The causality between chronic rhinosinusitis and inflammatory bowel disease: A bidirectional 2-sample Mendelian randomization analysis

Abstract

Prior observational studies suggested an association between chronic rhinosinusitis (CRS) and inflammatory bowel disease (IBD), but causality remains unclear. This study employed a bidirectional 2-sample Mendelian randomization (MR) analysis to investigate potential causal relationships. Genetic variants associated with CRS and IBD (including Crohn’s disease and ulcerative colitis) were sourced from the IEU Open genome-wide association study project. Primary analysis used inverse-variance weighted, supplemented by weighted median and MR-Egger methods. Sensitivity analyses included Cochran’s Q test, MR-Egger intercept, Mendelian Randomization Pleiotropy RESidual Sum and Outlier, and leave-one-out analysis. Inverse-variance weighted revealed CRS increased IBD risk (OR = 1.108, 95% confidence interval = 1.009–1.217, P = .032), while reverse MR indicated IBD increased CRS risk (OR = 1.035, 95% confidence interval = 1.001–1.071, P = .043). No significant causality was found between CRS and IBD subtypes, Crohn’s disease and ulcerative colitis. Sensitivity analyses supported robustness, with no evidence of pleiotropy (P > .05). This MR study suggests a bidirectional causal relationship between CRS and IBD, though not with subtypes. Further research is needed to elucidate underlying mechanisms.

1. Introduction

Inflammatory bowel disease (IBD) is a chronic and recurrent gastrointestinal inflammatory disease, caused by the interaction of multiple factors including host genetics, environmental factors, immune disorders, and the intestinal microbiota,^[1–4] mainly classified into Crohn’s disease (CD) and ulcerative colitis (UC) according to symptoms, endoscopic findings and pathological results.^[5] Currently, IBD has become a growing concern and serious public health problem as the increasing incidence and prevalence globally.^[6,7]

Chronic rhinosinusitis (CRS) is a common disease with chronic inflammation of the nose and paranasal sinuse characterized by symptoms including nasal obstruction, rhinorrhoea, smell loss and facial pressure or pain, and these symptoms last longer than 12 weeks,^[8] which severely impact the quality of life.^[9,10] Recent evidence has shown that epithelial barrier dysfunction and immune dysregulation play an important role in the development of CRS,^[11] which was similar to IBD.^[12] In addition, it has been briefly reported that a great improvement of UC after CRS treatment in some patients who suffered from both CRS and UC.^[13] Therefore, it is worthy of further studying the relationship between CRS and IBD. Actually, there are several studies having already evaluated the relationship between CRS and IBD,^[14–17] but the results are inconsistent. However, it is noteworthy that these evidence come only from observational studies, which are hard to infer causality and avoid the impact of confounding factors.^[18]

Mendelian randomization (MR) analysis is a novel epidemiological method for assessing the causality between an exposure and an outcome through using genetic variants as instrumental variables (IVs), which follow Mendel’s law of inheritance, and thus avoiding the reverse causation and the influence of confounding factors.^[18] Consequently, our study applied a bidirectional 2-sample MR analysis to better clarify the causality between CRS and IBD.

2. Materials and methods

This study was conducted in accordance with the STROBE-MR guidelines to ensure transparent reporting. For a causal interpretation of MR, the single nucleotide polymorphisms (SNPs) were selected as IVs and must satisfy 3 assumptions^[19]: SNPs must be strongly associated with the exposure; SNPs must not be influenced by confounders affecting the exposure–outcome relationship; SNPs affect the outcome solely through the exposure, with no horizontal pleiotropy. A schematic representation of the MR design is provided in Figure 1.

Figure 1.

Schematic diagram of the 2-sample Mendelian randomization hypothesis. CD = Crohn’s disease, CRS = chronic rhinosinusitis, IBD = inflammatory bowel disease, SNPs = single nucleotide polymorphisms, UC = ulcerative colitis.

2.1. Data source

As shown in Table 1, genetic variants and their associations with CRS and IBD were obtained from the IEU Open genome-wide association study (GWAS) project (https://gwas.mrcieu.ac.uk/datasets/), including CRS data (GWAS ID: finn-b-J10_CHRONSINUSITIS; N = 176,373; cases = 8524), IBD data (GWAS ID: ebi-a-GCST004131; N = 59,957; cases = 25,042), CD data (GWAS ID: ebi-a-GCST004132; N = 40,266; cases = 12,194), and UC data (GWAS ID: ebi-a-GCST004133; N = 45,975; cases = 12,366), all derived from European-ancestry populations to minimize population stratification bias.

Table 1.

Details of data sources used in this MR study.

Disease	Author and year	Sample	Case	Control	Population	GWAS ID
CRS	NA (2021)	1,76,373	8524	1,67,849	European	finn-b-J10_CHRONSINUSITIS
IBD	de Lange KM (2017)	59,957	25,042	34,915	European	ebi-a-GCST004131
CD	de Lange KM (2017)	40,266	12,194	28,072	European	ebi-a-GCST004132
UC	de Lange KM (2017)	45,975	12,366	33,609	European	ebi-a-GCST004133

CD = Crohn’s disease, CRS = chronic rhinosinusitis, GWAS = genome-wide association study, IBD = inflammatory bowel disease, UC = ulcerative colitis.

2.2. SNPs selection

To identify eligible SNPs as IVs, the following steps were carried out. Firstly, for SNP identification, we used a conventional threshold of P < 5 × 10⁻⁸ for IBD, CD, and UC, while adopting a relaxed threshold of P < 5 × 10⁻⁷ for CRS due to limited genome-wide significant SNPs.^[20] Secondly, to avoid the effect of linkage disequilibrium in SNPs, we set the clumping process with a distance of 10,000 kb and r² < 0.001.^[21] Thirdly, we harmonized exposure and outcome datasets to ensure consistent effect alleles and excluded palindromic and ambiguous SNPs with intermediate allele frequencies to prevent strand misalignment. Fourthly, we calculated the F statistic for indicating instrument strength of individual SNPs, and SNP was considered a strong instrument strongly associated with exposure when the F statistic was >10.^[22]

2.3. Statistical analysis

For our primary MR analysis, we employed inverse-variance weighted (IVW) regression as the main analytical approach. This method provides unbiased causal estimates under the assumption of no horizontal pleiotropy. We applied fixed-effects models when no heterogeneity was detected, while random-effects models were used in cases where heterogeneity was present. To complement these analyses and enhance robustness, we implemented supplementary methods including the weighted median method, which remains valid when ≥50% of the weight comes from valid SNPs, and MR-Egger method that accounts for directional pleiotropy through its intercept term.^[23,24] Besides, we conducted comprehensive sensitivity analyses to evaluate result reliability, beginning with heterogeneity assessment using Cochran’s Q test (P < .05 indicating significant heterogeneity and warranting random-effects IVW). Potential pleiotropy was examined through both MR-Egger intercept testing (where a nonzero intercept at P < .05 suggests bias) and Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR-PRESSO) for outlier SNP detection and correction.^[25] At last, leave-one-out analysis was performed to assess whether the possibility of result was driven by a single SNP.^[26] All statistical analyses were executed in R version 4.3.1, primarily utilizing the TwoSampleMR package for MR analyses and MR-PRESSO for outlier detection and correction.

3. Results

After SNP dataset filtering, we obtained all the detailed information on SNPs for this study, and all SNPs had an F-statistic >10, implying a low probability of weak SNPs. The detailed information of SNPs are presented in Tables S1–6, Supplemental Digital Content, https://links.lww.com/MD/P894.

3.1. Causal effects of CRS on IBD

As shown in Table 2, there were no significant heterogeneity observed across the all studies (all P-values for Cochran’s Q test > 0.05), and we used the IVW fixed-effects model and found genetically predicted CRS increased the risk of IBD (OR = 1.108, 95% CI = 1.009–1.217, P = .032), but MR-Egger method showed an opposite direction to IVW, indicating that the association was not robust (Fig. 2). In addition, there were no causal relationship between CRS and CD (OR = 1.054, 95% CI = 0.934–1.188, P = .394) or UC (OR = 1.125, 95% CI = 0.997–1.268, P = .055). Sensitivity analysis showed that there was no horizontal pleiotropy by MR-Egger regression and no SNP outliers indicated by MR-PRESSO analysis (Table 2). Moreover, leave-one-out plots showed that no single SNP drove the results (Figs. S1A–C, Supplemental Digital Content, https://links.lww.com/MD/P893). The scatter plots, funnel plots and forest plots were shown in Figures S2A–C, S3A–C, S4A–C, Supplemental Digital Content, https://links.lww.com/MD/P893.

Table 2.

Pleiotropy and heterogeneity test of MR analysis.

Expouse	Outcome	Cochran’s Q P-value	MR-Egger intercept P-value	MR-PRESSO global test P-value
CRS	IBD	.519	.198	.572
CRS	CD	.581	.358	.591
CRS	UC	.895	.64	.922
IBD	CRS	<.001	.303	<.0003
CD	CRS	<.001	.59	<.0003
UC	CRS	<.001	.278	<.0003

CD = Crohn’s disease, CRS = chronic rhinosinusitis, IBD = inflammatory bowel disease, MR = Mendelian randomization, MR-PRESSO = Mendelian Randomization Pleiotropy RESidual Sum and Outlier, UC = ulcerative colitis.

Figure 2.

Estimated causal effects of CRS on IBD and and its subtypes using different MR methods. CD = Crohn’s disease, CI = confidence interval, CRS = chronic rhinosinusitis, FE = fixed-effects, IBD = inflammatory bowel disease, IVW = inverse-variance weighted, MR = Mendelian randomization, OR = odds ratio, SNPs = single nucleotide polymorphisms, UC = ulcerative colitis, WM = weighted median.

3.2. Causal effects of IBD on CRS

As shown in Table 2, there was significant heterogeneity observed across the all studies (all P-values for Cochran’s Q test < 0.05), and we used the IVW random-effects model and found genetically predicted IBD increased the risk of CRS (OR = 1.035, 95% CI = 1.001–1.071 P = .043), and the weighted median and MR-Egger methods showed the same effect directions to IVW, indicating that the association was robust (Fig. 3). Both CD (OR = 1.021, 95% CI = 0.984–1.060, P = .261) and UC (OR = 0.974, 95% CI = 0.928–1.022, P = .278), we found no evidence for a causal relationship with CRS. Sensitivity analysis showed that there was no horizontal pleiotropy by MR-Egger regression, but there were outliers detected by MR-PRESSO analysis (Table 2). However, removal of these outliers did not significantly change the effect direction and statistical significance. Besides, leave-one-out plots showed that no single SNP drove the results (Fig. S1D–F, Supplemental Digital Content, https://links.lww.com/MD/P893). The scatter plots, funnel plots and forest plots were shown in Figs. S2D–F, S3D–F, S4D–F, Supplemental Digital Content, https://links.lww.com/MD/P893.

Figure 3.

Estimated causal effects of IBD and its subtypes on CRS using different MR methods. CD = Crohn’s disease, CI = confidence interval, CRS = chronic rhinosinusitis, FE =fixed-effects, IBD = inflammatory bowel disease, IVW = inverse-variance weighted, MR = Mendelian randomization, OR = odds ratio, SNPs = single nucleotide polymorphisms, UC = ulcerative colitis, WM = weighted median.

4. Discussion

To our knowledge, this study represents the first MR analysis to investigate the bidirectional causal relationship between CRS and IBD. Our findings suggested that genetically predicted CRS increases IBD risk (OR = 1.108, P = .032), while IBD conversely elevates CRS risk (OR = 1.035, P = .043). These findings offered novel insights into the shared pathophysiology of mucosal inflammatory disorders, though we observed no significant associations with IBD subtypes (CD or UC), likely due to limited statistical power or divergent disease mechanisms.

Previous observational studies have reported inconsistent associations between CRS and IBD. For instance, a retrospective study by Rai et al suggested that UC might predispose individuals to CRS, possibly as an extraintestinal manifestation of IBD.^[14] Similarly, Lin et al found that IBD patients, particularly those with UC, had an elevated risk of developing CRS.^[16] Conversely, Lee et al reported that CRS was associated with an increased incidence of UC but not CD,^[15] while other studies found no significant correlation.^[17,27] These discrepancies may stem from confounding factors inherent in observational designs, such as medication use, comorbidities, or environmental influences. Our MR analysis, by leveraging genetic variants as IVs, minimizes such biases and strengthens the evidence for a bidirectional causal link between CRS and IBD. The lack of association with UC or CD may reflect subtype-specific etiologies – for instance, UC’s mucosal-limited inflammation or CD’s granulomatous pathology – which may not uniformly interact with CRS-driven pathways. Importantly, sensitivity analyses confirmed the robustness of our primary findings, though residual pleiotropy cannot be entirely ruled out.

The observed bidirectional relationship may be mediated through several interconnected mechanisms: microbial dysbiosis, particularly Staphylococcus aureus superantigens from CRS that may impair gut barrier function^[13,28,29]; systemic inflammation propagated through shared cytokine networks (e.g., IL-6, TNF-α) that drive inflammatory responses across both systems^[11,12]; gut-lung axis interactions where microbial metabolites and immune cells traffic between these mucosal surfaces^[30]; and treatment-related effects whereby antibiotics used for CRS may disrupt gut microbiota homeostasis, potentially triggering IBD development.^[31,32] These pathways collectively create a vicious cycle of inflammation and barrier dysfunction that may sustain the CRS–IBD relationship.

Our study’s strengths include the MR design’s resistance to confounding, use of large European GWAS datasets, and comprehensive sensitivity analyses. However, limitations warrant consideration: the modest CRS sample size, potential population stratification, and possible residual pleiotropy. Notably, the lack of subtype associations may reflect true biological differences rather than methodological issues, as all MR methods consistently showed null effects.

To address these limitations and further validate our findings, future research should prioritize: larger-scale GWAS studies with enhanced statistical power; more diverse population cohorts to examine the generalizability of these associations across different ethnic groups; and comprehensive functional studies to elucidate the precise biological mechanisms underlying the observed bidirectional relationship. Such investigations would not only confirm our findings but also provide deeper insights into the shared pathophysiology of these chronic inflammatory conditions.

In summary, this MR study provided evidence for a bidirectional causal relationship between CRS and IBD, independent of confounding factors. While the exact mechanisms remain to be fully elucidated, our findings underscore the interconnected nature of chronic inflammatory diseases across mucosal surfaces. Further research is needed to translate these insights into preventive and therapeutic strategies.

Acknowledgments

We thank all the teams that contributed to the GWAS data involved in this study.

Author contributions

Data curation: Zan Liu, Junyu Huang.

Formal analysis: Zan Liu.

Investigation: Zan Liu.

Methodology: Zan Liu, Junyu Huang.

Resources: Zan Liu.

Software: Zan Liu.

Supervision: Junyu Huang.

Writing – original draft: Zan Liu.

Writing – review & editing: Junyu Huang.