Analyzing atopic and non-atopic asthma

There is a need to better define phenotypes of asthma. However, many studies have data available only on asthma and atopy, so they are often used to define 'atopic' and 'non-atopic' asthma. We discuss and illustrate the problems of analyzing such outcomes. We used the 31 year follow-up of the Northern Finland Birth Cohort 1966 (n = 5,429). 'Atopic asthma' and 'non-atopic asthma' were defined based on presence or absence of atopy (any skin prick test >= 3 mm) at age 31. Gender and ownership of cat in childhood were used as risk factors. Simple calculations on hypothetical datasets were used to support the conclusions. 'Atopic asthma' and 'non-atopic asthma', are not well separated disease entities. The association of a risk factor with 'atopic asthma' and 'non-atopic asthma' is determined both by its association with asthma and with atopy. E. g. if a risk factor is not associated with asthma, but is protective for atopy, this will produce a protective association with 'atopic asthma', but an opposite association with 'non-atopic asthma'. This is the result from the typical analysis, which uses all non-asthmatics as the comparison group. Valid results, unconfounded by atopy, can be gained by comparing asthmatics to non-asthmatics separately among atopics and non-atopics, i.e. by doing the analysis stratified by atopy. If data only on asthma and atopy are available, asthma and atopy should be analyzed at first as separate outcomes. If atopic and non-atopic asthma are used as additional outcomes, valid results can be gained by stratifying the analysis by atopy.