Study design and participants
In this cross-sectional study, adolescents aged 13–17 years with paediatric allergist-diagnosed allergy to one or more staple foods (cow’s milk, hen’s egg and/or wheat) were identified from medical records and recruited in 2010–2012 by a paediatric nurse from the outpatient allergy clinic at Sachs’ Children and Youth Hospital, Södersjukhuset, in Stockholm, Sweden.
Inclusion criteria were a convincing history of allergy to one or more of the above-mentioned staple foods ascertained either by a positive food challenge with evident symptoms, or by levels of food specific Immunoglobulin E (IgE) antibodies levels associated with a 95 % probability of a positive result in a double-blind placebo controlled food challenge . Exclusion criteria were an unclear allergy diagnosis to staple food(s), poor understanding of the Swedish language, or presence of coeliac disease, diabetes and/or a malignancy. Information on concomitant allergic disease (asthma, allergic rhinitis, allergic conjunctivitis, eczema) was also obtained. A total of 87 adolescents were eligible and invited to participate. These adolescents were mailed the FAQLQ-TF (described below, English version available as an Additional file 1), as well as an information letter and a postage-paid return envelope. Parents were mailed an information letter and consent form. Completed FAQLQ-TF and signed parental consent forms were received from 58 adolescents (67 % of those eligible; Fig. 1). Adolescents received two movie tickets following receipt of completed questionnaires. This study was approved by the Regional Ethical Review Board in Stockholm, Sweden (Dnr 2009/84-31/5). Personal data were treated according to the Swedish Personal Data Act.
Both the number of staple food allergies, as well as the number of offending foods (at least one staple food allergy and, participant-reported allergies to other foods) were considered as exposures.
Adolescents responded to 36 closed-ended questions on food allergy symptoms, from which we generated specific symptoms:
Gastrointestinal: stomach upset; vomiting; diarrhoea.
Oral: itchy tongue, mouth or lips; swollen tongue or lips.
Upper respiratory: runny or blocked nose; sneezing.
Lower respiratory: itchy or tight throat; difficulty swallowing; shortness of breath; wheeze; cough.
Cardiovascular/neurological: dizziness; tachycardia; blurred vision; inability to stand/collapse; loss of consciousness.
The most severe symptoms, including difficulty breathing, inability to stand/collapse, and/or loss of consciousness, involved the respiratory- and/or cardiovascular/neurological systems. In keeping with our previous publications on children  and adults , and approximating as best as possible the criteria outlined by Sampson et al. , such symptoms are collectively referred to as anaphylaxis. Adolescents were asked if they had been prescribed an adrenaline auto injector (AAI).
Food allergy quality of life questionnaire-teenager form (FAQLQ-TF)
The FAQLQ-TF  was translated into Swedish as per World Health Organization guidelines , and was piloted in 10 Swedish-speaking adolescents to ascertain comprehension. Following minor linguistic adjustments, the translation was deemed adequate. The FAQLQ-TF contains 28 questions on HRQL, each of which has corresponding closed-ended answers on a 7-point scale where 1 is best HRQL and 7 is worst HRQL . Overall HRQL established by taking the mean of the 28 questions. These questions were also designed to address three domains: allergen avoidance and dietary restrictions (AADR), emotional impact (EI) and risk of accidental exposure (RAE). The first domain, AADR, describes adolescents’ perceptions of limitations, hesitations and refusals of foods that they purchase or are offered in social situations. The second domain, EI, reflects adolescents’ fears of an allergic reaction or accidental consumption of the food(s) to which they are allergic and their disappointment when others do not take their food allergy seriously. The third domain, RAE, captures adolescents’ assessments of needing to be cautious about purchasing food or eating out in relation to changes in ingredients, incorrect disclosure of ingredients and touching certain foods.
Floor and ceiling effects (percentages of patients with minimal and maximum scores, respectively) of the FAQLQ-TF were calculated to verify discriminative capacity. These effects were considered present if >15 % of a sample of a minimum of 50 individuals achieved the lowest or highest possible scores, respectively. Absence of these effects demonstrates the efficacy of the questionnaire.
Descriptive statistics included sample sizes (n), percentages, means, parametric two-sample t-tests and 95 % CI. Statistical significance was set at p < 0.05. Overall and domain-specific HRQL scores were calculated for the entire study population and stratified by gender. To permit statistical comparisons, the number of staple food allergies was classified into 2 dichotomous groups: 1 vs. 2–3. The number of offending foods was classified into 4 groups: 1, 2, 3 or >3. As described above, adolescents reported on symptoms. Adolescents may forget or inaccurately report their symptoms. Thus, we performed intra-class correlations of adolescent-reported symptoms with those reported by their parents as part of a parallel study  to measure reliability. These analyses showed modest correlations between adolescent- and parent-reported symptoms, with increasing reliability with increasingly severe symptoms (results not shown). As such, we present the results herein based on adolescent-reported symptoms.
Univariable and multivariable linear regression analyses were performed to identify predictors of HRQL. Potential covariates were identified based on prior knowledge of the exposures and outcome. The covariates gender, number of symptoms, history of anaphylaxis, AAI prescription and concomitant allergic disease were included in the final model as they statistically and independently altered the prediction model. The same models were used for overall and domain-specific HRQL. In keeping with previous publications on HRQL assessed via the FAQLQ, a score of ≥ ±0.5 was considered to be clinically relevant [9, 16]. Analysis was performed with STATA Statistical Software (release 13.1; StataCorp, College Station, Texas, USA).