Background:

Food challenges (FCs) form the basis for assessing efficacy outcomes in interventional studies of food allergy; however, different studies have used a variety of similar but not identical criteria to define a challenge reaction, including subjective (nonobjective) symptoms occurring in a single-organ system as dose limiting.
Objective:

Our aim was to undertake a secondary analysis of 4 interventional studies to assess the impact of using less objective criteria to determine challenge-stop on reaction thresholds and their reproducibility.
Methods:

We analyzed individual participant data, including individual participant data meta-analysis, by using 3 different published challenge-stop criteria: (1) PRACTALL consesus criteria; (2) Consortium for Food Allergy Research version 3 (CoFAR v3) with at least 1 moderate- or severe-grade symptom; or (3) CoFAR v3 with at least 2 mild symptoms occurring in different organ systems. Reproducibility of challenge threshold was also assessed in participants undergoing subsequent repeat FCs.
Results:

Four studies, with detailed challenge data from a total of 592 participants, were included. Applying CoFAR v3 definitions for dose-limiting symptoms resulted in an underestimate of reaction thresholds compared with those in PRACTALL (P < .001) that is equivalent to almost a single dosing increment when using a semi-log dosing regimen. Reproducibility was also reduced when applying CoFAR v3 (P < .001 [n = 223]). Using the least conservative interpretation of CoFAR v3 (≥2 mild symptoms occurring in different systems) resulted in a significant overestimate of 15% when assessing oral immunotherapy efficacy. Applying a data-driven minor modification to CoFAR v3 resulted in a new set of challenge-stop criteria with validity similar to that of PRACTALL but one that is simpler to implement and in which significant gastrointestinal discomfort with observable decreased activity remains a dose-limiting symptom.
Conclusion:

The use of less objective symptoms to define challenge-stop compromises the reproducibility of the FC as a tool to assess efficacy outcomes in interventional studies, and potentially overestimates the efficacy of the intervention tested.