For several years, based on several large-scale cohort and twin studies of people well into older age, type 2 diabetes has been recognized to be strongly associated with a twofold increase in the risk of a “new” or incident dementia (1). The causality of this association and putative underlying pathways still remain uncertain.

Several features such as hyperglycemia, insulin resistance, inflammation, and vascular endothelial dysfunction could theoretically mediate a causal link between type 2 diabetes and dementia (2), and their putative effects may also be influenced by several modifiers (Fig. 1). It is plausible that these pathways could lead to brain injury indirectly via cerebrovascular disease, or by directly affecting the neurons (gray matter) and their connections (white matter), or both. The attractiveness of this overall causal hypothesis lies in the potential to find targeted interventions to reduce the risk of future dementia. Thus, it is important to generate strong evidence either supporting or disproving causality in the associations of type 2 diabetes and its related features with measures of brain health.

In their article in this issue of Diabetes, Garfield et al. (3) attempt to further clarify causality in the relationship of type 2 diabetes (and specifically its primary feature glycemia) with indicators of dementia risk (cognitive function, brain MRI) and a diagnosis of Alzheimer dementia (AD). They adopted Mendelian randomization to tackle confounding and reverse causation, which are hard to avoid in traditional cohort studies. They used the rich data collected on a large scale in the UK Biobank, a population-based study of adults at midlife (n = 349,326, aged 40–69 years), with a diabetes prevalence of ∼4% (n = 14,010) and relatively well-controlled HbA1c (mean 35.5 mmol/mol, ∼5.5%). They used available outcome data in specific cognitive functions (reaction time, visual memory in all participants), brain MRI (hippocampal volume and white matter hyperintensity volume in ∼32,000 participants), and primary care recording of AD based on ICD-10 codes (746 participants). Exposures of interest were 157 independent genetic variants for diabetes explaining ∼1.5% of the sample variance in people with diabetes, and 51 single nucleotide polymorphisms for HbA1c explaining 2.8% of the sample variance in HbA1c. In addition, they examined reverse directionality of relationships between 43 single nucleotide polymorphisms related to reaction time and diabetes/HbA1c. Their primary results showed no forward associations of either diabetes or HbA1c with reaction time, visual memory, hippocampal and white matter hyperintensity volumes, and AD diagnosis. There was also no evidence for a reverse directional association between reaction time and either diabetes or HbA1c. They concluded a lack of evidence to support causality of the relationships of diabetes and glycemia with the outcome measures that they used.

The conclusions should be viewed in light of certain limitations relating to the sample and measurement, and the authors acknowledge some of these in the article. Overall prevalence of diabetes in the sample was low, and with low variation in peripheral glycemia. Peripheral glycemia itself may not capture the phenomenon of brain insulin resistance. The range of cognitive tests used was small and may have lacked sufficient sensitivity to capture variation in a relatively young sample such as that in the UK Biobank. The use of more sensitive imaging markers for dementia in this scenario is therefore important. Although the authors used hippocampal volume and white matter hyperintensity volumes as reasonable intermediate measures of dementia, these measures may also be relatively insensitive in middle age. Changes in brain macrostructure are likely to become evident with impact in late old age, with brain amyloid or tau accumulation beginning earlier, and changes in white matter microstructure and brain metabolism potentially even earlier (4). Imaging modalities such as diffusion MRI, functional MRI, and positron emission tomography may be more valuable at such earlier ages (4). For example, in late middle-aged twins discordant for type 2 diabetes, substantial differences in functional MRI were found in the absence of measurable differences in several measures of cognitive tests or brain structure (5). Cognitive decline could also not be captured in the current study, as sufficient longitudinal cognitive data were not available. Finally, diagnostic accuracy of a specific diagnosis of AD is at best moderate for data reported from primary care, death records, and self-report, but somewhat better for all-cause dementia (6). The relationship between type 2 diabetes and neurodegeneration has been shown to be less likely related to amyloid-related pathways (considered essential for AD pathology) than with tau-related pathways (more nonspecific and related to multiple neurological disease processes) (7). Thus, it may have been interesting to see how the results of the study would have been affected by using a more general diagnosis of all-cause dementia to capture the broad impact of diabetes and glycemia gene variants, as has been demonstrated recently (8).

While this new Mendelian randomization study adds to the puzzle of causality in this field, it should also be viewed in context of an important issue complicating the understanding of dementia. Dementia is characterized by the presence of several cerebral pathologies that cumulatively result in neurodegeneration and expression of cognitive symptoms (9), with multiple direct or indirect causal mechanisms, each with small individual effects, acting or interacting at various stages of life, and modified by protective factors such as brain or cognitive reserve from early life (10). This may explain the paucity of evidence for cognitive benefit from trials in people with diabetes for individual targets such as glycemia (11) or blood pressure (12). It is also possible that a direct causal link does not exist between type 2 diabetes or dementia, and if this were the case, one could speculate on why there is such a consistent association between the two syndromes. The effect of diabetes may be indirect, for example, by leading to brain infarcts, which are known to lower the threshold for clinical expression of dementia (13). It is also possible that the two syndromes are merely more likely to coexist over the life span and that their association can be explained by unmeasured confounding, but this is best clarified using life-course approaches such as in birth cohorts. Nevertheless, future studies at earlier stages of life may benefit greatly from technologies that assist in the highly sensitive measurement of exposures (e.g., continuous glucose monitoring), outcomes (e.g., brain metabolism and function), and modifiers (e.g., functional imaging measures of brain reserve) while dynamically manipulating exposure states such as glycemia. Such studies would allow us to get closer to an understanding of the brain’s response to diabetes-related insults and to developing ways to protect it.

See accompanying article, p. 2313.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

1.
Biessels
GJ
,
Staekenborg
S
,
Brunner
E
,
Brayne
C
,
Scheltens
P
.
Risk of dementia in diabetes mellitus: a systematic review
.
Lancet Neurol
2006
;
5
:
64
74
2.
Biessels
GJ
,
Despa
F
.
Cognitive decline and dementia in diabetes mellitus: mechanisms and clinical implications
.
Nat Rev Endocrinol
2018
;
14
:
591
604
3.
Garfield
V
,
Farmaki
A-E
,
Fatemifar
G
, et al
.
The relationship between glycemia, cognitive function, structural brain outcomes and dementia: a Mendelian randomization study in the UK Biobank
.
Diabetes
2021
;
70
:
2313
2321
4.
Sperling
RA
,
Jack
CR
 Jr
,
Black
SE
, et al
.
Amyloid-related imaging abnormalities in amyloid-modifying therapeutic trials: recommendations from the Alzheimer’s Association Research Roundtable Workgroup
.
Alzheimers Dement
2011
;
7
:
367
385
5.
Wood
AG
,
Chen
J
,
Moran
C
, et al
.
Brain activation during memory encoding in type 2 diabetes mellitus: a discordant twin pair study
.
J Diabetes Res
2016
;
2016
:
3978428
6.
Wilkinson
T
,
Schnier
C
,
Bush
K
, et al.;
Dementias Platform UK and UK Biobank
.
Identifying dementia outcomes in UK Biobank: a validation study of primary care, hospital admissions and mortality data
.
Eur J Epidemiol
2019
;
34
:
557
565
7.
Moran
C
,
Beare
R
,
Phan
TG
,
Bruce
DG
,
Callisaya
ML
;
Alzheimer’s Disease Neuroimaging Initiative (ADNI)
.
Type 2 diabetes mellitus and biomarkers of neurodegeneration
.
Neurology
2015
;
85
:
1123
1130
8.
Benn
M
,
Nordestgaard
BG
,
Tybjærg-Hansen
A
,
Frikke-Schmidt
R
.
Impact of glucose on risk of dementia: Mendelian randomisation studies in 115,875 individuals
.
Diabetologia
2020
;
63
:
1151
1161
9.
Neuropathology Group of the Medical Research Council Cognitive Function and Ageing Study (MRC CFAS)
.
Pathological correlates of late-onset dementia in a multicentre, community-based population in England and Wales
.
Lancet
2001
;
357
:
169
175
10.
Biessels
GJ
,
Deary
IJ
,
Ryan
CM
.
Cognition and diabetes: a lifespan perspective
.
Lancet Neurol
2008
;
7
:
184
190
11.
Tuligenga
RH
.
Intensive glycaemic control and cognitive decline in patients with type 2 diabetes: a meta-analysis
.
Endocr Connect
2015
;
4
:
R16
R24
12.
Murray
AM
,
Hsu
FC
,
Williamson
JD
, et al.;
Action to Control Cardiovascular Risk in Diabetes Follow-On Memory in Diabetes (ACCORDION MIND) Investigators
.
ACCORDION MIND: results of the observational extension of the ACCORD MIND randomised trial
.
Diabetologia
2017
;
60
:
69
80
13.
Abner
EL
,
Nelson
PT
,
Kryscio
RJ
, et al
.
Diabetes is associated with cerebrovascular but not Alzheimer’s disease neuropathology
.
Alzheimers Dement
2016
;
12
:
882
889
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/content/license.