Patients with a history of immediate reaction to cow's milk were enrolled. Patients who were strongly suspected to have a cow's milk allergy, such as those who experienced an allergic reaction to cow's milk, such as eczema, that improved by elimination of cow's milk and who were positive for cow's milk-specific IgE, were also enrolled. Past history of anaphylaxis was not established as an exclusion criterion, and no exclusion criteria were established with respect to cow's milk-specific IgE antibody titers. We did not use the skin prick test because cow's milk-specific IgE is more useful for the diagnosis of heated cow's milk allergy.1

Oral food challenges (OFCs) were not performed on patients with a history of immediate reaction to cow's milk within the past 6 months or those with comorbid symptoms such as eczema or respiratory symptoms that would affect the determination of OFC results.

We used total and specific IgE values (Immuno CAPTM; Thermo Fisher Scientific, MA, USA) measured within 6 months before the OFC. Cow's milk-specific IgE titers were measured within 6 months of OFC.

The OFC was performed by an open challenge method during hospitalization. We first performed an OFC for butter and then performed an OFC for heated cow's milk, regardless of the test outcome.

In the OFC for butter, we made a sweet potato cake containing 10 g butter (equivalent of 2.9 mL cow's milk) and sweet potatoes. First, the potatoes were boiled and mashed, and then the butter was added. The potato cake was administered either all at once or for patients with a past history of severe reaction to cow's milk, in 2 portions administered 1 hour apart. For the heated cow's milk OFC, we used a pumpkin cake containing 25 mL of cow's milk. The pumpkin cake was heated in a microwave (1,000 W for 1 minute and 30 seconds at a temperature of 89℃). We performed the OFC for heated cow's milk by administering the cake in 2 separate portions 1 hour apart. Challenges were discontinued at the first objective sign of reaction.

The results of statistical analyses are expressed as the median value and 25-75th percentiles. To make a statistical comparison between 2 groups, we used the Mann-Whitney U test or Fisher's exact test, and P values of <0.05 were considered statistically significant. We created probability curves using the regression analysis method after logarithmic transformation of antibody values. Sensitivity, specificity, positive predictive values, and negative predictive value were calculated. The data were statistically analyzed using the statistical software SPSS 20.0 (IBM Corporation, NY, USA).

According to the Declaration of Helsinki, the study design and risks of symptom provocation were fully explained to patients' guardians both orally and in writing, and written informed consent was obtained from all participants. Approval by the institutional review board was not necessary because in Japan, food challenge tests are within the purview of health insurance treatment.

Sixty-eight children (median age, 3.9 years; 25-75th percentile, 2.1-7.2 years) underwent butter and heated milk challenges (Fig. 1, Table). Thirty-eight children tolerated butter only (butter-tolerant), 6 reacted to butter (butter-reactive), and 24 tolerated heated milk (heated milk-tolerant). Thirty-eight (86.4%) of 44 patients with a positive OFC for heated cow's milk could tolerate butter. All 6 patients with a positive OFC for butter also had a positive OFC for heated cow's milk. Four of 6 patients with a positive OFC for butter had minimal symptoms within 10 minutes and did not need to be treated. One patient developed a wheezing cough. This patient improved rapidly with administration of oral antihistamine and steroids and an inhaled bronchodilator. Another patient complained of symptoms of cough and itchiness in the throat and over the whole body. This patient's symptoms resolved rapidly after an oral antihistamine was taken (Supplementary Table 1). Patients were observed 4 hours after the symptoms disappeared, and no biphasic reaction was observed.

All subjects with serum milk-specific IgE <11.3 kUA/L tolerated the butter challenge (24/24); thus, both negative tests had excellent sensitivity (100%) and negative predictive value (100%). However, specificity (37.1%) and positive predictive value (13.3%) were poor. A total of 90.2% of patients with milk-specific IgE levels of 11.3 kUA/L to 100 kUA/L tolerated the butter challenge (37/41), and 60.0% of patients with milk-specific IgE levels over 100 kUA/L tolerated the butter challenge (3/5). We then created probability curves based on the results of the OFCs (Fig. 2). With a higher milk-specific IgE value, there was a higher positive probability of OFC for butter. Based on the statistical model, the positive probability of patients with a milk-specific IgE value of 100 kUA/L was 31.3%. We could not calculate milk-specific IgE values to indicate a positive predictive value over 95%. Based on this statistical model, a cut-off value of 17.8 kUA/L was proposed for identifying those who would tolerate the butter challenge (approximately 95% rate of tolerating butter). We created probability curves by patient age based on the results of the OFCs (Fig. 3). All patients aged ≤2 years passed the challenge test. Based on this statistical model, a cut-off value of 13.7 kUA/L was proposed for identifying those aged ≥2 years who would tolerate the butter challenge (approximately 95% rate of tolerating butter).