Study Type
The study design will include various types of randomized controlled trials (RCTs) such as clustered and crossover RCTs. This protocol is based on International Prospective Registered in the Register of Systematic Reviews (PROSPERO) (No. † ××××).
Participants
We will include adults and children of all ages from any worksite, community, household, school and university.
Intervention
We plan to include food-based interventions that focus on social marketing interventions including incentive strategies (e.g., discount coupon, point systems) implemented in worksite, school or university cafeterias, supermarkets, small shops, vending machines, or greengrocer etc.
Comparator(s)/control
Any other intervention of not incentive and social marketing, or no intervention
Outcome measures
Measuring method:
・Food and nutrient consumption (or intake) will be measured by questionnaires and purchasing data
・Obesity and blood pressure values are the actual measurement values of body measurements and self-reported values
・Cholesterol, triglycerides, haemoglobin A1c are measured values of blood tests
Primary outcomes (Continuous variables)
・Changes in fruits and vegetables/fruit/vegetables consumption (or intake) [g or Serving; SV]
・Changes in healthy food/healthy lunch/ healthy drinks consumption (or intake) [g or score]
・Changes in snacks or sweets or sugary beverages consumption (or intake) [g or score]
・Weight changes (including body mass index (BMI) [kg/m2], body fat[%])
Secondary outcomes
・Changes in blood pressure [mmHg]
・Changes in cholesterol (LDL cholesterol, HDL cholesterol, total cholesterol) [g]
・Changes in triglycerides [mg/dl]
・Changes in haemoglobin A1c [%]
・Changes in nutritional intake
・changes in fat and oil intake [g]
・changes in energy intake [kcal]
・Adverse effects that are defined by the authors
Electronic searches
We will search the following the electronic databases from inception to the present date.
・CENTRAL; monthly searches of the Cochrane Central Register of Controlled Trials
・MEDLINE; weekly searches of MEDLINE
・EMBASE; weekly searches of EMBASE
・CINAHL; weekly searches of CINAHL
・PsycINFO ; weekly searches of PsycINFO
Appendix 1 shows the keywords in searches. An experienced librarian at the National Center for Child Health and Development will implement these strategies according to the Cochrane methods and guidance.
Searching other resources
We will also utilize Google Scholar for related studies, and we will hand-search systematic reviews and reference lists included in these studies. Additionally, we will review the titles and summaries of these studies.
Data collection and analysis
Inclusion criteria
・Study design: RCTs (including cluster and cross-over RCTs).
・Participants: children and adults of any age.
・Intervention site: worksites, community sites (e.g. supermarket, small shop), and schools.
・Intervention: organization-based, incentive pricing strategies, or social marketing. We will include co-intervention.
Exclusion criteria
・Excluded designs: quasi-experimental design and pre-post and observational studies.
・Excluded participants: mental illness
・Excluded intervention programs:
・Fitness- and exercise-focused interventions
・Programs that do not use behavioural science theoretical approaches incorporating incentive pricing strategies for food and social marketing
・Taxation or subsidy studies
・Alcoholic beverage and drug studies
・Excluded publications:
・Non-academic journals and reports
・The abstracts of conferences
・Other than English
Data extraction
Review authors (KS and Y Y) will separately read titles, abstracts and extract studies. They will then independently decide whether to include or exclude the studies and reach a consensus on each other's extracted data. In case of a discrepancy, KS and YY will consult with other authors to come to a resolution. If there is unclear information at the time of data extraction, the preferred course of action will be to seek more information by contacting the author of the original paper.
Assessment of risk of bias in included studies
Assessment of risk of bias will be conducted through random sequence generation, concealment of allocation sequences, reporting of results and other bias criteria. [16]
In addition, KS and YY will independently assess each domain as high risk, low risk, or uncertain risk of bias and utilize the Cochrane Systematic Review Methodology Tool2 to determine whether to include the study in the meta-analysis.
If more than one study report has the same outcome and is sufficiently homogeneous conceptually, methodologically, and statistically, we will perform meta-analysis of these studies. If there are any disagreements between the two authors, the differences will be discussed and resolved, and if no agreement is reached, the opinion of the third or fourth author (KL, NW) will be sought. In case of any further conflict, a decision will be made in consultation with all authors.
Measures of treatment effect
Continuous data
Method of measuring intervention effect
For continuous data, we will use the Mean Difference (MD) and 95% Confidence Interval (CI) if the same method is used to measure results between trials. If the trials do not involve measurement of results based on the same method, we will use standardized mean difference (SMD) for the analysis. For dichotomous data, we will report a summary of results using risk rate (RR) and 95% CI.
Methods of unit of analysis
When we use similar intra-cluster correlation coefficient (ICC) estimates, we will check the sample size of each test, and when we utilize other sources’ ICCs, we will perform a sensitivity analysis to determine the ICC variability and analyse the effect. If both cluster RCTs and individual cluster RCTs are found, the required information will be included. We use both individual and cluster results when minimal heterogeneity exists between study designs and the effect of the intervention and the interaction between the randomization units are considered distant.
Dealing with missing data
In the case of missing data, we will contact the authors involved in the research of the main article and request the missing data or information. All subjects will be analysed in the already assigned group rather than based on whether they actually received the planned intervention program.
We will check for missing participant values in included trials and consider whether the results include trials with a substantial amount of missing data. Intention-to-treat (ITT) analysis is used whenever possible for all outcomes
Assessment of heterogeneity
We will determine heterogeneity in the meta-analysis based on if we statistically confirm I2 (≧30% ), T2 (>0 ), or χ2 (<0.1).
Assessment of reporting bias
If we are able to confirm more than 10 trials, we will use the funnel plot tool to evaluate publication bias. Plot effect size values and study accuracy will be utilized to assess bias.
Data synthesis
If two or more outcomes are the same and no heterogeneity is observed, we will use a fixed effects model. If there is more than one instance of the same outcome and heterogeneity is observed, we will use a random effects model if it is considered to be meaningful for the mean program effect. Further, if we use random- and meta-mixing effects, this will be displayed as the mean conditioning effect that is provided as the evaluation of I2 and T2 with a 95% CI.
We will use Review Manager V.5.3 (Cochrane Collaboration software) for statistical analysis when integrating data.
If quantitative synthesis is not appropriate, we will describe the summary using the evidence of table.
Subgroup analysis and investigation of heterogeneity
We will perform a subgroup analysis on the following items and investigate heterogeneity using sensitivity analysis:
・Type of intervention: social marketing or incentive strategies intervention only versus social marketing or incentive strategies intervention including other intervention program versus control
・Type of incentive: pricing versus not pricing (coupons or points) / all pricing (discount, coupons and points) versus free offer
・Location of intervention: worksite versus community (including home) versus school (including university)
・Intervention impact (discount rate): <20% / ≧20%~<30% / ≧30%
・Participants including obese participants or participants with other physical illness versus normal participants
・Participants: low-income versus others
Sensitivity analysis
We will perform a sensitivity analysis of the primary outcomes, excluding studies with a high risk of bias as the risk of bias can affect the meta-analysis. Therefore, randomization concealment and allocation and data with incomplete results will be judged to be high risk.
Recommendations
We will use the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) [17] to determine the quality of evidence and recommendations for this systematic review. The GRADE does not assess individual evidence but rather the quality of body of evidence that integrates multiple studies according to outcome. The GRADE approach includes five factors: risk of bias, inconsistency of results, indirectness of evidence, imprecision, and publication bias. We will GRADE the main outcomes based on the quality grade of the evidence (High, Moderate, Low, Very Low) and create a summary of findings using tables.