DISCUSSION
In the current study, meta-analysis indicated an increased risk of T1DM in children with asthma, but not increased risk for asthma in children with T1DM. However, strong heterogeneity among the included studies was noted. Subgroup analysis, meta-regression, and sensitivity analysis failed to identify the sources of heterogeneity.
The results of MR analysis were not completely consistent with that of the meta-analysis. Specifically, children with asthma had higher risk of T1DM, whereas children with T1DM had lower risk of asthma. The importance in sequential appearance of the two diseases was apparently illustrates. However, in the majority of the included studies, it was not clear whether an asthma diagnosis or symptoms preceded the T1DM diagnosis, which could lead to reverse causality.27-38Only 3 cohort studies documented the sequence of disease occurrence, but the results were still conflicting.6-8 The results of MR analysis are consistent with a previous cohort study by Metsala et al7. Similarly, Smew et al6indicated that childhood asthma was related to an increased risk of subsequent T1DM, whereas T1DM was not associated with a decreased risk of subsequent asthma. Hisiao et al8, however, reported a markedly higher incidence of asthma in patients with T1DM in the Asian population. It is important to note that the MR analysis data came from European population. Whether the results of MR analysis in the current study could be extrapolated to other populations remained unknown.
Both nurture (i.e. the environment) and nature (i.e. genetics) contribute to the etiology of both asthma and T1DM. Asthma and T1DM tend to be comorbid and congregate within families.6Siblings of children with one disease have higher risk of the other disease.6 For example, TLR2 rs3804100 T allele has been shown to be a susceptibility allele for both childhood asthma and T1DM.39 In contrast, the T allele of rs9273349 and rs1063355 in HLA-DQB1 seems to be a protective factor to both diseases.16 Adding to the complexity, susceptibility genes to one disease could be protective to the other disease. For example, GIMAP5 SNP (rs6965571) has been associated with increased risk for asthma but decreased risk for T1DM.40 Future GWAS studies are required to better characterize the potential link between them.41
At a mechanistic level, lower risk of asthma in children with T1DM could be attributed to the anti-inflammatory properties of insulin. High levels of insulin promote Airway smooth muscle (ASM) contraction, enhance contractile responses to methacholine, which are likely to result in increased airway hyperresponsiveness, bronchoconstriction, and airway remodeling.42 Insulin deficiency in T1DM could prevent airway remodeling.43 A recent study in a mouse model for asthma showed that insulin deficiency could prevent the development of allergic inflammation, eosinophilic pulmonary infiltration, and airway hyperresponsiveness in an asthma mouse model.44 These findings mainly explain the reduced risk of asthma in patients with T1DM, but lacking mechanistic research on the increased risk of T1DM in children with asthma. The use of some anti-asthmatic drugs (inhaled corticosteroids and inhaled β-agonists) was potentially associated with the risk of T1DM in a case-cohort study. Therefore, future research is required to explore the underlying biological mechanisms of this relationships.
MR analysis makes sense only when it uses IVs that are: (1) associated with the exposure; (2) independent of factors confounding the association of the exposure and outcome; and (3) associated with the outcome only through exposure. With these premises, MR analysis can work as an analogue to RCT, which is more plausible to identify causality than observational analysis.45 To satisfy the first assumption, we selected only SNPs with a genome-wide significant association (P < 5*10-8) and the F statistic at > 10.14 The second assumption is not verifiable. To satisfy the third assumption, MR-Egger regression was used to estimate horizontal pleiotropy.23 Since these premises were fulfilled in our MR analysis, we believe that our findings were reached with least likelihood of confounding effects and maximal avoidance of reverse causality.
The meta-analysis in the current study has several limitations. First, majority of the studies included in the analysis are of case-control design; only three cohort studies were included. Second, due to the low prevalence of T1DM, most included studies were based on T1DM and considered asthma as potential risk factors. Third, there was significant heterogeneity among the included studies, but we were unable to identify the source of heterogeneity. The MR analysis also has several important limitations. First, the MR analysis used only one dataset, and lacks validation. Second, some of the SNPs (such as rs4795399 and rs28407950) utilized in this study are potentially correlated with confounding factors which may also influence the risk of T1DM or childhood asthma. We were unable to completely rule out the possible impact of pleiotropic effects on the findings. Third, the MR analysis data came from European population only, and whether the findings could be extrapolated to other populations remains unknown.