In our cohort, we found that later formula introduction was associated with a lower odds of food allergy in children without eczema. There were similar trends for delayed introduction (>6 months) of rice/wheat cereal and other solid foods. However, the risk of developing of food allergy in high-risk children with eczema was not modified by early or late introduction of formula or food. Early introduction of formula did result in an earlier diagnosis of food allergy in children without eczema. This was not seen in children with eczema.
Not surprisingly, our findings were consistent with previous studies that found eczema to be a significant risk factor for the development of food allergen sensitization18,19,27
and food allergy.2
Unexpectedly, we did not find a protective effect associated with early introduction of formula or food. Instead, later introduction of both formula as well as solid foods (cut point of 1 year) seemed to have a protective effect, but only in children without eczema. It is possible that a proportion of the children with eczema may have had undiagnosed food allergy driving their disease before any change in feeding. However, the need to stratify by eczema or risk may be 1 factor leading to the inconsistent findings in previous literature.
The literature is not conclusive with regard to the effects of introduction of solid food, duration of exclusive breast feeding, and use of hydrolyzed formula on the development of food allergy in general.28
Some studies suggest a detrimental effect on eczema if multiple foods were introduced early in life.6,7
Also, early episodic introduction of cow's milk into infant diets may increase rates of milk allergy.10–13
In contrast, a number of studies and meta-analyses did not find any protective effect on atopic outcomes associated with delaying solid food introduction beyond 6 months of age.5,6,16
Others found an increased risk of food allergy4
with delayed food introduction. Some of these findings may have been due to reverse causation,7
as risk factors such as family history of allergy or early manifestations of allergic disorders may alter timing of food introduction.29
Clear answers to the questions around food introduction and tolerance30
may only become available when the results of ongoing clinical trials are known. However, our data suggest that it would be important to appropriately stratify individuals by eczema when evaluating the effect of food introduction on the development of food allergy.
There are a number of potential limitations to our study. First, we recognize that the cross-sectional nature of the data (derived from the baseline visit of our prospective cohort) does not allow us to definitively answer these questions. The age of introduction of foods is based on recall at this time point. On the other hand, our analysis of younger children (<3 years of age) showed no change in the direction or significance of our findings. Second, we have optimized our food allergy phenotype to include only those with clear objective symptoms, timing of symptoms close to ingestion, and corroborating test results. Although this is not equal to performing food challenges, we have previously corroborated diagnoses among a subset of study subjects with food allergy followed in the allergy clinic at our tertiary care center who have had either oral food challenges or a previous history of anaphylaxis.31
Also, using more stringent criteria that incorporated sIgE levels ≥95% PPV for a positive challenge (milk, egg or peanut), the findings remained significant with similar magnitude and direction of effect. Another important limitation is that we have fewer individuals with solid food introduction for nonrice/wheat cereals before 6 months of age. Thus, the analyses for solid food introduction was only adequately powered using a cut point of 1 year of age, which is beyond the age of solid food introduction of many children in the general population. However, the results using a cut point of 6 months showed similar magnitude and direction of findings. In those with an EMR record, there was only 75.4% agreement of EMR diagnostic coding with parent report of any previous physician diagnosis of eczema. Many children in the cohort are at an age when eczema may have remitted. As such, historical diagnoses of eczema may not be confirmed by the current EMR diagnoses, especially in those children not initially seen in the allergy clinic. We also cannot exclude residual confounding from unmeasured variables such as self-imposed maternal avoidance diets during pregnancy. This will need to be addressed in future longitudinal prospective birth cohort studies. It is also possible that some young children have undiagnosed food allergy concomitant with the onset of eczema. Since a significant subset (21%) had food allergy diagnosed before eczema being diagnosed, there may be a degree of error due to the retrospective nature of ascertainment of the time of onset of both eczema and food allergy. Both of these diseases tend to occur at a similar time, early in childhood. This potential recall bias also needs to be addressed in prospective well-phenotyped birth cohort studies. This does not change the impact of our findings. We do not hypothesize that skin manifestations of eczema led to the development of food allergy. Rather, eczema may identify a group of high-risk children with a different response to timing of food introduction. Finally, this is an affluent, sub-urban, largely white cohort, decreasing generalizability to other populations such as minorities or urban poor populations.
Our study suggests that delayed formula/solid food introduction may be beneficial in infants without eczema. However, a similar protective effect was not seen in children with eczema. These findings, if confirmed by prospective studies, would provide new evidence-based data which would suggest that important clinical questions around the risks and benefits of early food introduction must be stratified by the atopic status of the child.