Causal risk factors for asthma in Mendelian randomization studies: A systematic review and meta‐analysis

Abstract Background Several risk factors for asthma have been proposed; however, the causality of these associations is sometimes unclear. Mendelian randomization is a powerful epidemiological approach that can help elucidate the causality of risk factors. The aim of the present study was to identify causal risk factors for asthma through Mendelian Randomization studies. Methods A systematic search of PubMed and EMBASE was conducted, to identify studies investigating risk factors for asthma or respiratory allergies through Mendelian Randomization. When two or more studies investigated the same risk factor a meta‐analysis was conducted. Of 239 studies initially identified, 41 were included. Results A causal association between adiposity and adult asthma risk was found in 10 out of 12 studies with a summary risk ratio of 1.05 per kg/m2 increase in BMI (95% CI: 1.03–1.07). Puberty timing (n = 3), alcohol (n = 2), and linoleic acid (n = 1) had causal effects on asthma risk, while vitamins/minerals (n = 6) showed no consistent effect on asthma. The effect of smoking on adult asthma conflicted between studies. Several of the significant associations of asthma with immune related proteins (n = 5) and depression (n = 2) investigated through multiple traits analyses could generally benefit from replications in independent datasets. Conclusion This systematic review and meta‐analysis found evidence for causal effects of adiposity, puberty timing, linoleic acid, alcohol, immune related proteins, and depression on risk of asthma.

understood. There is an urgent need for disentangling risk factor causation from correlation in the case of asthma and respiratory allergies in order to obtain better understanding of the mechanisms and etiology of asthma. [3][4][5] This could allow targeted risk factor prevention for asthma and allergy with high impact on the asthma burden and the quality of life for patients with asthma.
Observational studies can be used to understand the extent and strength of association between risk factors and asthma, and to indicate novel associations; however, results from observational studies sometimes suffer from confounding and/or reverse causation. Randomized clinical trials can control for both confounding and reverse causation, but randomized trials can be unethical or infeasible to conduct -exemplified by asking people to smoke or drink alcohol. In these cases, Mendelian randomization (MR) studies provide a well-established design for investigating causal risk factors in diseases such as asthma independent of other risk factors (Supplementary Figure S1). 6 The concept of MR builds on the random segregation of alleles from parents to offspring. These randomly allocated types of alleles are consequently not expected to be associated with any confounders except those on the causal pathway between genotypes and outcome. Also, as risk factors encountered later in life cannot influence the genetic makeup of an individual, MR also avoids the problem of reverse causation.
MR thereby constitutes a naturally determined randomized controlled trial in which the two random allocated groups are divided on the basis of genetic variants comparable to treatment and placebo groups.
We conducted a systematic review and meta-analysis of MR studies investigating any causal risk factor for asthma and other respiratory allergies to (1) provide an overview of the current knowledge of causal risk factors for asthma and respiratory allergies, and (2) to clarify where more research is needed.

| Study selection and data collection
Studies using MR to investigate the association of any potential risk factor for asthma or other respiratory allergies were included. The use of MR as a method had to be specified for inclusion of the studies identified. Thus, studies using genetic variations as proxies without using MR design were not included. Studies not directly related to asthma or respiratory allergy were excluded. We additionally excluded reviews, statistical, methodological and theoretical papers, editorials, commentaries, letters and conference abstracts. The search was not language restricted. For further details see supplementary methods S1.

| Risk of bias/quality score assessments
The risk of bias of the MR studies was assessed through a quality assessment scheme built from references. [7][8][9][10] The quality assessment scheme takes both the MR limitations and the MR assumptions into account (Supplementary Table 1) and allows potentially biased studies to be evaluated on fair basis. The quality scores range from 0 to 13 points, with 13 points awarded to studies with high statistical power and low risk of bias. "High sample size" was defined based on a power of 0.8 or above in the study, a sample size comparable to similar studies with power of 0.8 or above, or an assessment of what sample size would approximately be needed to achieve a power of 0.8 or above. Biology of the instrument was included in the quality score in accordance with Burgess et al., 9 who recommended that variant selection for MR should be based on variants having biological relevance to the exposure. The quality score system was used to weight studies such that the results from studies of high quality, and those from meta-analysis, were reported earlier and described more elaborately, than were results from single studies with low quality score.
The certainty of evidence for each potential asthma risk factor was assessed using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) framework. 11 The MR studies were initially assigned high quality of evidence as in the paper by Kim et al., 12 the quality was downgraded if risk of bias, inconsistency, indirectness, imprecision or publication bias were causes of serious concerns (one level) or very serious concerns (two levels). Evaluation of risk of bias was adapted to the MR study design, whereas the remaining risks were assessed as recommended by Cochrane. 11

| Meta-analysis
Meta-analyses were performed on studies investigating risk factors for asthma using STATA version 12.0. When two or more studies investigating the same risk factor were identified, the data was pooled using the random-effects model in the metan package. Heterogeneity was assessed through the I 2 test.

| RESULTS
239 studies were initially identified in the electronic database search of PubMed and EMBASE. After removal of duplicates (65 studies), 174 studies were screened based on title and abstract. Of these, 112 were found to be eligible for full text screening and detailed evaluation, based on the inclusion and exclusion criteria. Forty one studies were included in the review. A summary of the search flow is shown in Figure 1.

| Risk of bias assessment
The risk of bias score ranged from 3 to 13 points, with the lowest scores awarded to some of the oldest MR studies (Supplementary Table 2). The oldest studies were carried out before sensitivity analyses were developed and so are low quality for this reason. As a case in Granell et al., 13

| Level of evidence
The level of evidence for each potential asthma risk factor identified from the MR studies was evaluated by the GRADE framework in Supplementary Table 3. About a quarter of the meta-analysis/studies were supported by high evidence certainty. About a third were graded with moderate and another third were graded with low evidence certainty. A low grading was often given to studies which screened for multiple outcomes followed by MR, or studies with serious concerns for risk of bias.

| Definition of asthma and subtypes
Most asthma definitions were based on self-reported physician diagnosed asthma, some included wheezing and/or asthma treatment, others relied on hospital contact. Classification of atopic and non-atopic asthma was based on different methods including fractional exhaled nitric oxide measurements, 17 having allergic rhinitis in combination with use of allergy medication or reported allergic symptoms, 18 positive skin prick test, 13 or investigated allergic sensitization defined as specific IgE positivity to one or more inhalant allergens 19 (Supplementary Results S1).

| CAUSAL RISK FACTORS FOR ASTHMA AND RESPIRATORY ALLERGY
Forty one studies examined a potential causal association between risk factors and asthma or respiratory allergy through MR analysis.
Most studies were based on European individuals with primarily European ancestry. Two studies used the Taiwan Children Health study and thereby an Asian population. Some of the studies used multiple cohorts to increase sample size. The oldest study was from 2007; however, most of the studies were relatively recent.
Au Yeung et al. (quality score 11) showed that high childhood BMI (before age of 18) tended to increase risk of lifetime asthma, and also found a significant strong effect of adult high BMI on risk of asthma, proposing the effect of childhood high BMI on asthma to be mediated via a higher risk of having a high BMI in adulthood for children with a high BMI. 15 Two MR studies (quality scores 8 and 8) found, respectively, a small tendency for a significant result showing that higher adiposity led to mid-childhood (7-11 years) asthma, with a trend towards a stronger association for non-atopic compared to atopic asthma. 17,23 Another study from Chen et al. (quality score 7) showed no causal effect of higher BMI from birth to 17 years on the risk of asthma at age 17. 26 Zeng et al. 27 (quality score 12, certainty of evidence graded high) investigated birth weight, as it is often used as a proxy for early life development and potentially an effect on adult diseases as well; however, Zeng et al. found no causal association between adult asthma and birth weight.

| Puberty and sex hormones
Two studies investigated the association between timing of puberty and asthma 28,29 (Table 2). Both MR studies showed that early pubertal maturation was causally associated with higher risk of asthma. 28,29 The pooled RR for asthma for early puberty compared to normal puberty was 1.

| Vitamins, minerals and fatty acid levels
Seven MR studies investigated the effect of differing vitamin, mineral and fatty acid levels on asthma/allergy risk (Table 3, Figure 2). Two MR studies examined the association between vitamin D (25OHD) levels and, respectively, risk of asthma and respiratory allergies. Feng et al. (quality score 12) found no evidence for a causal association between serum 25OHD levels and allergic rhinitis or sensitization risk. 32 In the case of asthma, no evidence was found for a casual role of low 25OHD on asthma or childhood onset asthma risk in the study by Manousaki et al. 33 (quality score 7).   the role of linoleic acid in asthma and found an inverse association of genetically predicted linoleic acid with asthma. 36 Thus, linoleic acid may have a protective role towards asthma (although certainty of evidence was graded low, supplementary Table 3).

| Lifestyle factors
Six studies investigated the effect of lifestyle factors on the risk of asthma 14,37-41 ( Table 4)  examined the casual effect of smoking on adult asthma and hay fever, finding the investigated smoking-increasing allele to be associated with lower hay fever risk and higher adult asthma risk in current smokers. 40 With these conflicting results, the causal effect of smoking on asthma development remains inconclusive.
Bryan et al. (quality score 7) investigated the potential causal association between metabolism of arsenic, and asthma, in participants consuming high amounts of rice, finding no effect of arsenic on lifetime or current asthma risk but finding an effect in never smokers with past asthma diagnoses. 41

| Biomarkers and other potential risk factors
Different biomarkers have been investigated as causal risk factors for asthma and respiratory allergic diseases (Table 5) Table 3).
Two MR studies investigated the association between major depressive disorder (MDD) and adult asthma among other disease outcomes. 46,47 The MR analyses were performed either through phenome wide association study (PheWAS) followed by MR or through generalized summary data MR. Both found a causal association between MDD and adult asthma risk, pooled RR 1.23 (1.13-  allergic rhinitis. BTN3A2 belongs to the butyrophilin family of proteins, which has diverse function in the immune system, including immune modulation, which helps in establishing self-tolerance. Valette et al. (quality score 9) investigated 431 blood expressed genes through MR and identified 50 blood expressed genes that were causally associated to asthma. 49   Strong evidence was found for a causal association between high BMI and adult asthma resulting in an increased summary risk ratio for asthma of 1.05 (95%CI, 1.03-1.07). The effect of a high adult BMI increasing asthma risk is supported by previous studies, including a meta-analysis of prospective epidemiologic studies, finding that overweight and obesity are associated with incident asthma in a dose dependent manner. 54 The pathophysiological mechanism from high BMI to increased risk of asthma is still unclear; however, a possible mechanism could be obesity-related increased inflammation or immune response, but restrictive physiology due to mechanical factors is also plausible. In support of the latter, Çolak et al. 55    IV and the risk of atopy was found, but this association did not withstand sensitivity analysis and must be re-examined.
Different immune related proteins were found to have causal effects on asthma risk. Among these, a protective role of altered levels of circulating ST2 proteins for risk of asthma and allergic rhinitis was found. 42 The ST2 is a product of the IL1RL1 gene, a known asthma locus. 70,71 The variants included in this MR study comprise both asthma risk reducing and increasing variants. The ST2-IL33 complex promotes a pro-inflammatory type 2 response when membrane bound. 71 However, when the ST2 is in soluble form the T A B L E 5 Studies investigating for causal association between biochemical factors and asthma using MR effect on asthma and inflammation is unclear. Previous studies have shown that the levels of serum ST2 increase proportionally with the severity of asthma exacerbations and is generally higher in asthma patients, suggesting an effect on asthma risk. 72 Folkersen et al. 42 found an association between CASP-8 and high asthma risk.
Few other studies have investigated the effect of CASP-8 on asthma.
One intervention study in mice reported that CASP-8 can mediate an IL1 signal, which promotes a type 2 response and thereby increases pulmonary inflammation and allergic asthma 73 ; whether this effect is the same in humans remain unresolved.
In addition, a protective effect of BTN3A2 expressed in resting Tcells on asthma and allergic rhinitis was found. 48 The included MR studies investigating immune related proteins all use methods investigating multiple traits simultaneously through either PheWAS, quantitative trait loci, or multiple-trait co-localization. These methods can be used to investigate large numbers of exposure-outcome combinations in a hypothesis-free manner. However, significant findings often should be treated more provisionally and may benefit from replication in an independent dataset with a more pre-defined hypothesis.
This review and meta-analysis provides an overview of the cur- Effects of investigated risk factors for respiratory allergy followed the trends from asthma risk. MR studies investigating risk factors for asthma and respiratory allergy would help further clarify the causes of asthma and respiratory allergy, and help extend the results from this review.

ACKNOWLEDGMENTS
This work was supported by the Region Zealand Research Foundation, Novo Nordisk Foundation and Alpha-1 Foundation. The sponsors of the study are public or non-profit organizations and support science in general. They had no role in gathering, analysing, or interpreting the data and could neither approve nor disapprove the submitted manuscript.

CONFLICTS OF INTEREST
None of the authors have any competing interests to declare.

DATA AVAILABILITY STATEMENT
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.