An accurate diagnosis of CMA is important to reduce the number of children on inappropriate diets. Many studies using DBPCFC with CMP to evaluate the incidence of CMA have been retrospective , and/or did not include late reactions . None included long-term follow up to assess if parents continued to follow the medical advice based on the DBPCFC. In our study DBPCFC led to the long-term use of an appropriate diet based on the presence or absence of CMA in 100 (88%) of 114 children tested (intention-to-treat analysis; information on long term diet was unavailable in 2 children).
Besides acute reactions, we also studied late reactions which typically develop within 24 to 72 hours after ingesting CMP. When late reactions are described in the literature, they form a substantial part of the positive test results, ranging from 20% to 60% [22, 26]. When only acute reactions are included, as in the study by Schade et al. , a much lower incidence of CMA is found than mentioned in the literature. In our study population, 37.5% of the 40 children with CMA developed a late reaction alone. A limitation of our study is, that the only information we have about late reactions comes from the parents; it is hardly feasible to hospitalise children for a total of 6 days to perform a DBPCFC. However, only symptoms that were identical to the original presenting symptoms that occurred after verum feeding but not after placebo feeding were interpreted as a positive DBPCFC. Systematically ignoring late reactions in DBPCFC would lead to an unjust rejection of the diagnosis of non-IgE-mediated CMA.
The presenting symptom of swelling was significantly more often present in the DBPCFC positive group. None of these children had a negative test. This is not surprising, since swelling is usually an immediate IgE-mediated hypersensitivity reaction. In these cases, the diagnosis is more easily made on clinical presentation. However, presenting symptoms of urticaria, erythema, vomiting and respiratory tract symptoms, which can also be interpreted as IgE-mediated reactions when occurring as an immediate reaction, were not significantly different between the two groups. This emphasizes the need for a DBPCFC for diagnosing or excluding CMA.
Reactions to placebo are described in the literature. Vlieg-Boerstra et al. found them in 12.9% of all their DBPCFC tests and in 5/43 (11,6%) of their cow’s milk DBPCFC’s . Hospers et al. describe a reaction to placebo feeding in 24% of their tests . We also found them in 17 children (22%). The precise cause for this is not known. We cannot fully exclude the possibility that placebo and verum feedings were accidentally exchanged in some of the cases with a reaction to placebo feeding alone, because 7 DBPCFC negative children in our study group who had developed a reaction after placebo feeding alone were later interpreted as having CMA based on recurrence of symptoms after the reintroduction of CMP. However, a reaction to placebo feeding alone occurred in 9 additional DBPCFC negative children in whom CMP was successfully reintroduced.
A possible explanation for a false negative result is the possibility that the threshold to respond is higher than the dose achieved during the challenge, i.e. larger quantities of allergen are needed to produce a reaction. Sicherer et al.  studied the quantity of food that elicited a reaction during DBPCFC in children with atopic dermatitis. Of 117 children (median age 5 years 9 months) with positive reactions to CMP, in 12% the reaction occurred after the final test dose of 2 to 2,5 grams or during open challenge. These children received a total of 8 to 10 grams of CMP. In our study the final dose consisted of 1,6 grams and a total of 4,5 grams of CMP was ingested during the test. So, it could be that some children in our study did not receive a high enough dose to produce a reaction.
Unfortunately, the data represented by Sicherer et al. is not detailed enough for us to determine a cumulative distribution of the MEDs of their study group for comparison with our study group. Recent research has explored the importance of having adequate MED-data available for population risk assessment purposes which makes optimal use of all available information, including the dose distribution of MEDs within the allergic population [29, 30]. However studies that are developed to determine MEDs often only describe the lowest MED within the population encountered, whereas a distribution of MEDs within that population is not established . Flinterman et al. , Baehler et al. , Caminiti et al.  and Patriarca et al.  however, do represent data in their studies describing allergic reactions to CMP which can be used for determining the cumulative distribution of MEDs in their population.
It is striking that our subgroup of infants aged ≤ 12 months has a higher cumulative MED distribution than the children aged > 12 months. This suggests that infants have a higher MED than older children, and therefore will react only to higher amounts of CMP. This is supported by the studies from the literature used for comparison. The cumulative MED distributions based on these studies are also lower than the cumulative MED distribution of our infant group. As can be seen in Table 5, the age distribution of the children described in the literature is comparable to our subgroup of children aged > 12 months. However, there are some important differences between our study and the studies in the literature, which makes comparison difficult. At first, we performed our study in a regional hospital. All the studies from the literature were performed in a tertiary referral centre, which can lead to a different patient selection. Patients visiting a tertiary referral centre may have a more severe CMA, and therefore a lower MED. Secondly, our study is the only study that included all children with suspected CMA, without selection based on the presence or absence of atopic dermatitis. Also no selection was made based on the severity of CMA.
The population of Flinterman et al.  is most similar to our subgroup of children considering the age distribution. However, they included only children with atopic dermatitis. The children described by Baehler et al.  are somewhat younger, however they excluded all children with co-existing atopic dermatitis. The children in the study groups of Caminiti et al.  and Patriarca et al.  are not only older than our population, but also consist of a selected patient group. These two studies were performed to investigate the effect of oral desensitization, and children with a ‘severe’ CMA were selected. It is not surprising that these children have a lower MED distribution. Therefore, our study seems more representative for the general population of children with CMA compared to the other studies mentioned above. Our study is the only one that included enough infants to allow a separate distribution for children aged ≤ 12 months. With the possible exception of the study by Baehler et al., in which the age distribution is not clearly described, none of the studies included children under the age of 12 months.
A recent paper by Brand et al.  states that the individual MED remains fairly constant over time, however we found no other studies in the literature to confirm this statement. They also state that 75% of infants with CMA are cow’s milk tolerant by the age of 1 year. 90% are cow’s milk tolerant by the age of 4 years. A study by Host et al.  in 2002 investigated the natural history of CMA. They found a recovery of CMA in 56% of patients at 1 year, 77% at 2 years, 87% at 3 years, 92% at 5 and 10 years and 97% at 15 years of age. In our study group, older children have a lower cumulative distribution of MED than the infants aged 0 – 12 months. An interesting discussion point is whether this means that the infants with a higher MED will become cow’s milk tolerant and infants with a lower MED will remain allergic to cow’s milk. Another hypothesis is that there is some kind of selection bias. Infants in general consume more milk than older children. Therefore both infants with a relatively high MED and a low MED might seek medical attention in contrast to older children who would just start consuming less cow’s milk products and don’t seek medical advice unless they have a lower MED.
Further follow-up studies are needed to confirm these hypotheses and to investigate whether the individual MED remains constant over time. At this point, the cumulative distribution of the MED in a population can only be used for population risk assessment purposes.