Diagnostic utility of allergy tests to predict baked egg and lightly cooked egg allergies compared to double‐blind placebo‐controlled food challenges

Abstract Background Double‐blind placebo‐controlled food challenges (DBPCFC) are the gold‐standard to diagnose food allergy. However, they can cause allergic reactions of unpredictable severity. We assessed accuracy of current and new diagnostic tests compared to DBPCFC to baked egg (BE) and to lightly cooked egg (LCE). Methods Children aged 6 months to 15 years were assessed for possible egg allergy as part of the BAT2 study (NCT03309488). They underwent clinical assessment, skin prick test (SPT), specific IgE (sIgE) and basophil activation test (BAT). The results of the tests were compared with DBPCFC outcomes to both BE and LCE. Results A total of 150 children underwent DBPCFC to BE, 60 (40%) reacted to and 85 (57%) tolerated BE and 5 (3%) had inconclusive oral food challenges (OFC). Seventy‐seven children tolerant to BE had DBPCFC to LCE and 16 reacted. The test within each modality with the best diagnostic performance for BE allergy was as follows: SPT to egg white (EW) (AUC = 0.726), sIgE to EW (AUC = 0.776) and BAT to egg (AUC = 0.783). BAT (AUC = 0.867) was the best test in the younger than 2 years age group. Applying 100% sensitivity and 100% specificity cut‐offs, followed by OFC, resulted in 100% diagnostic accuracy. BAT enabled the greatest reduction in OFC (41%). Using sIgE followed by BAT allowed to reduce the number of BATs performed by about 30% without significantly increasing the number of OFC. Conclusions The best diagnostic test was BAT to egg in terms of diagnostic accuracy and reduction in number of OFC. Using sIgE to EW followed by BAT required fewer BATs with sustained OFC reduction and diagnostic accuracy.


| INTRODUC TI ON
Egg allergy is one of the most common food allergies in childhood. 1,2 Most egg allergic children tolerate baked egg (BE) and resolve their egg allergy spontaneously over time. [3][4][5] However, at the time of diagnosis, most children with suspected egg allergy are either avoiding BE strictly or eating small amounts, and this leaves uncertainty in terms of the best advice to give about consumption of BE, namely the amount likely to be tolerated and whether the introduction of BE needs to be done as an oral food challenge (OFC) in the hospital setting or can be done at home in the community. 6 Previous studies have shown that allergy tests currently used in clinical practice, such as skin prick test (SPT) and specific IgE (sIgE) to allergen extracts, are not very precise predictors of the allergic status to BE, with sensitivity and specificity being 54/68% and 80/74% respectively, according to a recent meta-analysis. 7 Allergen components, namely egg major allergen ovomucoid (OVM), showed to be more informative than sIgE to egg extract to identify patients with persistent and BE allergy in some studies but not others. 8,9 Doubleblind placebo-controlled food challenges (DBPCFC) remain the gold-standard to diagnose BE allergy; however, they are resourceintensive and can cause allergic reactions of unpredictable severity.
Thus, improved tests are needed to enable an accurate diagnosis and reduce the number of referrals for OFC.
We aimed to assess accuracy of new diagnostic tests, namely the basophil activation test (BAT), as well as current tests, such as SPT and sIgE, compared to DBPCFC to BE and to lightly cooked egg (LCE). The BAT assesses the expression of activation marker CD63 on the surface of blood basophils by flow cytometry following stimulation with the allergen or controls. 10 Our hypothesis was that the BAT, being a functional test, was more accurate than tests that measure the presence of allergen-sIgE to confirm the diagnosis of BE and LCE allergies. This hypothesis follows on from our previous studies on peanut allergy, 10,11 where BAT showed to be a superior diagnostic test and to reflect the function of IgE in its ability to induce basophil activation following stimulation with the allergen and therefore to induce allergic symptoms.

| Study design
The BAT 2 Study (NCT03309488) was a cross-sectional diagnostic study in which children with suspected IgE-mediated allergy to either cow's milk, egg, sesame or cashew nut were prospectively recruited from specialised tertiary Paediatric Allergy clinics in London to undergo a diagnostic work-up to confirm or refute the diagnosis.
The present manuscript reports results for the egg study according to the STARD guidelines. Participants were referred by many different clinicians working in Paediatric Allergy specialist clinics at the Evelina London Children's Hospital or private clinics in London, and a screening telephone visit was undertaken by the study team to confirm eligibility. Children assessed for possible egg allergy were submitted to DBPCFC to BE and, if they passed this, to DBPCFC to LCE. Participants' parents or carers completed food frequency questionnaires (FFQ) and 7-day food diaries, and children underwent clinical evaluation, SPT, sIgE testing to allergen extracts and to allergen components, BAT and OFC. All study participants had DBPCFC, except for infants younger than 12 months who had open OFC. After completion of the DBPCFC, participants' parents or carers filled in FFQ to egg every 2 months for 2 years.

| Participants
Children aged 6 months or more and younger than 16 years at the time of consent were prospectively and sequentially recruited.

G R A P H I C A L A B S T R A C T
All participants had DBPCFC to baked egg and, if this was negative, also DBPCFC to lightly cooked egg; SPT, sIgE and BAT were performed in parallel. The test with the best diagnostic performance was BAT, followed by sIgE and SPT. Applying 100% sensitivity and 100% specificity cut-offs allowed 100% accuracy. Using tests sequentially reduced the number of OFCs (and BATs). Abbreviations: BAT, basophil activation test; DBPCFC, double-blind placebo-controlled food challenge; EW, egg white; NCT, national clinical trial; OVA, ovalbumin; OVM, ovomucoid; OFC, oral food challenge; OFC+, positive oral food challenge; SPT, skin prick test; sIgE, specific IgE Suspected IgE-mediated egg allergy was defined by a history of an immediate-type allergic reaction to any form of egg, or the presence of sIgE to egg as documented by SPT greater or equal to 1 mm and/or serum sIgE greater or equal to 0.10 KU/L. Infants and children with no history of regular consumption of an age-appropriate amount of egg were included as their allergic status to egg remained unclear in the absence of uneventful egg consumption. Children tolerating small amounts of BE were advised to avoid BE for at least 2 days prior to blood collection for BAT and sIgE. Exclusion criteria are listed in the Supporting Information. The study was approved by the London -Westminster Research Ethics Committee (reference 17/ LO/0296) and the UK Health Research Authority. Informed consent was obtained from parent or guardian and assent was obtained from the child before any study procedures.

| Skin prick testing
SPT was performed using a single-head metal lancet, a positive control (10 mg/mL histamine dihydrochloride), a negative control (50% glycerol and 50% buffered saline), egg white (EW) extract (ALK Abello) and fresh foods, including raw egg and BE (the latter using slurry made up of 1 g of the challenge food in 10 mL of saline). Skin weal diameter was recorded in milimeters after 15 min.
The size of the wheal was determined as the arithmetic average of two perpendicular diameters including the longest one. The positive control test needed to be ≥3 mm, and the negative control test needed to be 0 mm. If the saline negative control test was ≥1 mm or the histamine positive control was ≤3 mm, SPT was considered inconclusive.

| IgE and IgG4 testing
Venepuncture was performed prior to the OFC. Blood collection was repeated if the OFC was performed more than 6 months after blood collection. Total IgE, sIgE and IgG4 to egg, EW, ovalbumin and ovomucoid were determined using the standardised immunoenzymatic assay ImmunoCAP (Thermofisher). For IgE levels above 100 KU A /L, serial dilutions were performed to determine the exact serum IgE level.

| Basophil activation test
Blood was collected in lithium heparin-containing tubes, and the BAT was performed on the same day of blood collection. Per condition, 100 μL of whole blood was incubated with the same volume of egg extract (ALK-Abello) or baked EW (Sigma-Aldrich) diluted in RPMI (GIBCO). Baked EW was prepared as a single batch by heating in a oven at 180°C for 20 min and stored at −80°C thereafter. Anti-IgE (1 μg/mL; Sigma-Aldrich) and formyl-methionyl-leucylphenylalanine (fMLP, 1 μM; Sigma-Aldrich) were used as positive controls and RPMI alone as a negative control. Cells were stained with CD123-FITC, CD203c-PE, HLA-DR-PerCP, and CD63-APC (all Biolegend) which had been lyophilised in individual tubes as a single batch at the start of the study. Flow cytometry was performed using FACS Fortessa with FACSDiva software (BD Biosciences). The flow cytometry data were analysed using FlowJo software (version 7.6.1; TreeStar).

| Oral food challenges
DBPCFC were performed on one single day with placebo doses randomly interspersed between active doses, like what was done in the LEAP and EAT studies. 12,13 If a reaction to placebo occurred, a 2-day DBPCFC was performed. Participants who reacted to placebo on the 2-day DBPCFC, or who had a reaction to placebo but refused to have a 2-day DBPCFC, or who did not complete the OFC or who had an indeterminate OFC were considered as not having an outcome and were excluded from the diagnostic analyses.

| Statistical analyses
Categorial variables were represented as proportions and compared with chi-square or Fisher's exact test, as appropriate. Quantitative variables were represented as median and interquartile range and compared with Mann-Whitney U-test. ROC curve analyses were performed to assess the diagnostic accuracy of the various tests; simulation and resampling techniques were conducted to assess internal validity. Optimal cut-offs were determined by the Youden index, compared with the outcome of OFC. Positive and negative cut-offs were defined by the highest point in the ROC curve with 100% sensitivity and the lowest point in the same curve with 100% specificity. Sequential use of tests was considered for the patients whose results fell within positive and negative cut-offs and for every sequence ended with OFC to clarify the allergic status of equivocal cases. All statistical evaluations were performed by SPSS version 27, internal validation was performed by numerical computation and resampling performed using the R software. Statistical tests were two tailed, and type-I error rate was set to 5% (α = 0.05).

| Sample size adequacy and statistical power assessment
A formal power calculation was conducted using the power.roc.test function from the pRoc package using the R software version 4.0.2.
According to this calculation, a number of cases in the range of 50-60 and an equal number of non-cases would result in a probability of false negative (type-II error) from 1% to 5% (Statistical power in a range of 95%-99%) with an AUC of 0.75 or above, with a probability of false positive (type-I error) in a range of 1%-5%. With the same number of cases and non-cases and an AUC of 0.7, we calculated a statistical power in a range of 85% to 95%.

| Baseline demographic and clinical characteristics of participants
Out of the 265 children screened, 65 were not eligible, 40 declined participation and 10 refused to undergo OFC ( Figure 1). One hundred and fifty children were recruited: 60 (40%) reacted to BE, 85 (57%) tolerated BE and 5 (3%) had inconclusive BE OFC. Two patients reacted after a placebo dose, and a 2-day DBPCFC was undertaken. Table S3 describes the demographic and clinical characteristics of the study population. Out of the 60 children who reacted to BE, 13 (22% of positive OFC) required intra-muscular adrenaline to treat their allergic symptoms and three of them (5% of positive OFC) required two doses of adrenaline. All patients responded well to treatment; three patients required an extended period of observation but none needed an inpatient hospital admission.

| Basophil activation test and specific IgE to egg white are the best tests to predict the outcome of baked egg and lightly cooked egg challenges
The results of tests were compared between BE allergic and BE tolerant children ( Figure 2; Table S4). The diagnostic performance of the various tests was determined in relation to the gold-standard OFC. Among SPT, SPT to EW extract had the largest area under the ROC curve but was not statistically significantly different from SPT to raw egg or the BE slurry ( Figure S2). Although the performance was similar, the diagnostic cut-offs for SPT to raw egg were higher than the one for SPT to EW. We also assessed the diagnostic performance of the difference and the ratio between the results of SPT to EW and SPT to raw egg; however, neither the difference nor the ratio of SPT to EW and raw egg seemed useful ( Figure S3). SIgE to EW had a larger area under the ROC than sIgE to ovalbumin or ovomucoid ( Figure S4). In terms of BAT, 16 (10.8%) patients had nonresponder basophils. The stimulant that provided the largest area under the ROC curve for BAT was egg extract at 10 or 100 ng/mL using the activation marker CD63 ( Figure S5). Figure 3 shows the ROC curve for the best test for each test modality, namely BAT to egg, specific IgE to EW and SPT to EW. BAT to egg and sIgE to EW were the tests with the best performance, both for BE and LCE allergies. All of the above-mentioned ROC curves had AUC in agreement to resampling and simulation estimates pointing out for internal validity of our results.

| Optimal, 100% sensitivity and 100% specificity diagnostic cut-offs
Various cut-offs were identified for SPT to EW, EW-sIgE, OVA-sIgE, OVM-sIgE, BAT using CD63 and 100 ng/mL of egg extract and BAT using CD203c and 10 ng/mL of egg extract. These cutoffs included the optimal cut-off, positive and negative cut-offs defined by 100% sensitivity and 100% specificity, respectively. The sensitivity, specificity, PPV, NPV, accuracy as well as number of true and false positives and true and false negatives are shown in Table 1 for BE for children of all ages and for children younger than 2 years. The younger children had lower cut-offs for all tests, except for BAT. The performance of tests and the diagnostic cut-offs for LCE allergy was similar to the ones determined for BE allergy (Table S6).

| Combination of diagnostic tests enables optimisation of resources and improved patient outcomes
Using 100% sensitivity and 100% specificity cut-offs to identify non-allergic and allergic children, respectively, and equivocal cases as the ones falling between cut-offs, whom will need further testing and ultimately OFC to confirm the allergic status, avoids falsepositives and false-negatives, resulting in 100% accuracy. Applying this approach to single tests followed by OFC, BAT was the test that significantly reduced the number of OFC ( Table 2; Table S7) in both age groups (59% for all ages and 54% for younger than 2 years).
Applying the same approach sequentially, with BAT as a second step in the diagnostic work-up, allowed a similar reduction in patients requiring OFC and additionally a reduction in the number of BATs performed to about 70% considering children of all ages (Figure 4). The reduction in OFC and in the number of BATs was more marked in the younger age group with only 38% of children aged below 2 years needing an OFC and 58% being tested on BAT.

| DISCUSS ION
Most egg allergic children tolerate BE. 3 Being able to eat BE can improve children's diet and quality of life. 14    features throughout the study. The proportion of egg allergic patients tolerating baked egg in our cohort was lower than previously reported, 16 which is probably because this is a selected population of patients requiring OFC to BE. Egg allergic patients who tolerate an age-appropriate amount of BE when they attend clinic are often recommended to continue with such consumption and are not referred for OFC; however, if we had included such patients, the overall proportion of BE tolerant patients in our study would have probably been higher. We performed diagnostic analyses not only for BAT but also for tests that are currently available to clinicians. We defined optimal cut-offs and also 100% sensitivity and 100% specificity cut-offs. Cut-offs with high sensitivity are useful as screening tests to capture sensitisation. Cut-offs with high specificity are useful to confirm food allergy, especially in the presence of a history of reaction.
Our results for BAT in the diagnosis of egg allergy, like previously for peanut allergy, 10 support the incorporation of BAT in the diagnostic work-up for egg allergy in clinical practice. 17  another study looking at IgE to egg components did not report advantage of OVM-sIgE, 27 and a more recent study of IgE epitopes on F I G U R E 4 Sequential use of diagnostic tests to predict the outcome of challenges to baked egg with the basophil activation test (BAT) as a second step in the diagnostic process, after specific IgE to egg white (sIgE EW) or specific IgE to ovalbumin (sIgE OVA). Cut-offs with 100% specificity (i.e. 44.7 KU/L for sIgE EW; 21.1 KU/L for sIgE OVA; 48.8% CD63+ basophils for BAT for children of all ages; and 5.35 KU/L for sIgE EW and 65.5% CD63+ basophils for BAT for children younger than 2 years) were used to confirm baked egg allergy (BEA); Cut-offs with 100% sensitivity (i.e. 0.19 KU/L for sIgE EW; 0.21 KU/L for sIgE OVA; 2.2% CD63+ basophils for BAT for children of all ages; and 0.17 KU/L for sIgE EW and 2.3% CD63+ basophils for BAT for children younger than 2 years) were used to exclude allergy (NA, non-allergic). Patients with results between these two cut-offs would need an additional test (BAT) or an OFC-after OFC, patients were considered allergic or non-allergic depending on the OFC outcome, positive or negative, respectively. Best combinations of tests for all ages and younger than 2 years age group are shown. This approach resulted in 100% accuracy.
peptides from OVA and OVM showed that a combination of peptides from the two allergens was the best predictor or reactivity to BE. 28 Children were advised to avoid baked egg for 2 days prior to the challenge given due to the possible interference of egg allergens in circulation with the outcome of the OFC and/or the BAT.

CO N FLI C T O F I NTER E S T S TATEM ENT
Dr Radulovic reports salary support from grants from National

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.