Single- dose oral challenges to validate eliciting doses in children with cow’s milk allergy

Background: There is increasing interest in the use of eliciting doses (EDs) to inform allergen risk management. The ED can be estimated from the distribution of threshold doses for allergic subjects undergoing food challenges within a specified population. Estimated ED 05 values for cow's milk (the dose expected to cause objective allergic symptoms in 5% of the milk- allergic population) range from 0.5 mg to 13.9 mg cow's milk protein. We undertook a single- dose challenge study to validate a predicted ED 05 for cow's milk of 0.5 mg protein. Methods: Participants were recruited from 4 clinical centres. Predetermined criteria were used to identify patients reacting to 0.5 mg cow's milk protein (approximately 0.015 mL of fresh cow's milk). Children over 1 year underwent formal challenge to cow's milk to confirm clinical reactivity. Results: 172 children (median age 6.0 (IQR 0.7- 11) years, 57% male) were


| INTRODUC TI ON
There is increasing interest in the use of routinely collected clinical data from oral food challenges (OFC) to inform both patient management and allergen risk management in industry, in terms of the level of dietary allergen avoidance required. Eliciting doses (ED) for allergic reactions in 1% and/or 5% of the allergic population (ED 01 and ED 05 , respectively) can be used to inform 'reference doses', indicating a level of allergen presence above which additional risk management strategies (such as precautionary allergen labelling) are required to protect the allergic population. 1,2 In addition, it has been proposed that dietary advice to food-allergic consumers might be individualized if a particular level of tolerance can be demonstrated at clinical OFC. 3,4 ED values are generated from OFC data, 5 but many OFC protocols use a starting dose that will trigger symptoms in a significant proportion of patients. For example, the PRACTALL consensus recommends a starting dose of 3 mg food protein for OFC, 6 but data suggest that for cow's milk protein allergy (CMPA), this may cause objective symptoms in 10% of allergic individuals. 7 Thus, these data are 'left-censored' and cause greater uncertainty when estimating a level of exposure, which causes symptoms in a small proportion of the allergic population. 5 Conventional protocols, which use incremental doses given every 20-30 minutes also make it difficult to reliably determine the precise dose that has triggered symptoms, but this uncertainty can be reduced by using both discrete and cumulative doses to estimate ED values. 8 In addition, relying on OFC undertaken in routine clinical practice can result in selection bias, since subjects at high likelihood of true clinical reactivity or with a history of prior anaphylaxis may be excluded. 6 CMPA is a major cause of severe and even fatal allergic reactions. 9,10 Data from the United Kingdom have found that cow's milk is the confirmed trigger in over a quarter of anaphylaxis fatalities in children, 11 a pattern that has also been noted in North America and Israel. [12][13][14] This is probably due to a combination of factors: milk as an ingredient that is ubiquitous in our diets; milk as a high protein food; and lower levels of awareness amongst the public and food business operators that CMPA can cause severe reactions. 15 Reported estimated ED 05 values for cow's milk in the literature range from 0.21 mg to 13.9 mg cow's milk protein. 2,7,16,17 This variation may be partly explained by differences in patient selection, with some studies including younger patients (many of whom may be at various stages of outgrowing their milk allergy) or with lower levels of sensitization, which has been associated with higher levels of reactivity at challenge. 17 We have previously used a novel, single-dose challenge design to validate the ED 05 for peanut. 3 In this study, we sought to replicate this method in children with cow's milk protein allergy (CMPA), to assess whether current estimates for ED 05 for cow's milk are valid in terms of allergen risk management. and not necessarily those of the NHS, NIHR, or the Department of Health. The funders of the study had no role in study design, data collection, data analysis, data interpretation or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

| ME THODS
Editor: Motohiro Ebisawa that responded to a single dose of adrenaline, the remainder experienced only mild symptoms with no treatment required. We did not identify any baseline predictors of sensitization that were associated with objective reactivity to the single-dose challenge using 0.5 mg cow's milk protein.

Conclusions:
These data support an estimated ED 05 for cow's milk of 0.5 mg protein.
Values for ED 05 above 0.5 mg for cow's milk protein proposed for allergen risk management need to be reviewed.

K E Y W O R D S
cow's milk, eliciting dose, single-dose challenge, thresholds, Voluntary Incidental Trace Allergen Labelling (VITAL)

Key Message
In this multicentre study, we validated the eliciting dose (ED) for cow's milk at around 0.5 mg cow's milk protein (ED 05 ), and thus, higher estimates currently proposed to inform allergen risk management should be reviewed.
OFC; anaphylaxis of any cause in the 4 weeks prior to OFC (to exclude patients in an anergic state); and antihistamines within 5 days of OFC. In order to minimize selection bias, routine clinic lists were screened for patients with IgE-mediated allergy to cow's milk, and then, participation discussed with all potentially suitable participants and their families during routine clinic appointments. Subjects with a history of prior anaphylaxis were not excluded. The studies were registered at Clinicaltrials.gov (NCT02216175, NCT02295397).

| Food challenge
Protocols were aligned across the 4 centres in order to obtain the same clinical data following 0.5 mg cow's milk protein (approximately 0.015 mL of fresh cow's milk) administered as a single dose, using the same predefined case definition for objective allergic symptoms. In general, the single-dose challenge was administered as milk powder incorporated into an allergen-free chocolate dessert matrix (previously validated for double-blind challenges 19 ) or dissolved into flavoured rice 'milk' (Table 1). In participants under age 1 year at CUH, the dose was instead administered as diluted (1:7) fresh milk using a syringe (to reduce the risk of a contact reaction to the lips). Routine OFC monitoring was undertaken according to local practice. At two centres (Imperial and NJH), formal DBPCFC were conducted, where the 0.5 mg dose constituted the first dose of the OFC; subjects were observed for at least 1 hour prior to the next challenge dose being administered (and longer if there were any non-transient symptoms). At HCSC and CUH, subjects underwent a single (unblinded) administration of 0.5 mg protein and were observed for at least 2 hours thereafter.

| Criteria for a positive OFC result and case definition
Data collection and case definitions have been previously described. 3 In brief, detailed notes were taken recording all physical or behavioural changes observed or self-reported during the single-dose OFC. Predetermined objective criteria were used, since published ED 05 values are predicted on the basis of challengeassociated objective symptoms only. [1][2][3][4][5][6] The predetermined objective criteria for a positive single-dose OFC result were as follows: 3 or more concurrent wheals of non-contact urticaria persisting for at least 5 minutes; perioral or periorbital angioedema; rhinoconjunctivitis (including sneezing) for at least 5 minutes; diarrhoea; vomiting (excluding gag reflex); or anaphylaxis (with evidence of circulatory or respiratory compromise, such as persistent cough, wheeze, change in voice, stridor, difficulty breathing, and collapse). 20 Transient objective symptoms (rhinoconjunctivitis <5 minutes, transient mild erythema) were excluded. Subjective symptoms were also recorded. Following completion of the clinical studies, cases were reviewed by at least 2 senior independent investigators and the above criteria were applied to define OFC which met these predetermined objective criteria.

| Confirmation of clinical reactivity to cow's milk
In order to avoid the possibility of including participants without CMPA, clinical reactivity was confirmed in participants over 1 year of age at formal oral exposure, typically double-blind placebocontrolled challenge conducted according to international

Ireland
Madrid, Spain United Kingdom

Imperial College London (Imperial)
Inclusion criteria History of unequivocal exposure (including accidental) and typical acute allergic reaction within the preceding 2 mo and evidence of IgE sensitization (SPT or sIgE) to cow's milk. OR Positive OFC to cow's milk within 2 mo of the single-dose challenge.
History consistent with IgE-mediated allergy to CM and Positive DBPCFC to cow's milk immediately following single-dose challenge.

| IgE sensitization
Blood samples were collected from participants prior to OFC.
Samples were processed according to the manufacturers' instructions and snap-frozen at −80°C until analysis. Specific IgE to cow's milk and casein was measured using ImmunoCAP (Thermo Fisher Scientific). Skin prick testing was undertaken according to international guidelines using ALK lancets and commercial extracts (ALK-Abello) with 1% histamine as a positive control, and the mean wheal diameter noted.

| Statistical analyses
Analyses were planned prospectively. The proportion of participants reacting to 0.5 mg cow's milk protein was estimated with 2-sided exact 95% confidence intervals. Baseline characteristics across cohorts were compared using Kruskal-Wallis test since the data were not normally distributed. Receiver operating characteristic (ROC) curves were generated in order to identify possible predictors for reactivity to 0.5 mg cow's milk protein. A P value of < .05 was considered significant. Assuming a reaction rate of 5% to 0.5 mg cow's milk protein, an overall sample size of 150 and 250 would correspond to a lower 95% confidence limit of 2.1% and 2.8%, respectively, and an upper confidence limit of 9.8% and 8.7%, respectively, for the estimated ED 05 .

| RE SULTS
A total of 267 children were screened for inclusion between August 2015 and September 2020, of whom 182 underwent OFC to 0.5 mg cow's milk protein. Ten individuals went on to pass a formal food challenge (ie did not react to a minimum of 250 mL cow's milk) following the 0.5 mg challenge and were therefore excluded from the primary analysis. Baseline demographics are shown in  Table 2, and the dose distribution is summarized in Figure 1. There were no differences across the cohorts in terms of eliciting dose (P = .29), implying that the 4 cohorts were similar to each other in terms of clinical reactivity. We did not observe any correlation between age and eliciting dose at formal challenge (Spearman's r = .05, P = .59).
These data therefore broadly validate the estimated ED 05 for cow's milk of 0.5 mg protein (with potential reactions occurring in an interval between 3.7% and 11.9% of the milk-allergic population). In the latest analysis by the VITAL Scientific Expert Panel, over 21%

| D ISCUSS I ON
of data was left-censored (ie patients with CMPA reacted to the first OFC dose) and 75% of included data were derived from OFC where the initial dose was >1.5 mg protein (and often significantly more so). 2 In addition, current estimates rely on data from routine clinical challenges where subjects may be excluded (eg due to prior anaphylaxis or recent reaction) and so the resulting dose-distribution curves may not represent the true allergic population. Importantly, heterogeneity in patients with cow's milk allergy is much more likely to be a confounder compared to other food allergies. This is because many included subjects may be in a transition towards natural resolution; there is evidence that this dynamic nature of milk allergy in younger children will impact on reaction threshold levels at food challenge. 17 These are the pivotal justifications for single-dose challenges (such as this study) to validate the estimated ED at the lower end of the dose distribution curve where data have been lacking, using a patient cohort that includes subjects who are less likely to outgrow their milk allergy in early childhood.
It is particularly important to have certainty over EDs used for allergen risk management in CMPA. Cow's milk is increasingly ubiquitous in our diets; its protein fractions are soluble and both (liquid) milk and milk powder tend to distribute well in formulations resulting in a homogenous distribution throughout a food product (as opposed to particulate distribution associated with allergens such as nuts). 9,23 It is a frequent cause of severe and even fatal allergic reactions [9][10][11][12][13][14] and can be difficult to eliminate from food production lines (eg those used to produce chocolate-based products) to the extent that a significant proportion of dark chocolate products (made without cow's milk as an ingredient) contain significant levels of cow's milk protein due to shared production. 23,24 In validating the ED 05 for cow's milk as 0.5 mg protein, these data indicate that current estimates for ED 05 for cow's milk based on population modelling using existing data are too high. Additional, larger challenge data sets (based on dosing schedules that would allow for interval censoring) are needed to provide more precision to the population dose-distribution modelling around lower ED values.
These data are also relevant to the selection of appropriate protocols for clinical challenges to diagnosis CMPA. In general, the initial doses recommended for DBPCFC under the PRACTALL consensus are 3 mg protein, 6 which for most allergens will tend to cause objective symptoms in around 10% of individuals (ED 10 ). 1,7 If the ED 05 for cow's milk is closer to 0.5 mg, then well over 10% of individuals with CMPA would be expected to react to an initial dose of 3 mg.

TA B L E 3 Symptoms experienced to single-dose challenge to cow's milk
Furthermore, many challenge protocols used in clinical practice start with higher doses of 1 mL cow's milk (approximately 30 mg protein), 25,26 to which around 25% of allergic individuals will react. In the context of OFC where patients may have a higher likelihood of clinical reactivity (eg prior to commencing allergen immunotherapy), clinicians might therefore wish to choose a lower initial challenge dose to which objective symptoms are unlikely (eg to build confidence in the patient and their family).

| Strengths and limitations of this study
The international collaboration, robust protocol and the use of predetermined objective, challenge-positive criteria to demonstrate true clinical reactivity (including by OFC in 67%, of which 84% were DBPCFC) are strengths of this study. Infants in one of the Cork cohorts underwent challenges using liquid milk rather than milk powder; however, the estimated EDs for liquid milk and milk powder are equivalent. 7 We chose to recruit a significant proportion of participants under 1 year of age, since CMPA is more prevalent in this age group, but also included teenagers with persistent CMPA who are often excluded from challenge studies. It is possible that very young children with CMPA are more likely to outgrow this allergy and this may be reflected in eliciting doses at challenge. However, we did not identify any differences across the included cohorts in terms of eliciting dose, although the number of very young children in this subanalysis was limited.
To minimize selection bias, we utilized a recruitment strategy that did not involve the subjective selection of participants by healthcare professionals, nor did we exclude participants with a history of anaphylaxis. Furthermore, the distribution of eliciting doses at challenge in this study (as depicted in Figure 1) is not dissimilar to published data for cow's milk, 22 which strongly supports our assessment that our participants are very likely to represent the population with CMPA in Europe. We were unable to determine the proportion of subjects with tolerance to baked milk, as this had not been assessed in the majority of subjects. In any event, data suggest that eliciting doses in individuals with or without tolerance to the baked allergen one had no symptoms whatsoever to the 0.5 mg dose, but rapidly developed objective symptoms to the next dose administered; the second had subjective symptoms that completely resolved for at least 60 minutes prior to the next dose administered. Given these observations and published data that reactions to cow's milk occur more rapidly than for other allergens, 28 it is very unlikely that the reactions in these 2 participants were due to the 0.5 mg dose. Finally, these data apply to challenge conditions, where allergic individuals are clinically well and usually without the presence of co-factors, which might impact on reaction thresholds. Our aim in this study was not to assess the risk of participants reacting to 0.5 mg doses due to the presence of co-factors.

| CON CLUS IONS
In summary, we have demonstrated that the ED 05 for cow's milk is likely to be around 0.5 mg protein and certainly lower than some of the proposed values for ED 05 in the literature. These data demonstrate the need to validate estimated ED values derived from dose-distribution analyses of data in studies not limited by left censoring, in order to identify the most highly dose-sensitive population of patients with food allergy. This will assist regulators, public health agencies and food business operators in establishing evidence-based approaches to allergen management as means to protect the food-allergic consumer from accidental exposures.

ACK N OWLED G EM ENTS
We thank our study participants and their families, and our research support staff.