Prospects of Mendelian randomization in hepatology: a comprehensive literature review with practice guidance

Lanlan Chen; Qi Rao; Menghan Gao; Guoyue Lv; Frank Tacke

doi:10.3350/cmh.2025.0541

Articles

Page Path

Review

Prospects of Mendelian randomization in hepatology: a comprehensive literature review with practice guidance

Lanlan Chen^1,2,3

, Qi Rao^1,4, Menghan Gao⁵, Guoyue Lv^2,3

, Frank Tacke¹

Clinical and Molecular Hepatology 2025;31(4):1115-1138.
Published online: June 9, 2025

DOI: https://doi.org/10.3350/cmh.2025.0541

¹Department of Hepatology & Gastroenterology, Charité - Universitätsmedizin Berlin, Campus Virchow-Klinikum and Campus Charité Mitte, Berlin, Germany

²Department of Hepatobiliary and Pancreatic Surgery, General Surgery Center, First Hospital of Jilin University, Changchun, Jilin, China

³China-Singapore Belt and Road Joint Laboratory on Liver Disease Research, the First Hospital of Jilin University, Changchun, Jilin, China

⁴Department of Biology, Chemistry, Pharmacy, Free University of Berlin, Berlin, Germany

⁵Medical Publishing Center, The First Hospital of Jilin University, Jilin, China

Corresponding author : Lanlan Chen Department of Hepatology & Gastroenterology, Charité - Universitätsmedizin Berlin, Campus Virchow-Klinikum and Campus Charité Mitte, Augustenburger Platz 1, D-13353 Berlin, Germany Tel: +49-152-3643-6998, E-mail: lanlan.chen@charite.de

Guoyue Lv Department of Hepatobiliary and Pancreatic Surgery, General Surgery Center, First Hospital of Jilin University, No.1 Xinmin Street, 130021, Changchun, Jilin, China Tel: +86-136-8982-5113, E-mail: lvgy@jlu.edu.cn

Frank Tacke Department of Hepatology & Gastroenterology, Charité - Universitätsmedizin Berlin, Campus Virchow-Klinikum and Campus Charité Mitte, Augustenburger Platz 1, D-13353 Berlin, Germany Tel: +49-(30)-450-553-022, Fax: +49-(30)-450-553-902, E-mail: frank.tacke@charite.de

Editor: Won Kim, Seoul National University College of Medicine, Korea

• Received: May 16, 2025 • Revised: May 29, 2025 • Accepted: June 3, 2025

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

3,251 Views
205 Download
3 Crossref

prev next

Full Article

Download PDF

ABSTRACT
INTRODUCTION
CONCLUSION
FOOTNOTES
Abbreviations
REFERENCES

ABSTRACT

Mendelian randomization (MR), a powerful statistical tool for causal inference, has been widely applied in various fields of medical research, even extending to economics and psychology. In hepatology, MR has been utilized to identify risk factors and potential therapeutic targets for liver diseases, including metabolic dysfunction-associated steatotic liver disease, cholestatic and autoimmune liver diseases, and hepatobiliary cancer. MR can provide evidence of causation via associations between genetic variants, modifiable exposures and liver disease occurrence or outcomes, using large existing datasets. However, results from MR studies are sometimes scattered, biologically not plausible or even controversial between analyses, potentially reflecting a trend of inappropriate application of this method (e.g., inappropriate selection of genetic instruments, insufficient assessment of horizontal pleiotropy, compromised statistical power, and neglected genetic diversity among different populations), and thus hinder the translation of MR findings from bench to bedside. Assessing these critical issues and pinpointing bona fide evidence are essential but quite challenging for clinicians. In this review, we aim to introduce the MR method to hepatologists and provide a comprehensive overview of the current MR findings that are relevant for hepatologists. Furthermore, we will discuss how to evaluate the quality of MR publications, interpret MR findings, and illustrate good practice of using MR studies in hepatology.
Keywords: Mendelian randomization; Genome-wide association studies; Liver diseases; Metabolic dysfunctionassociated steatotic disease; Cholelithiasis

INTRODUCTION

“Causation” versus “correlation”

Chronic liver diseases claim millions of lives worldwide annually, hence, it is urgent to prevent their occurrence and alleviate the liver-associated public health burden [1]. While the rapid development in genomics, sequencing techniques and molecular epidemiology facilitates such research, large data-based studies often cannot differentiate between “correlation” and “causation”. Traditional observational studies, such as case-control design and cohort studies, can only offer the “correlation” rather than the “causation”, and even a sophisticated prospective study design can be biased by reverse causation and unmeasured confounders, e.g., undetected socioeconomic factors [2]. For instance, higher serum high-density lipoprotein cholesterol (HDL-C) was observed to be associated with lower risk of cardiovascular disease (CVD), but there was no significant reduction in CVD risk when elevating serum HDL-C [3]. Such discrepancy indicates the challenge when translating observational results to clinical interventions tested in randomized controlled trials (RCT).

MR can be a proxy for the impractical RCT

High-quality RCTs provide sufficient confidence on causality and the highest level of evidence in clinical practice. However, RCTs usually cost a lot of time and money, and sometimes it is unfeasible to conduct an RCT, as manipulating the exposure is extremely difficult or unethical, especially when the exposure is a disease [4]. For example, designing an RCT to assess the causal relationship between metabolic dysfunction-associated steatotic disease (MASLD) and type 2 diabetes mellitus requires observing the patients developing another disease without any intervention, which would be unethical. To address such issues, Mendelian randomization (MR) has been proposed to possibly fill this gap and help to provide evidence of causation without RCTs [5]. MR can give the causal estimate, which can be viewed as an estimate on a specific intervention to some extent [6]. Thus, the MR design may be a suitable proxy for RCT when carrying out an RCT is costly or unethical [4].

We performed a literature search on the topic of “Mendelian randomization” and limited the publication type to “Article” in the Web of Science database from 2020 to 2024 (Fig. 1). This reflects growing interest and recognition of the method, but perhaps also an increase in low-quality studies, contributing to controversial findings and a potential credibility crisis. Citation counts began to decline in medicine from 2021 and in hepatology from 2022, possibly due to the lower quality of recent studies or simply a shorter time frame for citations to accumulate. Given this surge in MR publications, careful evaluation involving statisticians, epidemiologists, and hepatologists is essential. This review aims to summarize current MR findings in hepatology, guide interpretation for hepatologists, and present a state-of-the-art example of MR analysis.

Mendelian randomization: concepts and principles

MR uses genetic variants, usually single nucleotide polymorphisms (SNPs), as the instrumental variable (IV, which will be referred to as genetic instruments) to assess the causal relationship between putative risk factors and the outcome [7]. The most important strength of MR is that it can estimate a causal relationship using observational data instead of interventional strategies. Primarily, it should be established on 3 core IV assumptions as follows: (a) relevance: genetic variants should be closely associated with the exposure; (b) exchangeability: there is no common confounder between the genetic variants and the outcome; (c) exclusion restriction: genetic variants can be linked to the outcome only via the way of exposure without different mechanisms (no horizontal pleiotropy) (Fig. 2A) [7,8]. Only when genetic variants satisfy all IV assumptions can they be called “valid” genetic instruments. Besides, a majority of MR investigations assume the linearity between the exposure and outcome, and no interaction between the exposure and mediators [9]. According to Mendel’s Law of Inheritance, gametes are randomly allocated at conception and the different levels of exposure determined by genetic variants can be viewed as being distributed randomly within a specific population; thus the unmeasured confounders are balanced between groups [10]. Additionally, since the genotype is fixed at conception, the results should not be biased by reverse causation [11]. Thus, the MR design, which uses genetic instruments to proxy for an exposure, can be assumed as a “natural” randomization executed by genetic instruments (Fig. 2B, Table 1) [12,13].

Development and recent advances in methodology

A historic overview on the development of MR

The idea of MR was initiated in 1986 when the apolipoprotein E isoforms could be used as a proxy for serum cholesterol to estimate the associations of serum cholesterol with cancer risk [14]. Subsequently, Smith and Ebrahim [2] formally proposed the term “Mendelian randomization” where they introduced the potential value of MR in understanding risk factors of diseases. Another milestone in MR research is the release of STROBE-MR guideline [7,15]. In 2021, Burgess et al. [10] released the newest book on MR and it was translated into Chinese in 2023, which may have contributed to the exponentially increased number of MR studies from China [16]. A prominent example in early MR research is that it rules out the causal relationship between HDL-C and myocardial infarction (MI) [3], indicating that elevating HDL-C may not be effective for preventing MI. However, such a design requires collecting individual-level data, which is expensive and time-consuming [17].

With the rapid development of genome-wide association studies (GWAS), which test the associations of millions of SNPs with a phenotype, MR investigation based on summary-level data, usually called “two-sample MR”, has gradually gained popularity due to the easy access to GWAS summary statistics and some easy-to-use statistical methods [18]. Although inverse variance weight (IVW) method, which combines Wald ratios estimated from multiple genetic variants and weights them by the inverse of their variance, is mainly applied in the preliminary analysis due to its good power, it requires that all genetic instruments be valid, which often does not hold [18]. Thus, IVW may not always be the best option, especially since its causal estimations can easily be biased, and other methods were proposed, such as MR-Egger method [19], median-based method [20], and mode-based method [21]. These methods relax the basic IV conditions. Therein, the MR-Egger method is robust under a weaker set of assumptions though its estimate is often not precise, and it is a method to detect horizontal pleiotropy as well [19]. The median-based method can make a valid causal inference even though up to 50% of genetic instruments are invalid [20]. Besides, the mode-based method can give the consistent estimate even if the majority of instruments are invalid [21].

Advanced statistical methods and extension to other areas

Additionally, several advanced methods have been developed to improve the MR analysis, and they can be categorized as follows: (i) assessing horizontal pleiotropy: MR-PRESSO is a statistical method to test pleiotropic outliers and give a causal estimate after removal of these outliers [22]. MR-CAUSE is a statistical method accounting for both correlated and uncorrelated horizontal pleiotropy [23]; (ii) overcoming weak instrument bias: it refers to the genetic instrument that has a relatively weak association with the exposure, often caused by a small sample size. Such instruments can still be used in MR analysis but should be treated cautiously. The “MR.RAPS” method can be used to assess the weak instrument bias [24]; (iii) evaluating “collider bias” or “index event bias”: it refers to the “bias induced in the association between two variables when conditioning on their common effect” [25,26] and it often happens when selecting a subset cohort for analyzing risk factors of disease progression (e.g., the subset with all diabetic patients where “diabetes” is the “collider” or “index event”). Several methods have been proposed to adjust for such bias such as “SlopeHunter” and “ColliderBias” [27,28]; (iv) handling multidimensional data: a two-sample MR approach based on Bayesian model averaging can handle multiple risk factors from high-throughput experiments (metabolome and proteome) in a multivariable design [29]. Besides, other methods are well summarized by Sanderson et al. [8] and a recent benchmark study implicates that MR-APSS, MR-Egger, MR-CAUSE, and mode-based method can reduce type I errors and be better choices in analysis [30]. The MR methodological research is in rapid development, and it has been recently incorporated into artificial intelligence [31] and single-cell data [32]; however, as these fields are currently emerging, caution should be taken when applying them to research projects.

Despite the abovementioned statistical methods, there are three key limitations of this method: (1) Untestable assumptions: MR can possibly yield misleading results even when the statistical analysis is technically sound because MR relies on some untestable assumptions (i.e., random mating, validity of independence assumption, validity of exclusion restriction assumption). Such untestable assumptions consist of inborn weaknesses of MR and should be carefully considered before analysis, especially in combination with biological plausibility and clinical relevance. (2) Weak instrument bias: since genetic instruments can only explain a very small proportion of phenotypic variance in complex traits, the sample sizes are usually very large in MR analysis and the small cohort-based MR studies are often underpowered. However, weak instrument can still be useful to proxy for the exposure as its association with exposure will increase with the enlarged sample size. A better approach to dealing with weak instrument bias is to select genetic instruments from small cohorts but use the summary statistics from independent large cohorts, which will also enhance the robustness of MR study. (3) Hypothesis multiplicity: the MR analysis can test almost an infinite number of “causal hypothesis” due to the explosion of publicly available GWAS datasets on multi-omics analysis (e.g., transcriptome, proteome, and metabolome); however, there are no recommended strategies to mitigate the impacts of increased false positive rates or even data contamination. A possible solution may be putting MR analysis as a supportive method of the whole study instead of using it alone (https://www.kidney-international.org/content/authorinfo#mendelian-randomization-studies-1-5).

Last but not least, sensitivity analyses (e.g., leave-one-out analysis, MR-Egger, weighted median, weighted mode, MR-PRESSO, MR-CAUSE, and MR.RAPS) are essential to guarantee the robustness of MR findings and strengthen the evidence of causal inference, especially when there is potential bias from heterogeneity, horizontal pleiotropy or weak instruments.

Applications of Mendelian randomization in liver diseases

Metabolic dysfunction-associated steatotic disease

MASLD, previously termed as nonalcoholic fatty liver disease (NAFLD) or metabolic dysfunction-associated fatty liver disease, is the most studied liver disease using MR method [33,34]. For clarity and simplicity, we will use MASLD to replace NAFLD throughout this review, although most of MR studies had been conducted using the “NAFLD” terminology.

In the context of MASLD as the outcome, several metabolic risk factors have been confirmed to be causally associated with an increased risk of MASLD (Fig. 3), namely higher body mass index (BMI), higher waist circumference, genetic liability to insulin resistance, genetic liability to type 2 diabetes (T2D), genetic liability to hypothyroidism, higher systolic blood pressure, genetic liability to hypertension, higher liver enzymes, higher triglycerides, lower HDL-C, higher serum ferritin and higher serum iron (Table 2) [35-40]. These findings have already been incorporated in the most recent MASLD guideline [41]. In addition, higher serum homocysteine and lower serum folate may increase the risk of MASLD [42]. Lifestyle is another important aspect of MASLD, and smoking and drinking are two established risk factors, while coffee consumption and vigorous physical activities reduce the risk of MASLD [35,36].

In addition to all those metabolic risk factors and lifestyles abovementioned, genetic liability to clonal hematopoiesis of indeterminate potential is a newly identified risk factor for MASLD and MASH (metabolic dysfunction-associated steatohepatitis) in large cohorts [43]. Diet-derived antioxidant selenium may also increase the risk of MASLD [44]. Although air pollution is not suitable to be a direct exposure in MR analysis, a recent study links air pollution, metabolites and MASLD, where the air pollution-associated metabolites are treated as the exposure and give insights into the impact of air pollution on MASLD, which appears plausible [45]. Great attention should be paid to ethnic differences, which can reflect a distinct genetic susceptibility to risk factors. For instance, vitamin D is protective for MASLD in Europeans [46] while not significant in Chinese population [47]. Besides, the selection of a GWAS dataset can also affect the results, as one study reported no association between serum vitamin D and MASLD, even though all samples were from Europeans [48]. This lack of association in Europeans may be explained by a relatively small sample size, as this study used only 1 MASLD GWAS (which may not have sufficient power), while a significant result was obtained from the combination of 3 MASLD datasets. An enlarged sample size will achieve high statistical power, and MR analysis usually needs much more samples than traditional cohorts to give reliable causal inference [17]. Thus, we recommend that investigators include more high-quality GWAS datasets to obtain reliable results when performing MR analysis.

When treating genetic liability to MASLD as the exposure, a critical issue is the selection of valid genetic instruments, as numerous MASLD-related SNPs are strongly associated with T2D and obesity [49]. It is difficult to dissect the shared genetics between MASLD and other cardiometabolic diseases, and it is almost impossible to select sufficiently valid instruments for this purpose. Two approaches can be used to select genetic instruments: one is a biological mechanism-driven approach and it obtained two variants, rs738409 (PNPLA3) and rs2228603 (NCAN) [37]; the other is a statistical significance-driven approach and it obtained up to 44 genetic instruments for MASLD [50]. Both implicate that genetic liability to MASLD could increase the risk of T2D and obesity [37,50]. In comparison, the significant result from a biological mechanism-driven approach may be more convincing. Another important aspect is to assess the heterogeneity of MASLD and its subtypes’ impacts on cardiometabolic diseases, which has been unveiled by recent advances [51]. Current MR evidence supports that genetic liability to MASLD would increase the risk of coronary artery disease [50,52]. In recent studies, MASLD heterogeneity has been dichotomized where one subtype is MASLD mainly leading to progressive liver damage and the other is MASLD mainly leading to systemic effects with increased risk of CVDs [53,54]. It will be of relevance to dissect the genetic variants associated with two subgroups and assess their intrahepatic and extrahepatic impacts using MR. Besides, genetic liability to MASLD may increase the serum uric acid levels but not vice versa [55,56]. All the evidence pinpointed that genetic liability to MASLD might trigger other severe metabolic disorders, emphasizing that preventing MASLD would be a very important strategy from a public health perspective. Last but not least, most MR studies are based on European ancestries, and one should be cautious when extrapolating these findings to other ethnicities.

Despite the meaningful findings from MR analyses in MASLD, there are still several critical questions, which can compromise the confidence and generalizability of MR results. One issue relates to the diagnostic criteria of MASLD: the most reliable one is MASLD derived from liver biopsy [57]; however, there are some MASLD GWAS using questionnaire or electronic health records to determine MASLD [58]. Besides, MASLD has also been proxied by (chronically elevated) ALT [59] or serum and anthropometric traits [60]. Although using proxy diagnostic criteria can increase the statistical power with an enlarged sample size, it may reduce the clinical relevance or hamper the interpretation and translation of MR findings. In Table 3, we summarize the characteristics of selected MASLD-related GWAS datasets and try to give recommendation on their application in MR investigation. The recommendation takes into consideration the definition of MASLD, ethnicity and sample size. Of note, use of rs738409 in MASLD-related MR is complex and depends on context: it is valid when MASLD is the exposure but problematic for MASLD related progression (i.e., liver cirrhosis and hepatocellular carcinoma [HCC]) being the exposure due to pleiotropy. Possible solutions include conditional analysis (using conditioned effects on cirrhosis or HCC in MR analysis), multivariable MR (treating MASLD, MASLD-related cirrhosis and HCC as exposures), colocalization to dissect the genetics between MASLD incidence and progression, or sensitivity analyses with/without the SNP to ensure robust results. Currently, there is no standard criterion but multiple strategies are recommended to ensure the robustness of the results.

Alcohol-related liver disease

Alcohol-related liver disease (ALD), mainly characterized by the excess consumption of alcohol and subsequent liver injury, steatosis or steatohepatitis, is a leading cause of liver cancer and liver transplantation worldwide, especially in Western countries [61]. In addition to the alcohol consumption per se, MR studies unveiled that the gut dysbiosis and circulating inflammatory cytokines could increase the risk of ALD as well [62-64]. Therein, Escherichia coli was a protective factor, while Roseburia hominis and Porphyromonadaceae were risk factors for ALD [62]. Two studies consistently discovered that genetically elevated serum interleukin-7 could increase the risk of ALD [63,64]. Besides, the genetically lowered serum sphingomyelins could increase the risk of ALD as well [65].

When treating ALD as the exposure, one study reported that genetic susceptibility to ALD can reduce the bone mineral density in the femoral neck, which was possibly mediated by reduced serum vitamin D [66]. As mentioned above, SNP rs738409 is closely associated with steatotic liver diseases and is implicated in both ALD and MASLD [67]. Thus, it is necessary to dissect the genetic effects between ALD and MASLD when selecting genetic instruments (e.g., determining SNPs associated with either or both), and possible solutions include but are not limited to colocalization and multivariable MR [11]. Although the multivariable MR was applied in assessing the impact of ALD on bone mineral density adjusting for MASLD, it was not powerful enough to dissect the shared genetics between ALD and MASLD. Future MR studies should consider dissecting the genetics of ALD and MASLD first and then apply MR method to study causal inference.

Viral hepatitis

Hepatitis B virus (HBV) and hepatitis C virus (HCV) are the two leading etiologies of chronic viral hepatitis investigated using MR method. HBV is much more investigated than HCV and most of relevant studies used genetic data derived from East Asian ancestry. One study investigated the causal relationship between HBV infection and extrahepatic cancer, and they found that genetic susceptibility to HBV infection could increase the risk of cervical cancer and gastric cancer [68]. This study employed 4 SNPs to proxy for the genetic susceptibility of HBV infection, including SNP rs1419881 in the TCF19 region, which was independently reported to be associated with HBV infection [69] but lacked experimental validation. Intriguingly, there was no significant SNP located near the NTCP region, a well-known receptor of HBV. Empirically, SNPs located near NTCP gene are more convincing as genetic risk factors for HBV infection than other SNPs. However, when there is no biologically plausible SNP for the exposure, the SNP whose association with the exposure was validated in independent datasets would be reasonable. Thus, we should avoid selecting a lot of agnostic SNPs simply via the threshold of GWAS P-value <5×10^-8. This consideration similarly applies to HCV as well. In addition, HBV infection was also found to increase the risk of osteoporosis [70] and rheumatoid arthritis [71], highlighting the potential liver-bone-joint axis which may be mediated by the imbalance of HBV X protein and trans-ferulic acid [71]. For HCV, one study reported a genetic susceptibility to an increased risk of diffuse large B cell lymphoma [72]. The results are summarized in Table 2.

Primary liver cancer

Primary liver cancer mainly includes HCC and intrahepatic cholangiocarcinoma (CCA) [73]. Due to various etiologies, liver cancer is quite heterogeneous and such heterogeneity poses a great challenge to related MR research as different causes of HCC usually indicate different genetic basis. For example, the viral HCC and MASLD-HCC are caused by different triggers, representing different molecular mechanisms, therapeutic options and prognosis, and it is necessary to separate them into different analyses when evaluating outcomes [74].

A Biobank Japan (BBJ) MR study discovered that alcohol consumption increases the risk of HCC while coffee, tea, milk, and yogurt consumption reduce its risk [75,76]. Another study implicated that low-to-moderate wine drinking may reduce the risk of HCC in Europeans [77]. Genetic liability to insomnia and daytime napping may be risk factors of primary liver cancer as well [78]. In addition to diets and lifestyles, the lower total basophil neutrophil counts (sum of basophil counts and neutrophil counts) might be an independent risk factor for HCC from the multivariable MR analysis [79], and lower circulating vitamin C could increase the risk of HCC occurrence [80]. Besides, a higher serum adiponectin may increase the HCC risk [81]. As for the causal effects of serum lipids on HCC, especially LDL-C, lower LDL-C was reported to increase the risk of HCC. MR also unveils that visceral and ectopic adiposity, instead of BMI, can increase the risk of liver cancer [82]. Additionally, genetic liability to inflammatory bowel disease (IBD) could increase the risk of HCC using BBJ data [83], and another BBJ MR study suggested that genetic liability to rheumatoid arthritis might decrease the risk of HCC [84]. The latter seems biologically implausible, and the shared genetics and horizontal pleiotropy are not well dissected in the analysis as most genetic variants with autoimmune diseases are located in the MHC region and are highly pleiotropic.

There is a limited number of MR research on CCA. Chen et al. [85] systematically tested the causal relationship between 26 putative risk factors and CCA, and found that genetic liability to MASLD and cholelithiasis are two independent risk factors of CCA. However, the causal relationship between primary sclerosing cholangitis (PSC) and CCA was not significant, and this may be attributed to the heterogeneity of CCA in the GWAS participants [85]. It is probable that the CCA GWAS does not contain enough PSC-related CCA cases. Besides, genetic liability to insomnia could increase the risk of CCA [78], and IBD is another risk factor as well [83]. Recent studies highlighted the role of gut microbiota in the pathogenesis of CCA and they were reported to affect the risk of CCA as well where the genetically elevated abundance of family porphyromonadaceae, genus bacteroidetes and phylum verrucomicrobia might be protective factors [86,87]. However, most evidence about CCA is established on European data, which limits their global application to clinical practice.

Furthermore, we should be very careful about the exposures related to human behavior and medical treatments such as milk and yogurt consumption as their associations with genetic variants are usually indirect [88,89]. Notably, the current MR studies usually rely on two HCC GWAS datasets: (1) one is from the BBJ (1,866 cases+195,745 controls); (2) the other is from UK Biobank (UKB) with HCC cases <500. The two datasets have distinct characteristics, including ethnicity difference and the discrepancy in the number of cases. Due to a relatively small sample size, a possible problem in current MR studies related to liver cancer may be insufficient statistical power, which was not reported by many studies. Besides, these data did not specify HCC subtypes. Furthermore, the current evidence from MR studies is still weak in liver cancer research, and we cannot solely use these to change clinical practice without solid RCTs.

Biliary diseases

For biliary diseases, we will mainly focus on cholelithiasis with or without cholecystitis, primary biliary cholangitis (PBC) and PSC, as these entities constitute a majority of MR publications related to biliary diseases.

The genetically predicted sedentary lifestyle, characterized by more leisure screen time and less moderate-to-vigorous intensity physical activities, can increase the risk of cholecystitis and cholelithiasis [90]. Also, smoking initiation, defined by the actual beginning of smoking, is another risk factor of both, while alcohol consumption seems not to affect either [91]. In contrast, higher levels of circulating omega-3 fatty acid concentrations have protective effects on cholelithiasis and cholecystitis [92], and higher serum campesterol can reduce the risk of cholelithiasis and cholecystitis as well, especially supported by the shared ABCG5/ABCG8 gene regions [93]. The genetic susceptibility to both cholecystitis and cholelithiasis can increase the risk of gastroesophageal reflux disease (GERD), and they may mediate the effects of BMI on GERD [94]. For cholelithiasis only, the genetic liability to obesity and T2D has been confirmed to robustly increase the risk of cholelithiasis [95,96]. An interesting finding of cholelithiasis is that the genetically elevated serum total cholesterol may reduce the risk of it while genetically elevated serum triglycerides might increase its risk [97]. Furthermore, analyses of lipid metabolism-associated targets unveil that genetically reduced serum cholesterol proxied by HMGCR inhibition might reduce the risk of cholelithiasis [97]. Besides, some metabolites have been proposed to be causally associated with cholelithiasis where genetically elevated sphingomyelin and unsaturated fatty acid could reduce the risk of it [98]. When treating cholelithiasis as the exposure, its strong association with CCA was confirmed [85]. In addition, genetic susceptibility to cholelithiasis is linked to the increased risk of colorectal cancer [99] and kidney cancer [100]. Currently, most cholelithiasis or cholecystitis-related MR are mainly derived from UKB and FinnGen, and these results are encouraged to be validated in other cohorts with more genetic diversity.

Several risk factors have been identified for PBC, including the genetic liability to IBD [101], genetic liability to systemic lupus erythematosus (SLE) [102], genetic liability to obesity [103,104] lower hemoglobin level [105], lower vitamin D [104]. The link between PBC and other autoimmune disorders such as IBD and SLE appears plausible, although it has no immediate therapeutic implications. Additionally, the higher bioavailable testosterone and lower sex hormone binding globulin may increase the risk of PBC in female specifically [106]. Also, genetic liability to PBC increases the risk of IBD [107], celiac disease [108], rheumatic arthritis [109], thyroid dysfunction [110], osteoporosis [111] while reducing the risk of gastric cancer [112]. As for PSC, the genetic liability to celiac disease may increase its risk and it could elevate the risk of celiac disease as well [108]. Additionally, the genetic liability to IBD and SLE could increase the risk of PSC as well [102,113]. Of note, when analyzing the genetically predicted associations of PBC and PSC with other autoimmune diseases (e.g., SLE and IBD), a critical issue is the shared genetics among different autoimmune or autoimmune-like diseases as most of genetic instruments are located in the major histocompatibility complex (MHC) region, and it is hard to guarantee that selected genetic variants affect the exposure first and then the outcome. Possibly, the genetic instruments may affect both exposure and outcome simultaneously. Also, the genetic effects of MHC variants are complicated and some GWAS analyses even remove this region directly. Thus, the GWAS results could be biased and would then lead to biased results in MR.

Target identification and drug repurposing

Target identification and drug repurposing are two new emerging applications of MR methods. Target identification refers to discovering novel targets for a specific disease, and drug repurposing means to decipher unknown effects of an existing drug [114]. In the drug-target MR, additional principles should be met: (i) traditionally, the genetic instruments should be located in or near genes that encode the drug targets, which is also called cis-MR [115]. Besides, the trans-instruments, which represent genetic variants far away from the target gene, can also be useful in target identification and drug repurposing, but the potential bias of horizontal pleiotropy should be carefully assessed when using trans-instruments [114]; (ii) the exposure should usually be a disease biomarker or a protein, which represents the intervention effects [115]; (iii) the genetic instruments should be causally associated with the disease biomarker or protein, usually reflecting the biological mechanism [115,116]. For instance, the genetic variants should be located within the HMGCR region when studying the effects of statins, and the exposure can be LDL-C as it is a biomarker reflecting the drug effects. Detailed information about drug-target MR and its application in liver diseases has been well depicted in a recent state-of-the-art review by Luo et al. [117].

Two recent studies have proposed two novel potential targets for MASLD, namely lipoprotein lipase (LPL) [118] and sodium-glucose cotransporter-1 (SGLT1) [119]. Therein, LPL-activation and SGLT1-inhibition are predicted to reduce the risk of MASLD. Moreover, the inhibition of SGLT2 has been identified to reduce the risk of MASLD and its related liver progression by MR, which was further corroborated in large cohorts, in which SGLT2 inhibitors have been used as a treatment for cardio-renal-metabolic diseases such as T2D [120]. Similarly, the agonism of glucagon-like peptide 1 (GLP-1) and glucose-dependent insulinotropic polypeptide (GIP) receptors was found to reduce the risk of developing MASLD via the effects of glycemia control and weight loss as well [121]. This prediction is supported by clinical trial results using the GLP-1/GIP coagonist tirzepatide in patients with MASLD [122]. Besides, another interesting study used drug-target MR to corroborate their experimental finding that inhibition of ATP-citrate lyase (ACLY) gene expression ameliorates MASH and dyslipidemia [123]. For ALD, the drug-target MR identified that ENPP2 (Autotaxin) could increase the risk of ALD, which might be mediated by effector memory CD8+ T cell, suggesting that targeting ENPP2 may be a therapy for ALD [124]. Besides, CFHR1 and CFHR² were reported to reduce the risk of ALD-related cirrhosis, and could be candidates to prevent the progression of ALD [125,126].

In addition to MASLD and ALD, drug-target MR has also been applied to other liver diseases. For instance, three proteins (ficolin-1, CD40, and FAM177A1) have been identified to be candidate targets for PBC [127], and AIF1 and HLA-DQA2 have been implicated in the pathogenesis of PSC [128]. Intriguingly, a recent drug-target MR systematically examined the effects of antihypertensive targets on HCC and discovered that thiazide diuretics should reduce the HCC risk in both Europeans and East Asians, but beta-adrenoceptor blockers might increase the HCC risk specifically in Europeans [129].

As described in the abovementioned studies, a downstream biomarker, such as LDL-C, can be the exposure proxy for the drug effects. Also, the gene expression and protein level can be exposures as well, and an SNP associated with gene expression is called eQTL (expression quantitative trait locus) and that associated with protein level is called pQTL (protein quantitative trait locus). If the effect of a specific target is discovered and confirmed by multi-level exposures, such an effect is more likely to have great potential for clinical applications. Usually, the significant drug-target MR results need verification by colocalization, a Bayesian model to find the shared mechanism between exposure and outcome [11,130]. Please note that although the role of ACLY and other targets (e.g., SGLT2 and HMGCR) in MASLD was corroborated using experimental validation [123,131,132], there is still a large gap between data-driven discoveries and experimental validation in the field of drug-target MR, which offers both opportunities and challenges in liver research.

Of note, MR is a predictive tool in drug repurposing and methodological research is trying to make the prediction more accurate. For instance, a recent study proposes a novel concept called “pathway-partitioned MR”, which offers possibility to explore the pathway from which the drug target affects the disease [133]. Such analyses can facilitate the design and conduct of functional experiments or real-world data analysis, avoiding unnecessary attempts. However, functional experiments and real-world data analyses are still necessary and cannot be replaced when translating drug repurposing findings from MR analysis into clinical implementation. Last but not least, the result of a drug-target MR study usually reflects a life-long effect while the RCT represents a shorter-duration effect, which needs specific attention in clinical implementation [134].

Take-home points to assess the quality of Mendelian randomization studies

There are some practical tips on how to assess MR results for hepatologists and other researchers:

(i) Genetic relevance of the exposure: Considering the potential misuse of MR in medical research, the first and easiest step to identify a high-quality MR study is to assess genetic relevance of exposure as higher genetic relevance means more credibility of the MR study. If an exposure cannot be reasonably explained by genetic variants, such as noodle consumption, chopstick use, and exposure to air pollution, such MR may be implausible or a reflection of misuse of this method [135,136].

(ii) Ethnicity and population: there are substantial genetic differences between European mainland and Finnish populations though they are both of Caucasian ancestry [137], let alone the genetic difference between Caucasian and non-Caucasian ancestries. Besides, “population” refers to a group of study participants and it should be either monoethnic or multiethnic. Usually, the study population should be ideally monoethnic or the majority of it should be monoethnic. Thus, a valid MR design should specify the ethnicity and the conclusion should not be extrapolated to other populations arbitrarily.

(iii) Total strength and average strength: total strength refers to the total variance of exposure explained by used genetic instrument, usually indicated by R², and average strength usually takes the number of instruments, and it is calculated by the formula:

F=N−k−1k×R21−R2

N is the sample size of continuous exposure or the effective sample size of binary exposure; k is the number of genetic instruments used in MR analysis; R² is the total variance of exposure explained by used genetic instruments [138]. If all genetic instruments are independent, the total R² equals the sum of each SNP’s R².

(iv) Effect size and confidence interval: The most important aspect of MR design is to determine the existence of causal relationship, and detecting the causation is much more important than estimating the causal effect sizes [16]. The main reason is that the effect size estimated from MR analysis usually does not represent the clinical effect size. In addition, the causal effect estimated from MR often means a chronic life-long effect of exposure on the outcome, as genetic variants usually have persisting effects throughout the life [6]. Accordingly, the MR might be useless in studying acute diseases or highly varied exposures. However, the direction of effect size and 95% confidence interval should be helpful to judge how exposure impacts the outcome and determine how to design the clinical trial.

(v) Consistency of results in different statistical methods: As mentioned in the section “Development and recent advances in methodology”, different MR methods rely on distinct assumptions and they can help to test horizontal pleiotropy and heterogeneity. Reporting MR results only derived from IVW method cannot be convincing, and may introduce inflated type 1 error. According to the benchmark study, four methods (i.e., MR-APSS, MR-Egger, MR-CAUSE, and weightedmode) display good performance in controlling type 1 errors, and researchers are encouraged to consider them in the analysis [30]. Only when results are significant in different methods can we claim a valid causal inference.

(vi) Consistency with evidence derived from other studies: Apart from evaluating MR results using different statistical methods, MR findings should be corroborated by evidence from different sources, such as cohort analyses, colocalization analyses, or even functional experiments. For instance, the negative correlation between serum cholesterol levels and cholelithiasis identified through MR was further validated by colocalization analyses and cohort studies [77]. While obtaining extensive individual-level data and conducting functional experiments remains challenging and resource-intensive, researchers can utilize diverse in silico approaches to support MR findings.

An illustration of performing the state-of-the-art Mendelian randomization study

(i) Defining a plausible causal hypothesis at first: Similar to what has been discussed in “Genetic relevance of the exposure” of the previous section, defining a research question which can be answered by a genetic instrument is the first and crucial step in MR research. Therein, the levels of exposure proxied by genetic instruments can have the same (equivalent) effects on the outcome as the levels of exposure changed by environmental factors [139]. In the research of MASLD, the impact of genetic variants can be proxied by PNPLA3-rs738409, which can have similar effect on MASLD as in unhealthy diet or other environmental factors [37]. When studying effects of serum lipids on cholelithiasis, it is plausible to use genetic instruments located in the gene involved in lipid metabolism (e.g., HMGCR, LDLR, and PSCK9) [97].

(ii) GWAS data selection: For one-sample MR design, a key point is to use SNP’s weight derived from an external cohort and then to perform causal inference based on the genotype, exposure and outcome from the same cohort, which can reduce the rate of false positive and avoid the overestimation of causal effects. When conducting two-sample MR design, it is necessary to ensure that the exposure GWAS and outcome GWAS are performed in different populations, which means no sample overlap when estimating SNP-exposure associations and SNP-outcome associations. For instance, the genetic associations with serum lipids can be extracted from global lipid genetics consortium, and genetic associations with cholelithiasis can be obtained from UKB or FinnGen, suggesting it is likely that there is no sample overlap [97]. Empirically, two key aspects should be considered, including phenotypic heterogeneity and effective sample size. Phenotypic heterogeneity refers to the variation within a phenotype. As previously mentioned above, MASLD can be divided into at least two groups (i.e., hepatic MASLD and systemic MASLD) [53] and the analysis should be performed in hepatic MASLD and systemic MASLD separately so that different modifiable risk factors can be identified for them precisely. Besides, if exposure is binary, a continuous exposure should be better used as an additional exposure, e.g., the levels of serum alanine aminotransferase should be considered as exposure as well if treating genetic liability to MASLD as the exposure, which could reduce the selection bias. Effective sample size can affect the study power, and please note that effective sample size equals the total sample size for continuous phenotype but it should be calculated by the formula below for binary phenotypes:

Neff=41Ncases+1Ncontrols

The N_eff is the effect sample size, N_cases is the number of cases and N_controls is the number of controls. Please note that although GWAS summary statistics are frequently used in MR studies for their statistical efficiency and public availability, MR can be conducted using any dataset where the associations between genetic variants and the exposure, as well as the outcome, are robustly characterized. This includes well-phenotyped cohort studies or other genotyped resources, provided that the assumptions of MR are adequately satisfied.

(iii) IV selection: Selecting valid IVs (also called genetic instruments) for MR analysis is a key step to produce convincing and reproducible results. Simply, two strategies are usually adopted in practice where one is called biological mechanism-driven approach, which chooses IVs based on biological and clinical evidence, and the other is statistical significance-driven approach, which is solely established on genome-wide significance (i.e., choosing SNPs with P-value <5×10-8) [89,138]. The former choice is more preferred as it can satisfy the IV conditions to the largest extent while the latter choice is an agnostic way and can be used as a complementary or sensitivity analysis. For instance, when choosing IVs for MASLD, the SNP rs1260326 from GCKR is a very strong signal, but GCKR is strongly associated with T2D and other metabolic pathways. Although a genome-wide selection approach will reserve this SNP as a valid IV in the MR analysis, whether it can be a valid IV should be specified in different scenarios. This means that one needs to consider whether T2D is a confounder via different mechanisms in the MR analysis. If so, we need to remove this SNP in the analysis. Additionally, a foreseeable problem for statistical significance-driven approaches is that it can yield a lot of IVs with the increase of sample size in GWAS study, however, such increment in the number of IVs may not provide valuable information but can possibly select more pleiotropic SNPs [140]. In addition, positive controls and negative controls are also useful to assess the validity of genetic instruments. The positive controls and negative controls have similar meanings in experiments, and they are used to assess the validity of selected IVs [5,141]. Therein, positive control refers to assessing the causal relationship between exposure and a well-established outcome. For instance, when assessing the effects of LDL-lowering targets on cholelithiasis (to be established), it is better to estimate the effects of these targets on CAD (well established) using selected IVs. The final result will be convincing if the effects of them on CAD are significant and consistent [142]. If the “positive control” does not hold, the “relevance” assumption may be violated, and there may be weak instrument bias in the results. The negative control refers to guaranteeing that the exposure cannot affect an impossible outcome (e.g., the birth weight, adulthood height or hair color) using the same IVs. However, if the “negative control” does not hold, the “exchangeability” or “exclusion restriction” assumption may be violated, and the results can be biased by unknown confounders or horizontal pleiotropy [141]. Usually, the researchers can apply both “positive control” and “negative control” to their MR analysis to make results convincing. Last but not least, the MR-Steiger directionality test can also be a good choice to assess the validity of genetic instruments [143].

(iv) Choice of statistical methods: As displayed in Figure 4, the IVW method is often the primary choice if there is neither heterogeneity nor horizontal pleiotropy. Despite its good power, IVW method can be easily biased because of invalid IVs and this is often the case. If there is any invalid IV, which can be assessed by heterogeneity or pleiotropy test, it is recommended to perform MR-PRESSO to detect outliers and re-analyze the data after removal of these outliers. It should be noted that there may be heterogeneity or horizontal pleiotropy even if the MR-PRESSO suggests no outliers. Under such circumstances, the median-based and mode-based methods are recommended to obtain robust associations if there is significant heterogeneity. For horizontal pleiotropy, the MR-Egger and MR-CAUSE methods are recommended. Besides, the MR-APSS is also recommended by a recent benchmark study [30]. All these abovementioned methods can be tried in the analysis and the evidence will be more robust if more statistical methods display consistency. Of note, different statistical methods rely on different assumptions and it is important to assess whether these assumptions are satisfied before application.

(v) Validation using independent datasets: Preliminary MR results should be, if possible, validated in independent datasets. A commonly used approach is choosing two or more independent outcome GWAS datasets (e.g., UKB, FinnGen and other genetics consortia) and estimating the genetically predicted association between exposure and outcome using the same IVs. However, this approach requires different outcome GWAS datasets, which should be derived from the same ancestry as the exposure, and the conclusion may not be extrapolated to other ancestries. The other approach is to validate MR results in different ancestries, which can also be called “trans-ancestry MR” (e.g., UKB and FinnGen for European ancestry and BBJ for East Asian ancestry). In this scenario, the IVs and GWAS datasets should be changed in the validation. If validated in other ancestries, the MR conclusion may have a broader application.

(vi) Comparison with evidence from other sources: It is always recommended to compare MR results with RCTs to strengthen the causal relationship. Functional experiments are an ideal approach to explore the mechanism or validate the causal relationship when molecular phenotype (e.g., the mRNA level or the protein level) is the exposure. If such experiments are not feasible, an alternative way may be colocalization analysis, especially in the context of drug-target MR analysis which is often accompanied by colocalization. For instance, colocalization analysis supports the preventive effects of HMGCR inhibition on cholelithiasis [97]. However, caution should be taken when comparing MR results with those derived from observational studies such as cohort analyses, because the MR estimates will be biased toward the direction of observational estimates if there are weak instruments in MR analysis [140]. Nonetheless, MR can be a cost-effective choice for external validation and enhancing generalizability, especially when combined with large-scale real-world data (e.g., biobanks, large-scale electronic health records or clinical databases), as we had recently used this approach to confirm the inverse associations of serum cholesterol with cholelithiasis [97].

CONCLUSION

As a method for causal inference in medical research, both methodology and application research of MR are experiencing rapid developments. Undoubtedly, these MR findings may provide novel and meaningful insights into the causal risk factors of liver diseases and may serve as guidance for clinicians. However, we need to judge the reliability of related MR results and perform MR analysis in a rigorous fashion, before considering translating MR findings to clinical practice.

FOOTNOTES

Authors’ contributions

L.C. and G.L. conceptualized this review, and L.C. drafted the manuscript. Q.R. revised the review and contributed to visualization. M.G. gave some statistical and epidemiological advice, and revised the manuscript. F.T. substantially revised this review from the perspective of hepatologists.

Acknowledgements

We would like to thank Prof. Stephen Burgess’s help and advice on this review.

This work is supported by National Key Research and Development Program of China (No. 2024YFE0213800), the German Research Foundation (DFG Ta434/8-1, CRC/ TR 412 and SFB1382, Project-ID 403224013). Lanlan Chen has been supported by the China Scholarship Council (2024.09 ~ 2027.09). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conflicts of Interest

FT’s lab has received research grants (funding to the institution) from AstraZeneca, MSD, Gilead. FT has received honoraria for consulting or lectures from Gilead, Abbvie, Falk, AstraZeneca, Boehringer, MSD, GSK, Ipsen, Pfizer, Novartis, Novo Nordisk, Sanofi. The other authors disclose no potential conflict of interest.

Figure 1.

The number of publications and citations in studies reporting Mendelian randomization (MR) from 2020 to 2024. (A) Displays all MR-related publications and citations, and (B) displays only those related to liver diseases. The keywords are listed at the bottom of figure.

Figure 2.

The illustration of Mendelian randomization (MR) principles and its comparison with randomized controlled trial (RCT).

Figure 3.

Illustration of key risk factors identified for MASLD. CHIP, clonal hematopoiesis of intermediate potential; MASLD, metabolic dysfunction-associated steatotic liver disease.

Figure 4.

An illustration of performing the state-of-the-art Mendelian randomization study. GWAS, genome-wide association study; IV, instrumental variable; IVW, inverse variance weighted method.

Table 1.

A glossary of key terms in Mendelian randomization for hepatologists

Table 1.
Instrumental variable (IV) assumptions
Relevance	Genetic variants should be closely associated with the exposure.
Independence	There is no common confounder between the genetic variants and the outcome
Exclusion restriction	Genetic variants can be linked to the outcome only via the way of exposure without different mechanisms (no horizontal pleiotropy)
Concepts related to genetics
Genetic variants	Differences in the DNA sequence among individuals, including single nucleotide variation and structural variations. Some variants are harmless, while others may influence traits or disease susceptibility.
SNP/SNV	SNP refers to a common genetic variant (usually with a minor allele frequency >0.01) where a single nucleotide is replaced with another. Example: A→G.
SNP/SNV	SNV (single nucleotide variant) refers to a broader term referring to any single nucleotide change, whether rare or common.
Genotype/phenotype	Genotype refers to an organism’s genetic makeup—the specific set of genes inherited from its parents, consisting of DNA sequences.
Genotype/phenotype	Phenotype is the observable physical, biochemical, and behavioral traits of an organism, which result from the interaction of the genotype with the environment.
CNV	CNV is a type of structural variant where a subset of DNA sequences are duplicated (gains) or deleted (losses). It can affect gene dosage, leading to changes in traits or disease risk.
QTL	It is a specific region in the genome that is associated with variation in a quantitative trait (e.g., serum lipids, fasting glucose)
eQTL	It is a type of QTL that affects gene expression levels rather than a physical trait. eQTLs can be cis-eQTLs (affecting genes nearby) or trans-eQTLs (affecting distant genes).
pQTL	It is a genetic locus associated with the variation in protein abundance. They are crucial for studying disease mechanisms and identifying drug targets. Similar to eQTL, there are cis-pQTL and trans-pQTL.
Pleiotropy	Pleiotropy refers to a single genetic variant influencing multiple, seemingly unrelated traits, e.g., SNVs in the TM6SF2 region affect the risk of liver steatosis and diabetes.
Vertical pleiotropy	It refers to a genetic variant influencing multiple traits through a shared causal pathway, e.g., a gene affecting cholesterol levels might also influence the risk of liver diseases due to its effect on cholesterol. “Vertical pleiotropy” is not problematic in MR.
Horizontal pleiotropy	It refers to a genetic variant influencing multiple traits through independent pathways, e.g., a gene may independently affect the metabolism of both glucose and lipids. “Horizontal pleiotropy” is problematic in MR analysis and should be avoided.
LD	LD refers to the non-random association of alleles at different loci in the genome. If two genetic variants are physically close on a chromosome, they are often inherited together.
Clumping/pruning	It refers to techniques to reduce redundancy in genetic analyses by removing variants that are highly correlated (in LD) and selecting one representative SNP per LD block.
GRS	It refers to a score that summarizes the cumulative effect of multiple genetic variants on a trait or disease risk. It can be divided into “unweighted GRS”, which counts the number of risk alleles, and “weighted GRS”, which accounts for the effect size of each allele from studies (e.g., GWAS). It is used to predict an individual’s genetic predisposition to diseases.
Population stratification	It refers to the systematic genetic differences between populations due to ancestry, which can lead to confounding in genetic studies and should should be accounted for using statistical methods like principal component analysis.
Genomic control	It refers to a statistical method correcting for population stratification or inflation in test statistics in genetic studies, e.g., adjusting the chi-square using a genomic inflation factor (lambda).
Concepts related to statistics
IV	IV refers to a variable used in causal inference that is associated with the exposure of interest but not directly with the outcome, except through the exposure. In MR analysis, genetic variants serve as IVs to study causal relationships between a risk factor and an outcome.
Weak instrument	It refers to IVs that are weakly associated with the exposure, leading to unreliable causal estimates.
F-statistic	It is a statistic used to assess the strength of an IV. According to the “Rule of Thumb”, an F-statistic >10 indicates a sufficiently strong instrument; lower values suggest weak instruments.
R² (coefficient of determination)	It represents the proportion of variance in the exposure explained by the IV. Higher R² indicates stronger instruments, improving the reliability of causal estimates.
Individual-level data	It refers to data that includes detailed genetic and phenotypic information for each individual in a study, e.g., data from cohort studies or biobanks like UK Biobank. In MR analysis, individual-level data allow more flexible analyses, such as non-linear MR or factorial MR.
Summary-level data	It refers to aggregated results, such as SNP-exposure and SNP-outcome associations from public GWAS summary statistics, rather than raw individual data. In MR analysis, it facilitates two-sample MR and meta-analyses without accessing individual-level data.
Two-sample MR	An MR approach using two independent datasets—one for SNP-exposure associations and another for SNP-outcome associations. No need for individual-level data; enables larger sample sizes.
One-sample MR	An MR approach using a single dataset with both SNP-exposure and SNP-outcome data. It allows robust sensitivity analyses but requires individual-level data.
Bidirectional MR	An MR approach to test causality in both directions (e.g., does X cause Y, or does Y cause X?). An example is examining whether serum cholesterol influences MASLD risk and vice versa.
Mediation MR	An MR approach to determine whether the causal effect of an exposure on an outcome is mediated through an intermediate variable, e.g., testing whether cholesterol mediates the effect of diet on liver diseases.
Multivariable MR	An MR approach includes multiple exposures simultaneously to account for confounding or to identify independent causal effects, e.g., analyzing the effects of serum HDL-C and triglycerides on the risk of liver diseases.
Factorial MR	An MR approach examines the combined effects of multiple exposures or interventions using genetic proxies for each exposure, e.g., studying the combined effects of lipid-modifying drugs and blood glucose-lowering drugs on liver diseases.
Nonlinear MR	An MR approach examines whether the relationship between exposure and outcome varies across the range of the exposure, e.g., investigating whether the effect of serum triglycerides on cholelithiasis differs at different ranges of triglycerides.
Drug-target MR	An MR approach uses genetic variants in drug target genes as instruments to predict the potential effects of drug interventions, e.g., testing the effect of HMGCR inhibition (via genetic instruments) on cholesterol levels and liver disease.
Colocalization	A statistical method to determine whether the same genetic variant affects both the exposure and outcome via a shared causal mechanism. It helps distinguish true causal effects from linkage artifacts.
Directionality test (MR-Steiger test)	It tests whether the causal relationship flows in the hypothesized direction (e.g., exposure→outcome) rather than the reverse.
Power analysis	It evaluates the ability of a study to detect a true causal effect based on sample size, genetic instrument strength, and effect size, which ensures the study design is adequately powered to avoid false negatives.
Winner’s curse	It refers to the overestimation of the effect size in SNPs that are strongly associated with traits in the discovery sample but weaker in replication studies, which can reduce the reliability of causal inference.
Collider bias	It refers to bias arising when both the exposure and outcome influence a third variable (collider), leading to spurious associations, e.g., adjusting for a common out-come of two variables (e.g., BMI and liver diseases) can create false associations between the two.
Concepts related to hepatology
GCKR (rs1260326)	A genetic variant strongly associated with MASLD and T2D.
PNPLA3 (rs738409)	A genetic variant strongly associated with MASLD, liver fibrosis, and HCC.
TM6SF2 (rs58542926)	A gene variant involved in hepatic lipid metabolism, influencing fatty liver disease risk and fibrosis progression.

BMI, body mass index; CNV, copy number variant; eQTL, expression quantitative trait locus; GCKR, glucokinase regulator; GRS, genetic risk score; GWAS, genome-wide association study; HCC, hepatocellular carcinoma; HMGCR, 3-hydroxy-3-methylglutaryl-CoA reductase; IV, instrumental variable; LD, linkage disequilibrium; MASLD, metabolic dysfunction-associated steatotic liver disease; MR, Mendelian randomization; PNPLA3, patatin like phospholipase domain containing 3; pQTL, protein quantitative trait locus; QTL, quantitative trait locus; SNP, single nucleotide polymorphism; SNV, single nucleotide variant; T2D, type 2 diabetes; TM6SF2, transmembrane 6 superfamily member 2.

Table 2.

The summary of Mendelian randomization results in the field of hepatology

Table 2.
Liver diseases	Risk factors identified by MR	Outcomes associated with genetic liability	References
MASLD	Metabolic factors: higher BMI, higher waist circumference, insulin resistance, genetic liability to T2D, genetic liability to hypothyroidism, higher SBP, higher liver enzymes, higher triglycerides, lower HDL-C, higher serum ferritin/iron, higher homocysteine, lower serum folate.	Increased risk of T2D, obesity, CAD; higher serum uric acid	[35-56]
MASLD	Non-metabolic factors: smoking, alcohol consumption, higher serum selenium, genetic liability to CHIP, air pollution-associated metabolites.	Increased risk of T2D, obesity, CAD; higher serum uric acid	[35-56]
ALD	Alcohol use, gut dysbiosis, higher serum interleukin 7, lower circulating sphingomyelins.	Lower bone mineral density in the femoral neck.	[62-66]
HBV	-	Cervical cancer, gastric cancer, osteoporosis, rheumatoid arthritis.	[68,70-71]
HCV	-	Diffuse large B-cell lymphoma	[72]
HCC	Alcohol consumption, lower neutrophil counts, lower serum vitamin C, higher serum adiponectin, higher visceral/ectopic adiposity (adjusted by BMI), genetic liability to IBD, genetic liability to insomnia, genetic liability to daytime napping.	-	[73-84]
CCA	Genetic liability to MASLD, cholelithiasis, insomnia, IBD; protective effect from gut microbiota (family porphyromonadaceae, genus bacteroidetes and phylum verrucomicrobia)	-	[78,83,85-87]
Cholelithiasis (cholecystitis)	Metabolic factors: obesity (higher BMI, higher waist circumference), genetic liability to T2D, lower circulating omega-3 fatty acid concentrations, lower serum sphingomyelin levels, lower serum campesterol levels; lower serum total cholesterol, higher serum triglycerides.	Increased risk of GERD, CCA, colorectal cancer, kidney cancer.	[90-100]
Cholelithiasis (cholecystitis)	Non-metabolic factors: increased leisure screen time; lower moderate-to-vigorous intensity physical activity; smoking.		[90-100]
PBC^*	Autoimmune conditions: genetic liability to IBD, SLE.	Increased risk of IBD, celiac disease, rheumatoid arthritis, thyroid dysfunction, osteoporosis; decreased risk of gastric cancer	[101-112]
PBC^*	Other factors: obesity, low serum hemoglobin, lower serum vitamin D, higher serum testosterone (females), lower sex hormone binding globulin.		[101-112]
PSC^*	Genetic liability to celiac disease, IBD, SLE.	Increased risk of celiac disease, IBD.	[102,108,113]

ALD, alcohol-related liver disease; BMI, body mass index; CAD, coronary artery disease; CHIP, clonal hematopoiesis of indeterminate potential; GERD, gastroesophageal reflux disease; HBV, hepatitis virus B; HCC, hepatocellular carcinoma; HCV, hepatitis virus C; HDL-C, high-density lipoprotein cholesterol; MASLD, metabolic dysfunction-associated liver disease; MR, Mendelian randomization; PBC, primary biliary cholangitis; PSC, primary sclerosing cholangitis; SBP, systolic blood pressure; SLE, systemic lupus erythematosus; T2D, type 2 diabetes.

^*For PBC and PSC, there is strong genetic overlap in autoimmune conditions, caution is needed when interpreting MR results.

Table 3.

Explanation and comparison of selected MASLD-related GWAS with summary statistics available

Table 3.
Author (yr)	Definition	Ethnicity	Sample size	Recommendation^*	GCST ID	PMID
Namjou et al. (2019)	MASLD from health records	Europeans	1,106 cases; 8,571 controls.	3 (low)	GCST008468	31311600
Anstee et al. [57] (2020)	MASLD from live biopsy	Europeans	1,483 cases; 17,781 controls.	1 (high)	GCST90011885	32298765
Fairfield et al. (2021)	MASLD from health records (UKB)	Europeans	4,761 cases; 373,227 controls.	3 (low)	GCST90054782	34535985
Ghodsian et al. [58] (2021)	MASLD from health records (GWAS meta-analysis of 4 cohorts)	Europeans	8,434 cases; 770,180 controls.	1 (high)	GCST90091033	34841290
Miao et al. (2021)	MASLD imputed from serum and anthropometric traits	Europeans	28,396 cases; 108,652 controls.	3 (low)	GCST90094908	35047847
Vujkovic et al. [59] (2022)	MASLD proxied by cALT	Mixed	Europeans: 95,472 cases; 68,725 controls.	2 (medium)	GCST90129601	35654975
			African Americans: 23,977 cases; 13,387 controls.
			Hispanics: 7,650 cases; 7,468 controls.
			Asians: 1,088 cases; 828 controls.
Sveinbjornsson et al. (2022)	MASLD from health records	Europeans	9,491 cases and 876,210 controls	1 (high)	NA	36280732
Chen et al. [60] (2023)	MASLD from CT-imaging and health records (GWAS meta-analysis of 5 cohorts)	Mixed	Imaging samples (n=66,814) and diagnostic code (3,584 cases, 621,081 controls).	2 (medium)	GCST90271622	37709864

cALT, chronic elevation of alanine aminotransferase; CT, computed tomography; GCST ID, the GCST number of each MASLD GWAS indexed in GWAS Catalog (https://www.ebi.ac.uk/gwas/); GWAS, genome-wide association study; MASLD, metabolic dysfunction-associated steatotic liver disease; UKB, UK Biobank.

^*Recommendation: “1 (high)” means this MASLD GWAS is highly recommended to be used in Mendelian randomization analysis of MASLD due to the large sample size and clear definition; “2 (medium)” means this MASLD GWAS has some limitations of mixed ancestries or unclear definitions of MASLD despite the large sample size. “3 (low)” means this MASLD GWAS is not suitable in Mendelian randomization analysis as it only represents a subset of existing GWAS datasets.

Abbreviations

ALD

alcohol-related liver disease

BMI

body mass index

CAD

coronary artery disease

cALT

chronic elevation of alanine aminotransferase

CHIP

clonal hematopoiesis of indeterminate potential

CNV

copy number variant

computed tomography

eQTL

expression quantitative trait locus

GCKR

glucokinase regulator

GCST ID

the GCST number of each MASLD GWAS indexed in GWAS Catalog (https://www.ebi.ac.uk/gwas/)

GERD

gastroesophageal reflux disease

GRS

genetic risk score

GWAS

genome-wide association study

HBV

hepatitis virus B

HCC

hepatocellular carcinoma

HCV

hepatitis virus C

HDL-C

high-density lipoprotein cholesterol

HMGCR

3-hydroxy-3-methylglutaryl-CoA reductase

instrumental variable

linkage disequilibrium

MASLD

metabolic dysfunction-associated steatotic liver disease

Mendelian randomization

PBC

primary biliary cholangitis

PNPLA3

patatin like phospholipase domain containing 3

pQTL

protein quantitative trait locus

PSC

primary sclerosing cholangitis

QTL

quantitative trait locus

SBP

systolic blood pressure

SLE

systemic lupus erythematosus

SNP

single nucleotide polymorphism

SNV

single nucleotide variant

T2D

type 2 diabetes

TM6SF2

transmembrane 6 superfamily member 2

UKB

UK biobank

REFERENCES

1. Karlsen TH, Sheron N, Zelber-Sagi S, Carrieri P, Dusheiko G, Bugianesi E, et al. The EASL-Lancet Liver Commission: protecting the next generation of Europeans against liver disease complications and premature mortality. Lancet 2022;399:61-116.
Article
PubMed
2. Smith GD, Ebrahim S. ‘Mendelian randomization’: can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol 2003;32:1-22.
Article
PubMed
3. Voight BF, Peloso GM, Orho-Melander M, Frikke-Schmidt R, Barbalic M, Jensen MK, et al. Plasma HDL cholesterol and risk of myocardial infarction: a mendelian randomisation study. Lancet 2012;380:572-580.
PubMed
PMC
4. Song Y, Ye T, Roberts LR, Larson NB, Winham SJ. Mendelian randomization in hepatology: a review of principles, opportunities, and challenges. Hepatology 2025;81:1836-1846.
Article
PubMed
5. Davies NM, Holmes MV, Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. Bmj 2018;362:k601.
Article
PubMed
PMC
6. Burgess S, Butterworth A, Malarstig A, Thompson SG. Use of Mendelian randomisation to assess potential benefit of clinical intervention. Bmj 2012;345:e7325.
Article
PubMed
7. Skrivankova VW, Richmond RC, Woolf BAR, Davies NM, Swanson SA, VanderWeele TJ, et al. Strengthening the reporting of observational studies in epidemiology using mendelian randomisation (STROBE-MR): explanation and elaboration. Bmj 2021;375:n2233.
Article
PubMed
PMC
8. Sanderson E, Glymour MM, Holmes MV, Kang H, Morrison J, Munafò MR, et al. Mendelian randomization. Nat Rev Methods Primers 2022;2:6.
Article
PubMed
PMC
PDF
9. Lawlor DA, Harbord RM, Sterne JA, Timpson N, Davey Smith G. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Stat Med 2008;27:1133-1163.
Article
PubMed
10. Burgess S, Thompson SG. Mendelian randomization: methods for causal inference using genetic variants. CRC Press; 2021.
11. Zuber V, Grinberg NF, Gill D, Manipur I, Slob EAW, Patel A, et al. Combining evidence from Mendelian randomization and colocalization: review and comparison of approaches. Am J Hum Genet 2022;109:767-782.
Article
PubMed
PMC
12. Little J, Khoury MJ. Mendelian randomisation: a new spin or real progress? Lancet 2003;362:930-931.
Article
PubMed
13. Timpson NJ, Lawlor DA, Harbord RM, Gaunt TR, Day IN, Palmer LJ, et al. C-reactive protein and its role in metabolic syndrome: mendelian randomisation study. Lancet 2005;366:1954-1959.
Article
PubMed
14. Katan MB. Apolipoprotein E isoforms, serum cholesterol, and cancer. Lancet 1986;1:507-508.
Article
15. Skrivankova VW, Richmond RC, Woolf BAR, Yarmolinsky J, Davies NM, Swanson SA, et al. Strengthening the reporting of observational studies in epidemiology using Mendelian randomization: the STROBE-MR statement. Jama 2021;326:1614-1621.
Article
PubMed
16. Chen L, Burgess S, Luo S, Lv G. First release of Mendelian randomisation book in Chinese. eGastroenterology 2023;1:e100043.
Article
PubMed
PMC
17. Burgess S, Thompson SG. Mendelian randomization: methods for using genetic variants in causal estimation. CRC Press; 2015.
18. Burgess S, Butterworth A, Thompson SG. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol 2013;37:658-665.
Article
PubMed
PMC
PDF
19. Burgess S, Thompson SG. Interpreting findings from Mendelian randomization using the MR-Egger method. Eur J Epidemiol 2017;32:377-389.
Article
PubMed
PMC
PDF
20. Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 2016;40:304-314.
Article
PubMed
PMC
PDF
21. Hartwig FP, Davey Smith G, Bowden J. Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. Int J Epidemiol 2017;46:1985-1998.
Article
PubMed
PMC
22. Verbanck M, Chen CY, Neale B, Do R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 2018;50:693-698.
Article
PubMed
PMC
PDF
23. Morrison J, Knoblauch N, Marcus JH, Stephens M, He X. Mendelian randomization accounting for correlated and uncorrelated pleiotropic effects using genome-wide summary statistics. Nat Genet 2020;52:740-747.
Article
PubMed
PMC
PDF
24. Zhao Q, Wang J, Hemani G, Bowden J, Small DS. Statistical inference in two-sample summary-data mendelian randomization using robust adjusted profile score. arXiv preprint arXiv 1801.09652 2018.
Article
25. Mitchell RE, Hartley AE, Walker VM, Gkatzionis A, Yarmolinsky J, Bell JA, et al. Strategies to investigate and mitigate collider bias in genetic and Mendelian randomisation studies of disease progression. PLoS Genet 2023;19:e1010596.
Article
PubMed
PMC
26. Clayton GL, Gonçalves A, Goulding N, Borges MC, Holmes MV, et al. A framework for assessing selection and misclassification bias in mendelian randomisation studies: an illustrative example between body mass index and covid-19. Bmj 2023;381:e072148.
Article
PubMed
27. Dudbridge F, Allen RJ, Sheehan NA, Schmidt AF, Lee JC, Jenkins RG, et al. Adjustment for index event bias in genome-wide association studies of subsequent events. Nat Commun 2019;10:1561.
Article
PubMed
PMC
PDF
28. Mahmoud O, Dudbridge F, Davey Smith G, Munafo M, Tilling K. A robust method for collider bias correction in conditional genome-wide association studies. Nat Commun 2022;13:619.
Article
PubMed
PMC
PDF
29. Zuber V, Colijn JM, Klaver C, Burgess S. Selecting likely causal risk factors from high-throughput experiments using multivariable Mendelian randomization. Nat Commun 2020;11:29.
Article
PubMed
PMC
PDF
30. Hu X, Cai M, Xiao J, Wan X, Wang Z, Zhao H, et al. Benchmarking Mendelian randomization methods for causal inference using genome-wide association study summary statistics. Am J Hum Genet 2024;111:1717-1735.
Article
PubMed
PMC
31. Yao M, Miller GW, Vardarajan BN, Baccarelli AA, Guo Z, Liu Z. Deciphering proteins in Alzheimer’s disease: A new Mendelian randomization method integrated with AlphaFold3 for 3D structure prediction. Cell Genom 2024;4:100700.
Article
PubMed
PMC
32. Haglund A, Zuber V, Abouzeid M, Yang Y, Ko JH, Wiemann L, et al. Cell state-dependent allelic effects and contextual Mendelian randomization analysis for human brain phenotypes. Nat Genet 2025;57:358-368.
Article
PubMed
PMC
PDF
33. Rinella ME, Lazarus JV, Ratziu V, Francque SM, Sanyal AJ, Kanwal F, et al. A multisociety Delphi consensus statement on new fatty liver disease nomenclature. J Hepatol 2023;79:1542-1556.
PubMed
34. Eslam M, Newsome PN, Sarin SK, Anstee QM, Targher G, Romero-Gomez M, et al. A new definition for metabolic dysfunction-associated fatty liver disease: An international expert consensus statement. J Hepatol 2020;73:202-209.
PubMed
35. Yuan S, Chen J, Li X, Fan R, Arsenault B, Gill D, et al. Lifestyle and metabolic factors for nonalcoholic fatty liver disease: Mendelian randomization study. Eur J Epidemiol 2022;37:723-733.
Article
PubMed
PMC
PDF
36. Xie J, Huang H, Liu Z, Li Y, Yu C, Xu L, et al. The associations between modifiable risk factors and nonalcoholic fatty liver disease: a comprehensive Mendelian randomization study. Hepatology 2023;77:949-964.
Article
PubMed
PDF
37. Liu Z, Zhang Y, Graham S, Wang X, Cai D, Huang M, et al. Causal relationships between NAFLD, T2D and obesity have implications for disease subphenotyping. J Hepatol 2020;73:263-276.
Article
PubMed
PMC
38. Yuan M, He J, Hu X, Yao L, Chen P, Wang Z, et al. Hypertension and NAFLD risk: Insights from the NHANES 2017-2018 and Mendelian randomization analyses. Chin Med J (Engl) 2024;137:457-464.
Article
PubMed
39. Liu C, Chen Y, Zhang Z, Xie J, Yu C, Xu L, et al. Iron status and NAFLD among European populations: a bidirectional two-sample Mendelian randomization study. Nutrients 2022;14.
Article
40. De Silva NMG, Borges MC, Hingorani AD, Engmann J, Shah T, Zhang X, et al. Liver function and risk of type 2 diabetes: bidirectional Mendelian randomization study. Diabetes 2019;68:1681-1691.
Article
PubMed
PMC
PDF
41. European Association for the Study of the Liver (EASL); European Association for the Study of Diabetes (EASD); European Association for the Study of Obesity (EASO). EASL-EASD-EASO Clinical Practice Guidelines on the management of metabolic dysfunction-associated steatotic liver disease (MASLD). J Hepatol 2024;81:492-542.
Article
PubMed
PDF
42. Yuan S, Chen J, Dan L, Xie Y, Sun Y, Li X, et al. Homocysteine, folate, and nonalcoholic fatty liver disease: a systematic review with meta-analysis and Mendelian randomization investigation. Am J Clin Nutr 2022;116:1595-1609.
Article
PubMed
PDF
43. Wong WJ, Emdin C, Bick AG, Zekavat SM, Niroula A, Pirruccello JP, et al. Clonal haematopoiesis and risk of chronic liver disease. Nature 2023;616:747-754.
PubMed
PMC
44. Chen L, Fan Z, Sun X, Qiu W, Mu W, Chai K, et al. Diet-derived antioxidants and nonalcoholic fatty liver disease: a Mendelian randomization study. Hepatol Int 2023;17:326-338.
Article
PubMed
PDF
45. Ran S, Zhang J, Tian F, Qian ZM, Wei S, Wang Y, et al. Association of metabolic signatures of air pollution with MASLD: observational and Mendelian randomization study. J Hepatol 2025;82:560-570.
Article
PubMed
46. Yuan S, Larsson SC. Inverse association between serum 25-hydroxyvitamin D and nonalcoholic fatty liver disease. Clin Gastroenterol Hepatol 2023;21:398-405.e394.
Article
PubMed
47. Wang N, Chen C, Zhao L, Chen Y, Han B, Xia F, et al. Vitamin D and nonalcoholic fatty liver disease: bi-directional Mendelian randomization analysis. EBioMedicine 2018;28:187-193.
Article
PubMed
PMC
48. Sheng Q, Shi H, Liu S, Zhuang L, Zhao Z, Xin Y. Serum 25-hydroxyvitamin D levels and the risk of non-alcoholic fatty liver: a two-sample Mendelian randomization study. Saudi J Gastroenterol 2023;29:39-46.
PubMed
49. Sookoian S, Rotman Y, Valenti L. Genetics of metabolic dysfunction-associated steatotic liver disease: the state of the art update. Clin Gastroenterol Hepatol 2024;22:2177-2187.e2173.
Article
PubMed
50. Au Yeung SL, Borges MC, Wong THT, Lawlor DA, Schooling CM. Evaluating the role of non-alcoholic fatty liver disease in cardiovascular diseases and type 2 diabetes: a Mendelian randomization study in Europeans and East Asians. Int J Epidemiol 2023;52:921-931.
Article
PubMed
PDF
51. Ahmed A, Cule M, Bell JD, Sattar N, Yaghootkar H. Differing genetic variants associated with liver fat and their contrasting relationships with cardiovascular diseases and cancer. J Hepatol 2024;81:921-929.
Article
PubMed
52. Ren Z, Simons P, Wesselius A, Stehouwer CDA, Brouwers M. Relationship between NAFLD and coronary artery disease: a Mendelian randomization study. Hepatology 2023;77:230-238.
Article
PubMed
PDF
53. Jamialahmadi O, De Vincentis A, Tavaglione F, Malvestiti F, Li-Gao R, Mancina RM, et al. Partitioned polygenic risk scores identify distinct types of metabolic dysfunction-associated steatotic liver disease. Nat Med 2024;30:3614-3623.
Article
PubMed
PMC
PDF
54. Raverdy V, Tavaglione F, Chatelain E, Lassailly G, De Vincentis A, Vespasiani-Gentilucci U, et al. Data-driven cluster analysis identifies distinct types of metabolic dysfunction-associated steatotic liver disease. Nat Med 2024;30:3624-3633.
Article
PubMed
PMC
PDF
55. Li S, Fu Y, Liu Y, Zhang X, Li H, Tian L, et al. Serum uric acid levels and nonalcoholic fatty liver disease: a 2-sample bidirectional Mendelian randomization study. J Clin Endocrinol Metab 2022;107:e3497-e3503.
Article
PubMed
PDF
56. Tang Y, Xu Y, Liu P, Liu C, Zhong R, Yu X, et al. No evidence for a causal link between serum uric acid and nonalcoholic fatty liver disease from the dongfeng-tongji cohort study. Oxid Med Cell Longev 2022;2022:6687626.
Article
PubMed
PMC
PDF
57. Anstee QM, Darlay R, Cockell S, Meroni M, Govaere O, Tiniakos D, et al. Genome-wide association study of non-alcoholic fatty liver and steatohepatitis in a histologically characterised cohort(☆). J Hepatol 2020;73:505-515.
Article
PubMed
58. Ghodsian N, Abner E, Emdin CA, Gobeil É, Taba N, Haas ME, et al. Electronic health record-based genome-wide meta-analysis provides insights on the genetic architecture of nonalcoholic fatty liver disease. Cell Rep Med 2021;2:100437.
Article
PubMed
PMC
59. Vujkovic M, Ramdas S, Lorenz KM, Guo X, Darlay R, Cordell HJ, et al. A multiancestry genome-wide association study of unexplained chronic ALT elevation as a proxy for nonalcoholic fatty liver disease with histological and radiological validation. Nat Genet 2022;54:761-771.
PubMed
PMC
60. Chen Y, Du X, Kuppa A, Feitosa MF, Bielak LF, O’Connell JR, et al. Genome-wide association meta-analysis identifies 17 loci associated with nonalcoholic fatty liver disease. Nat Genet 2023;55:1640-1650.
Article
PubMed
PMC
PDF
61. Fu Y, Maccioni L, Wang XW, Greten TF, Gao B. Alcohol-associated liver cancer. Hepatology 2024;80:1462-1479.
Article
PubMed
62. Li S, Zhou C, Liu T, Zhang L, Liu S, Zhao Q, et al. Causal relationships between the gut microbiota, inflammatory cytokines, and alcoholic liver disease: a Mendelian randomization analysis. Front Endocrinol (Lausanne) 2024;15:1442603.
Article
PubMed
PMC
63. Wu D, Hao O, Hu W, Wu Z, Bian L, Wang H, et al. Circulating cytokines and alcoholic liver disease: a two-sample bidirectional Mendelian randomization study. Scand J Gastroenterol 2024;59:325-332.
Article
PubMed
64. Chen Z, Ding C, Chen K, Lu C, Li Q. Exploring the impact of inflammatory cytokines on alcoholic liver disease: a Mendelian randomization study with bioinformatics insights into potential biological mechanisms. Am J Drug Alcohol Abuse 2024;50:643-658.
Article
PubMed
65. Thiele M, Suvitaival T, Trošt K, Kim M, de Zawadzki A, Kjaergaard M, et al. Sphingolipids are depleted in alcohol-related liver fibrosis. Gastroenterology 2023;164:1248-1260.
Article
PubMed
66. Huang Q, Guo J, Zhao H, Zheng Y, Zhang Y. The associations of alcoholic liver disease and nonalcoholic fatty liver disease with bone mineral density and the mediation of serum 25-Hydroxyvitamin D: A bidirectional and two-step Mendelian randomization. PLoS One 2023;18:e0292881.
Article
PubMed
PMC
67. Atkinson SR, Way MJ, McQuillin A, Morgan MY, Thursz MR. Homozygosity for rs738409:G in PNPLA3 is associated with increased mortality following an episode of severe alcoholic hepatitis. J Hepatol 2017;67:120-127.
Article
PubMed
68. Kamiza AB, Fatumo S, Singini MG, Yeh CC, Chikowore T. Hepatitis B infection is causally associated with extrahepatic cancers: a Mendelian randomization study. EBioMedicine 2022;79:104003.
Article
PubMed
PMC
69. Kim YJ, Kim HY, Lee JH, Yu SJ, Yoon JH, Lee HS, et al. A genome-wide association study identified new variants associated with the risk of chronic hepatitis B. Hum Mol Genet 2013;22:4233-4238.
Article
PubMed
70. Li Z, Cao J, Li K, Wu Y, Luo Z, Cao R, et al. Causal associations between osteoporosis and HBV infection across Asian and European populations: evidence from Mendelian randomization and colocalization analysis. Front Endocrinol (Lausanne) 2024;15:1419303.
Article
PubMed
71. Zhao N, Zhang X, Wang X, Liu Z, Zheng G, Zhang X, et al. Hidden liver-joint axis: HBV infection causes rheumatoid arthritis via TRAFD1 with imbalance of HBV X protein and transferulic acid. Virulence 2024;15:2422540.
Article
PubMed
PMC
72. Fu L, Yu J, Chen Z, Gao F, Zhang Z, Fu J, et al. Shared genetic factors and causal association between chronic hepatitis C infection and diffuse large B cell lymphoma. Infect Agent Cancer 2024;19:15.
Article
PubMed
PMC
PDF
73. Brouillet A, Lafdil F. Risk factors of primary liver cancer initiation associated with tumour initiating cell emergence: novel targets for promising preventive therapies. eGastroenterology 2023;1:e100010.
Article
PubMed
PMC
74. Cabibbo G, Celsa C, Rimassa L, Torres F, Rimola J, Kloeckner R, et al. Navigating the landscape of liver cancer management: Study designs in clinical trials and clinical practice. J Hepatol 2024;80:957-966.
Article
PubMed
75. Deng Y, Huang J, Wong MCS. Associations between six dietary habits and risk of hepatocellular carcinoma: a Mendelian randomization study. Hepatol Commun 2022;6:2147-2154.
Article
PubMed
PMC
PDF
76. Nie D, He X, Zheng H, Deng D, He F, Li R, et al. Association between green tea intake and digestive system cancer risk in European and East Asian populations: a Mendelian randomization study. Eur J Nutr 2024;63:1103-1111.
Article
PubMed
PDF
77. Liu Z, Song C, Suo C, Fan H, Zhang T, Jin L, et al. Alcohol consumption and hepatocellular carcinoma: novel insights from a prospective cohort study and nonlinear Mendelian randomization analysis. BMC Med 2022;20:413.
Article
PubMed
PMC
PDF
78. Yang X, Wang J, Wang H. Association between sleep traits and primary liver cancer: a Mendelian randomization analysis. Eur J Clin Invest 2023;53:e14002.
Article
PubMed
79. Pan GQ, Yang CC, Shang XL, Dong ZR, Li T. The causal relationship between white blood cell counts and hepatocellular carcinoma: a Mendelian randomization study. Eur J Med Res 2022;27:278.
Article
PubMed
PMC
PDF
80. Larsson SC, Mason AM, Vithayathil M, Carter P, Kar S, Zheng JS, et al. Circulating vitamin C and digestive system cancers: Mendelian randomization study. Clin Nutr 2022;41:2031-2035.
Article
PubMed
PMC
81. Jiang H, Hu D, Wang J, Zhang B, He C, Ning J. Adiponectin and the risk of gastrointestinal cancers in East Asians: Mendelian randomization analysis. Cancer Med 2022;11:2397-2404.
Article
PubMed
PMC
PDF
82. Xu FQ, Xu QY, Zhu ZJ, Jin L, Ye TW, Du CF, et al. Visceral and ectopic fat are more predictively associated with primary liver cancer than overall obesity from genetic sights: A Mendelian randomization study. Int J Cancer 2024;154:530-537.
Article
PubMed
83. Huang J, Li X, Hong J, Huang L, Jiang Q, Guo S, et al. Inflammatory bowel disease increases the risk of hepatobiliary pancreatic cancer: a two-sample Mendelian randomization analysis of European and East Asian populations. Cancer Med 2023;12:13599-13609.
Article
PubMed
PMC
84. Zhang Y, Zhang Y, He P, Ge F, Huo Z, Qiao G. The genetic liability to rheumatoid arthritis may decrease hepatocellular carcinoma risk in East Asian population: a Mendelian randomization study. Arthritis Res Ther 2023;25:49.
Article
PubMed
PMC
PDF
85. Chen L, Fan Z, Sun X, Qiu W, Mu W, Chai K, et al. Examination on the risk factors of cholangiocarcinoma: a Mendelian randomization study. Front Pharmacol 2022;13:900424.
Article
PubMed
PMC
86. Ma J, Li J, Jin C, Yang J, Zheng C, Chen K, et al. Association of gut microbiome and primary liver cancer: a two-sample Mendelian randomization and case-control study. Liver Int 2023;43:221-233.
Article
PubMed
PDF
87. Su Q, Jin C, Bo Z, Yang Y, Wang J, Wang J, et al. Association between gut microbiota and gastrointestinal cancer: a two-sample bi-directional Mendelian randomization study. Front Microbiol 2023;14:1181328.
Article
PubMed
PMC
88. Yuan S, Larsson SC, Gill D, Burgess S. Concerns about instrumental variable selection for biological effect versus uptake of proton pump inhibitors in Mendelian randomisation analysis. Gut 2024;74:e6.
Article
PubMed
89. Burgess S, Cronjé HT. Incorporating biological and clinical insights into variant choice for Mendelian randomisation: examples and principles. eGastroenterology 2024;2:e100042.
Article
PubMed
PMC
90. Chen J, Ruan X, Fu T, Lu S, Gill D, He Z, et al. Sedentary lifestyle, physical activity, and gastrointestinal diseases: evidence from mendelian randomization analysis. EBioMedicine 2024;103:105110.
Article
PubMed
PMC
91. Yuan S, Chen J, Ruan X, Sun Y, Zhang K, Wang X, et al. Smoking, alcohol consumption, and 24 gastrointestinal diseases: Mendelian randomization analysis. Elife 2023;12.
Article
PDF
92. Zagkos L, Dib MJ, Pinto R, Gill D, Koskeridis F, Drenos F, et al. Associations of genetically predicted fatty acid levels across the phenome: A mendelian randomisation study. PLoS Med 2022;19:e1004141.
Article
PubMed
PMC
93. Mi J, Jiang Q, Qi Z, Liu Z, Bai X, Zheng X, et al. Plasma campesterol and ABCG5/ABCG8 gene loci on the risk of cholelithiasis and cholecystitis: evidence from Mendelian randomization and colocalization analyses. Hum Genomics 2024;18:19.
Article
PubMed
PMC
PDF
94. Lin H, Lu R, Shang Q, Gu Y, Liu Y, Yang Y, et al. Exploring the causal relationships between cholelithiasis, cholecystitis, cholecystectomy, and gastroesophageal reflux disease: a bidirectional two-sample Mendelian randomization study. Int J Surg 2025;111:932-940.
Article
PubMed
95. Chen L, Yang H, Li H, He C, Yang L, Lv G. Insights into modifiable risk factors of cholelithiasis: a Mendelian randomization study. Hepatology 2022;75:785-796.
Article
PubMed
PDF
96. Yuan S, Gill D, Giovannucci EL, Larsson SC. Obesity, type 2 diabetes, lifestyle factors, and risk of gallstone disease: a Mendelian randomization investigation. Clin Gastroenterol Hepatol 2022;20:e529-e537.
Article
PubMed
97. Chen L, Qiu W, Sun X, Gao M, Zhao Y, Li M, et al. Novel insights into causal effects of serum lipids and lipid-modifying targets on cholelithiasis. Gut 2024;73:521-532.
Article
PubMed
98. Mi J, Jiang L, Liu Z, Wu X, Zhao N, Wang Y, et al. Identification of blood metabolites linked to the risk of cholelithiasis: a comprehensive Mendelian randomization study. Hepatol Int 2022;16:1484-1493.
Article
PubMed
PDF
99. Chen L, Fan Z, Sun X, Qiu W, Mu W, Chai K, et al. Associations of cholecystectomy with the risk of colorectal cancer: a Mendelian randomization study. Chin Med J (Engl) 2023;136:840-847.
Article
PubMed
PMC
100. Kharazmi E, Scherer D, Boekstegers F, Liang Q, Sundquist K, Sundquist J, et al. Gallstones, cholecystectomy, and kidney cancer: observational and Mendelian randomization results based on large cohorts. Gastroenterology 2023;165:218-227.e218.
Article
PubMed
101. Zhang H, Chen L, Fan Z, Lv G. The causal effects of inflammatory bowel disease on primary biliary cholangitis: a bidirectional two-sample Mendelian randomization study. Liver Int 2023;43:1741-1748.
Article
PubMed
102. Huang W, Jin T, Zheng W, Yin Q, Yan Q, Pan H, et al. Identifying the genetic association between systemic lupus erythematosus and the risk of autoimmune liver diseases. J Autoimmun 2024;145:103188.
Article
PubMed
103. Kim MS, Song M, Kim S, Kim B, Kang W, Kim JY, et al. Causal effect of adiposity on the risk of 19 gastrointestinal diseases: a Mendelian randomization study. Obesity (Silver Spring) 2023;31:1436-1444.
Article
PubMed
PMC
104. Xu H, Wu Z, Feng F, Li Y, Zhang S. Low vitamin D concentrations and BMI are causal factors for primary biliary cholangitis: A mendelian randomization study. Front Immunol 2022;13:1055953.
Article
PubMed
PMC
105. Ke B, Li C, Shang H. Hematologic traits and primary biliary cholangitis: a Mendelian randomization study. J Hum Genet 2023;68:559-563.
Article
PubMed
PDF
106. Wu J, Fan X, Song Y. The causal effect of bioavailable testosterone on primary biliary cholangitis in female patients: a Bidirectional Mendelian randomization analysis. Dig Liver Dis 2023;55:1091-1097.
Article
PubMed
107. Huang W, Jiang R, Li S, Zeng R, Li Y, Zhang Y, et al. Investigating shared genetic architecture between inflammatory bowel diseases and primary biliary cholangitis. JHEP Rep 2024;6:101037.
Article
PubMed
PMC
108. Li T, Feng Y, Wang C, Shi T, Huang X, Abuduhadeer M, et al. Causal relationships between autoimmune diseases and celiac disease: a Mendelian randomization analysis. Biotechnol Genet Eng Rev 2024;40:4611-4626.
Article
PubMed
109. Fan J, Jiang T, He D. Genetic link between rheumatoid arthritis and autoimmune liver diseases: a two-sample Mendelian randomization study. Semin Arthritis Rheum 2023;58:152142.
Article
PubMed
110. Huang P, Hou Y, Zou Y, Ye X, Yu R, Yang S. The causal effects of primary biliary cholangitis on thyroid dysfunction: a two-sample Mendelian randomization study. Front Genet 2021;12:791778.
Article
PubMed
PMC
111. Wu Y, Qian Q, Liu Q, Wang R, Pu X, Li Y, et al. Osteoporosis and primary biliary cholangitis: a trans-ethnic Mendelian randomization analysis. Clin Rev Allergy Immunol 2024;66:138-148.
Article
PubMed
PDF
112. Gu D, Zhang M, Wang Y, Bai Y, Wang X, Deng G. Causal effect of autoimmune liver diseases on cancer: meta-analyses of cohort studies and Mendelian randomization study. Liver Int 2022;42:2216-2226.
Article
PubMed
PDF
113. Xie Y, Chen X, Deng M, Sun Y, Wang X, Chen J, et al. Causal linkage between inflammatory bowel disease and primary sclerosing cholangitis: a two-sample Mendelian RANDOMIZATION ANALYSIS. Front Genet 2021;12:649376.
Article
PubMed
PMC
114. Burgess S, Mason AM, Grant AJ, Slob EAW, Gkatzionis A, Zuber V, et al. Using genetic association data to guide drug discovery and development: review of methods and applications. Am J Hum Genet 2023;110:195-214.
Article
PubMed
PMC
115. Walker VM, Davey Smith G, Davies NM, Martin RM. Mendelian randomization: a novel approach for the prediction of adverse drug events and drug repurposing opportunities. Int J Epidemiol 2017;46:2078-2089.
Article
PubMed
PMC
116. Holmes MV, Richardson TG, Ference BA, Davies NM, Davey Smith G. Integrating genomics with biomarkers and therapeutic targets to invigorate cardiovascular drug development. Nat Rev Cardiol 2021;18:435-453.
Article
PubMed
PDF
117. Luo S, Zheng MH, Wong VW, Au Yeung SL. Drug-target Mendelian randomisation applied to metabolic dysfunction-associated steatotic liver disease: opportunities and challenges. eGastroenterology 2024;2:e100114.
Article
PubMed
PMC
118. Li Z, Zhang B, Liu Q, Tao Z, Ding L, Guo B, et al. Genetic association of lipids and lipid-lowering drug target genes with nonalcoholic fatty liver disease. EBioMedicine 2023;90:104543.
Article
PubMed
PMC
119. Dobbie LJ, Cuthbertson DJ, Hydes TJ, Alam U, Zhao SS. Mendelian randomisation reveals sodium-glucose cotransporter-1 inhibition’s potential in reducing non-alcoholic fatty liver disease risk. Eur J Endocrinol 2023;188:K33-k37.
Article
PubMed
PMC
PDF
120. Chung SW, Moon HS, Shin H, Han H, Park S, Cho H, et al. Inhibition of sodium-glucose cotransporter-2 and liver-related complications in individuals with diabetes: a Mendelian randomization and population-based cohort study. Hepatology 2024;80:633-648.
Article
PubMed
121. Yan R, Liu L, Tzoulaki I, Fan J, Targher G, Yuan Z, et al. Genetic evidence for GLP-1 and GIP receptors as targets for treatment and prevention of MASLD/MASH. Liver Int 2025;45:e16150.
PubMed
122. Loomba R, Hartman ML, Lawitz EJ, Vuppalanchi R, Boursier J, Bugianesi E, et al. Tirzepatide for metabolic dysfunction-associated steatohepatitis with liver fibrosis. N Engl J Med 2024;391:299-310.
Article
PubMed
123. Morrow MR, Batchuluun B, Wu J, Ahmadi E, Leroux JM, Mohammadi-Shemirani P, et al. Inhibition of ATP-citrate lyase improves NASH, liver fibrosis, and dyslipidemia. Cell Metab 2022;34:919-936.e918.
Article
PubMed
124. Luo P, Yu X. ENPP2/Autotaxin: The potential drug target for alcoholic liver disease identified through Mendelian randomization analysis. Liver Int 2024;44:1624-1633.
PubMed
125. Shi Y, Dong H, Sun S, Wu X, Fang J, Zhao J, et al. Proteincentric omics analysis reveals circulating complements linked to non-viral liver diseases as potential therapeutic targets. Clin Mol Hepatol 2024;30:80-97.
PubMed
126. Rezaee-Zavareh MS, Kim N, Yang JD. Starting the journey: understanding the roles of complement proteins in liver diseases through mendelian randomization. Clin Mol Hepatol 2024;30:150-153.
Article
PubMed
PMC
PDF
127. Yang H, Chen L, Liu Y. A large-scale plasma proteome Mendelian randomization study identifies novel causal plasma proteins related to primary biliary cholangitis. Front Immunol 2023;14:1052616.
Article
PubMed
PMC
128. Chen L, Zhao Y, Li M, Lv G. Proteome-wide Mendelian randomization highlights AIF1 and HLA-DQA2 as targets for primary sclerosing cholangitis. Hepatol Int 2024;18:517-528.
Article
PubMed
PDF
129. Wang Z, Lu J, Hu J. Association between antihypertensive drugs and hepatocellular carcinoma: a trans-ancestry and drug-target Mendelian randomization study. Liver Int 2023;43:1320-1331.
Article
PubMed
130. Giambartolomei C, Vukcevic D, Schadt EE, Franke L, Hingorani AD, Wallace C, et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS Genet 2014;10:e1004383.
Article
PubMed
PMC
131. Shen Y, Cheng L, Xu M, Wang W, Wan Z, Xiong H, et al. SGLT2 inhibitor empagliflozin downregulates miRNA-34a-5p and targets GREM2 to inactivate hepatic stellate cells and ameliorate non-alcoholic fatty liver disease-associated fibrosis. Metabolism 2023;146:155657.
Article
PubMed
132. Park HS, Jang JE, Ko MS, Woo SH, Kim BJ, Kim HS, et al. Statins increase mitochondrial and peroxisomal fatty acid oxidation in the liver and prevent non-alcoholic steatohepatitis in mice. Diabetes Metab J 2016;40:376-385.
Article
PubMed
PMC
PDF
133. Leyden GM, Sobczyk MK, Richardson TG, Gaunt TR. Distinct pathway-based effects of blood pressure and body mass index on cardiovascular traits: comparison of novel Mendelian randomization approaches. Genome Med 2025;17:54.
Article
PubMed
PMC
PDF
134. Gill D, Georgakis MK, Walker VM, Schmidt AF, Gkatzionis A, Freitag DF, et al. Mendelian randomization for studying the effects of perturbing drug targets. Wellcome Open Res 2021;6:16.
Article
PubMed
PMC
PDF
135. Stender S, Gellert-Kristensen H, Smith GD. Reclaiming mendelian randomization from the deluge of papers and misleading findings. Lipids Health Dis 2024;23:286.
Article
PubMed
PMC
PDF
136. Chen L, Guillot A, Schneider CV. Attention to the misuse of Mendelian randomisation in medical research. eGastroenterology 2025;3:e100187.
Article
PubMed
PMC
137. Kurki MI, Karjalainen J, Palta P, Sipilä TP, Kristiansson K, Donner KM, et al. FinnGen provides genetic insights from a well-phenotyped isolated population. Nature 2023;613:508-518.
PubMed
PMC
138. Garfield V, Salzmann A, Burgess S, Chaturvedi N. A guide for selection of genetic instruments in mendelian randomization studies of type 2 diabetes and HbA1c: toward an integrated approach. Diabetes 2023;72:175-183.
Article
PubMed
PDF
139. Burgess S, Davey Smith G, Davies NM, Dudbridge F, Gill D, Glymour MM, et al. Guidelines for performing Mendelian randomization investigations: update for summer 2023. Wellcome Open Res 2019;4:186.
Article
PubMed
PDF
140. Hemani G, Bowden J, Davey Smith G. Evaluating the potential role of pleiotropy in Mendelian randomization studies. Hum Mol Genet 2018;27:R195-r208.
Article
PubMed
PMC
141. Sanderson E, Richardson TG, Hemani G, Davey Smith G. The use of negative control outcomes in Mendelian randomization to detect potential population stratification. Int J Epidemiol 2021;50:1350-1361.
Article
PubMed
PMC
PDF
142. Hamilton FW, Hughes DA, Spiller W, Tilling K, Davey Smith G. Non-linear Mendelian randomization: detection of biases using negative controls with a focus on BMI, Vitamin D and LDL cholesterol. Eur J Epidemiol 2024;39:451-465.
Article
PubMed
PMC
PDF
143. Hemani G, Tilling K, Davey Smith G. Orienting the causal relationship between imprecisely measured traits using GWAS summary data. PLoS Genet 2017;13:e1007081.
Article
PubMed
PMC

Figure & Data

References

Citations

Citations to this article as recorded by

Helicobacter pylori, peptic ulcer disease, and colorectal cancer: a prospective study with genome-wide interaction analysis and Mendelian randomization
Ziqi Wan, Jiarui Mi, Xiaoyin Bai, Dong Wu, Sunny Hei Wong
Infectious Agents and Cancer.2025;[Epub] CrossRef
Evaluating causal protective effect of dual GLP-1R/GIPR agonists on MASLD: A Mendelian randomization and colocalization study
Yangke Cai, Siyuan Xie, Liyi Xu, Jiamin Chen, Jianting Cai
European Journal of Pharmacology.2025; 1005: 178088. CrossRef
Integrative genome-wide analysis unveils the genetic landscape of gallstone disease and highlights novel loci with therapeutic potential
Haotian Chen, Zhengye Liu, Hanze Du, Mixue Zheng, Ziqi Wan, Nan Zhao, Guanqiao Li, Xiaoyin Bai, Dong Wu, Jiarui Mi
BMJ Open Gastroenterology.2025; 12(1): e001976. CrossRef

Cite

CITE

export

Copy Download

Format
XML Download

Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:

RIS — For EndNote, ProCite, RefWorks, and most other reference management software
BibTeX — For JabRef, BibDesk, and other BibTeX-specific software

Include:

Citation for the content below
Citation and abstract for the content below

Prospects of Mendelian randomization in hepatology: a comprehensive literature review with practice guidance

Clin Mol Hepatol. 2025;31(4):1115-1138. Published online June 9, 2025

DOI: https://doi.org/10.3350/cmh.2025.0541

Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:

RIS — For EndNote, ProCite, RefWorks, and most other reference management software
bib — For JabRef, BibDesk, and other BibTeX-specific software

Include:

Citation for the content below
Citation and abstract for the content below

Prospects of Mendelian randomization in hepatology: a comprehensive literature review with practice guidance

Clin Mol Hepatol. 2025;31(4):1115-1138. Published online June 9, 2025

DOI: https://doi.org/10.3350/cmh.2025.0541

Figure

Prospects of Mendelian randomization in hepatology: a comprehensive literature review with practice guidance

Figure 1. The number of publications and citations in studies reporting Mendelian randomization (MR) from 2020 to 2024. (A) Displays all MR-related publications and citations, and (B) displays only those related to liver diseases. The keywords are listed at the bottom of figure.

Figure 2. The illustration of Mendelian randomization (MR) principles and its comparison with randomized controlled trial (RCT).

Figure 3. Illustration of key risk factors identified for MASLD. CHIP, clonal hematopoiesis of intermediate potential; MASLD, metabolic dysfunction-associated steatotic liver disease.

Figure 4. An illustration of performing the state-of-the-art Mendelian randomization study. GWAS, genome-wide association study; IV, instrumental variable; IVW, inverse variance weighted method.

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Prospects of Mendelian randomization in hepatology: a comprehensive literature review with practice guidance

Instrumental variable (IV) assumptions
Relevance	Genetic variants should be closely associated with the exposure.
Independence	There is no common confounder between the genetic variants and the outcome
Exclusion restriction	Genetic variants can be linked to the outcome only via the way of exposure without different mechanisms (no horizontal pleiotropy)
Concepts related to genetics
Genetic variants	Differences in the DNA sequence among individuals, including single nucleotide variation and structural variations. Some variants are harmless, while others may influence traits or disease susceptibility.
SNP/SNV	SNP refers to a common genetic variant (usually with a minor allele frequency >0.01) where a single nucleotide is replaced with another. Example: A→G.
SNP/SNV	SNV (single nucleotide variant) refers to a broader term referring to any single nucleotide change, whether rare or common.
Genotype/phenotype	Genotype refers to an organism’s genetic makeup—the specific set of genes inherited from its parents, consisting of DNA sequences.
Genotype/phenotype	Phenotype is the observable physical, biochemical, and behavioral traits of an organism, which result from the interaction of the genotype with the environment.
CNV	CNV is a type of structural variant where a subset of DNA sequences are duplicated (gains) or deleted (losses). It can affect gene dosage, leading to changes in traits or disease risk.
QTL	It is a specific region in the genome that is associated with variation in a quantitative trait (e.g., serum lipids, fasting glucose)
eQTL	It is a type of QTL that affects gene expression levels rather than a physical trait. eQTLs can be cis-eQTLs (affecting genes nearby) or trans-eQTLs (affecting distant genes).
pQTL	It is a genetic locus associated with the variation in protein abundance. They are crucial for studying disease mechanisms and identifying drug targets. Similar to eQTL, there are cis-pQTL and trans-pQTL.
Pleiotropy	Pleiotropy refers to a single genetic variant influencing multiple, seemingly unrelated traits, e.g., SNVs in the TM6SF2 region affect the risk of liver steatosis and diabetes.
Vertical pleiotropy	It refers to a genetic variant influencing multiple traits through a shared causal pathway, e.g., a gene affecting cholesterol levels might also influence the risk of liver diseases due to its effect on cholesterol. “Vertical pleiotropy” is not problematic in MR.
Horizontal pleiotropy	It refers to a genetic variant influencing multiple traits through independent pathways, e.g., a gene may independently affect the metabolism of both glucose and lipids. “Horizontal pleiotropy” is problematic in MR analysis and should be avoided.
LD	LD refers to the non-random association of alleles at different loci in the genome. If two genetic variants are physically close on a chromosome, they are often inherited together.
Clumping/pruning	It refers to techniques to reduce redundancy in genetic analyses by removing variants that are highly correlated (in LD) and selecting one representative SNP per LD block.
GRS	It refers to a score that summarizes the cumulative effect of multiple genetic variants on a trait or disease risk. It can be divided into “unweighted GRS”, which counts the number of risk alleles, and “weighted GRS”, which accounts for the effect size of each allele from studies (e.g., GWAS). It is used to predict an individual’s genetic predisposition to diseases.
Population stratification	It refers to the systematic genetic differences between populations due to ancestry, which can lead to confounding in genetic studies and should should be accounted for using statistical methods like principal component analysis.
Genomic control	It refers to a statistical method correcting for population stratification or inflation in test statistics in genetic studies, e.g., adjusting the chi-square using a genomic inflation factor (lambda).
Concepts related to statistics
IV	IV refers to a variable used in causal inference that is associated with the exposure of interest but not directly with the outcome, except through the exposure. In MR analysis, genetic variants serve as IVs to study causal relationships between a risk factor and an outcome.
Weak instrument	It refers to IVs that are weakly associated with the exposure, leading to unreliable causal estimates.
F-statistic	It is a statistic used to assess the strength of an IV. According to the “Rule of Thumb”, an F-statistic >10 indicates a sufficiently strong instrument; lower values suggest weak instruments.
R² (coefficient of determination)	It represents the proportion of variance in the exposure explained by the IV. Higher R² indicates stronger instruments, improving the reliability of causal estimates.
Individual-level data	It refers to data that includes detailed genetic and phenotypic information for each individual in a study, e.g., data from cohort studies or biobanks like UK Biobank. In MR analysis, individual-level data allow more flexible analyses, such as non-linear MR or factorial MR.
Summary-level data	It refers to aggregated results, such as SNP-exposure and SNP-outcome associations from public GWAS summary statistics, rather than raw individual data. In MR analysis, it facilitates two-sample MR and meta-analyses without accessing individual-level data.
Two-sample MR	An MR approach using two independent datasets—one for SNP-exposure associations and another for SNP-outcome associations. No need for individual-level data; enables larger sample sizes.
One-sample MR	An MR approach using a single dataset with both SNP-exposure and SNP-outcome data. It allows robust sensitivity analyses but requires individual-level data.
Bidirectional MR	An MR approach to test causality in both directions (e.g., does X cause Y, or does Y cause X?). An example is examining whether serum cholesterol influences MASLD risk and vice versa.
Mediation MR	An MR approach to determine whether the causal effect of an exposure on an outcome is mediated through an intermediate variable, e.g., testing whether cholesterol mediates the effect of diet on liver diseases.
Multivariable MR	An MR approach includes multiple exposures simultaneously to account for confounding or to identify independent causal effects, e.g., analyzing the effects of serum HDL-C and triglycerides on the risk of liver diseases.
Factorial MR	An MR approach examines the combined effects of multiple exposures or interventions using genetic proxies for each exposure, e.g., studying the combined effects of lipid-modifying drugs and blood glucose-lowering drugs on liver diseases.
Nonlinear MR	An MR approach examines whether the relationship between exposure and outcome varies across the range of the exposure, e.g., investigating whether the effect of serum triglycerides on cholelithiasis differs at different ranges of triglycerides.
Drug-target MR	An MR approach uses genetic variants in drug target genes as instruments to predict the potential effects of drug interventions, e.g., testing the effect of HMGCR inhibition (via genetic instruments) on cholesterol levels and liver disease.
Colocalization	A statistical method to determine whether the same genetic variant affects both the exposure and outcome via a shared causal mechanism. It helps distinguish true causal effects from linkage artifacts.
Directionality test (MR-Steiger test)	It tests whether the causal relationship flows in the hypothesized direction (e.g., exposure→outcome) rather than the reverse.
Power analysis	It evaluates the ability of a study to detect a true causal effect based on sample size, genetic instrument strength, and effect size, which ensures the study design is adequately powered to avoid false negatives.
Winner’s curse	It refers to the overestimation of the effect size in SNPs that are strongly associated with traits in the discovery sample but weaker in replication studies, which can reduce the reliability of causal inference.
Collider bias	It refers to bias arising when both the exposure and outcome influence a third variable (collider), leading to spurious associations, e.g., adjusting for a common out-come of two variables (e.g., BMI and liver diseases) can create false associations between the two.
Concepts related to hepatology
GCKR (rs1260326)	A genetic variant strongly associated with MASLD and T2D.
PNPLA3 (rs738409)	A genetic variant strongly associated with MASLD, liver fibrosis, and HCC.
TM6SF2 (rs58542926)	A gene variant involved in hepatic lipid metabolism, influencing fatty liver disease risk and fibrosis progression.

Liver diseases	Risk factors identified by MR	Outcomes associated with genetic liability	References
MASLD	Metabolic factors: higher BMI, higher waist circumference, insulin resistance, genetic liability to T2D, genetic liability to hypothyroidism, higher SBP, higher liver enzymes, higher triglycerides, lower HDL-C, higher serum ferritin/iron, higher homocysteine, lower serum folate.	Increased risk of T2D, obesity, CAD; higher serum uric acid	[35-56]
MASLD	Non-metabolic factors: smoking, alcohol consumption, higher serum selenium, genetic liability to CHIP, air pollution-associated metabolites.	Increased risk of T2D, obesity, CAD; higher serum uric acid	[35-56]
ALD	Alcohol use, gut dysbiosis, higher serum interleukin 7, lower circulating sphingomyelins.	Lower bone mineral density in the femoral neck.	[62-66]
HBV	-	Cervical cancer, gastric cancer, osteoporosis, rheumatoid arthritis.	[68,70-71]
HCV	-	Diffuse large B-cell lymphoma	[72]
HCC	Alcohol consumption, lower neutrophil counts, lower serum vitamin C, higher serum adiponectin, higher visceral/ectopic adiposity (adjusted by BMI), genetic liability to IBD, genetic liability to insomnia, genetic liability to daytime napping.	-	[73-84]
CCA	Genetic liability to MASLD, cholelithiasis, insomnia, IBD; protective effect from gut microbiota (family porphyromonadaceae, genus bacteroidetes and phylum verrucomicrobia)	-	[78,83,85-87]
Cholelithiasis (cholecystitis)	Metabolic factors: obesity (higher BMI, higher waist circumference), genetic liability to T2D, lower circulating omega-3 fatty acid concentrations, lower serum sphingomyelin levels, lower serum campesterol levels; lower serum total cholesterol, higher serum triglycerides.	Increased risk of GERD, CCA, colorectal cancer, kidney cancer.	[90-100]
Cholelithiasis (cholecystitis)	Non-metabolic factors: increased leisure screen time; lower moderate-to-vigorous intensity physical activity; smoking.		[90-100]
PBC^*	Autoimmune conditions: genetic liability to IBD, SLE.	Increased risk of IBD, celiac disease, rheumatoid arthritis, thyroid dysfunction, osteoporosis; decreased risk of gastric cancer	[101-112]
PBC^*	Other factors: obesity, low serum hemoglobin, lower serum vitamin D, higher serum testosterone (females), lower sex hormone binding globulin.		[101-112]
PSC^*	Genetic liability to celiac disease, IBD, SLE.	Increased risk of celiac disease, IBD.	[102,108,113]

Author (yr)	Definition	Ethnicity	Sample size	Recommendation^*	GCST ID	PMID
Namjou et al. (2019)	MASLD from health records	Europeans	1,106 cases; 8,571 controls.	3 (low)	GCST008468	31311600
Anstee et al. [57] (2020)	MASLD from live biopsy	Europeans	1,483 cases; 17,781 controls.	1 (high)	GCST90011885	32298765
Fairfield et al. (2021)	MASLD from health records (UKB)	Europeans	4,761 cases; 373,227 controls.	3 (low)	GCST90054782	34535985
Ghodsian et al. [58] (2021)	MASLD from health records (GWAS meta-analysis of 4 cohorts)	Europeans	8,434 cases; 770,180 controls.	1 (high)	GCST90091033	34841290
Miao et al. (2021)	MASLD imputed from serum and anthropometric traits	Europeans	28,396 cases; 108,652 controls.	3 (low)	GCST90094908	35047847
Vujkovic et al. [59] (2022)	MASLD proxied by cALT	Mixed	Europeans: 95,472 cases; 68,725 controls.	2 (medium)	GCST90129601	35654975
			African Americans: 23,977 cases; 13,387 controls.
			Hispanics: 7,650 cases; 7,468 controls.
			Asians: 1,088 cases; 828 controls.
Sveinbjornsson et al. (2022)	MASLD from health records	Europeans	9,491 cases and 876,210 controls	1 (high)	NA	36280732
Chen et al. [60] (2023)	MASLD from CT-imaging and health records (GWAS meta-analysis of 5 cohorts)	Mixed	Imaging samples (n=66,814) and diagnostic code (3,584 cases, 621,081 controls).	2 (medium)	GCST90271622	37709864

Table 1. A glossary of key terms in Mendelian randomization for hepatologists

Table 2. The summary of Mendelian randomization results in the field of hepatology

For PBC and PSC, there is strong genetic overlap in autoimmune conditions, caution is needed when interpreting MR results.

Table 3. Explanation and comparison of selected MASLD-related GWAS with summary statistics available

Recommendation: “1 (high)” means this MASLD GWAS is highly recommended to be used in Mendelian randomization analysis of MASLD due to the large sample size and clear definition; “2 (medium)” means this MASLD GWAS has some limitations of mixed ancestries or unclear definitions of MASLD despite the large sample size. “3 (low)” means this MASLD GWAS is not suitable in Mendelian randomization analysis as it only represents a subset of existing GWAS datasets.

Articles

Page Path

Prospects of Mendelian randomization in hepatology: a comprehensive literature review with practice guidance

Full Article

ABSTRACT

INTRODUCTION

A historic overview on the development of MR

Advanced statistical methods and extension to other areas

Metabolic dysfunction-associated steatotic disease

Primary liver cancer

Biliary diseases

CONCLUSION

FOOTNOTES

Abbreviations

REFERENCES

Figure & Data

References

Citations

CITE

Download Citation

Format:

Include:

Figure

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Table 1.

Table 2.

Table 3.

ABOUT

BROWSE ARTICLES

FOR CONTRIBUTORS