Frequency, severity and causes of unexpected allergic reactions to food: a systematic literature review

Food allergic patients have to deal with an avoidance diet. Confusing labelling terms or precautionary labels can result in misinterpretation and risk‐taking behaviour. Even those patients that strictly adhere to their diet experience (sometimes severe) unexpected allergic reactions to food. The frequency, severity and causes of such reactions are unknown. The objective of this review was to describe the frequency, severity and causes of unexpected allergic reactions to food in food allergic patients aged > 12 years, in order to develop improved strategies to deal with their allergy. A systematic review was carried out by two researchers, in six electronic databases (CINAHL, Cochrane, EMBASE, Medline, Psychinfo and Scopus). The search was performed with keywords relating to the frequency, severity and causes of unexpected allergic reactions to food. This resulted in 24 studies which met the inclusion criteria; 18 observational and six qualitative studies. This review shows that knowledge about the frequency of unexpected reactions is limited. Peanut, nuts, egg, fruit/vegetables and milk are the main causal foods. Severe reactions and even fatalities occur. Most reactions take place at home, but a significant number also take place when eating at friends' houses or in restaurants. Labelling issues, but also attitude and risky behaviour of patients can attribute to unexpected reactions. We conclude that prospective studies are needed to get more insight in the frequency, severity, quantity of unintended allergen ingested and causes of unexpected allergic reactions to food, to be able to optimize strategies to support patients in dealing with their food allergy. Although the exact frequency is not known, unexpected reactions to food occur in a significant number of patients and can be severe. For clinical practice, this means that patient education and dietary instructions are necessary.


Introduction
The prevalence of food allergy is estimated to affect more than 2% and possibly up to 10% of the population [1]. At present, no curative treatment is available. This implies that patients can only avoid symptoms by strictly adhering to their avoidance diet. Patients are not always adequately advised how to deal with their diet, and confusing labelling terms and rampant use of multiple forms of precautionary labels (i.e. May Contain X, Produced in a facility that processes X) has resulted in risktaking behaviour by some allergic patients [2]. In daily practice, it is well-known that even those patients that strictly adhere to their diet experience unexpected allergic reactions to food during their life. Symptoms can be mild and limited to the oral cavity, but also generalized and severe allergic reactions can occur, sometimes involving multiple organ systems such as the skin and mucosal tissues and the gastrointestinal, respiratory and cardiovascular tract [3]. The frequency of unexpected allergic reactions to food is unknown. The overall prevalence rate of anaphylaxis is estimated to be 10 per 100 000 inhabitants per year and is primarily caused by food, drugs and insect venoms [4].
Appropriate labelling of food allergens is essential to help people manage their allergy and prevent food allergic reactions [5]. Although labelling of well-known allergens or primary food allergens is adequately regulated in many countries throughout the world, crosscontamination during food processing can cause the unintended presence of food allergens in a variety of packaged foods. Regulation of cross-contact of allergens is not specifically addressed under the mandates of food allergen labelling laws throughout the world [6,7]. The labelling laws specifically address the accurate labelling of major food allergens in terms that can easily be understood by consumers (i.e. casein would be labelled as 'milk' on the packaged food product label) when the allergenic source is used as a direct ingredient or processing aid in the food product [5]. Labelling laws are different in different countries [8]. In the European Union, wheat/cereals, eggs, milk, peanut, fish, crustaceans, soy, tree nuts, sesame, shellfish/molluscs, mustard, celery and lupine are considered major allergens. In the United States, these are wheat/cereals, eggs, milk, peanut, fish, crustaceans, soy and tree nuts [8]. Regulatory authorities require proper management of food allergens to minimize the chance of allergen cross-contamination in the processing facility by utilizing good manufacturing practices, including the development of a robust allergen control programme. Some manufacturers voluntarily use precautionary labelling to alert consumers to products that might be subject to such adventitious contamination [5]. However, the increased use of this type of labelling, along with the inconsistent and non-transparent way by which various companies decide to use these labels, has resulted in devaluation of its intended message to relay potential risk to allergic consumers. As a result, up to 40% of allergic individuals ignore these recommendations and taking risks by consuming these products [2,9]. These products have been shown to contain sporadic and varying levels of undeclared food allergens at levels that could result in adverse allergic reactions. It is highly recommended that allergic consumers avoid products bearing precautionary labels. However, it is unknown how often an unexpected allergic reaction occurs due to ignoring food labels or due to other causes. It is important to advise patients on how to recognize the first signs of an allergic reaction so that they can treat themselves and seek qualified medical attention [10,11].
The aim of this review was to summarize the current evidence about frequency, severity and causes of unexpected allergic reactions to food in food allergic patients.

Design
A systematic literature review was carried out, following the recommendations of the preferred reporting items for systematic reviews and meta-analyse statement [12].

Eligibility criteria
Studies which contributed to the aim and were published between January 2001 and April 2013 were included. Inclusion criteria were as follows: articles from peer reviewed journals that were written in English, German or Dutch, and participants of 12 years or older who had an indication/diagnosis of food allergy. In case of mixed populations of children and adults, the study was included when > 50% of the participants were at least 12 years of age. Children < 12 years of age were excluded, because in younger children parents or other caregivers take mainly responsibility for their food choice and a safe environment [13,14]. Around 12 years of age, children start to develop independence and take their own responsibility for managing their food allergy [15,16]. Case studies were excluded.

Information sources en search methods
The search was performed by two researchers independently in April 2013. The studies were identified by searches of six electronic databases (CINAHL, Cochrane, EMBASE, Medline, Psychinfo and Scopus), and by using the snowball method, by screening the reference lists of the included articles and through authors' knowledge about relevant studies.
The following keywords were used based on disease characteristics (e.g. type of food allergy) and possible determinants (e.g. attitude, labelling, place of reaction and allergens): (food allergy) AND (reactions OR anaphylaxis OR anaphylactic OR accidental OR reactions OR exposure OR ingestion OR eating OR labelling OR labelling OR (food labelling) OR (food labelling) OR restaurant OR (consumer attitudes) OR (food products) OR information OR (quality of life) OR kiss OR (soy OR soybean OR soya) OR milk OR egg OR crustacean OR shellfish OR fish OR lupin OR mustard OR celery OR molluscs OR peanut OR [sesame OR (sesame seeds)] OR [nuts OR almond OR hazelnut OR walnut OR (cashew nut) OR (pecan nut) OR [(brazil nut) OR (para nut)] OR (pistachio nut) OR [(macadamia nut) OR (queensland nut)] OR (kemiri nut)]). The limit title/abstract was used.
The articles were first screened for relevance to the stated study aims by reading the title and abstract. Of the articles that appeared to fit the criteria for the study after the primary review, the full text was then critically reviewed for relevance and quality. The entire process was performed independently by two researchers who then came to a consensus on the articles that fit the criteria for inclusion in the study.

Quality appraisal
The methodological quality of the included studies was evaluated to get insight in the methods and to assess the risk of bias of the studies. Because of the various designs, two different quality appraisal tools were used. The quality of observational studies (e.g. cohort, case-control or cross-sectional studies) was evaluated with the criteria of the 'strengthening the reporting of observational studies in epidemiology' (STROBE) statement [17]. The evaluation of qualitative studies (e.g. grounded theory practice or phenomenology and data collection by, e.g. interviewing, participant observation and focus groups) was performed with the 'consolidated criteria for reporting qualitative research' (COREQ) checklist [18].

Data abstraction and synthesis
The included articles were analysed by two researchers. The characteristics of the studies were recorded on a data extraction form, comprising of the following items: author and year; design, participants and setting; and results relating to the research question of interest: frequency, severity, causes and locations of unexpected reactions, labelling issues and attitude or behaviour related to unexpected reactions. The severity of unexpected reactions was classified according to an adapted version of the Mueller classification. Reactions with local symptoms (Mueller 0) were classified as mild, with skin and mucosal (Mueller 1) or gastro-intestinal symptoms (Mueller 2) as moderate and with respiratory (Mueller 3) or cardiovascular symptoms (Mueller 4) as severe [19,20].
In studies with mixed populations (adults and children), we only reported results of patients of at least 12 years of age. Since different designs and outcomes were used in the study, it was not possible to pool the data in a meta-analysis. Therefore, the findings were reported by using a narrative summary technique.

Quality of the studies and potential risk on bias
The observational studies, assessed with STROBE [17], had an average score of 17 out of 22 items, with a range of 13-21. Most articles clearly described the rationale and objective of the study, the study population, data measurements and outcomes. Less than half of the studies reported about potential sources of bias      Problems related to readability and interpretability of ingredients lists No allergy advice box could be incorrectly considered to be a signal that there was nothing to worry about Some avoided products where the product name raised concern, even if the ingredients list/allergy advice box indicated the product to be safe Risk assessment is based on product brand or name, prior experiences with the product; if this strategy does not lead to a confident decision, printed packet information is used  12% reported that dining hall food was always labelled to identify allergens. 37% reported main course alternative was available 33% of the group with anaphylactic symptoms was concerned about potential exposure 40% always avoided ingesting the food items of concern 60% did not always avoid the food items of concern. A significantly higher number of students without symptoms of anaphylaxis continued eating a food to which they identified to be allergic to (continued)   in the methods section, how the study size was arranged, reasons for non-participation and funding. Seven studies had a self-report questionnaire [15,23,[26][27][28][29]33] (Table 3). The qualitative studies, assessed with COREQ [18], scored an average of 16 of the 33 items of interest, with a range of 11 to 21. Most articles described the sample, data collection and outcomes. Frequent limitation concerns were lack of information about the personal characteristics of the interviewers and their relationship to the participants. Only three studies gave a minimal description about this domain [39,41,42]. None of the studies reported about field notes or carried out repeat interviews to get participant feedback on the finding. Two study reported about non-participation [39,42]. Only one study reported about the methodological orientation of the study [41] ( Table 4).
Overall, the methodological quality of the observational and qualitative studies was moderate. The moderate methodological quality and differences in designs and outcomes used in the included studies leads to potential risk of bias. Possible selection bias could not be determined when reasons for non-participating and study size were not reported. The use of self-reported questionnaires could have led to information bias or less accurate data.

Frequency and severity of unexpected reactions
Little is known about the frequency of unexpected reactions in patients with a known food allergy and all studies describe different parameters, making it difficult to elucidate the frequency among different populations. Añ ıbarro [24] studied unexpected reactions due to hidden allergens in a sample of 436 patients over a 5-year period and showed a mean of 1.98 (range 1-10) unexpected reactions per patient. Comstock [25] reported about allergic reactions aboard airliners. Among 471 patients with peanut, tree nut or seed allergy, 9% reported reactions during a flight, of which 10% had more than one reaction [25]. Kanny [28] and Sicherer [34] reported that 53-58% had recurrent reactions, whereof 12-20% had more than six recurrent reactions. Kalogeromitros [22] studied patients with allergy to grape. Among 11 patients, all had recurrent reaction, but none more than six reactions. Kalogeromitros [27] reported that 75% had a recurrent reaction, after re-exposure to a suspected allergic food.
Four studies [30,[35][36][37] reported about fatalities due to food allergy. Bock described in 2001 and 2007, 26 and 23 fatal reactions, respectively, in patients older than 12 years of age over a 5 years period [35,36]. Liew [30] described 3 fatal reactions in patients of older than 10 years of age over a 9-year period. Malmheden [37] reported 6 fatal reactions; three of which were older than 15 years.
In most studies, unexpected reactions occurred because patients ingested food containing the allergen. However, Comstock [25] and Eriksson [33] also reported unexpected reactions by inhalation or indirect contact as kissing.

Labelling issues related to unexpected allergic reactions
Unexpected reactions could also be related to the way of food labelling. Labelling issues can be divided into three categories, namely readability, clarity and interpretation of labels [29,32,38,43]. Readability problems, like the use of difficult words or terms or a number of languages without the national language of the consumer, were mentioned by 40% [32] to 70% [29] of the patients [43]. Clarity issues were due to the layout of the ingredients list or allergy advice box, the amount of information and the changes of recipes [32,43] were reported by 30% [32] of the patients.

Attitude and behaviour and unexpected allergic reactions
Attitude and behaviour are clearly related to the occurrence of unexpected reactions in patients with a known food allergy. This seems to be influenced by the severity of symptoms (Table 2).
According to Barnett [38], patients estimate the risk of a product based on product brand or name and prior experiences with that product or brand. When this strategy does not lead to a confident decision, printed packet information is used. Patients with severe allergic reactions seemed to take less risk than patients with less severe or inconsistent symptoms [26,43]. According to L€ ammel [29], 82% of allergic individuals surveyed would not buy a product that may contain the allergen; however, it was not studied whether severity of earlier reactions played a role. Greenhawt [26] reported that 60% of allergic individuals surveyed did not always avoid a product with the allergen, but a higher proportion of students (57%) with less severe symptoms indicated eating food containing the allergen vs. 42% of students with a prior history of anaphylaxis.
Four studies specifically addressed attitude and risktaking behaviour of food allergic individuals during puberty and adolescence (11-21 years), the majority of which had a peanut or tree nut allergy [15,39,41,42]. These studies reported that teenagers and adolescents take risks with precautionary labelled food [15,39,41,42]. Mackenzie [41] and Sampson [15] showed that not more than 43% to 64% always read labels, were very strict in managing their food allergy and avoided food with precautionary labelling; thereby, severity of symptoms and the risk of an unexpected reaction seem to play a role. In another study, it was reported that more than 75% of allergic individuals surveyed consumed food with 'may contain' labelling [42]. Gallager [39] reported that all adolescents attempt to avoid allergen by checking the label. Reasons for risk-taking behaviour among students [26] and teenagers [15] can be summarized as follows: no previous severe symptoms or inconsistent symptoms, the possibility to self-treat reactions, lack of concern and social circumstances.
In managing their risks, patients with peanut or tree nut allergy are balancing between communication and fear of potential social embarrassment to disclosure of their allergic status [40]. Patients restricted themselves in terms of eating out [39,40] and foreign travel [39]. Patients avoided particular restaurants that were considered to pose a high risk through self-evident and hidden presence of nuts, and sought familiarity to reduce uncertainty and risks. In addition, availability of medical care nearby played a role in their decision making [40]. Teenagers felt that other people could not be trusted to take precautions on their behalf [41].

Discussion
Most food allergic patients are confronted with unexpected allergic reactions to food. Seven studies investigated the frequency of unexpected reactions, but results differ because studies used different time periods and figures were reported in different ways. Peanuts, nuts, vegetables/fruit, egg and milk are the main causal foods. Severe reactions and even fatalities occur. Many reactions occur in patients with a well-known food allergy and with products known to be at risk. A major factor attributing to the risk of unexpected allergic reactions is the eating location. Most reactions take place at home, but a significant number also take place when eating at friends' houses or in restaurants. But also labelling issues, and attitude and risky behaviour of patients play an important role.
This systematic review gives a state-of-the art overview of current knowledge of unexpected reactions to food using an extensive search strategy in six relevant databases. However, generalizability of the results is limited because of the heterogeneity of study populations and methodology and potential bias due to moderate methodological quality. However, we still feel that it is possible to draw important conclusions. Some studies included participants with self-reported food allergy [15, 25-27, 29, 32, 33, 39, 43], whereas other studies only included people with a doctor-diagnosed allergy [21-24, 38, 40-42, 44]. It is known that the selfreported rate of food allergy varies from 1% to 35% [45] in contrast to 3% after doctor diagnosis using the double-blind food challenge which is the gold standard [46]. Inclusion of patients with self-reported food allergies as well as lack of information about non-responders and lack of information about sample size calculation could have led to selection bias in individual studies. This could therefore result in a lower generalizability of our outcomes for the population of patients with a doctor-diagnosed food allergy. It was not possible to analyse if the frequency and severity of reactions differs between studies of patients with selfreport vs. doctors-diagnosed food allergy, because in none of these studies all patients were diagnosed by food challenge.
The included studies used different ways of classification of severity. To make it easier to compare the results, we classified the severity of the reactions with the adapted version of the Mueller classification [19,20]. However, it remained hard to compare the results. A third limitation is the lack of prospective studies with a quantitative design. Most studies used a retrospective or cross-sectional design with self-reported data. It is often hard for patients to remember the exact frequency and severity of their allergic reactions. This increased the risk of recall bias and information bias.
Figures about the frequency of unexpected reactions to food were reported in different ways. Some authors described reactions on specific foods [22,25,34], sometimes in a specific environment, like during flight [25] and others only described fatal reactions [30,35,36]. Besides, some studies reported about the actual percentage of patients with recurrent reactions [25,27,34], while others reported the number of unexpected reactions with sometimes the percentage of patients [22,24,28,35,36]. For example, Sicherer [34] reported that 58% had a recurrent reaction on sea food, whereas Añ ıbarro [26] showed a frequency of two reactions per patient over 5 years. Therefore, it is not possible to make a very precise conclusion about the frequency of unexpected reactions. Nevertheless, these data clearly illustrate that unexpected reactions are a significant problem.
The severity of unexpected reactions varied between the included studies. However, it can be concluded that severe reactions do occur frequently; varying from 16% to 96% of the reactions (Table 1) and even fatalities are reported. The one study suggesting that 96% of unexpected reactions are severe [22], included only eleven patients with doctor-diagnosed reactions, which is likely a significant over-estimation of severe reactions due to the limited sample size and selection bias. It might be that there is an underestimation of mild reactions, which might be easier forgotten than severe reactions, increasing the risk of recall bias.
Peanuts, nuts, egg and milk seem to be the most frequent cause of unexpected reactions, which is probably due to the frequency of these food allergies and the fact that these allergens are frequently used in prepared foods [47]. Moreover, cross-contamination during production of packaged food products can occasionally occur [47].
Unexpected reactions occur at home regularly, for example, due to errors in preparation, cross-contamination [48] and a variety of problems in readability and clarity of food labelling. Information on food labels is overwhelming [43], products with advisory labels with tree nut frequently do not specify the type of tree nut, and non-specific terms (such as spices and natural flavours) are frequently used but are not linked to an allergen or ingredient, in case they do not belong to the 14 allergens to be labelled [49]. Besides the compulsory information about allergens on labels, 17-65% of all manufactured food products, contains precautionary labels [49,50]. Chocolate, candy and cookies have such a description on more than 50% of the labels [49]. But only 10% of the packaged products [2] and 25% of the packaged cookies/chocolates [51] with a precautionary statement about peanuts had a detectable level of this allergen. Such an unwanted restriction of food choices can seriously affect the quality of life [50]. In addition, products can contain undeclared allergen concentration reaching levels that trigger allergic reactions [52].
Unexpected reactions often occur in restaurants. This is not surprising, because restaurant personnel deal with a large number of different and potentially allergenic foods in the same facility which could lead to crosscontamination during food preparation [53]. Moreover, there are deficits in their knowledge about how to provide safe meals to allergic consumers [54].
Attitude also influences the risk of unexpected reactions. Not everyone is as strict in avoiding allergens [15,26,29,41,42]. The percentage of patient with risk-taking behaviour differs between the included studies, and to some extent it depends on disease severity [15,43], the possibility to self-treat reactions or patients' estimation if it is a risky action [26,39]. Allergic individuals regularly feel embarrassed about their allergy [40]. Therefore, they sometimes prefer to take risks instead of asking the restaurant or retail staff about the presence of allergens in food [40]. Teenagers and adolescents might show risk-taking behaviour more often than adults [41,43]. This is consistent with two studies which reported that teenagers are at the greatest risk of food-induced anaphylaxis [36,55]; Bock [36] showed that among 32 fatalities, 69% occurred in patients aged between 13 and 21 years. It is reported that adolescence is a period of heightened vulnerability to risk-taking behaviour because of a disjunction between novelty and sensation seeking (both of which increase dramatically at puberty) and the development of self-regulatory competence (which does not fully mature until early adulthood) [56].
To conclude, the exact frequency is not known but unexpected reactions to food occur in a significant number of patients and can be severe and even fatal. Major factors contributing to the risk of unexpected reactions are eating location, labelling issues and attitude and risk-taking behaviour of patients.
For clinical practice, this means that patient education about the risks of unexpected reactions, emergency medication and instructions when and how to use it are essential and preferably should be tailored to the specific age groups. Moreover, dietary instructions on how to read labels and how to deal with may contain labelling are necessary. Further, attention to develop a more transparent way of precautionary labelling is needed. In Australia and New Zealand, the Vital (Voluntary Incidental Trace Allergen Labelling) system is in use which is a promising approach to address this issue [57]. Prospective studies are needed to get more insight in the frequency, severity, quantity of unintended allergen ingested and causes of unexpected allergic reactions to food, to be able to optimize strategies to support patients in dealing with their food allergy; to prevent unexpected reactions as much as possible and to increase awareness and knowledge in food industry and among retail and restaurant staff.