found contradictory evidence: interventions using theory had smaller intervention effects than those not using theory.17 However, with study quality taken into account in the current analyses, theory use was positively related to V consumption only and the effect size was small. This finding suggests that theory may have a role in increasing intervention effectiveness but the effect sizes owing to theory were not large and were confounded by study quality. Theorybased studies were equally split between high quality and low quality, but the majority of non-theory studies were of low quality. In addition, in the current meta-analysis, some interventions without theory did really well and other interventions with theory did not do well. Thus, theory by itself does not appear to be the primary answer. Having a compilation of which behavior change procedures are effective for which theory construct, dietary behavior, and/or population group may provide clarity and insight as to
how to optimally operationalize theory in practice.
Theory does not appear to provide a primary answer to increasing intervention effectiveness. There was no relationship between number of theories in an intervention and intervention effectiveness, with or without controlling for study quality. These findings contradict findings from the meta-analysis among adults,17 revealing that interventions with a single theory had larger effects than interventions with multiple theories. Despite current beliefs that polytheoretical interventions may be more effective,32 the number of theories may not be as important as how well theory is used, or theories may make contradictory assumptions about how to change behavior.89 Most interventions (78.6%) reported a theoretical basis. The most commonly reported theory was Social Cognitive Theory, which is also most widely used for designing nutrition education programs in general.10 However, when examining interventions by effect size, no theory was consistently used in the interventions with the largest effect sizes (ie, no theory was better than others).32 This is similar to findings from the meta-analysis among
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
Funnel Plot of Standard Error by Hedges' g 0.0
Standard Error
0.1
0.2
0.3
0.4
0.5 -2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
Hedges's g
Funnel Plot of Standard Error by Hedges' g
Standard Error
0.0
0.1
0.2
0.3
0.4 -2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
Hedges's g
Funnel Plot of Standard Error by Hedges' g
Standard Error
0.0
0.1
0.2
0.3
0.4
0.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
Hedges's g
Figure 3. Funnel plot to explore publication bias: (A) fruit, (B) vegetables, and (C) fruit and vegetables combined. Note: The vertical line is at the mean of effect size. Not all 29 studies were included in (A), (B), and (C) because not all reported fruit, vegetables, and fruit and vegetables results, respectively.
adults,17 which examined mainly interventions based on Social Cognitive Theory and the Transtheoretical Model and also found no differences in their effectiveness. Neither the causal effi-
cacy nor strength of relationships of theoretical constructs to dietary behaviors has been well established in longitudinal or experimental research,17,90–92 especially among
Diep et al 541 children. Such research is needed to better guide efforts to target causal relationships14 and establish which theory or variables predict or work best in given situations or with children. Studies with a formal planning process had a higher effect size for F but a lower effect size for FV combined than did studies without a formal planning process. There was no relationship for V. When taking into account study quality, there were no significant relationships between formal planning process and F, V, or FV consumption. These results contradict what might be expected and suggest that formal planning processes may not be as effective as previously thought, the variables in these processes may lack causal efficacy or have low predictive strength, or interventionists may not be implementing these processes well. Study quality may be a confounder, but there were too few studies with a formal planning process to assess the relationship between formal planning process and study quality in this meta-analysis. Similarly, higher-quality studies had higher effect sizes for F and V than lower-quality studies, but higher-quality studies had a lower effect size for FV combined. These results also contradict what might be expected (ie, lower effect sizes for F, V, and FV in higher-quality studies). In addition, when assessing the effect of theory while covarying out the quality of studies, theory use had a relationship with only V consumption. This suggests that the biases, not the intervention, may have accounted for the effects of theory on F, V, and FV consumption. Future research will need to identify which biases account for these enhanced outcomes and attempts should be made to minimize their occurrence in future interventions. The meta-analysis among adults17 found randomization to be associated with greater intervention effectiveness, but more research is needed to determine whether this is also true for dietary interventions in children.
More emphasis is needed on study quality in behavioral interventions.
542 Diep et al The overall pattern of results is mixed and confusing. It is possible that behavioral theory and intervention planning have extensively diffused throughout the nutrition education community, and published results do not adequately reflect the planning and theory that were employed to design the interventions. Alternatively, more effective intervention groups may report using theory for social desirability reasons, but the use of theory had no effect on their intervention (eg, rationalizing their intervention design and implementation in light of theory). It is also possible that the effect of theory may have been overshadowed or offset by study quality. Behavioral interventionists need to establish intervention design and implementation reporting conventions (as others have done for other types of interventions and studies93–95) to provide clarity and permit comparability of studies in reviews and meta-analyses. Recent and current efforts to consistently classify behavioral intervention procedures96–98 provide an important component for consistent reporting, but these efforts need to be extended to considerations of theory and non–theory based aspects in interventions.99 The Transparent Reporting of Evaluations with Nonrandomized Designs (TREND) statement100 offers guidelines for reporting of theories used in evaluation studies that use nonrandomized designs, but not all authors provide the TREND statement checklist in their publications or have it available for readers to access. Researchers should be encouraged to use and report the TREND statement.101
Behavioral interventionists should establish and follow intervention design and reporting conventions. Methods are also needed that facilitate formalizing and reporting non–theory based considerations (ie, practical or experience-based understanding) when designing dietary change interventions. Some interventions without theory did well in this meta-analysis; thus, understanding
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 what aspects of these practical or non–theory based procedures were effective may provide insight for future strategies. It is possible that interventions not mentioning theory actually employed theory-based procedures. Some practical procedures may be mapped onto theory constructs as implementation strategies, but there was inadequate detail in these studies to determine whether these procedures were rooted in theory or apart from theory. For example, the non–theory based intervention by Govula et al74 had the second highest effect sizes for F and FV combined, and incorporated 6 nutrition education lessons on MyPyramid, food math, FV, healthy eating, and the Medicine Wheel. Some of these lessons could be linked to theory constructs (eg, setting goals to eat more FV and identifying benefits of healthy activities), but more careful reporting of procedures is needed to differentiate theory and practical procedures and determine whether they are the same or one is more effective than another, or whether they are equally important. Limitations of the current meta-analysis should be noted. First, the search strategies used by the authors may have resulted in unintentional exclusion of articles, especially those not indexed in the selected databases, not fitting the search terms used, or excluded from the literature searches/meta-analyses reviewed. Some interventions with negative results may not have been published (ie, publication bias), thus inflating effect sizes reported in the literature, but the Egger's tests indicated that there were no statistically significant risks of publication bias. The authors excluded interventions for overweight and/or obese children (because these focus on caloric intake reduction). Future reviews and meta-analyses of studies with obese children may reveal additional insights for nutrition education and interventions. Another limitation is the exclusion of non–FV targeted interventions. Obesity prevention and dietary interventions have targeted an array of other dietary behaviors from sweetened beverage consumption to energy intake. To maximize the number of studies with comparable variables across interventions, this review and metaanalysis assessed only FV-related inter-
ventions, but investigating dietary studies targeting other dietary behaviors is important for nutrition education professionals. Next, this meta-analysis did not examine SES, education, and other factors that may affect FV intake and/or weight. For example, children and adolescents from lower-SES households consume less FV and a more limited variety of them.102 Not all studies in this meta-analysis adjusted for SES and demographic factors, and of those that did, not all investigated the specific impact of these factors. To differentiate the effects of theory vs confounding factors (such as SES) on FV intake and weight status, research on dietary interventions that report effects of SES and demographics is needed. Finally, because the subject group was composed of children aged 2–18 years, the issue arises of the developmentally appropriate selection and employment of theory. This review and metaanalysis included interventions targeting FV consumption, which is a behavior partly determined by taste preferences, familiarity/exposure, and sensory attributes such as flavor, texture, smell, and visual appeal.103– 106 The interventions included in this review were based on behavioral theories emphasizing cognitive factors, not sensory-affective factors, so the theories used may not have been appropriate for different age groups and should incorporate affective, emotional factors. Developmental appropriateness was not clearly addressed in any of the studies, so it could not be assessed in this meta-analysis, but it requires closer scrutiny in future research.
IMPLICATIONS FOR RESEARCH AND PRACTICE This literature review and metaanalysis provided a systematic, indepth review of behavioral theory use in the design of child FV interventions and their effectiveness. There was little or mixed support for enhanced dietary change with use of theory, multiple theories, or a formal planning process in dietary change interventions when taking into account study quality. Research is needed on the causality and strength of relationships between
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 theoretical variables and dietary change behaviors in experiments and longitudinal research to provide guidance to future intervention design efforts. Research is also needed to identify what may be effective practical or experience-based procedures to incorporate into interventions.
ACKNOWLEDGMENTS Cassandra Diep was supported by Primary Care Research Training Grant from National Research Service Award T32 HP10031. Tzu-An Chen was supported by several sources of grant funding, none specifically related to this systematic review and metaanalysis. Vanessa Davies was supported by the CAPES–Brazilian Federal Agency for Support and Evaluation of Graduate Education. Janice and Tom Baranowski were supported in part by federal funds from the US Department of Agriculture/Agricultural Research Service under Cooperative Agreement 58-6250-6001 with the Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine.
REFERENCES 1. Boeing H, Bechthold A, Bub A, et al. Critical review: vegetables and fruit in the prevention of chronic diseases. Eur J Nutr. 2012;51:637-663. 2. Alinia S, Hels O, Tetens I. The potential association between fruit intake and body weight—a review. Obes Rev. 2009;10:639-647. 3. Ledoux T, Hingle M, Baranowski T. Relationship of fruit and vegetable intake with adiposity: a systematic review. Obes Rev. 2011;12:e143-e150. 4. Newby PK. Are dietary intakes and eating behaviors related to childhood obesity? A comprehensive review of the evidence. J Law Med Ethics. 2007;35:35-60. 5. Newby PK. Plant foods and plantbased diets: protective against childhood obesity? Am J Clin Nutr. 2009; 89:1572S-1587S. 6. Rolls BJ, Ello-Martins JA, Tohill BC. What can intervention studies tell us about the relationship between fruit and vegetable consumption and weight management? Nutr Rev. 2004;62:1-17.
7. Tohill B, Seymour J, Serdula M, Kettel-Khan L, Rolls BJ. What epidemiologic studies tell us about the relationship between fruit and vegetable consumption and body weight. Nutr Rev. 2004;64:365-374. 8. Bourke M, Whittaker PJ, Verma A. Are dietary interventions effective at increasing fruit and vegetable consumption among overweight children? A systematic review. J Epidemiol Commun Health. 2014;68: 485-490. 9. Wolfenden L, Wyse RJ, Britton BI, et al. Interventions for increasing fruit and vegetable consumption in children aged 5 years and under. Cochrane Database Syst Rev. 2012;11: CD008552. 10. Contento IR. Nutrition Education: Linking Research, Theory, and Practice. 2nd ed. Sudbury, MA: Jones and Bartlett; 2011. 11. Contento IR. Nutrition education: linking research, theory, and practice. Asia Pac J Clin Nutr. 2008;17(suppl 1):176-179. 12. Goodson P. Theory in Health Promotion Research and Practice: Thinking Outside the Box. Sudbury, MA: Jones and Bartlett; 2010. 13. Hochbaum GM, Sorenson JR, Lorig K. Theory in health education practice. Health Educ Q. 1992;19:295-313. 14. Baranowski T. Understanding the behavioral linkages needed for designing effective interventions to increase fruit and vegetable intake in diverse populations. J Am Diet Assoc. 2011;111:1472-1475. 15. Contento I, Balch GI, Bronner YL, et al. The effectiveness of nutrition education and implications for nutrition education policy, programs, and research: a review of research. J Nutr Educ. 1995;27:277-418. 16. Noar SM, Mehrotra P. Toward a new methodological paradigm for testing theories of health behavior and health behavior change. Patient Educ Couns. 2011;82:468-474. 17. Prestwich A, Sniehotta FF, Whittington C, Dombrowski SU, Rogers L, Michie S. Does theory influence the effectiveness of health behavior interventions? Meta-analysis. Health Psychol. 2014;33:465-474. 18. Noar SM. Behavioral interventions to reduce HIV-related sexual risk behavior: review and synthesis of meta-analytic evidence. AIDS Behav. 2008;12:335-353.
Diep et al 543 19. Herbst JH, Sherba RT, Crepaz N, et al. A meta-analytic review of HIV behavioral interventions for reducing sexual risk behavior of men who have sex with men. J Acquir Immune Defic Syndr. 2005;39:228-241. 20. Johnson BT, Carey MP, Marsh KL, Levin KD, Scott-Sheldon LA. Interventions to reduce sexual risk for the human immunodeficiency virus in adolescents, 1985-2000: a research synthesis. Arch Pediatr Adolesc Med. 2003; 157:381-388. 21. Johnson BT, Carey MP, Chaudoir SR, Reid AE. Sexual risk reduction for persons living with HIV: research synthesis of randomized controlled trials, 1993 to 2004. J Acquir Immune Defic Syndr. 2006;41:642-650. 22. Crepaz N, Lyles CM, Wolitski RJ, et al. Do prevention interventions reduce HIV risk behaviours among people living with HIV? A metaanalytic review of controlled trials. AIDS. 2006;20:143-157. 23. Ward DJ, Rowe B, Pattison H, Taylor RS, Radcliffe KW. Reducing the risk of sexually transmitted infections in genitourinary medicine clinic patients: a systematic review and meta-analysis of behavioural interventions. Sex Transm Infect. 2005;81: 386-393. 24. Herbst JH, Kay LS, Passin WF, et al. A systematic review and meta-analysis of behavioral interventions to reduce HIV risk behaviors of Hispanics in the United States and Puerto Rico. AIDS Behav. 2007;11:25-47. 25. Albarracin J, Albarracin D, Durantini M. Effects of HIVprevention interventions for samples with higher and lower percents of Latinos and Latin Americans: a metaanalysis of change in condom use and knowledge. AIDS Behav. 2008;12: 521-543. 26. Legler J, Meissner HI, Coyne C, Breen N, Chollette V, Rimer BK. The effectiveness of interventions to promote mammography among women with historically lower rates of screening. Cancer Epidemiol Biomarkers Prev. 2002;11:59-71. 27. Webb TL, Joseph J, Yardley L, Michie S. Using the internet to promote health behavior change: a systematic review and meta-analysis of the impact of theoretical basis, use of behavior change techniques, and mode of delivery on efficacy. J Med Internet Res. 2010;12:e4.
544 Diep et al 28. Noar SM, Benac CN, Harris MS. Does tailoring matter? Meta-analytic review of tailored print health behavior change interventions. Psychol Bull. 2007;133:673-693. 29. Baranowski T, Cullen KW, Nicklas T, Thompson D, Baranowski J. Are current health behavioral change models helpful in guiding prevention of weight gain efforts? Obes Res. 2003;11(suppl): S23-S43. 30. Sallis JF, Owen N, Fisher EB. Ecological models of health. In: Glanz K, Rimer BK, Viswanath K, eds. Health Behavior and Health Education: Theory, Research, and Practice. 4th ed. San Francisco, CA: JosseyBass; 2008:465-486. 31. Bandura A. Health promotion by social cognitive means. Health Educ Behav. 2004;31:143-164. 32. Achterberg C, Miller C. Is one theory better than another in nutrition education? A viewpoint: more is better. J Nutr Educ Behav. 2004;36: 40-42. 33. Bartholomew LK, Parcel GS, Kok G. Intervention mapping: a process for developing theory- and evidencebased health education programs. Health Educ Behav. 1998;25:545-563. 34. Riley BL, Macdonald J, Mansi O, et al. Is reporting on interventions a weak link in understanding how and why they work? A preliminary exploration using community heart health exemplars. Implement Sci. 2008;3:27. 35. Bhattacharyya O, Reeves S, Garfinkel S, Zwarenstein M. Designing theoretically-informed implementation interventions: fine in theory, but evidence of effectiveness in practice is needed. Implement Sci. 2006;1:5. 36. Branscum P, Sharma M. After-school based obesity prevention interventions: a comprehensive review of the literature. Int J Environ Res Public Health. 2012;9:1438-1457. 37. Thomson CA, Ravia J. A systematic review of behavioral interventions to promote intake of fruit and vegetables. J Am Diet Assoc. 2011;111:1523-1535. 38. AlMarzooqi MA, Nagy MC. Childhood obesity intervention programs: a systematic review. Life Sci J. 2011; 8:45-60. 39. Jepson RG, Harris FM, Platt S, et al. The effectiveness of interventions to change six health behaviours: a review of reviews. BMC Public Health. 2010; 10:538.
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 40. Zenzen W, Kridli S. Integrative review of school-based childhood obesity prevention programs. J Pediatr Health Care. 2009;23:242-258. 41. Howerton MW, Bell BS, Dodd KW, et al. School-based nutrition programs produced a moderate increase in fruit and vegetable consumption: meta and pooling analyses from 7 studies. J Nutr Educ Behav. 2007;39:186-196. 42. Cole K, Waldrop J, D’Auria J, Garner H. An integrative research review: Effective school-based childhood overweight interventions. J Spec Pediatr Nurs. 2006;11:166-177. 43. Williams AJ, Henley WE, Williams CA, Hurst AJ, Logan S, Wyatt KM. Systematic review and meta-analysis of the association between childhood overweight and obesity and primary school diet and physical activity policies. Int J Behav Nutr Phys Act. 2013;10:101. 44. Stice E, Shaw H, Marti CN. A metaanalytic review of obesity prevention programs for children and adolescents: the skinny on interventions that work. Psychol Bull. 2006;132:667-691. 45. Evans CE, Christian MS, Cleghorn CL, Greenwood DC, Cade JE. Systematic review and meta-analysis of school-based interventions to improve daily fruit and vegetable intake in children aged 5 to 12 y. Am J Clin Nutr. 2012;96:889-901. 46. Knai C, Pomerleau J, Lock K, McKee M. Getting children to eat more fruit and vegetables: a systematic review. Prev Med. 2006;42:85-95. 47. Kremers SP, de Bruijn GJ, Droomers M, van Lenthe F, Brug J. Moderators of environmental intervention effects on diet and activity in youth. Am J Prev Med. 2007;32: 163-172. 48. Tedstone A, Aviles M, Shetty P, et al. Effectiveness of Interventions to Promote Healthy Eating in Preschool Children Aged 1 to 5 Years: A Review. London, UK: Health Education Authority; 1998. 49. Thomas J, Sutcliffe K, Harden A, et al. Children and Healthy Eating: A Systematic Review of Barriers and Facilitators. London, UK: EPPI-Centre; 2003. Social Science Research Unit, Institute of Education, University of London. 50. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and metaanalyses of studies that evaluate health care interventions: explanation and
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
elaboration. J Clin Epidemiol. 2009;62: e1-e34. Higgins JPT, Altman DG. Assessing risk of bias in included studies. In: Higgins JPT, Green S, eds. Cochrane Handbook for Systematic Reviews of Interventions. West Sussex, UK: John Wiley & Sons; 2008:187-242. Juni P, Altman DG, Egger M. Systematic reviews in health care: assessing the quality of controlled clinical trials. BMJ. 2001;323:42-46. Hedges LV, Olkin I. Statistical Methods for Meta-analysis. Orlando, FL: Academic Press; 1985. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med. 2002;21:1539-1558. Shadish WR, Haddock CK. Combining estimates of effect size. In: Cooper H, Hedges LV, eds. The Handbook of Research Synthesis. New York, NY: Russell Sage Foundation; 1994:261-281. Lipsey MW, Wilson DB. Applied Social Research Methods. Thousand Oaks, CA: Sage; 2001. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses. BMJ. 2003;327:557-560. Egger M, Davey Smith G, Schneider M, et al. Bias in metaanalysis detected by a simple, graphical test. BMJ. 1997;315:629-634. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7:177-188. Anderson AS, Porteous LE, Foster E, et al. The impact of a school-based nutrition education intervention on dietary intake and cognitive and attitudinal variables relating to fruits and vegetables. Public Health Nutr. 2005;8: 650-656. Angelopoulos PD, Milionis HJ, Grammatikaki E, Moschonis G, Manios Y. Changes in BMI and blood pressure after a school based intervention: The CHILDREN study. Eur J Public Health. 2009;19:319-325. Baranowski T, Davis M, Resnicow K, et al. Gimme 5 fruit, juice and vegetables for fun and health: outcome evaluation. Health Educ Behav. 2000;27: 96-111. Birnbaum AS, Lytle LA, Story M, Perry CL, Murray DM. Are differences in exposure to a multicomponent school-based intervention associated with varying dietary outcomes in adolescents? Health Educ Behav. 2002;29:427-443.
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014 64. Black MM, Hager ER, Le K, et al. Challenge! Health promotion/obesity prevention mentorship model among urban, black adolescents. Pediatrics. 2010;126:280-288. 65. Branscum P, Sharma M, Wang LL, Wilson BR, Rojas-Guyler L. A true challenge for any superhero: An evaluation of a comic book obesity prevention program. Fam Commun Health. 2013;36:63-76. 66. Chen JL, Weiss S, Heyman MB, Cooper B, Lustiq RH. The efficacy of the web-based childhood obesity prevention program in Chinese American adolescents (Web ABC study). J Adolesc Health. 2011;49:148-154. 67. Cullen KW, Bartholomew LK, Parcel GS. Girl Scouting: an effective channel for nutrition education. J Nutr Educ. 1997;29:86-91. 68. Davis JN, Ventura EE, Cook LT, Gyllenhammer LE, Gatto NM. LA Sprouts: a gardening, nutrition, and cooking intervention for Latino youth improves diet and reduces obesity. J Am Diet Assoc. 2011;111: 1224-1230. 69. Day ME, Strange KS, McKay HA, Naylor PJ. Action schools! BC– Healthy Eating: effects of a wholeschool model to modifying eating behaviours of elementary school children. Can J Public Health. 2008;99: 328-331. 70. Domel S, Baranowski T, Davis H, et al. Development and evaluation of a school intervention to increase fruit and vegetable consumption among 4th and 5th grade students. J Nutr Educ. 1993;25:345-349. 71. Dzewaltowski DA, Estabrooks PA, Welk G, et al. Healthy Youth Places: a randomized controlled trial to determine the effectiveness of facilitating adult and youth leaders to promote physical activity and fruit and vegetable consumption in middle schools. Health Educ Behav. 2009;36:583-600. 72. Evans AE, Dave J, Tanner A, et al. Changing the home nutrition environment: effects of a nutrition and media literacy pilot intervention. Fam Commun Health. 2006;29:43-54. 73. Gentile DA, Welk G, Eisenmann JC, et al. Evaluation of a multiple ecological level child obesity prevention program: switch what you Do, View, and Chew. BMC Med. 2009;7:49. 74. Govula C, Kattelman K, Ren C. Culturally appropriate nutrition lessons increased fruit and vegetable con-
75.
76.
77.
78.
79.
80.
81.
82.
83.
sumption in American Indian children. Top Clin Nutr. 2007;22: 239-245. Gribble LS, Falciglia G, Davis AM, Couch SC. A curriculum based on social learning theory emphasizing fruit exposure and positive parent childfeeding strategies: a pilot study. J Am Diet Assoc. 2003;103:100-103. Haerens L, Deforche B, Maes L, Cardon G, Stevens V, De Bourdeaudhuij I. Evaluation of a 2year physical activity and healthy eating intervention in middle school children. Health Educ Res. 2006;21: 911-921. Johnson BA, Kremer PJ, Swinburn BA, de SilvaSanigorski AM. Multilevel analysis of the Be Active Eat Well intervention: Environmental and behavioural influences on reductions in child obesity risk. Int J Obes (Lond). 2012;36: 901-907. Mangunkusumo RT, Brug J, de Koning HJ, van der Lei J, Raat H. School-based internettailored fruit and vegetable education combined with brief counselling increases children’s awareness of intake levels. Public Health Nutr. 2007;10:273-279. Parcel GS, Simons-Morton B, O’Hara NM, Baranowski T, Wilson B. School promotion of healthful diet and physical activity: Impact on learning outcomes and self-reported behavior. Health Educ Q. 1989;16:181-199. Ransley JK, Greenwood DC, Cade JE, et al. Does the school fruit and vegetable scheme improve children’s diet? A non-randomised controlled trial. J Epidemiol Commun Health. 2007;61: 699-703. Reinaerts E, de Nooijer J, Candel M, et al. Increasing children’s fruit and vegetable consumption: distribution or a multicomponent programme? Public Health Nutr. 2007;10:939-947. Reinaerts E, Crutzen R, Candel M, De Vries NK, De Nooijer J. Increasing fruit and vegetable intake among children: comparing long-term effects of a free distribution and a multicomponent program. Health Educ Res. 2008; 23:987-996. Reynolds KD, Bishop DB, Chou CP, Xie B, Nebeling L, Perry CL. Contrasting mediating variables in two 5-a-day nutrition intervention programs. Prev Med. 2004;39:882-893.
Diep et al 545 84. Rosenkranz RR, Behrens TK, Dzewaltowski DA. A grouprandomized controlled trial for health promotion in Girl Scouts: healthier troops in a SNAP (Scouting Nutrition & Activity Program). BMC Public Health. 2010;10:81. 85. Smolak L, Levine MP, Schermer F. A controlled evaluation of an elementary school primary prevention program for eating problems. J Psychosom Res. 1998;44:339-353. 86. Tak NI, Te Velde SJ, Brug J. Ethnic differences in 1-year follow-up effect of the Dutch Schoolgruiten Project— promoting fruit and vegetable consumption among primary-school children. Public Health Nutr. 2007;10: 1497-1507. 87. Thompson D, Baranowski T, Baranowski J, et al. Boy Scout 5-aDay Badge: outcome results of a troop and Internet intervention. Prev Med. 2009;49:518-526. 88. Wilson DK, Friend R, Teasley N, Green S, Reaves IL, Sica DA. Motivational versus social cognitive interventions for promoting fruit and vegetable intake and physical activity in African American adolescents. Ann Behav Med. 2002;24:310-319. 89. Dombrowski SU, Sniehotta FF, Avenell A, et al. Current issues and future directions in Psychology and Health: towards a cumulative science of behaviour change: do current conduct and reporting of behavioural interventions fall short of best practice? Psychol Health. 2007;22: 869-874. 90. Cerin E, Barnett A, Baranowski T. Testing theories of dietary behavior change in youth using the mediating variable model with intervention programs. J Nutr Educ Behav. 2009;41: 309-318. 91. Guillaumie L, Godin G, VezinaIm LA. Psychosocial determinants of fruit and vegetable intake in adult population: a systematic review. Int J Behav Nutr Phys Act. 2010;7:12. 92. Shaikh AR, Yaroch AL, Nebeling L, Yeh MC, Resnicow K. Psychosocial predictors of fruit and vegetable consumption in adults: a review of the literature. Am J Prev Med. 2008;34: 535-543. 93. Chauhan SP, Blackwell SC, Saade GR. A suggested approach for implementing CONSORT guidelines specific to obstetric research. Obstet Gynecol. 2013;122:952-956.
546 Diep et al 94. Samaan Z, Mbuagbaw L, Kosa D, et al. A systematic scoping review of adherence to reporting guidelines in health care literature. J Multidiscip Healthc. 2013;6:169-188. 95. Sorensen AA, Wojahn RD, Manske MC, Calfee RP. Using the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement to assess reporting of observational trials in hand surgery. J Hand Surg Am. 2013; 38:1584-1589.e2. 96. Michie S, Richardson M, Johnston M, et al. The Behavior Change Technique Taxonomy (v1) of 93 hierarchically clustered techniques: building an international consensus for the reporting of behavior change interventions. Ann Behav Med. 2013;46:81-95. 97. Montgomery P, Grant S, Hopewell S, et al. Protocol for CONSORT-SPI:
Journal of Nutrition Education and Behavior Volume 46, Number 6, 2014
98.
99.
100.
101.
an extension for social and psychological interventions. Implement Sci. 2013;8:99. Woodside JV, Koletzko BV, Patterson CC, Welch RW. Scientific standards for human intervention trials evaluating health benefits of foods, and their application to infants, children and adolescents. World Rev Nutr Diet. 2013;108:18-31. Michie S, West R, Spring B. Moving from theory to practice and back in social and health psychology. Health Psychol. 2013;32:581-585. Des Jarlais DC, Lyles C, Crepaz N. TREND Group. Improving the reporting quality of nonrandomized evaluations of behavioral and public health interventions: the TREND statement. Am J Public Health. 2004; 94:361-366. Treasure E. The TREND statement. Evid Based Dent. 2004;5:88-91.
102. Darmon N, Drewnowski A. Does social class predict diet quality? Am J Clin Nutr. 2008;87:1107-1117. 103. Birch LL. Development of food preferences. Annu Rev Nutr. 1999;19:41-62. 104. Rasmussen M, Krolner R, Klepp KI, et al. Determinants of fruit and vegetable consumption among children and adolescents: a review of the literature. Part I: quantitative studies. Int J Behav Nutr Phys Act. 2006;3:1-19. 105. Krolner R, Rasmussen M, Brug J, Klepp KI, Wind M, Due P. Determinants of fruit and vegetable consumption among children and adolescents: a review of the literature. Part II: qualitative studies. Int J Behav Nutr Phys Act. 2011;8:112. 106. Ventura AK, Worobey J. Early influences on the development of food preferences. Curr Biol. 2013;23: R401-R408.
