Glucose, insulin, and brain health in the UK Biobank: A Mendelian randomisation study

Previous research suggests that HbA _1c and diabetes are unlikely to be causally related to brain health and dementia outcomes. With the availability of better genetic instruments for additional glycaemic-related markers (i.e., insulin resistance (IR), fasting glucose (FPG), fasting insulin (FI) and 2hr post-glucose (2hPG), there is scope to more thoroughly examine how diabetes-related mechanisms are causally related to specific neuroimaging brain health outcomes, and/or all-cause dementia and/or Alzheimer’s dementia (AD). Data were from the UK Biobank (max n = 349,288) with our sample consisting of an average age of 56 years (54% women). We constructed genetic instruments for 2hPG (15 variants), FPG (109 variants), FI (48 variants) and examined their relationships with neuroimaging outcomes such as total brain volume (TBV), hippocampal volume (HV), white matter hyperintensity volume (WMHV), Alzheimer’s dementia (AD) and all-cause dementia. We ran a two-sample MR and used conventional inverse-variance–weighted (IVW), weighted median estimator (WME), as well as MR-Egger intercept for horizontal pleiotropy. We found an association between 2hPG and higher risk of AD in both the IWV (OR: 1.49 [1.17, 1.90]) and WME (OR: 1.54 [1.12, 2.10]) models. However, we did not find that this association replicated in the International Genomics of Alzheimer’s Project (IGAP) data. We also observed an association between higher FI and higher TBV (β 15.25 (3.43, 27.07) in the IVW model. MR analyses of FPG and 2hPG with volumetric brain measures showed no evidence of a causal relationship with TBV, WMHV or hippocampal volume. These findings provide interesting results in how diabetes-related markers are causally related to brain health albeit the precise mechanisms through which postprandial hyperglycaemia increases AD risk and higher insulin preserves brain volumes warrant further research.