Type 1 diabetes mellitus and non-alcoholic fatty liver disease: a two-sample Mendelian randomization study

Abstract

Background:

NAFLD (Nonalcoholic fatty liver disease) is becoming an increasingly common cause of chronic liver disease. Metabolic dysfunction, overweight/obesity, and diabetes are thought to be closely associated with increased NAFLD risk. However, few studies have focused on the mechanisms of NAFLD occurrence in T1DM.

Methods:

We conducted a two-sample Mendelian randomization (MR) analysis to assess the causal association between T1DM and NAFLD with/without complications, such as coma, renal complications, ketoacidosis, neurological complications, and ophthalmic complications. Multiple Mendelian randomization methods, such as the inverse variance weighted (IVW) method, weighted median method, and MR-Egger test were performed to evaluate the causal association of T1DM and NAFLD using genome-wide association study summary data from different consortia, such as Finngen and UK biobank.

Results:

We selected 37 SNPs strongly associated with NAFLD/LFC (at a significance level of p < 5 × 10−8) as instrumental variables from the Finnish database based on the T1DM phenotype (8,967 cases and 308,373 controls). We also selected 14/16 SNPs based on with or without complications. The results suggest that the genetic susceptibility of T1DM does not increase the risk of NAFLD (OR=1.005 [0.99, 1.02], IVW p=0.516, MR Egger p=0.344, Weighted median p=0.959, Weighted mode p=0.791), regardless of whether complications are present. A slight causal effect of T1DM without complications on LFC was observed (OR=1.025 [1.00, 1.03], MR Egger p=0.045). However, none of the causal relationships were significant in the IVW (p=0.317), Weighted median (p=0.076), and Weighted mode (p=0.163) methods.

Conclusion:

Our study did not find conclusive evidence for a causal association between T1DM and NAFLD, although clinical observations indicate increasing abnormal transaminase prevalence and NAFLD progression in T1DM patients.

1 Introduction

Non-alcoholic fatty liver disease (NAFLD) has a global prevalence of as high as 25% and poses a significant threat to human health, placing a tremendous economic burden on society (1). However, no pharmacological treatments have been approved for NAFLD worldwide. Metabolic dysfunction, overweight/obesity, and diabetes are thought to be closely associated with increased NAFLD risk. Thus, the latest 2020 international expert consensus statement redefined it as metabolic-associated fatty liver disease (MAFLD) (2, 3). In the progression of type 2 diabetes (T2D) complications, NAFLD can manifest as ectopic triglyceride accumulation through insulin resistance, impaired fatty acid oxidation, and altered lipid metabolism. However, few studies have examined the relationship between T1DM and NAFLD.

T1DM is an autoimmune disease characterized by β-cell destruction and absolute insulin deficiency (4). Treatment of T1DM mainly involves adjusting exogenous insulin supplementation based on the patient’s insulin needs. During T1DM progression, increasing insulin demand and poor glycemic control can lead to indirect insulin resistance (IR) (5). IR induces dyslipidemia through excessive lipolysis, increased lipogenesis, and lipotoxic changes in adipose tissue. Dysregulation of lipid metabolism can cause multi-organ dysfunction including the liver, which is key for NAFLD development. However, the association between T1DM and NAFLD incidence, particularly in advanced T1DM with complications, and its genetic predisposition remains contentious. Conversely, NAFLD is the most common cause of transaminase elevation in T1DM (6, 7), though the underlying mechanisms remain unclear.

Mendelian randomization (MR) analysis uses common genetic variations, typically single nucleotide polymorphisms (SNPs), as instrumental variables (IVs) to explore potential causal relationships between exposures and outcomes (8–10). Since SNPs are randomly assigned at conception and unaffected by confounders (11), they can largely avoid biases from confounding factors. Thanks to rapid GWAS developments and accumulating publicly available GWAS summary statistics, two-sample MR with significantly improved statistical power has become more accessible and flexible. Previous MR studies have elucidated causal relationships between NAFLD, T2D, and obesity (12, 13). However, no MR study has examined the causal association between T1DM and NAFLD.

This article analyzed the causal relationships between T1DM with various complications and NAFLD using a two-sample MR framework, with validation in separate databases.

2 Materials and methods

We performed two-sample MR analyses to investigate the causal relationship between type 1 diabetes and non-alcoholic fatty liver disease derived from different large-scale cohorts.

2.1 Overview of Mendelian randomization study

The study design is shown in (Figure 1).

Figure 1

We employed the statistical analysis technique known as the two-sample Mendelian randomization (MR) analysis method. This approach utilizes genetic variants as instrumental variables to assess causal relationships between exposure and outcome variables. Mendelian randomization is based on three assumptions: (i) genetic variation is associated with the risk factor, (ii) genetic variation is independent of confounding factors, and (iii) genetic variation influences the outcome only through the risk factor (Figure 1A).

The overlap of participants in two-sample Mendelian randomization (MR) may lead to an increased type 1 error rate due to weak instrument bias (14, 15). Non-overlap between exposure and outcome is crucial for the two-sample MR framework. Causal estimates from single-sample analysis (using data from a single data source) tend to be biased toward the observed association between the risk factor and the outcome. In contrast, estimates from two-sample analysis (using data from non-overlapping datasets of the risk factor and the outcome) have a smaller estimation bias, tending towards zero. As most exposure factors were derived from the FinnGen team (https://r9.finngen.fi/), the NAFLD GWAS based on the GWAS catalog (https://www.ebi.ac.uk/) was excluded to avoid bias due to sample overlap.

This analysis was conducted using summary-level data from published genome-wide association studies (GWAS) and followed the STROBE-MR guidelines (9). All GWAS studies included in the analysis had obtained approval from the relevant review committees, and all participants had provided informed consent. The FinnGen biobank GWAS was conducted by the FinnGen team and ethical approval was obtained from the original studies (Figure 1B).

2.2 Sources of genome-wide association studies

The GWAS summary statistics of NAFLD were obtained from a recently published GWAS catalog (GCST90054782), with 4,761 European NAFLD cases and 373,227 genetically matched controls. Considering NAFLD is closely associated with LFC, a recent LFC GWAS (GCST90016676) which included 25,617 European participants was also selected. The GWAS data for T1DM were obtained from the FinnGen team, which included 8,967 European T1DM cases and 308,373 control cases from the r9 version of the database. Further searches were conducted based on the occurrence of T1DM complications, resulting in GWAS data for the complication group (6,234 cases and 308,280 controls) and the non-complication group (4,918 cases and 183,185 controls). Specific complications were further analyzed, and GWAS data related to coma (2,050 cases), renal dysfunction (4,918 cases), ketoacidosis (2,102 cases), neurological complications (1,077 cases), and optic nerve complications (5,202 cases) were selected (Table 1).

The coordination of encoding and reference alleles was employed to eliminate ambiguous SNPs with inconsistent alleles in the exposure and outcome datasets. Palindromic SNPs with minor allele frequencies between 0.45 and 0.55 were defined as having ambiguous minor alleles and all potentially ambiguous SNPs were excluded in sensitivity analyses. Some missing tools were not imputed by proxy SNPs, as the impact of a small fraction of missing tools on the results is considered negligible.

2.3 Statistical analysis

Mendelian randomization (MR) assumes that the genetic variants used as instruments are closely related to the exposure of interest and are unrelated to any potential confounding factors that might influence the outcome. The F-statistic quantifies the strength of association of each assumed risk factor with its genetic instrument. The power of each SNP was evaluated by its F-statistic (F = beta2/se2) and the overall F-statistic for all SNPs was calculated (excluding SNPs with F-values less than 10) (Table 2). Statistical power was calculated using mRnd (https://shiny.cnsgenomics.com/mRnd/). Additionally, the PhenoScanner (V2) database was utilized for filtering genetic variants associated with confounding factors. Subsequently, these rigorously filtered SNPs were employed as final instrumental variables for subsequent Mendelian randomization analysis. In this study, no instrumental variables related to phenotypes such as insulin resistance and lipid metabolism alterations were identified.

We conducted and reported the results of a range of sensitivity analyses that allowed for effective estimation in the presence of horizontal pleiotropy. First, the random-effects inverse variance weighted method was used to relax the exclusion restriction assumption by allowing all SNPs to exhibit random pleiotropic effect while maintaining asymptotic unbiasedness even if all SNPs show horizontal pleiotropy (16). The random-effects inverse variance weighted method has been reported as the most common and best-performing method in various scenarios that violate the exclusion restriction principle (17). Second, if at least 50% of the selected SNPs are valid, the weighted median estimator can provide an unbiased causal effect (18). Third, the MR-PRESSO method was performed, which can detect pleiotropy outliers in the multiple instrument summary-level MR analysis, and obtain causal effect estimates using the inverse variance weighted method after excluding the outliers (19). Finally, leave-one-out sensitivity tests were conducted to examine the influence of peripheral and pleiotropic SNPs on causal estimates (20). Heterogeneity and pleiotropy of individual SNPs were assessed using Cochran’s Q statistic and MR-PRESSO. Finally, a fixed-effects model was used to combine the MR results from the discovery and replication stages. All statistical analyses were performed using R 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria).

3 Results

3.1 SNP selection

Firstly, several SNPs were identified as IVs (instrumental variables) related to different phenotypes in various databases, with a strong correlation (at a significance level of p < 5 × 10−8). Among the wide-definition T1DM phenotypes, 37 SNPs were identified. Based on the T1DM with/without complications, 14/16 SNPs were selected. Additionally, 5-15 SNPs were identified as related phenotypes of different T1DM complications. Secondly, the variances of all the aforementioned SNPs ranged from 1.85% to 28.84%. Eventually, the F statistics of all IVs were greater than 10, indicating the absence of potential weak instrument bias (Table 2).

3.2 The causal association between T1DM with/without complications and NAFLD

Generally, our MR study suggested that there might be no causal relationship between NAFLD and T1DM. Preliminary findings indicate that the genetic susceptibility of T1DM does not increase the risk of NAFLD (OR=1.005 [0.99, 1.02], IVW p=0.516). Consistent conclusions were obtained through additional statistical methods such as MR Egger (OR=0.982 [0.95, 1.02], p=0.344), Weighted median (OR=1.000 [0.98, 1.02], p=0.959), and Weighted mode (OR=1.002 [0.98, 1.02], p=0.791). Considering that T1DM is often classified into the presence or absence of complications, subgroup analysis for these two phenotypes also suggests that the genetic predisposition of both phenotypes does not influence the risk of NAFLD, both for T1DM with complications (OR=1.007 [0.99, 1.03], IVW p=0.519) and without complications (OR=1.020 [0.98, 1.06], IVW p=0.284) (Figure 2A, Supplementary Table 1).

Figure 2

A slight causal effect of T1DM without complications on LFC was observed (OR=1.025 [1.00, 1.03], MR Egger p=0.045). However, none of the causal relationships were significant in the IVW (p=0.317), Weighted median (p=0.076), and Weighted mode (p=0.163) methods. Similarly, genetically-driven T1DM (OR=1.003 [0.99, 1.02], IVW p=0.693) and T1DM with complications (OR=0.998 [0.98, 1.01], IVW p=0.818) did not alter LFC. In the MR-Egger regression and weighted median methods, none of the causal relationships were significant (Table 3). Despite detecting slight heterogeneity in the T1DM-NAFLD and T1DM-LFC analyses, the conclusions remained valid after outlier removal (Figure 2B, Supplementary Table 2).

3.3 The causal association between common T1DM with different complications and NAFLD

3.3.1 E4_DM1COMA is type 1 diabetes with coma

In the set of IVs (p < 5 × 10−8), 8 SNPs related to type 1 diabetes with coma were identified. It was found that no genetic liability to type 1 diabetes with coma was causally associated with NAFLD (OR=1.009 [0.99, 1.03], IVW p=0.433). Consistent conclusions were obtained through additional statistical methods such as MR Egger (OR=1.010 [0.97, 1.06], p=0.693), Weighted median (OR=1.010 [0.98, 1.03], p=0.463), and Weighted mode (OR=1.009 [0.98, 1.03], p=0.489) (Figure 3, Supplementary Table 1).

Figure 3

In the other database, consistent conclusions were obtained between type 1 diabetes with coma and LFC (OR=1.003 [0.99, 1.01], IVW p=0.604) (Figure 4, Supplementary Table 2).

Figure 4

3.3.2 E4_DM1REN is type 1 diabetes with renal complications

In the set of IVs (p < 5 × 10−8), 5 SNPs related to type 1 diabetes with renal complications were identified. It was found that no genetic liability to type 1 diabetes with renal complications was causally associated with NAFLD (OR=1.003 [0.98, 1.03], IVW p=0.788). Consistent conclusions were obtained through additional statistical methods such as MR Egger (OR=1.021 [0.98, 1.06], p=0.341), Weighted median (OR=1.007 [0.98, 1.03], p=0.557), and Weighted mode (OR=1.008 [0.98, 1.03], p=0.564) (Figure 3, Supplementary Table 1).

In the other database, consistent conclusions were obtained between type 1 diabetes with renal complications and LFC (OR=1.003 [0.99, 1.02], IVW p=0.692) (Figure 4, Supplementary Table 2).

3.3.3 E4_DM1KETO is type 1 diabetes with ketoacidosis

In the set of IVs (p < 5 × 10−8), 8 SNPs related to type 1 diabetes with ketoacidosis were identified. It was found that no genetic liability to type 1 diabetes with ketoacidosis was causally associated with NAFLD (OR=1.012 [0.99, 1.03], IVW p=0.279). Consistent conclusions were obtained through additional statistical methods such as MR Egger (OR=1.012 [0.98, 1.05], p=0.528), Weighted median (OR=1.010 [0.99 1.03], p=0.416), and Weighted mode (OR=1.010 [0.99, 1.03], p=0.433) (Figure 3, Supplementary Table 1).

In the other database, consistent conclusions were obtained between type 1 diabetes with ketoacidosis and LFC (OR=1.003 [0.99, 1.01], IVW p=0.624) (Figure 4, Supplementary Table 2).

3.3.4 E4_DM1NEU is type 1 diabetes with neurological complications

In the set of IVs (p < 5 × 10−8), 4 SNPs related to type 1 diabetes with neurological complications were identified. It was found that no genetic liability to type 1 diabetes with neurological complications was causally associated with NAFLD (OR=1.008 [0.98, 1.03], IVW p=0.526). Consistent conclusions were obtained through additional statistical methods such as MR Egger (OR=0.960 [0.89, 1.03], p=0.384), Weighted median (OR=1.010 [0.98 1.03], p=0.667), and Weighted mode (OR=1.004 [0.97, 1.04], p=0.818) (Figure 3, Supplementary Table 1).

In the other database, consistent conclusions were obtained between type 1 diabetes with neurological complications and LFC (OR=1.003 [0.99, 1.01], IVW p=0.610) (Figure 4, Supplementary Table 2).

3.3.5 E4_DM1OPTH is type 1 diabetes with ophthalmic complications

In the set of IVs (p < 5 × 10−8), 15 SNPs related to type 1 diabetes with ophthalmic complications were identified. It was found that no genetic liability to type 1 diabetes with ophthalmic complications was causally associated with NAFLD (OR=1.002[0.97, 1.03], IVW p=0.885). Consistent conclusions were obtained through additional statistical methods such as MR Egger (OR=1.027 [0.98, 1.07], p=0.266), Weighted median (OR=1.008 [0.98 1.04], p=0.621), and Weighted mode (OR=1.011 [0.98, 1.04], p=0.459) (Figure 3, Supplementary Table 1).

In the other database, consistent conclusions were obtained between type 1 diabetes with ophthalmic complications and LFC (OR=1.004 [0.99, 1.02], IVW p=0.666) (Figure 4, Supplementary Table 2).

4 Discussion

Given the escalating global burden of metabolic disorders, conducting high-quality prospective research to evaluate the correlation between the genetic condition T1DM and metabolic disorder NAFLD presents significant challenges. Our study extensively examined the correlation between non-alcoholic fatty liver disease (NAFLD) and type 1 diabetes mellitus (T1DM). Through an in-depth analysis of a substantial and independent cohort, our findings suggest that there may not be a direct causal relationship between T1DM and NAFLD. Prior studies have posited a plausible link between NAFLD and T1DM, potentially attributed to metabolic disruptions stemming from the evolution of T1DM, including secondary metabolic derangements induced by T1DM complications (21). We further investigated the relationship between NAFLD and T1DM with complications. Our research found that the presence or absence of complications such as renal dysfunction, coma, ketoacidosis, neurological complications, and retinopathy did not increase the probability of NAFLD development in individuals with T1DM. In other words, regardless of complications, the genetic susceptibility of T1DM does not increase the likelihood of NAFLD. Our findings carry noteworthy implications for understanding the T1DM-NAFLD relationship, potentially facilitating the clinical management of T1DM patients with concurrent NAFLD.

The observed causal effect of T1DM on LFC in this study may be a false positive, as this result was not significant in both the IVW and weighted median methods. However, we cannot completely rule out this causal relationship, as the genetic susceptibility of T1DM, high glucose-induced oxidative stress, and cellular metabolic dysfunction may impact the pathogenesis of NAFLD (22, 23). Additionally, reports suggest that as the disease progresses, the degree of hepatic steatosis increases further in T1DM with complications (21, 24, 25). Therefore, further investigations are warranted to elucidate this association.

Regarding the null associations, several factors can explain them. Metabolic syndrome, characterized by obesity, hyperglycemia, dyslipidemia, and systemic hypertension, currently represents the most prominent risk factor for NAFLD (26). More than 70% of NAFLD patients typically manifest elevated levels of triglycerides (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), and reduced levels of high-density lipoprotein cholesterol (HDL-C), primarily synthesized in the liver, indicating dysregulated lipid metabolism in these individuals. In contrast to type 2 diabetes mellitus (T2DM), well-controlled patients with type 1 diabetes mellitus (T1DM) show no significant alterations in lipid levels (LDL-C, HDL-C, TG) and blood pressure levels (27–29). Furthermore, the majority of T1DM patients tend to develop a lean phenotype rather than obesity (29). Hence, we postulate that T1DM may induce a specific metabolic state that may not predispose individuals to NAFLD.

Interestingly, there is direct evidence suggesting that the absolute insulin deficiency in T1DM patients, leading to the absence of portal vein/peripheral insulin gradients, may reduce insulin stimulation of hepatic lipogenesis and impede the development of NAFLD (30). Additionally, Perseghin et al. (31) reported significantly lower hepatic fat content in T1DM patients compared to non-diabetic individuals, with no significant association between NAFLD and T1DM (32). However, due to the lifelong subcutaneous insulin injections required in T1DM, relative peripheral hyperinsulinemia and hepatic hypoinsulinemia may occur, leading to insulin resistance (33). This relative insulin-resistant state can affect glucose and lipid hepatic metabolism, triggering pro-inflammatory cascades and resulting in hepatic steatosis, fibrosis, and cirrhosis (34–36). The onset and progression of NAFLD can be promoted following insulin resistance in T1DM. Therefore, the observed coexistence of NAFLD and T1DM may be attributed to the effects of treatment, particularly the impact of insulin resistance on NAFLD. In conclusion, based on previous studies, it is challenging to demonstrate a direct promotion of NAFLD development by T1DM.

Furthermore, recent studies indicate an association between NAFLD and complications in patients with T1DM (21). In the general population, the prevalence of hepatic steatosis (defined as HSI > 36) is 37.1%, but it increases to 49% in T1DM patients with complications and to 32% in those without complications. This suggests an independent correlation of NAFLD with cardiovascular disease, chronic kidney disease, retinopathy, and neuropathy in T1DM patients, regardless of established risk factors (24, 25, 37, 38). Some researchers propose that poor blood glucose control in patients with chronic diabetes complications leads to the conversion of glucose into fat in the liver via GLUT2 transport, thereby promoting hepatic lipid accumulation (39). However, these studies are based on populations of poorly controlled T1DM patients with complications, and there is no difference in prevalence compared to normal overweight populations. From a genetic susceptibility perspective, it cannot be inferred that the occurrence of T1DM (including T1DM with complications) is causally related to the development of NAFLD.

The selection of instrumental variables is a critical step in Mendelian randomization studies. To ensure validity, any genetic variation must influence the outcome solely through its effect on the exposure, without horizontal pleiotropy (40, 41). To address potential pleiotropic factors, MR-Egger regression can be used to reveal a statistically significant intercept when these genetic variants are excluded, thereby reducing heterogeneity and eliminating horizontal pleiotropy (42). In patients with T1DM, metabolic disturbances resulting from insulin resistance and alterations in glucose and lipid metabolism may significantly contribute to the development of NAFLD (43, 44). In this study, according to the fundamental assumption of MR studies, we adopted SNPs with a significance threshold of P < 5 × 10^-8 from the GWAS data as instrumental variables. To address potential linkage disequilibrium effects on the analysis outcomes, we imposed criteria of r^2 < 0.001 and a window size of 10,000 kb. To ensure robust associations between instrumental variables and endogenous variables and to mitigate bias from weak instrumental variables, we computed R^2 [R^2 = 2 × EAF × (1 - EAF) × b^2], which represents the proportion of variance explained by instrumental variable SNPs, and the F-statistic [F = R^2 × (N - 2)/(1 - R^2)] to evaluate the strength of instrumental variables, each with its distinct characteristics. Furthermore, we utilized the PhenoScanner (V2; http://www.phenoscanner.medschl.cam.ac.uk/upload/) database to filter out genetic variants linked to potential confounding factors like insulin resistance. These meticulously screened SNPs were subsequently employed as the definitive instrumental variables for subsequent MR analyses.

To minimize potential weak instrument bias amplification, separate databases with minimal sample overlap between exposure and outcome data sources were utilized. Despite our efforts, limitations of this study have been acknowledged that must be taken into account. One limitation is the challenges of pleiotropy in the MR framework, as found in all MR studies (45, 46). To address this, various sensitivity analyses with different assumptions were conducted.

Our research boasts several noteworthy strengths. Firstly, this analysis enrolled the largest number of phenotypes of T1DM with complications in different datasets to date, including discovery and validation stages, thereby enhancing the persuasiveness of the findings (47). Secondly, we conducted rigorous sensitivity analyses to validate the IVs pertinent to our hypothesis. After considering horizontal pleiotropy, outliers, and sample overlap, the confidence of the results was increased. Thirdly, our study was restricted to individuals of European descent, which could reduce population stratification bias (48, 49).. Finally, in our study, we maintained a minimal degree of overlap between exposure and outcome data sources, so that inflation of the weak instrument bias could be avoided as much as possible.

5 Conclusion

In this study, a two-sample Mendelian randomization framework was utilized to investigate the potential causal relationship between T1DM and NAFLD. The analysis was validated subsequent to the integration and scrutiny of selected databases. Our findings suggest that T1DM, as well as combined with complications, does not directly lead to an increased incidence of NAFLD. Mechanisms linking the two beyond genetics remain unclear and warrant further investigation in future studies.

References

• 1

  YuenLSaeediPRiazMKarurangaSDivakarHLevittNet al. Projections of the prevalence of hyperglycaemia in pregnancy in 2019 and beyond: Results from the International Diabetes Federation Diabetes Atlas, 9th edition. Diabetes Res Clin Pract. (2019) 157:107841. doi: 10.1016/j.diabres.2019.107841

  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  CraneHGoftonCSharmaAGeorgeJ. MAFLD: an optimal framework for understanding liver cancer phenotypes. J Gastroenterol. (2023) 58:947–64. doi: 10.1007/s00535-023-02021-7

  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  GoftonCUpendranYZhengM-HGeorgeJ. MAFLD: How is it different from NAFLD? Clin Mol Hepatol. (2023) 29:S17–31. doi: 10.3350/cmh.2022.0367

  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  BluestoneJABucknerJHHeroldKC. Immunotherapy: Building a bridge to a cure for type 1 diabetes. Science. (2021) 373:510–6. doi: 10.1126/science.abh1654

  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  SimsEKBesserREJDayanCGeno RasmussenCGreenbaumCGriffinKJet al. Screening for type 1 diabetes in the general population: A status report and perspective. Diabetes. (2022) 71:610–23. doi: 10.2337/dbi20-0054

  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  MertensJDe BlockCSpinhovenMDriessenAFrancqueSMKwantenWJ. Hepatopathy associated with type 1 diabetes: distinguishing non-alcoholic fatty liver disease from glycogenic hepatopathy. Front Pharmacol. (2021) 12:768576. doi: 10.3389/fphar.2021.768576

  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  AbdelmalekMF. Nonalcoholic fatty liver disease: another leap forward. Nat Rev Gastroenterol Hepatol. (2021) 18:85–6. doi: 10.1038/s41575-020-00406-0

  • CrossRef
  • Google Scholar

• 8

  SekulaPDel GrecoMFPattaroCKöttgenA. Mendelian randomization as an approach to assess causality using observational data. J Am Soc Nephrol JASN. (2016) 27:3253–65. doi: 10.1681/ASN.2016010098

  • CrossRef
  • Google Scholar

• 9

  SkrivankovaVWRichmondRCWoolfBARDaviesNMSwansonSAVanderWeeleTJet al. Strengthening the reporting of observational studies in epidemiology using mendelian randomisation (STROBE-MR): explanation and elaboration. BMJ. (2021) 375:n2233. doi: 10.1136/bmj.n2233

  • CrossRef
  • Google Scholar

• 10

  ThanassoulisGO’DonnellCJ. Mendelian randomization: nature’s randomized trial in the post-genome era. JAMA. (2009) 301:2386–8. doi: 10.1001/jama.2009.812

  • CrossRef
  • Google Scholar

• 11

  Davey SmithGHemaniG. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet. (2014) 23:R89–98. doi: 10.1093/hmg/ddu328

  • CrossRef
  • Google Scholar

• 12

  LiuZZhangYGrahamSWangXCaiDHuangMet al. Causal relationships between NAFLD, T2D and obesity have implications for disease subphenotyping. J Hepatol. (2020) 73:263–76. doi: 10.1016/j.jhep.2020.03.006

  • CrossRef
  • Google Scholar

• 13

  XieJHuangHLiuZLiYYuCXuLet al. The associations between modifiable risk factors and nonalcoholic fatty liver disease: A comprehensive Mendelian randomization study. Hepatol Baltim Md. (2023) 77:949–64. doi: 10.1002/hep.32728

  • CrossRef
  • Google Scholar

• 14

  RasoolyDPelosoGM. Two-sample multivariable mendelian randomization analysis using R. Curr Protoc. (2021) 1:e335. doi: 10.1002/cpz1.335

  • CrossRef
  • Google Scholar

• 15

  ZhangMChenJYinZWangLPengL. The association between depression and metabolic syndrome and its components: a bidirectional two-sample Mendelian randomization study. Transl Psychiatry. (2021) 11:633. doi: 10.1038/s41398-021-01759-z

  • CrossRef
  • Google Scholar

• 16

  LambertJERamos-RomanMABrowningJDParksEJ. Increased de novo lipogenesis is a distinct characteristic of individuals with nonalcoholic fatty liver disease. Gastroenterology. (2014) 146:726–35. doi: 10.1053/j.gastro.2013.11.049

  • CrossRef
  • Google Scholar

• 17

  DonnellyKLSmithCISchwarzenbergSJJessurunJBoldtMDParksEJ. Sources of fatty acids stored in liver and secreted via lipoproteins in patients with nonalcoholic fatty liver disease. J Clin Invest. (2005) 115:1343–51. doi: 10.1172/JCI23621

  • CrossRef
  • Google Scholar

• 18

  BrouwersMCGJSimonsNStehouwerCDAKoekGHSchaperNCIsaacsA. Relationship between nonalcoholic fatty liver disease susceptibility genes and coronary artery disease. Hepatol Commun. (2019) 3:592–3. doi: 10.1002/hep4.1319

  • CrossRef
  • Google Scholar

• 19

  MantovaniACsermelyAPetraccaGBeatriceGCoreyKESimonTGet al. Non-alcoholic fatty liver disease and risk of fatal and non-fatal cardiovascular events: an updated systematic review and meta-analysis. Lancet Gastroenterol Hepatol. (2021) 6:903–13. doi: 10.1016/S2468-1253(21)00308-3

  • CrossRef
  • Google Scholar

• 20

  LiuJAu YeungSLLinSLLeungGMSchoolingCM. Liver enzymes and risk of ischemic heart disease and type 2 diabetes mellitus: A mendelian randomization study. Sci Rep. (2016) 6:38813. doi: 10.1038/srep38813

  • CrossRef
  • Google Scholar

• 21

  de VriesMWesterinkJKaasjagerKHAHde ValkHW. Prevalence of nonalcoholic fatty liver disease (NAFLD) in patients with type 1 diabetes mellitus: A systematic review and meta-analysis. J Clin Endocrinol Metab. (2020) 105:3842–53. doi: 10.1210/clinem/dgaa575

  • CrossRef
  • Google Scholar

• 22

  RégnierMCarbinattiTParlatiLBenhamedFPosticC. The role of ChREBP in carbohydrate sensing and NAFLD development. Nat Rev Endocrinol. (2023) 19:336–49. doi: 10.1038/s41574-023-00809-4

  • CrossRef
  • Google Scholar

• 23

  KawamuraSMatsushitaYKurosakiSTangeMFujiwaraNHayataYet al. Inhibiting SCAP/SREBP exacerbates liver injury and carcinogenesis in murine nonalcoholic steatohepatitis. J Clin Invest. (2022) 132:e151895. doi: 10.1172/JCI151895

  • CrossRef
  • Google Scholar

• 24

  AydınFGerenliNDursunFAtasoyTÖKalınSKırmızıbekmezH. Hepatopathies in children and adolescents with type 1 diabetes. J Pediatr Endocrinol Metab JPEM. (2019) 32:121–6. doi: 10.1515/jpem-2018-0255

  • CrossRef
  • Google Scholar

• 25

  CasteraLFriedrich-RustMLoombaR. Noninvasive assessment of liver disease in patients with nonalcoholic fatty liver disease. Gastroenterology. (2019) 156:1264–1281.e4. doi: 10.1053/j.gastro.2018.12.036

  • CrossRef
  • Google Scholar

• 26

  FriedmanSLNeuschwander-TetriBARinellaMSanyalAJ. Mechanisms of NAFLD development and therapeutic strategies. Nat Med. (2018) 24:908–22. doi: 10.1038/s41591-018-0104-9

  • CrossRef
  • Google Scholar

• 27

  ChiesaSTCharakidaMMcLoughlinENguyenHCGeorgiopoulosGMotranLet al. Elevated high-density lipoprotein in adolescents with type 1 diabetes is associated with endothelial dysfunction in the presence of systemic inflammation. Eur Heart J. (2019) 40:3559–66. doi: 10.1093/eurheartj/ehz114

  • CrossRef
  • Google Scholar

• 28

  ChaitAGoldbergI. Treatment of dyslipidemia in diabetes: recent advances and remaining questions. Curr Diabetes Rep. (2017) 17:112. doi: 10.1007/s11892-017-0942-8

  • CrossRef
  • Google Scholar

• 29

  MarxNFedericiMSchüttKMüller-WielandDAjjanRAAntunesMJet al. 2023 ESC Guidelines for the management of cardiovascular disease in patients with diabetes. Eur Heart J. (2023) 44:4043–140. doi: 10.1093/eurheartj/ehad192

  • CrossRef
  • Google Scholar

• 30

  XuWCuiCCuiCChenZZhangHCuiQet al. Hepatocellular cystathionine γ lyase/hydrogen sulfide attenuates nonalcoholic fatty liver disease by activating farnesoid X receptor. Hepatol Baltim Md. (2022) 76:1794–810. doi: 10.1002/hep.32577

  • CrossRef
  • Google Scholar

• 31

  PerseghinGLattuadaGDe CobelliFEspositoACostantinoFCanuTet al. Reduced intrahepatic fat content is associated with increased whole-body lipid oxidation in patients with type 1 diabetes. Diabetologia. (2005) 48:2615–21. doi: 10.1007/s00125-005-0014-5

  • CrossRef
  • Google Scholar

• 32

  LlauradóGSevastianovaKSädevirtaSHakkarainenALundbomNOrho-MelanderMet al. Liver fat content and hepatic insulin sensitivity in overweight patients with type 1 diabetes. J Clin Endocrinol Metab. (2015) 100:607–16. doi: 10.1210/jc.2014-3050

  • CrossRef
  • Google Scholar

• 33

  DugbarteyGJAlornyoKKAdamsIAtuleSObeng-KyeremehRAmoahDet al. Targeting hepatic sulfane sulfur/hydrogen sulfide signaling pathway with α-lipoic acid to prevent diabetes-induced liver injury via upregulating hepatic CSE/3-MST expression. Diabetol Metab Syndr. (2022) 14:148. doi: 10.1186/s13098-022-00921-x

  • CrossRef
  • Google Scholar

• 34

  RousselRTravertFPasquetBWilsonPWFSmithSCGotoSet al. Metformin use and mortality among patients with diabetes and atherothrombosis. Arch Intern Med. (2010) 170:1892–9. doi: 10.1001/archinternmed.2010.409

  • CrossRef
  • Google Scholar

• 35

  RandriamboavonjyVMannWAElgheznawyAPoppRRogowskiPDornaufIet al. Metformin reduces hyper-reactivity of platelets from patients with polycystic ovary syndrome by improving mitochondrial integrity. Thromb Haemost. (2015) 114:569–78. doi: 10.1160/TH14-09-0797

  • CrossRef
  • Google Scholar

• 36

  CarvajalRRosasCKohanKGablerFVantmanDRomeroCet al. Metformin augments the levels of molecules that regulate the expression of the insulin-dependent glucose transporter GLUT4 in the endometria of hyperinsulinemic PCOS patients. Hum Reprod Oxf Engl. (2013) 28:2235–44. doi: 10.1093/humrep/det116

  • CrossRef
  • Google Scholar

• 37

  ClementeMGMandatoCPoetaMVajroP. Pediatric non-alcoholic fatty liver disease: Recent solutions, unresolved issues, and future research directions. World J Gastroenterol. (2016) 22:8078–93. doi: 10.3748/wjg.v22.i36.8078

  • CrossRef
  • Google Scholar

• 38

  CaralloCTripolinoCDe FranceschiMSIraceCXuXYGnassoA. Carotid endothelial shear stress reduction with aging is associated with plaque development in twelve years. Atherosclerosis. (2016) 251:63–9. doi: 10.1016/j.atherosclerosis.2016.05.048

  • CrossRef
  • Google Scholar

• 39

  ShearerAMWangYFletcherEKRanaRMichaelESNguyenNet al. PAR2 promotes impaired glucose uptake and insulin resistance in NAFLD through GLUT2 and Akt interference. Hepatol Baltim Md. (2022) 76:1778–93. doi: 10.1002/hep.32589

  • CrossRef
  • Google Scholar

• 40

  YinK-JHuangJ-XWangPYangX-KTaoS-SLiH-Met al. No genetic causal association between periodontitis and arthritis: A bidirectional two-sample mendelian randomization analysis. Front Immunol. (2022) 13:808832. doi: 10.3389/fimmu.2022.808832

  • CrossRef
  • Google Scholar

• 41

  LiPWangHGuoLGouXChenGLinDet al. Association between gut microbiota and preeclampsia-eclampsia: a two-sample Mendelian randomization study. BMC Med. (2022) 20:443. doi: 10.1186/s12916-022-02657-x

  • CrossRef
  • Google Scholar

• 42

  BowdenJDel GrecoMFMinelliCDavey SmithGSheehanNAThompsonJR. Assessing the suitability of summary data for two-sample Mendelian randomization analyses using MR-Egger regression: the role of the I2 statistic. Int J Epidemiol. (2016) 45:1961–74. doi: 10.1093/ije/dyw220

  • CrossRef
  • Google Scholar

• 43

  LiuDJPelosoGMYuHButterworthASWangXMahajanAet al. Exome-wide association study of plasma lipids in >300,000 individuals. Nat Genet. (2017) 49:1758–66. doi: 10.1040/ng.3977

  • CrossRef
  • Google Scholar

• 44

  BrouwersMCGJSimonsNStehouwerCDAIsaacsA. Non-alcoholic fatty liver disease and cardiovascular disease: assessing the evidence for causality. Diabetologia. (2020) 63:253–60. doi: 10.1007/s00125-019-05024-3

  • CrossRef
  • Google Scholar

• 45

  SmithMLBullCJHolmesMVDavey SmithGSandersonEAndersonELet al. Distinct metabolic features of genetic liability to type 2 diabetes and coronary artery disease: a reverse Mendelian randomization study. EBioMedicine. (2023) 90:104503. doi: 10.1016/j.ebiom.2023.104503

  • CrossRef
  • Google Scholar

• 46

  BrownBCKnowlesDA. Welch-weighted Egger regression reduces false positives due to correlated pleiotropy in Mendelian randomization. Am J Hum Genet. (2021) 108:2319–35. doi: 10.1016/j.ajhg.2021.10.006

  • CrossRef
  • Google Scholar

• 47

  XiangKWangPXuZHuY-QHeY-SChenYet al. Causal effects of gut microbiome on systemic lupus erythematosus: A two-sample mendelian randomization study. Front Immunol. (2021) 12:667097. doi: 10.3389/fimmu.2021.667097

  • CrossRef
  • Google Scholar

• 48

  TeoKAbeysekeraKWMAdamsLAignerEAnsteeQMBanalesJMet al. rs641738C>T near MBOAT7 is associated with liver fat, ALT and fibrosis in NAFLD: A meta-analysis. J Hepatol. (2021) 74:20–30. doi: 10.1016/j.jhep.2020.08.027

  • CrossRef
  • Google Scholar

• 49

  MantovaniAMorieriMLPalmisanoLMasulliMCossuEBaroniMGet al. Hepatic steatosis with significant fibrosis is associated with an increased 10-year estimated risk of cardiovascular disease in adults with type 1 diabetes mellitus. Cardiovasc Diabetol. (2023) 22:204. doi: 10.1186/s12933-023-01945-x

  • CrossRef
  • Google Scholar