Abstract
Purpose
This Mendelian randomization (MR) analysis study aimed to investigate the genetic causal relationship between non-thyroidal autoimmune diseases (ADs) and Graves’ ophthalmopathy (GO).
Materials
Single nucleotide polymorphisms (SNPs) associated with inflammatory bowel disease (IBD), multiple sclerosis (MS), psoriasis vulgaris (PV), type 1 diabetes (T1D), systemic lupus erythematosus (SLE), and rheumatoid arthritis (RA) were obtained from the IEU Open genome-wide association studies (GWAS) database, GWAS data for GO were obtained from the FinnGen database. Bidirectional MR analysis was conducted using inverse variance weighted (IVW) method, weighted median (WM) method and MR-Egger test. Cochran’s Q statistic was used to assess the heterogeneity between SNP estimates. MR-Egger regression was used to evaluate horizontal pleiotropy and MR pleiotropy residual sum and outlier (MR-PRESSO) test was used to detect the outliers.
Results
For non-thyroidal ADs, the forward MR results using the IVM method showed that T1D (OR = 1.259, 95%CI: 1.026–1.5465, P = 0.028) and SLE (OR = 1.807, 95%CI: 1.229–2.655, P = 0.003) were correlated with the risk of GO at the genetic level, while there was no evidence showing that IBD, MS, PV and RA were correlated with GO. In the reverse MR study, there was a significant increase in the risk of developing T1D in GO (OR = 1.135, 95%CI: 1.018–1.265, P = 0.022), but pleiotropy and heterogeneity existed.
Conclusions
In the European population, there is strong genetic evidence that patients with T1D and SLE have a higher risk of developing GO, whereas the effect of GO on ADs is unclear.
Highlights
- •
T1D and SLE could increase the risk of getting GO.
- •
No evidence substantiate a significant causal association of GO on non-thyroidal autoimmune diseases.
- •
Offers new insights, emphasizing targeted interventions for GO prevention and treatment.
1
Introduction
Graves’ ophthalmopathy (GO), also known as thyroid-associated ophthalmopathy and thyroid eye disease, is an autoimmune inflammatory disease related to thyroid diseases such as Graves’ disease (GD), causing hyperthyroidism, and Hashimoto’s disease, causing hypothyroidism. A small proportion of patients can be euthyroid (approximately 5%). It is an organ-specific disorder caused by immunological imbalance of the thyroid gland. The antibodies produced by T and B cells, mainly thyroid receptor antibodies and insulin like growth factor antibodies attack orbital soft tissue such as orbital fat and extraocular muscle. The symptoms of GO mainly include eyelid retraction, redness of the eye, and exophthalmos in severe cases, there might be optic nerve damage caused by the impression of soft tissue. GO changes the appearance and visual function of patients, which causes mental burden and affects their quality of life. Although it is believed to be associated with abnormal immune response, the pathogenesis of GO has not yet been fully elucidated.
Expect for GO, there are also other autoimmune diseases (ADs) such as systemic lupus erythematosus (SLE), type 1 diabetes (T1D), and inflammatory bowel disease (IBD), which cause lifelong recurrent symptoms and bring socio-economic burden for both the patients and society. Owing to shared genetic or environmental factors, ADs tend to cluster among individuals. It has been reported that non-thyroidal ADs may be related to some thyroid disease such as thyroditis. However, most of the studies were observational cohort studies that only reported the prevalence and the underlying genetic association between those non-thyroidal autoimmune diseases remains unclear.
With the advantage of minimizing residual confounding factors, Mendelian randomization (MR) has been used to assess the potential causal association between exposure factors and outcomes. This approach uses independent single-nucleotide polymorphisms (SNPs) extracted from genome-wide association studies (GWASs) to evaluate the impact of exposure. Since genetic makeup is determined randomly at an early stage of life, MR can eliminate acquired factors such as lifestyle and environment. Several studies have focused on the causal relationships between autoimmune diseases and GD, suggesting potential links between non-thyroidal autoimmune diseases and GO. In this study, we aimed to reveal the possible relationship between other immune-related diseases and GO, and to explore more on the complex causes of GO.
2
Materials and methods
2.1
MR study design
We used the public GWAS summary data for the analysis. IBD, multiple sclerosis (MS), psoriasis vulgaris (PV), rheumatoid arthritis (RA), SLE, and T1D were included in this study based on the previously published studies. Bidirectional MR analysis was conducted to assess the causal association between these ADs and GO. Fig. 1 shows a schematic flow chart of the study. The MR analysis meets three main assumptions: (1) correlation assumption: the included instrumental variables (IVs) must be closely related to exposure (non-thyroidal autoimmune diseases); (2) independence assumption: IVs are independent of the confounding factors of non-thyroidal autoimmune diseases and GO; and (3) exclusion assumption: IVs only affect GO via non-thyroidal autoimmune diseases. Only GWASs obtained from the European population were used to minimize the bias caused by ethnic variety.
2.2
Data sources of non-thyroidal ADs and GO
The GWAS data for 6 non-thyroidal ADs were procured from the Integrative Epidemiology Unit (IEU) Open GWAS database ( https://gwas.mrcieu.ac.uk/ ). The summary data for GO were obtained from the FinnGen database (www.finngen.fi/en/, version: R9), with the endpoint name “E4_GRAVES_OPHT_STRICT”. Details of the data sources were shown in Table 1 .
| Exposure | GWAS-ID or FinnGen endpoint name | Sample size | Number of SNPs | Year |
|---|---|---|---|---|
| Inflammatory bowel disease | ieu-a-292 | 75000 | 14378 | 2012 |
| Multiple sclerosis | ieu-a-1025 | 38589 | 156632 | 2013 |
| Psoriasis vulgaris | ebi-a-GCST90018907 | 483174 | 24191364 | 2021 |
| Type 1 diabetes | ebi-a-GCST90018925 | 457695 | 24182422 | 2021 |
| Systemic lupus erythematosus | ebi-a-GCST90018917 | 482911 | 24198877 | 2021 |
| Rheumatoid arthritis | ebi-a-GCST90018910 | 417256 | 24175266 | 2021 |
| Graves’ ophthalmopathy | E4_GRAVES_OPHT_STRICT | 377277 | 20170236 | 2023 |
2.3
Genetic instrument selection
The SNPs were selected based on the MR assumptions mentioned above: (1) each selected SNP was associated with non-thyroidal ADs at a genome-wide significance threshold of P < 5 × 10 −8 , and if the number of included SNPs was insufficient for subsequent MR analysis, the threshold value was set to P < 5 × 10 −6 ; (2) the linkage disequilibrium analysis clumped the SNPs further ( r 2 < 0.001 and clumping distance = 10000 kb); (3) The F -statistics were used to assess the weak instrumental variable bias to ensure that there was a robust correlation between IVs and exposure ( F > 10 deemed to be valid).
2.4
Statistical analysis
Statistical analysis was performed using the R software (version 4.3.1). The “Two SampleMR” package (version 0.6.0) was used for the MR analysis. Three MR methods were used to evaluate the causal relationship between non-thyroidal ADs and GO: the variance weighted (IVW) method, weighted median (WM) method and MR-Egger test. The IVM method was selected as the primary approach due to its ability to provide precise effect estimates under the assumption of no horizontal pleiotropy, as it calculates the Wald ratio for each SNP, ensuring optimal statistical power in this context. To enhance robustness, we supplemented IVM with the WM and MR-Egger methods. The WM approach offers greater resilience in causal effect estimation, performing reliably even if up to 50% of the IVs are invalid. Meanwhile, MR-Egger provides an intercept test to detect horizontal pleiotropy and remains applicable even when all IVs are invalid. Odds ratios (OR) and 95% confidence intervals (CI) were used to determine effect size.
Cochran’s Q statistic was used to assess the heterogeneity between SNP estimates. A random-effects model was used if heterogeneity was present ( P < 0.05). Otherwise, the fixed-effects model was used. We used the MR-Egger regression to evaluate horizontal pleiotropy. MR pleiotropy residual sum and outlier (MR-PRESSO) test was used to confirm the results and detect the outliers. After removing the outliers, we re-conducted the MR analysis. Finally, leave-one-out analysis was employed to assess the influence of individual SNP.
Since all data were obtained from publicly shared databases, no additional ethical approval was required for this study.
3
Results
3.1
The forward MR analyses
Under a threshold of P < 5 × 10 −8 , 144 IVs for IBD, 49 for MS, 15 for PV, 25 for RA, 5 for SLE and 18 for T1D were strongly correlated with these exposures. All the F-statistics of the IVs were larger than 10, ranging from 45.52 to 179.09, indicating that there was no weak instrumental bias, and the SNPs had adequate validity. We employed the MR-PRESSO method and found outlier SNPs, including rs11889341, rs35139284, rs6679677, rs7731626 for RA and rs13204736, rs6679677, rs9273363 for T1D. A subsequent MR analysis was done after removing these outliers. We then re-ran the MR-PRESSO test. Global p values for IBD, MS, PV, RA, SLE and T1D were 0.013, <0.001, 0.142, 0.069, 0.132, and 0.336, respectively.
The IVM method showed that in European populations, there was a significant increase in the risk of developing GO in SLE (OR = 1.807, 95%CI: 1.229–2.655, P = 0.003) and T1D (OR = 1.259, 95%CI: 1.026–1.5465, P = 0.028) patients. There is insufficient evidence to show that the other four ADs are associated with GO. ( Fig. 2 ). The SNPs included in the MR analysis are listed in Supplementary Table 1 .
