MRC Unit for Lifelong Health and Ageing at UCL, University College London, 1-19 Torrington Place, London, WC1E 7HB, UK.
Diabetologia. 2019 Oct;62(10):1891-1900. doi: 10.1007/s00125-019-4949-3. Epub 2019 Jul 29.
AIMS/HYPOTHESIS: Type 2 diabetes, hyperglycaemia and insulin resistance are associated with cognitive impairment and dementia, but causal inference studies using Mendelian randomisation do not confirm this. We hypothesised that early-life cognition and social/educational advantage may confound the relationship.
From the population-based British 1946 birth cohort, a maximum number of 1780 participants had metabolic variables (type 2 diabetes, insulin resistance [HOMA2-IR] and HbA) assessed at age 60-64 years, and cognitive state (Addenbrooke's Cognitive Examination III [ACE-III]) and verbal memory assessed at age 69 years. Earlier-life measures included socioeconomic position (SEP), cognition at age 8 years and educational attainment. Polygenic risk scores (PRSs) for type 2 diabetes were calculated. We first used a PRS approach with multivariable linear regression to estimate associations between PRSs and metabolic traits and later-life cognitive state. Second, using a path model approach, we estimated the interrelationships between earlier-life measures, features of mid-life type 2 diabetes and cognitive state at age 69 years. All models were adjusted for sex.
The externally weighted PRS for type 2 diabetes was associated with mid-life metabolic traits (e.g. HOMA2-IR β = 0.08 [95% CI 0.02, 0.16]), but not with ACE-III (β = 0.04 [-0.02, 0.90]) or other cognitive outcomes. While there was an association between HOMA2-IR and subsequent ACE-III (β = -0.09 [-0.15, -0.03]), path modelling showed no direct effect (β = -0.01 [-0.06, 0.03]) after accounting for the association between childhood SEP and education with HOMA2-IR. The same pattern was observed for later-life verbal memory.
CONCLUSIONS/INTERPRETATION: Associations between type 2 diabetes and mid-life metabolic traits with subsequent cognitive state do not appear causal, and instead they may be explained by SEP in early life, childhood cognition and educational attainment. Therefore, glucose-lowering medication may be unlikely to combat cognitive impairment in older age.
目的/假设：2 型糖尿病、高血糖和胰岛素抵抗与认知障碍和痴呆有关，但使用孟德尔随机化的因果推断研究并未证实这一点。我们假设，早年认知和社会/教育优势可能会混淆这种关系。
在基于人群的英国 1946 年出生队列中，最多有 1780 名参与者在 60-64 岁时评估了代谢变量（2 型糖尿病、胰岛素抵抗[HOMA2-IR]和 HbA），在 69 岁时评估了认知状态（阿登布鲁克认知测验 III [ACE-III]）和言语记忆。早年的测量包括社会经济地位（SEP）、8 岁时的认知和教育程度。计算了 2 型糖尿病的多基因风险评分（PRS）。我们首先使用 PRS 方法结合多变量线性回归来估计 PRS 与代谢特征和晚年认知状态之间的关联。其次，使用路径模型方法，我们估计了早年测量、中年 2 型糖尿病特征与 69 岁时认知状态之间的相互关系。所有模型均调整了性别。
外部加权的 2 型糖尿病 PRS 与中年代谢特征相关（例如，HOMA2-IRβ=0.08 [95%CI 0.02, 0.16]），但与 ACE-IIIβ=0.04 [-0.02, 0.90]）或其他认知结果。虽然 HOMA2-IR 与随后的 ACE-III 之间存在关联（β=-0.09 [-0.15, -0.03]），但在考虑到儿童时期 SEP 和教育与 HOMA2-IR 的关联后，路径模型显示没有直接影响（β=-0.01 [-0.06, 0.03]）。对于晚年的言语记忆，也观察到了同样的模式。
结论/解释：2 型糖尿病与中年代谢特征与随后的认知状态之间的关联似乎并非因果关系，而是可以通过早年的社会经济地位、儿童认知和教育程度来解释。因此，降低血糖的药物可能不太可能在老年时对抗认知障碍。
