Physician's appraisal vs documented signs and symptoms in the interpretation of food challenge tests
Grabenhenrich, L. B., Reich, A., McBride, D., Sprikkelman, A., Roberts, G., Grimshaw, K. E. C., ... Beyer, K. (2017). Physician's appraisal vs documented signs and symptoms in the interpretation of food challenge tests: The EuroPrevall birth cohort. Pediatric Allergy and Immunology. DOI: 10.1111/pai.12811
BACKGROUND: Blinded food challenges are considered the current gold standard for the diagnosis of food allergies. We used data from a pan-European multicenter project to assess differences between study centers, aiming to identify the impact of subjective aspects for the interpretation of oral food challenges.
METHODS: Nine study centers of the EuroPrevall birth cohort study about food allergy recruited 12 049 newborns and followed them for up to 30 months in regular intervals. Intensive training was conducted and every center visited to ensure similar handling of the protocols. Suspected food allergy was clinically evaluated by double-blind, placebo-controlled food challenges using a nine dose escalation protocol. The primary challenge outcomes based on physician's appraisal were compared to documented signs and symptoms.
RESULTS: Of 839 challenges conducted, study centers confirmed food allergy in 15.6% to 53.6% of locally conducted challenges. Centers reported 0 to 16 positive placebo challenges. Worsening of eczema was the most common sign when challenged with placebo. Agreement between documented objective signs and the challenge outcome assigned by the physician was heterogeneous, with Cohen's kappa spanning from 0.42 to 0.84.
CONCLUSIONS: These differences suggest that the comparison of food challenge outcomes between centers is difficult despite common protocols and training. We recommend detailed symptom assessment and documentation as well as objective sign-based challenge outcome algorithms to assure accuracy and comparability of blinded food challenges. Training and supervision of staff conducting food challenges is a mandatory component of reliable outcome data.