Timing of Allergenic Food Introduction and Risk of Immunoglobulin E–Mediated Food Allergy

This systematic review and meta-analysis examines whether introducing allergenic foods to infants aged 2 to 12 months is associated with risk of developing immunoglobulin E–mediated food allergy at 1 to 5 years of age.


GRADE Evaluation of Evidence
The certainty of evidence in this report was rated using the GRADE approach, which has four levels of certainty: high, moderate, low, and very low. 12,13 Evidence can be downgraded by one or two levels if there is serious or very serious risk of bias, inconsistency of results, indirectness of evidence, imprecision, or likely or very likely publication bias. The interpretation of GRADE certainty ratings is that for high certainty evidence further research is very unlikely to change our confidence in the estimate of effect; for moderate certainty evidence further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate; for low certainty evidence further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate; and for very low certainty evidence any estimate of effect is very uncertain.

eFigure 1. PRISMA Flow Diagram of Study Selection
Search results and study selection procedure. Searches were run on May 11, 2021, and updated on June 28, 2022                  Heated egg powder (50mg containing 25mg egg protein daily from 6 to 9 months; 250mg containing 125mg egg protein daily from 9 to 12 months; n=73) vs placebo (pumpkin powder) from 6 to 12 months (n=74). One teaspoon of pasteurized raw whole egg powder (0.9g egg protein; n=49) vs rice flour powder (n=37) daily until 8 months. Introduction of cooked egg in both groups at 8 months.
Egg allergy: if no obvious prior reaction, single-dose (if low-risk) or graded (if high-risk) pasteurized raw egg challenge in sensitized children at 12 months Egg sensitization: egg-SPT ≥3mm at 12 months Withdrawal: participants who withdrew from study for a reason other than "moving overseas", or who ceased powder due to an allergic reaction Women's and Children's Hospital Foundation and Ilhan Food Allergy Foundation. Senior author reported being on the Boards for Nestlé, Danone, and ALK-Abelló. Intervention foods were peanut, soya, almond, cashew, hazelnut, pecan, pistachio, walnut, wheat, oat, milk, egg, cod, shrimp, salmon, and sesame (1 trial; 705 participants); egg, milk, wheat, soya, buckwheat, and peanut (1 trial; 163 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1303 participants); milk and egg, or egg and peanut, or milk and peanut, or milk, egg, peanut, cashew, almond, shrimp, walnut, wheat, salmon, and hazelnut (1 trial; 135 participants); peanut, milk, wheat, and egg (1 trial; 2397 participants). g Most information is from studies at high risk of bias for one domain or some concerns for multiple domains. The main reason for risk of bias is that withdrawal could have been influenced by knowledge of intervention received. h I 2 =89%; heterogeneity appears to be explained by lower compliance in two largest studies (both high allergen intake for multiple allergenic foods, using normal foods rather than powders). i Intervention foods were egg, milk, wheat, soya, buckwheat, and peanut (1 trial; 150 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1173 participants); peanut, milk, wheat, and egg (1 trial; 1504 participants). j All information is from studies at low risk of bias. k I 2 =73%; one study (Skjerven et al, 2022) which used skin prick test (SPT) ≥3mm showed reduced sensitization and two studies which used SPT >0mm and specific immunoglobulin E test showed no effect. l Although the estimate indicates reduced allergic sensitization, the 95% CI includes the possibility of no effect. There is therefore little meaningful information about the effect of earlier egg introduction on risk of allergy to any food. e The 95% CI excludes a RR of no effect or effect sizes that are not clinically meaningful.

eTable 4. Summary of Findings for Earlier Introduction of Egg
f Intervention foods were egg (8 trials; 2739 participants); peanut, soya, almond, cashew, hazelnut, pecan, pistachio, walnut, wheat, oat, milk, egg, cod, shrimp, salmon, and sesame (1 trial; 705 participants); egg, milk, wheat, soya, buckwheat, and peanut (1 trial; 163 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1303 participants); egg, or milk and egg, or egg and peanut, or milk, egg, peanut, cashew, almond, shrimp, walnut, wheat, salmon, and hazelnut (1 trial; 135 participants); peanut, milk, wheat, and egg (1 trial; 2397 participants). g Most information is from studies at high risk of bias, but findings appear to be similar to studies at low risk of bias or some concerns for one domain only. h I 2 =90%; heterogeneity appears to be explained by lower compliance in two large, pragmatic studies of high-dose multiple allergenic food introduction using normal foods, together with reports of allergic reactions to egg intervention in some smaller, single-food intervention trials. i Populations and interventions are similar to those which might be targeted outside of trial settings. Some studies used food powder, but findings were similar to studies using normal foods. Populations varied in risk factors such as eczema at baseline, in whether and how egg allergy was excluded prior to the intervention and in the nature of the intervention. j The 95% CI excludes a RR of no effect but includes effect sizes that are both trivial or uncertain and clinically meaningful. k Funnel plot asymmetrical and Egger's test P=0.004; however, likely due to lower compliance in larger studies of multiple, high-dose allergenic food introduction, compared with smaller studies of single allergen (egg) introduction. l Intervention foods were egg (6 trials; 1646 participants); egg, milk, wheat, soya, buckwheat, and peanut (1 trial; 161 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1165 participants); peanut, milk, wheat, and egg (1 trial; 1839 participants). m I 2 =0%; mostly similar RRs and overlapping CIs. No obvious difference in outcome between single and multiple allergenic food introduction. Borderline interaction for high versus low dose egg introduction (P=0.02), with increased effect in low dose group. n In one further high risk of bias trial which could not be included in meta-analysis, egg yolk was introduced before 3 weeks versus after 6 months in a cohort of 1753 infants; 9 participants in the earlier and 4 in the later introduction group developed symptoms before 4 and by 7 months, respectively, but precise outcome definition and denominators were not available. o Intervention foods were egg (5 trials; 1504 participants); egg, milk, wheat, soya, buckwheat, and peanut (1 trial ; 1839 participants). b Most or all information is from studies at low risk of bias or some concerns for one domain only. c I 2 =81%; heterogeneity appears to be partly (but not wholly) explained by a lack of effect in the single peanut-only intervention trial but reduced allergy in the multiple food intervention trials. This interaction was statistically significant (P=0.02). d Some data contributing to this effect estimate are from trials of multiple allergenic food introduction. The single trial of peanut only shows no effect. Populations varied in risk factors such as eczema at baseline, in whether and how peanut allergy was excluded prior to the intervention and in the nature of the intervention. e The 95% CI includes both a meaningful reduction and a trivial reduction or small increase in RR.
f Intervention foods were peanut (1 trial; 640 participants); peanut, soya, almond, cashew, hazelnut, pecan, pistachio, walnut, wheat, oat, milk, egg, cod, shrimp, salmon, and sesame (1 trial; 705 participants); egg, milk, wheat, soya, buckwheat, and peanut (1 trial; 163 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1303 participants); peanut, or egg and peanut, or milk and peanut, or milk, egg, peanut, cashew, almond, shrimp, walnut, wheat, salmon, and hazelnut (1 trial; 135 participants); peanut, milk, wheat, and egg (1 trial; 2397 participants). g Most information is from studies at high risk of bias for one domain or some concerns for multiple domains. The main reason for risk of bias is that withdrawal could have been influenced by knowledge of intervention received. h I 2 =91%; heterogeneity appears to be partly explained by lower compliance in two large, pragmatic studies of high-dose multiple allergenic food introduction using normal foods. i The 95% CI excludes a RR of no effect or effect sizes that are not clinically meaningful. j Intervention foods were peanut (1 trial; 629 participants); egg, milk, wheat, soya, buckwheat, and peanut (1 trial; 150 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1173 participants); peanut, milk, wheat, and egg (1 trial; 1504 participants). k I 2 =61%; one study (Skjerven et al, 2022) of multiple allergenic food introduction which used skin prick test (SPT) ≥3mm showed reduced sensitization and appears to explain the heterogeneity, but the reason for different findings in this trial is unclear. l Most of the events contributing to this effect estimate are from a trial of single allergenic food introduction. Populations varied in risk factors such as eczema at baseline, in whether and how peanut allergy was excluded prior to the intervention and in the nature of the intervention. m Intervention foods were peanut (1 trial; 628 participants); egg, milk, wheat, soya, buckwheat, and peanut (1 trial; 161 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1168 participants); peanut, milk, wheat, and egg (1 trial; 1839 participants). n I 2 =21%; mostly similar RRs and overlapping CIs.
o Intervention foods were peanut (1 trial; 617 participants); egg, milk, wheat, soya, buckwheat, and peanut (1 trial; 145 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1168 participants); peanut, milk, wheat, and egg (1 trial; 1504 participants). p I 2 =77%; one study (Skjerven et al, 2022) of multiple allergenic food introduction which used SPT ≥3mm showed reduced sensitization and appears to explain the heterogeneity, but the reason for different findings in this trial is unclear. Post hoc subgroup analysis showed a possible interaction for method of assessment (P=0.  ; 161 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1162 participants); milk, or milk and egg, or milk and peanut, or milk, egg, peanut, cashew, almond, shrimp, walnut, wheat, salmon, and hazelnut (1 trial; 133 participants); peanut, milk, wheat, and egg (1 trial; 1839 participants). b Most information is from studies at high risk of bias, but findings appear to be similar to studies at low risk of bias or some concerns for one domain only. c I 2 =83%; heterogeneity appears to be partly, but not completely, explained by one high risk of bias study. d Interventions were not all representative of typical ways to introduce cow's milk products to an infant's diet. Some studies used food powder, others used formula milk compared with a control group using soya formula, but often using very early and specific exposure regimes. e The 95% CI includes both a meaningful reduction and increase in risk or no effect. f In one further high risk of bias study which could not be included in meta-analysis, none of 120 children who were given cow's milk (formula) and none of 115 children who were given a soya formula for the first 9 months of life developed allergy to foods such as soya, fish, orange, and nuts, and one child became clinically sensitive to chocolate after a 10-year follow-up. g Intervention foods were milk (6 trials; 3192 participants); peanut, soya, almond, cashew, hazelnut, pecan, pistachio, walnut, wheat, oat, milk, egg, cod, shrimp, salmon, and sesame (1 trial; 705 participants); egg, milk, wheat, soya, buckwheat, and peanut (1 trial; 163 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1303 participants); milk, or milk and egg, or milk and peanut, or milk, egg, peanut, cashew, almond, shrimp, walnut, wheat, salmon, and hazelnut (1 trial; 135 participants); peanut, milk, wheat, and egg (1 trial; 2397 participants). h I 2 =94%; heterogeneity is extreme and is fully explained by high rates of withdrawal in two pragmatic trials of multiple, high-dose, real food intervention trials, and a trial where many participants withdrew from the soya milk intervention due to a preference for cow's milk. Without these three trials included, the RR is 0.86 (95% CI, 0.70-1.06; I 2 =0%). i Funnel plot asymmetrical and Egger's test P=0.01; however, likely due to lower compliance in larger studies of multiple, high-dose allergenic food introduction, compared with smaller studies of single allergen (milk) introduction. j In one further high risk of bias trial which could not be included in meta-analysis, 32/249 participants in the cow's milk group were given soya milk and 41/238 participants in the cow's milk avoidance group were given cow's milk. However, total number of participants randomized in each group was not clear as these numbers excluded some post-randomization exclusions and the number of events in the cow's milk avoidance group was inconsistent between study publications. k Intervention foods were milk (3 trials; 734 participants); egg, milk, wheat, soya, buckwheat, and peanut (1 trial; 161 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1166 participants); peanut, milk, wheat, and egg (1 trial; 1839 participants). l Most information is from studies at high risk of bias for at least one domain and/or some concerns for multiple domains, with high heterogeneity in the high risk of bias information. The main reason for risk of bias is that outcome assessors were not blinded. m I 2 =36%; heterogeneity appears to be explained by risk of bias, with low risk of bias studies showing no heterogeneity and RR 0.32 (95% CI, 0.09-1.18). n One further high risk of bias trial (Sakihara et al, 2021) reported milk allergy in 2/242 in the earlier and 17/249 in the later introduction group, assessed at 6 months, so could not be included in meta-analysis. o Intervention foods were milk (4 trials; 2071 participants); egg, milk, wheat, soya, buckwheat, and peanut (1 trial; 145 participants); milk, peanut, egg, sesame, white fish, and wheat (1 trial; 1167 participants); peanut, milk, wheat, and egg (1 trial; 1504 participants). p Most information is from studies at low risk of bias or some concerns for one domain only. q I 2 =45%; heterogeneity may be partly explained by method of outcome measurement. Post hoc subgroup analysis showed a possible interaction for method of assessment (P=0. CI, confidence interval; RR, risk ratio. a Prespecified subgroup analyses for single vs multiple allergenic food introduction; high vs low allergen intake, using a cut-off of 2g per week of the relevant individual food protein(s); milk feeding status at enrolment -exclusively breastmilk, with or without non-milk foods, vs mixed breastmilk/other milk or non-breastmilk fed; and a post hoc subgroup analysis of high vs standard/low risk for atopic disease based on family or personal history. Subgroup analyses were conducted using study-level variables; one small study (Quake et al, 2022) which used a mix of high and low allergen interventions within the study could not be included in the high vs low allergen intake subgroup analysis. a Prespecified subgroup analyses for single vs multiple allergenic food introduction; high vs low allergen intake, using a cut-off of 2g per week of the relevant individual food protein(s); milk feeding status at enrolment -exclusively breastmilk, with or without non-milk foods, vs mixed breastmilk/other milk or non-breastmilk fed; and a post hoc subgroup analysis of high vs standard/low risk for atopic disease based on family or personal history. Subgroup analyses were conducted using study-level variables; one small study (Quake et al, 2022) which used a mix of single and multiple allergenic food and high and low allergen interventions within the study could not be included in the single vs multiple allergenic food introduction and high vs low allergen intake subgroup analyses.