Garfield Victoria, Farmaki Aliki-Eleni, Fatemifar Ghazaleh, Eastwood Sophie V, Mathur Rohini, Rentsch Christopher T, Denaxas Spiros, Bhaskaran Krishnan, Smeeth Liam, Chaturvedi Nish
MRC Unit for Lifelong Health and Ageing at UCL, Institute of Cardiovascular Science, University College London, London, U.K.
Institute of Health Informatics, University College London, London, U.K.
Diabetes. 2021 Oct;70(10):2313-2321. doi: 10.2337/db20-0895. Epub 2021 Feb 25.
We investigated the relationship between glycemia and cognitive function, brain structure and incident dementia using bidirectional Mendelian randomization (MR). Data were from the UK Biobank ( = ∼500,000). Our exposures were genetic instruments for type 2 diabetes (157 variants) and HbA (51 variants) and our outcomes were reaction time (RT), visual memory, hippocampal volume (HV), white matter hyperintensity volume (WMHV), and Alzheimer dementia (AD). We also investigated associations between genetic variants for RT (43 variants) and diabetes and HbA We used conventional inverse-variance-weighted (IVW) MR alongside MR sensitivity analyses. Using IVW, genetic liability to type 2 diabetes was not associated with RT (exponentiated β [expβ] = 1.00 [95% CI 1.00; 1.00]), visual memory (expβ = 1.00 [95% CI 0.99; 1.00]), WMHV (expβ = 0.99 [95% CI 0.97; 1.01]), HV (β-coefficient mm = -2.30 [95% CI -12.39; 7.78]) or AD (odds ratio [OR] 1.15 [95% CI 0.87; 1.52]). HbA was not associated with RT (expβ = 1.00 [95% CI 0.99; 1.02]), visual memory (expβ = 0.99 [95% CI 0.96; 1.02]), WMHV (expβ = 1.03 [95% CI 0.88; 1.22]), HV (β = -21.31 [95% CI -82.96; 40.34]), or risk of AD (OR 1.09 [95% CI 0.42; 2.83]). IVW showed that reaction time was not associated with diabetes risk (OR 0.94 [95% CI 0.54; 1.65]), or with HbA (β-coefficient mmol/mol = -0.88 [95% CI = -1.88; 0.13]) after exclusion of a pleiotropic variant. Overall, we observed little evidence of causal association between genetic instruments for type 2 diabetes or peripheral glycemia and some measures of cognition and brain structure in midlife.
我们使用双向孟德尔随机化（MR）研究了血糖与认知功能、脑结构和新发痴呆之间的关系。数据来自英国生物银行（约500,000人）。我们的暴露因素是2型糖尿病的遗传工具（157个变异）和糖化血红蛋白（HbA，51个变异），我们的结局指标是反应时间（RT）、视觉记忆、海马体积（HV）、白质高信号体积（WMHV）和阿尔茨海默病（AD）。我们还研究了RT的遗传变异（43个变异）与糖尿病和HbA之间的关联。我们使用传统的逆方差加权（IVW）MR以及MR敏感性分析。使用IVW方法，2型糖尿病的遗传易感性与RT（指数化β[expβ]=1.00[95%置信区间1.00；1.00]）、视觉记忆（expβ=1.00[95%置信区间0.99；1.00]）、WMHV（expβ=0.99[95%置信区间0.97；1.01]）、HV（β系数mm=-2.30[95%置信区间-12.39；7.78]）或AD（优势比[OR]1.15[95%置信区间0.87；1.52]）均无关联。HbA与RT（expβ=1.00[95%置信区间0.99；1.02]）、视觉记忆（expβ=0.99[95%置信区间0.96；1.02]）、WMHV（expβ=1.03[95%置信区间0.88；1.22]）、HV（β=-21.31[95%置信区间-82.96；40.34]）或AD风险（OR 1.09[95%置信区间0.42；2.83]）均无关联。IVW显示，排除一个多效性变异后，反应时间与糖尿病风险（OR 0.94[95%置信区间0.54；1.65]）或与HbA（β系数mmol/mol=-0.88[95%置信区间=-1.88；0.13]）均无关联。总体而言，我们几乎没有发现2型糖尿病或外周血糖的遗传工具与中年时期的某些认知和脑结构指标之间存在因果关联的证据。
