Selecting likely causal risk factors from high-throughput experiments using multivariable Mendelian randomization

In The Lancet Diabetes & Endocrinology, Jon White and colleagues1White J Sofat R Hemani G et al.Plasma urate concentration and risk of coronary heart disease: a Mendelian randomisation analysis.Lancet Diabetes Endocrinol. 2016; (published online Jan 15.)http://dx.doi.org/10.1016/S2213-8587(15)00386-1Google Scholar consider the causal role of plasma urate in coronary heart disease, using the technique known as mendelian randomisation. They assess the association between genetic predictors of the risk factor (plasma urate) and the disease outcome (coronary heart disease), noting a statistically robust association between a genetic score based on 31 variants associated with plasma urate and coronary heart disease risk, with a straightforward application of Mendelian randomisation estimating an 18% (95% CI 8-29) relative increase in coronary heart disease risk per 1 SD increase in plasma urate. This positive finding from the mendelian randomisation investigation was in line with the observational epidemiological analysis also reported by the investigators. However, the association of the genetic score with the outcome seems to be at least partly explained by pleiotropic (off-target) associations of the genetic score, and the pattern of associations of the individual genetic variants does not show the characteristic dose–response association that would be expected if the genetic associations were solely driven by the effect of the risk factor on the outcome. Mendelian randomisation is a relatively new, but well established epidemiological technique to assess whether a modifiable risk factor is a worthwhile target for clinical or pharmacological intervention.2Davey Smith G Ebrahim S ‘Mendelian randomization’: can genetic epidemiology contribute to understanding environmental determinants of disease?.Int J Epidemiol. 2003; 32: 1-22Crossref PubMed Scopus (2488) Google Scholar, 3Burgess S Thompson SG Mendelian randomization: methods for using genetic variants in causal estimation. Chapman & Hall, London2015Crossref Scopus (195) Google Scholar An association between a genetic predictor of the risk factor and the outcome is more likely to reflect that the risk factor has a causal relation with the outcome, compared with an association of the risk factor itself from a traditional epidemiological analysis, for several reasons. First, risk factors tend to be mutually correlated. It is difficult to distinguish whether the risk factor itself is the cause, or if a correlated risk factor is driving the identified association with an outcome. By contrast, most genetic variants tend to be uncorrelated with conventional epidemiological risk factors.4Davey Smith G Lawlor D Harbord R Timpson N Day I Ebrahim S Clustered environments and randomized genes: a fundamental distinction between conventional and genetic epidemiology.PLoS Med. 2007; 4: e352Crossref PubMed Scopus (329) Google Scholar Second, genetic variants are fixed at conception. Hence, a genetic association cannot be affected by external factors that lead to confounded associations. Finally, the associations of genetic variants are not subject to reverse causation. The gene can only ever be the cause, and never the effect. In an ideal mendelian randomisation analysis, the genetic variants have a clear biological link with the risk factor (for example, genetic variants in the CRP gene region for the analysis of C-reactive protein5CRP CHD Genetics CollaborationAssociation between C reactive protein and coronary heart disease: Mendelian randomisation analysis based on individual participant data.BMJ. 2011; 342: d548Crossref PubMed Scopus (210) Google Scholar), and are associated with the risk factor, but not with alternative risk factors. The analysis proceeds analogously to a randomised trial,6Hingorani A Humphries S Nature's randomised trials.Lancet. 2005; 366: 1906-1908Summary Full Text Full Text PDF PubMed Scopus (141) Google Scholar and compares groups of individuals with (on average) genetically raised levels of the risk factor versus those with genetically lowered levels to assess whether the risk factor is a cause of the outcome, and therefore a promising target for prioritisation in drug development. White and colleagues' analysis1White J Sofat R Hemani G et al.Plasma urate concentration and risk of coronary heart disease: a Mendelian randomisation analysis.Lancet Diabetes Endocrinol. 2016; (published online Jan 15.)http://dx.doi.org/10.1016/S2213-8587(15)00386-1Google Scholar is far from an ideal mendelian randomisation investigation. This is not a criticism, but a consequence of their investigation of a risk factor that is not a protein biomarker, and so does not have a coding gene region, as in the case of C-reactive protein. However, the consideration of multiple gene regions gives some robustness in the analysis, in that the consistency of any causal finding can be assessed. If multiple independent genetic variants associated with the risk factor are all concordantly associated with the outcome (as in the case of LDL cholesterol with coronary heart disease7Ference BA Yoo W Alesh I et al.Effect of long-term exposure to lower low-density lipoprotein cholesterol beginning early in life on the risk of coronary heart disease: a Mendelian randomization analysis.J Am Coll Cardiol. 2012; 60: 2631-2639Summary Full Text Full Text PDF PubMed Scopus (544) Google Scholar), the separate genetic associations each provide independent evidence for a causal effect of the risk factor on the outcome, strengthening the evidence for a causal finding. Another deficiency is that the composite genetic instrument considered by White and colleagues1White J Sofat R Hemani G et al.Plasma urate concentration and risk of coronary heart disease: a Mendelian randomisation analysis.Lancet Diabetes Endocrinol. 2016; (published online Jan 15.)http://dx.doi.org/10.1016/S2213-8587(15)00386-1Google Scholar is not solely associated with urate concentrations, but also shows associations with other risk factors, including blood pressure and lipid fractions. This finding is not necessarily a problem provided that the genetic associations with these factors are mediated by urate concentrations, and so one single causal pathway is represented. However, these associations might reflect pleiotropy of the genetic variants, meaning that there are multiple causal pathways from gene to disease, thereby violating the assumption of mendelian randomisation that an association between gene and disease reflects a causal effect of the risk factor. White and colleagues1White J Sofat R Hemani G et al.Plasma urate concentration and risk of coronary heart disease: a Mendelian randomisation analysis.Lancet Diabetes Endocrinol. 2016; (published online Jan 15.)http://dx.doi.org/10.1016/S2213-8587(15)00386-1Google Scholar show a statistically robust association between their 31 variant composite instrument and disease risk. However, the authors are to be commended in not stopping here, as others have done previously, but instead seeking to assess whether potential violations of the instrumental variable assumptions invalidate a causal interpretation for their findings. They use two main statistical approaches: multivariable mendelian randomisation8Burgess S Thompson S Multivariable Mendelian randomization: the use of pleiotropic genetic variants to estimate causal effects.Am J Epidemiol. 2015; 181: 251-260Crossref PubMed Scopus (434) Google Scholar and Egger regression.9Bowden J Davey Smith G Burgess S Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression.Int J Epidemiol. 2015; 44: 512-525Crossref PubMed Scopus (1958) Google Scholar In multivariable mendelian randomisation, alternative risk factors are accounted for by adjusting for the genetic associations with these risk factors in the analysis model. In Egger regression, the assessment is not simply of whether genetic variants that are associated with the risk factor are also associated with the outcome, but whether there is a dose–response relation in these associations—that is, are genetic variants that are more strongly associated with the risk factor also more strongly associated with the outcome? Egger regression also encompasses a test for directional pleiotropy—whether this dose–response relation suggests that a variant having zero association with the risk factor also has zero association with the outcome (as expected if there is no pleiotropy, or balanced pleiotropy). Although multivariable mendelian randomisation does not suggest that measured pleiotropy completely accounts for the causal finding, Egger regression detects pleiotropy among the genetic variants, and provides a weaker causal estimate (a 5% [95% CI −8 to 20] relative increase coronary heart disease risk per 1 SD increase in plasma urate) that is compatible with the null once this pleiotropy is accounted for. Additionally, the genetic associations with the outcome are more variable than would be expected by chance alone; this heterogeneity suggests that the genetic associations with coronary heart disease risk are not solely mediated via plasma urate concentrations. The overall finding is not unequivocal. There is an undeniable association between genetic predictors of plasma urate concentrations and risk of coronary heart disease. However, there are suggestions that the association might represent pleiotropic effects of genetic variants included in the model, rather than a causal effect of urate. In summary, the evidence presented honestly by White and colleagues1White J Sofat R Hemani G et al.Plasma urate concentration and risk of coronary heart disease: a Mendelian randomisation analysis.Lancet Diabetes Endocrinol. 2016; (published online Jan 15.)http://dx.doi.org/10.1016/S2213-8587(15)00386-1Google Scholar suggests that urate lowering should be prioritised as a potential mechanism to improve cardiovascular outcomes, but without providing a smoking gun that puts the question of causality beyond doubt. I declare no competing interests. Plasma urate concentration and risk of coronary heart disease: a Mendelian randomisation analysisConventional and multivariate Mendelian randomisation analysis implicates a causal role for urate in the development of coronary heart disease, but these estimates might be inflated by hidden pleiotropy. Egger Mendelian randomisation analysis, which accounts for pleiotropy but has less statistical power, suggests there might be no causal effect. These results might help investigators to determine the priority of trials of urate lowering for the prevention of coronary heart disease compared with other potential interventions. Full-Text PDF Open Access

Background: Tobacco smoking and alcohol consumption have been associated with frailty in observational studies. We sought to examine whether these associations reflect causality using the two-sample Mendelian randomization (MR) design. Methods: We used summary genome-wide association statistics for smoking initiation (N = 2,669,029), alcohol consumption (N = 2,428,851), and the frailty index (FI, N = 175,226) in participants of European ancestry. Both univariable and multivariable MR were performed to comprehensively evaluate the independent effects of smoking and alcohol consumption on the FI, accompanied by multiple sensitivity analyses. Results were verified using lifetime smoking and alcohol use disorder. Reverse direction MR was undertaken to assess the potential for reverse causation. Results: Genetic predisposition to smoking initiation was significantly associated with increased FI (univariable MR: β = 0.345; 95% confidence interval [CI] = 0.316 to 0.374; p = 1.36E-113; multivariable MR: β = 0.219; 95% CI = 0.197 to 0.241; p = 2.44E-83). Genetically predicted alcohol consumption showed a suggestive association with the FI (univariable MR: β = -0.090; 95% CI = -0.151 to -0.029; p = 0.003; multivariable MR β = -0.153; 95% CI = -0.212 to -0.094; p = 2.03E-07), with inconsistent results in sensitivity analyses. In complementary analysis, genetic predicted lifetime smoking, but not alcohol use disorder was associated with the FI. There is no convincing evidence for reverse causation. Conclusion: The present MR study supported smoking as a causal risk factor of frailty. Further research is warranted to investigate whether alcohol consumption has a causal role in frailty.

Cachexia-related consequences of glycemic metabolism: A multivariable and two-step Mendelian randomization study

People with university degrees have a lower incidence of Alzheimer's Disease (AD). However, the relationship between education and AD could be due to selection, collider, and ascertainment biases, such as lower familiarity with cognitive testing or the fact that those with degrees are more likely to participate in research. Here, we use two-sample Mendelian randomization (MR) to investigate the causal relationships between education, participation, and AD. We used genome-wide association study (GWAS) summary statistics for educational attainment, three different measures of participation, AD (clinically diagnosed AD), and AD/ADRD (clinical diagnosis and family history of AD and related dementias). Independent genome-wide significant single nucleotide polymorphisms (SNPs) were extracted from the exposure summary statistics and harmonized with the outcome SNPs. Fixed-effects inverse variance weighted meta-analysis was the primary MR method; Cochran's Q statistic and MR Egger intercept were used to test for heterogeneity and pleiotropy, and Radial-MR was used to identify outliers. Sensitivity analyses included MR Egger, Weighted Median, and Weighted mode. Bidirectional analyses were used to test if AD pathology affects participation and multivariable MR (MVMR) assessed independent exposure-outcome effects. Educational attainment reduced the risk of AD (OR IVW 95% CI= 0.70 [0.63, 0.79], p = 8e-10), and the results were robust based on sensitivity analyses. However, education increased the risk of AD/ADRD, though the results were not robust to sensitivity analyses (OR IVW 95% CI= 1.09 [1.02, 1.15], p = 0.006). Participation in MHQ reduced the odds of AD (OR IVW 95% CI= 0.325 [0.128, 0.326], p = 0.01). When adjusting for participation in MVMR, education remained associated with a reduced risk of AD (OR IVW 95% CI= 0.76 [0.62, 0.92], p = 0.006). Univariate MR analyses indicated that education and participation reduced the risk of AD. However, MR also suggested that education increased the risk of AD/ADRD, highlighting the inconsistencies between clinical and proxy diagnoses of AD, as proxy-AD may be affected by selection, collider, or ascertainment bias. MVMR indicated that participation is unlikely to explain the effect of education on AD identified in MR, and the protective effect of educational attainment may be due to other biological mechanisms, such as cognitive reserve.

Background: Previous observational studies have suggested a potential association between Epstein-Barr virus (EBV) infection and the development of multiple myeloma (MM), but this relationship is not clear. Therefore, we conducted a systematic Mendelian randomization (MR) analysis to investigate the causal relationship between EBV infection and the risk of MM, while exploring the possible mediating role of immune cells in this association. Methods: The study first conducted a two-sample MR analysis using the MM R11 dataset from the FinnGen Consortium to evaluate the causal relationship between five EBV infection-related antibodies (AEB-IgG, EA-D, EBNA-1, VCA-p18, and ZEBRA) and MM, with validation in the MM R10 dataset. A reverse MR analysis was then performed. For significant results, multivariable MR (MVMR) was used to adjust for the effects of confounding risk factors. Next, a two-step MR mediation analysis was applied to investigate the potential mediating role of 731 immune cell types between positive exposure and MM. Multiple sensitivity analyses were conducted to assess the robustness of the findings. Results: A two-sample MR study found that EBNA-1 antibodies (OR = 1.36, 95% CI: 1.06–1.73; P = 0.015) were associated with an increased risk of MM, with similar results observed in the FinnGen Consortium R10 replication study. Although the association did not remain statistically significant after false discovery rate (FDR) adjustment (P_fdr = 0.075), further adjustment for relevant confounders using multivariable MR (MVMR) demonstrated that EBNA-1 antibodies (OR = 1.33, 95% CI: 1.01–1.75; P = 0.041) were still significantly associated with an increased risk of MM. Reverse MR analysis indicated no causal effect of MM on EBV-related antibodies. A two-sample MR analysis involving 731 immune cell phenotypes identified 27 potential mediating cell types. Ultimately, two-step MR confirmed that HLA-DR on myeloid dendritic cells (HLA-DR⁺ mDC) serves as a mediating factor, with EBNA-1 antibodies downregulating HLA-DR⁺ mDC, thereby increasing MM risk. Multiple sensitivity analyses supported the robustness of these findings. Conclusion: The findings of this study suggest that EBNA-1 antibodies may increase the risk of MM by downregulating HLA-DR⁺ mDC. This indicates that chronic EBV infection may contribute to an elevated risk of MM. We hope these results provide new insights for future research on the prevention and treatment of MM.

Periodontitis drives irreparable destruction of periodontal tissue and possesses a tendency to aggravate inflammatory disorders. Systemic immunomodulation management remains an attractive approach in periodontal care within a context of “predictive, preventive, and personalized” periodontics. Leveraging a comprehensive Mendelian randomization (MR), the present study aims to evaluate the causal relationships between circulating immune cells and the risk of periodontitis.Genetic proxies for circulating immune cells and periodontitis were obtained from genome-wide association studies. We performed a two-sample bidirectional univariable Mendelian Randomization, followed by sensitivity testing, multivariable MR, subgroup analysis, and Bayesian model averaging (MR-BMA) to explore the causal association between them. The transcriptome-wide association study (TWAS) and colocalization analysis were further conducted to identify responsible genes as immunomodulation candidates for periodontal care.MR study indicated that circulating neutrophils, Natural Killer T cells, and plasmacytoid Dendritic Cells were relative to a higher risk of periodontitis, with no evidence of heterogeneity or pleiotropy. The MR-BMA identified neutrophils as the primary factor responsible for periodontitis. The TWAS recognized five cross-trait genes to be involved in their interaction. Two high-confidence genes on 1q21.3, S100A9 and S100A12, could represent immunomodulation targets for neutrophils in periodontitis.The present study suggests the causative association between circulating immune cells and the risk of periodontitis. Our findings highlight the prospect of systemic immunomodulation management in periodontal care, which can be valuable for early diagnostics, risk assessment, targeted prevention, and personalized management of periodontitis. More research is required to comprehend the biological plausibility and clinical applicability of our findings.This work was supported by the Major Program of National Natural Science Foundation of China (No. 81991500 & 81991502), the Fundamental Research Funds for the Central Universities (No. 226-2023-00121 & 226-2022-00213), Zhejiang University Global Partnership Fund (No. 188170 & 194452307/004) and the Joint Funds of the Zhejiang Provincial Natural Science Foundation of China (No. LHDMD23H300001)

BackgroundAlzheimer’s disease (AD) is predicted to affect 132 million people by 2050. Targeting modifiable lifestyle risk factors that are associated with an increased risk of AD could prevent a large proportion of dementia cases, allowing people to reach the end of their life dementia free. However, evidence obtained from observational studies does not take into account how risk factors are correlated with one another, and whether they causally contribute to increased AD risk. In this study, we determine whether the relationship between twelve previously speculated AD risk factors and AD susceptibility is consistent with causality using large‐scale genetic data (table 1).MethodWe conduct multivariable Mendelian Randomization (MR) analyses to identify causal risk factors of AD while controlling for genetic pleiotropy, using genetic instruments from the most well powered GWAS studies currently available. We also use GWAS‐by‐subtraction implemented with genomic structural equation modelling (genomic SEM) to understand the cognitive and non‐cognitive components of the causal association of EA on AD risk.ResultWe identified significant positive correlations between AD and physical activity and loneliness, and significant negative correlations with educational attainment (EA), intelligence and household income. Using GWAS‐by‐subtraction to partition EA into cognitive (intelligence) and noncognitive genetic factors, we showed a significant genetic correlation between AD and the cognitive factor, but not the noncognitive factor (figure 1).Univariable MR found EA, intelligence, household income and the cognitive factor to all be causally protective against AD. Multivariable MR analysis showed that higher intelligence lowers AD risk, independent of EA and household income. There was no evidence that EA lowers AD risk, independent of intelligence. In addition the cognitive (intelligence) component of EA decreased AD risk with no significant influence of the noncognitive factor (table 2).ConclusionFrom the risk factors tested only the cognitive component of EA (i.e., intelligence) is an independent causal risk factor for AD. To reduce incidence of AD this cognitive aspect of EA would need to be modified. It is questionable whether substantial modification of intelligence is achievable through simple interventions such as improving access to extended education.

BackgroundCharcot foot or Charcot neuropathic joint disease (CN) is a rare and complex foot disease with unknown pathogenesis, hindering early identification and intervention. The study aimed to clarify the causal association between all predominant risk factors and CN.MethodsTwo-sample Mendelian Randomization (MR), Multivariate MR, and Bidirectional MR analyses investigated the causal association between 36 modifiable risk factors and CN. The causal relationship between CN and Inflammatory cytokine and immune cells was also analyzed.ResultsGenetic factors associated with obesity and genetic susceptibility to various autoimmune diseases and non-cancerous thyroid diseases increased the risk of CN (P < .05), genetically associated high basal metabolic rate and high total cholesterol decreased the risk of CN (P < .05). In addition, we found a bi-directional causal relationship between CN and diabetes. In further immune cell analysis, we found 8 CN related immune cells, and in inflammatory cytokine analysis, we found 2 CN related inflammatory cytokines.ConclusionsThis comprehensive MR Study supports the causal role of Obesity-related factors, diabetes, autoimmune-related factors, and smoking in the development and progression of CN. This study identifies a potential cause of CN that has not been identified in previous studies and provides a new direction for further research.

To disentangle tobacco constituents in psoriasis, we contrasted nicotine exposure-proxied by the nicotine metabolite ratio (NMR)-with smoking intensity (cigarettes per day, CPD) and evaluated cross-ancestry effects. This study applied multivariable Mendelian randomization (MR) jointly modeling genetically proxied NMR (instrumented using variants from a European-ancestry GWAS) and CPD to estimate their independent effects on liability to psoriasis and psoriatic arthritis (PsA). Chronic obstructive pulmonary disease (COPD) served as a positive control. Cross-ancestry generalizability was tested using a trans-ethnic MR (TEMR) framework under conditional likelihood with Nelder-Mead optimization. Sensitivity analyses assessed pleiotropy, heterogeneity, directionality (Steiger), MRLap, RadialMR, and multiple testing (Benjamini-Hochberg). NMR showed an independent association with higher PsA risk irrespective of CPD (OR = 1.104, 95% CI: 1.039-1.174). CPD retained an independent effect on overall psoriasis after conditioning on NMR (OR = 1.305, 95% CI: 1.082-1.573), while the NMR effect on psoriasis attenuated (P > 0.05). In univariable MR, genetically predicted NMR increased psoriasis risk in Europeans (EUR; OR = 1.032, 95% CI: 1.013-1.051). CPD associated with psoriasis in EUR (OR = 1.130, 95% CI: 1.031-1.239) and strongly in Hispanics (HIS; OR = 1.448, 95% CI: 1.434-1.463), with suggestive evidence in East Asians. Reverse-direction MR indicated psoriasis liability correlated with lower CPD across EUR, EAS, AFR, and HIS (β < 0, Padj < 0.05). This study supports ancestry-specific differences and suggests distinct roles of nicotine-related versus non-nicotine tobacco smoke constituents in psoriasis and its subtypes, while the underlying biological mechanisms remain to be clarified.

Phenotypic heterogeneity at genomic loci encoding drug targets can be exploited by multivariable Mendelian randomization to provide insight into the pathways by which pharmacological interventions may affect disease risk. However, statistical inference in such investigations may be poor if overdispersion heterogeneity in measured genetic associations is unaccounted for. In this work, we first develop conditional F statistics for dimension-reduced genetic associations that enable more accurate measurement of phenotypic heterogeneity. We then develop a novel extension for two-sample multivariable Mendelian randomization that accounts for overdispersion heterogeneity in dimension-reduced genetic associations. Our empirical focus is to use genetic variants in the GLP1R gene region to understand the mechanism by which GLP1R agonism affects coronary artery disease (CAD) risk. Colocalization analyses indicate that distinct variants in the GLP1R gene region are associated with body mass index and type 2 diabetes (T2D). Multivariable Mendelian randomization analyses that were corrected for overdispersion heterogeneity suggest that bodyweight lowering rather than T2D liability lowering effects of GLP1R agonism are more likely contributing to reduced CAD risk. Tissue-specific analyses prioritized brain tissue as the most likely to be relevant for CAD risk, of the tissues considered. We hope the multivariable Mendelian randomization approach illustrated here is widely applicable to better understand mechanisms linking drug targets to diseases outcomes, and hence to guide drug development efforts.

Understanding the etiological risk factors for retinal vascular occlusion (RVO) is critical for prevention and treatment. While the effects of cardiovascular events, hypertension, glaucoma, obesity and glycemic risk factors on RVO are still controversial. This study employed two-sample Mendelian randomization (MR) analysis to investigate these causal risk factors. Single-nucleotide polymorphisms (SNPs) were used as instrumental variables (IVs). Genetic instruments for hypertension, glaucoma, obesity, cardiovascular events and glycemic risk factors were obtained from published genome-wide association studies (GWASs). Summary-level data for RVO and hypertension were obtained from the FinnGen consortium. MR analysis primarily utilized the inverse variance weighted (IVW) method, with MR-Egger and weighted median as supplementary approaches. Multivariable MR (MVMR) adjusting for hypertension or glaucoma of RVO were conducted. Heterogeneity was assessed using Cochrane’s Q test and I2, while MR-Egger intercept and MR-PRESSO tested horizontal pleiotropy. All MR analyses were performed within R software (4.1.3) using the R packages “TwoSampleMR” and “MR-PRESSO.” Genetic instruments for hypertension and glaucoma were significantly associated with RVO risk. A one-standard deviation (SD) increase in hypertension was associated with a higher risk of RVO [OR = 1.577, 95% CI = (1.342, 1.854), P < .001], while a one-SD increase in the log odds of genetically predicted glaucoma was associated with a higher risk of RVO [OR = 1.24, 95% CI = (1.115, 1.379), P < .001]. Meanwhile, hypertension and glaucoma were still significant in multivariable MR. There was not sufficient evidence to suggest cardiovascular events and obesity were associated with RVO risk. This MR study provided genetic evidence supporting that hypertension and glaucoma were causally associated with the risk of RVO. It may help guide clinical decisions in the management of RVO patients with hypertension and glaucoma.

ObjectiveThis study aimed to investigate the causal associations between several liver traits (liver iron content, percent liver fat, alanine transaminase levels, and liver volume) and colorectal cancer (CRC) risk using a Mendelian randomization (MR) approach to improve our understanding of the disease and its management.MethodsGenetic variants were used as instrumental variables, extracted from genome-wide association studies (GWAS) datasets of liver traits and CRC. The Two-Sample MR package in R was used to conduct inverse variance weighted (IVW), MR Egger, Maximum likelihood, Weighted median, and Inverse variance weighted (multiplicative random effects) MR approaches to generate overall estimates of the effect. MR analysis was conducted with Benjamini-Hochberg method-corrected P values to account for multiple testing (P < 0.013). MR-PRESSO was used to identify and remove outlier genetic variants in Mendelian randomization (MR) analysis. The MR Steiger test was used to assess the validity of the assumption that exposure causes outcomes. Leave-one-out validation, pleiotropy, and heterogeneity testing were also conducted to ensure the reliability of the results. Multivariable MR was utilized for validation of our findings using the IVW method while also adjusting for potential confounding or pleiotropy bias.ResultsThe MR analysis suggested a causal effect between liver volume and a reduced risk of CRC (OR 0.60; 95% CI, 0.44–0.82; P = 0.0010) but did not provide evidence for causal effects of liver iron content, percent liver fat, or liver alanine transaminase levels. The MR-PRESSO method did not identify any outliers, and the MR Steiger test confirmed that the causal direction of the analysis results was correct in the Mendelian randomization analysis. MR results were consistent with heterogeneity and pleiotropy analyses, and leave-one-out analysis demonstrated the overall values obtained were consistent with estimates obtained when all available SNPs were included in the analysis. Multivariable MR was utilized for validation of our findings using the IVW method while also adjusting for potential confounding or pleiotropy bias.ConclusionThe study provides tentative evidence for a causal role of liver volume in CRC, while genetically predicted levels of liver iron content, percent liver fat, and liver alanine transaminase levels were not associated with CRC risk. The findings may inform the development of targeted therapeutic interventions for colorectal liver metastasis (CRLM) patients, and the study highlights the importance of MR as a powerful epidemiological tool for investigating causal associations between exposures and outcomes.

This study uses Mendelian randomisation (MR) to investigate the causal link between SLE and osteoporosis across different ethnic groups. Genetic variants associated with SLE were identified from publicly available genome-wide association studies in European and East Asian populations. Two-sample MR (TSMR) analysis and meta-analysis with inverse variance weighting (IVW) assessed their effects on bone mineral density (BMD) and fracture risk. Multivariable MR (MVMR) analysis in East Asians adjusted for potential mediators, and two-step mediation analysis evaluated mediation effects of independent covariates. A meta-analysis of IVW results from TSMR in East Asian populations revealed a significant positive genetic association of SLE with osteoporosis (OR=1.023, CI 1.007 to 1.040, p<0.01). A similar, although weaker, association was observed in the European population (OR=1.001, CI 1.000 to 1.001, p<0.01). Furthermore, SLE was identified as a risk factor for reduced BMD in both East Asian (β=-0.0690, p<0.05) and European (β=-0.0109, p<0.05) populations, and for fracture risk in European (OR=1.002, p<0.05) populations, while no significant association was observed in the East Asian population (OR=1.010, p=0.705). MVMR analysis of East Asian data assessed mediation effects and found that the SLE-osteoporosis association was nullified after adjusting for cardiovascular disease and health status. Mediation analysis identified low-density lipoprotein cholesterol (LDL-C) and anti-inflammatory medication use as independent mediators, with mediation effects of 0.1170 and 0.0510, respectively. No significant heterogeneity or pleiotropy was detected. SLE appears to be a causal risk factor for osteoporosis. LDL-C and anti-inflammatory medication use mediate this relationship, suggesting the importance of managing these factors in patients with SLE to reduce osteoporosis risk.

Background Observational evidence indicates that patients with atrial fibrillation (AF) are at a four- to fivefold increased risk of stroke. By two-sample mendelian randomization (MR), we have shown in our previous work that specifically the cardioembolic stroke subtype is caused by AF. This risk elevation is however variable and appears to be modulated by other risk factors such as hypertension traits and coronary artery disease (CAD). Purpose This study seeks to determine (1) whether genetic predisposition to cardioembolic stroke by AF is independent from modulators, and (2) whether the other stroke subtypes are causally linked with different exposures. Methods The associations between AF, hypertension traits, and CAD as exposures and stroke subtypes as outcomes were individually tested in a two-sample mendelian randomisation (MR) analysis using genetic instruments from publicly available genome-wide association studies (GWAS) in large cohorts. For atrial fibrillation, genetic instruments were obtained from the dataset by Nielsen et al. with 60,620 atrial fibrillation cases and 970,216 controls, collected from six contributing studies including the AFGen Consortium. Associations with stroke subtypes were evaluated in the MEGASTROKE genome-wide association study data set (67, 162 cases; 454, 450 controls). The genetic markers for hypertension traits were taken from a GWAS by Evangelou et al. which combines data from the UK Biobank (UKB) and the International Consortium of Blood Pressure (ICBP) with a total sample size of 757,601 individuals. Genetic instruments for CAD were obtained from the CARDIoGRAMplusC4D consortium (60,801 cases and 123,504 controls from 48 studies). Several MR regression methods such as MR-Egger regression and inverse-variance weighted regression as well as sensitivity analyses were applied. A multivariable MR model served to determine which of the exposures had an independent causal effect on each of the stroke subtypes. Results Atrial fibrillation turned out to be the sole independent causal risk factor for cardioembolic stroke with a beta coefficient of 0.75 (SE = ± 0.04; P = 8.59 x 10e-100) and an OR of 2.16 (1.97-2.27) in a multivariable model adjusted for AF, CAD, systolic blood pressure (SBP) and diastolic blood pressure. In this model, SBP and CAD were non-causative for cardioembolic stroke, but exhibited significant, albeit weak causal associations with small vessel stroke (SBP: OR 1.04 (1.02-1.06); CAD: OR 1.14 (1.02-1.28)) and large artery stroke (SBP: OR 1.09 (1.06-1.11); CAD: OR 1.59 (1.40-1.79). Conclusions Causality of the association between the cardioembolic stroke subtype and AF is independent from modulators such as SBP and CAD. However, SBP and CAD play a complementary role as causal exposures for small vessel and large artery stroke which in turn are not causally linked with AF.

Previous observational studies have indicated that omega-3 may reduce the risk of various cancers. However, the relationship between omega-3 and the incidence of multiple myeloma (MM) remains unclear. Therefore, we conducted a systematic Mendelian randomization (MR) analysis to investigate the causal relationship between omega-3 and the risk of developing MM, while also exploring the potential mediating role of plasma lipids in this association. First, we conducted a two-sample MR study with MM using the omega-3 GWAS data from Richardson TG. We then repeated the validation with the other three omega-3 GWAS data and performed a meta-analysis of the MR results for a total of four omega-3 data. In the second step, we used multivariate Mendelian randomization (MVMR) analyses to adjust for the effects of confounders and explore the direct causal effects of omega-3 with MM. In the third step, we employed a two-step MR to investigate the potential mediating roles of 179 plasma lipids in the association between omega-3 and the risk of MM. Multiple sensitivity analyses were used to assess the robustness of the results. A two-sample MR analysis found that omega-3 can reduce the risk of MM (OR = 0.80, 95% CI 0.69–0.94; P = 0.005). In subsequent validation, omega-3 data from both Kettunen J and Davyson E yielded similar results. However, data from Zhang S indicated that omega-3 was not associated with MM risk. Ultimately, the meta-analysis results demonstrated that omega-3 can lower the risk of MM (OR = 0.80, 95% CI 0.72–0.88; P < 0.001). Furthermore, MVMR analysis, after adjusting for relevant risk factors such as obesity and type 2 diabetes, confirmed that omega-3 still reduces the risk of MM. Finally, two-step MR identified phosphatidylcholine (18:2_20:4) as a potential mediator of the causal relationship between omega-3 and MM. Various sensitivity analyses validated the robustness of these findings. Our study suggests that omega-3 may reduce the incidence risk of MM by increasing the levels of phosphatidylcholine (18:2_20:4). We hope that these findings will provide new insights for the prevention and treatment of MM.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-12804-y.