Frequency of food allergy in Europe: An updated systematic review and meta‐analysis

Food allergy (FA) is increasingly reported in Europe, however, the latest prevalence estimates were based on studies published a decade ago. The present work provides the most updated estimates of the prevalence and trends of FA in Europe. Databases were searched for studies published between 2012 and 2021, added to studies published up to 2012. In total, 110 studies were included in this update. Most studies were graded as moderate risk of bias. Pooled lifetime and point prevalence of self‐reported FA were 19.9% (95% CI 16.6–23.3) and 13.1% (95% CI 11.3–14.8), respectively. The point prevalence of sensitization based on specific IgE (slgE) was 16.6% (95% CI 12.3–20.8), skin prick test (SPT) 5.7% (95% CI 3.9–7.4), and positive food challenge 0.8% (95% CI 0.5–0.9). While lifetime prevalence of self‐reported FA and food challenge positivity only slightly changed, the point prevalence of self‐reported FA, sIgE and SPT positivity increased from previous estimates. This may reflect a real increase, increased awareness, increased number of foods assessed, or increased number of studies from countries with less data in the first review. Future studies require rigorous designs and implementation of standardized methodology in diagnosing FA, including use of double‐blinded placebo‐controlled food challenge to minimize potential biases.


| INTRODUC TI ON
The frequency of food allergy (FA) in Europe has been increasingly reported over the past decades. However, the data supporting an increase are mainly anecdotal, considering that the latest systematic report on FA epidemiology was published by the European Academy of Allergy and Clinical Immunology (EAACI) in 2014 based on the articles published between 2000 and 2012. 1,2 That report provided a detailed overview of the epidemiology of FA, including estimates of the incidence, prevalence, and time trends of any FA, as well as the so-called eight big foods, i.e., cow's milk, egg, wheat, soy, peanut, tree nuts, fish, and shellfish.
It is now 10 years since the EAACI-commissioned systematic review was completed. Several studies have been published since then, indicating that an update of the previous review is now warranted. By bringing together the evidence generated from the previous systematic review together with more recent studies, we have the opportunity to make clearer estimates of the incidence, prevalence, and time trends of FA in Europe. The update will also give greater opportunity to estimate the epidemiological burden of FA across various population subgroups (e.g., age and regions). Furthermore, this update is an excellent opportunity to identify and estimate the epidemiological burden of potentially "new" and "emerging" food allergy in Europe, beyond the so-called eight big foods. The aim of the current work was to update the previously EAACI-commissioned systematic review on the incidence, prevalence, and time trends of FA in Europe by identifying, critically appraising, and synthesizing evidence from studies now published since the previous systematic review was completed (2012). The current article reports on the estimates of the frequency of any FA.

| Protocol registration
The protocol for this review was registered with the International Prospective Register of Systematic Reviews (PROSPERO; reference CRD42021266657) prior to undertaking this review.

| Search strategy
The search strategy was adapted from the previously published EAACI review. The two concepts of FA and epidemiology were combined to identify all relevant literature (including both articles, conference abstracts or posters, and theses) from the electronic databases. Six databases were searched: MEDLINE, EMBASE, CINAHL, Web of Science, Cochrane Library, and Scopus. Compared with the EAACI review from 2014, Cochrane Library and Scopus were added among the databases in acknowledgement of the advancements that have occurred in the indexing of studies on the topic since the first review was done. A few more keywords were also included in the current review to ensure that it was updated with all the new and emerging keywords on FA. Experts on the topic of FA were consulted to ensure that the study identification procedure did not miss any relevant work. No language restrictions were applied in the database searches. When possible, studies published in languages different from English were translated by researchers fluent in the language to permit data extraction. When it was not possible to translate the article, but an English abstract was available, data extraction from the abstract was performed. The few cases for which neither abstract nor full article data extraction was possible have been reported. Detailed description of the search strategies employed are available in Box S1 of the Supporting Information section of the online version of this article.

| Inclusion and exclusion criteria
The studies included in the current review comprised studies published from January 1, 2000 to June 30, 2021 (i.e., studies published in the previous systematic review, and studies identified in the current update). All studies that examined subjects with suspected FA, of any age and gender, and of any European country as defined by the United Nations (see Appendix 1) were considered eligible.
Studies from Greenland and Turkey were also included, similarly to what was done in the previous EAACI review. The following types of studies were considered for inclusion: systematic reviews and meta-analyses, prospective and retrospective cohort studies, crosssectional studies, case-control studies, clinical trials, and routine sIgE and SPT positivity increased from previous estimates. This may reflect a real increase, increased awareness, increased number of foods assessed, or increased number of studies from countries with less data in the first review. Future studies require rigorous designs and implementation of standardized methodology in diagnosing FA, including use of double-blinded placebo-controlled food challenge to minimize potential biases.

K E Y W O R D S
epidemiology, Europe, food allergy, sensitization, systematic review healthcare studies. Expert reviews or other reviews that are not systematic reviews, discussion papers, non-research letters and editorials, qualitative studies, case studies, case series, and animal studies were excluded from the present work.

| Study selection
All records obtained from the databases searches were exported to EndNote 20 (Clarivate Analytics, 2020) for de-duplication. Following this, all relevant articles were exported to Rayyan (https://rayyan. ai) for titles and abstracts screening, and to manage all the retrieved records. Titles and abstracts screening was performed by four independent reviewers (SN/GS and YA/MA), working in pairs. Disagreements between reviewers were resolved with consensus when possible or by consultation with the project PI (BN). The full texts of the potentially eligible studies were then assessed by the same four independent reviewers, similarly to what was done for the titles and abstracts selection. The Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) flow diagram was used to document the screening process.

| Risk of bias assessment
Risk of bias in individual studies was independently assessed for each study by the four reviewers (SN/GS and YA/MA), working in pairs, by employing the Critical Appraisal Skills Programme (CASP; http://www.casp-uk.net) quality assessment tool. The same tool was also used in the previous EAACI systematic review. Accordingly, all studies were assigned an overall rating, along with a separate rating for each different components of the individual studies (i.e., appropriateness of the study design for the research question, risk of selection bias, exposure measurement, and outcome assessment).
Any discrepancy was resolved by consensus or arbitrated by the project PI (BN).

| Data extraction
Data were collected from included studies using a customized data extraction form. All data extracted were reported in a standardized and reproducible fashion. The developed form was first piloted with a small number of included studies and approved by all reviewers before it was employed to extract data from all studies. The form was stored on a Google Drive (Alphabet Inc., Mountain View, CA, USA) repository online to make all the information promptly available to all reviewers. Similar to the selection process, the extraction of all data was performed by four reviewers, working in pairs (SN/GS and YA/MA). Each pair conducted an independent extraction of the assigned records. After crosschecking, all disagreement were addressed and further arbitrated by the project PI (BN).

| Data analysis, synthesis, and reporting
We recalculated all the frequency estimates of FA occurrence if adequate data were provided by authors. If any discrepancies were observed between our recalculated estimates and those of the authors, we reported our recalculated estimates. Our recalculated estimates were based on minimal measured events rather than the extrapolated ones. The 95% confidence intervals (95% CI) were obtained by employing the Wilson score method without continuity correction. 3 Heterogeneity was assessed using I 2 statistics. For studies that presented missing data, thus not allowing estimates recalculation, we reported the estimates provided by the authors. Where needed and possible, we contacted authors of primary studies for clarifications.
Countries outside the Organization for Economic Co-operation and Development (OECD) definition of Europe were included in the systematic review but were not included in meta-analysis, similarly to what was done in the previous version of the systematic review and meta-analysis. An exception was made for Lithuania and Russia, which had recorded FA data for meta-analysis also in the previous study. According to the criteria above, Albania, Bulgaria, Croatia, and Ukraine were included in the systematic review, but not in the current meta-analysis.
Random-effects meta-analysis was performed for all studies that provided numerical data in order to derive pooled estimates across studies. The meta-analysis was conducted using the soft- those obtained from the current updated searches. As was done in the previous EAACI review, data were also stratified by age category, in children (0-17 years) and adults (18 years and over), and by European region (Northern-Eastern-Southern-Western Europe) following the classification by the United Nations (see Appendix 1). In case of overlap between the two age categories, or between groups, the estimate was included in either age group if the age distribution was skewed to that age group, following the approach used in the 2014 EAACI review. An exception was made for the United Kingdom, which was assigned to Western Europe instead of Northern Europe, as was done in the previous EAACI review. In the meta-analysis, we estimated the updated prevalence of FA for the period 2000-2021. In addition, we also performed and reported meta-analysis separately for the studies published during 2012-2021, which were compared with the estimate obtained in the previous review for the period 2000-2012.

| Study selection and characteristics
The study selection and screening process of the current update are illustrated in the PRISMA flow chart presented in Figure 1. A total of 38,903 records were retrieved from the databases searched. After de-duplication, 33,875 records were selected for screening. Based on titles and abstracts, 33,625 records were excluded due to being clearly ineligible or not fulfilling the inclusion criteria. Out of the remaining 250 records, two full-text articles could not be retrieved, and the abstracts did not include any relevant information. They were therefore excluded. Of the remaining 248 records, 72 reports were included in this review. The new reports included were based on 54 newly identified studies, and on one study already included in the previous review but presenting updated data for the cohort enrolled in the study. Putting together the number of reports (and studies) included in the first systematic review from EAACI, with the F I G U R E 1 PRISMA flow diagram for updated systematic review on prevalence of food allergy in Europe, 2000-2021.

| Risk of bias assessment
Overall, the risk of bias assessment for individual studies graded by the CASP quality assessment tool indicated that most of the studies had a moderate risk of bias (91 out of 110 studies). Table S2 summarizes the grading of the main CASP quality assessment features for all studies.

| Frequency of any FA
The ranges of estimates for any FA categorized by age groups, and by different methods of assessment are presented in Table 1 Figure 6).

| Time trends of frequency of FA
Data on time trends of FA in Europe are reported in Table 2. In addition to the three studies originally reported in the previous systematic review, two more studies were identified and included in this update, giving a total of five studies. 29,[55][56][57]65,76,[136][137][138]141 Four out of the five studies were undertaken in the United Kingdom, 29,65,76,[136][137][138]141 while the remaining one took place in Finland. 65 We also added updated trends on the frequency of FA for one of the studies already reported in the review from 2014. [136][137][138]141 Three of the studies reported trends for both any FA and specific FA. One study reported the trends of hospital admission rate for FA, 55-57 while one study reported trends for doctor-diagnosed peanut allergy. 76 While the available data remained limited to allow clear conclusion on the current time trends in the incidence or prevalence of FA, the additional data from this updated review may suggest a slight but progressive increase of clinician-diagnosed FA in the United Kingdom, with reported prevalence going from 0.6% in 2000 to 1.3% in 2015, 29   Point prevalence of physician-diagnosed FA was also measured for the following specific foods: nuts and eggs.  Children attending the first year of elementary school at 29 different schools in the Tampere (Finland) district were screened for the study. The objective was to assess the prevalence of self-reported FA in the Tampere district.
Basic foods according to the authors' definition include milk, eggs, and grains. Point prevalence for the following specific foods was also measured: cow milk, eggs, grain, nuts, fruits and vegetables, and fish.

Kotz et al. 2011, United Kingdom 76
All ages Lifetime prevalence physician diagnosed peanut allergy per 1000 patients: All estimates were age-and sex-standardized. During the study period, while the lifetime prevalence of peanut allergy doubled, the incidence rate of peanut allergy remained fairly stable. Sex-specific, age-specific, and SES-specific estimates are also reported in the table.
Only data regarding the prevalence trends of peanut allergy were reported by the authors.
Year The data presented come from three different birth cohorts of children, which were born in the Isle of Wight 1989Wight -1990Wight , 1994Wight -1996Wight , and 2001Wight -2002 Year 1990-1991 Self-reported FA: any FA: 8.5%

| Strengths, limitations, and implications of the current review update
As with the previous EAACI-led systematic review 1,2 we followed recommended rigor in undertaking this updated review, which in- As observed in the previous review, 1 most studies in this update also failed to make any distinction between IgE or non-IgE FA phenotypes, thus it was impossible to present estimates of prevalence of FA by its IgE-mediated and non-IgE-mediated phenotypes.
Overall, the quality of studies included in the review remained moderate as it was in the previous review, indicating that in going forward, the quality of the underlying evidence needs to be improved. with Southern and Western regions. As there seemed to be important methodological and diagnostic differences within and across the European regions, interpretation of the findings requires caution, especially considering the high heterogeneity among the studies still observed in this updated review.
Overall, there was no improvement in the design of studies and diagnostic approaches used between the current update and the previous review. There is still a need to improve this evidence base in order to better understand the frequency of FA across Europe, through which its healthcare and societal burden can be clearer explained. As indicated in the previous review, future studies still require rigorous designs and implementation of standardized methodology in diagnosing FA, including use of DBPCFC to minimize potential biases.

ACK N OWLED G EM ENTS
The study was funded through an unrestricted grant from DBV