Most AD risk derives from non-modifiable risk factors. Aging is a major risk factor for sporadic AD, the focus of this review, and usually occurs over the age of 65 years. AD is also highly heritable, and genetic factors account for approximately 60 to 80% of risk based on twin studies [3, 17]. The common apolipoprotein E4 (APOE4) allele explains a substantial portion of AD heritability (about 9%) [18], increasing risk an estimated 3- to 4-fold. On the other hand, besides the highly penetrant APP (amyloid precursor protein), PSEN1 (presenilin-1), and PSEN2 (presenilin-2), most of the other over 80 identified risk loci are associated with odds ratios of only between 1.05 and 1.2 [19]. However, risk from these more minor contributing alleles is additive; a polygenic risk score can estimate an individual’s AD risk through a weighted sum of all risk-associated variants they harbor based on their effect sizes [20]. Dementias more broadly exhibit a complex genetic architecture (Table 1) with polygenic risk [21].

The balance in variation after accounting for 60 to 80% of AD heritability, i.e., the remaining 20 to 40%, is explained by individual environmental and lifestyle factors that differ between people (e.g., lifestyle, diet, etc.) [17], formalized by the concept of the exposome [22]. The AD exposome is the cumulative of environmental exposures, spanning toxicant (chemical, biological), lifestyle, social, and ecosystem factors [22], that a person experiences over their lifetime, which influences disease risk and progression. Components of the exposome can be either harmful or protective. The AD exposome is wide-ranging, but poor diet and impaired metabolic health, such as T2D and (central) obesity, are key detrimental components [23]. The Lancet Commission on Dementia prevention, intervention, and care was launched to generate authoritative, evidence-based guidance for reducing the global burden of dementia by identifying potentially modifiable risk factors [6]. In their latest 2024 report, they listed diabetes along with various additional cardiometabolic variables, as dementia risk factors.

Besides environmental and lifestyle risk factors, gene-environment interactions may influence a person’s individual AD risk [24]. For AD risk genes that are not highly penetrant, adverse and protective environmental and lifestyle factors may interact with genetic risk to raise and lower the odds of disease, respectively. Besides APOE4, gene-environment interactions are relatively understudied in the context of diabetes and dementia and AD. Nevertheless, individuals that harbor genetic risk from non-penetrant mutations can reduce their personal chances of developing dementia through appropriate protective measures.

The China Health and Retirement Longitudinal Study observed diabetes significantly correlated to cognitive decline after adjusting for numerous covariates [31]; stratified by sex, diabetes was linked to cognitive decline in females but not males. These sex dimorphic results are echoed by other studies. Sex-stratified analysis of UK Biobank identified excess AD, but not all-cause or VaD, risk in females versus males [32], while a large meta-analysis of 14 studies found 19% excess VaD risk in T2D females versus T2D males [33]. Dementia prevalence is approximately 1.3 times higher in females versus males [34], attributed in large part to longer female life expectancy [2]; however, biological factors, such as menopause and hormones (e.g., estrogen, which exerts neuroprotective properties [35]) or social factors (e.g., education level) may contribute to real sex differences in dementia risk in T2D and remains an active research area. Besides sex, one meta-analysis examined dementia risk from diabetes by race but concluded cautious interpretation due to sample size [36].

Finally, and worryingly, T2D in youths may also predispose patients to poorer cognition and brain structural changes; although studies are presently sparse, they indicate an area of potentially grave concern [37]. A large nationwide, population-based study (n = 971,677 of which 3,570 developed T2D) found that adolescents in the lowest versus highest quintile of global cognitive Z-scores were at highest risk for T2D, with hazard ratios of 2.46 in males and 2.33 in females, adjusted for age, various metabolic factors, and socioeconomic status [38]. The relationship also held for cognitive subdomain scores. This finding is recapitulated by relatively small studies that show, versus obese youths without T2D [39] or lean controls [40] matched for various metrics, youths with T2D and obesity perform poorer in several cognitive domains, estimated intellectual functioning [39], working memory [40], verbal memory [39, 40], psychomotor efficiency [39], and processing speed [40]. Brain structural findings are also concerning, finding variably across diverse studies, lower global and regional grey and white matter volumes and microstructural integrity, with altered functional connectivity [37, 41]. Further study is needed in this critical area with emphasis on prevention of T2D and related cognitive impairment (CI).

As highlighted above, mounting evidence supports a link between diabetes and dementia and AD risk (Table S1). However, one prominent limitation has been lack of stratification by dementia subtype, restricting mechanistic interpretation of pathophysiology [25]. For instance, Zhou et al. [32], from examining UK Biobank, found that T2D was significantly linked to AD with a hazard ratio of 2.38 but to VaD with hazard ratio of 3.93, although they did not compute relative ratios. Thus, granular analysis is essential. Additionally, many observational studies relied on electronic health records and International Classification of Diseases codes rather than on clinical examination and disease biomarkers, e.g., amyloid-beta (Aβ), which may misclassify cases and dementia subtypes. Importantly, although several analyses adjusted for APOE4 status as a covariate [26–28, 32], many lacked the information. However, one investigation found that accounting for APOE4 status did not change their main findings on the T2D-dementia relationship [30].

Furthermore, several studies acknowledged that they lacked complete information on all potential covariates that could influence the T2D-dementia relationship, such as metabolic comorbidities, medication use and diabetes control, neuropsychological conditions (e.g., depression, sleep disorders), lifestyle habits, and socioeconomic factors. It is crucial to adjust for all covariates to derive meaningful insight. It is especially important to consider T2D is highly comorbid with many other cardiometabolic risk factors (e.g., obesity, dyslipidemia, hypertension) [15], which the Lancet Commission lists as dementia risk factors [6]. Thus, although the weighted population attributable fraction of diabetes to dementia is only 2.3, accounting for comorbid cardiometabolic risk factors increases the total weighted population attributable fraction, which is consequently more impactful on dementia and AD burden.

Insulin is a peptide hormone released by pancreatic β-cells in response to the post-prandial rise in blood glucose levels. Insulin acts on tissues throughout the body to promote cellular glucose uptake as well as fatty acid and amino acid uptake and energy storage. At the cellular level, insulin binding to insulin receptors activates several downstream signaling cascades; the insulin-IRS-AKT axis promotes translocation of glucose transporters (GLUTs) to the membrane for glucose uptake and regulates phosphorylation of various kinases, including GSK3, overall activating metabolic and anabolic processes, including DNA, protein, and lipid synthesis [42]. Insulin binding also engages the insulin-MAPK axis, which regulates various transcription factors and elements controlling transcription, translation, and post-translational modifications.

Although neuronal glucose uptake in the brain occurs primarily by insulin-independent GLUT3, the brain is nonetheless an insulin sensitive organ [42]. Insulin, entering via selective transport across blood-brain barrier (BBB) endothelial cells, binds to insulin receptors widely expressed across the brain, especially at pre- and postsynaptic compartments. Binding activates neurite outgrowth, modulates neurotransmitter release and uptake, and regulates expression, trafficking, and localization of excitatory and inhibitory neurotransmitter receptors. Insulin also mediates glucose uptake via GLUT4, also expressed on neurons, during periods of intense activity, acting in concert and supplementing uptake via insulin-independent GLUT3. Under homeostatic conditions, glia aid neurons in their activities; astrocytes recycle neurotransmitters, transfer energy substrates to neurons, and maintain BBB integrity [43]. Microglia, the resident macrophages of the brain, surveil their surroundings and conduct housekeeping activities, such as synaptic remodeling and clearing debris [44]. Astrocytes and microglia express insulin receptors and are insulin responsive, modulating their actions in supporting neuronal activity [42].

Insulin resistance in peripheral tissues is a defining and early characteristic during T2D progression, especially in muscle and adipose tissues, with failure of β-cells to compensate. A similar concept occurs centrally, defined as the development of brain insulin resistance in T2D and lack of a response of brain cells to insulin [42] (Figure 1). Consequently, neurons and their supportive glia fail to perform their usual activities prompting cognitive dysfunction in T2D. Drawing parallels to the periphery, it is tempting to speculate that brain insulin resistance is the initiating or an early process in the brain leading to CI. Indeed, subtle CI occurs in individuals with T2D early in the disease course [39, 45], suggesting this state of brain insulin resistance may underlie these cognitive deficits. Furthermore, brain-specific knockout of insulin signaling in animal models promotes CI along with disrupted hippocampal synaptic plasticity, mitochondrial dysfunction, reduced energy production, and oxidative stress [46, 47] (see sections below), but confirmatory investigation is needed. Additionally, further comorbid risk factors, e.g., central obesity, and pathophysiological processes also contribute to CI progression, including vascular pathology. The human T2D brain is associated with poor glucose metabolism [48, 49], also an early occurrence likely to coincide with the development of brain insulin resistance.

In parallel, insulin resistance [42, 50] is also a facet of the human AD brain, independent of diabetes, showing altered levels and phosphorylation of insulin signaling pathway proteins, potentially early in the disease course [51, 52]. Glucose hypometabolism is also a key feature in AD and other dementias thought to arise downstream of neurodegeneration and synaptic failure, and tracking with disease progression [53, 54].

In T2D, insulin resistance and consequent hyperglycemia trigger a series of inflammatory responses and mitochondrial dysfunction (Figure 1) [43]. In vitro models suggest insulin may modulate inflammatory cytokine secretion from astrocytes and microglia [55]; in vivo models of T2D [56] and high-fat diet (HFD)-induced prediabetes and obesity [57, 58] exhibit astrogliosis and microgliosis underscoring a neuroinflammatory state. Molecularly, hyperglycemia promotes the formation of advanced glycation end products (AGEs), which activate receptor for AGE (RAGE), promoting inflammation through NF-κB in diabetes [59]. Microglia express RAGE [60] and may potentially constitute inflammatory pathways in diabetes, although this remains understudied specifically in the context of T2D-related CI. T2D also compromises the BBB, prompting trafficking of immune cells and inflammatory factors from chronic peripheral inflammation [8] into the CNS [61].

Hyperglycemia raises mitochondrial fission [62] and depolarization [63] and lowers levels of mitochondrial complex components [64] with consequent reactive oxygen species (ROS) and aggravated insulin resistance [64] in vitro in hippocampal neurons [62–64] and in vivo [64]. Furthermore, hyperglycemia reduces mitochondrial motility in hippocampal neurons [65], which, in the brain, would impair their ability to direct mitochondria to areas of high energy demand to support cognitive function. Dyslipidemia, a frequent comorbidity in T2D, modeled by treating neurons with palmitate in vitro, also drives mitochondrial fission [66] and depolarization [66, 67], alters mitochondrial respiration [68], diminishes cellular ATP levels [69], and raises ROS generation [66, 67, 69].

In AD, neuroinflammation is a disease hallmark with both microglial and astrocytic dimensions (Figure 1) [44, 70]. Initially, during aging, acute microglial inflammatory responses via Toll-like receptor binding promote phagocytosis and clearance of amyloid aggregates, maintaining CNS homeostasis [71]; however, triggers, such as genetic predisposition, peripheral inflammation, or other risk factors, promote sustained microglial activation, which eventually fails to resolve and segues into chronic inflammation with ensuing synapse loss and neurodegeneration [44]. RAGE signaling via microglia [72, 73] may also contribute to neuroinflammation, Aβ aggregation, and CI in AD [74, 75]. Astrogliosis also occurs in AD, even prior to amyloid deposition; with progression, in tandem with activated microglia, reactive astrocytes also surround amyloid aggregates [76]. However, astrocytes more distant from amyloid may instead exhibit atrophy [77], with resulting deficits in neurotransmitter recycling and impaired neurotransmission [78].

Like in the T2D brain, mitochondrial dysfunction is also a potentially central AD characteristic [79, 80]. Mitochondria from AD patients show altered structure and function versus those in healthy neurons. AD mitochondria are smaller in size with swollen cristae, found associated with oxidatively damaged neurons [81]. Mitochondrial dynamics are impaired early in the AD disease course, with defects in biogenesis, fission:fusion balance, and mitophagy [82]. Functionally, based on human brain autopsy tissues, AD mitochondria are deficient in mitochondrial enzyme activities, such as pyruvate dehydrogenase complex [83], 2-ketoglutarate dehydrogenase [84], glutamate dehydrogenase, and ATP synthase [85]. Mitochondrial ATP production has been reported both as inhibited [86] and unaffected [82], requiring clarification. Nevertheless, these mitochondrial abnormalities may either contribute to or drive AD pathologies, which have birthed alternate theories on their involvement in disease development. From one predominant perspective, Aβ promotes mitochondrial dysfunction in AD, which is viewed as a neuropathology byproduct, a view aligned with the amyloid cascade hypothesis. Alternatively, a substantial body of evidence suggests mitochondrial dysfunction can occur independent of Aβ, potentially upstream, formulated as the “mitochondrial cascade hypothesis” [79], which has garnered interest [87, 88] but has not become mainstream; thus, questions persist on the precise contribution of mitochondrial dysfunction in AD. Oxidative stress is also present in AD, evidenced by elevated brain levels of oxidatively damaged proteins, lipids, DNA, and RNA [75].

Diabetes is long-known to cause vascular damage, giving rise to its diabetic complications, either as microvascular (small vessels) or macrovascular (large vessels) complications. Microvascular small vessel damage to peripheral nerves gives rise to diabetic peripheral neuropathy [15]. In the brain, diabetes injures the small vessels, promoting the development of white matter hyperintensities, lacunae, and microbleeds, and also damages the large vessels [25, 61]. Mechanisms underlying vascular dysfunction in diabetes are multifactorial; systemic hyperglycemia promotes AGE production [89], atherosclerosis development, buildup of proteinaceous deposits, and ROS formation along with oxidized lipotoxic species [90]. Brain endothelial-specific knockout of genes involved in removing proteinaceous deposits and inflammatory modulation compromises cognition and increases string vessel pathology, BBB disruption, and neuroinflammation, underscoring the contribution of vascular dysfunction to neurodegeneration [91, 92].

Vascular pathology in T2D manifests with slowed blood flow [93, 94], stiffening of the large vasculature that increases pressure and damage to the microvasculature, reduced vasoreactivity, basement membrane thickening, and increased capillary and BBB permeability [61]. Decreased perfusion leads to chronic ischemia and loss of neurovascular coupling, with consequent neuronal injury, CI, and eventual neurodegeneration. Indeed, the T2D brain exhibits both grey [95] and white matter atrophy in various regions, including hippocampus linked to memory and learning [96–98], although studies are not universally concordant [99]. T2D also increases vulnerability to stroke and post-stroke dementia [100, 101].

Likewise, the AD brain also exhibits vascular dysfunction [102, 103] and reduced cerebral blood flow [104], built from a body of epidemiological, clinical, and preclinical evidence. Aβ has vasoactive and vasculotoxic properties; Aβ interacts with blood vessel endothelial cells [105], BBB RAGE [106], and various other receptors [107], generating ROS and inflammatory responses that impair endothelial vasoactivity and cerebral blood flow. Additionally, cerebral amyloid angiopathy, the cerebrovascular deposition of Aβ, is a common age-related small vessel pathology associated with intracerebral hemorrhage and CI [108], that frequently occurs as an AD co-pathology [109]. Brain ischemia may also be a contributing factor to AD [110, 111].

Overall, studies demonstrate that similar multifactorial pathological processes occur in the T2D and AD brains, spanning insulin resistance, impaired glucose metabolism, mitochondrial dysfunction, inflammation, oxidative stress, and vascular dysfunction, although the interrelationship of these pathways, if any, in T2D to AD remains unclear. Moreover, the precise sequence of events and whether specific pathways are causal or associated with CI in T2D remain uncertain, and more mechanistic studies are needed. A deeper understanding of CI pathogenesis in T2D may yield therapeutic avenues warranting further investigation. The next sections discuss some points regarding the neuropathological hallmarks of AD, amyloid-β plaques and neurofibrillary tangles of tau protein, and their relationship to T2D.

Although T2D patients are at elevated risk of developing AD (see “Diabetes and comorbid risk factors raise dementia and AD risk” section), studies do not show a correlation between T2D measures, such as HbA1c [112], and amyloid burden, or a higher amyloid or neurofibrillary tangle burden in T2D patients [113, 114]. Rather, dementia in T2D patients is heterogeneous, encompassing vascular, AD, inflammatory, and metabolic processes with substantial inter-individual variability, which arises from a confluence of superimposing non-modifiable and modifiable risk factors spanning aging, genetics, extent of diabetes control, and lifestyle factors and other environmental exposures (Figure 2; see Biessels and Despa [25] for a review). Thus, T2D patients are at increased AD risk but not of AD neuropathology [25, 42]. In contrast (see section below), T2D in animal models and mechanistic studies does show a link to AD neuropathology. Herein, we discuss possible concepts to reconcile these differences.

As outlined in the “Diabetes and comorbid risk factors raise dementia and AD risk” section, individuals with earlier T2D diagnosis are at greater risk of brain atrophy [28] and dementia [28, 30]. Analysis of UK Biobank at 12-year follow-up found excess all-cause, AD, and VaD risk in T2D overall [28]; however, in age stratified analysis, excess VaD risk was fully attributable to T2D onset prior to 50 years of age, while all-cause dementia and AD excess risk were most evident in T2D onset < 50 years of age although other onset age groups were also at risk. In analysis of Whitehall II at 32-year follow-up, compared to individuals without T2D aged 70 years, every 5-year increment in earlier T2D onset increased dementia risk [30]. Therefore, preventing or delaying T2D onset is imperative for better long-term brain health outcomes; T2D prevention and/or delaying approaches span lifestyle interventions through diet and exercise as well as pharmacotherapy [8]. Wearable technologies may track cardiometabolic metrics, e.g., weight, BMI, and lifestyle, e.g., dietary habits, exercise duration, helping individuals gauge their health status [8]. Technology through online coaching, social media, and gamification may also help. Methods that identify individuals most at risk of T2D can direct more intensive efforts or allocation of resources and public health policy can implement national and community programs [8].

Although the optimal approach would be prevention, some individuals nevertheless develop T2D, even despite efforts. This could be attributable to genetic predisposition to T2D [148], inability to adhere to recommended guidelines [8], or life factors [149] that thwart implementing ideal prevention approaches, e.g., socioeconomic status, lack of access to health resources, lack of access to green spaces or exercise. Nevertheless, some research shows that, even in individuals with T2D, adopting a healthy lifestyle can curb the risk of dementia development. Exercise, both resistance and endurance training, may exert salutary effects on cognition and brain structure and function in individuals with T2D [150]; however, further investigation, especially of exercise intensity and duration, is required to identify the optimal regimen or personalized approaches, and determine whether it can prevent dementia onset and development. The situation is similar regarding dietary habits; although there is evidence that diverse diets, both in composition [151] and time-restricted eating [152], may be beneficial to brain health, the mechanisms are incompletely understood, and the ideal approach or personalized strategies for preventing dementia and/or AD onset remain uncertain [153]. Nonetheless, T2D patients should be highly encouraged to adopt a healthy lifestyle; working from UK Biobank data, Zhou et al. [154] found that persons with diabetes who managed to adhere to more favorable healthy lifestyle habits were only at twice the risk while those not adhering were at three times the risk of all-cause dementia versus diabetes-free participants adhering to a healthy lifestyle. Adopted healthy lifestyles dose-dependently lowered VaD risk in diabetes cases but did not impact AD risk. Although the results are encouraging, reduced dementia risk from healthy lifestyle adoption was most pronounced in patients with shorter diabetes duration (< 10 years), suggesting the need to early implementation, or those without insulin use [154].

Just as a higher level of disadvantage can impact T2D [149] and dementia [155, 156] onset, similarly, these circumstances can also potentially raise dementia risk in individuals with T2D, although this is a less studied aspect. Social determinants of health (SDoH) are modifiable risk factors that constitute an avenue for mitigating future burden of dementia. Importantly, most individuals living with T2D [7] and/or dementia [2] reside in lower and lower-middle incomes countries, making SDoH a topic of global concern.

Adverse SDoH lower individuals’ access to high-quality care, which delays diabetes diagnosis, leads to suboptimal medication use and poor disease control, and lack of comorbidity management, e.g., obesity, hypertension [157]. Chronic stress from socioeconomic pressures increases allostatic load, promoting insulin resistance, elevated cortisol, and inflammation [158]. Socially disadvantaged people have limited access to healthy foods and green spaces for exercise and fewer opportunities for higher education and more cognitively stimulating jobs. All these conditions, as dementia risk factors, can predispose socioeconomically disadvantaged T2D patients to future AD and other dementia development.

Zhong et al. [159] analyzed T2D cases from UK Biobank (n = 17,321), extracting data on 17 SDoH to generate a composite [159]. Versus T2D participants in the highest SDoH tertile, those in the middle and lowest tertile had hazard ratios of 1.35 and 1.76, respectively, for incident dementia, adjusted for age, sex, diabetes duration, Hb1Ac, BMI, systolic blood pressure, comorbidities, and various lifestyle factors. This study represents emerging evidence that adverse SDoH may negatively impact brain health in T2D diabetes patients.

Antidiabetic medications and interventions can mitigate dementia and AD risk by controlling systemic cardiometabolic risk factors, for instance via weight loss and improved glucose handling. Moreover, antidiabetic medications, if BBB-permeable, and interventions can potentially exert neuroprotective mechanisms in the brain microenvironment, ameliorating cognition. Several investigations have examined whether antidiabetic medications and interventions exert salutary benefits to cognition and risk of dementia, including AD (Table 3).