Apathy Mediates the Association Between Diabetes Risk and Alzheimer’s Disease and Related Dementias in Older Adults: The Health and Retirement Study

1. Introduction

Apathy is an increasingly recognized neuropsychiatric syndrome characterized by diminished motivation, goal-directed behavior, and emotional responsiveness.[1,2] Apathy is a potentially treatable condition and is common among older adults with an estimated prevalence of 20–30% that increases with advancing age.[3,4]

The clinical significance of apathy lies in its association with multiple adverse health outcomes among older adults, including falls, frailty, hospitalizations and an increased risk of cognitive decline and Alzheimer’s disease and related dementias (ADRD). [4,5,6,7] Prior studies show that individuals with apathy experience more rapid cognitive decline and have an approximately two-fold increased risk of developing dementia. [8,9,10,11] This elevated cognitive risk may reflect both direct mechanisms, such as underlying neural pathology, as well as indirect mechanisms, such as reduced social engagement and physical activity due to apathy. Importantly, emerging evidence indicates that apathy is associated with cognitive decline, independent of depression. While apathy and depression are have some overlapping symptoms and can co-occur, apathy has emerged as a stronger predictor of cognitive decline than depression, among individuals with mild cognitive impairment.[12,13,14,15,16]

Type 2 diabetes mellitus (hereafter referred to as diabetes) is another well-established risk factor for cognitive impairment[17,18,19,20,21] and is associated with an approximately 60% increased risk of ADRD.[21] Existing research report a higher prevalence of apathy among older adults with diabetes compared to healthy controls,[22,23] however, longitudinal evidence for relationships between diabetes and apathy remains limited, particularly on the directionality of these conditions in relation to cognitive decline and ADRD.

The aim of this study was to examine whether apathy mediates the association between diabetes risk (measured by glycosylated hemoglobin) and incident self-reported ADRD and cognitive impairment— using data from the Health and Retirement Study (HRS). Given that in prior studies apathy has been associated with vascular disease in older adults,[24] we hypothesized that hyperglycemia accelerates microvascular damage, leading to apathy which in turns mediates the association between diabetes and incident ADRD and cognitive impairment. As both diabetes and apathy are potentially modifiable through medication and lifestyle interventions, examining these relationships has important implications for delaying the onset of cognitive impairment and subsequently, ADRD.

2. Methods

2.1. Study Participants

We used data from 4,571 adults aged 60 and older enrolled in the U.S. HRS, a nationally representative longitudinal study of adults aged 45 and older. Participants included in the present analysis had no diagnosis of memory-related problems at baseline and completed the 2006 HRS Enhanced Face-to-Face (EFTF) interview which collected data on physical performance, anthropometric measurements and blood and saliva samples that were used to determine glycosylated hemoglobin levels (see below). Prior to enrollment in HRS, informed consent was obtained from all participants. Data were accessed through the Gateway to Global Aging Data (https://g2aging.org/) and the Health and Retirement Study public data repository (https://hrsdata.isr.umich.edu). Ethical approval for the HRS was granted by local institutional review boards, and all participants provided informed consent prior to enrollment.

2.2. Diabetes Risk

Blood-based glycosylated hemoglobin (HbA1c), collected during 2006 HRS EFTF interview, was used as the primary measure of diabetes risk. Based on this variable, participants without diabetes, prediabetes, and diabetes were all included in the analyses and the variable was treated as continuous. HbA1c reflects average blood glucose levels over approximately the prior 120 days, corresponding to the average lifespan of red blood cells. It represents the proportion of hemoglobin molecules that have undergone non-enzymatic glycation due to circulating glucose concentrations and provides a stable indicator of long-term glycemic exposure that is less sensitive to short-term fluctuations in blood glucose[25].

2.3. Self-Reported Diagnosis of ADRD

A self-reported question on whether participants had ever been diagnosed with a memory problem (2006 and 2008), Alzheimer’s disease or dementia (2010 onwards). The date on which an individual first reported a diagnosis during the follow-up was defined as the date of incident ADRD.

2.4. Incident Cognitive Impairment

Cognitive impairment was defined as scoring >1.5 SD below the age-adjusted cut-scores on at least one of the cognitive tests (total recall, orientation, serial 7s and counting backwards)[26], combined with any impairment in Instrumental Activities of Daily Living (IADL). IADL impairment included self-reported difficulty in making phone calls, managing money and medication, reading a map, shopping for groceries, and preparing a hot meal. The date on which an individual first met criteria for cognitive impairment during the follow-up was defined as the date of incident cognitive impairment.

2.5. Apathy

Apathy was derived as an average of four items of the 8-item Center for Epidemiologic Studies Depression Scale (CES-D), that assesses affective and motivational symptoms experienced during the past week. The items included “felt happy”, “enjoyed life”, “felt that it took a lot to go on” and “everything was an effort”. Positively worded items were reverse-coded such that higher scores reflected greater apathy. These items were chosen to capture core features of apathy, including diminished positive affect and reduced motivation. We derived apathy measures from the CES-D, in the HRS core sample rather than using the Neuro Psychiatric Inventory available in Aging, Demographics, and Memory Study (ADAMS) in order to maximize sample size and ensure broader population representation. This approach was supported by that in the absence of a gold-standard instrument for apathy, prior large-scale cohort studies such as Baltimore Epidemiologic Catchment Area Study and the Prevention of Dementia by Intensive Vascular Care trial have similarly operationalized apathy using depression measures [2,3,27,28] .

2.6. Covariates

Age, sex, race/ethnicity and level of education were used as covariates. History of physician diagnosed medical conditions (i.e., cardiovascular risk factors and events), depression, obesity (calculated using body mass index) were used to characterize the study sample.

2.7. Statistical Analysis

2.7.1. Survival Analyses

All analyses were conducted in STATA version 19 (StataCorp LLC, College Station, TX, United States). The association between diabetes and incident self-reported ADRD and incident cognitive impairment was assessed using the Accelerated Time to Failure (ATF), assuming a Weibull distribution. ATF models estimate the extent to which an exposure accelerates (or decelerates) the time-to-event and are well suited to use in mediation analysis with survival outcomes compared to Cox Proportional Hazard models. Results are presented as time ratios (model coefficients) with corresponding 95% confidence intervals (CIs). All models were adjusted for baseline age, sex and education. Post-stratification sample weights provided by the HRS were applied to all models to account for the survey design. Person-time was calculated in days from the date of first assessment to the date of first reporting an ADRD diagnosis or meeting criteria for incident cognitive impairment, or to the date of last assessment for those who did not develop cognitive impairment or ADRD.

2.7.2. Mediation

To determine whether apathy mediated the association between diabetes and incident ADRD and cognitive impairment, we used a product-of-coefficients approach that appropriately accounted for the survey design. First, the association between diabetes (exposure) and apathy (mediator) were examined in a survey-weighted linear regression adjusted for covariates. Next, apathy was included in separate ATF Weibull models examining the association between diabetes and time to incident ADRD and time to cognitive impairment. The indirect (mediated) effect was calculated as the product of the coefficient for the exposure–mediator association (a) and the coefficient for the mediator–outcome association (b). The variance of the indirect effect was estimated using Sobel’s delta method, and statistical significance was assessed using a normal approximation.

3. Results

Table 1 summarizes participant characteristics at baseline, reported as raw sample sizes and weighted percentages. The mean age was 71.0 (SD 8.1 years) and 56% of participants were female.

Table 1. Characteristics of participants at baseline (n=4,571).

	Means or n	SD or percentages
Demographic factors
Age (years)	71.0	8.1
Female	2671	56
Higher levels of education (years)	12.5	2.9
Race/ethnicity
Non-Hispanic White	3882	89
Non-Hispanic Black	547	7.8
Other	142	3.3
Medical conditions
Hypertension	2655	56
Diabetes	939	20
Heart problems	1133	24
Stroke	314	7.1
Depression1	627	14
Obesity	1349	30
Cognitive
Global cognition	21.9	4.8
Lifestyle, psychosocial and neuroinflammatory factors
Sedentary behavior	2017	43
Smoking	2573	58
Poor sleep1	880	19

Footnote: * The characteristics are reported with raw sample n and weighted sample %. We present the raw sample n and weighted sub-sample n to show the number of observations the estimates are based on, the percentages are the weighted proportion estimates (the population estimates of the proportions). 1. Requiring medication for this condition.

Table 1. Characteristics of participants at baseline (n=4,571).

	Means or n	SD or percentages
Demographic factors
Age (years)	71.0	8.1
Female	2671	56
Higher levels of education (years)	12.5	2.9
Race/ethnicity
Non-Hispanic White	3882	89
Non-Hispanic Black	547	7.8
Other	142	3.3
Medical conditions
Hypertension	2655	56
Diabetes	939	20
Heart problems	1133	24
Stroke	314	7.1
Depression1	627	14
Obesity	1349	30
Cognitive
Global cognition	21.9	4.8
Lifestyle, psychosocial and neuroinflammatory factors
Sedentary behavior	2017	43
Smoking	2573	58
Poor sleep1	880	19

Footnote: * The characteristics are reported with raw sample n and weighted sample %. We present the raw sample n and weighted sub-sample n to show the number of observations the estimates are based on, the percentages are the weighted proportion estimates (the population estimates of the proportions). 1. Requiring medication for this condition.

Table 2. Risk of incident self-reported Alzheimer’s Disease and related dementias (ADRD) and cognitive impairment in individuals with elevated diabetes risk.

	Time coefficients (95% CI)	p	Follow-up time	events
Self-reported ADRD	-0.095(-0.165, -0.026)	0.008	9.23, 4.65	583
With adjustments for apathy	-0.076 (-0.145, -0.006)	0.032
Cognitive impairment	-0.162 (-0.231, -0.092)	<0.001	8.89, 4.71	938
With adjustments for apathy	-0.144 (-0.210, -0.078)

Note: The survival estimates are based on Accelerated Failure Time (AFT) with Weibull distribution.

Table 2. Risk of incident self-reported Alzheimer’s Disease and related dementias (ADRD) and cognitive impairment in individuals with elevated diabetes risk.

	Time coefficients (95% CI)	p	Follow-up time	events
Self-reported ADRD	-0.095(-0.165, -0.026)	0.008	9.23, 4.65	583
With adjustments for apathy	-0.076 (-0.145, -0.006)	0.032
Cognitive impairment	-0.162 (-0.231, -0.092)	<0.001	8.89, 4.71	938
With adjustments for apathy	-0.144 (-0.210, -0.078)

Note: The survival estimates are based on Accelerated Failure Time (AFT) with Weibull distribution.

When using self-reported diagnoses of ADRD as the outcome, 583 out of 4,571 older adults reported a diagnosis over a mean follow-up of 9.2, SD 4.6 years. In the ATF model, adjusted for age, sex, and education, diabetes risk (higher HbA1c) was associated with shorter time to ADRD diagnosis (time-to-event coefficient-0.09; 95%CI: -0.16, -0.02; =p=0.008). For the analyses of incident cognitive impairment, we excluded 346 adults who met criteria for cognitive impairment at baseline. In the remaining 4,225 older adults, 938 developed cognitive impairment over the follow-up. In the ATF model, adjusted for age, sex, and education, diabetes (higher HbA1c) was associated with shorter time to incident cognitive impairment (time-to-event coefficient -0.16; 95% CI:-0.23, -0.09; p<0.001).

3.1. Mediation

In the survey-weighted linear regression model, diabetes (higher HbA1c levels) was significantly associated with greater apathy (β = 0.01; 95% CI: 0.007, 0.03; p = 0.002). In Weibull ATF models, after including apathy as a mediator the association between diabetes and time to ADRD diagnosed changed but remained significant (coefficient: -0.07; 95C%CI: -0.14, -0.006; p=0.032). Mediation analysis using a product-of-coefficients approach indicated a statistically significant indirect effect of diabetes on incident ADRD through apathy (indirect effect = 1.31; SE = 0.39; z = 0.30; p = 0.0009).

Similarly in the analysis for incident cognitive impairment, the association between diabetes risk and time to cognitive impairment changed as: coefficient = −0.14; 95% CI: −0.21, −0.07; p<0.001; but remained significant, after the inclusion of apathy in the model. Mediation analysis demonstrated a statistically significant indirect effect of diabetes on incident cognitive impairment through apathy (indirect effect = 1.32; SE = 0.43; z = 0.78; p = 0.002), indicating that apathy partially mediates the association between diabetes and time to cognitive impairment.

4. Discussion

Diabetes and apathy are known risk factors for Alzheimer’s disease and related dementias (ADRD). In this nationally representative sample of older adults, we found that diabetes risk as measured by glycosylated hemoglobin, was associated with a) a shorter time to incident ADRD —defined by self-reported diagnosis — and b) shorter time to incident cognitive impairment — defined as impairment in cognitive testing and instrumental activities of daily living. Importantly, both these associations were partially mediated by apathy suggesting a clinically observable and a potentially modifiable intermediate pathway linking metabolic dysfunction to cognitive decline.

4.1. Diabetes, Apathy and Cognitive Decline

The neural correlates of T2 diabetes and apathy may overlap. Type 2 diabetes is associated with cerebrovascular pathology, including white matter hyperintensities, lacunar infarcts, and cerebral microbleeds.[29,30,31,32] These associations likely reflect the effects of chronic hyperglycemia, insulin resistance, and advanced glycation end-products contributing to endothelial dysfunction, blood–brain barrier disruption, and impaired cerebral perfusion. Over time, vascular changes may affect brain networks that regulate motivation, goal-directed behavior, and reward processing such as fronto-striato-thalamic circuits —manifesting apathy.[33,34] Apathy may subsequently accelerate cognitive decline through either or a combination of direct and indirect pathways. Directly, apathy has been associated with structural and functional brain changes in frontal and limbic regions implicated in executive function and memory.[35,36] Indirectly, apathy can reduce participation in cognitively stimulating, social, and physical activities—behaviors known to promote cognitive reserve and neuroplasticity.[37] Further, altered behaviors such as reduced physical activity may exacerbate worsening of glycemic control and cerebrovascular health, creating a self-reinforcing cycle.[38]

Beyond vascular mechanisms, diabetes is characterized by chronic systemic low-grade inflammation.[39,40] Similarly, apathy is associated with elevated inflammatory markers as well.[41,42] Elevated inflammatory markers—such as C-reactive protein, serum amyloid A, interleukin-6, and tumor necrosis factor-alpha—can cross the blood–brain barrier and promote neuroinflammation resulting in synaptic dysfunction, neuronal injury, as well as the accumulation of amyloid-β and tau pathology.[39,40,41,43] Inflammation may also affect dopaminergic and other neurotransmitter systems involved in motivation and reward processing, potentially providing a mechanistic link between metabolic dysregulation and apathy.[44]

4.2. Apathy vs Depression

Apathy is a long-recognized neuropsychiatric syndrome, yet it has often been conflated with depression due to overlapping symptoms such as reduced motivation and interest. Recent evidence, however, shows that apathy and depression are dissociable constructs with unique neurobiological underpinnings and clinical trajectories.[45] Importantly, apathy has been identified as an independent risk factor for cognitive decline and ADRD, even after accounting for depressive symptoms. Prior research —using the HRS—has established an association between type 2 diabetes and worsening episodic memory mediated by depression. [46,47] In the present study, we specifically examined apathy as a candidate mediator as it represents a plausible yet underexplored pathway linking diabetes to cognitive decline. Identifying apathy as a distinct mediator is clinically meaningful, since it may have different therapeutic implications than depression, including behavioral activation tailored to motivational deficits and specific pharmacologic targets.

Despite growing recognition of apathy’s importance, its measurement remains a major challenge in epidemiologic research. The most widely used assessment tools, the Neuropsychiatric Inventory (NPI) and Apathy Evaluation Scale (AES), are designed for informant-report or may have been omitted from many large-scale, population-based studies due to practical considerations. As a result, many cohorts—including the HRS (except for the subset of ADAMS)—must rely on subscales derived from depression scales, which may not adequately capture the emotional, behavioral, and cognitive motivational domains of apathy.[1,2] This limits comparison across studies and may obscure the true magnitude of apathy’s role in cognitive aging.

4.3. Clinical Implications

Our findings suggest apathy partially mediates the association between increased diabetes risk and cognitive impairment, highlighting it as a clinically observable marker for accelerated cognitive decline in individuals with diabetes that should be incorporated into routine screening. It is also a modifiable mechanistic target for ADRD prevention. In prior clinical trials, cholinesterase inhibitors, particularly Donepezil, have shown promise in reducing apathy symptoms across dementia severity levels.[48] Other agents, such as psychostimulants and dopaminergic medications have been explored, yet, not robustly tested in randomized controlled trials of older adults without Alzheimer’s disease. Conversely, evidence for behavioral interventions specifically targeting apathy are limited, despite recommendations in prior reviews to combine pharmacologic and non-pharmacologic approaches.[50] This represents an important opportunity for behavioral intervention development, potentially engaging caregiver-participant dyads to identify individuals’ prior interests to direct them for tailored interventions. For example, art-based interventions—such as visual art activities or storytelling—potentially delivered in intergenerational contexts, may help re-engage motivational systems while also strengthening social connections.

4.4. Future Research

A few directions for future research emerge from our findings. First, to maximize sample size, in the current study, apathy was derived from items of the CES-D; however, future studies should incorporate more robust, multidimensional apathy-specific measures to improve construct precision. Second, although apathy partially mediated the association between diabetes and cognitive decline, a substantial proportion of the relationship remained unexplained. Identifying additional mechanistic pathways—such as polypharmacy, frailty, other cardiovascular diseases—may reveal complementary targets for personalized interventions. Third, given that apathy was a mediator in our analyses and prior research has consistently identified apathy as a predictor of adverse health outcomes, there is a need to better understand its underlying biology. Integrating proteomic and other biomarker approaches could clarify the molecular and metabolic mechanisms linking apathy, diabetes, and cognitive aging.

4.5. Strengths and Limitations

Key strengths of this study include the use of a nationally representative sample followed for an average of nearly 10 years, the incorporation of objective biomarkers of diabetes (glycated hemoglobin), and the examination of associations with incident ADRD and incident cognitive impairment. The longitudinal design strengthens temporal inference while the sample enhances generalizability. Several limitations warrant consideration. As mentioned, apathy was derived from CES-D items rather than assessed using a validated, multidimensional apathy-specific instrument. Although this approach has been used in prior longitudinal cohorts (e.g., the Baltimore Epidemiologic Catchment Area Study and the Prevention of Dementia by Intensive Vascular Care study), it may not fully capture the construct of apathy or adequately distinguish it from depressive symptoms. Although glycosylated hemoglobin provides an objective marker of chronic glycemic exposure, it reflects average glucose levels over an extended time window and does not capture short-term glycemic fluctuations. Similarly, CES-D responses were self-reported, and cognitive assessments were episodic rather than continuous. Future naturalistic studies leveraging high-frequency digital measures—such as continuous glucose monitoring and ecological momentary assessment—could better characterize short-term dynamics and temporal coupling among metabolic dysfunction, motivational changes, and cognitive fluctuations. Finally, we did not examine alternative mechanistic pathways linking diabetes and cognitive decline, as our analyses focused specifically on apathy.

5. Conclusions

This study adds to the existing evidence on diabetes and dementia, by showing that elevated diabetes risk is associated with accelerated time to onset of ADRD and cognitive impairment and further demonstrates that apathy partially mediates these relationships. These findings highlight that routine screening for apathy symptoms in individuals with diabetes may 1) improve risk stratification and 2) as both diabetes and apathy are modifiable risk factors, inform the development of interventions.