Observational and MR mediation analyses
According to both observational and MR analyses, the two mediators with
the highest ranking of mediation proportions were FEV1/FVC and physical
fitness (Figure 4). For all the mediators, the mediation proportions of
the MR analysis (Figure 4, right) were higher than that of the
observational analysis (Figure 4, left). FeNO did not
appear to mediate the influence of obesity on asthma, because the
mediation coefficient was in the opposite direction from obesity to
FeNO (negative in Figure 2) and from
FeNO to active asthma (positive in Figure 3), making the
indirect pathway coefficient negative; hence, the mediation proportion
was unable to be calculated. Furthermore, SDB was not considered a
strong mediator because the MR analysis showed that obesity-induced SDB
was insignificant (Figure 2). The mediators presented as consistent
causal relationships from obesity to the mediators and then from the
mediators to asthma were FEV1/FVC, physical fitness, and early puberty
in mediation proportions of 61.57%, 91.42%, and 28.28%, respectively,
based on the MR methods; and 24.88%, 21.05%, and 10.32% based on the
observational analysis. The mediation proportion implies the proportion
of the mediating pathway compared with the total effect from obesity to
asthma, considering one mediator at a time. Therefore, the mediation
proportion from all five mediators does not necessarily have to sum to
100%. The more substantial mediation proportion indicates that the
mediator might play a more critical role.
The mediation analysis assumes that the exposure and mediators had no
interaction effects on the outcome.14 We did not
observe an obvious interaction effect of BMI or BMI-GRS and the
mediators on predicting active asthma after correction for multiple
comparisons (Table E17).