DAY 2 READING A MEDICAL ARTICLE

This is the first comprehensive population-based study on food allergies in the United States. Such data provide a more accurate estimate of the temporal trends in the incidence of food allergy at a population level (outpatients and inpatients combined) as compared with hospital-based case series or patient self-reported surveys, which are distorted by referral bias and ascertainment bias.¹⁶ Consistent with findings from a previous study of this population,¹⁷ this retrospective review found an increase in incident cases of all food allergies in Olmsted County, Minnesota, over a 10-year period from 2002-2011. Males had a higher incidence of food allergy, and other coexistent atopic diseases were common. In particular, 324 of the 578 patients (56.1) with food allergy also had atopic dermatitis, consistent with findings of previous studies. The types of food allergies were similar between those delivered by cesarean section vs vaginally. In this cohort, 30 patients (5.2%) had associated penicillin/amoxicillin allergy. This is an area that is often overlooked and underaddressed by clinicians and may need further attention to address the potential role of antibiotic allergies in this population^18,19 and vice versa.

Our population-based study is unique because it was not based on registry data, which may underestimate or overestimate the true incidence of food allergies based on reporting bias. In addition, the availability of 10 years of data has made it possible for us to examine changes in the incidence of food allergies over time with changing recommendations from the AAP.

Interestingly, despite changes in AAP recommendations regarding the uncertainty surrounding the introduction of allergenic foods in 2008, food allergy incidence remained high in the subsequent years. This issue may be due to the delay in overall implementation of new recommendations by primary care practices or difficulties in changing established practices, especially when concrete evidence was lacking. Our study is strengthened by the unique REP medical records linkage system,²⁰ which allowed for accurate ascertainment of cases. This system helped minimize ascertainment and misclassification bias.

Our study has several limitations. We included only cases that were clinically diagnosed as food allergy because we were reliant on diagnosed cases. The number of undiagnosed cases is unknown because no population-based screen is available. With the severe, usually life-threatening clinical manifestations of food allergies, it is unlikely that these cases would not have come to medical attention; hence, the possibility of missing cases was relatively low.

In addition, the generalizability of the study findings is limited largely to white people because the Olmsted County population is mainly white (∼90%-95% during the study period). However, studies comparing various chronic diseases in Olmsted County with those in other communities in the United States indicate that data from this population can be extrapolated to a large part of the population of the country.¹⁴Finally, the use of a retrospective study design is subject to several biases, including reviewer bias. A reliability study was conducted in a smaller sample (25 patients) of the study patients. There was complete agreement between the 2 investigators (E.K.W. and A.Y.J.).

The incidence and prevalence of food allergy will remain a topic of interest as recommendations for the introduction of allergenic foods evolves. Although current recommendations for early introduction include only peanut, this situation may change in the future as we learn more about the etiology of milk and egg sensitization. We hope that with this population-based study, our understanding of food allergies will improve, which in turn will help guide avenues for primary and secondary prevention of food allergies.