Complementary and Allergenic Food Introduction in Infants: An Umbrella Review

The protocol was created based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2015 checklist.¹⁸  Reporting has been checked using the Preferred Reporting Items for Overviews of Systematic Reviews with harms 2017 checklist.¹⁹  The protocol was published previously on PROSPERO (CRD42020158525). Significant protocol amendments have been explained.

We included systematic reviews of human participants that reported age of first introduction of complementary or allergenic food by age 1 and outcomes after and including age 1. A systematic review was defined as in the Cochrane handbook,²⁰  with full compliance assessed in the quality assessment. Complementary food was defined as any solid, semisolid, or soft food, excluding breastmilk, infant formula, supplements, vitamins, or water. Allergenic foods included peanuts, hen’s egg, cow’s milk (not including through introduction of infant formula), shellfish, tree nuts, soybean, and wheat. Outcomes included common noncommunicable diseases, which could plausibly be affected by diet including allergic, inflammatory, and autoimmune diseases, and growth and nutritional outcomes (Table 1). Narrative reviews, original studies, breastmilk or formula exposures, and nonpeer review literature were excluded. The full inclusion and exclusion criteria is presented in Table 1.

We searched Medline, Embase, Cochrane Database of Systematic Reviews, and the last 24 months of PubMed to capture reviews not yet available in Medline, from inception to July 25, 2022 (Supplemental Table 4). We included systematic reviews in any language. Reference lists of included studies were hand-checked to ensure we did not miss key reviews. Conference abstracts were checked to see if results had been published in paper form. Duplicates were removed using Endnote X8. A librarian and expert on the topic helped create the search strategy. Additional primary studies were not included but were discussed where appropriate in the discussion.

Article screening and extraction was performed in duplicate. Two independent reviewers screened titles and abstracts (V.X.S. and D.C., Y.W., or G.G.) and read selected full texts for eligibility (V.X.S. and D.C.). Conflicts were resolved by discussion or a third reviewer (J.J.K.). The title and abstract screening checklist is in Supplemental Table 5 and full texts were assessed as per the inclusion and exclusion criteria in Table 1. We piloted the selection process with 50 titles and abstracts, 10 full texts, and 2 extractions and quality assessments. Extraction was conducted using the form in Supplemental Figure 4. Agreement was measured using random agreement probability and Cohen’s κ.

Two independent reviewers (V.X.S. and D.C. or Y.W.) appraised the quality and risk of bias of systematic reviews using the A Measurement Tool to Assess Systematic Reviews (AMSTAR 2) tool (2017).²¹  AMSTAR 2 evaluates the overall confidence in reviews as high, moderate, low, or very low, according to critical domains (Supplemental Methods). Low and very low confidence reviews may not present an accurate or comprehensive summary of the data. Reviews were included regardless of quality. Where strength of evidence was not assessed with Grading of Recommendations Assessment, Development, and Evaluation (GRADE) as part of the original systematic review, we performed our own GRADE assessment using the information available within the review to assist with between-review outcome comparisons (Supplemental Methods, Supplemental Fig 5). Heterogeneity (I²) and selection bias outcomes were extracted in duplicate where available.

Results were grouped by health outcome and presented in a narrative manner as decided a priori because of expected heterogeneity of results and according to recommendations in the Cochrane Handbook.²⁰  We had planned to separate outcomes by region of the world; this was not possible as few systematic reviews distinguished regions. All reviews were presented even if primary studies within the reviews overlapped. Overlap of primary studies presented in multiple reviews was documented. Where data were missing, authors were contacted to obtain further information. Covidence was used to manage articles, screen, and extract data. The statistical software package Stata (release 17.0; StataCorp, Tx) for data analysis.

We identified 8427 records, of which 4015 title and abstracts and 156 full texts were screened in duplicate (Fig 1). We included 32 systematic reviews for extraction and qualitative synthesis. The reviewers had good agreement both for the title and abstract screening (95%) and full-text review (95%; Supplemental Table 6). A list of the 124 excluded full texts and exclusion reasons are in Supplemental Table 7. In total, 410 index publications (the first occurrence of a primary publication in the included reviews) were included in the 32 reviews. Overlap of primary studies presented in multiple reviews and number of duplicate articles (articles also published in another review) are available in Supplemental Table 8.

Characteristics of the 32 reviews can be found in Table 2. Twenty-two reviews included both randomized controlled trial (RCT) and observational study designs. Observational study design were commonly included because of the lack of published intervention trials. Most reviews (n = 19) included age of introduction of complementary food as an exposure of interest. The review by Vissers 2018²²  was the only one focused solely on studies of preterm infants. Medline, PubMed, Embase, and Cochrane were among the most common databases searched by included reviews. Further details can be found in Supplemental Table 9 and 10.

The AMSTAR 2 ratings of included systematic reviews are shown in Fig 2, with further details in Supplemental Table 10. Of the 32 included reviews, 7 were high quality (only 1 noncritical weakness each).^23–29  English 2019a,b^24,27  and Obbagy 2019a,b,c^23,25,26  did not explicitly justify their restriction to English language only searches. Ierodiakonou 2016²⁸  did not provide a list of the studies they excluded. De Silva 2020²⁹  only provided basic details on their inclusion criteria, but not extensive details. We have low to very low confidence that 20 of the 32 reviews accurately and objectively summarize the literature in question.^22,30–48 

About half of the reviews contained detailed, publicly available protocols established before the review (question #2).^{22–31,49–53}  Two of those did not specify their meta-analysis plans or plans to investigate heterogeneity.^31,50  Only 7 reviews used a comprehensive literature search strategy (question #4), which was unlikely to be biased.^{28,29,38,41,46,50,53} 

Sixteen of the 22 reviews including all study designs assessed the risk of bias adequately in all study designs (question #9).^{22–28,32,36,37,41,43,50–53}  Two^39,48  of 7 reviews,^{30,38–40,46–48}  including only observational studies did not adequately assess risk of bias.

Lanigan 2001³³  and Pinto-Sanchez 2016⁴¹  had a potential conflict as they were both funded by the baby food company, Nestle Ltd (question #16). Most studies did not report whether the funding sources of included studies was checked (n = 25; question #10).

Thirteen studies conducted meta-analyses. Two^32,42  did not use appropriate meta-analysis methods. They either did not investigate causes of heterogeneity or they combined RCT and observational studies within their meta-analysis. Eight of the 13 meta-analyses failed to assess or discuss potential publication bias and the effects on their results (question #15).^{32,35,36,38,42,49,52,53} 

The key findings from the 32 included reviews are summarized by outcome in Table 3 and the reviews cover 10 main outcomes (Supplemental Figure 3). The following sections will describe the findings according to childhood outcome. Areas explored by the included reviews but found to be lacking evidence are in Supplemental Table 11.

Eleven of the 32 included reviews investigated food sensitization and/or food allergy as an outcome. Four reviews agreed there was moderate quality evidence to support no association between general complementary food and food allergy later in life. Burgess 2019³¹  was the only food allergy review which performed meta-analyses. Authors found that introduction of complementary food at or after 4 months increased the likelihood of food sensitization compared with before 4 months. However, their meta-analysis only included 3 cohorts. The other 21 original studies were lacking the correct exposure or outcome data.

Eight reviews examined the exposure to individual foods, such as peanut, egg, milk, and wheat, and food allergy outcomes. Ierodiakonou 2016 was the highest quality meta-analysis conducted on early egg (6 RCTs), peanut (2 RCTs), and milk (2 RCTs) introduction in association with egg, peanut, and milk allergy, respectively. The more recent Obbagy 2019a²⁵  presented consistent findings without a meta-analysis. De Silva 2020²⁹  and Dai 2021³⁵  did not find any additional original studies (Supplemental Table 8).

Two reviews^44,45  conducted before 2015 (when an RCT showing peanut introduction prevented peanut allergy was published⁵⁴ ) indicated insufficient evidence for food allergy. After 2015, there was moderate to high certainty evidence from high quality reviews that peanut introduction between 4 and 11 months was protective of peanut allergy (n = 5 reviews, 2 RCTs), and egg introduction between 4 and 6 months was protective of egg allergy (n = 6 reviews, 6 RCTs). There was consensus between reviews that age of cow’s milk (n = 3) and wheat introduction (n = 2) was not associated with milk or wheat allergy, respectively, although the evidence was low to very low, from few, mostly observational studies.

Six reviews reported asthma outcomes. Four of them found no association between age of introduction of complementary food, cow’s milk, egg, peanut, tree nuts, or sesame in the first year of life and asthma later in life, although the certainty of the evidence for the allergenic foods was low to very low. Ierodiakonou 2016²⁸  and Papamichael 2018³⁸  found a decreased risk of current wheeze and asthma in children under 4 years old who had been introduced to fish and eaten frequently (at least weekly) from 6 to 9 months of age in meta-analyses of 11 RCTs and 3 cohorts, respectively.^28,38 

Four^25,28,36,49  of 6 reviews that assessed eczema found no association between age of introduction of complementary food or allergenic food introduction and eczema. Griebler 2015³²  and Tarini 2006⁴⁴  (very low confidence reviews) found very low certainty evidence linking introduction of cow’s milk by age 1 and eczema and the introduction of complementary food before 4 months and eczema. A high quality review²⁵  in 2019 did not support earlier findings indicating there was moderate certainty evidence of no effect. Authors found limited evidence that introduction of fish from 5 to 11 months, compared with delaying until 1 year, may reduce risk of atopic dermatitis. Though evidence was conflicting, most studies reported that age of introduction of complementary or allergenic foods did not affect eczema outcomes.

Ierodiakonou 2016²⁸  found that fish introduction before 6 months versus 6 to 12 months decreased the risk of allergic rhinoconjunctivitis. The authors indicated the evidence was low certainty from 4 prospective cohort studies with high heterogeneity (I2 = 59.2%).²⁸  Obbagy 2019a²⁵  determined there was low certainty evidence supporting no effect of introduction of fish, vegetables, and meat on allergic rhinitis, but there was not enough evidence for other foods. Two other reviews^44,49  found insufficient evidence on the association between any food introduction and allergic rhinitis.

Three^28,43,53  of 4 reviews presented high certainty evidence that introduction of gluten or wheat products before age 1 was not associated with developing celiac disease or autoimmunity later in life. Contrarily, Pinto-Sanchez 2016⁴¹  (low confidence review) reported gluten introduction past 6 months could increase the risk of celiac disease compared with earlier. This data were from 5 cohort studies (n = 100 224 participants), whereas data from Szajewska 2015⁵³  and Ierodiakonou 2016²⁸  was based on RCT data (4 RCTs; n = 1822).

Three reviews found low to very low certainty evidence that introduction of complementary food from 4 months onwards was not associated with overweight or obesity. Five reviews found low certainty evidence that introducing complementary food before 4 months could increase the risk of overweight and obesity in childhood and adolescence.^{27,40,46–48}  English 2019a²⁷  (the highest quality review) included 2 RCTs and 71 prospective cohort studies to support this finding. Wang 2016⁴⁶  performed a meta-analysis of eight studies on overweight and 5 studies on obesity, but heterogeneity was moderate to high and the review itself was low quality. Authors found an overall risk ratio of 1.18 (95% confidence interval [CI], 1.06–1.31) for risk of overweight and 1.33 (95% CI, 1.07–1.64) for risk of obesity if complementary food were introduced before 4 months compared with later.⁴⁶  One review found insufficient evidence on the introduction of complementary foods in preterm infants and obesity outcomes.²²  No systematic reviews examined the association between age at introduction of allergenic foods and risk of overweight or obesity.

Four reviews^27,32,33,42  presented data on the introduction of complementary food or cow’s milk on childhood weight and growth (weight, height, BMI, etc.), rather than the searched terms for obesity and underweight. None reported an association. Lanigan³³  found insufficient evidence on the topic in 2001. Later reviews^27,32,42  provided low to moderate evidence that age of introduction of complementary food or cow’s milk has no effect on growth outcomes.

Four reviews examined the association between timing of introduction of cow’s milk^28,32,39  (n = 3) or gluten^28,52  (n = 2) and type 1 diabetes. Patelarou 2012³⁹  reported an increased risk of type 1 diabetes in infants introduced to cow’s milk early (“early” age: by 3, 5, or 7 months, versus later), however the data were inconsistent and very low certainty (n = 6 observational studies). Piescik-Lech 2017⁵²  found very low certainty evidence (n = 2 observational studies) that gluten introduction before 4 months may increase the risk of type 1 diabetes later in life. Ierodiakonou 2016²⁸  (high-quality review) did not find any effect of age of cow’s milk introduction or gluten on type 1 diabetes.

Two reviews examined iron deficiency and/or anemia.^26,42  Qasem 2016⁴²  found very low certainty evidence that introducing complementary food at 4 months compared with 6 months could increase hemoglobin status (1 RCT) in developing countries (mean difference, 5.0g/L; 95% CI, 1.5–8.5g/L; P = .005) and plasma ferritin concentration (2 RCTs) in both developing (P = .050) and developed (P = .040) countries. However, a higher quality review²⁶  in 2019 found moderate certainty evidence of no association with iron status in childhood.

Two reviews examine a variety of other health outcomes in relation to age of introduction of complementary food, cow’s milk, or other allergenic foods.^28,32  Griebler 2016³²  found very low certainty evidence of no association between introduction of cow’s milk and gastrointestinal blood loss (3 non-RCTs, 1 cohort). There was insufficient evidence to examine the effect of age of introduction of cow’s milk and dehydration or gastrointestinal diseases. Ierodiakonou 2016²⁸  found that age of introduction of allergenic solids by age 1 was not associated with the risk of a range of autoimmune diseases (eg, inflammatory bowel disease, autoimmune thyroid disease, and juvenile rheumatoid arthritis), from 48 observational studies.

Two studies determined there was insufficient evidence on bone health or osteoporosis.^23,32  No reviews were found on the outcomes: eosinophilic esophagitis, pernicious anemia, diabetes type 2, dental caries, neurodevelopment, cardiovascular disease, respiratory tract infection, or diarrhea.

Overall, there was evidence to support that introduction of complementary and allergenic solids in the first year of life, from around 6 months of age, was safe for a range of health outcomes. We found 19 of the systematic reviews covered food allergy or obesity. There was high certainty RCT evidence that the introduction of peanut and egg in the first year of life reduces the risk of peanut and egg allergy, respectively. Infants who are introduced to complementary food before 4 months may have a higher risk of overweight and obesity, though reviews including alternate growth outcomes did not see an effect. There was no evidence introduction of gluten was associated to celiac disease, supporting the fact that the immunologic pathway is different in celiac disease than in food allergies.

The evidence supporting allergen introduction for the prevention of food allergy was based on RCTs of peanut or egg. All 5 reviews for peanut are based on the same 2 RCTs from primarily Caucasian, high-income countries.^54,55  Therefore, the ability to extrapolate the findings to different populations is limited. More original research is required for cow’s milk, tree nuts, sesame, and shellfish.

Reviews on obesity examined a range of food as exposures, but not allergenic foods specifically. The finding that food introduction before 4 months may increase the risk of obesity should be examined with caution. Most reviews agreed that results were of low certainty and may be biased. Early interest in food could be a marker for genetic susceptibility to increased appetite or hormonal pathways which increase the risk of obesity.⁵⁶  Early feeding by parents may indicate a family dynamic and environment that promotes eating larger portions and leads to obesity.⁵⁷ 

The WHO recommends a varied diet including “meat, poultry, fish or eggs, as well as vitamin-A rich fruits and vegetables”⁷  from 6 to 24 months, while continuing to breastfeed. Nutritionally, the types of food introduced may affect some health outcomes more than the timing. Iron-rich foods in infancy prevent iron deficiency in breastfed infants.²⁶  To the contrary, sugar-sweetened drinks may increase the likelihood of obesity in childhood.⁵⁸ 

Parents can get stressed and overwhelmed during the weaning period (when transitioning infants to a solid diet).⁵⁹  Some families note that changing and conflicting advice can lack practical suggestions and make this time more difficult.⁵⁹  Some health professionals now recommend weaning infants when they are “developmentally ready,” rather than at a specific age. One systematic review found that baby led weaning could have growth, health, and developmental benefits, though the original articles define baby led weaning inconsistently, with a potentially biased methodology.⁶⁰  Contrarily, delaying the introduction of “lumpy” foods until too late could lead to feeding problems later in childhood.⁶¹ 

Most infant feeding guidelines recommend beginning introducing complementary food to a child’s diet around 6 months of age (eg, WHO, Europe, US, Australia).^7,12,13,62  This overview has found evidence to support 6 months as a safe age to introduce foods. Interestingly, WHO, European, and Australian general infant feeding guidelines do not include recommendations on how or when to introduce allergenic solids, even though there is now considerable evidence that this could prevent food allergy.^15–17  Furthermore, earlier allergen introduction has not been found to have a negative effect on breastfeeding rates in trials or in population-based studies.^55,63 

There is a lack of evidence between complementary or allergenic food and bone health, iron deficiency, gastrointestinal diseases, autoimmune disorders, asthma, eczema, or allergic rhinitis. Evidence was very low regarding cow’s milk or wheat introduction and milk or wheat allergy, respectively.

The introduction of fish around 6 months compared with later may prevent asthma and allergic rhinitis in childhood, however the quality of evidence is limited and conflicting. This association may be because of omega-3 fatty acid, which has been linked to protection from asthma in an RCT.⁶⁴  There was low certainty evidence supporting no effect of introduction of fish, vegetables, and meat on allergic rhinitis, but not enough evidence for other foods.²⁵ 

The data available for overweight and obesity and type 1 diabetes were low certainty. Preterm infants are a group of interest for obesity prevention, however the only review to examine this topic found insufficient evidence.²²  There were no systematic reviews examining the effect of age at introduction of allergenic foods and risk of overweight or obesity.

No reviews were found on timing of any food introduction and: eosinophilic esophagitis, pernicious anemia, diabetes type 2, dental caries, neurodevelopment, cardiovascular disease, respiratory tract infection, or diarrhea.

A strength of the current overview is that we followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. We have collated systematic reviews to obtain the highest quality data available to date on age of food introduction to infants. Conducting an overview decreases bias in the results, whereas identifying large gaps in infant feeding research. Observational studies were very common, likely because it would be unethical to randomize complementary food introduction to infants outside of the recommended age bracket of 4 to 6 months. Over 410 original papers were included within the 32 reviews.

A potential limitation was not including gray literature because we were interested in reviews that had undergone peer review, however we did hand search government documents for systematic reviews. Because only 1 review focused on preterm infants, we are not able to extrapolate out findings to this subgroup of infants. Further research into preterm infants would be warranted because preterm infants being highly susceptible to health problems. This overview depends on the quality of the included reviews, so under-researched topics may not have a systematic review published yet.

This review supports current recommendations to introduce complementary and allergenic solids by age 1, from around 6 months of age, as it seems safe for a range of childhood health outcomes. Nevertheless, infant food introduction may occur at different ages for a variety of complex reasons. Complementary feeding is influenced by a range of socioeconomic and cultural factors, so there is unlikely a “1-size fits all” solution to feeding methods. There is a need to provide higher quality evidence from original studies and systematic reviews regarding introduction of allergenic foods and the outcomes of asthma, eczema, and allergic rhinitis. Original studies are also needed to examine the effect of complementary or allergenic foods on bone health, iron deficiency, gastrointestinal diseases, or other autoimmune disorders.

We thank the librarian Ms. Poh Chua from the Royal Children’s Hospital Library for assistance with the search strategy and Prof. Anne-Louise Ponsonby for her support as a PhD supervisor to V.X.S.

AMSTAR

A Measurement Tool to Assess Systematic Reviews

RCT

randomized controlled trial

WHO

World Health Organization