This two-sample and multivariable MR study was designed to explore the causal effect of T2DM and glycemic traits on the risk of AD. The ethics committee at each institutional review board review board authorized all participants’ written informed consent in separate studies. Further ethical approval or consent was deemed unnecessary. Our MR study was predicated on three foundational assumptions: (1) the IVs are related to exposures of interest; (2) the IVs are not related to the confounders of the exposure-outcome relation; (3) the IVs’ links with the outcome are only via the exposure of interest (15). To eliminate potential confounding factors, specifically the impact of hypertension, two types of secondary analyses were conducted. To address potential confounders, particularly hypertension's influence, we conducted two secondary analyses. First, a two-sample MR study was used to exclude Single Nucleotide Polymorphisms (SNPs) linked to hypertension. Then, drawing upon epidemiological information on T2DM and AD, we conducted two-sample MR analysis to explore factors that may influence the onset of these two diseases. In this context, we examined factors such as blood pressure, lipids, diet, sleep, physical activity, smoking, alcohol, socioeconomic elements, and stress. Factors related to both T2DM and AD were deemed confounders and included in further multivariate MR studies to assess T2DM's independent role. We used coronary artery disease (CAD) as a positive control to ensure IVs showed expected associations with a common disease, reducing selection bias risk (16). Additionally, the UK Biobank's AD database served as a validation set. Subsequent sections detail the data sources, genetic instrument selection, and statistical analyses (Figure 1; Table 1; Supplementary Table S1).

We sourced genetic data for T2DM from the DIAbetes Meta-ANalysis of Trans-Ethnic association studies (DIAMANTE) consortium's GWAS meta-analysis, encompassing 74,124 cases and 824,006 controls, all of European ancestry (17). T2DM in the original GWAS was defined by diagnostic fasting glucose (FG), fasting insulin (FI), 2 h plasma glucose or hemoglobin A_1c (HbA_1c) levels; use of glucose-lowering medication (by Anatomical Therapeutic Chemical code or self-report); or T2DM history from electronic medical records, self-report and varying combinations of each, depending on the contributing cohort.

For glycemic traits like FI, FG, and HbA_1c, we referred to the Meta-Analyses of Glucose and Insulin-Related Traits Consortium (MAGIC) data, which includes information from up to 196,991 European participants.

To confirm potential confounding factors, we included variables such as Systolic Blood Pressure (SBP), Diastolic Blood Pressure (DBP), Pulse Pressure (PP), High-Density Lipoprotein cholesterol (HDL-C) (18), Low-Density Lipoprotein cholesterol (LDL-C) (18), Total Cholesterol (TC) (18), Triglycerides (TG) (18), dietary intakes (protein, fat, sugar, carbohydrates) (19), socioeconomic factors [educational level (20), annual household income, Townsend Deprivation Index (TDI)], lifestyle factors (skipping breakfast (21), napping (22), daytime sleepiness (23), sleep duration (24), insomnia (25), leisure screen time (26), moderate-to-vigorous intensity physical activity during leisure time (MVPA) (26), sedentary behavior at work (26), environmental stress and adversity (27), smoking (smoking initiation, daily smoking, smoking cessation, smoking initiation) (28), drinking (weekly alcohol consumption data (28), alcohol abuse (29). Detailed database sources for each variable are provided in Supplementary Table S2.

We included SNPs correlating with the selected traits at a genome-wide significance threshold (p < 5 × 10⁻⁸). We ensured the avoidance of weak instrument bias by prioritizing SNPs with F-statistics exceeding 10 (30). The absence of rsid in the DIAMANTE's T2DM database necessitated its determination via SNP-nexus (https://www.snp-nexus.org/v4/) and the National Library of Medicine (https://www.ncbi.nlm.nih.gov/snp/?term=), reliant on chromosome and position data. In the presence of linkage disequilibrium (LD) (r²> 0.001), preference was given to the SNP boasting the strongest association, thereby safeguarding the independence assumption. This yielded 150 SNPs for T2DM, 38 for FI, 70 for FG, and 75 for HbA1c collectively termed as primitive groups. The proportion of variance in the exposure factors explained by the instrumental variables for T2DM, FG, FI, and HbA1c are 7.32%, 4.32%, 1.42%, and 5.76%, respectively (Supplementary Table S1).

Each SNP was scanned with PhenoScanner to identify and mitigate potential pleiotropic effects. SNPs correlating with established risk factors of AD, including hypertension within the European population, were filtered out utilizing a threshold of p = 1 × 10⁻⁵ and R²= 0.8. Consequently, 17 SNPs of T2DM, 7 SNPs of FI, 11 SNPs of FG, and 6 SNPs of HbA_1c were excluded (Supplementary Table S3). The remaining SNPs constituted the adjustment groups.

GWAS summary statistics for AD were obtained from FinnGen Release 10, a project mapping genotype-phenotype correlations using Finnish biobank data. This included genome and health data from 1,069 AD individuals and 377,277 controls, characterized by 473,681 genotyped SNPs. Aortic dissection diagnosis was based on hospital records, using ICD-10 codes I71.00, I71.01, I71.09, and ICD-9 code 4410. Endpoint and control definitions are available at https://www.finngen.fi/en/researchers/clinical-endpoints. The UK Biobank database, with 127 cases, served as a validation set (http://www.nealelab.is/uk-biobank).

The genetic data for coronary artery disease were obtained from a meta-analysis of UK Biobank SOFT CAD GWAS (interim release) with CARDIoGRAMplusC4D (Coronary ARtery DIsease Genome wide Replication and Meta-analysis (CARDIoGRAM) plus The Coronary Artery Disease (C4D) Genetics) 1,000 Genomes-based GWAS and the Myocardial Infarction Genetics and CARDIoGRAM Exome (31).

Statistical significance was set at a two-sided 0.05 level, if the p-value is less than the Bonferroni-corrected threshold of 1.11 × 10⁻³, we consider it indicative of a strong correlation. The selection of instrumental variables (IVs) is detailed in the “Exposure” section. For the primary analysis, we combine Wald ratio together in fixed effect meta-analysis, where the weight of each ratio is the inverse of the variance of the SNP-outcome association. The IVW method provides the most precise and robust estimates when three pivotal assumptions regarding instrumental variables are satisfied (32). MR-Presso, MR-Egger, weighted median, simple mode, and Mode based estimate method were used for sensitivity analysis. Heterogeneity in our analysis was assessed using the Cochran Q test, with a p-value of less than 0.05 deemed significant. A Q value substantially exceeding its degrees of freedom indicates evidence of heterogeneity and suggests the presence of invalid instruments (33). MR-Egger regression was utilized to analyze potential unbalanced horizontal pleiotropy, combining the Wald ratio into a meta-regression with an intercept and slope parameter. This approach estimates the causal effect while adjusting for any directional pleiotropy (34). Leave-one-out analyses were used to verify result reliability. Multivariable MR analyses also utilized five methods (IVW, MR-Egger, MR-Lasso, median, and robust), which used to learn about the causal effect of two or more exposures on an outcome (35, 36). Statistical analyses were performed using R software (version 4.3.1) and packages including TwoSampleMR (37), vroom, MVMR (38), MRPRESSO (39), and others.

Standard IVW analysis showed no convincing effect of genetic predisposition to T2DM on AD, with an odds ratio (OR) of 0.92 [95% confidence interval (CI), 0.80–1.04; p = 0.186]. The MR-Egger and weighted median models, which are more robust to directional pleiotropy, showed similar findings (MR-Egger OR = 0.90; 95%CI, 0.67–1.19; p = 0.448; weighted median OR = 0.91, 95%CI, 0.75–1.11; p = 0.359) compared with the IVW model. This consistency in outcomes was maintained even in the adjustment group (IVW OR = 0.87; 95%CI, 0.75–1.00; p = 0.055). The outcomes in the UK Biobank database were similarly concluded (IVW p = 0.290; MR-Egger p = 0.295; weighted median p = 0.412). The results are tabulated for further review (Figures 2, 3; Supplementary Tables S4, S7).

Through Cochran Q test, it was found that there was no significant heterogeneity among SNPs (Supplementary Table S5). Through MR-Egger regression analysis, it is confirmed that there is no significant horizontal pleiotropy in the above analysis data (Supplementary Table S6). A comprehensive sensitivity analysis, facilitated by the leave-one-out method, affirmed the stability of the results, indicating that the exclusion of any individual SNP would not materially influence the overall conclusion (Supplementary Figure S2). For an enhanced visual representation and interpretation of the data, scatter and funnel plots have been provided (Supplementary Figure S1), and the forest plots were presented in Supplementary Figure S2.

Regarding FG, FI, and HbA_1c levels, the standard IVW analysis did not reveal a significant impact on AD susceptibility (FG: OR = 0.91; 95%CI, 0.41–2.01; p = 0.821; FI: OR = 3.08; 95%CI, 0.98–9.65; p = 0.053; HbA_1c: OR = 0.40; 95%CI, 0.14–1.09; p = 0.074) (Figures 2, 3; Supplementary Tables S4, S7).

The Cochran Q test revealed significant heterogeneity within HbA_1c and the primitive group of FG, while other SNP groups displayed no notable heterogeneity (Supplementary Table S5). Additionally, the data analysis did not indicate significant horizontal pleiotropy (Supplementary Table S6). Sensitivity analysis, conducted via the leave-one-out method, suggested the overall conclusion remained stable when any given SNP was omitted from the study (Supplementary Figures S4, S6, S8). Comprehensive visualizations of the study's findings, including scatter plots and funnel plots, are depicted in Supplementary Figures S3, S5, S7, while detailed forest plots are illustrated in Supplementary Figures S4, S6, S8.

A comprehensive two-sample MR study was conducted to identify confounding factors and it was found that SBP, DBP, PP, TG, insomnia, napping, daytime sleepiness, leisure screen time, and smoking initiation were positively correlated with T2DM. Conversely, HDL-C, LDL-C, TC, educational level, annual household income, and the TDI were negatively correlated with T2DM. SBP, DBP, age of smoking initiation, smoking cessation, and environmental stress and adversity were positively correlated with AD, whereas PP and MVPA were negatively correlated with AD (Figure 4; Supplementary Tables S4, S8).

In light of the aforementioned findings, we conducted multivariate MR analysis using SBP, DBP, and PP as confounding factors alongside the exposure variables. Considering that PP is essentially determined by both SBP and DBP, we repeated the multivariate MR analysis using only SBP and DBP as confounding factors with the exposure variables. This approach was adopted to prevent the introduction of bias.

The results demonstrated that T2DM, FG, FI, and HbA1c did not exhibit a significant correlation with AD in both instances of the multivariate MR studies (Figure 5; Supplementary Table S8).

The robustness of the IVs was further assessed against CAD to confirm whether it yielded significant positive associations as previously established. Genetic liability to T2DM and glycemic traits was significantly associated with CAD under IVW models (T2DM: OR = 1.12; 95%CI, 1.08–1.16; p < 0.001; FG: OR = 1.26; 95%CI, 1.04–1.53; p = 0.018; FI: OR = 1.91; 95%CI, 1.31–2.79; p < 0.001; HbA_1c: OR = 1.44; 95%CI, 1.09–1.90; p = 0.010) (Supplementary Table S4). Comprehensive visualizations of the study's findings, including scatter plots and funnel plots, are depicted in Supplementary Figures S9, S10.