. 2019 Feb 19;2019(2):CD012818. doi: 10.1002/14651858.CD012818.pub2

1. Unused methods.

Measures of treatment effect	Dichotomous data We will present dichotomous data as OR with 95% CI (Deeks 2011).
Measures of treatment effect	Continuous data We will use the SMD with 95% CI to combine trials that measure the same outcome using different measurement methods.
Unit of analysis issues	Studies with more than two treatment groups If a control group is shared by two or more study arms, we will divide the control group over the number of relevant categories using the methods described in the Cochrane Handbook for Systematic Reviews of Interventions so as to avoid double counting study participants (Higgins 2011).
Dealing with missing data	We will explore the impact of including studies with high levels of missing data in the overall assessment of treatment effect by conducting a 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 we judge that missing data may not be missing at random, we will aim to impute missing summary data using other statistical information (e.g. CI, standard errors) provided in the primary paper and impute the SD from other studies in the review.
Assessment of reporting biases	If more than 10 studies reporting the same outcome of interest are available, we will generate funnel plots in Review Manager 5 and visually examine them for asymmetry (Review Manager 2014).
Data synthesis	If continuous measures are not available for primary outcomes (such as LAZ scores), and we are unable to obtain the data from the study authors, we will use dichotomous outcomes and 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).
Subgroup analysis and investigation of heterogeneity	We will conduct subgroup analyses by: age (6 to 23 months versus 24 to 59 months versus mixed); and 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 visually explore the forest plots 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. Using Review Manager 2014 (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 consider the impact of removing studies at high risk of bias (due to allocation concealment or baseline imbalances in outcomes between groups). We will also carry out a sensitivity analysis for quasi‐RCTs using a range of ICC values.

CI: confidence interval  LAZ: length‐for‐age z score  MD: mean difference  OR: odds ratio  SD: standard deviation  SMD: standardized mean difference