Causal Association Between Inflammatory Bowel Disease and Psoriasis: A Two-Sample Bidirectional Mendelian Randomization Study

Background Previous observational studies have found an association between inflammatory bowel disease (IBD) and psoriasis. Using the mendelian randomization (MR) approach, we aim to determine whether there was a causal association between IBD and psoriasis. Methods We performed a two-sample MR with the genetic instruments identified for IBD and its main subtypes, Crohn’s disease (CD) and ulcerative colitis (UC), from a genome-wide association study (GWAS) involving 25,042 cases with an IBD diagnosis and 34,915 controls. Summarized data for psoriasis were obtained from different GWAS studies which included 4510 cases and 212,242 controls without psoriasis. Causal estimates are presented as odds ratios (ORs) with 95% confidence intervals (CIs). Results The overall outcome of MR analysis was to demonstrate that genetic predisposition to IBD was associated with an increased risk of psoriasis (OR: 1.1271; 95% CI: 1.0708 to 1.1864). Psoriatic arthritis (PsA) had a significant association with total IBD (OR: 1.1202; 95% CI: 1.0491 to 1.1961). Casual relationship was also identified for CD-psoriasis (OR: 1.1552; 95% CI: 1.0955 to 1.2182) and CD-PsA (OR: 1.1407; 95% CI: 1.0535 to 1.2350). The bidirectional analysis did not demonstrate that a genetic predisposition to psoriasis was associated with total IBD, although psoriasis showed association with CD (OR: 1.2224; 95% CI: 1.1710 to 1.2760) but not with UC. A genetic predisposition to PsA had a borderline association with IBD (OR: 1.0716; 95% CI: 1.0292 to 1.1157) and a suggestive association with CD (OR: 1.0667; 95% CI: 1.0194 to 1.1162). Conclusion There appears to be a causal relationship between IBD and psoriasis, especially for PsA, but for psoriasis and IBD, only total psoriasis and PsA were associated with CD. Understanding that specific types of psoriasis and IBD constitute mutual risk factors facilitates the clinical management of two diseases.


INTRODUCTION
Psoriasis is an immune-related, chronic inflammatory skin disorder of common occurrence. Clearly demarcated areas of erythematous plaques with overlying silvery scales appear on the skin. There are estimated to be 60 million psoriasis patients worldwide with the prevalence ranging from 0.1% to 1.99%. The conditions occur more frequently in high-income countries and approximately 2% of the population in Western and Central Europe are affected (1,2). Psoriasis vulgaris is the most common form, accounting for 90% of cases (3). Psoriatic arthritis (PsA), involving inflammation, swelling, and pain in affected joints, may precede or succeed skin manifestations, and PsA affects 5% to 30% of psoriasis patients (4)(5)(6). Psoriatic disease, consisting of both psoriasis (only skin manifestation) and PsA, is considered a multifactorial disease that has been associated with multiple systemic disorders. Numerous studies involving diverse populations and settings (including both the western and eastern countries) have associated psoriatic diseases with autoimmune, metabolic, gastrointestinal, and mental health disorders (4,5,(7)(8)(9)(10), which has caused a worldwide disease burden.
Inflammatory bowel disease (IBD), including Crohn's disease (CD) and ulcerative colitis (UC), involves chronic relapsing episodes of immune-mediated inflammation. Evidence from meta-analyses and observational studies supports an underlying bidirectional relationship between psoriasis and inflammatory bowel disease (IBD), although mechanisms remain unclear (11,12). Many factors, such as common genetic susceptibility loci, clinical course, and immunologic features, are shared by psoriatic diseases and IBD (5,(13)(14)(15)(16), and IBD was considered a subsequent extra-articular manifestation of PsA which should be specially monitored in PsA patients according to the previous studies. However, basic and observational studies are susceptible to confounding factors, such as demographics or environmental exposure (17,18). Thus, any causal bidirectional role in the respective development of psoriasis/PsA and IBD remains controversial.
Causal relationships between exposure and outcome can be assessed by Mendelian randomization (MR) analysis introducing genetic variation as an instrumental variable. This approach exploits the random distribution of genetic variation to eliminate confounding factors and reverse causality, simulating the randomization process of a randomized controlled experiment. The current study applied MR to examine the causal relationship, the strength of association, and the direction of causality between psoriasis and IBD (19)(20)(21).

MATERIALS AND METHODS
Multiple single-nucleotide polymorphisms (SNP) representing genetic variation were selected as instrumental variables and subjected to two-sample MR analysis. Three key hypotheses were adopted as follows ( Figure 1): 1. instrumental variables are directly related to exposure; 2. instrumental variables are independent of any confounding variables; 3. genetic variants only affected results via exposure (22). MR analysis was used to assess bidirectional causal relationships between psoriasis and IBD, including subtypes of psoriasis, psoriasis vulgaris, and PsA, and of IBD, UC and CD.

Genome-Wide Association Analysis
Databases resulting from genome-wide association studies (GWAS), including GWAS Catalog, IEU openGWAS and NealELab, were searched and eligible datasets extracted. Since all data used was already in the public domain, no additional ethical approval was required. The study population's genetic background was restricted to that of European ancestry to reduce the resulting bias caused by ethnically related confounding factors. The FinnGen Biobank Analysis Consortium 2021 (23) was used to identify genetic risk variants for psoriasis. The database included 4510 cases with a psoriasis diagnosis plus a control group of 212,242 individuals without psoriasis. A dataset identifying 334 cases of psoriasis vulgaris and 1,637 of PsA in addition to 212,242 controls was used for subtype analysis. The diagnosis of psoriasis and its subtypes was according to the ICD-10 (International Classification of diseases) criteria.
Similarly, the European Bioinformatics Institute (EBI) database included 25,042 cases with an IBD diagnosis and 34,915 controls. Datasets identifying 12,366 cases of UC with 33,609 controls and 12,194 cases of CD with 28,072 controls were used to assess secondary outcomes. The diagnosis of IBD and its subtypes was based on accepted endoscopic, histopathological, and radiological criteria.

SNP Selection
A genome-wide significance level of p < 5 · 10 -8 and a clumping algorithm with a cutoff of r 2 = 0.001 and kb = 10000 were used to avoid linkage disequilibrium (LD). For exposure analyses of IBD and its two main subtypes, 95 SNPs related to total IBD, 76 SNPs related to CD, and 51 SNPs related to UC were selected. There were an additional 10 SNPs associated with total psoriasis selected for coanalysis with IBD. At a threshold of p < 5 x 10 -8 , fewer than 10 SNPs were identified for the two subtypes of psoriasis thus failing to meet the minimum requirements for MR studies (24,25). At a looser threshold of p < 5 · 10 -6 (26, 27), 10 SNPs for psoriasis vulgaris and 13 SNPs for PsA were identified. Due to the reduced significance threshold, F statistics were used to evaluate the risk of weak instrument bias, producing a bias level of F <10. Selected SNPs were also matched to databases for phenome-wide association studies (pheWAS) to avoid the potential association between the SNPs and outcomes confounders with a threshold of p < 5 x 10 -6 (26,28).
Exposures and outcomes were harmonized in terms of effect allele and subsequent analyses based on the merged exposureoutcome dataset. Detailed information on IVs is shown in Supplementary Tables 1-6.

Statistical Analysis
A meta-analytical approach involving an inverse variance weighted model (IVW) was used to combine Wald estimates of causality for each IV, an approach expected to be stable with balanced pleiotropy. Several statistical approaches were combined to evaluate associations between cause and effect. Weighted median estimator model, weighted model-based method, MR-Egger regression model, MR-Robust Adjusted Profile Score (MRAPS), and MR pleiotropy residual sum and outlier (MR-PRESSO) were also established to estimate causal associations under different conditions. The weighted median estimator method examined the median effects of all available SNPs when half the IVs were valid, resulting in unbiased estimates of effects (29). The weighted model-based method obtained robust overall causal estimates when the majority of individual estimates were from valid IVs (30). The MR-Egger regression model provided a relatively robust estimate independent of IV validity and an adjusted result by existing horizontal pleiotropy via the regression slope and intercept (31,32). MRAPS has the potential to derive a more accurate causal assessment on the condition of ideal independence of IVs (33). MR-PRESSO used global and SNP-specific observed residual sum of squares to test for potential outlying IVs to obtain a corrected causal effect by excluding outliers with p < 0.05 in the further distortion test (34). Heterogeneity of IVs was assessed by Cochrane's Q-statistic. A value of p < 0.05 indicated significant heterogeneity in which case a random-effect model was adopted for subsequent analyses. Otherwise, a fixed-effect model was used (35). IVs were excluded one by one to judge the stability of the MR results (36). Directional pleiotropy was assessed and corrections were made based on the intercept obtained from the MR-Egger regression model analysis (32). However, compared to the IVW method, other methods produced wide confidence intervals (CI) (37) and were only used as complementary methods. Thus, the MR-Egger Regression Model was used in the event of significant pleiotropy and the MR-PRESSO model for the detection of final outliers. Otherwise, IVW results were given priority.
Causal estimates are presented as odds ratios (ORs) with 95% confidence intervals (CIs). A Bonferroni correction was made to avoid false-positive results of various comparison methods in bidirectional multiple tests. A two-sided p-value <0.0083 was regarded as statistically significant after Bonferroni correction while associations with p-values ≥0.0083 and <0.05 were regarded as suggesting significance. Bidirectional two-sample MR analysis was conducted using open-source statistical software R (version 4.1.2, R Foundation for Statistical Computing, Vienna, Austria) with TwoSampleMR (version 0.5.6) and MR-PRESSO packages (version 1.0.0). Results are reported following the STROBE-MR (Strengthening the Reporting of Observational Studies in Epidemiology-Mendelian Randomization) statement (38).
The scatter plots of causal relationships of primary MR analysis are shown in Figure 2. The forest plots of causal relationships of genetically predicted IBD and the risk of psoriasis and its subtypes are shown in Figure 3, and the forest plots of causal effects of psoriasis on IBD and its subtypes are shown in Figure 4.

Effect of IBD or Its Main Subtypes on Psoriatic Diseases
Analyses of total IBD on psoriasis vulgaris or PsA showed significant heterogeneity but no directional pleiotropy. Hence, multiplicative random-effect models were applied in the IVW analysis and a significant association emerged between total IBD and PsA (IVW: OR: 1.1217; 95% CI: 1.0047 to 1.2043; p=1.54E-03). There was no association with psoriasis vulgaris. MR-PRESSO confirmed the IBD-PsA association (outlier-corrected: OR: 1.1202; 95% CI: 1.0491 to 1.1961; p=1.02E-03).
Analyses of IBD's two main subtypes, UC and CD, were also conducted. The association of CD with psoriatic diseases showed significant heterogeneity but no directional pleiotropy. CD was causally associated with total psoriasis (IVW However, no causal association was found between UC and total psoriasis or the two subtypes.

Effect of Psoriatic Diseases on IBD and Its Main Subtypes
The association of total psoriasis with CD and UC showed heterogeneity but no directional pleiotropy and multiplicative random-effect models were applied for the final IVW analysis. Psoriasis was significantly associated with CD (IVW: OR: 1.2689; 95% CI: 1.1504 to 1.3996; p=1.93E-06). Two outliers (rs12713428 and rs2021511) were removed and the association remained significant (outlier-corrected: OR: 1.2224; 95% CI: 1.1710 to 1.2760; p=9.54E-05). However, there was no evidence of a causal effect of psoriasis on UC.
Significant heterogeneity but no directional pleiotropy was found in the MR analyses of the relationships between PsA and IBD plus subtypes. The IVW model did not show a causal effect of PsA on IBD and its subtypes. However, after removing all outliers, the MR-PRESSO model did reveal a causal effect for PsA on IBD (outlier-corrected: OR: 1.0716; 95% CI: 1.0292 to 1.1157; p=7.28E-03) and suggested a significant causal effect of PsA on CD (outlier-corrected: OR: 1.0667; 95% CI: 1.0194 to 1.1162; p=1.91E-02). No heterogeneity, no directional pleiotropy, and no significant causal association were found between psoriasis vulgaris and IBD plus subtypes.
The forest plots and scatter plots of secondary MR analysis are shown in Supplementary Figures 1-4. Details of sensitivity analysis are shown in Table 1 and Supplementary Figures 5, 6.

DISCUSSION
The overall outcome of MR analysis was to demonstrate that genetic predisposition to IBD was associated with an increased risk of psoriasis (OR:1.1268; 95% CI: 1.0662 to 1.1908). Total  There has been much discussion in the medical literature, including two meta-analyses, on the relationship between psoriasis and IBD. A meta-analysis of 9 observational studies (5 casecontrol or cross-sectional and 4 cohorts) (11), including 7,794,087 study participants, concluded that patients with psoriasis had increased odds of CD  (12) analyzed 93 studies and reported bidirectional associations between psoriasis and IBD. IBD was associated with a 1.8-fold increased odds for psoriasis and associations between psoriasis and CD (OR: 2.0; 95% CI: 1.4 to2.9) and UC (OR: 1.5; 95% CI: 1.2 to2.0) were also observed. Similarly, patients with psoriasis also had increased odds of developing total IBD (OR: 2.0; 95% CI: 1.6 to 2.5), CD (OR: 2.2; 95% CI: 1.6 to 3.1) and UC (OR: 1.6, 95% CI: 1.3 to 2.0). There are some discrepancies between these reported estimates and our results. We believe that the difference may be caused by the two different analysis methods themselves. The observational research might be affected by unavoidable clinical confounding factors, these confounding factors could affect both exposure and outcome and weaken the ability of observational results to make an accurate causal judgment. Therefore, even if an observational study reported a strong correlation, it could not prove the existence of a direct causal correlation. Mendelian randomization could avoid the influence of these confounding factors by introducing genetic instrumental variables, to obtain a relatively accurate causal assessment.
The results of the current work demonstrate that the observational estimates reported in the literature can be explained by a causal effect of IBD on psoriasis. Thus, the presence of IBD was partly responsible for the development of psoriasis in some patients with the relationship being especially strong for PsA. It is noteworthy that, when the exposure was psoriasis, there was some evidence of causality in the case of CD but not in that of UC. For both psoriasis and IBD, disease subtypes differed in terms of their bidirectional associations. UC was not genetically linked to any subtype of psoriasis and psoriasis vulgaris was not linked to any type of IBD. Subtypes of the same broad group of diseases may share some similarities in the genetic background but there remain significant differences in individual genes, genetics, and immunity between PsA and psoriasis vulgaris (1,39). The same may be said for UC and CD (40)(41)(42). Further study is required to elucidate whether such apparent differences have their origins in environmental factors, differences in immune response, changes in intestinal flora or genetic factors that influence different disease subtypes.
We would like to highlight some strengths of our study and acknowledge some limitations. Firstly, it is the first study to assess the bidirectional causal association of IBD and psoriasis by using a 2-sample MR approach. This method is less susceptible to confounders, reverse causation, and exposures non-differentially compared to observational studies. Secondly, the subtypes of the disease are strictly defined to avoid the impact on the results of the co-existence of diseases. Thirdly, sensitivity analysis was performed to ensure the consistency of causal estimation and the robustness of the results. There are also some limitations. Due to the paucity of IVs, a lower p threshold was chosen for the study of IBD with PsA and psoriasis vulgaris. Although no significant risk of weak instrument bias was found in our F-statistic tests, this negative result should be interpreted with caution. In addition, the sample size of psoriasis vulgaris patients was relatively small due to the strict definition. The current study was conducted on individuals of European ancestry and cannot be reliably generalized to other races. For example, some gastrointestinal involvements (such as IBD) were also identified as extra-articular manifestations of Chinese patients with PsA (9). Although previous studies revealed that the prevalence of PsA among Asian psoriasis patients might be less than that of western patients (43,44), the exploration of the causal association between psoriatic diseases and IBD in the Asian population in the future studies remained to be of great significance. Last but not the least, although we have matched all selected SNPs to the pheWAS database to avoid the potential confounders and associated horizontal pleiotropy with a strict threshold, this measure can not completely avoid the impact of horizontal pleiotropy as the exact biological function of many genetic variants is still unknown. Future studies have been planned to address these limitations when higher quality GWAS studies become available.

CONCLUSIONS
In conclusion, there appears to be a causal impact of IBD on psoriasis, especially for PsA. Notable differences between the two subtypes of IBD, UC, and CD, occurred in their respective associations with psoriasis. Total psoriasis and PsA were associated with CD. Understanding that specific types of psoriasis and IBD constitute mutual risk factors facilitates early diagnosis enabling more efficient targeting of therapy. Further study is required to elucidate the pathophysiology of the causal relationship between IBD and psoriasis.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

AUTHOR CONTRIBUTIONS
ZC and JW conceived the study, participated in its design and coordination, and critically revised the manuscript. YL and ZC preparation, review, or approval of the manuscript; and decisionto submit the manuscript for publication.