Effectiveness of provision of animal‐source foods for supporting optimal growth and development in children 6 to 59 months of age

Abstract

This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:

To assess the effectiveness of animal‐source foods compared to other feeding interventions or no intervention in improving growth and developmental outcomes in children aged 6 to 59 months.

Background

Exclusive breastfeeding is recommended during the first six months of life followed by continued breastfeeding with appropriate complementary foods for up to two years or beyond (Kramer 2002; WHO 2003). Complementary foods provide calories and nutrients beyond what is provided in breast milk (PAHO/WHO 2003). Adequate nutrients early in life promote cognitive development and are critical for proper growth and functioning. Growth faltering is seen across global contexts and usually occurs between the ages of three months and two years of age (Victora 2010). Most diets, particularly in low‐ and middle‐income countries (LMIC), are nutritionally poor, based on staple foods like rice, wheat, maize (corn), millet, sorghum, roots and tubers (FAO 1995). Animal products, such as eggs, meat, fish and dairy, are energy dense and contain multiple micronutrients (particularly iron, zinc, vitamin A, vitamin B12, and choline) and essential fatty acids in a highly bioavailable form (Leroy 2007). Their consumption is associated with improved growth and developmental outcomes in observational studies but, as a result of availability, access or sociocultural norms, may not be practical for the lowest‐income consumers (Leroy 2007).

The World Health Organization (WHO) Global Strategy on Diet, Physical Activity and Health, endorsed by the 57th World Health Assembly recognizes the need to draft, update and implement national food‐based dietary and physical activity guidelines (WHO 2004). The Brazilian Dietary Guidelines of 2014 (Brazilian MoH 2015), for example, emphasize the importance of understanding nutrition in terms of food and meals rather than individual nutrients. As countries develop economically, animal‐source foods, vegetable oils, and sugars begin to replace a larger portion of calories (Popkin 2001). In high‐income contexts, meat consumption is associated with obesity and its sequelae in adults, but not children (Loring Bradlee 2010; Wang 2009). For this reason, it is important to understand the impact of animal‐source food consumption on growth and development outcomes in children across global contexts.

Description of the condition

Malnutrition in children encompasses both undernutrition and overweight and obesity. Undernutrition includes stunting (low height‐for‐age), wasting (low weight‐for‐height), and micronutrient deficiencies. In 2011, undernutrition contributed to 45% of all child deaths (Black 2013). Stunting in children affects 156 million; a further 50 million children are wasted and 42 million are overweight (WHO 2016). Of the major micronutrient deficiencies, vitamin A, zinc, iron, and iodine are responsible for the largest proportion of years of life lost (YLLs) and disability‐adjusted life years (DALYs) (Black 2008). Deficiencies of vitamin A and zinc result in increases in all‐cause morbidity and mortality; deficiencies in iron and iodine, in addition to omega‐3 fatty acids, impair children’s ability to reach their development potential (Nyaradi 2013).

In 2015, global estimates found that 42 million children under five years of age, or 6.2%, were classified as overweight (weight‐for‐height score greater than 2 z‐scores above the median WHO standard) (WHO 2006; WHO 2016). Overweight in children under five years of age may result in type 2 diabetes and high blood pressure, and is a risk for adult obesity and its sequelae. Although stunting is less prevalent among overweight or obese children, deficiencies in micronutrients and essential fatty acids ‐ 'hidden hunger' ‐ may persist, with negative impacts on neurocognitive development (Black 2013).

Evidence for point‐of‐use multiple micronutrient powder supplementation or supplementary feeding interventions on growth and development outcomes are unclear. A Cochrane Review of eight trials found that a micronutrient powder containing at least iron, zinc and vitamin A, provided for home fortification, was associated with reduced risk of anaemia and iron deficiency in children under two years of age, but had no impact on growth (De‐Regil 2011). A Cochrane Review of community‐based supplementary feeding for promoting growth of children under five years of age in LMICs found a small but statistically significant effect on length in children less than 12 months old but, due to the variance in outcomes between studies, no firm conclusions were reached (Sguassero 2012).

Strategically developed and implemented food‐based strategies that take into account relevant ecological, cultural and socioeconomic factors could be sustainable and acceptable forms of intervention (FAO/WHO 1998). Animal‐source foods in particular are energy dense and contain multiple micronutrients (particularly iron, zinc, vitamin A, vitamin B12, and choline) and essential fatty acids in a highly bioavailable form (Leroy 2007). To date, the impact of animal‐source foods on growth and development outcomes in children under five years of age has not been systematically reviewed.

Description of the intervention

The effect of individual nutrients consumed through food is often not the same as consuming the same nutrients in supplementary form. This is due to 'food synergy', the biological and chemical interrelations that occur between nutrients when consumed in foods rather than in supplement form (Jacobs 2009). When consumed in food form, nutrients may work in concert with each other to improve absorption and likely have a different impact than their technologically‐produced counterparts.

This review will incorporate interventions that include provision of animal‐source foods or foods containing an animal‐source food component. Animal‐source foods include eggs, meat, fish and dairy, prepared with any cooking method. Foods containing animal‐source components will be considered if they account for 75% of the energy density in the food provided. We will only consider interventions where the food is actually given to infants and children or their caretakers, and not interventions where only counselling promoting consumption of animal‐source foods is given. Animal‐source foods may be provided via the intervention, produced within the home (through milking, egg collection, or animal slaughter), or purchased locally, either from local markets or supermarket‐style stores.

To examine the role of processing of animal‐source foods, we will conduct subgroup analyses of interventions, separating unprocessed, minimally‐processed, and processed foods from ultra‐processed foods. Drawing on the Brazilian Dietary Guidelines for 2014 (Brazilian MoH 2015), we define degree of processing in the following ways:

1. unprocessed: foods that have undergone no alteration (except cooking) following their removal from nature;

2. minimally‐processed: foods that have been submitted to cleaning, fractioning, grinding, fermentation, pasturization, or other processes intended to preserve shelf life or facilitate consumption, but which subtract minimally from the original food, such as long‐life and powdered milk, dried eggs, and whole wheat flours;

3. processed: foods that are easily recognizable as a modified version of the original food, typically a minimally‐processed or natural food with the addition of salt or sugar, such as breads; meat that is salted, cured, or smoked; and foods canned with preservatives such as salt and oil; and

4. ultra‐processed: foods that are industrial products that may incorporate natural or minimally‐processed foods with the addition of modified food constituents (such as hydrogenated oils or modified starch), in addition to flavor enhancers, colors, and other additives. These foods are not recognizable from their original products and usually come in some form of packaging.

There is also growing concern, particularly in high‐income countries (HIC) of allergies associated with some animal‐source foods, particularly eggs and shellfish, although there is currently no evidence to suggest that restrictive diets after six months of age have an allergy‐preventing effect (PAHO/WHO 2003). Exposure to livestock‐borne pathogens in areas of high human‐to‐animal contact are also a concern (Headey 2016). We will include adverse effects, such as allergies and zoonotic illness associated with livestock proximity, in our outcome measures.

How the intervention might work

Animal‐source foods, including milk and meat, have been shown to improve both growth and cognitive outcomes in intervention trials across a range of international contexts, predominantly in school‐aged children (Dror 2011). The role that animal‐source foods play during the complementary feeding window, however, is less well researched.

Animal‐source foods are energy dense and rich sources of protein and fatty acids, vitamins, and minerals. Milk, for example, is intended to support the growth and development of nursing mammals, and thus may have a positive impact on linear growth (Dror 2011). This may be due to energy or protein content, a combination of micronutrients, or other factors present in milk. Eggs are considered a perfect protein source, and a good source of essential fatty acids, choline, vitamins A and B12, and selenium (Iannotti 2014).

Importantly, animal‐source foods have the benefit of food synergy (Jacobs 2009). The vitamins and minerals found in animal‐source foods are more highly bioavailable than when consumed in plant‐based foods, particularly when consumed in concert with other ingredients. For example, animal‐source foods are typically good sources of fat, critical to absorption of fat‐soluble vitamins like vitamin A. Moreover, consuming critical nutrients in naturally found forms minimizes risk of excess consumption. In addition, although fortified staple foods may be cheaper than animal‐source foods, they are often consumed in conjunction with antinutrients, which may inhibit absorption. In particular, phytic acid found in fortified staples like wheat and corn, already high in diets in LMIC, binds to nutrients such as zinc and calcium, inhibiting absorption (Michaelsen 1998).

Processed foods, specifically fortified products, have the advantage of being able to address site‐specific nutrient deficiencies and can include many of the key limiting nutrients found in commonly consumed complementary foods such as staple grains. They may also present a lower risk for food contamination. However, there are also numerous disadvantages. The impact of the level of food processing in children is not yet well‐studied. A 2015 study from Brazil showed that consumption of ultra‐processed products is associated with increase in total cholesterol and low density lipoprotein cholesterol from pre‐school to school age (Rauber 2015). Most epidemiological studies have not taken level of food processing into account (Fardet 2015). Particularly in rural areas, access to processed foods also requires an external supply chain and source of funding that locally‐raised animal‐source foods do not.

Although the benefits of animal‐source foods for children in LMIC have been reported, the role that animal‐source foods play in the development of overweight and obesity in older children has not been well studied. Animal‐source foods are energy‐dense foods, which have been implicated in the development of obesity across contexts. However, unlike processed foods, animal‐source foods provide a wide range of nutrients and may also promote feelings of satiety, which can help prevent obesity (Jacobs 2009; Speakman 2013). Separating the role that animal‐source foods play in proper growth and cognitive development versus non‐communicable, diet‐related disease is critical in moving nutritional policy and programming forward.

Why it is important to do this review

To date, the impact of animal‐source foods on growth and development in infants and children has not been systematically reviewed. Dror and Allen conducted a narrative review in 2011, which included both observational studies and interventions (Dror 2011). That review found evidence for animal‐source foods in improving child growth and cognition, but did not apply meta‐analysis and was less strict in study eligibility. Previous systematic reviews of complementary feeding have included studies of animal‐source foods (Dewey 2008), but none have conducted an exclusive analysis. Additional studies may have been published in the years after the Dror 2011 review, which this study will include.

A growing body of research has examined the impact of increasing the intake of energy, protein, vitamins and minerals through fortified infant and child foods, oral micronutrient supplements, or lipid‐based nutrient supplements on growth and development in the case of moderate or severe malnutrition. While these interventions provide key nutrients, they usually rely on external suppliers, may be highly processed, and contain other ingredients that may be detrimental in the diet if consumed in excess, such as sugar (Popkin 2014). In addition, many interventions incorporate an animal‐based ingredient in a processed form, such as skimmed milk powder. Understanding if differences exist between natural and minimally‐processed ingredients compared to processed and ultra‐processed products is also of importance

Barriers related to local availability, affordability and accessibility, in addition to cultural preferences against animal‐source feeding in some contexts, have meant that, to date, animal‐source‐food‐based approaches to nutrition have received little research and programming attention (Demment 2003). However, as animal‐source food consumption increases worldwide, as the result of Westernization of diets and rising incomes, it is likely that animal‐source foods will grow increasingly more accessible and accepted across country contexts (Pingali 2007; Popkin 2014). This review will help inform future policy and programming related to animal‐source foods.

Objectives

To assess the effectiveness of animal‐source foods compared to other feeding interventions or no intervention in improving growth and developmental outcomes in children aged 6 to 59 months.

Methods

Criteria for considering studies for this review

Types of studies

Randomized control trials (RCTs), both individually‐ and cluster‐randomized, as well as quasi‐RCTs.

Types of participants

Infants and children of any sex, aged between 6 and 59 months, independently of their breastfeeding history, living in any country, and not more than three standard deviations (SD) above or below the WHO growth standards for length/height‐for‐age, weight‐for‐age, and weight‐for‐length/height.

We will exclude interventions for severe malnutrition (children below three SD of WHO growth standards for weight‐for‐length/height) and obesity (children above three SD of WHO growth standards for weight‐for‐length/height) (WHO 2006).

Types of interventions

Provision of animal‐source foods or foods containing an animal‐source food component.

Animal‐source foods include eggs, meat, fish and dairy, prepared with any cooking method. We will consider foods containing animal‐source components if they account for 75% of the energy density in the food provided. We will only consider interventions where the food is actually provided to infants and children or their caretakers, and not interventions where counselling or nutrition education promoting consumption of animal‐source foods is provided.

We will consider all durations of interventions. Where possible, we will pool results and conduct subgroup analyses at the following time points: less than 6 months; 6 to 12 months; greater than 12 months.

Types of outcome measures

Primary and secondary outcomes will be assessed at 6 and 12 months' duration of intervention.

Primary outcomes

1. Linear growth (as measured by height‐for‐age z‐scores or length‐for‐age z‐scores)

2. Weight gain (as measured by weight‐for‐age z‐scores)

3. All‐cause morbidity (number of participants with at least one episode of any disease during the trial)

Secondary outcomes

1. Anaemia (defined as haemoglobin lower than 110 g/L for children aged 6 to 59 months, adjusted by altitude where appropriate)

2. Iron deficiency (as measured by serum/plasma ferritin below WHO cut‐off adjusted for inflammation of 12 for both boys and girls under five years of age)

3. Developmental outcomes (motor skills (e.g. Movement Assessment of Infants (Chandler 1980) or Peabody Developmental Gross Motor Scale (Folio 1983)), visual and cognitive ability (e.g. Forced Preferential Looking), and others as assessed by trialists)

4. Allergic reaction (e.g. rash, angioedema, diarrhea)

Search methods for identification of studies

Electronic searches

We will search Ovid MEDLINE using the search strategy in Appendix 1, which uses the sensitivity and precision maximising version of the Cochrane Highly Sensitive Search Strategy for identifying RCTs in MEDLINE (Lefebvre 2011). We will modify this strategy, as appropriate, to search the other databases and trials registers listed below. We will not restrict the search by date, publication status or language, and will seek translations as necessary.

International databases and trial registers

1. Cochrane Central Register of Controlled Trials (CENTRAL; current issue) in the Cochrane Library, which includes the Cochrane Developmental, Psychosocial and Learning Problems Specialized Register

2. MEDLINE Ovid (1946 onwards)

3. MEDLINE In Process and Other Non‐Indexed Citations Ovid (current issue)

4. MEDLINE Epub Ahead of Print Ovid (current issue)

5. Embase Ovid (1974 onwards)

6. CINAHL EBSCOhost (Cumulative Index to Nursing and Allied Health Literature; 1981 onwards)

7. Science Citation Index Web of Science (SCI; 1980 onwards)

8. Social Science Citation Index Web of Science (SSCI; 1980 onwards)

9. Conference Proceedings Citation Index ‐ Science Web of Science (CPCI‐S; 1990 onwards)

10. Conference Proceedings Citation Index ‐ Social Science & Humanities Web of Science (CPCI‐SS&H; 1990 onwards)

11. Cochrane Database of Systematic Reviews (CDSR; current issue) part of the Cochrane Library

12. Epistemonikos (www.epistemonikos.org/en/advanced_search)

13. POPLINE (www.popline.org; 1970 onwards)

14. ClinicalTrials.gov (clinicaltrials.gov)

15. WHO International Clinical Trials Registry Platform (ICTRP; apps.who.int/trialsearch)

16. UK Clinical Trials Gateway (www.ukctg.nihr.ac.uk)

Regional databases

1. IBECS (ibecs.isciii.es)

2. SciELO (Scientific Electronic Library Online; www.scielo.br)

3. LILACS (Latin American and Caribbean Health Sciences Literature; lilacs.bvsalud.org/en)

4. PAHO (Pan American Health Library; www1.paho.org/english/DD/IKM/LI/library.htm)

5. WHOLIS (WHO Library; dosei.who.int)

6. WPRO (Western Pacific Region Index Medicus; www.wprim.org)

7. Index Medicus for the South‐East Asia Region (IMSEAR; imsear.hellis.org)

8. IndMED (Indian medical journals; indmed.nic.in; 1985 onwards)

9. Native Health Research Database (hscssl.unm.edu/nhd/)

Searching other resources

We will contact authors and known experts for assistance in identifying ongoing or unpublished data. We will also contact the Department of Nutrition for Health and Development, and regional offices of the WHO, Centers for Disease Control and Prevention (CDC), the United Nations Children's Fund (UNICEF), the World Food Programme (WFP), Nutrition International (formerly the Micronutrient Initiative (MI)), Helen Keller International (HKI), Home Fortification Technical Advisory Group (HFTAG), and the Global Alliance for Improved Nutrition (GAIN).

We will handsearch all included studies for references to other trials that may not have been captured by the Electronic searches.

Data collection and analysis

Selection of studies

Two review authors (JE and PR‐P) will independently assess all titles and abstracts retrieved by the search strategy for eligibility. When a title or abstract cannot be rejected with certainty, we will obtain the full text of the report for further review. We will resolve any disagreement through discussion or, if required, we will consult a third review author (PR‐S). We will place no restrictions on publication date. We will record our decisions in a PRISMA diagram (Moher 2009).

Data extraction and management

JE will design a data extraction form on which to record the following information: location of intervention, method of random allocation to treatment and control groups, details about participants, description and length of the intervention (including nutritional characteristics of the food provided), description of cointerventions, data on outcomes related to child growth and development, and rates of withdrawals. JE and PR‐P will independently extract all data using the form. Any discrepancies will be resolved through discussion or, if necessary, in consultation with PR‐S.

Assessment of risk of bias in included studies

Using Cochrane's tool for assessing risk of bias (Higgins 2011a), two review authors (JE and PR‐P) will independently assess and rate the risk of bias in each included study as low, high or unclear (uncertain), across the following seven domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other sources of potential bias. They will resolve any disagreement by discussion or, if necessary, by involving a third assessor (PR‐S). Review authors will not be blinded to the study authors, institution or journal. The criteria for assigning 'Risk of bias' judgments are detailed in Appendix 2.

Measures of treatment effect

Dichotomous data

We will present results as odds ratios (OR) with 95% confidence intervals (CI) (Deeks 2011, Section 9.2.2).

Continuous data

We will analyze continuous data if means and SD are available and there is no clear evidence of skew in the distribution (i.e. the data approximate a normal distribution). For outcomes measured in the same way between trials, we will use the mean difference (MD) with 95% CI. We will use the standardized mean difference (SMD) with 95% CI to combine trials that measured the same outcome using different measurement methods (Deeks 2011, Section 9.2.3).

Where some studies have reported endpoint data and others have reported change from baseline data (with errors), we will combine these in the meta‐analysis providing the outcomes have been reported using the same scale.

Unit of analysis issues

Cluster‐randomized trials

We will combine results from both individually‐ and cluster‐randomized studies providing there is little heterogeneity between the studies. We will label cluster‐randomized trials with a (C). If the authors of cluster‐randomized trials have conducted their analyses at a different level to that of allocation, and they have not appropriately accounted for the cluster design in their analyses, we will calculate trials' effective sample sizes to account for the effect of clustering in the data. Using the intra‐cluster correlation coefficient (ICC), derived from the trial (if available) or from another source (for example, using the ICCs derived from other similar trials), we will calculate the design effect with the formula provided in theCochrane Handbook for Systematic Reviews of Interventions (Deeks 2011, Section 9.3.1). We will report this and undertake a sensitivity analysis to investigate the effect of variations in ICC (see Sensitivity analysis).

Studies with more than two treatment groups

If we identify studies that compare more than two intervention groups (multi‐arm studies), where possible, we will combine groups to create a single pair‐wise comparison, or else use the methods set out in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b, Section 16.5), so as to avoid double counting study participants. For example, if the control group is shared by two or more study arms, we will divide the control group over the number of relevant subgroup categories; for dichotomous data, we will divide the events and the total population, while for continuous data, we will assume the same mean and SD but divide the total population.

Dealing with missing data

We will note missing outcome data and levels of attrition for included studies on the data extraction form. We will assess studies as having a low risk of attrition bias if study participation is described adequately, the proportion of missing data is balanced between groups, and the reasons for missing data are provided. We will explore the impact of including studies with high levels of missing data in the overall assessment of treatment effect by using a sensitivity analysis (see Sensitivity analysis). The denominator for each outcome in each trial will be the number randomized minus any participants whose outcomes are known to be missing.

For missing summary data, we will first contact the lead study authors for clarification. If this information is not available, and it is judged that missing data may not be missing at random, we will aim to impute missing summary data using other statistical information (for example, CI, standard errors) provided in the primary paper and impute the SD from other studies in the review.

Assessment of heterogeneity

We will assess included studies for clinical, methodological, and statistical heterogeneity. We will assess clinical heterogeneity by comparing the distribution of study participants, study setting, dose and duration of the intervention. We will evaluate methodological heterogeneity on the basis of factors such as the method of sequence generation, allocation concealment, blinding of outcome assessment, and losses to follow‐up. We will examine forest plots from a meta‐analysis to visually determine the level of heterogeneity (in terms of the size or direction of treatment effect) between studies. We will use the Chi² statistic to quantify the level of heterogeneity of intervention effects, with significant heterogeneity assessed as P value less than 0.10 in the Chi² test (Deeks 2011, Section 9.5). To assess the impact that heterogeneity, if present, has on the meta‐analysis, we will use the I² statistic (Higgins 2003), considering a value greater than 40% as significant heterogeneity (Deeks 2011, Section 9.5). Finally, if heterogeneity cannot be explained, we will use Tau² to quantify between‐study variance in a random‐effects meta‐analysis. We will consider substantial or considerable heterogeneity as T² greater than 0. We will note this in the text and explore it using prespecified subgroup analyses to determine whether heterogeneity is clinical or methodological (see Subgroup analysis and investigation of heterogeneity). We will interpret results with high levels of unexplained heterogeneity with caution.

Assessment of reporting biases

Statistical methods for identifying within‐study selective reporting are not yet well developed (Sterne 2011, Section 10). We will conduct a matrix of reported outcomes to examine patterns in reporting between studies. If we suspect that key outcomes have not been reported, we will contact the study authors to request that information.

We do not anticipate that there will be sufficient studies contributing data for any outcome to examine possible publication bias. However, if more than 10 studies reporting the same outcome of interest are available, we will generate funnel plots in Review Manager 5 (RevMan 5) (Review Manager 2014), and visually examine them for asymmetry.

Data synthesis

We will synthesize data using the latest version of Review Manager 2014.

We will use a random‐effects meta‐analysis for combining data, as we anticipate finding natural heterogeneity between studies attributable to the different doses, durations, populations and types of interventions.

Where different studies have reported the same outcomes using both continuous and dichotomous measures (such as z‐scores versus stunting), we will request continuous data from the study authors. If these are unavailable, we will re‐express ORs as SMD (or vice versa) and combine the results using the generic inverse variance method, as described in the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011). This is based on the assumption that continuous measurements in each intervention group follow a logistic distribution and that the variability of outcomes is the same in treated and control populations, thereby allowing the OR to be re‐expressed as an SMD (Chinn 2000).

'Summary of findings' table

We will present the findings of the review for our main comparisons in a 'Summary of findings' table, prepared using GRADE profiler software (GRADEpro GDT 2015). We will list the primary outcomes for each comparison with estimates of relative effects along with the number of participants and studies contributing data for those outcomes. Two review authors (JE and PR‐P) will assess the quality of the evidence for each outcome as high, moderate, low or very low, using the GRADE approach (Balshem 2011), which takes into consideration: study limitations, imprecision, inconsistency, indirectness, and publication bias. PR‐S will resolve differences of opinion.

We will include the outcomes listed below in the 'Summary of findings' table.

1. Linear growth (as measured by height‐for‐age z‐score or length‐for‐age z‐score)

2. Weight gain (as measured by weight‐for‐age z‐score)

3. All‐cause morbidity

4. Anaemia (defined as haemoglobin lower than 110 g/L for children aged 6 to 59 months, adjusted by altitude where appropriate)

Subgroup analysis and investigation of heterogeneity

We will conduct subgroup analyses by:

1. age: 6 to 23 months versus 24 to 59 months of age versus mixed; and

2. type of animal‐source foods: eggs versus meat versus fish versus dairy versus mixed.

We will use the primary outcomes for our subgroup analyses (see Primary outcomes).

We will not conduct subgroup analyses for those outcomes with 10 or fewer trials. We will explore the forest plots visually and identify where CIs do not overlap to identify differences between subgroup categories. We will also formally investigate differences between two or more subgroups by conducting t‐tests or F‐tests to calculate the significance of the ratio of MD to standard error (Borenstein 2009). Using Review Manager 2014, we will compute an I² statistic to describe variability in effect estimates from different subgroups that is due to genuine subgroup differences. The main focus of the analysis will be comparing magnitudes of effects across the different subgroups.

Sensitivity analysis

We will carry out a sensitivity analysis to examine the effects of removing studies at high risk of bias (those with high or unclear risk of bias for allocation concealment, similarity of baseline outcome measurements, incomplete outcome data) from the meta‐analysis. If we include cluster‐randomized trials or quasi‐RCTs, we will carry out a sensitivity analysis for type of study design using a range of ICC values.

Appendices

Appendix 1. Search strategy for Ovid MEDLINE

1 Meat/ 2 Meat Products/ 3 Red Meat/ 4 (beef or chicken$ or goat$ or meat or pork or poultry or venison).tw,kf. 5 exp Seafood/ 6 (fish$ or shellfish$ or seafood$ or sea‐food$).tw,kf. 7 Insects/ 8 (insect$ or caterpillar$ or spider$ or beetle$ or termite$ or ant or ants).tw,kf. 9 exp Eggs/ 10 exp Egg Proteins, Dietary/ 11 exp Dairy Products/ 12 exp Milk Proteins/ 13 (butter or cheese$ or dairy or eggs or milk or yog?urt).tw,kf. 14 (kefir or kephir or bulgaros).tw,kf. 15 ((buffalo$ or camel$ or cattle or cow$ or deer$ or donkey$ or goat$ or horse$ or pig$ or sheep$ or swine or reindeer$ or yak$) adj3 (protein$ or product$)).tw,kf. 16 ((animal$ or livestock) adj2 source$ adj2 (diet$ or feed$ or food$ or nutrition$ or protein$)).tw,kf. 17 or/1‐16 18 infant/ 19 Child, Preschool/ 20 child/ 21 (infan$ or baby or babies or toddler$ or preschoo$ or pre‐school$ or child$ or schoolage$ or school‐age$).tw. 22 or/18‐21 23 exp Child Development/ 24 infant nutrition disorders/ 25 malnutrition/ 26 child nutrition disorders/ 27 growth disorders/ 28 nutrition disorders/ 29 ANTHROPOMETRY/ 30 exp Body Composition/ 31 exp body height/ 32 exp body weight/ 33 body mass index/ 34 muscle, skeletal/ 35 Z score$.tw,kf. 36 (stunted or stunting).tw,kf. 37 body fat$.tw,kf. 38 length for age.tw,kf. 39 weight for age.tw,kf. 40 weight for length.tw,kf. 41 weight for height.tw,kf. 42 lean mass.tw,kf. 43 growth.tw,kf. 44 (BMI or body mass index).tw,kf. 45 or/23‐44 46 17 and 22 and 45 47 randomized controlled trial.pt. 48 controlled clinical trial.pt. 49 randomi#ed.ab. 50 placebo.ab. 51 clinical trials as topic.sh. 52 randomly.ab. 53 trial.ti. 54 or/47‐53 55 exp animals/ not humans.sh. 56 54 not 55 57 46 and 56

Appendix 2. 'Risk of bias' domains

Random sequence generation (checking for possible selection bias)

We will describe the method used to generate the allocation sequence in sufficient detail to allow an assessment of whether it produces comparable groups.

1. Low risk of bias: any truly random process; for example, random number table, computer random number generator.

2. High risk of bias: any process that is not strictly random; for example, odd or even date of birth, hospital or clinic record number.

3. Unclear risk of bias: information about the randomization process not available.

Allocation concealment (checking for possible selection bias)

We will describe the method used to conceal the allocation sequence in sufficient detail to determine whether intervention allocations could have been foreseen in advance of, or during, enrolment.

1. Low risk of bias: telephone or central randomization; consecutively numbered, sealed, opaque envelopes.

2. High risk of bias: open random allocation, unsealed or non‐opaque envelopes.

3. Unclear risk of bias: information about the allocation process not available.

Blinding of participants and personnel (checking for possible performance bias)

We will describe all measures used, if any, to blind study participants and personnel from knowledge of which intervention a participant received.

We will assess the risk of performance bias associated with blinding of participants as follows.

1. Low risk of bias: blinding of caregivers.

2. High risk of bias: non‐blinding is likely to influence care received throughout the study; for example, if mothers are aware their child is not receiving the treatment intervention.

3. Unclear risk of bias: inadequate information to assess the risk of bias as low or high.

We will assess the risk of performance bias associated with blinding of personnel as follows.

1. Low risk of bias: blinding of all personnel.

2. High risk of bias: non‐blinding is likely to influence care throughout the study, such as through nutrition counselling at follow‐up visits.

3. Unclear risk of bias: inadequate information to assess the risk of bias as low or high.

Whilst assessed separately, we will combine the results into a single evaluation of risk of bias associated with blinding of participants and personnel (Higgins 2011a, Section 8.5).

Blinding of outcome assessment (checking for possible detection bias)

We will describe all measures used, if any, to blind outcome assessors from knowledge as to which intervention a participant received.

1. Low risk of bias: if outcomes are objective, or if participants and key personnel were not blinded but the outcome assessment was blinded and the non‐blinding of others is unlikely to introduce bias.

2. High risk of bias: no blinding of outcome assessment, where measurement is likely to be influenced by lack of blinding, or where blinding could have been broken.

3. Unclear risk of bias: insufficient information to permit a judgement of low or high risk of bias.

Incomplete outcome data (checking for possible attrition bias through withdrawals, dropouts, protocol deviations)

We will assess outcomes in each included study.

1. Low risk of bias: either there were no missing outcome data or the missing outcome data were unlikely to bias the results based on the following considerations; study authors provided transparent documentation of participant flow throughout the study, the proportion of missing data was similar in the intervention and control groups, the reasons for missing data were provided and balanced across intervention and control groups, or the reasons for missing data were not likely to bias the results (for example, moving house).

2. High risk of bias: if missing outcome data were likely to bias the results. Studies will also receive this rating if an 'as‐treated (per protocol)' analysis was performed with substantial differences between the intervention received and that assigned at randomization, or if potentially inappropriate methods for imputation were used.

3. Unclear risk of bias: insufficient information to assess the risk of bias as low or high.

Selective reporting (checking for possible reporting bias)

We will state how the possibility of selective outcome reporting was examined and what was found.

1. Low risk of bias: it is clear that all of the study’s prespecified outcomes and all expected outcomes of interest to the review have been reported.

2. High risk of bias: not all the study’s prespecified outcomes have been reported, one or more reported primary outcomes were not prespecified, outcomes of interest are reported incompletely and so cannot be used, study fails to include results of a key outcome that would have been expected to have been reported.

3. Unclear risk of bias: insufficient information to assess the risk of bias as low or high.

Other bias (checking for other potential sources of bias)

We will assess if the study is free of other potential biases.

1. Low risk of bias: there is similarity between outcome measures at baseline, similarity between potential confounding variables at baseline, or adequate protection of study arms against contamination.

2. High risk of bias: there is no similarity between outcome measures at baseline, similarity between potential confounding variables at baseline, or adequate protection of study arms against contamination.

3. Unclear risk of bias: insufficient information to assess the risk of bias as low or high.

History

Protocol first published: Issue 10, 2017

Date	Event	Description
10 January 2017	Feedback has been incorporated	Amended for JP's latest edits and proofread for content

Contributions of authors

JE wrote the protocol, with extensive input from PR‐S. All review authors read and reviewed the protocol. JE is the guarantor for the review.

Sources of support

Internal sources

• World Health Organization (WHO), Switzerland.

  Pura Rayco‐Solon is a full‐time staff member of the WHO.

External sources

• World Health Organization (WHO), Switzerland.

  The Evidence and Programme Guidance Unit, Department of Nutrition for Health and Development, provided financial support to Jacob Eaton for his work in the preparation of this protocol.

• Bill & Melinda Gates Foundation, USA.

  The World Health Organization acknowledges the financial support of Bill & Melinda Gates Foundation for its work in building and maintaining updated systematic reviews on the effects of nutrition interventions.

Declarations of interest

Jacob Eaton received financial support from the World Health Organization (WHO) for his work completing this protocol. Pamela Rothpletz‐Puglia ‐ none known. Margaret R Dreker ‐ none known. Joyceline Kaganda ‐ none known. Lora Iannotti ‐ none known. Chessa Lutter ‐ none known. Pura Rayco‐Solon (PR‐S) is a full‐time staff member of the WHO.

Disclaimer: the authors alone are responsible for the views expressed in this publication and they do not necessarily represent the official position, decisions, policy or views of the WHO. The WHO gratefully acknowledges the financial contribution of the Bill & Melinda Gates Foundation, Nutrition International (NI; formerly Micronutrient Initiative (MI)), the Centers for Disease Control and Prevention (CDC), the US Agency for International development (USAID), and the Global Alliance for Improved Nutrition (GAIN) towards work in the area of nutrition. Donors do not fund specific guidelines and do not participate in any decision related to the guideline development process, including the composition of research questions, membership of the guideline groups, conduct and interpretation of systematic reviews, or formulation of recommendations.

New