Modifiable factors for irritable bowel syndrome: evidence from Mendelian randomisation approach.

Irritable bowel syndrome (IBS) is a common gastrointestinal disorder that affects approximately 5–10% of the global population. 1 3 IBS is characterised by symptoms such as abdominal pain, diarrhoea and constipation, which diminish the quality of life and impose a considerable economic burden. 1 The heterogeneous disorder may arise from a complex interplay between environmental, psychological and genetic factors. 4 8 Despite ongoing research efforts, the exact pathogenesis of IBS and effective prevention strategies remain largely elusive. Modification of risk factors has been proposed as a promising strategy to tackle the global challenge of IBS. 1 3 5 Epidemiological studies have identified various modifiable factors associated with IBS, including diet, alcohol consumption, sleep disorders and frailty. 9 12 However, a comprehensive exploration of these modifiable factors in relation to IBS remains lacking. Moreover, observed associations between modifiable factors and IBS may be influenced by residual confounding and reverse causality biases. Unmeasured confounding variables, such as living environment and unaccounted lifestyle factors, could lead to misleading associations. Additionally, changes in modifiable factors that occur after the onset of the disease may create spurious connections with IBS. Mendelian randomisation (MR) approach is a widely used genetic epidemiological technique that leverages genetic variants as instrumental variables (IVs) to estimate unbiased associations between exposures and outcomes, thereby reducing biases from unmeasured confounding and reverse causality. 13 14 With the widespread availability of genome-wide association study (GWAS) data, MR has become an essential tool for investigating causal associations between exposures and outcomes. 15 18 However, it is crucial to recognise that the estimates derived from MR analyses cannot be directly compared with those obtained from observational studies. To mitigate this limitation, we conducted a comprehensive review of the existing observational evidence before performing MR analyses. IBS frequently coexists with other disorders, sharing common risk factors and genetic backgrounds, with bidirectional relationships typically observed between IBS, other gastrointestinal disorders and psychiatric disorders. 7 19 20 It is crucial to assess the impact of various exposures on IBS accounting for these related disorders, as this knowledge can significantly enhance IBS management. To tackle the challenge posed by multiple related exposures and outcomes, a novel MR approach known as multiresponse MR (MR 2 ) has been developed to identify the shared and distinct exposures across multiple related disorders. 21 In our study, we explored the causal relationship between a diverse array of factors and IBS by leveraging comprehensive epidemiological evidence. We also employed MR 2 to evaluate the effects of exposures on IBS while accounting for the presence of multiple coexisting disorders.

To identify reliable modifiable factors for targeted IBS prevention strategies, our study addressed three key questions: (1) summarising important modifiable factors for IBS; (2) mitigating the effect of unobserved confounding; (3) accounting for the influence of coexisting disorders ( online supplemental figure S1 ). As illustrated in figure 1 , we systematically compiled a list of modifiable factors and coexisting disorders associated with IBS by conducting a thorough search of the PubMed database up to May 2024 for cohort studies or meta-analyses (detailed search strategies in online supplemental appendix 1 and the selection criteria in online supplemental method 1 ). Subsequently, we conducted MR and genetic correlation analyses. We further employed MR 2 analysis to investigate the observed causal associations while accounting for the influence of coexisting disorders. Additional details regarding the MR approach, including related assumptions and biases, are provided in online supplemental method 2 . While we presented our findings in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization guideline ( online supplemental appendix 2 ), 22 we acknowledge that this standard may not be fully align with our study design. Future research endeavours should aim to improve reporting standards for studies like ours, incorporating advanced MR methodologies. We compiled a list of modifiable factors associated with IBS based on the cohort studies and meta-analysis included in our review ( online supplemental table S1 ). We excluded ecological indicators, treatments, and factors without available GWAS data. Our focus was on potentially modifiable factors, broadly categorised into lifestyle behaviours, social determinants, infections, body symptoms, behaviour disorders, autoimmune diseases and gynaecological diseases ( online supplemental table S2 and online supplemental method 3 ). Detailed definitions and sources of the GWAS data are provided in online supplemental table S3 , with further information accessible from previously published studies. 23 35 By integrating insights from observational studies and genetic evidence, 19 20 we examined nine gastrointestinal diseases and 11 psychiatric disorders as coexisting disorders ( online supplemental table S2 ). Detailed definitions and sources of the GWAS data are provided in online supplemental table S3 , with additional information accessible from previously published studies. 2426 32 36 41 These coexisting disorders were considered modifiable factors in MR analyses, whereas they were treated as outcomes alongside IBS in MR 2 analysis. We obtained summary GWAS data for IBS from the UK Biobank and international collaborative Bellygenes initiative as the discovery dataset, 7 which included 53 400 IBS cases and 433 201 controls (refered to as “GWAS meta” hereinafter; online supplemental table S3 ). We used the 10th version of FinnGen Biobank comprising 10 329 IBS cases and 329 381 controls as the replication dataset. 32 The diagnosis of IBS in GWAS meta was based on the Rome III symptom criteria and the Patient Health Questionnaire 12 Somatic Symptom score, as detailed in a previous GWAS paper. 7 The diagnosis of IBS in the FinnGen Biobank was based on International Classification of Diseases 10th Revision code K58. Genome-wide linkage disequilibrium score regression (LDSC) was employed to evaluate the genetic associations of modifiable factors (including coexisting disorders) with IBS. The linkage disequilibrium (LD) proxies were determined using 1000 genome European samples, and only single nucleotide polymorphisms (SNPs) with available data for both SNP-factor and SNP-IBS associations were included in the analysis. Results were presented as genetic correlation (rg) along with SE. The primary MR analysis used inverse-variance weighted (IVW) method to investigate the causal relationships between target modifiable factors and IBS. SNPs that met genome-wide significance (p<5×10 −8 ) and exhibited low LD (r 2 <0.001) were selected as IVs. When no eligible SNPs were identified, a more lenient threshold (p<5×10 −6 ) was also considered. The R 2 [R 2 =2×EAF×(1-EAF)×Beta 2 ] of each SNP was estimated, and then summed up to calculate the overall R 2 . Then F-statistic was calculated as Beta 2 /SE 2 . Higher R 2 and F-statistic indicate a reduced risk of weak IV bias. MR-Egger analysis was conducted to evaluate pleiotropy based on the intercept. Sensitivity analyses, including weighted median, penalised weighted median, MR-Egger, MR-Pleiotropy Residual Sum and Outlier and MR-Robust Adjusted Profile Score, were conducted to quantify the causal relationships. The MR Steiger test was also used to validate the assumption that exposure causes the outcome. Furthermore, the Causal Analysis Using Summary Effect estimates (CAUSE) 42 was employed as an alternative primary MR analysis to explore the causal effects of target modifiable factors on IBS. For this analysis, SNPs that achieved genome-wide significance (p<1×10 −3 ) were selected as IVs, with low LD (r 2 <0.01). Results were presented as ORs with their 95% CIs for IBS per genetically predicted increase in factors. Considering various sources of evidence, we adopted a relatively lenient p value threshold of <0.05 to identify potential associations. Based on a combination of genetic correlation and the primary MR analysis, we classified each association into distinct categories: convincing (‘class I’), highly suggestive (‘class II’), suggestive (‘class III’), weak (‘class IV’), controversial (‘class V’) or no evidence (‘class VI’). The detailed criteria for these classifications were outlined in online supplemental table S4 . For simplicity, we designated the IVW analysis in discovery dataset as MR-1, the IVW analysis in replication dataset as MR-2, the CAUSE analysis in discovery dataset as MR-3 and the CAUSE analysis in replication dataset as MR-4. MR 2 was used to further refine causal estimates while accounting for the influence of coexisting disorders. The involved hypotheses have been previously described, 21 and we used the marginal posterior probability of inclusions (mPPIs) to quantify support for these hypotheses. An mPPIs level exceeding a 5% false discovery rate was considered suggestive evidence of shared modifiable factors contributing to both IBS and coexisting disorders. Data cleaning and statistical analysis were performed using R V.4.1.3 ( https://www.r-project.org/ ) and Python V.3.11.2 ( https://www.python.org/ ).

As shown in figure 2 , the genetic correlations between modifiable factors and IBS ranged from 0.0005 to 0.635 in the discovery dataset and from 0.001 to 0.718 in the replication dataset. Significant genetic correlations were observed between IBS and various modifiable factors, including lifestyle behaviours (lifetime smoking index, alcohol frequency, sugar-sweetened beverage consumption, protein intake, coffee consumption, sleep duration and insomnia), social determinants (college or university degree, education years, intelligence, childhood maltreatment and average total household income), infections (COVID-19 infection and its severe illness), body symptoms (frailty index, multistate chronic pain, migraine and fibromyalgia), behaviour disorders (alcohol use disorder and restless legs syndrome), autoimmune diseases (osteoarthritis, allergic disease and asthma) and gynaecological diseases (uterine fibroids and endometriosis) in both the discovery and replication datasets (all p values <0.05). Due to the low heritability of factors such as bacterial enteritis and viral enteritis, their correlations with IBS could not be established. As shown in figure 2 , the genetic correlations between coexisting disorders and IBS ranged from 0.040 to 0.619 in the discovery dataset and from 0.082 to 0.743 in the replication dataset. In addition to inflammatory bowel disease and coeliac disease, significant genetic correlations were identified between IBS and several other gastrointestinal disorders, including gastro-oesophageal reflux disease, peptic ulcer disease, dyspepsia, non-alcoholic fatty liver disease, primary biliary cholangitis, diverticular disease of the intestine and haemorrhoidal disease, in both the discovery and replication datasets (all p values <0.05). With the exception of anorexia nervosa, the significant genetic correlations were found between IBS and various psychiatric disorders, such as well-being spectrum, life satisfaction, positive affect, neuroticism score, depression, anxiety, bipolar disorder, schizophrenia, attention deficit hyperactivity disorder and post-traumatic stress disorder, in both datasets (all p values <0.05). All the F-statistics for IVs related to modifiable factors and coexisting disorders were greater than 10, indicating that the included SNPs satisfied the strong relevance assumption. Detailed information about the IVs was provided in online supplemental table S5 . When considering modifiable factors in relation to IBS, supported by genetic correlation evidence, we discovered a convincing causal association between multisite chronic pain and IBS through combined IVW and CAUSE analyses in both the discovery and replication datasets (MR-1: OR=2.20, 95% CI 1.82 to 2.66; MR-2: OR=1.77, 95% CI 1.24 to 2.51; MR-3: OR=1.65, 95% CI 1.51 to 1.81; MR-4: OR=1.48, 95% CI 1.23 to 1.76). Furthermore, we also found highly suggestive causal associations between IBS and factors such as lifetime smoking index, alcohol frequency, college or university degree, intelligence, childhood maltreatment and frailty index ( figures3 4 ). These results were consistent across several sensitivity analyses ( online supplemental tables S6 and S7 ). The MR Steiger results suggested that the causal directions were likely accurate. MR-Egger regression intercept, which was close to zero with a p value larger than 0.05, suggested no significant evidence of horizontal pleiotropy for the above association of modifiable factors with IBS. Some factors, such as coffee consumption, sleep duration, insomnia, average total household income, fibromyalgia, alcohol use disorder, osteoarthritis, allergic disease, asthma and endometriosis, exhibited genetic correlation with IBS, but did not show consistent associations across specific datasets and MR methods. The evidence for these factors ranged from suggestive to weak. The potential causal effects of some factors on IBS remain controversial, such as body mass index, birth weight, overall healthy diet, fresh fruit consumption, fruit consumption, vegetables consumption, cooked vegetables consumption, non-oily fish consumption, sugar-sweetened beverage consumption, fat intake, sugar intake, protein intake, tea intake, sleep duration, morning person, shift work, Townsend deprivation index at recruitment, COVID-19 infection and severity, bacterial enteritis, migraine, alcohol dependence, restless legs syndrome, polycystic ovary syndrome and uterine fibroids. Overall, we found no genetic evidence supporting a causal effect of these factors on IBS, including physical activity (including vigorous, moderate, moderate to vigorous), salad/raw vegetable intake, oily fish consumption, fish consumption, processed meat intake, carbohydrate intake, infection, viral enteritis, atrial fibrillation and alopecia areata (all p values >0.05; figure 5 ). When considering genetic evidence for coexisting disorders and IBS, we found convincing causal associations of gastro-oesophageal reflux disease (MR-1: OR=1.31, 95% CI 1.23 to 1.39; MR-2: OR=1.37, 95% CI 1.22 to 1.54; MR-3: OR=1.23, 95% CI 1.20 to 1.31; MR-4: OR=1.25, 95% CI 1.14 to 1.37) and well-being spectrum (MR-1: OR=0.17, 95% CI 0.13 to 0.21; MR-2: OR=0.28, 95% CI 0.17 to 0.47; MR-3: OR=0.35, 95% CI 0.30 to 0.40; MR-4: OR=0.49, 95% CI 0.34 to 0.70) with IBS. Well-being spectrum, which included life satisfaction, positive affect, neuroticism score and depression, showed convincing causal associations with IBS. In addition, we found highly suggestive causal associations of intestinal diverticula and schizophrenia with IBS. Similar results were also observed in some sensitivity analyses ( online supplemental tables S6 and S7 ). While haemorrhoidal disease and bipolar disorder exhibited genetic correlation with IBS, their associations found in specific datasets and MR methods, provided only weak evidence. The causal effects of these factors on IBS remain controversial, including peptic ulcer disease, dyspepsia, non-alcoholic fatty liver disease, primary biliary cholangitis, inflammatory bowel disease, anxiety disorder, attention deficit hyperactivity disorder, post-traumatic stress disorder and anorexia nervosa. Overall, there was no genetic evidence supporting the association of coeliac disease with IBS ( figure 5 ). Modifiable factors and coexisting disorders with at least a high level of evidence were included in the further MR 2 analysis. In the discovery dataset, we found the causal factors of IBS, including college or university degree, multisite chronic pain and frailty index, were also associated with gastro-oesophageal reflux disease and intestinal diverticula and well-being spectrum. However, lifetime smoking index, intelligence and childhood maltreatment were not found to be associated with IBS but were linked to coexisting disorders ( figure 6 ). In the replication dataset, multisite chronic pain was again causally associated with IBS, gastro-oesophageal reflux disease and intestinal diverticula. However, lifetime smoking index, college or university degree, intelligence, childhood maltreatment and frailty index were not independently associated with IBS, but were instead linked to coexisting disorders. Overall, both datasets consistently showed that the lifetime smoking index, intelligence and childhood maltreatment were not associated with IBS, despite their links to coexisting disorders. However, the associations of college or university degree and frailty index with IBS and coexisting disorders were inconsistent across the discovery and replication datasets. In addition, we conducted sensitivity analyses focusing on factors with convincing evidence of associations with IBS. We found robust associations of multisite chronic pain with both IBS and gastro-oesophageal reflux disease ( online supplemental figure S2 ). Lifetime smoking index and childhood maltreatment were not independently associated with IBS but were linked to coexisting disorders.

Through a systematic review of factors related to IBS, we observed an evolution in the studies exploring these associations between factors and IBS. Early research focused on bacteria, digestive system disorders, psychiatric disorders, diet, air pollution, medication and surgical treatments. Recently, there has been an increased emphasis on COVID-19 infection and lifestyle factors. Future research should consider the comprehensive impact of treatments on IBS. Since the absence of Food and Drug Administration-approved drugs for IBS, modifying factors, especially in recent years attention to the change of lifestyle, emerge as key strategies for preventing IBS. At the same time, we need to explore joint prevention and treatment of diseases that primarily coexist with IBS. Further research is required to elucidate the mechanisms linking these modifications and IBS and its coexisting disorders.

Our study provided the most comprehensive evidence to date on the causal association between modifiable factors, coexisting disorders and IBS. This evidence was based on extensive observational studies that encompassed a broad range of exposure factors, and was supported by robust genetic analyses. The majority of the modifiable factors and coexisting disorders exhibited genetic associations with IBS. Notably, factors such as multisite chronic pain, gastro-oesophageal reflux disease, neuroticism score, depression, well-being spectrum, life satisfaction and positive affect demonstrated convincing causal associations with IBS in MR analyses. Additionally, smoking, alcohol frequency, education, intelligence, childhood maltreatment, frailty index, diverticular disease of intestine and schizophrenia showed highly suggestive causal associations with IBS. We also found that multisite chronic pain was causally associated with IBS and shared with coexisting diseases, while lifetime smoking index, intelligence and childhood maltreatment were not independently associated with IBS but were linked to coexisting disorders. Compared with previous MR studies focusing on single factor and multiple factors for IBS, 1043 49 we first systematically summarised the existing observational studies, including cohort and meta-analyses, that investigated the associations between factors (modifiable factors and coexisting disorders) and IBS. We then conducted comprehensive MR analyses to explore the genetic correlations between these factors and IBS, employing multiple methods to assess the overall stability of the MR results. Additionally, we used MR 2 approach to account for the potential impact of coexisting disorders on the associations between modifiable factors and IBS. Our study provided a relatively updated framework for MR analyses and offered insights into the exploration of modifiable factors for IBS. IBS is widely recognised as a functional gastrointestinal disorder caused by brain-gut axis dysfunction, and found to coexist with gastrointestinal and psychiatric disorders, sharing common risk factors and genetic backgrounds. 7 19 50 In our study, we identified stronger genetic correlations between IBS and coexisting disorders than with modifiable factors. Incorporating MR analyses, we established convincing causal factors of IBS, including gastro-oesophageal reflux disease, well-being spectrum including life satisfaction, positive affect, neuroticism score, and depression, and highly suggestive causal associations of diverticular disease of intestine and schizophrenia with IBS. In addition, MR 2 results indicated that IBS and coexisting disorders shared modifiable factors. This suggested that gastrointestinal and psychiatric disorders had causal associations with IBS and exhibited strong correlations possibly due to shared underlying factors. Therefore, it is crucial to account for the coexistence of gastrointestinal and psychiatric disorders when investigating modifiable risk factors for IBS. Our study found a convincing causal association of multisite chronic pain with IBS, along with highly suggestive causal associations of smoking, alcohol frequency, education, intelligence, childhood maltreatment and frailty index with IBS. Importantly, multisite chronic pain remained associated with IBS even after accounting for the impact of coexisting disorders, while lifetime smoking index, intelligence and childhood maltreatment were not independently associated with IBS but were linked to coexisting disorders in both the discovery and replication datasets. These findings suggested that multisite chronic pain may be an important factor contributing to the development of IBS, whereas the associations of lifetime smoking index, intelligence and childhood maltreatment with IBS may be mediated by coexisting disorders. Therefore, identifying modifiable factors for IBS necessitates careful consideration of the impact of coexisting conditions, highlighting the importance of targeted interventions that address both IBS and coexisting disorders. The impact of lifestyle factors on IBS is a topic that cannot be overlooked. Two recent studies have examined the associations of lifestyles, including smoking, alcohol consumption, physical activity, diet and sleeping, with IBS. 9 11 However, their findings appear to be inconsistent, with one study reporting no significant associations of diet and alcohol consumption with IBS and the other suggesting a potential association. In our study, we found that, apart from smoking and alcohol consumption, other lifestyle factors did not yield consistent or significant results. The discrepancies between our findings and those reported in observational studies on associations of coffee consumption, tea intake, physical activity and unhealthy diet with IBS 9 51 may be attributed to confounding factors that are challenging to control for even with advanced statistical adjustments. These confounding factors may include socioeconomic status and medical comorbidities associated with unhealthy lifestyle choices including dietary habits. In addition to lifestyle factors, social determinants have also been shown to be associated with IBS in observational studies. 52 We found highly suggestive causal associations of education, intelligence and childhood maltreatment with IBS. These factors were associated with a higher risk of psychiatric disorders and mortality 53 55 and warrant attention for their associations with IBS. In addition, our MR 2 results indicated that intelligence and childhood maltreatment were associated with IBS, potentially influenced by coexisting disorders. This finding needs to be further verified. Ongoing studies suggested that bacterial and viral infections may be triggers for IBS. The immune inflammatory response is an important mechanism of IBS. 6 56 A meta-analysis showed that C reactive protein could serve as a promising predictor for IBS. 57 Although we observed a genetic correlation between COVID-19 and IBS, we found no evidence of causal relationship of COVID-19 with IBS in MR analysis. During the 3 years of COVID-19 pandemic, gastrointestinal disorders including IBS were the main symptom of Long COVID. 58 This raises concerns about the possibility for IBS to emerge as a post-infectious complication, warranting continued vigilance. Extraintestinal somatic symptoms are closely related to IBS, possibly reflecting the bidirectional nature of the gut-brain axis. 7 19 50 Observational studies have reported associations of migraines and atrial fibrillation with IBS, 59 60 but our MR results did not reveal strong relationships. However, the robust association of multistate pain with IBS is an innovative finding in our study. Our MR 2 results confirmed this association even after accounting for the presence of coexisting disorders. Additionally, we observed another intriguing factor: frailty status may have a causal relationship with IBS based on genetic evidence, aligning with previous observational studies. 12 Further investigation into whether interventions targeting multisite pain and frailty could help prevent IBS is warranted. Although the association between autoimmune diseases and gynaecological diseases with IBS is well-recognised, the specific mechanisms underlying these connections remain unclear. Our study provided genetic correlation but was limited in establishing a reliable causal relationship. Further research is necessary to deepen our understanding of the interplay between autoimmune diseases, gynaecological disorders and IBS, which could lead to improved diagnostic and treatment strategies for affected patients. There were several strengths in our study. A comprehensive systematic review of observational studies (cohort study and meta-analysis) and a comprehensive MR approach were used to consolidate our findings. Overall, we combined genetic correlation and MR methods as analytical frameworks and accounted for the potential shared effect of modifiable factors on IBS and coexisting disorders to ensure robust findings. Our study reduced bias by addressing the issues of weak IVs, horizontal pleiotropy, outliers and sample overlap, making our estimates of the causal associations more precise and less prone to bias compared with previous estimates. Our study still faced some limitations. First, the selection of exposure factors was confined to cohort studies and meta-analyses, which may have resulted in the omission of some relevant factors. However, we also referred to several systematic reviews to supplement any potentially missed important factors. We identified some important factors that were not investigated in our further MR approach, such as environmental pollution, drug use and surgery. Aside from COVID-19, summary-level GWAS data for specific bacterial or viral infections were unavailable. Second, one distinctive feature of our study is that we considered the impact of coexisting disorders on IBS, focusing primarily on gastrointestinal and psychiatric disorders as coexisting disorders. Genetic correlation analysis also revealed that infections and autoimmune diseases could also be coexisting disorders. Third, due to power limitations, some exposure factors could not be included with sufficient SNPs as IVs in the MR analysis. For some exposure factors, the variance explained by the genetic instruments was relatively modest, which might limit the power to detect causality. Fourth, although we attempted to assess and address the biases of weak IV and genetic pleiotropy, our MR analysis still could not entirely eliminate these concerns. While the CAUSE method could consider both related and unrelated pleiotropy and provide some new insights, there are some inherent instability factors associated with its use, such as complexity of model assumptions, dependence on sample size and computational burden. Fifth, modifiable factors, especially behavioural patterns and dietary preferences, can be influenced by many factors, with genetic factors likely playing a relatively modest role. To avoid the misuse of genetic evidence, more research is needed in the future to explore the genetic background of modifiable factors in depth. Sixth, the bias risk assessment tools for MR study are still lacking, and future studies need to be explored. Given our consideration of various evidence, we adopted a relatively lenient p value threshold of <0.05 to identify potential associations, which may increase the risk of multiple false positives. Seventh, while MR 2 provides a valuable approach for the shared and unique exposures of coexisting disorders, it may also face challenges related to result stability. Although we employed sensitivity analyses to enhance the stability of the results, caution was warranted in their interpretation. Finally, many GWAS have predominantly focused on European populations, which may limit the generalisation of our conclusions to other racial and ethnic groups. Future research should prioritise validating our findings across diverse populations to enhance their applicability.

In conclusion, our study identified a wide range of potential modifiable factors and coexisting disorders associated with IBS through genetic evidence. These factors included multisite chronic pain, smoking, alcohol frequency, education, intelligence, childhood maltreatment, frailty index, gastro-oesophageal reflux disease, well-being spectrum, life satisfaction, positive affect, neuroticism score, depression, diverticular disease of intestine and schizophrenia. The presence of multisite chronic pain may offer opportunities for concurrent prevention strategies targeting both IBS and its coexisting disorders.