Bidirectional Mendelian randomization analysis of the causal relationship between inflammatory bowel disease and Hashimoto thyroiditis

Abstract

This study aimed to explore the bidirectional causal relationship between inflammatory bowel disease (IBD) and Hashimoto thyroiditis (HT) using Mendelian randomization (MR) analysis. While both conditions are characterized by persistent inflammation and immune dysregulation, the direction of causality remains unclear. We performed a 2-sample bidirectional MR analysis using summary statistics from genome-wide association studies for IBD and HT. Genetic variants strongly associated with IBD and HT were selected as instrumental variables for forward and reverse MR analyses, respectively. Various MR methods, including inverse-variance weighted (IVW), MR-Egger regression, weighted median, and weighted mode, were employed to assess causal relationships. Sensitivity analyses were conducted to evaluate the reliability of results, including tests for pleiotropy and heterogeneity. Data were sourced from individuals of European descent to minimize population stratification bias. In the forward MR analysis, no strong evidence of a causal effect of IBD on HT was found, with the IVW method yielding an odds ratio (OR) of 0.9617 (95% CI: 0.7549–1.2251; P = .7519) Similarly, in the reverse MR analysis, no significant causal effect of HT on IBD was observed, with the IVW method showing an OR of 0.9991 (95% CI: 0.9693–1.0298, P = .9527). Sensitivity analyses confirmed the absence of heterogeneity or pleiotropic effects that could influence the results. The MR-PRESSO analysis did not detect any outlier SNPs. This bidirectional MR analysis provides no evidence for a causal relationship between IBD and HT in either direction.

1. Introduction

Inflammatory bowel disease (IBD) exhibits considerable geographical heterogeneity in incidence. In Europe, the annual incidence ranges from approximately 10.5 to 46.14 cases per 100,000 population, whereas significantly lower rates, around 1.37 to 1.5 per 100,000, are reported in Asia and the Middle East.^[1] However, a notable upward trend has been observed in newly industrialized and developing regions, including Asia, Latin America, and Africa, suggesting a progressive global emergence and dissemination of IBD.^[2,3] An abnormal immune response to intestinal microbiota in genetically predisposed individuals underlies the pathogenesis of IBD. Both innate and adaptive immune dysfunctions, notably involving Th17 cells and their associated cytokines, as well as macrophages, dendritic cells, and neutrophils, contribute to persistent intestinal inflammation and subsequent tissue injury.^[4,5] The dynamic interaction between the host immune system and the gut microbiota plays a pivotal role in both the initiation and progression of the disease.^[4,5]

Hashimoto thyroiditis (HT) is a prevalent autoimmune thyroid disorder, with an estimated global adult prevalence of approximately 7.5%.^[6] Notably, prevalence rates show substantial geographical variation, being highest in Africa (14.2%), followed by Oceania (11.0%), South America and Europe (8.0%), and North America (7.8%), with the lowest rates observed in Asia (5.8%).^[6] In addition, regions classified as low- to middle-income exhibit a higher prevalence (11.4%) compared to upper-middle-income (5.6%) and high-income (8.4%) regions.^[6] HT arises from a complex interplay of genetic susceptibility and environmental triggers that compromise immune tolerance to thyroid-specific antigens.^[7] This breakdown initiates the activation of autoreactive T and B lymphocytes, the generation of thyroid autoantibodies, and a cytokine-driven inflammatory cascade. Subsequent immune-mediated apoptosis of thyroid follicular epithelial cells leads to the development of hypothyroidism.^[7,8] The immunopathogenesis of HT is characterized by coordinated disruptions in both cellular (T cell and cytokine-mediated) and humoral (autoantibody-mediated) immunity, alongside local thyroidal alterations such as ectopic lymphoid neogenesis and aberrant antigen presentation.^[7–9] Collectively, these mechanisms underpin the chronic, progressive course and clinical presentation of the disease.

Although the coexistence of IBD and HT is relatively uncommon,^[10] both conditions are characterized by persistent inflammation and dysregulated immune activation, implying the possibility of shared pathogenic mechanisms. Genetic predisposition and immunological dysregulation are thought to play contributory roles in the concomitant development of these disorders.^[11] However, the causal relationship between IBD and HT remains unclear. Although observational studies have hinted at potential associations,^[11,12] they are often confounded by environmental factors, reverse causality, and other biases inherent in traditional epidemiological approaches. In this context, Mendelian randomization (MR) offers a powerful alternative method to infer causality by leveraging genetic variants as instrumental variables (IVs). Since alleles are randomly assorted during meiosis, MR analyses are less susceptible to confounding and reverse causation, thus providing more robust evidence regarding the direction and strength of associations between traits.

To date, no comprehensive bidirectional MR analysis has been conducted to systematically evaluate the causal relationship between IBD and HT. Elucidating this relationship may inform clinical management strategies, including risk stratification and early screening in high-risk individuals. This study aims to employ a 2-sample bidirectional MR approach to explore the potential causal association between IBD and HT, thereby providing novel evidence to guide future research and clinical practice.

2. Materials and methods

2.1. Study design

This study adhered to the STROBE-MR guidelines, which are designed to strengthen the methodological rigor of epidemiological research utilizing MR approaches.^[13] All datasets utilized in this study were obtained from publicly available genome-wide association studies (GWAS) repositories, thereby exempting the requirement for additional ethical approval. In MR analyses, single nucleotide polymorphisms (SNPs) employed as IVs must satisfy 3 fundamental assumptions: assumption 1: relevance-genetic instruments are strongly associated with the exposure; assumption 2: independence-genetic instruments are not associated with confounders; assumption 3: exclusion restriction-genetic instruments affect the outcome solely through the exposure of interest (Figure S1, Supplemental Digital Content, https://links.lww.com/MD/P777).

2.2. Sources of GWAS summary statistics

All summary-level data for both cases and controls were sourced from individuals of European descent. Participants were recruited from multiple research centers across Europe, which helped to reduce potential biases arising from population stratification and ancestry-related confounding.

2.3. Exposure data acquisition

The genetic association data for IBD were obtained from the FinnGen consortium (release 12). The GWAS summary statistics (GWAS ID: finngen_R12_K11_IBD_STRICT) included a total of 500,348 participants, comprising 10,960 cases and 489,388 controls, all of European ancestry. A total of 21,327,062 SNPs were analyzed. The summary-level data were accessed via the public FinnGen repository (https://storage.googleapis.com/finngen-public-data-r12/summary_stats/release/finngen_R12_K11_IBD_STRICT.gz) (Table S1, Supplemental Digital Content, https://links.lww.com/MD/P778).

2.4. Outcome data acquisition

The genetic association data for HT were extracted from the GWAS Catalog (Study ID: GCST90435698). This dataset included summary statistics from 391,639 individuals of European ancestry. A total of 28,184,541 SNPs were analyzed. Detailed information and full summary statistics are publicly available through the GWAS Catalog (https://www.ebi.ac.uk/gwas/studies/GCST90435698). The use of participants of predominantly European descent helped to reduce the potential for confounding due to population stratification (Table S1, Supplemental Digital Content, https://links.lww.com/MD/P778).

2.5. Selection and validation of instrumental variables

The identification of robust IVs is a fundamental prerequisite in MR analyses. For the forward MR, SNPs strongly associated with IBD were selected from the exposure dataset using a genome-wide significance threshold (P < 5 × 10⁻⁸). In the reverse MR analysis, SNPs associated with HT at a less stringent significance level (P < 5 × 10⁻⁸) were selected. Subsequently, linkage disequilibrium clumping was performed using the “clump_data” function with parameters set to r² < 0.001 and a window size of 10,000 kb to eliminate correlated variants and remove non-biallelic SNPs. All SNPs associated with the exposure were identified through the PhenoScanner database (http://www.phenoscanner.medschl.cam.ac.uk/), and those showing associations with confounding variables or outcomes (P < 1 × 10⁻⁵) were subsequently removed to reduce the risk of horizontal pleiotropy.

To assess the strength of the selected IVs, F-statistics were calculated according to the formula F = (β/se)², with an F-statistic >10 considered indicative of sufficiently strong instruments, thereby minimizing weak instrument bias. Furthermore, the “harmonise_data” function was utilized to align effect alleles across exposure and outcome datasets, resolve strand ambiguities, remove palindromic SNPs, and retain only those SNPs with consistent orientation (“TRUE” harmonization) for subsequent MR analysis.

2.6. Bidirectional 2-sample Mendelian randomization analysis

Data analyses were performed using the “TwoSampleMR” package within R software. To estimate the causal effect between the exposure and outcome, multiple complementary MR methods were applied, including inverse-variance weighted (IVW), MR-Egger regression, weighted median, simple mode, and weighted mode approaches, with IVW regarded as the primary and most robust method. Subsequently, sensitivity analyses were conducted to evaluate the reliability of the findings. Cochran Q-statistic was utilized to assess heterogeneity among the IVs within both the IVW and MR-Egger frameworks, with a P-value <.05 indicating significant heterogeneity and informing the stability of the causal estimates.^[14] Furthermore, a funnel plot of the IVs was generated to visually assess heterogeneity. The MR-PRESSO method was employed to detect potential outlier SNPs that might introduce bias into the analysis and to estimate the causal effect after correction by excluding these outliers.^[15] In cases where MR-PRESSO detected outlier SNPs, these variants were removed, and the MR analysis was subsequently re-performed. Horizontal pleiotropy, defined as the influence of IVs on the outcome through pathways independent of the exposure, was recognized as a potential source of bias that could compromise the validity of the inferred causal relationship.^[16] To assess and quantify the presence of horizontal pleiotropy, the MR-Egger intercept test was employed. A P-value <.05, as determined by the “MR_pleiotropy_test” function, was considered indicative of significant directional pleiotropy.^[17] Finally, a leave-one-out analysis was conducted to assess the impact of each individual SNP on the outcome. This was achieved by iteratively excluding each SNP and re-running the IVW analysis. This approach allows for the identification of any single SNP that may drive the observed causal relationship.

3. Results

3.1. Selection and assessment of instrumental variables

In the forward MR analysis, we identified 53 SNPs with a strong association to IBD. In the reverse MR analysis, 6 SNPs were found to be associated with HT as the exposure related to IBD. All of these SNPs met the criteria for strong IVs (F-statistic > 10). We queried the PhenoScanner database to assess potential associations of these SNPs with confounders or outcomes (P < 1 × 10⁻⁵); however, no SNPs were excluded based on this search. A detailed summary of the selected IVs is provided in Tables S2 and 3, Supplemental Digital Content, https://links.lww.com/MD/P778.

3.2. Forward Mendelian randomization analysis: impact of IBD on HT

We performed a forward MR analysis to assess the potential causal effect of IBD on HT. A total of 53 SNPs with F-statistics >10 were used as IVs for IBD. The results of the MR analysis, conducted using various methods, are summarized in Table 1.

Table 1.

Two-sample MR analysis evaluating the causal association between IBD and HT, along with results from corresponding sensitivity tests.

Forward MR	SNP	F	MR				Heterogeneity
			Method	OR	95% CI	P	Q	Q_P
Exposure: inflammatory bowel disease, Outcome: HT	53	>10	MR-Egger	0.6578	0.3475–1.2452	.2041	47.7580	0.6032
			Weighted median	0.8207	0.5659–1.1901	.2974	–	–
			IVW	0.9617	0.7549–1.2251	.7519	49.3478	0.5788
			Simple mode	0.9576	0.4757–1.9277	.9038	–	–
			Weighted mode	0.7735	0.4816–1.2425	.2931	–	–
			MR-PRESSO	–	–	.7468	–	–
			Pleiotropy	–	–	.2131	–	–

CI = confidence interval, HT = Hashimoto thyroiditis, IBD = inflammatory bowel disease, IVW = inverse-variance weighted, MR = Mendelian randomization, MR-PRESSO = Mendelian randomization pleiotropy residual sum and outlier, OR = odds ratio, SNP = single nucleotide polymorphism.

The MR-Egger method yielded an odds ratio (OR) of 0.6578 (95% CI: 0.3475–1.2452, P = .2041), suggesting no strong evidence of a causal effect between IBD and HT. Similarly, the weighted median approach produced an OR of 0.8207 (95% CI: 0.5659–1.1901, P = .2974), further supporting the null hypothesis. The IVW method showed an OR of 0.9617 (95% CI: 0.7549–1.2251, P = .7519), indicating no significant causal relationship (Fig. 1). Additional sensitivity analyses using the simple mode and weighted mode methods revealed ORs of 0.9576 (95% CI: 0.4757–1.9277, P = .9038) and 0.7735 (95% CI: 0.4816–1.2425, P = .2931), respectively, further corroborating the lack of evidence for a causal effect (Figure S2, Supplemental Digital Content, https://links.lww.com/MD/P777).

Figure 1.

Forward MR analysis assessing the causal effect of IBD on HT. Scatter plots display the associations between genetic instruments for IBD and HT across 5 MR methods: MR-Egger, weighted median, IVW, simple mode, and weighted mode. The slopes of the lines represent the estimated causal effects under each method. HT = Hashimoto thyroiditis, IBD = inflammatory bowel disease, IVW = inverse-variance weighted, MR = Mendelian randomization.

The MR-PRESSO analysis did not detect any outliers, and the pleiotropy test revealed a P-value of .2131, suggesting no substantial pleiotropic effects influencing the results. Heterogeneity between SNPs, assessed using Cochran Q-statistic, was observed with a Q value of 47.7580 (P = .6032), indicating no significant heterogeneity in the IVs (Figure S3, Supplemental Digital Content, https://links.lww.com/MD/P777).

In summary, the forward MR analysis does not provide strong evidence to support a causal association between IBD and HT (Fig. 2).

Figure 2.

Forest plot of the estimated causal effects from the forward MR analysis (IBD → HT). Each SNP is shown with its corresponding odds ratio (OR) and 95% CI, as calculated using the IVW method. Summary estimates from multiple MR methods are also included for comparison. 95% CI = 95% confidence interval, HT = Hashimoto thyroiditis, IBD = inflammatory bowel disease, IVW = inverse-variance weighted, OR = odds ratio, SNP = single nucleotide polymorphism.

3.3. Reverse Mendelian randomization analysis: impact of HT on IBD

We conducted a reverse MR analysis to investigate the causal effect of HT on IBD. Six SNPs with F-statistics >10 were selected as IVs for HT. The results of the MR analysis using different methods are presented in Table 2.

Table 2.

Assessment of the causal effect of HT on IBD using 2-sample MR, accompanied by relevant sensitivity analyses.

Reverse MR	SNP	F	MR				Heterogeneity
			Method	OR	95% CI	P	Q	Q_P
Exposure: Hashimoto thyroiditis, outcome: Inflammatory bowel disease	6	>10	MR-Egger	1.0031	0.9489–1.0603	.9193	11.0089	0.0265
			Weighted median	0.9956	0.9687–1.0233	.7546	–	–
			IVW	0.9991	0.9693–1.0298	.9527	11.0944	0.0495
			Simple mode	0.9956	0.9576–1.0352	.8333	–	–
			Weighted mode	0.9853	0.9570–1.0145	.3661	–	–
			MR-PRESSO	–	–	.9550 (outlier: rs7308811)	–	–
			Pleiotropy	–	–	.8686	–	–

CI = confidence interval, HT = Hashimoto thyroiditis, IBD = inflammatory bowel disease, IVW = inverse-variance weighted, MR = Mendelian randomization, MR-PRESSO = Mendelian randomization pleiotropy residual sum and outlier, OR = odds ratio, SNP = single nucleotide polymorphism.

The MR-Egger method yielded an OR of 1.0031 (95% CI: 0.9489–1.0603, P = .9193), suggesting no significant causal effect of HT on IBD. Similarly, the weighted median approach produced an OR of 0.9956 (95% CI: 0.9687–1.0233, P = .7546), further supporting the null hypothesis. The IVW method showed an OR of 0.9991 (95% CI: 0.9693–1.0298, P = .9527), providing additional evidence that HT is not causally associated with IBD. Other methods, including simple mode (OR = 0.9956, 95% CI: 0.9576–1.0352, P = .8333) and weighted mode (OR = 0.9853, 95% CI: 0.9570–1.0145, P = .3661), yielded similar results, reinforcing the absence of a causal relationship (Fig. 3).

Figure 3.

Reverse MR analysis evaluating the causal effect of HT on IBD. Scatter plots illustrate the relationship between genetic instruments for HT and risk of IBD, using MR-Egger, weighted median, IVW, simple mode, and weighted mode methods. HT = Hashimoto thyroiditis, IBD = inflammatory bowel disease, IVW = inverse-variance weighted, MR = Mendelian randomization.

The MR-PRESSO analysis did not identify any outliers, and the pleiotropy test revealed a P-value of .8686, indicating no substantial pleiotropic effects on the results (Figure S4, Supplemental Digital Content, https://links.lww.com/MD/P777). Heterogeneity between SNPs, assessed using Cochran Q-statistic, was observed with a Q value of 11.0089 (P = .0265), indicating some level of heterogeneity in the IVs, but this did not affect the overall interpretation (Figure S5, Supplemental Digital Content, https://links.lww.com/MD/P777). In conclusion, the reverse MR analysis did not provide strong evidence to support a causal effect of HT on IBD (Fig. 4).

Figure 4.

Forest plot of the estimated causal effects from the reverse MR analysis (HT→IBD). Causal estimates and 95% CIs are shown for each SNP and MR method. Results provide insight into the potential reverse causality. 95% CIs = 95% confidence intervals, HT = Hashimoto thyroiditis, IBD = inflammatory bowel disease, IVW = inverse-variance weighted, MR = Mendelian randomization, OR = odds ratio, SNP = single nucleotide polymorphism.

4. Discussion

4.1. Key findings overview

In this bidirectional MR study, we examined the causal relationship between IBD and HT using genetic data from large European cohorts. Our findings suggest that there is no substantial evidence to support a causal relationship between IBD and HT in either direction.

Forward MR Analysis (IBD → HT): the analysis using genetic variants associated with IBD did not indicate a causal effect on the risk of developing HT. Various MR methods, including IVW, MR-Egger, and weighted median, all yielded consistent results with ORs close to 1 (OR = 0.96, 95% CI: 0.75–1.23, P = .75), further supporting the null hypothesis of no causal effect. Sensitivity analyses also confirmed the robustness of these findings, with no evidence of pleiotropy or significant heterogeneity between IVs. Reverse MR analysis (HT → IBD): Similarly, additional analyses also confirmed the absence of pleiotropy and substantial heterogeneity.

Taken together, our results suggest that although IBD and HT share certain immunological characteristics, there is no direct causal relationship between these 2 conditions based on genetic evidence. These findings contrast with previous observational studies, highlighting the potential limitations of confounding and reverse causality in nongenetic studies.

4.2. Comparison with previous studies

Compared to observational studies, MR offers distinct advantages by leveraging genetic variants as IVs that are randomly assigned at conception according to Mendel laws. This natural randomization closely parallels the allocation process in randomized controlled trials, thereby minimizing confounding biases that often compromise observational research.^[18,19] Furthermore, as genetic variants are determined at conception and remain unaffected by subsequent environmental or behavioral influences, MR studies are inherently less susceptible to confounding from both measured and unmeasured variables.^[18,19] Because genetic variants are determined prior to disease development and remain unaffected by disease progression, MR studies effectively avoid reverse causation – a frequent limitation of observational studies, where the outcome may influence the exposure instead of the exposure influencing the outcome.^[19,20]

IBD and HT are both classified as autoimmune disorders marked by dysregulated immune responses against self-antigens. In HT, autoreactivity is directed towards thyroid-specific antigens, including thyroid peroxidase (TPO), thyroglobulin (Tg), and the thyroid-stimulating hormone receptor, resulting in thyroid tissue destruction driven by both humoral mechanisms (autoantibody production) and cellular pathways (infiltration of T and B lymphocytes).^[8,12] A fundamental commonality between the 2 diseases is the dysregulation of T helper cell subsets, notably Th1, Th2, and Th17 cells, along with impaired regulatory T cell (Treg) function. In HT, an enhanced Th17 response coupled with diminished Treg activity fosters persistent inflammation and the development of autoimmunity.^[12] Similarly, in IBD, disruption of the Th17/Treg balance plays a critical role, with Th17 cells significantly contributing to intestinal inflammatory processes.^[12]

Previous observational studies have suggested associations not only between IBD and HT but also between IBD and other forms of thyroiditis, such as Graves’ disease and subacute thyroiditis.^[21] For example, some population-based cohort studies have reported an increased prevalence of various thyroid autoimmune diseases among IBD patients, implicating shared immune dysregulation or environmental triggers.^[21] However, these observational associations may be confounded by factors such as medication use, surveillance bias, or comorbid conditions. Our MR study, which controls for such confounding via genetic instruments, did not find evidence supporting a direct causal link. This discrepancy underscores the necessity of integrating genetic epidemiology with traditional observational data to clarify these complex interrelations.

Moreover, inflammatory mechanisms such as molecular mimicry between intestinal bacterial antigens and thyroid tissue antigens may partly explain the clinical co-occurrence of these diseases without necessitating a causal genetic relationship.^[11] The heterogeneity in reported comorbidity prevalence further suggests that environmental factors, epigenetics, or yet unidentified mediators may influence the observed associations, warranting further investigation.

In this bidirectional MR analysis, we found no evidence supporting a causal relationship between IBD and HT in either direction. The primary analysis using the IVW method, along with multiple complementary approaches (MR-Egger, weighted median, simple mode, and weighted mode), consistently yielded nonsignificant results. Sensitivity analyses, including Cochran Q-test, MR-Egger intercept test, MR-PRESSO global test, and leave-one-out analysis, further supported the robustness of our findings by demonstrating no substantial heterogeneity, horizontal pleiotropy, or outlier effects. Taken together, these results suggest that despite the immunological similarities and potential observational associations reported in previous studies, there is no genetic evidence to support a direct causal effect of IBD on HT, or vice versa. Our findings highlight the importance of employing robust causal inference methodologies to disentangle complex relationships between autoimmune diseases, and suggest that any co-occurrence of IBD and HT observed in clinical settings may be attributable to shared environmental exposures, comorbid immune dysregulation, or other confounding factors, rather than a direct causal link. Inflammatory responses in IBD may transcend the gastrointestinal tract through molecular mimicry between intestinal bacterial antigens and host self-antigens in distant tissues, such as the thyroid gland.^[10,11] This mechanism could account for the observed association between IBD and HT, although the prevalence of this comorbidity varies and continues to be a subject of ongoing research.^[10,11]

4.3. Implications for clinical practice

Our bidirectional MR analysis provides novel insights into the potential causal relationship between IBD and HT, which may have important clinical implications. First, the absence of a strong causal effect from IBD on HT suggests that routine thyroid function screening in patients with IBD may not be universally warranted, particularly in the absence of clinical symptoms suggestive of thyroid dysfunction.

Conversely, given that no robust causal effect of HT on IBD was identified, systematic gastrointestinal evaluation in patients diagnosed with HT does not appear to be necessary based solely on the presence of thyroid autoimmunity. However, it remains important to recognize that coexisting autoimmune conditions, though not directly causal, may still co-occur more frequently in certain individuals due to shared genetic or immunologic susceptibility.

Our findings underscore the importance of personalized risk assessment rather than blanket screening approaches in clinical practice. Patients with either IBD or HT who present with atypical or multisystem symptoms should be carefully evaluated for the possibility of overlapping autoimmune disorders. Future clinical guidelines could incorporate genetic risk profiles and immune-related biomarkers to identify individuals at higher risk for multiple autoimmune diseases, thereby enabling earlier intervention and tailored management strategies.

4.4. Limitations

First, although we employed multiple complementary MR methods and conducted extensive sensitivity analyses to minimize bias, the possibility of residual confounding and horizontal pleiotropy cannot be completely excluded. Despite the MR-Egger intercept and MR-PRESSO tests indicating limited pleiotropy, unmeasured or unknown pleiotropic pathways may still influence the results. Second, the genetic instruments used in this study were derived exclusively from individuals of European ancestry. Although this approach helps minimize bias from population stratification, it also limits the generalizability of our findings to other ethnic groups.

5. Conclusion

Our study found no significant evidence of a causal relationship between IBD and HT, nor did the analysis indicate any causal effect of HT on the onset of IBD.

Acknowledgments

Thanks to all authors for their contributions to this article.

Author contributions

Writing – original draft: Xiaoling Ding, Xuebing Zhou.

Writing – review & editing: Xuebing Zhou, Xuerong Jin, Xue Wang, Banglong Wu, Chaomeng Wu, Xin Ma, Xia Yang, Jie Yang, Tong Wu, Qi Liang, Lijun Yin.