Evaluating the Causal Effects of Gestational Diabetes Mellitus, Heart Disease, and High Body Mass Index on Maternal Alzheimer's Disease and Dementia: Multivariable Mendelian Randomization

Abstract

Introduction: Gestational diabetes mellitus (GDM), heart disease (HD) and high body mass index (BMI) are strongly related to Alzheimer's disease (AD) dementia in pregnant women. Therefore, we aimed to determine the total effects of GDM, heart disease, and high BMI on maternal AD dementia. Methods: We used data from the genome-wide association studies of European populations including more than 30,000 participants. We performed two-sample Mendelian randomization (MR) and multivariable MR (MVMR) to systematically estimate the direct effects of GDM, HD, and high BMI on maternal AD and dementia. Multiple sensitivity analyses involving classical MR approaches and expanded MR-pleiotropy residual sum and outlier analysis. Results: In two-sample MR analysis, the inverse-variance weighted method in our study demonstrated no significant causality between GDM and maternal dementia (β = -0.006 ± 0.0026, p = 0.82). This method also revealed no significant causality between high BMI and maternal dementia (β = 0.0024 ± 0.0043, p = 0.57), and it was supported by the MR-Egger regression results, which showed no causal effect of high BMI on maternal Alzheimer's disease and dementia (β = 0.0027 ± 0.0096, p = 0.78). The IVW method showed no significant causal relationship between maternal HD and maternal Alzheimer's disease and dementia (β = -0.05 ± 0.0042, p = 0.117) and MR-Egger regression analysis gave a similar result (β = -0.12 ± 0.0060, p = 0.079). In MVMR analysis, we found no significant causal relationship between GDM, high BMI, or HD and maternal Alzheimer's disease and dementia (p = 0.94, 0.82, and 0.13, respectively). Thus, the MVMR estimates were consistent with our results from the two-sample MR analysis. We confirmed that these results showed no horizontal pleiotropy and enhanced the robustness of our results through multiple sensitivity analyses. Conclusion: In two-sample MR analysis, we found no significant causal relationship between GDM, HD, high BMI and maternal AD and dementia. These results differed from previous observational studies showing HD is a significant predictor of dementia. MVMR analysis supported no significant causal relationship between GDM, HD, high BMI and maternal AD and dementia. Sensitivity analysis broadly increased the robustness of two-sample MR and MVMR analysis results.

Keywords: gestational diabetes mellitus (GDM); heart disease (HD); high body mass index; maternal alzheimer’s disease and dementia; multivariable mendelian randomization (MVMR).

FIGURE 1

Research framework. GDM, gestational diabetes mellitus; BMI, body mass index; HD, heart disease; GWAS, genome-wide association study; IVW, inverse-variance weighted; MR-PRESSO, pleiotropy residual sum and outlier.

FIGURE 2

Results of “leave-one-out” sensitivity analysis of the association between high BMI, HD, and maternal dementia in two-sample MR analysis. (A) Results for the association between high BMI and maternal dementia. (B) Results for the association between HD and maternal dementia. Each black line represents the MR effect after eliminating each SNP at a 95% confidence interval. Each black point represents the median value (that is, the β value) across each black line. The red line at the bottom represents the total MR effect in the sensitivity analysis.

FIGURE 3

Visualization of the MR analysis results for GDM, high BMI, HD, and maternal dementia. (A) Scatter plot for MR analyses of the causal association between GDM and maternal dementia mainly via IVW method. The slope of the blue line in the middle indicates the MR effect using the IVW estimation. (B) Funnel plot of the causal association between high BMI and maternal dementia. We performed analyses by the classical IVW and MR-Egger tests. The two vertical lines (from left to right) in the middle represent the MR effect using IVW method and MR Egger methods, respectively. (C) Scatter plot of the MR effect of HD on maternal dementia determined using multiple methods. Analyses include the IVW and MR-Egger methods, the simple mode, and the weighted median and weighted mode. The slope of each line in the middle indicates the MR effect for each method.