Introduction

Chronic rhinosinusitis (CRS) is a complex chronic inflammatory disease of the sinonasal mucosa, defined by at least 3 months of nasal symptoms and objective evidence of chronic sinonasal mucosal inflammation [1, 2]. CRS can be classified into chronic rhinosinusitis with nasal polyps (CRSwNP) and chronic rhinosinusitis without nasal polyps (CRSsNP), and the main symptoms include nasal obstruction, nasal discharge, facial pain/pressure, and reduction or loss of sense of smell [3]. CRS significantly reduces patients’ quality of life and imposes a huge global economic and social burden [4, 5]. Therefore, it is necessary to investigate the potentially modifiable risk factors of CRS from an etiological standpoint and provide evidence for primary prevention and mechanism studies of CRS.

Recently, epidemiological studies have reported the association between smoking, allergic rhinitis (AR), asthma, and obesity and an increased incidence of CRS [6,7,8,9], symptoms of which can be aggravated by alcohol consumption [10]. However, observational research results are susceptible to residual confounders and reverse causality, which do not allow assertion on the causal effects of associations [11]. The causal relationships between these potential risk factors and CRS remain elusive and thus need verification. Although randomized controlled trials (RCTs) are the gold standard for causal inference, the implementation of this method is often limited by medical ethics, trial design, and research costs. It is impractical and unethical to conduct RCTs for potential risk factors and CRS. In this case, the Mendelian randomization (MR) study can overcome the above research defects and provide high-quality evidence for causal effects.

The MR study is an emerging method of investigating the causal effects of exposures (potential risk factors) on health or diseases [11]. It stimulates the design of RCTs using single-nucleotide polymorphisms (SNPs) associated with exposure as instrumental variables to investigate causal effects between exposures and disease risk [12]. It can provide reliable evidence for causal inference while coherently avoiding the effects of residual confounding and reverse causality in traditional observational studies [13]. MR studies have been gradually applied in clinical fields (such as heart diseases, diabetes, COVID-19, etc.) and have obtained reliable causal evidence. However, there is still a lack of MR studies related to CRS.

Therefore, we conducted this MR study to investigate the bidirectional causal relationships between CRS and potentially modifiable risk factors, including daily cigarette consumption, age of smoking initiation, weekly alcohol consumption, AR, asthma, and body mass index (BMI).

Materials and methods

Study design

We first identified the potential CRS-related risk factors that piqued our interest, including daily cigarette consumption, age of smoking initiation, weekly alcohol consumption, AR, asthma, and BMI. We evaluated the causal relationships between these six potential risk factors and CRS using a bidirectional two-sample MR analysis.

The theory behind the MR study design was that genetic variants are randomly assigned at meiosis and are independent of many other confounders, avoiding potential confounders and reverse causality and allowing reliable causal relationships to be inferred [14]. MR study design depended on the following assumptions: (1) the instrument is related with the exposure; (2) it influences the outcome only through the exposure; and (3) it is not associated with other confounders [11].

This report was based on the Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization (STROBE-MR) [15].

Data sources

The data of this study were drawn from results of the recent large-sample genome-wide association studies (GWAS) on people of European ancestry.

Data for CRS were obtained from the FinnGen consortium with 8524 CRS cases, obtained from hospital records (ICD-8, ICD-9, or ICD-10 codes), and 167,849 noncases. Data for daily cigarette consumption were obtained from the GWAS and Sequencing Consortium of Alcohol and Nicotine use (GSCAN) with 337,334 participants and were based on self-reports. Data for age of smoking initiation were obtained from the GSCAN with 341,427 participants and were based on self-reports. Data for weekly alcohol consumption were obtained from the GSCAN with 335,394 participants and were based on self-reports. Data for AR were obtained from the UK Biobank (UKB) with 25,486 AR cases and 87,097 noncases. The AR cases were identified from the doctor-diagnosed cases: those who answered “Yes” to the question, “Has a doctor ever told you that you have hay fever or allergic rhinitis?” were classified as AR cases. Data for asthma were obtained from the European Molecular Biology Laboratory’s European Bioinformatics Institute (EMBL–EBI) with 56,167 asthma cases and 352,555 noncases. The asthma cases were identified from the hospital records (ICD-9 or ICD-10 codes) or primary-care medical records as well as those with self-reported asthma. Data for BMI were obtained from the Within family GWAS consortium with 99,998 participants. All these original studies obtained the ethical approvals, and informed consent of all participants was collected by original researchers.

Instrumental variables (genetic variants) selection

According to the basic assumptions of the MR study, genetic variants (SNPs) associated with exposure and not associated with other confounders and outcome (except through the exposure) were selected as instrumental variables. SNPs associated with each exposure and not associated with the outcome were selected as instrumental variables at the genome-wide significance threshold of p < 5 × 10–8 when CRS was identified as the outcome. SNPs associated with the exposure and not associated with each outcome were selected as instrumental variables at the genome-wide significance threshold of p < 5 × 10–6 when CRS was identified as the exposure, since fewer SNPs reached the genome-wide significance levels of p < 5 × 10–8. We excluded all SNPs with palindromic structures and high linkage disequilibrium (r2 > 0.001). The MR pleiotropy residual sum and outlier (MR-PRESSO) test was used to detect and correct for horizontal pleiotropic outliers [16]. When instrumental variables (SNPs) independently associated with the exposure were detected in the outcome database using MR statistical methods, the exposure was assumed to have a causal effect on the outcome.

Statistics analysis

Inverse-variance weighted (IVW) method with random effects was applied as the major analysis method to estimate causal effects, which effectively overcomes the impact of heterogeneity on causal effects [17]. Weighted median (WM) and MR–Egger methods were applied for sensitivity analyses in addition to IVW [18]. When valid instrumental variables account for > 50% of the weight in the analysis, the WM method provides a reliable assessment of the causal effects between the exposure and the outcome [18]. Horizontal pleiotropy can be detected and corrected by MR–Egger method, albeit the accuracy of estimation is low [19]. We used Cochran Q statistics to assess heterogeneity, and horizontal pleiotropy was defined using the p value for the intercept of the MR–Egger model [17]. In case of no horizontal pleiotropy, the IVW method evaluation results with random effects were used as the main causal effects. In case of horizontal pleiotropy, the MR–Egger method evaluation results were used as the main causal effects. We used leave-one-out sensitivity analysis to reassess the overall causal effects by removing one SNP in succession each time.

Causal effects of exposures on outcomes were described using odds ratio (OR) and 95% confidence intervals (95% CI). To further ensure the reliability of our findings, the Bonferroni adjusted p value of significance was < 0.004 (0.05/12 = 0.004). In this study, p < 0.004 was considered as strong evidence of causal effects of exposure on outcome, whereas 0.004 < p < 0.05 was considered suggestive evidence of causal effects of exposure on outcome.

The R software version 4.1.3 (R Foundation, Vienna, Austria) was used for the selection of instrumental variables, statistics analysis, and visualization of our findings.

Results

In this study, we investigated the bidirectional causal relationships between daily cigarette consumption, age of smoking initiation, weekly alcohol consumption, AR, asthma, BMI, and CRS through the bidirectional two-sample MR analysis.

The instrumental variables we selected were at different loci and had no linkage disequilibrium (r2 < 0.001), and they were independently associated with each exposure at the genome-wide significance threshold. The F statistic of each instrumental variable we selected exceeded the threshold value of 10, indicating sufficient statistical power and effectiveness. Supplementary Tables S1 through S12 demonstrate the complete instrumental variables (SNPs) results for each trait.

MR analysis revealed suggestive causal relationships between daily cigarette consumption, AR, BMI, and CRS in forward direction (IVW, 0.004 < p < 0.05) and a significant bidirectional causal relationship between asthma and CRS (IVW, p < 0.004). The results of IVW, WM, and MR–Egger analyses are presented in Tables 1 and 2.

Table 1 Causal effects of potential risk factors on chronic rhinosinusitis
Full size table
Table 2 Causal effects of chronic rhinosinusitis on potential risk factors
Full size table

Causal relationships between potential risk factors and CRS in forward direction

We observed that daily cigarette consumption, AR, and BMI had suggestive causal effects on CRS, and asthma had a significant causal effect on CRS. However, no causal effect of age of smoking initiation and weekly alcohol consumption was observed on CRS. As presented in Fig. 1, the risk of CRS increased with the increase in daily cigarette consumption (IVW, OR = 1.15, 95% CI 1.00−1.32, p = 0.046). Figure 3a depicts the relationship between daily cigarette consumption and the risk of CRS. The MR–Egger model revealed no evidence of horizontal pleiotropy for the causal effect of daily cigarette consumption on CRS (intercept p = 0.201). Leave-one-out sensitivity analysis revealed that no individual SNP significantly affected the causal effect of daily cigarette consumption on CRS (Fig. S1).

Fig. 1
figure 1

Causal effects of potential risk factors on chronic rhinosinusitis. CI confidence interval, OR odds ratio

Full size image

The risk of CRS increased with AR (IVW, OR = 4.77, 95% CI 1.61 − 14.13, p = 0.005). Figure 3b depicts the relationship between AR and the risk of CRS. The MR-Egger model revealed no evidence of horizontal pleiotropy for the causal effect of AR on CRS (intercept p = 0.702). Leave-one-out sensitivity analysis revealed that no individual SNP significantly affected the causal effect of AR on CRS (Fig. S2).

The risk of CRS increased with asthma (IVW, OR = 1.45, 95% CI 1.31 − 1.60, p < 0.001). Figure 3c depicts the relationship between asthma and the risk of CRS. MR–Egger model revealed no evidence of horizontal pleiotropy for the causal effect of asthma on CRS (intercept p = 0.202). Leave-one-out sensitivity analysis revealed that no individual SNP significantly affected the causal effect of asthma on CRS (Fig. S3).

The risk of CRS increased with the increase in BMI (IVW, OR = 1.05, 95% CI 1.00 − 1.09, p = 0.028). Figure 3d depicts the relationship between BMI and the risk of CRS. The MR–Egger model revealed no evidence of horizontal pleiotropy for the causal effect of BMI on CRS (intercept p = 0.309). Leave-one-out sensitivity analysis revealed that no individual SNP significantly affected the causal effect of BMI on CRS (Fig. S4).

Causal relationships between potential risk factors and CRS in backward direction

We observed that CRS had a significant causal effect on asthma. No causal effect of CRS on daily cigarette consumption, age of smoking initiation, weekly alcohol consumption, AR, and BMI was observed. As presented in Fig. 2, the risk of asthma increased with CRS (IVW, OR = 1.08, 95% CI 1.05−1.12, p < 0.001). Figure 3e depicts the relationship between CRS and the risk of asthma. MR–Egger model revealed no evidence of horizontal pleiotropy for the causal effect of CRS on asthma (intercept p = 0.506). Leave-one-out sensitivity analysis revealed that no individual SNP significantly affected the causal effect of CRS on asthma (Fig. S5).

Fig. 2
figure 2

Causal effects of chronic rhinosinusitis on potential risk factors. CI confidence interval, OR odds ratio

Full size image
Fig. 3
figure 3

Causal relationships between potential risk factors and chronic rhinosinusitis. a Causal effect of daily cigarette consumption on chronic rhinosinusitis. b Causal effect of allergic rhinitis on chronic rhinosinusitis. c Causal effect of asthma on chronic rhinosinusitis. d Causal effect of body mass index on chronic rhinosinusitis. e Causal effect of chronic rhinosinusitis on asthma. AR allergic rhinitis, BMI body mass index, CRS chronic rhinosinusitis, SNP single-nucleotide polymorphism, MR Mendelian randomization

Full size image

Discussion

This MR study demonstrated that daily cigarette consumption, AR, and BMI had suggestive causal effects on CRS and revealed a significant bidirectional causal relationship between asthma and CRS. To our knowledge, this is the first MR study to explore bidirectional causal relationships between potentially modifiable risk factors and CRS.

A systematic review by Alkholaiwi et al. presented that cigarette consumption increased the prevalence of CRS [7]. Studies have suggested a potential dose-dependent relationship between smoking and self-reported CRS [20], and smoking negatively affected the structure and function of the sinonasal mucosa [21, 22]. Our findings suggested that daily cigarette consumption had a causal effect on CRS. Smoking has been reported to increase the expression of oxidative stress and inflammatory factors [23]. In addition, studies showed that smoking was associated with increased systematic expression of IL (interleukin)-4, IL-5, and IL-13 [24, 25]. As important type 2 cytokines, IL-4, IL-5, and IL-13 play important roles in the pathogenesis of CRS, especially CRSwNP, which could be an important mechanism for the causal effect of smoking on CRS. Furthermore, sinonasal mucosa cilia injury, epithelial cell apoptosis, Staphylococcus aureus colonization and biofilm formation could also participate in the causal effect of smoking on CRS [26,27,28,29].

AR is a common comorbidity in patients with CRS. Studies reported that the positivity rate of skin prick test in patients with CRS was higher than that in control group, and patients with AR had an increased incidence of CRS [30,31,32]. Data from Geisinger Clinic primary care patients indicated a higher incidence of premorbid AR in patients with CRS [6]. Our findings suggested that AR had a causal effect on CRS, which may be through the following mechanisms: (1) the swelling of sinonasal mucosa, especially ostiomeatal complex, owing to AR leads to obstruction of the sinus orifice, causing obstructive inflammation and bacterial growth [31]. (2) AR decreases the ciliary clearance and causes ciliary loss in sinonasal mucosa [33]. (3) AR causes the increase in eosinophils, mast cells, and type 2 inflammatory factors in the nasal and sinus mucosa, participating in the occurrence of CRS, especially CRSwNP [34].

Several studies showed that asthma was closely associated with CRS. (1) A growing number of studies have reported the relation between asthma and CRS in people of all age groups [6, 8, 35]. The association was bidirectional, suggesting increased incidences of CRS in patients with asthma and asthma in patients with CRS. (2) Nasal symptoms of CRS patients with asthma were more severe and sinus radiological severity increased [36]. (3) CRS affected the severity of asthma, and treatment of CRS can reduce asthma attacks and improve symptoms [37]. However, observational studies often have many confounders that are difficult to remove, and the association of the two diseases is susceptible to reverse causality and genetic predispositions, making it hard to obtain causal inference between them. Our study provided robust evidence for the bidirectional causal relationship between asthma and CRS. With the advancement of research into airway diseases, united airways disease is becoming more widely recognized [38, 39], it is believed that there is a dynamic association between the upper and the lower airways, and respiratory inflammation may begin at the site of onset and gradually spread to other sites. It can also induce a systemic immune response by stimulating the bone marrow to produce inflammatory cells and mediators, causing a non-contact association. Based on the important roles of type 2 cytokines (e.g. IL-4, IL-5, IL-13, etc.) in the development and progression of asthma and type 2 CRS, these cytokines could be an important bridge between the two diseases in a bidirectional causal role. Furthermore, CRS may also have a causal effect on asthma through naso-bronchial reflex, respiratory pattern change, postnasal drip, etc.

A cross-sectional analysis of 229 million U.S. adults revealed the association between the increase in BMI and the increased incidence of CRS [9]. A prospective population-based study demonstrated that the incidence of new CRS was 53% higher in the obese group (BMI ≥ 30) than in the normal weight group (18.5 ≤ BMI < 25) in the Norwegian population, and higher BMI was associated with an increased risk of CRS [40]. Our study also suggested that BMI had a causal effect on CRS. Studies have revealed that obesity can cause metabolic inflammation, leading to the body being in a state of systemic low inflammation [41]. Obesity causes airway inflammation, decreased lung function, and aggravated asthma through the adipocytokines, leptin, adiponectin, other inflammatory cytokines, and mediators induced by adipose tissue [42]. According to the united airways disease theory, obesity may participate in the pathogenesis of CRS through the above-mentioned mechanisms, causing a causal effect on CRS.

Studies have reported that alcohol consumption caused upper respiratory reactions, particularly nasal obstruction and discharge, and patients with CRS often reported worsening of nasal symptoms following alcohol consumption [10, 43]. We did not observe a causal relationship between weekly alcohol consumption and CRS in either direction. Although an epidemiological survey reported that patients with CRSwNP consume more alcohol than healthy controls [44], this may be owing to uncontrollable confounders, such as increased smoking and relatively poor hygiene in patients who drink heavily. Nasal symptoms after alcohol consumption may be caused by non-specific mechanisms, such as alcohol-induced temporary vasodilation.

Although previous epidemiological studies have found associations between smoking, AR, asthma, obesity, and CRS, however, observational results have limited validity in causal inference, and are susceptible to confounders that are difficult to eliminate, making it difficult to assess causal effect sizes [12]. For example, CRS, AR, and asthma are both chronic airway diseases, the upper and lower airways are exposed to the same respiratory irritants, such as occupational exposures and environmental pollution, which may have an impact on causal effects; similar genetic predispositions between them could also influence causal inferences. Besides, smoking, drinking, and BMI often imply different occupations, health and economic conditions, and lifestyle habits, which could have an impact on the epidemiological results. Furthermore, the cross-sectional design does not allow assertion on the causal direction of associations. For the association between BMI and CRS, cross-sectional studies make it difficult to identify whether high BMI impacts CRS or BMI is impacted by CRS. CRS and asthma are also the association susceptible to bidirectional causal effects in observational studies. As an emerging research method, MR studies could avoid the limitations of the aforementioned studies and provide high-quality evidence for causality when RCTs are difficult to conduct [11]. Compared to observational studies, our MR study obtained more reliable evidence for causal effects between potential risk factors and CRS, and in addition, we have successfully assessed the causal effect sizes of these associations, which is necessary for prevention and etiological studies of CRS.

Furthermore, this study had the following strengths: we applied the two-sample MR study to investigate bidirectional causal relationships between potentially modifiable risk factors and CRS. MR studies can avoid confounders that are impossible to measure in observational studies and reverse causality, providing evidence for causal inference. We carefully checked the MR assumptions, and the data used in our study were drawn from suitable GWAS data sources with large sample sizes. The F statistic of each SNP as an instrumental variable exceeded the threshold value of 10, ensuring sufficient statistical power. We investigated causal effects using the IVW method with random effects, which effectively overcomes the impact of heterogeneity. The WM and MR–Egger methods were used as the sensitivity analyses to test the consistency of the results. In addition, the MR–PRESSO test and MR–Egger regression analysis were used to detect and correct for horizontal pleiotropy.

There were several limitations: (1) the major limitation of MR studies is that pleiotropy may affect the reliability of causal inference, although horizontal pleiotropy was not detected in our study using the MR–Egger model. (2) When evaluating the causal effect of BMI on CRS, the MR–Egger method displayed the opposite effect as the IVW and WM methods, but the MR–Egger method was not statistically significant. Since horizontal pleiotropy was not observed in the MR–Egger model, the IVW method was considered effective; however, the findings should still be explained with caution. (3) Owing to the limitation of GWAS data, we only evaluated the bidirectional causal relationships between potential risk factors and CRS and did not differentiate between CRSwNP and CRSsNP and different inflammatory endotypes. (4) This study relied on data from adults of European descent. Further research on other ethnic groups is needed. Causal inference of this study may not be applicable to children.

Conclusions

Conclusively, we demonstrated and measured the causal effects of daily cigarette consumption (OR = 1.15), AR (OR = 4.77), asthma (OR = 1.45), and BMI (OR = 1.05) on CRS, and the causal effect of CRS on asthma (OR = 1.08). However, we found no evidence supporting the causal relationship between age of smoking initiation, weekly alcohol consumption, and CRS. These findings are expected to provide high-quality causal evidence for clinical practice of rhinologists and provide a causal inference basis for the pathogenesis and prevention of CRS and asthma.