Statistics from Altmetric.com
As death certification of allergy deaths is problematic, we collected reports through the prospective BPSU mechanism and other sources as listed in our article. No additional deaths in the study period have been brought to our attention since publication. Our finding of no deaths due to peanut under the age of 13 is supported rather than refuted by the study of Bock,3 cited by Clark and Ewan. In Bock’s study, which looked at all ages, there were 10 childhood deaths (compared with 8 in ours) and they found no deaths due to peanut allergy under the age of 12. The death of a child aged two years was due to brazil nut, not peanut.
Clark and Ewan were concerned that we did not put sufficient weight on respiratory presentations. Of the 55 non-fatal severe cases, 30 had only upper or lower airway problems. Of the 173 children admitted to hospital not classified by us as severe, 60 had an upper or lower airway presentation for which adrenaline or bronchodilator or both were administered. Even if all these are added to our severe group, the rate of non-fatal severe events only changes from 0.2 to 0.4 per 100 000 children per year.
Allergic deaths indistinguishable from asthma are potentially important but if a child’s symptoms are only asthmatic and no allergen is suspected, then there is no means for attributing such reactions to food or for knowing if a causal link exists. The possibility of an unrecognised link is a key area for further research.
It is misleading to state that the mechanisms of the British Paediatric Surveillance Unit will miss children referred to allergy centres not run by paediatricians. We studied hospital admissions, not outpatient referrals, and all such children should be admitted under the care of paediatricians. The effectiveness of our reporting mechanisms is supported rather than refuted by the study of Alves and Sheikh,4 cited by Clark and Ewan. Alves and Sheik report all hospital admissions for anaphylaxis in children aged 0–15 years over a four year period in England and Wales. There were 60 for food allergy and a further 240 with no cause coded. If, as Clark and Ewan suggest, 94% of the uncoded cases were due to food then there were 288 admissions from 10.6 million children (denominator from Office of National Statistics, personal communication) over four years or 0.68 admissions per 100 000 children per year. This is comparable to our reported 231 admissions over 2 years in a population of 13 million, equivalent to 0.89 admissions per 100 000 children per year.
The inclusion of Clark and Ewan’s own data in their table 1 is inappropriate. Two concerns are that their data are extrapolations and use a completely different definition of severity. But more importantly, their figures are numbers of children with severe allergy per 100 000 child population; whereas our figures, and those of Alves,4 are numbers of severe reactions per year per 100 000 child population. Clark and Ewan therefore confuse prevalence of severely allergic children with incidence of severe allergic reactions. Such data cannot be tabulated against each other and their suggestion that our figures are a 62-fold underestimate is therefore without meaning.
We all want to know which children should have an adrenaline auto-injector and what the indications should be for administering it. However in seeking such knowledge, we must be careful not to extrapolate from adult studies. In Clark and Ewan’s own study,5 the high percentage (70%) of children in the study is stated but the median age of a moderate or severe reaction on follow up was 18 years and the three severe cases referred to in which adrenaline was used were actually aged between 27 and 41 years.
We agree with Clark and Ewan’s assertion that problems over the decision to prescribe adrenaline “highlight the need for good data on the clinical features and natural history of nut allergy in children” but would add that studying the epidemiology is also fundamental.