Skip to main content
NCBI home page
Alzheimer's & Dementia logoLink to Alzheimer's & Dementia
. 2023 Oct 13;20(2):975–985. doi: 10.1002/alz.13480

Heterogeneous treatment effects of metformin on risk of dementia in patients with type 2 diabetes: A longitudinal observational study

Huilin Tang 1, Jingchuan Guo 1,2,✉, C Elizabeth Shaaban 3,4, Zheng Feng 5, Yonghui Wu 5, Tanja Magoc 6, Xia Hu 7, William T Donahoo 8, Steven T DeKosky 9,10, Jiang Bian 5
PMCID: PMC10917005  NIHMSID: NIHMS1929890  PMID: 37830443

Abstract

INTRODUCTION

Little is known about the heterogeneous treatment effects of metformin on dementia risk in people with type 2 diabetes (T2D).

METHODS

Participants (≥ 50 years) with T2D and normal cognition at baseline were identified from the National Alzheimer's Coordinating Center database (2005–2021). We applied a doubly robust learning approach to estimate risk differences (RD) with a 95% confidence interval (CI) for dementia risk between metformin use and no use in the overall population and subgroups identified through a decision tree model.

RESULTS

Among 1393 participants, 104 developed dementia over a 4‐year median follow‐up. Metformin was significantly associated with a lower risk of dementia in the overall population (RD, –3.2%; 95% CI, –6.2% to –0.2%). We identified four subgroups with varied risks for dementia, defined by neuropsychiatric disorders, non‐steroidal anti‐inflammatory drugs, and antidepressant use.

DISCUSSION

Metformin use was significantly associated with a lower risk of dementia in individuals with T2D, with significant variability among subgroups.

Keywords: dementia, diabetes, heterogeneous treatment effect, metformin

1. BACKGROUND

Alzheimer's disease (AD) and AD‐related dementias (ADRD) are characterized by a decline in cognitive function 1 and affect ≈ 55 million individuals worldwide. 2 In the United States, ≈ 6.5 million older adults had a diagnosis of dementia in 2020, and the number is predicted to increase to 14 million by 2060. 3 , 4 Dementia was the sixth leading cause of death among adult Americans, with estimated annual costs of $305 billion in 2020. 3 , 5 However, no effective interventions have been found to improve function in a clinically meaningful way; thus, finding such therapeutic strategies is important and urgently needed. 6

Type 2 diabetes (T2D) is a well‐known risk factor for dementia. 7 , 8 Several hypothesized mechanisms by which diabetes increases dementia risk include brain insulin resistance, hyperinsulinemia, inflammation, oxidative stress, vascular changes, and impaired amyloid metabolism. 9 , 10 , 11 Thus, glucose‐lowering drugs (GLDs) might be potential treatments for dementia. Metformin is the most commonly used GLD as the first‐line treatment for T2D. 12 Numerous population studies have evaluated the association between metformin and the risk of dementia, yet they produce conflicting results. 13 , 14 , 15 While some studies found a lower risk of dementia associated with metformin, 13 , 14 , 16 not all studies supported this finding. 15 A recent meta‐analysis of 14 population‐based cohort studies showed that metformin was significantly associated with a reduced risk of all‐cause dementia, but with statistically significant heterogeneity between studies (P < 0.001), 17 indicating that certain individuals or subgroups may experience varying treatment effects from metformin. Although some studies have conducted subgroup analyses, such conventional “one‐variable‐at‐a‐time” subgroup analysis is susceptible to erroneous inferences and has limited ability to address intricate interactions among multiple patient characteristics that are believed to influence a patient's response to the treatment. 18

Doubly robust estimation is an extension of the inverse probability of treatment weighting (IPTW), also known as the augmented IPTW method. This approach combines the propensity score model (e.g., IPTW) and the outcome regression model to estimate the causal effect of an exposure (e.g., metformin) on a specific outcome (e.g., dementia risk). 19 , 20 The key advantage of doubly robust estimation lies in its ability to offer an unbiased estimation of the average treatment effect (ATE) if either the propensity score model or the outcome regression model is correctly specified. 20 Thus, if either the propensity score model or the outcome model is incorrectly specified, an unbiased estimate of the ATE will still be obtained. Moreover, the use of doubly robust learning (which involves using machine learning approaches) enables us to perform inference on the best conditional ATE (CATE) function based on a subset of relevant features that may be contributing to heterogeneity, thus identifying the subgroups and estimating the optimal treatment effect. 21 Therefore, in this study, we used doubly robust learning to estimate ATE and explore the heterogeneous treatment effects (HTEs) of metformin on the risk of dementia in the population with T2D.

2. METHODS

2.1. Data source

This retrospective cohort study was performed using the National Alzheimer's Coordinating Center (NACC) dataset, which was initiated by the National Institute on Aging (NIA) in 1999 and encompassed participant information from all of the NIA's US Alzheimer's Disease Research Centers (ADRCs). This study used data from the NACC Uniform Data Set (UDS) between September 2005 and June 2021. The ADRCs collected data from participants and their study partners using prospective, longitudinal clinical examinations by trained clinicians and clinic personnel. The cognitive performance of study participants at the initial visit and the annual follow‐up visits were assessed by trained clinicians. Other detailed information on personal characteristics, demographics, health behaviors, current health conditions and disease history, medication use, functional abilities, and depressive symptoms were also collected. Informed consent forms were approved by the individual ADRCs’ institutional review boards (IRBs), and consent was obtained from participants and study partners. Research studies using the NACC database were also approved by the University of Washington IRB.

2.2. Study population

This study included individuals who met the following criteria: (1) they had at least two visits at the corresponding ADRCs between September 2005 and June 2021, (2) they were aged ≥ 50 years at the initial visit (baseline), (3) they had a report of diabetes at baseline, and (4) they had normal cognition at baseline. Individuals were excluded if they met any of the following conditions: (1) taking a US Food and Drug Administration (FDA)‐approved medication for AD symptoms; and (2) having a diagnosis of type 1 diabetes, diabetes insipidus, latent autoimmune diabetes, or gestational diabetes. The patients were followed from the date of the first visit until they were diagnosed with dementia, dropped out, died, or until the end of follow‐up.

Normal cognition 22 was defined by (1) Clinical Dementia Rating (CDR) = 0 (no dementia); 23 (2) no deficits in activities of daily living directly attributable to cognitive impairment; and (3) no evidence of objective cognitive impairment defined as performance falling less than 1.5 standard deviations below the age‐adjusted normative mean on neuropsychological tests assessing language, attention, memory, executive, and visuospatial functioning (this domain was added in 2015). 24 , 25

2.3. Outcome definition and measurement

The outcome of interest in this study was the initial diagnosis of dementia. Dementia 22 was defined as meeting the criteria for AD 26 or other dementias 27 , 28 , 29 , 30 , 31 , 32 defined as (1) objective cognitive impairment (i.e., performance falling greater than 1.5 standard deviations below the age‐adjusted normative mean) in at least two cognitive domains (i.e., memory, language, attention, executive, and visuospatial functioning) and (2) cognitive impairment contributes directly to impaired activities of daily living.

2.4. Primary predictor: metformin exposure

Participants were asked to report or bring all medication (including prescription, non‐prescription, and over‐the‐counter medication) and supplements being used currently or within 2 weeks before baseline and follow‐up visits. A trained ADRC staffer or physician documented the medication use using a standard medication inventory. The participants were classified as metformin users if they reported taking metformin during the baseline visit.

2.5. Covariates

We adjusted for covariates (including demographic factors, health behaviors, comorbidities, medication use, and genetic biomarkers) that were shown to be related to dementia based on findings in previous studies 33 , 34 , 35 and clinical experience. The demographic characteristics included age (≥ 65 vs. < 65 years), sex (male vs. female), education level (≥ college school level vs. < college school level), race (Black vs. non‐Black), ethnicity (Hispanic vs. non‐Hispanic), and family history of dementia (yes/no). The health behaviors (based on self‐report) that affect cognition were obesity (body mass index ≥ 30 vs. < 30 kg/m2), smoking status (ever smoking vs. never smoking), alcohol abuse (yes vs. no), and drug abuse (yes vs. no). Comorbidities were classified as presence versus absence of a self‐reported history of diseases including cardiovascular disease (including heart attack/cardiac arrest, angioplasty/endarterectomy/stent, cardiac bypass procedure, pacemaker and/or defibrillator, congestive heart failure, atrial fibrillation, angina, heart valve replacement or repair, and other cardiovascular diseases), cerebrovascular disease (including stroke and transient ischemic attack), neurological diseases (including Parkinson's disease, other Parkinson's disease disorders, traumatic brain injury, seizures, and other neurological conditions), neuropsychiatric disorders (including post‐traumatic stress disorder, bipolar disorder, schizophrenia, depression, anxiety, obsessive‐compulsive disorder, developmental neuropsychiatric disorders, and other psychiatric disorders), hypercholesterolemia, hypertension, and vitamin B12 deficiency. Medication use (classified as yes vs. no based on self‐reported medication use at the baseline visit) included lipid‐lowering drugs; anti‐hypertensive drugs; non‐steroidal anti‐inflammatory drugs (NSAIDs); antidepressants; antipsychotic drugs; anti‐Parkinson's drugs; anxiolytic, sedative, or hypnotic agents; sodium‐glucose co‐transporter 2 (SGLT2) inhibitors; dipeptidyl peptidase 4 (DPP‐4) inhibitors; glucagon‐like peptide‐1 receptor agonists (GLP‐1 RAs); alpha‐glucosidase inhibitors; sulfonylureas; thiazolidinediones; and insulin. We also controlled for apolipoprotein E ε4, the strongest known genetic risk factor for late‐onset AD and cognitive impairment. 35

RESEARCH IN CONTEXT

  1. Systematic review: We systematically reviewed the literature using PubMed. Current evidence regarding the association between metformin use and the risk of dementia remains inconclusive. Furthermore, little is known about the heterogeneous treatment effects of metformin on the risk of dementia in patients with type 2 diabetes (T2D).

  2. Interpretation: Metformin use was significantly associated with a lower risk of dementia compared to no use. We identified four heterogeneous treatment effect subgroups based on the following three important factors: neuropsychiatric disorders, non‐steroidal anti‐inflammatory drugs, and antidepressant use.

  3. Future directions: Our findings will support personalized decision making regarding the use of metformin in individuals with T2D. Future research should confirm our findings and investigate the generalizability of these findings to individuals without T2D.

2.6. Statistical analysis

We performed descriptive analyses to describe the sociodemographic, health behaviors, comorbidities, and medication use of participants at baseline. The data were reported as the number of participants (and percentage). We assessed the balance of all covariates between metformin use and no‐use groups using standardized mean differences (SMD) before and after IPTW. An SMD of ≤ 0.10 was considered indicative of a negligible imbalance in the baseline covariates. 36

We estimated the CATEs of metformin on the risk of dementia following the doubly robust learning framework. 21 In this framework, we developed three models, including the propensity score model, an outcome regression model, and a doubly robust model. For the propensity model, we ran a classification to predict the probability of receiving the metformin for each subject using a multivariate logistic regression of covariates. For the outcome regression model, we estimated dementia outcomes by applying a multivariate logistic regression model including treatment and covariates. A cross‐validated model based on mean squared error (MSE) was used to select the optimal hyperparameters for the above two models. The best model for the propensity score model and the outcome regression model, respectively, were chosen and incorporated into the final doubly robust model. In the final model, we applied a debiased Lasso regression (“SparseLinearDRLearner” within EconML package) to estimate CATEs and individualized treatment effects (ITEs, treatment effect on person level). 21 , 37 The treatment effects were measured using the absolute risk differences (RDs) with a 95% confidence interval (CI) for the risk of dementia between metformin use and no‐use groups. Negative values (e.g., ITEs) indicated a reduced absolute risk of dementia (benefit from metformin), whereas positive values (e.g., ITEs) indicated increased absolute risk of dementia (harm from metformin).

The doubly robust model also yielded coefficients for each covariate, and a single decision tree model was deployed to pick the most important factors and discover the drivers of heterogeneity. To ensure reliable results, the samples in each leaf (subgroup) of the decision tree were required to be > 5% of the overall study sample. 38 The decision tree model was trained to maximize the treatment effect difference between leaves (subgroups), thus identifying the subgroup of patients that respond to metformin differently from other subgroups.

In this study, the sample was randomly divided into training (70%) and test (30%) sets. The training set was applied to train the machine learning models and optimize the hyperparameters. To validate the models, we used MSE as the evaluation metric, conducting a 5‐fold cross‐validation for internal validation in the training set. Additionally, the performance of the model was assessed using the testing set for external validation.

To account for the potential influence of other classes of GLD, we conducted further analyses according to whether other GLDs were added to metformin. Specifically, we compared the risk of dementia of the following two subgroups: individuals on metformin monotherapy and those on metformin combination therapy (i.e., metformin in combination with other GLDs), separately against non‐users. Furthermore, we considered the following two sensitivity analyses: (1) restricting the individuals aged ≥ 60 years; (2) using a valid doubly robust estimation method, which was developed through a SAS procedure (Proc causaltrt) and was specifically designed to estimate the CATEs rather than HTEs. 37 All analyses were performed using SAS version 9.4 (SAS Institute Inc.) and Python version 3.7 (Python Software Foundation).

3. RESULTS

Among the 43,999 participants included in the NACC dataset from September 2005 to June 2021, we included 1393 participants with T2D and normal cognitive function at baseline. The participant selection and the reasons for exclusion are shown in Figure 1. Seven hundred fifty‐four participants (54.1%) reported taking metformin (metformin use group), while 639 (45.9%) were not taking metformin (no‐use group). Among patients receiving metformin, 396 were on metformin monotherapy and 358 were using metformin combination therapy.

FIGURE 1.

FIGURE 1

Open in a new tab

Flowchart of study participant inclusion. NACC, National Alzheimer's Coordinating Center; FDA, Food and Drug Administration.

The demographic and clinical characteristics of all participants at baseline are presented in Table 1. The mean age of all participants was 71.8 ± 8.0 years; 38.8% were male, 31.4% were Black, and 10.8% were Hispanic. One hundred four participants (7.5%) developed dementia over a median follow‐up of 4.0 years (interquartile range: 2.1–7.1 years). All covariates were well balanced after IPTW (all SMDs < 0.1 36 ; Table 1). Also, the distributions of the estimated propensity score for the metformin use and no‐use groups overlapped quite well (Figure S1 in supporting information). The MSE of the final model for the training set and testing set was 0.33 and 0.38, respectively.

TABLE 1.

Participant characteristics.

All, n (%) Metformin users, n (%) Non‐users, n (%) SMD *
N 1,393 754 639 Unweighted Weighted *
Demographic factors
Age ≥ 65 years 1150 (82.6) 611 (81.0) 539 (84.4) −0.088 −0.006
Sex, male 540 (38.8) 285 (37.8) 255 (39.9) −0.043 0.007
Education level, ≥ college school 708 (50.8) 395 (52.4) 313 (49.0) 0.068 0.005
Hispanic/Latino ethnicity 151 (10.8) 103 (13.7) 48 (7.5) 0.201 −0.016
Race, Black 438 (31.4) 197 (26.1) 241 (37.7) −0.250 0.004
Family history of dementia 666 (47.8) 385 (51.1) 281 (44.0) 0.142 −0.015
APOE ε4 346 (24.8) 179 (23.7) 167 (26.1) −0.055 0.018
Health behaviors
Ever smoker 657 (47.2) 350 (46.4) 307 (48.0) −0.033 0.009
Alcohol abuse 81 (5.8) 47 (6.2) 34 (5.3) 0.039 0.019
Drug abuse 29 (2.1) 19 (2.5) 10 (1.6) 0.068 0.016
Obesity 674 (48.4) 377 (50.0) 297 (46.5) 0.071 −0.003
Comorbidities
Cardiovascular disease 442 (31.7) 214 (28.4) 228 (35.7) −0.157 −0.028
Cerebrovascular disease 120 (8.6) 54 (7.2) 66 (10.3) −0.112 −0.002
Neurological diseases 143 (10.3) 72 (9.5) 71 (11.1) −0.051 0.026
Neuropsychiatric disorders 425 (30.5) 238 (31.6) 187 (29.3) 0.050 −0.003
Hypertension 1079 (77.5) 579 (76.8) 500 (78.2) −0.035 −0.023
Hypercholesterolemia 1070 (76.8) 590 (78.2) 480 (75.1) 0.074 −0.008
Vitamin B12 deficiency 91 (6.5) 44 (5.8) 47 (7.4) −0.061 −0.004
Medication use
Lipid‐lowering drugs 895 (64.2) 563 (74.7) 332 (52.0) 0.485 0.007
Anti‐hypertensive drugs 1062 (76.2) 629 (83.4) 433 (67.8) 0.371 0.020
Non‐steroidal anti‐inflammatory medication 623 (44.7) 385 (51.1) 238 (37.2) 0.281 0.013
Antidepressants 256 (18.4) 142 (18.8) 114 (17.8) 0.026 0.002
Antipsychotic drugs 8 (0.6) 7 (0.9) 1 (0.2) 0.105 0.042
Anti‐Parkinson's drugs 40 (2.9) 26 (3.4) 14 (2.2) 0.076 0.030
Anxiolytic, sedative, or hypnotic agents 130 (9.3) 62 (8.2) 68 (10.6) −0.083 −0.011
Other GLDs
SGLT2 inhibitors 13 (0.9) 9 (1.2) 4 (0.6) 0.060 0.020
DPP‐4 inhibitors 76 (5.5) 52 (6.9) 24 (3.8) 0.140 −0.027
GLP‐1RAs 27 (1.9) 22 (2.9) 5 (0.8) 0.159 −0.039
Alpha‐glucosidase inhibitors 7 (0.5) 2 (0.3) 5 (0.8) −0.072 −0.010
Sulfonylureas 371 (26.6) 215 (28.5) 156 (24.4) 0.093 0.038
Thiazolidinediones 136 (9.8) 80 (10.6) 56 (8.8) 0.062 −0.008
Insulin 200 (14.4) 86 (11.4) 114 (17.8) −0.183 0.042
Open in a new tab

Notes: Cardiovascular disease consisted of heart attack/cardiac arrest, angioplasty/endarterectomy/stent, cardiac bypass procedure, pacemaker and/or defibrillator, congestive heart failure, atrial fibrillation, angina, heart valve replacement or repair, and other cardiovascular diseases. Cerebrovascular disease included stroke and transient ischemic attack. Neurological disease involved PD, other PD disorders, traumatic brain injury, seizures, and other neurological conditions; neuropsychiatric disorders included post‐traumatic stress disorder, bipolar disorder, schizophrenia, depression, anxiety, obsessive‐compulsive disorder, developmental neuropsychiatric disorders, and other psychiatric disorders.

Abbreviations: APOE, apolipoprotein E; DPP‐4, dipeptidyl peptidase 4; GLD, glucose‐lowering drug; GLP‐1RAs, glucagon‐like peptide‐1 receptor agonist; PD, Parkinson's disease; SGLT2, sodium‐glucose transport protein 2; SMD, standardized mean difference.

*

After inverse probability of treatment weighting with an SMD ≤ 0.1 indicating a balance between the two groups.

In the overall cohort, metformin users were significantly associated with a lower risk of dementia than non‐users (RD, −3.2%; 95% CI, −6.2% to −0.2%; Figure 2). We estimated the HTEs of metformin on the risk of dementia. Deciles of predicted ITEs of metformin on incident dementia are presented in Figure 3. Among all participants included, 966 (69.4%) had a decreased risk of dementia. HTE analysis using a single decision tree model identified neuropsychiatric disorders, antidepressant use, and NSAIDs, which could be used to divide the sample into four subgroups (Figure 4). The characteristics of participants among the four HTE subgroups are shown in Table S1 in supporting information. Several covariates were unbalanced between the metformin use and no‐use groups in each of the HTE subpopulations, such as metformin users having a lower frequency of cardiovascular disease/cerebrovascular disease than non‐users. Metformin use was significantly associated with a decreased risk of dementia in the subgroup with no neuropsychiatric disorders and no NSAIDs (subgroup 1; RD, −8.7%; 95% CI, −13.4% to −4.1%); however, in participants with neuropsychiatric disorders and no antidepressant use, metformin was significantly associated with an increased risk of dementia (subgroup 3; RD, 8.6%; 95% CI, 1.8% to 15.5%). No significant decrease in the risk of dementia associated with metformin was observed in the subpopulation with no neuropsychiatric disorders and NSAIDs (subgroup 2; RD, −2.0%; 95% CI, −7.0% to 3.0%) or the subpopulation with neuropsychiatric disorders and antidepressant use (subgroup 4; RD, −4.6%; 95% CI, −12.3% to 3.1%).

FIGURE 2.

FIGURE 2

Open in a new tab

Absolute risk difference for dementia for metformin use versus no use in the overall population and subgroups identified using a single decision tree model. CI, confidence interval; NSAID, non‐steroidal anti‐inflammatory drug.

FIGURE 3.

FIGURE 3

Open in a new tab

Predicted individualized treatment effect (ITE) of metformin use versus no use on the risk of dementia onset. The predicted ITE is stratified into deciles (x axis). The predicted ITE is presented as a risk difference in the risk of dementia between metformin use versus no use (y axis). The predicted ITE was negative (benefit from metformin) among 966 (69.4%) participants.

FIGURE 4.

FIGURE 4

Open in a new tab

A single decision tree model was developed to identify the most important factors and estimate the absolute risk difference in risk of dementia between metformin use and no use among the subgroups. Negative values indicate reduced absolute risk of dementia (benefit from metformin), whereas positive values indicate increased absolute risk of dementia (harm from metformin). NSAIDs, non‐steroidal anti‐inflammatory drugs.

To determine whether there were any HTEs within the four subgroups, we examined the frequency of predictive ITE deciles (derived from Figure 3) and negative ITE across the four subgroups (derived from Figure 4; see Figure 5). The participants in subgroups 1, 2, and 4 were more likely to have a reduced risk of dementia associated with metformin, as demonstrated by their negative ITE: subgroup 1 (91.1%), subgroup 2 (66.6%), and subgroup 4 (78.8%). However, only 14.5% of participants in subgroup 3 had a negative ITE, indicating a lower probability of experiencing a decreased risk.

FIGURE 5.

FIGURE 5

Open in a new tab

The frequency of predicted individualized treatment effect deciles (derived from Figure 3) across four heterogeneous treatment effect subgroups (derived from Figure 4). ‐ITE, negative individualized treatment effect.

3.1. Additional analyses

We performed additional analyses to discern the impact of metformin monotherapy and metformin combination therapy on the risk of dementia (Table S2 in supporting information). Metformin monotherapy was significantly associated with a decrease in the risk of dementia compared to no use (RD, −3.9%; 95% CI, −7.3% to −0.7%), while metformin combination therapy was marginally associated with a decreased risk (RD, −3.1%; 95% CI, −6.5% to 0.4%). However, we observed no significant difference between metformin monotherapy and metformin combination therapy regarding the risk of dementia (RD, −0.1%; 95% CI, −1.9% to 1.6%), further reinforcing the evidence for metformin's potential in reducing the risk of dementia. In the sensitivity analyses, the estimated RDs in the overall cohort were consistent with the primary analysis, with an RD of −3.5% (95% CI, −6.6% to −0.3%) and −2.9% (95% CI, −5.8% to −0.03%) for those aged ≥ 60 years and using a valid doubly robust estimation method, respectively.

4. DISCUSSION

In this longitudinal observational study using the NACC database, we found that metformin users were significantly associated with a decreased risk of dementia in a population with T2D compared to non‐users. There was no significant difference between metformin monotherapy and metformin combination therapy regarding the risk of dementia. Additionally, we identified the three most influential factors (neuropsychiatric disorders, antidepressants, and NSAIDs), that contributed to the HTEs. Participants were classified into four HTE subgroups with the maximum differences in dementia risk attributable to metformin. The metformin‐using subgroup with no neuropsychiatric disorders and no NSAIDs (subgroup 1) had a lower risk of dementia, while those with neuropsychiatric disorders and no antidepressant use (subgroup 3) had a higher risk associated with metformin.

Our findings were in line with previous studies indicating that metformin conferred a protective effect against dementia in people with T2D. 13 , 14 , 16 , 17 , 39 For instance, one national cohort study involving 210,237 showed that new users of metformin had a reduced risk of dementia compared to non‐users, with an adjusted hazard ratio (HR) of 0.88 (95% CI, 0.84 −0.92). 16 Another retrospective cohort study involving 28,640 older veterans with T2D showed that metformin initiators had a lower risk of dementia than new users of sulfonylureas over an average follow‐up of 5 years (HR, 0.89; 95% CI, 0.79 −0.99). 14 While the exact mechanisms of action underlying the decreased risk of dementia associated with metformin are yet to be fully understood, several potential mechanisms have been proposed. First, metformin has been shown to inhibit advanced glycation end products, 40 which contribute to degenerative processes by accelerating the deposition of amyloid beta (Aβ) in various regions of the brain. 41 Second, metformin exerts inhibition of tau phosphorylation and Aβ by activating the adenosine monophosphate–activated protein kinase (AMPK) pathway. 42 , 43 Third, metformin may improve memory function by decreasing neuronal insulin resistance and neuroinflammation. 44

Although there was a lower risk of dementia associated with metformin, we identified HTEs of metformin on the risk of dementia. Neuropsychiatric disorders, antidepressant use, and NSAIDs were identified as top risk factors that could influence the association between metformin use and the risk of dementia. In our decision tree model, we observed a significantly decreased risk of dementia associated with metformin in the subgroup with no neuropsychiatric disorders and no use of NSAIDs (subgroup 1), which suggested that this subgroup could get cognitive benefit from metformin. A non‐significant decrease in risk of dementia was detected in the subgroup with no neuropsychiatric disorders and use of NSAIDs (subgroup 2). It should be noted that among the subgroup with neuropsychiatric disorders and no antidepressant use (subgroup 3), metformin was significantly associated with a higher risk of dementia than the no‐use group, indicating metformin would not be expected to aid this sub‐population of T2D with neuropsychiatric disorders. However, this association was not found in the treated subgroup with neuropsychiatric disorders and antidepressant use (subgroup 4). Thus, the increased risk of dementia among neuropsychiatric disorders might be offset by using antidepressants. The identification of HTE of metformin on dementia via machine learning approaches may provide information to guide further precision medicine approaches to evaluate drug repurposing of metformin for dementia. However, additional research is needed to assess the dementia risk associated with metformin in the above HTE subgroups using electronic health record data.

Our findings showed that neuropsychiatric disorders and antidepressant use were the important factors that would influence the association between metformin and the onset of dementia. Neuropsychiatric disorders, especially depression and anxiety, are common in the population with predementia and are risk factors for developing dementia. 45 Among patients newly diagnosed with T2D, depression was significantly associated with accelerated cognitive decline and an increased risk of AD. 46 , 47 One population‐based nested case‐control study showed that metformin was significantly associated with an increased risk of AD in diabetes patients with depression, 46 which was consistent with our finding that a higher risk of dementia associated with metformin was observed in the subpopulation with neuropsychiatric disorders and no antidepressant use (subgroup 3). Antidepressant use, another significant variable identified in this study, was used to relieve symptoms of depression and anxiety disorder. 48 Relief of symptoms of neuropsychiatric disorders through antidepressants may be associated with a decreased risk of dementia. However, evidence regarding the association between antidepressant treatment and the risk of dementia remains inconsistent. 49 , 50 An imbalance in the frequency of several covariates between metformin use and no‐use groups might be an issue. For example, the participants in the metformin group are more likely to have older age, be of Hispanic ethnicity, and have a family history of dementia than those in the no‐use group. In this study, we did not observe an increased risk of dementia associated with metformin in the subgroup with neuropsychiatric disorders and antidepressant use (subgroup 4). Currently, little is known about the interaction among metformin, neuropsychiatric disorders, and antidepressants on the development of dementia.

Previous studies have indicated an inverse association between NSAIDs and the risk of dementia; 51 , 52 , 53 however, this association differs depending on the specific type of NSAIDs. 54 In this study, we observed a lower risk of dementia associated with metformin among the subgroup with no neuropsychiatric disorders or no use of NSAIDs (subgroup 1). However, this risk reduction was not evident in the subgroup with no neuropsychiatric disorders and use of NSAIDs (subgroup 2). This study highlighted the complexity of the association between metformin use and NSAIDs in relation to the risk of dementia. However, no studies have explored this intricate association, suggesting the need for further investigations to better understand the impact of these medications on dementia risk.

This study has several strengths. First, the application of doubly robust learning offered several advantages over traditional methods like survival analysis. (1) This approach achieves an unbiased estimation of the average treatment effect by combining a propensity score model and an outcome regression model. (2) Using RD provides a more intuitive and clinically meaningful measure of the treatment effect compared to relative risk (i.e., HR). (3) Data‐driven multivariate subgroup analysis is a more sophisticated and statistically robust approach, enabling a deeper understanding of the complex relationships between variables and their impact on the outcome by overcoming the limitations of conventional “one‐variable‐at‐a‐time” analysis including the issues of spurious findings and multiple hypotheses testing. Second, this study benefited from the gold standard dementia diagnoses made by a team of experienced clinicians based on standardized diagnostic criteria. 22 We also acknowledge several limitations in our study. First, the NACC dataset was a clinic‐based sample that was subject to selection bias. The results were derived from the subjects who came from clinician referral, self‐referral or family member referral, active recruitment, or volunteering. Also, the sample size included in the study represents 3.4% of the full NACC sample, indicating a possible poor internal validity within the NACC sample. Thus, the findings of this study had limited generalizability to the diabetic population within the community; additional research is needed to assess the performance of the HTE subgroups in patients with diabetes in a general population. Second, metformin use and other variables were based on self‐reported medical history and treatment information; thus, the actual status and disease remained unknown, which may bias the classification of the exposure and covariates. Third, glycated hemoglobin (HbA1c) was an important confounder, impacting the selection of GLD and influencing the strength of the T2D–dementia association. However, HbA1c was not available for this study, preventing us from controlling for its effects on the association between metformin and the risk of dementia. 55 , 56 Interestingly, one recent population‐based study suggested that metformin reduced the risk of dementia independently of its effect on HbA1c. 16 Future studies are needed to quantify the association of metformin treatment with ADRD dementia risk after isolating the effect of HbA1c. Fourth, the HTE subgroups from the tree model were challenging to interpret in the context of physiology and clinical outcomes and it would be difficult to apply them directly to clinical practice with current knowledge. The imbalance in the distribution of covariates between metformin use and no‐use groups might have unintentionally introduced bias into the study results. Finally, we included the participants and defined their treatment status based on the baseline information; however, their status might change during the follow‐up period, leading to potential immortal time bias. For example, participants who took metformin at baseline might switch to other classes of GLDs during the study. We did not assess the cumulative duration of metformin use and its association with the risk of dementia in this study.

In conclusion, this study supports a significant association between metformin use and a reduced risk of dementia in individuals with T2D. Our findings suggest that important factors, such as neuropsychiatric disorders, antidepressant use, and NSAIDs can assist in making personalized decisions regarding the use of metformin in people with T2D, particularly for those at high risk of dementia. To further validate these findings, future research should investigate a more extensive and diverse population.

CONFLICT OF INTEREST STATEMENT

H. Tang was supported by a PhRMA Predoctoral Fellowship. J. Guo was supported by the National Institutes of Health (NIH)/National Institute of Diabetes and Digestive and Kidney Diseases (R01DK133465) and PhRMA Foundation; she also received a consulting fee from Pfizer which was outside of the current work. C. E. Shaaban was supported by NIA/National Institute on Aging (NIA; K01AG071849); she is also the Chair of the ISTAART Professional Interest Area to Elevate Early Career Researchers and Co‐Chair of the ISTAART Sex and Gender Interest Group, Diversity and Disparities Professional Interest Area. Y. Wu was supported by Patient‐Centered Outcomes Research Institute (PCORI; ME‐2018C3‐14754) and NIH/NIA (R56AG069880). S. T. DeKosky received royalties from UpToDate and a consulting fee from Brainstorm Cell Therapeutics; he also participated in Acumen Pharmaceuticals Medical Advisory Board, Biogen Data Safety Monitoring Board (DSMB), Cognition Therapeutics Medical Advisory Board, Prevail Pharmaceuticals DSMB, and Vaccinex Medical Advisory Board; all were outside of the current work. J. Bian was supported by NIH/NIA (1R01AG076234). All other authors declare no conflicts of interest. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All human subjects provided informed consent.

Supporting information

Supporting Information

ALZ-20-975-s002.docx (62KB, docx)

Supporting Information

ALZ-20-975-s001.pdf (2.3MB, pdf)

ACKNOWLEDGMENTS

The NACC database is funded by NIA/NIH grant U24 AG072122. NACC data are contributed by the NIA‐funded ADRCs: P30 AG062429 (PI James Brewer, MD, PhD), P30 AG066468 (PI Oscar Lopez, MD), P30 AG062421 (PI Bradley Hyman, MD, PhD), P30 AG066509 (PI Thomas Grabowski, MD), P30 AG066514 (PI Mary Sano, PhD), P30 AG066530 (PI Helena Chui, MD), P30 AG066507 (PI Marilyn Albert, PhD), P30 AG066444 (PI John Morris, MD), P30 AG066518 (PI Jeffrey Kaye, MD), P30 AG066512 (PI Thomas Wisniewski, MD), P30 AG066462 (PI Scott Small, MD), P30 AG072979 (PI David Wolk, MD), P30 AG072972 (PI Charles DeCarli, MD), P30 AG072976 (PI Andrew Saykin, PsyD), P30 AG072975 (PI David Bennett, MD), P30 AG072978 (PI Neil Kowall, MD), P30 AG072977 (PI Robert Vassar, PhD), P30 AG066519 (PI Frank LaFerla, PhD), P30 AG062677 (PI Ronald Petersen, MD, PhD), P30 AG079280 (PI Eric Reiman, MD), P30 AG062422 (PI Gil Rabinovici, MD), P30 AG066511 (PI Allan Levey, MD, PhD), P30 AG072946 (PI Linda Van Eldik, PhD), P30 AG062715 (PI Sanjay Asthana, MD, FRCP), P30 AG072973 (PI Russell Swerdlow, MD), P30 AG066506 (PI Todd Golde, MD, PhD), P30 AG066508 (PI Stephen Strittmatter, MD, PhD), P30 AG066515 (PI Victor Henderson, MD, MS), P30 AG072947 (PI Suzanne Craft, PhD), P30 AG072931 (PI Henry Paulson, MD, PhD), P30 AG066546 (PI Sudha Seshadri, MD), P20 AG068024 (PI Erik Roberson, MD, PhD), P20 AG068053 (PI Justin Miller, PhD), P20 AG068077 (PI Gary Rosenberg, MD), P20 AG068082 (PI Angela Jefferson, PhD), P30 AG072958 (PI Heather Whitson, MD), P30 AG072959 (PI James Leverenz, MD). This study is supported by grants from NIA (R01AG076234 and R56AG069880), NIDDK (R01DK133465), and PhRMA Predoctoral Fellowship.

Tang H, Guo J, Shaaban CE, et al. Heterogeneous treatment effects of metformin on risk of dementia in patients with type 2 diabetes: A longitudinal observational study. Alzheimer's Dement. 2024;20:975–985. 10.1002/alz.13480

REFERENCES

  • 1. Livingston G, Huntley J, Sommerlad A, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet. 2020;396:413‐446. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. World Health Organization . Dementia. Accessed May 9, 2022. Available at https://www.who.int/news‐room/fact‐sheets/detail/dementia
  • 3. 2022 Alzheimer's disease facts and figures. Alzheimers Dement. 2022;18:700‐789. [DOI] [PubMed] [Google Scholar]
  • 4. Centers for Disease Control and Prevention . Alzheimer's disease and healthy aging. 2019. Accessed May 13, 2022. Available at https://www.cdc.gov/aging/dementia/index.html
  • 5. Wong W. Economic burden of Alzheimer disease and managed care considerations. Am J Manag Care. 2020;26:S177‐S183. [DOI] [PubMed] [Google Scholar]
  • 6. Ghazarian AL, Haim T, Sauma S, Katiyar P. National institute on aging seed funding enables Alzheimer's disease startups to reach key value inflection points. Alzheimers Dement. 2022;18:348‐359. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Chatterjee S, Peters SA, Woodward M, et al. Type 2 diabetes as a risk factor for dementia in women compared with men: a pooled analysis of 2.3 million people comprising more than 100,000 cases of dementia. Diabetes Care. 2016;39:300‐307. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Xue M, Xu W, Ou YN, et al. Diabetes mellitus and risks of cognitive impairment and dementia: a systematic review and meta‐analysis of 144 prospective studies. Ageing Res Rev. 2019;55:100944. [DOI] [PubMed] [Google Scholar]
  • 9. Akter K, Lanza EA, Martin SA, Myronyuk N, Rua M, Raffa RB. Diabetes mellitus and Alzheimer's disease: shared pathology and treatment? Br J Clin Pharmacol. 2011;71:365‐376. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. 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. [DOI] [PubMed] [Google Scholar]
  • 11. Lee AK, Rawlings AM, Lee CJ, et al. Severe hypoglycaemia, mild cognitive impairment, dementia and brain volumes in older adults with type 2 diabetes: the Atherosclerosis Risk in Communities (ARIC) cohort study. Diabetologia. 2018;61:1956‐1965. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. American Diabetes Association Professional Practice C , Draznin B, Aroda VR, Bakris G, et al, American Diabetes Association Professional Practice C. 9 . Pharmacologic approaches to glycemic treatment: standards of medical care in diabetes‐2022. Diabetes Care. 2022;45:S125‐S143. [DOI] [PubMed] [Google Scholar]
  • 13. Samaras K, Makkar S, Crawford JD, et al. Metformin use is associated with slowed cognitive decline and reduced incident dementia in older adults with type 2 diabetes: the Sydney Memory and Ageing Study. Diabetes Care. 2020;43:2691‐2701. [DOI] [PubMed] [Google Scholar]
  • 14. Orkaby AR, Cho K, Cormack J, Gagnon DR, Driver JA. Metformin vs sulfonylurea use and risk of dementia in US veterans aged >/= 65 years with diabetes. Neurology. 2017;89:1877‐1885. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Imfeld P, Bodmer M, Jick SS, Meier CR. Metformin, other antidiabetic drugs, and risk of Alzheimer's disease: a population‐based case‐control study. J Am Geriatr Soc. 2012;60:916‐921. [DOI] [PubMed] [Google Scholar]
  • 16. Zheng B, Su B, Ahmadi‐Abhari S, et al. Dementia risk in patients with type 2 diabetes: comparing metformin with no pharmacological treatment. Alzheimers Dement. 2023. doi: 10.1002/alz.13349. Online ahead of print. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Ji S, Zhao X, Zhu R, Dong Y, Huang L, Zhang T. Metformin and the risk of dementia based on an analysis of 396,332 participants. Ther Adv Chronic Dis. 2022;13:20406223221109454. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Dahabreh IJ, Hayward R, Kent DM. Using group data to treat individuals: understanding heterogeneous treatment effects in the age of precision medicine and patient‐centred evidence. Int J Epidemiol. 2016;45:2184‐2193. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Funk MJ, Westreich D, Wiesen C, Sturmer T, Brookhart MA, Davidian M. Doubly robust estimation of causal effects. Am J Epidemiol. 2011;173:761‐767. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Robins JM, Rotnitzky A, Zhao LP. Estimation of regression coefficients when some regressors are not always observed. J Am Stat Assoc. 1994;89:846‐866. [Google Scholar]
  • 21. Microsoft Research . Doubly robust learning. Accessed May 23,2022. Available at https://econml.azurewebsites.net/spec/estimation/dr.html
  • 22. Gifford KA, Liu D, Lu Z, et al. The source of cognitive complaints predicts diagnostic conversion differentially among nondemented older adults. Alzheimers Dement. 2014;10:319‐327. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Durazzo TC, Mattsson N, Weiner MW. Alzheimer's disease neuroimaging I. Smoking and increased Alzheimer's disease risk: a review of potential mechanisms. Alzheimers Dement. 2014;10:S122‐S145. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Weintraub S, Salmon D, Mercaldo N, et al. The Alzheimer's disease centers' uniform data set (UDS): the neuropsychologic test battery. Alzheimer Dis Assoc Disord. 2009;23:91‐101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Weintraub S, Besser L, Dodge HH, et al. Version 3 of the Alzheimer disease centers' neuropsychological test battery in the uniform data set (UDS). Alzheimer Dis Assoc Disord. 2018;32:10‐17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer's disease: report of the NINCDS‐ADRDA work group under the auspices of department of health and human services task force on Alzheimer's disease. Neurology. 1984;34:939‐944. [DOI] [PubMed] [Google Scholar]
  • 27. McKeith IG, Galasko D, Kosaka K, et al. Consensus guidelines for the clinical and pathologic diagnosis of dementia with lewy bodies (DLB): report of the consortium on DLB international workshop. Neurology. 1996;47:1113‐1124. [DOI] [PubMed] [Google Scholar]
  • 28. Roman GC, Tatemichi TK, Erkinjuntti T, et al. Vascular dementia: diagnostic criteria for research studies. Report of the NINDS‐AIREN international workshop. Neurology. 1993;43:250‐260. [DOI] [PubMed] [Google Scholar]
  • 29. Neary D, Snowden JS, Gustafson L, et al. Frontotemporal lobar degeneration: a consensus on clinical diagnostic criteria. Neurology. 1998;51:1546‐1554. [DOI] [PubMed] [Google Scholar]
  • 30. Litvan I, Agid Y, Calne D, et al. Clinical research criteria for the diagnosis of progressive supranuclear palsy (Steele‐Richardson‐Olszewski syndrome): report of the NINDS‐SPSP international workshop. Neurology. 1996;47:1‐9. [DOI] [PubMed] [Google Scholar]
  • 31. Litvan I, Bhatia KP, Burn DJ, et al. Movement disorders society scientific issues committee report: SIC task force appraisal of clinical diagnostic criteria for Parkinsonian disorders. Mov Disord. 2003;18:467‐486. [DOI] [PubMed] [Google Scholar]
  • 32. Mesulam MM. Primary progressive aphasia. Ann Neurol. 2001;49:425‐432. [PubMed] [Google Scholar]
  • 33. Aschwanden D, Aichele S, Ghisletta P, et al. Predicting cognitive impairment and dementia: a machine learning approach. J Alzheimers Dis. 2020;75:717‐728. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Yang X, Li Q, Wu Y, et al. Early prediction of Alzheimer's disease and related dementias using electronic health records. medRxiv. 2020:2020.06.13.20130401. [DOI] [PMC free article] [PubMed]
  • 35. Safieh M, Korczyn AD, Michaelson DM. ApoE4: An emerging therapeutic target for Alzheimer's disease. BMC Med. 2019;17:64. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity‐score matched samples. Stat Med. 2009;28:3083‐3107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Li X, Shen C. Doubly robust estimation of causal effect: upping the odds of getting the right answers. Circ Cardiovasc Qual Outcomes. 2020;13:e006065. [DOI] [PubMed] [Google Scholar]
  • 38. Basu S, Raghavan S, Wexler DJ, Berkowitz SA. Characteristics associated with decreased or increased mortality risk from glycemic therapy among patients with type 2 diabetes and high cardiovascular risk: machine learning analysis of the ACCORD trial. Diabetes Care. 2018;41:604‐612. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Campbell JM, Stephenson MD, de Courten B, Chapman I, Bellman SM, Aromataris E. Metformin use associated with reduced risk of dementia in patients with diabetes: a systematic review and meta‐analysis. J Alzheimers Dis. 2018;65:1225‐1236. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Zhou Z, Tang Y, Jin X, et al. Metformin inhibits advanced glycation end products‐induced inflammatory response in murine macrophages partly through AMPK activation and RAGE/NFkappaB pathway suppression. J Diabetes Res. 2016;2016:4847812. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Vitek MP, Bhattacharya K, Glendening JM, et al. Advanced glycation end products contribute to amyloidosis in Alzheimer disease. Proc Natl Acad Sci U S A. 1994;91:4766‐4770. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. Chiang MC, Cheng YC, Chen SJ, Yen CH, Huang RN. Metformin activation of AMPK‐dependent pathways is neuroprotective in human neural stem cells against amyloid‐beta‐induced mitochondrial dysfunction. Exp Cell Res. 2016;347:322‐331. [DOI] [PubMed] [Google Scholar]
  • 43. Salminen A, Kaarniranta K, Haapasalo A, Soininen H, Hiltunen M. AMP‐activated protein kinase: a potential player in Alzheimer's disease. J Neurochem. 2011;118:460‐474. [DOI] [PubMed] [Google Scholar]
  • 44. Madhu LN, Kodali M, Shetty AK. Promise of metformin for preventing age‐related cognitive dysfunction. Neural Regen Res. 2022;17:503‐507. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45. Acosta I, Borges G, Aguirre‐Hernandez R, Sosa AL, Prince M, Dementia Research Group . Neuropsychiatric symptoms as risk factors of dementia in a Mexican population: a 10/66 dementia research group study. Alzheimers Dement. 2018;14:271‐279. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46. Ha J, Choi DW, Kim KJ, et al. Association of metformin use with Alzheimer's disease in patients with newly diagnosed type 2 diabetes: a population‐based nested case‐control study. Sci Rep. 2021;11:24069. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47. Sullivan MD, Katon WJ, Lovato LC, et al. Association of depression with accelerated cognitive decline among patients with type 2 diabetes in the ACCORD‐MIND trial. JAMA Psychiatry. 2013;70:1041‐1047. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48. Kendrick T, Pilling S, Mavranezouli I, et al. Management of depression in adults: summary of updated NICE guidance. BMJ. 2022;378:o1557. [DOI] [PubMed] [Google Scholar]
  • 49. Kessing LV, Sondergard L, Forman JL, Andersen PK. Antidepressants and dementia. J Affect Disord. 2009;117:24‐29. [DOI] [PubMed] [Google Scholar]
  • 50. Lee CW, Lin CL, Sung FC, Liang JA, Kao CH. Antidepressant treatment and risk of dementia: a population‐based, retrospective case‐control study. J Clin Psychiatry. 2016;77:117‐122. quiz 22. [DOI] [PubMed] [Google Scholar]
  • 51. Broe GA, Grayson DA, Creasey HM, et al. Anti‐inflammatory drugs protect against Alzheimer disease at low doses. Arch Neurol. 2000;57:1586‐1591. [DOI] [PubMed] [Google Scholar]
  • 52. Szekely CA, Breitner JC, Fitzpatrick AL, et al. NSAID use and dementia risk in the cardiovascular health study: role of APOE and NSAID type. Neurology. 2008;70:17‐24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53. in t' Veld BA, Ruitenberg A, Hofman A, et al. Nonsteroidal antiinflammatory drugs and the risk of Alzheimer's disease. N Engl J Med. 2001;345:1515‐1521. [DOI] [PubMed] [Google Scholar]
  • 54. Zhang C, Wang Y, Wang D, Zhang J, Zhang F. NSAID exposure and risk of Alzheimer's disease: an updated meta‐analysis from cohort studies. Front Aging Neurosci. 2018;10:83. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55. Rawlings AM, Sharrett AR, Albert MS, et al. The association of late‐life diabetes status and hyperglycemia with incident mild cognitive impairment and dementia: the ARIC study. Diabetes Care. 2019;42:1248‐1254. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56. Dove A, Shang Y, Xu W, et al. The impact of diabetes on cognitive impairment and its progression to dementia. Alzheimers Dement. 2021;17:1769‐1778. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting Information

ALZ-20-975-s002.docx (62KB, docx)

Supporting Information

ALZ-20-975-s001.pdf (2.3MB, pdf)

Articles from Alzheimer's & Dementia are provided here courtesy of Wiley

RESOURCES