Double-blind placebo-controlled food challenges in children with alleged cow’s milk allergy: prevention of unnecessary elimination diets and determination of eliciting doses

Between October 2005 and June 2009, all children with suspected CMA under our regional hospital care who had been using a cow’s milk free diet for at least 4 weeks were prospectively enrolled in a DBPCFC if informed consent was given. Concomitant diseases e.g. atopic dermatitis were no exclusion criterion. Also, no selection was made based on the severity of CMA. At enrollment, symptoms and signs that had led to the suspicion of CMA were obtained from the parents and medical files. Because it is important to prevent unnecessary and to prescribe necessary elimination diets irrespective of the underlying immunological mechanism, and because a positive specific IgE does not prove CMA, the Dutch national protocol for the diagnosis of CMA does not advise specific IgE or skin prick test in the diagnostic evaluation of suspected CMA. Therefore, these were not performed. This study was approved by the medical ethical committee of the Jeroen Bosch Hospital.

The DBPCFC protocol (Table 1) was performed in our pediatric day care ward on two separate days with at least 1 week in between using placebo feeding A and verum feeding B on randomly assigned days. For placebo feeding A the child’s own commercially available hydrolyzed infant formula was used with which the child was symptom free. This formula consisted of either Nutramigen (Mead Johnson, Woerden, the Netherlands), Nutrilon Pepti (Nutricia, Zoetermeer, the Netherlands) or Neocate (Nutricia, Zoetermeer, the Netherlands). Feeding B contained a quantity of cow's milk protein equal to regular infant feedings (1.8 gram per 100 ml); it consisted of the hydrolyzed infant formula used for feeding A and Protifar (Nutricia, Zoetermeer, the Netherlands) in a proportion of 11:3 [18]. Placebo and verum feedings were equal in taste, color and smell. The feedings were packed in identical blinded bottles. The randomization code was packed in sealed, non-transparent envelopes that remained closed until at least one week after the second test day when the reactions had been assessed. The parents, nursing staff, doctors and investigators were unaware of the administered formula’s nature. Acute reactions (within 2 hours after the test) were immediately checked by the physician. Parents noted reactions (within 72 hours after the test) in a home diary. At home, patients continued their CMP free diet until their visit to the outpatient clinic at least one week after the second test day. There, the test was interpreted according to the DBPCFC protocol and a dietary advice was given (Table 1). Parents were contacted by phone several months later about the symptoms and diet of their child.

Statistical analysis was performed using SPSS 19.0. Two-sided chi-square tests with continuity correction were performed for every single presenting symptom. P-values <0,05 were considered significant.

For determination of the MED, only children with a positive DBPCFC with symptoms within 2 hours after ingestion of CMP were analyzed. The highest dose of CMP at which no symptoms occur (‘no observed adverse effect level’; NOAEL) and the dose of the first reaction (‘lowest observed adverse effect level’; LOAEL) were determined. Individual NOAELs and LOAELs were used to fit a cumulative distribution using the LIFEREG procedure of the SAS system (version 9.1) and applying interval censoring of the data (explained in Taylor et al., 2009) [19‐21]. Interval censoring is used when the exact dose that provokes a reaction in an individual is not known, but is known to fall into a particular interval. Using a cumulative distribution makes direct comparison of the distribution of MEDs in populations of variable size possible. The result is a cumulative distribution of the MEDs in which the probability (between 0 and 100%) of the allergic response is presented that occurs at a certain dose (amount of protein intake) less than or equal to this certain dose. For comparison, data from the literature [5, 22‐24] were fitted using the same method. In determining the individual MED, we used the discrete dose at which the symptoms occur instead of the cumulative dose for several reasons. The interval time between the steps of the protocol is 20 minutes. However, when there is a subjective or mild objective symptom, the next dose is awaited until the symptoms have disappeared. Thereafter, the previous dose is repeated before the next dose is given. This makes the contribution of previous doses to the development of symptoms unclear. Secondly, in a risk assessment a worst case scenario is preferred. Therefore, we assume the discrete dosage is the eliciting dose for the individual patient.

One-hundred-and-twenty-four patients were eligible, of which 51 (41%) were girls. Age at DBPCFC varied from 2.5 months to 134 months (median 7.5 months). In 6 children the DBPCFC was not completed. In 5 of these children, this was because they refused to drink the test feedings. In one child the DBPCFC failed because of an intercurrent infection. One child didn’t meet the inclusion criterion of a CMP free diet for at least 4 weeks. Therefore 116 children are available for further analysis.

The presenting symptoms of the 124 eligible children are listed in Table 2. Feeding problems were significantly more often present in the DBPCFC-negative group (p=0.046). Swelling was significantly more often present in the DBPCFC-positive group (p=0.023), all four tested patients with this presenting symptom had a positive DBPCFC. The remaining presenting symptoms were not significantly more often present in the DBPCFC-positive or in the DBPCFC-negative children. The number of presenting symptoms was also not significantly different between the two groups. None of the patients presented with an anaphylactic reaction.

Test results were interpreted according to the type and timing of reactions (Table 3). In 76 children (66%) the diagnosis of CMA was rejected (n=56 <1 year, n=20 1–3 years of age). In 38 (33%) the diagnosis was confirmed (n= 27 <1 year, n= 7 1–3 years, n=4 >3 years of age). Details about the MED and the type of reactions at this dose are listed in Table 4. In 12 of these 38 children (32%) an acute reaction occurred. In 15 of the 38 children (39%) there was a late reaction, and 11 (29%) showed both an acute and a late reaction. In two children an objective acute reaction to placebo feeding alone was observed. The first of these two children (nr 74) developed an acute itching rash on the arms and face after the first dose (18 mg), the second child (nr 118) developed itching eyes and swelling of the mouth after the second dose (180 mg). Although reactions to placebo have been described in the literature, these children were interpreted as having CMA with accidentally exchanged placebo and verum feedings by their treating physician and parents were advised to continue the CMP free diet. They were excluded from further analysis of the MED. In 17 of the 76 children (22%) with a negative DBPCFC a reaction occurred on placebo feeding alone (an acute reaction in 1 child (vomiting), a late reaction in 14 and an acute plus late reaction in 2).

Follow-up was performed by contacting parents by telephone several months after the reintroduction. When we were unable to reach parents by telephone, information was gathered from the medical file. Information on reintroduction was available in 74 of the 76 DBPCFC negative children (Table 3) with a mean follow-up of 12 months (range 1–31 months). In 60 of the 74 children (81%) in whom the diagnosis of CMA was rejected, cow’s milk was successfully reintroduced permanently. In 9 children (12%) symptoms reappeared upon reintroduction of cow’s milk. Symptoms included crying, worsening of eczema and diarrhea. The symptoms disappeared with re-elimination of cow’s milk from the diet. Therefore, they were interpreted by the treating physician as an allergic reaction to cow’s milk, with a false negative DBPCFC, probably due to reactions that only become apparent upon prolonged feeding with high doses of CMP. In 7 of these 9 children a reaction to placebo feeding alone had been observed that was equal to the presenting symptoms. These reactions consisted of colic, diarrhea, vomiting, eczema, erythema, itching, crying and sleeping problems. In one there was a comparable reaction to both feedings (diarrhea, colic), the other child had no reaction to either of the feedings. In 5 of the 74 children (7%) parents refused or stopped reintroduction.

23 children showed a positive DBPCFC with an acute reaction within 2 hours after ingestion of CMP. This excludes the two children with an acute reaction to placebo feeding who were interpreted as having CMA by their treating physician. Information on NOAEL and LOAEL was unavailable for 2 children due to incomplete medical files. Therefore, 21 children were available for determining the MEDs. Figure 1 shows the number of allergic responders at each MED in this population of 21 children (the Jeroen Bosch Hospital population; JBH population). For comparison between age groups, the total population was divided in two subgroups according to age (0–12 months, n=14; >12 months, n=7). To enable comparison of the two populations the individual MED are presented as a cumulative probability distribution (Figure 2). Data from the literature derived from the study populations of Flinterman et al. (n=11) [5], Baehler et al. (n=16) [22], Caminiti et al. (n=13) [23] and Patriarca et al. (n=8) [24] are represented in the same figure. Detailed information on the populations represented in Figure 2 is shown in Table 5. Figure 2 shows that the highest cumulative distribution of MEDs was in our study subgroup of children aged 0 – 12 months. The cumulative MED distribution of our population of children aged > 12 months group is comparable to the patient groups described in the literature.