Eczema presence and eczema phenotypes
AD was operationalized as self-/guardian-reported eczema. Within each age bracket, eczema presence was determined using three parameters: presence, medication and treatment by a doctor. Eczema presence was confirmed if presence was reported or if presence was denied, but either medication or treatment were reported. Eczema was not confirmed if all three parameters were reported negatively. Ever eczema was defined as eczema presence in at least one of the age brackets.
Eight longitudinal phenotypes were defined, based on presence, age of onset and persistence of eczema (Table 1) . This stratification of eczema subclasses had been first described by Paternoster et al. , who identified them from two longitudinal birth cohorts using a latent-class analysis.7 Age of eczema onset was defined as “Very early” if the first presence was reported before the age of 6 months, “Early” between 6 months and 4 years and “Late” after the age of 4 years (Supplementary Figure 1) . Persistence/remission of eczema was defined by assessing eczema presence in the most recent age bracket available. When persistence/remission could not be determined due to incomplete data, subjects were classified into phenotypes with “no info on persistence”. Very young children (age < 4 years at baseline) provided no information about persistence/remission and were excluded.