Table I.

Overview of sensitivity analyses

Sensitivity analysis	Justification	Findings
Primary analyses
The primary analysis was repeated on an incident (newly active) atopic eczema cohort (exposed patients, defined as those joining the cohort when they first fulfill our diagnostic criteria and after the start of the study period)	To reduce potential overadjustment, as covariates measured at entry precede atopic eczema onset in the incident atopic eczema cohort so will not be on the causal pathway between atopic eczema and mortality	Associations were partially attenuated for most causes of death relative to those in the primary analysis (see Table E6 in the Online Repository at www.jacionline.org)
The primary analysis was repeated on individuals with at least 1 consultation with their GP in the year before cohort entry	To reduce the possibility that the control group members are healthier purely because of the study design (controls are not required to have had a recent GP consultation in the primary analysis, whereas atopic eczema is defined on the basis of diagnosis and relevant treatment)	Associations were partially attenuated for all causes of death relative to those in the primary analysis (see Table E7 in the Online Repository at www.jacionline.org)
The primary analysis was repeated on a first redefined cohort (“first redefined cohort”), where the pool of unexposed persons also included patients with an atopic eczema diagnosis but without 2 further treatments for the entire duration of their follow-up and patients in the exposed cohort (with an atopic eczema diagnosis and 2 further treatments) were included as unexposed up until their cohort entry (ie, the latest of their atopic eczema diagnosis and their 2 further treatments)	To explore the sensitivity of the results to the definition of the exposure	Associations were very similar to those in the primary analysis for all causes of death (see Table E8 in the Online Repository at www.jacionline.org)
The primary analysis was repeated on a second redefined cohort (“second redefined cohort”), where the exposed patients were those with an atopic eczema diagnosis only (ie, without requiring 2 atopic eczema treatments), and these patients were eligible for the unexposed cohort up until their atopic eczema diagnosis (some patients may have had childhood atopic eczema but may not have had treatment codes recorded if registered at GP during adulthood, and therefore may have been erroneously excluded from the exposed cohort in the primary analysis)	To explore the sensitivity of the results to the definition of the exposure	Associations were partially attenuated for all causes of death relative to those in the primary analysis (see Table E9 in the Online Repository at www.jacionline.org)
The primary analysis was repeated on a subset of patients registered from 2006 onward, with additional adjustment for ethnicity	To examine whether the omission of this covariate in the primary analysis may have introduced bias	Associations adjusted for ethnicity (restricting to the 339,734 people who had such data available and who remained in valid matched sets) were consistent with those that were unadjusted for ethnicity in the same reduced cohort (see Table E10 in the Online Repository at www.jacionline.org)
The primary analysis was repeated with the exposure redefined to represent time-varying time since diagnosis (unexposed vs exposed at 0-4 years since diagnosis vs exposed at 5-9 years since diagnosis vs exposed at ≥10 years since diagnosis)	To examine whether the association between atopic eczema and all-cause and cause-specific mortality differed by time since diagnosis	The association with all-cause mortality appeared less strong within 5 years of diagnosis than it was more than 5 years after diagnosis, but the patterns with individual causes of death were heterogeneous (see Table E11 in the Online Repository at www.jacionline.org)
Secondary analyses
The atopic eczema severity analysis was repeated on individuals with at least 1 consultation with their GP in the year before cohort entry	To reduce the possibility that the control group members were healthier purely owing to the study design (controls were not required to have had a recent GP consultation in the primary analysis, whereas atopic eczema was defined on the basis of diagnosis and relevant treatment)	Associations were only slightly attenuated relative to those in the unrestricted analysis (see Table E12 in the Online Repository at www.jacionline.org)
The atopic eczema activity analysis was repeated on individuals with at least 1 consultation with their GP in the year before cohort entry	To reduce the possibility that the control group was healthier purely on account of to the study design (controls were not required to have had a recent GP consultation in the primary analysis, whereas atopic eczema was defined on the basis of diagnosis and relevant treatment)	Associations were only slightly attenuated relative to those in the unrestricted analysis (see Table E13 in the Online Repository at www.jacionline.org)
The atopic eczema activity analysis (assessed during the first 12 months of follow-up, assuming active eczema for 3 months after a single health care contact) was repeated, without excluding the first year of follow-up	To explore any potential bias caused by conditioning on survival up to 12 months in the analysis in which the first 12 months of follow-up are excluded. However, bias due to inclusion of the exposure definition period in follow-up may instead be induced	Associations were generally very similar, although with greater evidence of a protective association with moderately active atopic eczema for some causes of death (see Table E14 in the Online Repository at www.jacionline.org)