Exploring and mitigating potential bias when genetic instrumental variables are associated with multiple non-exposure traits in Mendelian randomization

AbstractBackgroundOur aim is to produce guidance on exploring and mitigating possible bias when genetic instrumental variables (IVs) associate with traits other than the exposure of interest in Mendelian randomization (MR) studies.MethodsWe use causal diagrams to illustrate scenarios that could result in IVs being related to (non-exposure) traits. We recommend that MR studies explore possible IV-non-exposure associations across a much wider range of traits than is usually the case. Where associations are found, confounding by population stratification should be assessed through adjusting for relevant population structure variables. To distinguish vertical from horizontal pleiotropy we suggest using bidirectional MR between the exposure and non-exposure traits and MR of the effect of the non-exposure traits on the outcome of interest. If vertical pleiotropy is plausible, standard MR methods should be unbiased. If horizontal pleiotropy is plausible, we recommend using multivariable MR to control for observed pleiotropic traits and conducting sensitivity analyses which do not require prior knowledge of specific invalid IVs or pleiotropic paths.ResultsWe applied our recommendations to an illustrative example of the effect of maternal insomnia on offspring birthweight in the UK Biobank. We found little evidence that unexpected IV-non-exposure associations were driven by population stratification. Three out of six observed non-exposure traits plausibly reflected horizontal pleiotropy. Multivariable MR and sensitivity analyses suggested an inverse association of insomnia with birthweight, but effects were imprecisely estimated in some of these analyses.ConclusionsWe provide guidance for MR studies where genetic IVs associate with non-exposure traits.Key messagesGenetic variants are increasingly found to associate with more than one social, behavioural or biological trait at genome-wide significance, which is a challenge in Mendelian randomization (MR) studies.Four broad scenarios (i.e. population stratification, vertical pleiotropy, horizontal pleiotropy and reverse causality) could result in an IV-non-exposure trait association.Population stratification can be assessed through adjusting for population structure with individual data, while two-sample MR studies should check whether the original genome-wide association studies have used robust methods to properly account for it.We apply currently available MR methods for discriminating between vertical and horizontal pleiotropy and mitigating against horizontal pleiotropy to an example exploring the effect of maternal insomnia on offspring birthweight.Our study highlights the pros and cons of relying more on sensitivity analyses without considering particular pleiotropic paths versus systematically exploring and controlling for potential pleiotropic paths via known characteristics.

Download Full-text

Evaluation of glycemic traits in susceptibility to COVID-19 risk: a Mendelian randomization study

BMC Medicine ◽

10.1186/s12916-021-01944-3 ◽

2021 ◽

Vol 19 (1) ◽

Author(s):

Shiu Lun Au Yeung ◽

Jie V Zhao ◽

C Mary Schooling

Keyword(s):

Type 2 Diabetes ◽

Genetic Predisposition ◽

Mendelian Randomization ◽

Fasting Glucose ◽

Association Studies ◽

Sensitivity Analyses ◽

Genome Wide Association Studies ◽

Inverse Association ◽

Wide Confidence Interval

Abstract Background Observational studies suggest poorer glycemic traits and type 2 diabetes associated with coronavirus disease 2019 (COVID-19) risk although these findings could be confounded by socioeconomic position. We conducted a two-sample Mendelian randomization to clarify their role in COVID-19 risk and specific COVID-19 phenotypes (hospitalized and severe cases). Method We identified genetic instruments for fasting glucose (n = 133,010), 2 h glucose (n = 42,854), glycated hemoglobin (n = 123,665), and type 2 diabetes (74,124 cases and 824,006 controls) from genome wide association studies and applied them to COVID-19 Host Genetics Initiative summary statistics (17,965 COVID-19 cases and 1,370,547 population controls). We used inverse variance weighting to obtain the causal estimates of glycemic traits and genetic predisposition to type 2 diabetes in COVID-19 risk. Sensitivity analyses included MR-Egger and weighted median method. Results We found genetic predisposition to type 2 diabetes was not associated with any COVID-19 phenotype (OR: 1.00 per unit increase in log odds of having diabetes, 95%CI 0.97 to 1.04 for overall COVID-19; OR: 1.02, 95%CI 0.95 to 1.09 for hospitalized COVID-19; and OR: 1.00, 95%CI 0.93 to 1.08 for severe COVID-19). There were no strong evidence for an association of glycemic traits in COVID-19 phenotypes, apart from a potential inverse association for fasting glucose albeit with wide confidence interval. Conclusion We provide some genetic evidence that poorer glycemic traits and predisposition to type 2 diabetes unlikely increase the risk of COVID-19. Although our study did not indicate glycemic traits increase severity of COVID-19, additional studies are needed to verify our findings.

Download Full-text

Causal association of adipokines with osteoarthritis: a Mendelian randomization study

Rheumatology ◽

10.1093/rheumatology/keaa719 ◽

2020 ◽

Author(s):

Jiayao Fan ◽

Jiahao Zhu ◽

Lingling Sun ◽

Yasong Li ◽

Tianle Wang ◽

...

Keyword(s):

Mendelian Randomization ◽

Association Studies ◽

Genome Wide Association ◽

Sensitivity Analyses ◽

European Ancestry ◽

Genome Wide Association Studies ◽

Knee Oa ◽

Genome Wide ◽

A Genome ◽

Weighted Median

Abstract Objective This two-sample Mendelian randomization study aimed to delve into the effects of genetically predicted adipokine levels on OA. Methods Summary statistic data for OA originated from a meta-analysis of a genome-wide association study with an overall 50 508 subjects of European ancestry. Publicly available summary data from four genome-wide association studies were exploited to respectively identify instrumental variables of adiponectin, leptin, resistin, chemerin and retinol-blinding protein 4. Subsequently, Mendelian randomization analyses were conducted with inverse variance weighted (IVW), weighted median and Mendelian randomization-Egger regression. Furthermore, sensitivity analyses were then conducted to assess the robustness of our results. Results The positive causality between genetically predicted leptin level and risk of total OA was indicated by IVW [odds ratio (OR): 2.40, 95% CI: 1.13–5.09] and weighted median (OR: 2.94, 95% CI: 1.23–6.99). In subgroup analyses, evidence of potential harmful effects of higher level of adiponectin (OR: 1.28, 95% CI: 1.01–1.61 using IVW), leptin (OR: 3.44, 95% CI: 1.18–10.03 using IVW) and resistin (OR: 1.18, 95% CI: 1.03–1.36 using IVW) on risk of knee OA were acquired. However, the mentioned effects on risk of hip OA were not statistically significant. Slight evidence was identified supporting causality of chemerin and retinol-blinding protein 4 for OA. The findings of this study were verified by the results from sensitivity analysis. Conclusions An association between genetically predicted leptin level and risk of total OA was identified. Furthermore, association of genetically predicted levels of adiponectin, leptin and resistin with risk of knee OA were reported.

Download Full-text

Bias in two-sample Mendelian randomization when using heritable covariable-adjusted summary associations

10.1101/816363 ◽

2019 ◽

Cited By ~ 5

Author(s):

Fernando Pires Hartwig ◽

Kate Tilling ◽

George Davey Smith ◽

Deborah A Lawlor ◽

Maria Carolina Borges

Keyword(s):

Waist Circumference ◽

Simulation Study ◽

Genetic Variants ◽

Mendelian Randomization ◽

Real Data ◽

Effect Estimate ◽

List Type ◽

Genome Wide Association Studies ◽

Genome Wide ◽

Residual Confounding

AbstractBackgroundTwo-sample Mendelian randomization (MR) allows the use of freely accessible summary association results from genome-wide association studies (GWAS) to estimate causal effects of modifiable exposures on outcomes. Some GWAS adjust for heritable covariables in an attempt to estimate direct effects of genetic variants on the trait of interest. One, both or neither of the exposure GWAS and outcome GWAS may have been adjusted for covariables.MethodsWe performed a simulation study comprising different scenarios that could motivate covariable adjustment in a GWAS and analysed real data to assess the influence of using covariable-adjusted summary association results in two-sample MR.ResultsIn the absence of residual confounding between exposure and covariable, between exposure and outcome, and between covariable and outcome, using covariable-adjusted summary associations for two-sample MR eliminated bias due to horizontal pleiotropy. However, covariable adjustment led to bias in the presence of residual confounding (especially between the covariable and the outcome), even in the absence of horizontal pleiotropy (when the genetic variants would be valid instruments without covariable adjustment). In an analysis using real data from the Genetic Investigation of ANthropometric Traits (GIANT) consortium and UK Biobank, the causal effect estimate of waist circumference on blood pressure changed direction upon adjustment of waist circumference for body mass index.ConclusionsOur findings indicate that using covariable-adjusted summary associations in MR should generally be avoided. When that is not possible, careful consideration of the causal relationships underlying the data (including potentially unmeasured confounders) is required to direct sensitivity analyses and interpret results with appropriate caution.Key messagesSummary genetic associations from large genome-wide associations studies (GWAS) have been increasingly used in two-sample Mendelian randomization (MR) analyses.Many GWAS adjust for heritable covariates in an attempt to estimate direct genetic effects on the trait of interest.In an extensive simulation study, we demonstrate that using covariable-adjusted summary associations may bias MR analyses.The bias largely depends on the underlying causal structure, specially the presence of unmeasured common causes between the covariable and the outcome.Our findings indicate that using covariable-adjusted summary associations in MR should generally be avoided.

Download Full-text

Adipokines and risk of rheumatoid arthritis: a Mendelian randomization study

10.21203/rs.3.rs-36474/v1 ◽

2020 ◽

Author(s):

Jiahao Zhu ◽

Haiyan Zheng ◽

Yasong Li ◽

Tianle Wang ◽

Yaohong Zhong ◽

...

Keyword(s):

Rheumatoid Arthritis ◽

Large Scale ◽

Mendelian Randomization ◽

Association Studies ◽

Statistical Significance ◽

Sensitivity Analyses ◽

European Ancestry ◽

Genome Wide Association Studies ◽

Primary Analysis ◽

Genome Wide

Abstract Background: Circulating adipokines levels have been reported to be associated with the risk of rheumatoid arthritis (RA). However, it is still unclear whether these associations are causal or biased by reverse causation or residual confounding. This study aimed to assess potential causal roles of five adipokines (namely, adiponectin, leptin, resistin, chemerin, and retinol-blinding protein 4 [RBP4]) in the occurrence of RA.Methods: We conducted a two-sample Mendelian randomization analysis to investigate these associations. We used summary-level data from genome-wide association studies (GWASs) for adipokines in individuals of European ancestry as the exposure, and a separate large-scale meta-analysis of a GWAS which included 14,361 RA cases and 43,923 controls of European ancestry as the outcome. Genetic variants were selected as instrumental variables if robustly genome-wide significant in their associations with adipokines. The causal effects were estimated using the inverse-variance weighted method in the primary analysis. Sensitivity analyses were performed to warrant that bias due to genetic pleiotropy was unlikely.Results: The circulating resistin was found to be the only adipokinetic factor having statistical significance, with higher levels causally associated with the risk of RA (odds ratio: 1.28; 95% confidence interval: [1.07, 1.53] per unit increase in the natural log-transformed resistin). In contrast, associations of adiponectin, leptin, chemerin, and RBP4 with risk of RA were not statistically significant. The MR assumptions did not seem to be violated. Sensitivity analyses yielded consistent findings.Conclusions: Genetically predicted circulating resistin levels were positively associated with RA risk. Our analysis suggested that resistin may play a notable causal role in RA pathogenesis. It would be beneficial for the development of clinical as well as public health strategies that target appropriate levels of resistin for future RA intervention.

Download Full-text

The use of negative control outcomes in Mendelian randomization to detect potential population stratification

International Journal of Epidemiology ◽

10.1093/ije/dyaa288 ◽

2021 ◽

Author(s):

Eleanor Sanderson ◽

Tom G Richardson ◽

Gibran Hemani ◽

George Davey Smith

Keyword(s):

Genetic Variants ◽

Population Stratification ◽

Mendelian Randomization ◽

Association Studies ◽

Epidemiological Studies ◽

Negative Control ◽

Mr Studies ◽

Genome Wide Association Studies ◽

Potential Population ◽

Mr Study

Abstract A key assumption of Mendelian randomization (MR) analysis is that there is no association between the genetic variants used as instruments and the outcome other than through the exposure of interest. One way in which this assumption can be violated is through population stratification, which can introduce confounding of the relationship between the genetic variants and the outcome and so induce an association between them. Negative control outcomes are increasingly used to detect unobserved confounding in observational epidemiological studies. Here we consider the use of negative control outcomes in MR studies to detect confounding of the genetic variants and the exposure or outcome. As a negative control outcome in an MR study, we propose the use of phenotypes which are determined before the exposure and outcome but which are likely to be subject to the same confounding as the exposure or outcome of interest. We illustrate our method with a two-sample MR analysis of a preselected set of exposures on self-reported tanning ability and hair colour. Our results show that, of the 33 exposures considered, genome-wide association studies (GWAS) of adiposity and education-related traits are likely to be subject to population stratification that is not controlled for through adjustment, and so any MR study including these traits may be subject to bias that cannot be identified through standard pleiotropy robust methods. Negative control outcomes should therefore be used regularly in MR studies to detect potential population stratification in the data used.

Download Full-text

Testing Causal Bidirectional Influences between Physical Activity and Depression using Mendelian Randomization

10.1101/364232 ◽

2018 ◽

Cited By ~ 1

Author(s):

Karmel W. Choi ◽

Chia-Yen Chen ◽

Murray B. Stein ◽

Yann C. Klimentidis ◽

Min-Jung Wang ◽

...

Keyword(s):

Physical Activity ◽

Protective Factor ◽

Mendelian Randomization ◽

Significant Snps ◽

Sensitivity Analyses ◽

Inverse Association ◽

Effective Prevention ◽

Genome Wide ◽

Average Acceleration ◽

Genetic Instruments

AbstractBackground:Burgeoning evidence from randomized controlled trials and prospective cohort studies suggests that physical activity protects against depression, pointing to a potential modifiable target for prevention. However, the direction of this inverse association is not clear: physical activity may reduce risk for depression, and/or depression may result in decreased physical activity. Here, we used bidirectional two-sample Mendelian randomization (MR) to test causal influences between physical activity and depression.Methods:For genetic instruments, we selected independent top SNPs associated with major depressive disorder (MDD, N = 143,265) and two physical activity phenotypes—self-reported (N = 377,234) and objective accelerometer-based (N = 91,084)—from the largest available, non-overlapping genome-wide association results. We used two sets of genetic instruments: (1) only SNPs previously reported as genome-wide significant, and (2) top SNPs meeting a more relaxed threshold (p < 1×10-7). For each direction of influence, we combined the MR effect estimates from each instrument SNP using inverse variance weighted (IVW) meta-analysis, along with other standard MR methods such as weighted median, MR-Egger, and MR-PRESSO.Results:We found evidence for protective influences of accelerometer-based activity on MDD (IVW odds ratio (OR) = 0.74 for MDD per 1 SD unit increase in average acceleration, 95% confidence interval (CI) = 0.59-0.92, p =.006) when using SNPs meeting the relaxed threshold (i.e., 10 versus only 2 genome-wide significant SNPs, which provided insufficient data for sensitivity analyses). In contrast, we found no evidence for negative influences of MDD on accelerometer-based activity (IVW b = 0.04 change in average acceleration for MDD versus control status, 95% CI = −0.43-0.51, p =.87). Furthermore, we did not see evidence for causal influences between self-reported activity and MDD, in either direction and regardless of instrument SNP criteria.Discussion:We apply MR for the first time to examine causal influences between physical activity and MDD. We discover that objectively measured—but not self-reported—physical activity is inversely associated with MDD. Of note, prior work has shown that accelerometer-based physical activity is more heritable than self-reported activity, in addition to being more representative of actual movement. Our findings validate physical activity as a protective factor for MDD and point to the importance of objective measurement of physical activity in epidemiological studies in relation to mental health. Overall, this study supports the hypothesis that enhancing physical activity is an effective prevention strategy for depression.

Download Full-text

Mendelian Randomization Study on Amino Acid Metabolism Suggests Tyrosine as Causal Trait for Type 2 Diabetes

Nutrients ◽

10.3390/nu12123890 ◽

2020 ◽

Vol 12 (12) ◽

pp. 3890

Author(s):

Susanne Jäger ◽

Rafael Cuadrat ◽

Clemens Wittenbecher ◽

Anna Floegel ◽

Per Hoffmann ◽

...

Keyword(s):

Type 2 Diabetes ◽

Amino Acids ◽

Amino Acid ◽

Mendelian Randomization ◽

Association Studies ◽

Genome Wide Association ◽

Genome Wide Association Studies ◽

Inverse Association ◽

Genome Wide

Circulating levels of branched-chain amino acids, glycine, or aromatic amino acids have been associated with risk of type 2 diabetes. However, whether those associations reflect causal relationships or are rather driven by early processes of disease development is unclear. We selected diabetes-related amino acid ratios based on metabolic network structures and investigated causal effects of these ratios and single amino acids on the risk of type 2 diabetes in two-sample Mendelian randomization studies. Selection of genetic instruments for amino acid traits relied on genome-wide association studies in a representative sub-cohort (up to 2265 participants) of the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam Study and public data from genome-wide association studies on single amino acids. For the selected instruments, outcome associations were drawn from the DIAGRAM (DIAbetes Genetics Replication And Meta-analysis, 74,124 cases and 824,006 controls) consortium. Mendelian randomization results indicate an inverse association for a per standard deviation increase in ln-transformed tyrosine/methionine ratio with type 2 diabetes (OR = 0.87 (0.81–0.93)). Multivariable Mendelian randomization revealed inverse association for higher log10-transformed tyrosine levels with type 2 diabetes (OR = 0.19 (0.04–0.88)), independent of other amino acids. Tyrosine might be a causal trait for type 2 diabetes independent of other diabetes-associated amino acids.

Download Full-text

Causal Relationship Between Parathyroid Hormone and the Risk of Osteoarthritis: A Mendelian Randomization Study

Frontiers in Genetics ◽

10.3389/fgene.2021.686939 ◽

2021 ◽

Vol 12 ◽

Author(s):

Guiwu Huang ◽

Yanlin Zhong ◽

Wenchang Li ◽

Weiming Liao ◽

Peihui Wu

Keyword(s):

Parathyroid Hormone ◽

Confidence Interval ◽

Odds Ratio ◽

Mendelian Randomization ◽

Causal Effect ◽

Sensitivity Analyses ◽

Inverse Association ◽

Genome Wide ◽

Inverse Variance ◽

Weighted Median

BackgroundPrevious studies have demonstrated an inverse association between parathyroid hormone (PTH) and the risk of osteoarthritis (OA). However, it remains unknown whether such association reflects causality. We aimed to apply a Mendelian randomization (MR) approach to investigate the causal association between PTH and OA.Materials and MethodsWe performed a two-sample MR analysis using summary statistics from 13 cohorts (PTH, N = 29,155) and a recent genome-wide association study meta-analysis (OA, N = 455,221) by the UK Biobank and Arthritis Research UK OA Genetics (arcOGEN). MR analyses were carried out mainly using the inverse-variance-weighted method. Sensitivity analyses were performed to test the robustness of the associations using the weighted median method, the MR–Egger method, and “leave-one-out” analysis. Analyses were performed again to test whether the associations remained statistically significant after excluding any outlier variants that were detected using the MR-PRESSO (Mendelian Randomization Pleiotropy RESidual Sum and Outlier) test.ResultsFive single-nucleotide polymorphisms (SNPs) were selected as instrumental variables at the genome-wide significance threshold (p < 5 × 10–8). The causal effect between PTH and OA was genetically predicted using the inverse-variance-weighted method (odds ratio = 0.67, 95% confidence interval: 0.50–0.90; p = 0.008). This result was borne out using the weighted median method (odds ratio = 0.73, 95% confidence interval: 0.60–0.90; p = 0.004). The causality remained robust after discarding the outlier variants as well as SNPs associated with confounding factors.ConclusionMR analysis supported a potential causative relationship between decreased serum circulating PTH and a higher risk of hip and knee OA.

Download Full-text

Causal Relationship between Educational Attainment and the Risk of Rheumatoid Arthritis: A Mendelian Randomization Study

10.21203/rs.3.rs-134634/v1 ◽

2021 ◽

Author(s):

Guiwu Huang ◽

Jiahao Cai ◽

Wenchang Li ◽

Yanlin Zhong ◽

Weiming Liao ◽

...

Keyword(s):

Rheumatoid Arthritis ◽

Educational Attainment ◽

Instrumental Variables ◽

Mendelian Randomization ◽

Meta Analysis ◽

Sensitivity Analyses ◽

European Ancestry ◽

Genome Wide Association Studies ◽

Weighted Median ◽

Leave One Out

Abstract Background Educational attainment is moderately heritable and positively associated with the risk of rheumatoid arthritis. However, the causality from educational attainment on rheumatoid arthritis remained unknown. Here, we aimed to determine whether educational attainment is causally associated with rheumatoid arthritis (RA) by using a Mendelian randomization (MR) approach.Methods Summary statistics data for RA were obtained from an available, published meta-analysis of genome-wide association studies (GWAS) that included 14,361 RA cases and 43,923 controls of European ancestry. The instrumental variables for educational attainment were obtained from a GWAS meta-analysis that included over 1 million individuals (N = 1,131,881) of European ancestry. MR analyses were performed using the inverse-variance weighted (IVW), weighted median, and MR-Egger methods. Sensitivity analyses were performed to test the robustness of the association using the Cochran Q test, MR Egger intercept test, “leave-one-out” analysis and MR-PRESSO test.Results A total of 387 SNPs were employed as instrumental variables in our MR analysis. Genetically predicted higher educational attainment was associated with a significantly lower risk of RA using the IVW method (odds ratio [OR] = 0.42, 95% confidence interval [CI]: 0.34–0.52; p = 1.78×10−14). The weighted median and MR Egger methods yielded consistent results. The causality remained robust after discarding the outlier variants and SNPs associated with the confounding factors. "Leave-one-out" analysis confirmed the stability of our results. Additionally, the results demonstrated the absence of the horizontal pleiotropy.Conclusions The MR analysis supported a potential inverse causative relationship between educational attainment and the risk of RA.

Download Full-text

Genetic Contribution of Endometriosis to the Risk of Developing Hormone-Related Cancers

International Journal of Molecular Sciences ◽

10.3390/ijms22116083 ◽

2021 ◽

Vol 22 (11) ◽

pp. 6083

Author(s):

Aintzane Rueda-Martínez ◽

Aiara Garitazelaia ◽

Ariadna Cilleros-Portet ◽

Sergi Marí ◽

Rebeca Arauzo ◽

...

Keyword(s):

Mendelian Randomization ◽

Clear Cell ◽

Association Studies ◽

Epidemiological Data ◽

Epidemiological Studies ◽

Genome Wide Association Studies ◽

Genome Wide ◽

Genomic Analyses ◽

Gynecological Disorder ◽

Robust Evidence

Endometriosis is a common gynecological disorder that has been associated with endometrial, breast and epithelial ovarian cancers in epidemiological studies. Since complex diseases are a result of multiple environmental and genetic factors, we hypothesized that the biological mechanism underlying their comorbidity might be explained, at least in part, by shared genetics. To assess their potential genetic relationship, we performed a two-sample mendelian randomization (2SMR) analysis on results from public genome-wide association studies (GWAS). This analysis confirmed previously reported genetic pleiotropy between endometriosis and endometrial cancer. We present robust evidence supporting a causal genetic association between endometriosis and ovarian cancer, particularly with the clear cell and endometrioid subtypes. Our study also identified genetic variants that could explain those associations, opening the door to further functional experiments. Overall, this work demonstrates the value of genomic analyses to support epidemiological data, and to identify targets of relevance in multiple disorders.

Download Full-text