The relation of diabetes to memory function

Abstract

Purpose of review:

Research has consistently shown that type 2 diabetes (T2D) is associated with increased risk of all-cause dementia. Because one of the most common clinical presentations of early stage dementia is memory impairment, we examined the relationship of T2D with memory function, using the recently published scientific literature.

Recent findings:

We conducted a structured review to identify studies of “T2D and memory” published since 2015. After review of the 129 articles retrieved, we identified 14 studies meeting the inclusion and exclusion criteria. Among the eight studies with a single assessment of memory function in time (mostly cross-sectional), six found an association of T2D with lower memory function, but mostly in select subgroups of persons. Separately, six studies included repeated measures of memory (longitudinal design). Four out of six longitudinal studies found that T2D was related with a faster decline in memory, while two did not. Among the four studies showing a relation with memory decline, two had sample sizes of 9,000–10,000 persons. Further, three longitudinal studies controlled for hypertension and stroke as co-variates, and results suggested that common vascular risk factors and diseases do not account for the relation. While mechanistic studies clearly support a role for cerebrovascular disease in the relation of T2D with cognition, emerging data suggest that insulin resistance in the brain itself may also play a role.

Summary:

Most, but not all recently published studies suggest that T2D is associated with a lower level and faster decline in memory function. This association does not appear to be fully accounted for by common vascular processes. More research will clarify the mechanisms linking T2D to memory and dementia.

INTRODUCTION

The number of people diagnosed with type 2 diabetes mellitus (T2D) continues to rise globally. In 2015, the global prevalence of T2D was estimated to be 8.8%, and is predicted to rise to about 10.4% by 2040 (1). Older adults (age older than 65 years) are the population with the highest prevalence of T2D (2). Further, T2D has been found to be associated with many long-term complications, especially in older adults (2). This population has been found to have the highest rate of myocardial infarction, visual impairment, and kidney failure (3, 4). T2D is also a well-known risk factor for neurologic complications, including peripheral disease such as peripheral neuropathy (5) and central (brain) disease such as stroke and cognitive impairment. While complications from T2D are often classified into macro- vs micro-vascular complications, T2D may have several types of effects on the brain. Indeed, T2D is associated with macrovascular disease including atherosclerosis, with increased activity of plasminogen activator inhibitor, an enzyme that increases the risk of hypercoagulability (6, 7, 8), microvascular disease including arteriolosclerosis and capillary changes, as well as other possible changes some of which are only beginning to be elucidated, such as brain insulin resistance (8, 9).

Recent research has been exploring the relationship between T2D and cognitive impairment. Many sources of data have demonstrated that individuals with T2D have a two-fold increased risk of developing all-cause dementia (10). Though it can be difficult to definitively determine the underlying dementia etiologies, T2D has been associated with an increased risk of both vascular dementia and Alzheimer’s disease dementia (11, 12, 8). Even in patients without a diagnosis of dementia, T2D has been found to be associated with cognitive impairment and decline (13). Studying associations of T2D with specific cognitive domains may further shed light on potential pathophysiologic mechanisms, whether vascular, neurodegenerative, or otherwise. In particular, with some data suggesting that T2D is associated with Alzheimer’s disease dementia, testing for a relation with memory, the most common early presentation of Alzheimer’s disease, will be informative.

In this review, we will examine the published scientific literature using both cross-sectional and longitudinal cognitive data on the relationship of T2D with memory function (based on objective cognitive testing), and summarize some of the recent work on mechanisms relating diabetes to cognition based on data from the Rush Alzheimer’s Disease Center. Improved understanding the relationship between T2D and memory may help elucidate therapeutic approaches for the treatment and prevention of cognitive impairment and dementia, since T2D is a modifiable risk factor for cognitive impairment and dementia and since treatments are already widely available and in use for diabetes.

METHODS

We conducted a review of the published literature on July 8th 2020, using PubMed. The following search terms in the title and/or abstract were considered: “type 2 diabetes and memory”. Inclusion criteria were: 1) publication date since (and including) 2015 to identify articles in the last five and a half years, and 2) ages 65 + years. We reviewed citation titles and abstracts, and the full manuscript when needed, to exclude publications which were not directly relevant to T2D and memory function or were with small sample size (less than 100 persons).

RESULTS

The literature search yielded 129 citations. After review of each citation, we excluded 115 which were not directly relevant to this review: 9 were interventional studies, 24 were neuroimaging studies, one study had fewer than 100 subjects, and the remainder 75 were on other topics (e.g., mechanisms of cognitive impairment). Thus, 14 articles were included in this review, of which eight had a single measure of memory and the other six had repeated measures (longitudinal design).

Studies of the association of T2D with level of memory function

Eight studies examined the association of diabetes with a single measure of memory function based on cognitive testing (see Table 1). Most studies were cross-sectional in design, and all used biomarkers to define diabetes, usually in blood (14, 15, 16, 17, 18, 19, 20, 21). The first study was among population-based persons with and without T2D, and examined the relation of a particular vascular risk factor, change in 24 hour blood pressure, with a single assessment of memory (14). Persons with T2D had a complex relation between diastolic blood pressure and lower cognition, including memory in particular. Relations were not noted among persons without T2D. These results point to the role of short-term change in blood pressure with lower memory scores among persons with T2D. Studies with cognitive follow-up will provide more clarity on this complex relationship of blood pressure in T2D with memory.

Table 1.

Association of T2D^* with level of memory function

Author, Year	Sample Size (total) and Age	Study Groups	Source of Subjects	Covariates in Model Reported	Results
Spauwen et al., 2015(14)	713 participants Mean age (SD), T2D group : 63.7 (7.0); non-T2D: 58.3 (8.5)	201 with T2D; 512 without T2D	Population-based cohort, Maastricht study	Age, sex, education, smoking, alcohol, waist circumference, cholesterol ratio, triglycerides, antihypertensive medications, lipid-modifying medications, eGFR, cardiovascular disease, depression	In participants with T2D, an inverted U-shaped relationship was found between diastolic blood pressure and memory and information processing speed (immediate word recall: β =−0.0180, p < 0.05; delayed word recall: β =−0.0076, p < 0.01). No relationship of blood pressure and cognition was observed in people without T2D.
Spauwen, et al. 2015 (18)	764 participants, 215 with and 549 without T2D Mean Age (SD) with low skin autofluorescence: 54.9 (8.7) Mean Age (SD) with middle skin autofluorescence: 60.2 (7.3) Mean Age (SD) with high skin autofluorescence: 63.8 (6.9)	254 with low skin autofluorescence; 255 with mild skin autofluorescence; 255 with high skin autofluorescence	Population-based cohort, Maastricht Study	Age, sex, diabetes, educational level, smoking, alcohol, waist circumference, total cholesterol/HDL ratio, triglycerides, lipid-lowering medication use	Skin autofluorescence (SAF) was used as a measurement of advanced glycation end-products (AGEs). In the total group, higher SAF was associated with worse delayed word recall (β=‒ 0.44, p=0.04)
Gao et al. 2015 (19)	492 participants Mean age (SD) T2D with mild cognitive impairment (MCI): 73.92 (2.17) Mean age (SD) T2D with normal cognition: 74.15 (3.92)	246 with T2D and MCI; 246 with T2D and normal cognition	Community-dwelling residents ≥65yo	Sex, age, educational level	Participants with T2DM-MCI did worse on word learning delayed recall, a test of memory, compared to those with T2DM-normal cognition (p=0.015).
Greenbaum et al., 2016 (17)	848 participants Mean Age (SD): 71.98 (4.72)	All 848 had T2D, and BIN1 testing was done	Israel Diabetes and Cognitive Decline sample of patients with T2D	Age, sex, education, disease duration, ancestry	Among participants with T2D, differences in a SNP near the BIN1 gene suggested associations with episodic memory. The TT genotype was associated with lower scores than the CT or CC genotype (F(1,834)=7.664, p=0.00576).
Marseglia et al., 2016 (16)	2305 participants Mean Age (SD) with T2D: 73.9 (9) Mean Age (SD) with pre-T2D: 74 (10) Mean Age (SD) without T2D: 71 (10)	196 with T2D: 144 with uncontrolled T2D (A1C > 6.4%) and 52 with controlled T2D (A1C 5.7–6.4%); 571 with pre-T2D (A1C 5.7–6.4% in T2D-free participants); 1538 without T2D	Population-based cohort, Swedish national study on Aging and Care-Kungsholmen study	Age, sex, education	T2D was not associated with episodic memory (p = 0.418), semantic memory (p =0.379), or a test of working memory (p = 0.119).
Geijsalaers et al. 2017 (21)	806 participants Mean age (SD): 62 (8)	All 806 participants with T2D	Population-based cohort, Maastricht Study	Age, sex, education, A1c, fasting glucose, waist circumference, total/HDL cholesterol ratio, use of lipid-modifying medications, systolic and diastolic blood pressure, use of antihypertensives, depression	Among participants with T2D, demographic factors accounted for 18% of variance in performance on memory function tests. In addition to these factors, fasting insulin and C-peptide did not significantly contribute to the variance in memory function test scores.
Geijsalaers et al. 2017 (20)	2531 participants Mean age (SD) normal glucose metabolism (NGM): 58 (8) Mean age (SD) prediabetes: 62 (7) Mean age (SD) T2D: 63 (8)	1479 with NGM; 386 with prediabetes; 666 with T2D	Population-based cohort, Maastricht Study	Age, sex, education	Participants with T2D did worse on measures of memory than those with NGM (p< 0.05), but participants with prediabetes did not do worse compared to those with T2D (p = 0.035).
van Gemert, et al., 2018 (15)	353 participants Mean Age (SD) with recent diagnosis of T2D: 59 (9) Mean Age (SD) with T2D >12months: 59 (10)	201 with recently diagnosed diabetes:119 with T2D; 110 with known diabetes for five years: 65 with T2D; 42 without diabetes	Population-based cohort, German Diabetes Study	M value, hsCRP, A1C, crystallized intelligence, BMI, age, sex	In participants with a recent diagnosis of T2D (within the last 12 months), verbal memory scores were lower compared to participants without T2D (β= −0.49 (−0.93; – 0.05) p= 0.029).

^*T2D defined by clinical testing and/or biomarkers in all studies listed in table

The next study, by the same research group, included 800 study participants, of which 28% had diabetes (18). Advanced glycation end-products were assessed using skin autofluorescence (18). In the total group of persons with and without diabetes, a higher level of autofluorescence (e.g., more advanced glycation end-products) was associated with lower cognitive function, including on a measure of memory (delayed recall). The core model adjusted for demographics, diabetes, and other factors (see Table 1). An additional analysis with an interaction term of skin autofluorescence and diabetes was not significant, suggesting that the association was independent of the presence of diabetes. Data on advanced glycation end-products measured in plasma did not show an association of pentosidine or other measures with memory. While these results suggest that advanced glycation end-products are associated with lower memory, further research will need to validate these findings and explore the independence of these products from diabetes status.

Another study identified 492 community-dwelling individuals with T2D, of which half had mild cognitive impairment (MCI) and half had normal cognition (19). Not surprisingly, in cross-sectional analyses, participants with T2D-MCI had lower scores on word learning delayed recall tests, than the participants with T2D-normal cognition (see Table 1). However, multivariate regression analyses showed that increasing duration of T2D was associated with lower scores in multiple cognitive domains but not memory. A limitation of the study is that there was no control group of persons without T2D.

In another study, 848 older Israeli Jewish persons were studied (17). All had diabetes and normal cognition (see Table 1). Analyses controlled for demographics and other covariates, such as age, sex, education, disease duration and ancestry. Persons with a specific genotype in the BIN1 gene, had lower levels of episodic memory, based on a composite measures of word list immediate and delayed recall, and recognition. Findings were stronger in a subgroup of study subjects (the Ashkenazi sub-sample). There were no relationships with other cognitive domains including executive function, attention/working memory, and semantic categorization, or with overall cognition. Implications of results are limited by the findings being present only in secondary analyses of a subgroup of persons, the Ashkenazi sub-sample. Furthermore, the study included only persons with T2D. Yet, results do suggest a possible relation of an AD-associated genotype with worse memory in a subset of persons with diabetes, and further study will help clarify these findings.

The Swedish National Study on Aging and Care – Kungsholmen study included 2,305 participants (16). In primary analyses, adjusted for age, sex, and education, there was no evidence for an association of T2D with different types of memory (see Table 1). Results were consistent in stratified analyses of uncontrolled T2D (A1C > 6.4%) and controlled T2D. Strengths of this study include the large sample size, the use of multivariable linear regressions, and detailed data on the diabetes parameters.

Another study examined the relationship between insulin resistance and memory in about 800 participants, all with T2D (21). The results showed that insulin-related variables, including insulin levels and C-peptide levels, did not contribute to variation among scores on a variety of cognitive domains including memory function test scores (see Table 1). Instead, variation in score could be accounted for by variation in demographic factors including age, sex, education, A1c levels, fasting glucose, waist circumference, total/HDL cholesterol ratio, use of lipid-modifying medications, systolic and diastolic blood pressure, use of antihypertensives, and the presence of depression. A strength of this study is the detailed examination of intersecting factors that may lead to variation in cognitive function scores. A limitation of the study may be the generalizability of findings due to the relatively young mean age of participants.

A larger study of 2,531 participants compared T2D with prediabetes, and controls (normal metabolism) on cognitive performance in a variety of domains, including memory (20). This study found that participants with T2D did worse on measures of memory than those with normal glucose metabolism. Notably, participants with prediabetes did not do worse compared to those with T2D (see Table 1). Secondary analyses revealed an interaction between hyperglycemia and the use of antihypertensive medication on memory. It was found that the presence of T2D exacerbated the effects of hypertension on memory test performance, and vice versa. Strengths of this study include the large number of participants. However, a limitation of the study is that participants with T2D had generally well-controlled hyperglycemia and vascular risk factors, and more research is needed in persons with a range in severity of T2D and related conditions.

Another study demonstrated that persons with a recent diagnosis of T2D within the past 12 months showed lower verbal memory scores compared to those who did not have T2D (15). Similar to the Kungsholmen study, this study did not identify significant results in primary analyses (see Table 1). The primary analyses compared participants with a longstanding diagnosis of T2D and participants with a new diagnosis of T2D. A possible limitation was the study’s small sample size of 353 participants. Nonetheless, to aid in analyses, patients were thoroughly categorized based on their metabolic phenotype, using variables such as whole-body insulin sensitivity (M value).

In summary, six of the eight studies which measured memory at a single time point, suggest an association between T2D and lower memory function, while two including a study with a large sample size, did not (16). Of particular note however, is that the associations were mostly only significant in specific subgroups or secondary analyses. Longitudinal studies have the distinguishing strength compared to cross-sectional studies, of decreased chance for measurement and other sources of errors and bias, and thus may better inform on the association of interest.

Studies of the relation of T2D with memory decline

Longitudinal study design provides the opportunity for more information that is biologically relevant. We identified six recent studies which used a longitudinal design to examine the relation of T2D with change in memory function, using repeated measures of cognitive testing at two or more time points. Studies were further separated into two groups, according to whether objective tests were used to define T2D.

T2D defined by clinical test and/or T2D biomarkers

Four longitudinal studies defined T2D by clinical testing such as with the oral glucose tolerance test, or T2D biomarkers such as with the A1c (see Table 2) (14). One study conducted analyses of data derived from a clinical trial on physical function in T2D. The sample size included almost 900 individuals, all with T2D who were stratified by A1C into high, moderate, and low levels (23). While there was a decline in physical function among those with higher A1C levels compared to those with the low levels, there was no observed decline in memory at the 8/9 year visit compared to baseline when comparing high to low A1C (see Table 2). Additional analyses of the same groups were conducted with further stratification by age, and there were no differences in results, suggesting that age does not play a role in the relationship of A1C with memory. One limitation of the study is that longitudinal results were based on only two measures of cognitive function, once at baseline and once at the 8/9 year visit, thus limiting the ability to more accurately model change over time.

Table 2.

Longitudinal studies of the relation of T2D with change in memory function

Author, Year	Sample Size (total), Age, and Follow-up	Study Groups	Source of Subjects	Covariates in Model Reported	Results
T2D defined by clinical testing and/or biomarkers
Beavers et al., 2017 (23)	879 participants Mean age (SD): 58.9 (6.79) Length of follow-up in years: 8	All 879 with T2D: 500 with A1C < 7%; 237 with A1C 7–8%; 142 with A1C >8%	Participants randomly selected among the Look AHEAD trial and Look AHEAD ancillary study, from across the US, all with T2D	Age, sex, race, education, smoking, alcohol, knee pain, metabolic equivalents, BMI, diabetes medications and statins, ancillary year visit, study arm and site	Among a group of participants with T2D, those with higher A1C levels had decreased physical function (p=0.03), but no decline in cognitive function (p=0.41 on global score) including on two measures of memory. Results did not differ when stratified by the mean baseline age.
Marden et al., 2018 (24)	8888 participants Mean age (SD): 67.4 (8.8) Mean follow-up in years: 5.2	1837 with T2D; 7051 without T2D	Population-based cohort of adults age 50+ from the Health and Retirement Study	Age, and several health and social confounders, including hypertension and stroke	People with T2D had faster rates of memory decline, compared to people without T2D (β=−0.04 per decade; 95% CI: −0.06,−0.01). Even among those without T2D, higher A1C level was related to faster memory decline (β=−0.05 per decade; 95% CI: −0.08,−0.03).
Cholerton et al. 2018 (22)	790 participants Mean Age (SD) with T2D: 53.0 (5.9); Mean Age (SD) with impaired fasting glucose:51.7 (5.5); Mean Age (SD) with normal glucose 51.3 (5.5) Mean follow-up in years: 21.1	185 with T2D; 344 with impaired fasting glucose; 261 with normal glucose	American Indian participants in the Strong Heart Study	Baseline age, sex, education, site, fluency of native language, baseline BMI, smoking, alcohol, APOE E4 allele presence	Participants with T2D were found to have a negative association when assessing verbal fluency (β = −3.583, p<0.001) and working memory/processing speed (β = −4.413, p=0.001) when compared to participants with normal fasting glucose.
Callisaya et al. 2019 (25)	705 participants Mean age (SD): 68.2 (7.0) Mean follow-up in years (SD): 4.6 (0.53)	348 with T2D; 357 without T2D	Participants >55yo from the National Diabetes Service Scheme registered in the Cognition and Diabetes in Older Tasmanians (CDOT) longitudinal study	Baseline stroke, hypertension, high cholesterol and obesity, and their interactions with time	Participants with T2D showed a significant decline over time in visual memory (β=0.11; 95% CI: 0.08, 0.14) and verbal memory (β=−0.06; 95% CI: −0.09, −0.02) compared to participants without T2D, but did not show a significant decline in working memory (β=−0.02; 95% CI: −0.05, 0.01).
T2D defined by self-report and/or medication
Bangen et al., 2015 (26)	1493 participants Mean age (SD) with T2D: 75.4 (5.8) Mean Age (SD) without T2D: 76.3 (6.7) Mean follow-up in years (SD): 6.05 (3.02)	378 with T2D; 1115 without T2D	Washington Heights-Inwood Columbia Aging Project (WHICAP) cohort	Age, sex, education, ethnicity, ApoE genotype	Compared to those without T2D, participants with T2D had lower baseline levels of cognitive function. However, participants with and without T2D showed similar rates of decline in memory (β = − 0.020, p = 0.603), language, speed/executive functioning, and visuospatial abilities.
Demakakos et al., 2017 (27)	10524 participants Mean Age (SD) without T2D with elevated depressive symptoms: 64.4 (9.9) Mean Age (SD) with T2D without elevated depressive symptoms: 68.4 (9.1) Mean Age (SD)without T2D with elevated depressive symptoms: 65.9 (10.8) Mean Age (SD) with T2D with elevated depressive symptoms: 67.4 (9.3) Length of follow-up in years: 10	8275 without T2D, with elevated depressive symptoms; 554 with T2D, without elevated depressive symptoms; 1526 without T2D, with elevated depressive symptoms; 169 with T2D, with elevated depressive symptoms	Community-dwellers aged ≥50 years in 2002–2003 from the English Longitudinal Study of Ageing	Age, sex, marital status, self-reported chronic conditions (heart disease, stroke, hypertension, chronic lung disease), education, occupational class, physical activity, smoking, alcohol consumption, BMI, adjustment for exposure*time interaction term	Participants with T2D have faster decline in memory, compared to participants without T2D, as determined by a word recall test. Memory declined faster among participants with both T2D and depressive symptoms, in mid-adult life, but not late-adult life (over age 65) (β= − 0.27, 95% CI, − 0.45 to − 0.08, per study wave).

The Health and Retirement Study included a sample size of almost 9,000 participants, of which 21% had T2D (24). In this observational study, participants with T2D showed a faster decline in memory per decade, when compared to those without T2D (see Table 2). Analyses controlled to hypertension and stroke, suggesting that results were not accounted for by these other vascular factors. Furthermore, the study showed higher A1C levels were associated with a faster decline in memory in participants with T2D compared to those without T2D. Strengths of this study include the large sample size and power to detect associations, as well as complementary analyses with consistent findings, including stratifying within the T2D sample. However, the memory measure was a composite of the word recall test and a proxy questionnaire, and did not allow to disentangle specific memory functions more precisely.

In a recent study of 790 American Indian participants (22) originally enrolled in the Strong Heart Study, participants with T2D were found to have lower subsequent scores (mean follow-up 21 years) on working memory/processing speed (executive function) compared to those with normal glucose metabolism (see Table 2). One strength of this study is the sizable cohort in a group that is not often represented in the literature. One of the study weaknesses is that only a single measure of memory was collected.

A study of Tasmanian persons with and without T2D, used oral glucose tolerance testing to classify subjects based on biomarker data. Those with T2D had a faster decline in visual memory and verbal memory, but not working memory, when compared to participants without T2D (25). In addition, there was no significant change in level of brain atrophy on neuroimaging in patients with T2D, compared to those without. Further study will need to investigate what mechanisms are involved in linking T2D with memory, using neuroimaging and other tools. Indeed, neuroimaging is useful to identify brain structural and functional abnormalities, including white matter changes and brain infarcts on magnetic resonance imaging (MRI), which are both commonly found in T2D. Much research, mostly cross-sectional in design, uses neuroimaging to explore these and other mechanistic pathways which may link T2D to brain function.

T2D defined by self-report and/or medication

Two of the longitudinal studies defined T2D based on participants’ self-report or medication data (see Table 2). The identification of T2D in these studies is not as rigorous as in above discussed studies, with increased possibility for misclassification of subjects and less opportunity to explore the role of disease severity and other factors.

Memory decline was examined in a study of about 1,500 participants with and without T2D (26). First, compared to those without T2D, participants with T2D showed an overall lower baseline level of cognitive function in a variety of domains including memory (see Table 2). Next however, over a mean follow-up period of six years, those with T2D demonstrated similar rates of decline in memory compared to participants without T2D. Results were similar for decline in other cognitive domains, with no significant differences noted among those with and without T2D. Notable strengths of this study include the large sample size and the repeated measure of cognition approximately every 18 months, with an average of 4 measures, which increases the accuracy of the estimate for the trajectory of change in cognition.

A very large study of more than 10,000 community-dwelling participants, showed that participants with T2D demonstrated a faster decline in memory on a test of word recall, compared to those without T2D (27). Analyses accounted for a range of self-reported covariates, including hypertension and stroke, among others (see Table 2). Furthermore, in secondary stratified analyses, those with both T2D and depressive symptoms in mid-adult life (50–64 years) demonstrated a faster decline in memory compared to those with only T2D or depressive symptoms, and compared to late-adult life. Strengths of the study include the large sample size and the community-based sample, enhancing the generalizability of the findings. Also, the memory assessments were conducted every two years over a 10 year follow-up period, strengthening the ability to model change in cognition. Further research is necessary to separate confounders associated with depressive symptoms, including antidepressant medication use.

In summary, four of six longitudinal studies showed that T2D is related to a faster decline in memory function, even when accounting for common vascular risk factors and diseases related to cognition such as hypertension and stroke. Confidence in these results are heightened by some of the robust study design and methods used in the studies. In particular, two of the longitudinal studies had a sample size in the range of 9,000–10,000 persons and several were population or community-based. In addition to all studies using two or more measures of memory over time to model change in cognition, three of the five studies used clinical or biomarker data to characterize T2D, further strengthening confidence in the validity of the outcome and predictor variables examined.

Mechanistic studies on T2D, cognition, and neuropathology from the Rush Alzheimer’s Disease Center

Using data from epidemiologic studies of aging conducted at the Rush Alzheimer’s Disease Center, we have leveraged detailed cognitive data across several cognitive domains including three memory domains, to help elucidate brain systems affected by T2D. We have shown that T2D is cross-sectionally associated with lower semantic memory (28) and also other memory and cognitive domains (29), including among both White and Black community-based research participants (30). In keeping with the recent literature, we did not find that controlling for hypertension changed our results. Further, while we did not find that hypertension modified the association, we found that other vascular factors, and smoking in particular, may play a role (28). In longitudinal studies, T2D was related to faster cognitive decline, and particularly in a composite measure of perceptual speed, but not in episodic memory or other cognitive domains (29, 31).

To further examine possible underlying pathophysiologic mechanisms for cognitive impairment in T2D, we leveraged postmortem neuropathologic biospecimens (human brain tissues from persons with and without T2D) and data from the Rush cohort studies which have high autopsy rates and internal validity. One plausible mechanism linking T2D to cognition is through cerebrovascular disease, since stroke is a well-established risk factor, and brain infarcts are a major etiology, for cognitive impairment and dementia. In keeping with the literature on T2D and cerebrovascular disease, we have consistently found evidence that T2D is associated with brain infarcts, including gross and subcortical infarcts but also lacunar infarcts (31, 32, 33). Further, A1C was associated with brain infarcts, in particular gross infarcts (34). Yet, cerebrovascular disease may not account for all of the relation of T2D to cognition. AD-related mechanisms are another possibility in the mechanistic pathway linking T2D with cognition and dementia. However, we have found no clear association of T2D with AD pathology, including in the mesial temporal lobe (31, 32, 34, 35). While mechanisms linking T2D to cognition are complex (36), recent literature point to a possible role of brain insulin resistance (37). Indeed, earlier work which included subjects from a Rush cohort showed that insulin receptor substrate 1 (IRS1) was associated with AD pathology and amyloid in particular, as well as with brain function and episodic memory specifically (38). More recently, using a novel ex vivo stimulation paradigm with insulin in human postmortem tissue (39), we further examined the relation of insulin resistance and related markers to AD pathology and cognition. In 150 deceased participants from the Religious Orders Study with and without T2D, we found that AKT phosphorylation by ELISA in middle frontal gyrus cortex was associated with more AD pathology overall, and with both amyloid and tangles separately (40). Also, AKT was associated with lower cognitive function proximate to death, including on a global score and also on domain scores of working memory and episodic memory separately (40). More work is underway to examine these and other factors in insulin resistance, such as the role of adipokines, as well as associations with other common neuropathlogies of aging and dementia including cerebrovascular diseases.

CONCLUSIONS

Recent literature informs on the relationship between T2D and memory function. Ten of the 14 recent studies showed associations of T2D with lower memory function and with memory decline, while the other four did not. These ten studies included six with only one measurement of memory, and four with repeated measures over time allowing to model change in cognition. This review benefits from the use of an inclusive search strategy to identify all recent studies on the topic in older persons. Another strength is the categorization of studies by design, facilitating our ability to interpret the results.

The published studies shared strengths. Many of them used a community-based cohort, while several more used a population-based cohort. This increased the generalizability of the findings. Another strength that allows for greater generalizability is the large sample sizes. All of the studies included sample sizes of 350 persons and more, with many including 1,000 or more and two longitudinal studies having about 9,000–10,000 persons each. Further, longitudinal studies decrease the possibility for selection, information, and other sources of bias and error.

The association of T2D with memory function is likely influenced by the age of the groups studied, among many other factors. The studies included in this review varied in participants’ mean ages (e.g., “young-old” vs “old-old”), and challenges our ability to compare results. Another limitation is the specific cognitive assessment used to measure memory function. Most studies used specific memory tests which allow a more direct observation of the relation of T2D with memory and provide more insight into the question. However, some studies used a single memory test, or a test of overall cognition which included a component of memory. Also, factors related to T2D such as duration and severity of disease, as well as medications used and level of glycemic control achieved, also likely play a role in the association of T2D with memory and needs further elucidation.

Study design issues will address recruitment of research participants representative of the general population, and rigorous scientific methodologies such as high follow-up and complete data.

While many studies in this review controlled for co-variates such as vascular risk factors including hypertension, and vascular disease including stroke, further inquiry into these factors as effect modifiers will provide valuable insight, given the known relation of T2D with cerebrovascular disease. More research is needed, particularly on mechanisms linking T2D to cognition and dementia (33), before recommendation for change in clinical care of T2D can be considered for the treatment and prevention of cognitive impairment or for use of T2D therapies for dementia (41). Also, a more refined characterization of T2D patients at risk for cognitive impairment and dementia is needed (36), particularly given recent research showing that T2D may cluster into five groups (of which, three are related to insulin resistance), using a soft clustering of genetic loci approach (42). Indeed, modern tools and methods, such as practical biomarkers (e.g., in blood) using epigenetic inquiry and other approaches (e.g., computational neurobiology), show promise in moving the field of metabolism and dementia forward. The application of precision medicine will facilitate in recognizing individuals with T2D who are most at risks for complications (e.g., brain dysfunction) and most likely to respond to treatments (e.g., preventive and symptomatic), and is expected to result in improved health outcomes (41).