Different definitions of atopic dermatitis: impact on prevalence estimates and associated risk factors

BACKGROUND: There is no objective test that can unequivocally confirm the diagnosis of atopic dermatitis (AD), and no uniform clinical definition. OBJECTIVES: To investigate to what extent operational definitions of AD cause fluctuation in the prevalence estimates and the associated risk factors. METHODS: We first reviewed the operational definitions of AD used in the literature. We then tested the impact of the choice of the most common definitions of ‘cases’ and ‘controls’ on AD prevalence estimates and associated risk factors (including filaggrin mutations) among children aged 5 years in two population‐based birth cohorts: the Manchester Asthma and Allergy Study (MAAS) and Asthma in Ashford. Model performance was measured by the percentage of children within an area of clinical indecision (defined as having a posterior probability of AD between 25% and 60%). RESULTS: We identified 59 different definitions of AD across 45 reviewed studies. Of those, we chose four common ‘case’ definitions and two definitions of ‘controls’. The prevalence estimates using different case definitions ranged between 22% and 33% in MAAS, and between 12% and 22% in Ashford. The area of clinical indecision ranged from 32% to 44% in MAAS and from 9% to 29% in Ashford. Depending on the case definition used, the associations with filaggrin mutations varied, with odds ratios (95% confidence intervals) ranging from 1·8 (1·1–2·9) to 2·2 (1·3–3·7) in MAAS and 1·7 (0·8–3·7) to 2·3 (1·2–4·5) in Ashford. Associations with filaggrin mutations also differed when using the same ‘case’ definition but different definitions of ‘controls’. CONCLUSIONS: Use of different definitions of AD results in substantial differences in prevalence estimates, the performance of prediction models and association with risk factors. What's already known about this topic? There is no objective test that can unequivocally confirm the diagnosis of atopic dermatitis (AD) and no uniform clinical definition. This results in different definitions utilized in AD studies, raising concerns on the generalizability of the results and comparability across different studies. What does this study add? This study has shown that different definitions of ‘cases’ and ‘controls’ have major impacts upon prevalence estimates and associations with risk factors, including genetics, in two population‐based birth cohorts. These findings suggest the importance of developing a consensus on AD definitions of both ‘controls’ and ‘cases’ to minimize biases in studies.