Mendelian Randomization versus Path Models: Making Causal Inferences in Genetic Epidemiology

Causal association between plasma metabolites and neurodegenerative diseases

BACKGROUND

Establishing causal relationships between metabolic biomarkers and neurodegenerative diseases such as Alzheimer's disease (AD) and Parkinson's disease (PD) is a challenge faced by observational studies. In this study, our aim was to investigate the causal associations between plasma metabolites and neurodegenerative diseases using Mendelian Randomization (MR) methods.

METHODS

We utilized genetic associations with 1400 plasma metabolic traits as exposures. We used large-scale genome-wide association study (GWAS) summary statistics for AD and PD as our discovery datasets. For validation, we performed repeated analyses using different GWAS datasets. The main statistical method employed was inverse variance-weighted (IVW). We also conducted enrichment pathway analysis for IVW-identified metabolites.

RESULTS

In the discovered dataset, there are a total of 69 metabolites (36 negatively, 33 positively) potentially associated with AD, and 47 metabolites (24 negatively， 23 positively) potentially associated with PD. Among these, 4 significant metabolites overlap with significant metabolites (PIVW < 0.05)in the validation dataset for AD, and 1 metabolite overlaps with significant metabolites in the validation dataset for PD. Three metabolites serve as common potential metabolic markers for both AD and PD, including Tryptophan betaine, Palmitoleoylcarnitine (C16:1), and X-23655 levels. Further pathway enrichment analysis suggests that the SLC-mediated transmembrane transport pathway, involving tryptophan betaine and carnitine metabolites, may represent potential intervention targets for treating AD and PD.

CONCLUSION

This study offers novel insights into the causal effects of plasma metabolites on degenerative diseases through the integration of genomics and metabolomics. The identification of metabolites and metabolic pathways linked to AD and PD enhances our comprehension of the underlying biological mechanisms and presents promising targets for future therapeutic interventions in AD and PD.

Mendelian randomization analysis of 34,497 German Holstein cows to infer causal associations between milk production and health traits

BACKGROUND

Claw diseases and mastitis represent the most important health issues in dairy cattle with a frequently mentioned connection to milk production. Although many studies have aimed at investigating this connection in more detail by estimating genetic correlations, they do not provide information about causality. An alternative is to carry out Mendelian randomization (MR) studies using genetic variants to investigate the effect of an exposure on an outcome trait mediated by genetic variants. No study has yet investigated the causal association of milk yield (MY) with health traits in dairy cattle. Hence, we performed a MR analysis of MY and seven health traits using imputed whole-genome sequence data from 34,497 German Holstein cows. We applied a method that uses summary statistics and removes horizontal pleiotropic variants (having an effect on both traits), which improves the power and unbiasedness of MR studies. In addition, genetic correlations between MY and each health trait were estimated to compare them with the estimates of causal effects that we expected.

RESULTS

All genetic correlations between MY and each health trait were negative, ranging from - 0.303 (mastitis) to - 0.019 (digital dermatitis), which indicates a reduced health status as MY increases. The only non-significant correlation was between MY and digital dermatitis. In addition, each causal association was negative, ranging from - 0.131 (mastitis) to - 0.034 (laminitis), but the number of significant associations was reduced to five nominal and two experiment-wide significant results. The latter were between MY and mastitis and between MY and digital phlegmon. Horizontal pleiotropic variants were identified for mastitis, digital dermatitis and digital phlegmon. They were located within or nearby variants that were previously reported to have a horizontal pleiotropic effect, e.g., on milk production and somatic cell count.

CONCLUSIONS

Our results confirm the known negative genetic connection between health traits and MY in dairy cattle. In addition, they provide new information about causality, which for example points to the negative energy balance mediating the connection between these traits. This knowledge helps to better understand whether the negative genetic correlation is based on pleiotropy, linkage between causal variants for both trait complexes, or indeed on a causal association.

Causal association of gastroesophageal reflux disease on irritable bowel syndrome: a two-sample Mendelian randomization study

Background

Recently, observational studies have reported that gastroesophageal reflux disease (GERD) is commonly associated with irritable bowel syndrome (IBS), but the causal relationship is unclear.

Methods

We conducted a two-sample Mendelian randomization study using summary data from genome-wide association studies (GWASs) to explore a causal relationship between GERD (N cases = 129,080) and IBS (N cases = 4,605) of European ancestry. Furthermore, the inverse-variance weighted (IVW) method and a series of sensitivity analyses were used to assess the accuracy and confidence of our results.

Results

We found a significant association of GERD with IBS (NSNP = 74; OR: 1.375; 95% CI: 1.164-1.624; p < 0.001). Reverse MR analysis showed no evidence of a causal association for IBS with GERD (NSNP = 6; OR: 0.996; 95% CI: 0.960-1.034; p = 0.845).

Conclusion

This study provides evidence that the presence of GERD increases the risk of developing IBS, and it is observed from the reverse MR results that IBS did not increase the risk of GERD.

Depression and hepatobiliary diseases: a bidirectional Mendelian randomization study

Background

More and more evidence suggests a close association between depression and hepatobiliary diseases, but its causal relationship is not yet clear.

Method

Using genome-wide association studies (GWAS) to summarize data, independent genetic variations associated with depression were selected as instrumental variables. Firstly, we designed a univariate Mendelian randomization (UVMR) analysis with two samples and simultaneously conducted reverse validation to evaluate the potential bidirectional causal relationship between depression and various hepatobiliary diseases. Secondly, we conducted a multivariate Mendelian randomization (MVMR) analysis on diseases closely related to depression, exploring the mediating effects of waist to hip ratio, hypertension, and daytime nap. The mediating effects were obtained through MVMR. For UVMR and MVMR, inverse variance weighted method (IVW) is considered the most important analytical method. Sensitivity analysis was conducted using Cochran'Q, MR Egger, and Leave-one-out methods.

Results

UVMR analysis showed that depression may increase the risk of non-alcoholic fatty liver disease (OR, 1.22; 95% CI, 1.03-1.46; p=0.0248) in liver diseases, while depression does not increase the risk of other liver diseases; In biliary and pancreatic related diseases, depression may increase the risk of cholelithiasis (OR, 1.26; 95% CI, 1.05-1.50; p=0.0120), chronic pancreatitis (OR, 1.61; 95% CI, 1.10-2.35; p=0.0140), and cholecystitis (OR, 1.23; 95% CI, 1.03-1.48; p=0.0250). In addition, through reverse validation, we found that non-alcoholic fatty liver disease, cholelithiasis, chronic pancreatitis, cholecystitis, or the inability to increase the risk of depression (p>0.05). The waist to hip ratio, hypertension, and daytime nap play a certain role in the process of depression leading to non-alcoholic fatty liver disease, with a mediating effect of 35.8%.

Conclusion

Depression is a susceptibility factor for non-alcoholic fatty liver disease, and the causal effect of genetic susceptibility to depression on non-alcoholic fatty liver disease is mediated by waist-hip ratio, hypertension, and daytime nap.

Inferring Causal Relationships Between Metabolites and Polycystic Ovary Syndrome Using Summary Statistics from Genome‑Wide Association Studies

Polycystic ovary syndrome (PCOS) is a disorder characterized by hyperandrogenism, ovulatory dysfunction, and polycystic ovarian morphology. Previous studies have suggested that metabolites may play a pivotal mediating role in the progression of phenotypic variations. Although several metabolites had been identified as potential markers for PCOS, the relationship between blood metabolites and PCOS was not comprehensively explored. Previously, Pickrell et al. designed a robust approach to infer evidence of a causal relationship between different phenotypes using independently putative causal SNPs. Our previous paper extended this approach to make it more suitable for cases where only a few independently putative causal SNPs were identified to be significantly associated with the phenotypes (i.e., metabolites). When the most significant SNPs in each independent locus (the independent lead SNPs) with p-values of < 1 × 10-5 were used, 3 metabolites (2-tetradecenoyl carnitine, threitol, 1-docosahexaenoylglycerophosphocholine) causally influencing PCOS and 2 metabolites (asparagine and phenyllactate) influenced by PCOS were identified, (relative likelihood r < 0.01). Under a less stringent threshold of r < 0.05, 7 metabolites (trans-4-hydroxyproline, glutaroyl carnitine, stachydrine, undecanoate, 7-Hoca, N-acetylalanine and 2-hydroxyisobutyrate) were identified. Taken together, this study can provide novel insights into the pathophysiological mechanisms underlying PCOS; whether these metabolites can serve as biomarkers to predict PCOS in clinical practice warrants further investigations.

Connecting atrial fibrillation to digestive neoplasms: exploring mediation via ischemic stroke and heart failure in Mendelian randomization studies

Background

Notwithstanding the acknowledged interplay between atrial fibrillation (AF) and the emergence of digestive system neoplasms, the intricacies of this relationship remain ambiguous. By capitalizing univariable Mendelian Randomization (MR) complemented by a mediated MR tactic, our pursuit was to elucidate the causative roles of AF in precipitating digestive system malignancies and potential intermediary pathways.

Method

This research endeavor seeks to scrutinize the causal clinical implications of whether genetic predispositions to AF correlate with an increased risk of digestive system malignancies, employing MR analytical techniques. Utilizing a dataset amalgamated from six studies related to AF, encompassing over 1,000,000 subjects, we performed univariable MR assessments, employing the random-effects inverse-variance weighted (IVW) methodology as our principal analytical paradigm. Subsequently, a mediated MR framework was employed to probe the potential mediating influence of AF on the nexus between hypertension (HT), heart failure (HF), ischemic stroke (IS), coronary artery disease (CAD), and digestive system neoplasms.

Result

The univariable MR evaluation unveiled a notable causal nexus between the genetic inclination toward AF and the genetic susceptibility to colon, esophageal, and small intestine malignancies. The mediated MR scrutiny ascertained that the genetic inclination for AF amplifies the risk profile for colon cancer via IS pathways and partially explains the susceptibility to esophageal and small intestine tumors through the HF pathway.

Conclusion

Our investigative endeavor has highlighted a definitive causative association between genetic inclination to AF and specific digestive system neoplasms, spotlighting IS and HF as instrumental mediators. Such revelations furnish pivotal perspectives on the complex genetic interconnections between cardiovascular anomalies and certain digestive tract tumors, emphasizing prospective therapeutic and diagnostic worthy of pursuit.

Metabolic syndrome and risk of colorectal cancer: A Mendelian randomization study

Background

Observational studies have previously demonstrated a significant relationship among both metabolic syndrome (Mets) and colorectal cancer (CRC). Whether there is a causal link remains controversial.

Objective

To clarify whether Mets and their components have a causal effect on colorectal cancer, we have carried out a bidirectional Mendelian randomization analysis (MR).

Methods

This study started from genome-wide association data for Mets and its 5 components (hypertension, waist circumference, fasting blood glucose, serum triglycerides, and serum high-density lipoprotein cholesterol) and colorectal cancer. Mendelian randomization (MR) techniques were used in the study to examine their associations.

Results

After Benjamini-Hochberg multiple corrections, genetically predicted significant causal link exists between WC (waist circumference) and CRC. The OR was 1.35 (95 % CI: 1.08-1.69; p = 0.0096). Other Mets components (HBP, FBG, TG, HDL), on the other hand, found no evidence of a genetic link between CRC and Mets. In addition, MR results showed that CRC was not causally related to either Mets or the components. We get the same result in the validated dataset.

Conclusion

According to the bidirectional MR investigation shows a significant causal relationship among obesity and CRC in the Mets component but no causal relationship in the opposite direction.

The causality of physical activity status and intelligence: A bidirectional Mendelian randomization study

BACKGROUND

Observational studies suggest physical activity (PA) enhances intelligence, while sedentary behavior (SB) poses a risk. However, causality remains unclear.

METHODS

We extracted genetic instruments from large genome-wide association studies summary data and employed an inverse-variance weighted (IVW) approach within a random-effects model as the primary method of Mendelian randomization (MR) analysis to estimate the overall effect of various physical activity statuses on intelligence. To assess IVW stability and MR sensitivity, we also utilized supplementary methods including weighted median, MR-Egger, and MR-PRESSO. Furthermore, multivariable MR analysis was conducted to examine the independent effects of each physical activity trait on intelligence.

RESULTS

The MR primary results indicated that LST was negatively associated with intelligence (β = -0.133, 95%CI: -0.177 to -0.090, p = 1.34×10-9), while SBW (β = 0.261, 95% CI: 0.059 to 0.463, p = 0.011) may have a positive effect on intelligence; however, MVPA and SC did not show significant effects on intelligence. Inverse causality analyses demonstrated intelligence significantly influenced all physical activity states.

CONCLUSIONS

Our study highlights a bidirectional causal relationship between physical activity states and intelligence.

Association between COVID-19 and telomere length: A bidirectional Mendelian randomization study

Several traditional observational studies suggested an association between COVID-19 and leukocyte telomere length (LTL), a biomarker for biological age. However, whether there was a causal association between them remained unclear. We aimed to investigate whether genetically predicted COVID-19 is related to the risk of LTL, and vice versa. We performed bidirectional Mendelian randomization (MR) study using summary statistics from the genome-wide association studies of critically ill COVID-19 (n = 1 388 342) and LTL (n = 472 174) of European ancestry. The random-effects inverse-variance weighted estimation method was applied as the primary method with several other estimators as complementary methods. Using six single-nucleotide polymorphisms (SNPs) of genome-wide significance as instrumental variables for critically ill COVID-19, we did not find a significant association of COVID-19 on LTL (β = 0.0075, 95% confidence interval [CI]: -0.018 to 0.021, p = 0.733). Likewise, using 97 SNPs of genome-wide significance as instrumental variables for LTL, we did not find a significant association of LTL on COVID-19 (odds ratio = 1.00, 95% CI: 0.79-1.28, p = 0.973). Comparable results were obtained using MR-Egger regression, weighted median, and weighted mode approaches. We did not find evidence to support a causal association between COVID-19 and LTL in either direction.

The association between depression and metabolic syndrome and its components: a bidirectional two-sample Mendelian randomization study

Observational studies suggested a bidirectional correlation between depression and metabolic syndrome (MetS) and its components. However, the causal associations between them remained unclear. We aimed to investigate whether genetically predicted depression is related to the risk of MetS and its components, and vice versa. We performed a bidirectional two-sample Mendelian randomization (MR) study using summary-level data from the most comprehensive genome-wide association studies (GWAS) of depression (n = 2,113,907), MetS (n = 291,107), waist circumference (n = 462,166), hypertension (n = 463,010) fasting blood glucose (FBG, n = 281,416), triglycerides (n = 441,016), high-density lipoprotein cholesterol (HDL-C, n = 403,943). The random-effects inverse-variance weighted (IVW) method was applied as the primary method. The results identified that genetically predicted depression was significantly positive associated with risk of MetS (OR: 1.224, 95% CI: 1.091-1.374, p = 5.58 × 10^-4), waist circumference (OR: 1.083, 95% CI: 1.027-1.143, p = 0.003), hypertension (OR: 1.028, 95% CI: 1.016-1.039, p = 1.34 × 10^-6) and triglycerides (OR: 1.111, 95% CI: 1.060-1.163, p = 9.35 × 10^-6) while negative associated with HDL-C (OR: 0.932, 95% CI: 0.885-0.981, p = 0.007) but not FBG (OR: 1.010, 95% CI: 0.986-1.034, p = 1.34). No causal relationships were identified for MetS and its components on depression risk. The present MR analysis strength the evidence that depression is a risk factor for MetS and its components (waist circumference, hypertension, FBG, triglycerides, and HDL-C). Early diagnosis and prevention of depression are crucial in the management of MetS and its components.

Presidential address: Six open questions to genetic epidemiologists

Given the rapid pace with which genomics and other ‐omics disciplines are evolving, it is sometimes necessary to shift down a gear to consider more general scientific questions. In this line, in my presidential address I formulate six questions for genetic epidemiologists to ponder on. These cover the areas of reproducibility, statistical significance, chance findings, precision medicine and related fields such as bioinformatics and data science. Possible hints at responses are presented to foster our further discussion of these topics.

F, König IR. Evaluating the current state of Mendelian randomization studies: a protocol for a systematic review on methodological and clinical aspects using neurodegenerative disorders as outcome

BACKGROUND

Mendelian randomization (MR) is fast becoming a popular method to judge causality from routinely conducted observational studies. However, stringent underlying statistical assumptions, missing biological information, and high sample size requirement might make it prone to misuse. Furthermore, rapidly updating methodologies and increasingly available datasets to researchers are making the interpretations of heterogeneous results even more complicated. In this protocol, we provide our design for a multifaceted systematic review on MR studies using neurodegenerative disease as an example outcome. The planned systematic review which has already passed the pilot stage will help to develop an in-depth understanding of how various MR methods have been applied, what has been achieved, and what can be done in future for to arrive at true causal risk factors.

METHODS

During the pilot phase of this systematic review, several versions of questionnaires and frequent consultations between reviewers helped us to finalize a comprehensive list of questions. This will be used to extract information on systematically searched MR articles investigating causality underlying neurodegenerative diseases. A literature search of the electronic databases (Embase, MEDLINE, Web of Science, Scopus, and databases listed in the Cochrane library) will be conducted. The search strategy will include terms related to MR and the spectrum of neurodegenerative diseases. Two independent reviewers will screen the studies, and three will extract the data. The included studies will be further judged by two reviewers for accuracy and completeness of available information. We will perform descriptive and quantitative synthesis using sensitivity analyses of causal association by study design, selection of genetic instrument, validity of MR assumptions, MR method, and sensitivity analysis based on exclusion of potential pleiotropic variants. The quality of conduct as well as quality of reporting in the included studies will be assessed and reported. A meta-analysis will be conducted, if effect estimates on identical genetic instruments are available for both exposure and outcome in the studies using data from participants from ethnically similar populations.

DISCUSSION

This systematic review protocol utilizes a unique comprehensive data abstraction tool based on recent methodological advancements in the field of MR. The planned systematic review will further integrate information on methodological details with clinical findings in latest available large-scale genome-wide association study datasets. Our findings aim to help raising awareness and promoting transparent reporting of MR studies.

SYSTEMATIC REVIEW REGISTRATION

PROSPERO CRD42018091434 .

Inferring causal relationships between phenotypes using summary statistics from genome-wide association studies

Genome-wide association studies (GWAS) have successfully identified numerous genetic variants associated with diverse complex phenotypes and diseases, and provided tremendous opportunities for further analyses using summary association statistics. Recently, Pickrell et al. developed a robust method for causal inference using independent putative causal SNPs. However, this method may fail to infer the causal relationship between two phenotypes when only a limited number of independent putative causal SNPs identified. Here, we extended Pickrell's method to make it more applicable for the general situations. We extended the causal inference method by replacing the putative causal SNPs with the lead SNPs (the set of the most significant SNPs in each independent locus) and tested the performance of our extended method using both simulation and empirical data. Simulations suggested that when the same number of genetic variants is used, our extended method had similar distribution of test statistic under the null model as well as comparable power under the causal model compared with the original method by Pickrell et al. But in practice, our extended method would generally be more powerful because the number of independent lead SNPs was often larger than the number of independent putative causal SNPs. And including more SNPs, on the other hand, would not cause more false positives. By applying our extended method to summary statistics from GWAS for blood metabolites and femoral neck bone mineral density (FN-BMD), we successfully identified ten blood metabolites that may causally influence FN-BMD. We extended a causal inference method for inferring putative causal relationship between two phenotypes using summary statistics from GWAS, and identified a number of potential causal metabolites for FN-BMD, which may provide novel insights into the pathophysiological mechanisms underlying osteoporosis.

A comparison of methods for inferring causal relationships between genotype and phenotype using additional biological measurements

Genome wide association studies (GWAS) have been very successful over the last decade at identifying genetic variants associated with disease phenotypes. However, interpretation of the results obtained can be challenging. Incorporation of further relevant biological measurements (e.g. ‘omics’ data) measured in the same individuals for whom we have genotype and phenotype data may help us to learn more about the mechanism and pathways through which causal genetic variants affect disease. We review various methods for causal inference that can be used for assessing the relationships between genetic variables, other biological measures, and phenotypic outcome, and present a simulation study assessing the performance of the methods under different conditions. In general, the methods we considered did well at inferring the causal structure for data simulated under simple scenarios. However, the presence of an unknown and unmeasured common environmental effect could lead to spurious inferences, with the methods we considered displaying varying degrees of robustness to this confounder. The use of causal inference techniques to integrate omics and GWAS data has the potential to improve biological understanding of the pathways leading to disease. Our study demonstrates the suitability of various methods for performing causal inference under several biologically plausible scenarios.

Mendelian Randomization

Confounding and reverse causality have prevented us from drawing meaningful clinical interpretation even in well-powered observational studies. Confounding may be attributed to our inability to randomize the exposure variable in observational studies. Mendelian randomization (MR) is one approach to overcome confounding. It utilizes one or more genetic polymorphisms as a proxy for the exposure variable of interest. Polymorphisms are randomly distributed in a population, they are static throughout an individual's lifetime, and may thus help in inferring directionality in exposure-outcome associations. Genome-wide association studies (GWAS) or meta-analyses of GWAS are characterized by large sample sizes and the availability of many single nucleotide polymorphisms (SNPs), making GWAS-based MR an attractive approach. GWAS-based MR comes with specific challenges, including multiple causality. Despite shortcomings, it still remains one of the most powerful techniques for inferring causality.With MR still an evolving concept with complex statistical challenges, the literature is relatively scarce in terms of providing working examples incorporating real datasets. In this chapter, we provide a step-by-step guide for causal inference based on the principles of MR with a real dataset using both individual and summary data from unrelated individuals. We suggest best possible practices and give recommendations based on the current literature.