Combined Associations of Diabetes and Depressive Symptoms on Cognitive Impairment Risk in the Population Aged 60 or Older: Cross-Cohort Comparisons From NHANES and CHARLS
Article information
Abstract
Objective
Diabetes mellitus (DM) and cognitive impairment (CoI) are correlated, but the combined impact of depressive symptoms on CoI risk in older adults remains unclear.
Methods
This study included 1,200 U.S. adults aged ≥60 years (National Health and Nutrition Examination Survey, NHANES) and 1,500 Chinese adults (China Health and Retirement Longitudinal Study, CHARLS). DM was defined by laboratory measures or self-report; depressive symptoms were assessed via the Patient Health Questionnaire-9 (PHQ-9) (NHANES) and CES-D-10 (CHARLS); CoI was measured using the CERAD battery (NHANES) and situational memory/mental integrity scores (CHARLS). Multivariate logistic regression estimated associations of DM and depressive symptoms with CoI. Additive interaction was evaluated by the relative excess risk due to interaction (RERI), attributable proportion (AP), and synergy index (S). Restricted cubic spline (RCS) analyses examined nonlinear DM–depressive symptoms interactions.
Results
Both DM and depressive symptoms were independently and significantly associated with CoI. After multivariate adjustment, diabetic patients exhibited significantly increased CoI risk; depressive symptoms similarly elevated CoI risk. Interaction analysis revealed an additive effect when DM co-occurred with depressive symptoms in CHARLS but not in NHANES. Mediation analyses further suggest bidirectional associations: in CHARLS, diabetes and depressive symptoms appear to mediate each other’s effects on CoI. Additionally, RCS analysis in NHANES indicated a nonlinear interaction between depressive symptom severity (PHQ-9) and DM on CoI (p<0.05).
Conclusion
The observed associations varied across cohorts, with a significant additive interaction between diabetes and depressive symptoms found only in the CHARLS cohort. Integrating metabolic and mental health screening and interventions may optimize cognitive outcomes in high-risk older adult populations.
INTRODUCTION
Major depressive disorder (MDD) is a highly prevalent mental disorder and the third leading cause of the global burden of mental illness. From 2010 to 2021, the age-standardized disability-adjusted life year rate for depressive symptoms exhibited the most substantial increase, with a 16.4% rise (95% eISSN 1976-3026 OPEN ACCESS confidence interval [CI]: 11.9% to 21.3%) among the 25 leading Tier 3 causes [1]. Individuals grappling with depressive symptoms often manifest symptoms such as persistent low moods, diminished interest in activities they typically find pleasurable, and a reduced capacity to engage in daily activities [2]. These symptoms can have significant health implications, including a decline in physical well-being, and hinder an individual’s ability to function optimally in their professional life. Research has demonstrated a correlation between depressive symptoms and suboptimal performance on various cognitive functioning assessments. Even when controlling for confounding variables, depressive symptoms show a significant association with lower scores in language, processing speed, and memory [3]. Cognitive impairment (CoI) is a common feature of MDD, and even after depressive symptoms subside, patients may still experience problems with memory, executive functioning, and attention [4]. It is worth noting that depressive symptoms is more prevalent in the older adult population, with a prevalence of up to 13.2% in later life [5,6]. In addition, studies have shown a significant correlation between depressive symptoms and CoI [7]. Individuals with depressive symptoms have been observed to have a higher risk of dementia, in part due to the higher prevalence of CoI in this population [8].
Diabetes mellitus (DM) is a metabolic disease characterized by elevated blood glucose levels and has traditionally been thought to involve primarily abnormalities in glucose metabolism. In recent years, the medical community has recognized that diabetes is a multifaceted disease with potential implications that go far beyond the simple regulation of blood sugar [9]. A mounting body of research has underscored the potential impact of diabetes on cognitive function, as the global incidence of the disease continues to rise [10]. Clinical practice has shown that DM is often associated with a variety of complications, including cardiovascular disease, neuropathy, and retinopathy [11]. However, a less explored aspect is its potential association with cognitive performance. The potential for CoI to complicate the management of disease in older patients with diabetes has been demonstrated, with observed difficulties in following treatment regimens. This, in turn, has been shown to correlate with glycemic control, increase the risk of complications, and ultimately compromise the patient’s overall health and quality of life [12]. Given the established association between diabetes and cognitive function, it is imperative to undertake a comprehensive exploration of the relationship between diabetes and CoI.
CoI is typified by a decline in one or more cerebral functions, including memory, learning, attention, decision-making, and executive functioning. The severity of this decline can range from mild to severe [13]. On a global scale, the prevalence of CoI is increasing, particularly among individuals over the age of 65. CoI is a prevalent feature among patients diagnosed with CoI. Although it does not fully align with the diagnostic criteria for dementia, its prevalence increases significantly with age, reaching as high as 19.7% in certain cases [14]. If CoI persists and worsens, patients are often diagnosed with dementia, which is currently the fifth leading cause of death worldwide [15]. This condition significantly impairs patients’ ability to care for themselves, resulting in a substantial disease burden. The financial burden on patients and society is significant, with healthcare expenditures for individuals with low cognitive function exceeding $300 billion [16]. Therefore, ascertaining the factors associated with CoI and how they interact is imperative for the development and implementation of effective interventions that may help reduce the incidence of CoI and enhance patient quality of life [17-19].
Although DM and depressive symptoms are each well-established as independent risk factors for CoI, the elusive nature of their synergistic interplay and underlying mechanisms remains poorly characterized. This study postulates that the co-occurrence of DM and depressive symptoms shows statistical associations with CoI in older adults, and that these two conditions may influence each other. To rigorously evaluate this interaction hypothesis, we harnessed data from two nationally representative cohorts spanning Eastern and Western populations—the China Health and Retirement Longitudinal Study (CHARLS) and the National Health and Nutrition Examination Survey (NHANES)—conducting a comprehensive cross-cohort analysis.
METHODS
Study design and population
This study utilized data from two nationally representative prospective cohort studies: the CHARLS and the NHANES, conducted in China and the United States, respectively. CHARLS focuses on the Chinese population aged 45 years and older, collecting comprehensive household and individual data from over 10,000 households across 150 counties and 450 villages nationwide. Launched in 2011, CHARLS conducts a survey every 2–3 years; the most recent national follow-up data was scheduled for release on 16 November 2023. NHANES evaluates the health and nutritional status of the U.S. population by interviewing and examining approximately 5,000 nationally representative individuals from 15 counties annually. In this analysis, we included CHARLS data from 2015 and NHANES data from 2011 to 2014. Exclusion criteria were consistent across both datasets and included: 1) participants younger than 60 years of age; 2) individuals without diabetes data, cognitive functioning test results, or depressive symptoms scale scores; and 3) individuals with missing covariates. After applying these criteria, 10,178 participants from CHARLS and 3,632 participants from NHANES were ultimately included (Figure 1).
Flowchart of participants in this study. CHARLS, China Health and Retirement Longitudinal Study; NHANES, National Health and Nutrition Examination Survey.
Assessment of depressive symptoms
In the NHANES cohort, depressive symptoms was assessed using the Patient Health Questionnaire-9 (PHQ-9), a self-report instrument based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition [20]. The PHQ-9 evaluated nine depressive symptoms over the past 2 weeks, with each item scored on a 4-point scale (0=“not at all,” 1=“a few days,” 2=“more than half the days,” 3=“almost every day”). Total scores ranged from 0 to 27, with higher scores indicating more severe depressive symptoms [21]. In the CHARLS cohort, depressive symptoms were measured using the 10-item Centre for Epidemiological Studies Depression Scale (CES-D-10), which evaluates participants’ self-reported experiences over the past week. The scale comprises items assessing feelings of being bothered by normally non-distressing things, inattention, low mood, perceiving tasks as burdensome, hopelessness, fearfulness, sleep disturbance, happiness, loneliness, and difficulty initiating action. Each item offers four response options: little or no time, some or very little, occasional or moderate time, and most or all the time. Responses are scored on a 0–3 scale, except for items 5 and 8, which are reverse-scored [22]. For both instruments, depressive symptoms severity was categorized as follows: normal (<10), mild to moderate (10–19), and severe (≥20).
Diabetes status
The diagnostic criteria for DM include either a glycated hemoglobin (HbA1c) level of ≥6.5% (48 mmol/mol) or a fasting plasma glucose (FPG) level of ≥126 mg/dL (measured after 8–24 hours of fasting). The diagnosis of prediabetes is established when HbA1c levels range from 5.7% to 6.4% (39–46 mmol/mol) or when FPG levels are between 100 and 125 mg/dL. In contrast, normoglycemia is defined as an HbA1c level of less than 5.7% (<39 mmol/mol) and an FPG level of less than 100 mg/dL. Additionally, the diagnosis of diabetes may be based on clinical self-reporting by a physician or healthcare professional [23].
Cognitive function examination and composite scoring
In the NHANES cohort, four core tests are commonly used to assess cognitive function: the Consortium for Alzheimer’s Disease Word Learning Registry (CERAD-WL) test, the Animal Fluency (AF) test, the Digit Symbol Substitution Test (DSST), and the CERAD Delayed Recall (CERAD-DR) test [24,25]. We used Z scores, a globally standardized total cognitive functioning score incorporating the CERAD-WL, CERAD-DR, AF, and DSST tests, to eliminate unequal variations in individual cognitive scores. The Z scores were determined using the formula Z=(x–m)/σ, where σ is the overall standard deviation, m is the overall mean, and x is the raw score. The Z score of an individual is less than -1 if that person exhibits CoI [26,27].
In the CHARLS cohort, cognitive functioning was assessed across two primary domains. For situational memory, participants were required to listen to the interviewer read ten words aloud and then recall them immediately (immediate recall) and again approximately ten minutes later (delayed recall). Scores were calculated as the average of immediate and delayed recall performance (range: 0–10). As for mental completeness, this domain included tasks such as drawing an overlapping pentagon, answering questions about the current date/season/day of the week, and consecutively subtracting 7 from 100 five times. One point was awarded for each correctly answered task, with a maximum total score of 11. The composite score across all items (0–21 points) represented participants’ overall cognitive status, where higher scores indicated better cognitive functioning. Following previous research, cognitive dysfunction was defined as cognitive function scores ≥1.0 standard deviation below the mean [28].
Assessment of covariates
A series of covariates was incorporated into this study to account for potential confounders in the CHARLS and NHANES cohorts. Common covariates across both datasets encompassed age (categorized as 60–69 years and ≥70 years), sex, education level (less than high school, high school, college and above), smoking status, drinking status, body mass index (BMI; <25, 25–30, ≥30), ethnicity, and history of hypertension. However, notable discrepancies existed between the two cohorts regarding categorization and specific variables. In CHARLS, race was dichotomized as Han Chinese and other races, whereas NHANES utilized a more comprehensive classification: Mexican American, other Hispanic, non-Hispanic white, non-Hispanic black, and other races. Concerning drinking status, CHARLS distinguished between current non-drinkers and current drinkers, while NHANES defined excessive drinking based on consumption patterns in the 12 months before the interview—individuals consuming an average of five or more alcoholic beverages per day during this period were categorized as excessive drinkers [29].
Statistical analysis
This study analyzed data from the U.S. NHANES and the CHARLS, with data from the 2015 CHARLS wave selected for analysis. NHANES utilizes a complex stratified multistage probability sampling design; by established guidelines, accounting for sample weights, stratification, and primary sampling units was critical to addressing the issue of oversampled ethnic minority populations. Categorical variables were presented as proportions. A chi-square test was performed on the survey data to evaluate differences in categorical characteristics between the non-CoI and CoI groups.
This study utilized multivariate logistic regression models to examine the associations among depressive symptoms, DM, and CoI. Three models were constructed: Model 1 was an unadjusted crude model without confounder correction; Model 2 was adjusted for age and education; and Model 3 included additional adjustment for marital status, hypertension, and smoking on top of the variables in Model 2. Additive interaction models were developed to investigate potential interactions. For the CoI, the additive interaction between diabetes and depressive symptoms was evaluated by assessing whether their joint effect exceeded the sum of their individual effects. This was operationalized using the relative excess risk (RERI), attributable proportion (AP), and synergy index (S). An additive interaction was considered present when the 95% CIs of RERI and AP did not include 0, and the CI of S did not include 1. Mediation analyses were employed to explore bidirectional mediating effects between DM, depressive symptom severity, and CoI. Restricted cubic spline (RCS) curve-fitting analyses were used to characterize non-linear associations and interactions between depressive symptom severity—measured by the PHQ-9 scale in the NHANES cohort and the CES-D-10 scale in the CHARLS cohort—and DM on the outcome (CoI).
All analyses were performed using R version 4.4.2 (http://www.R-project.org) and Empower software (X&Y Solutions); considered statistically significant at two-tailed p<0.05.
Ethics statement
CHARLS was approved by the Biomedical Ethics Review Committee of Peking University, and all participants provided informed consent. The National Center for Health Statistics Ethics Review Committee approved the survey protocols and research procedures for NHANES, with informed consent obtained from all participants/patients.
This study was approved by the Ethics Committee of the Xijing Hospital of the Fourth Military Medical University, approval number KY20242298-C-1.
RESULTS
Description of study participants
Tables 1 and 2 present the demographic characteristics of participants in the NHANES and CHARLS cohorts, respectively. Following screening, 10,178 participants from CHARLS and 3,632 participants from NHANES were included in the analysis.
Baseline characteristics of participants by category of CoI (NHANES)
Baseline characteristics of participants by category of CoI (CHARLS)
In the NHANES cohort, the CoI group was older, with a higher proportion of individuals with less than a high school education (60.08%), a greater proportion of non-Hispanic Black participants (29.41%), and a slightly higher proportion of widowed/divorced/separated/unmarried individuals (46.64%). This group also exhibited significantly higher prevalences of hypertension (70.17%), diabetes (45.80%), and depressive symptoms (18.91%) compared to the non-CoI group. In summary, CoI in the U.S. cohort was associated with lower educational attainment, a higher proportion of non-Hispanic Black individuals, and greater burdens of metabolic and mental health conditions.
In the CHARLS cohort, the CoI group similarly comprised older individuals, with a higher proportion of those with less than a high school education (91.81%) and a significantly greater proportion of widowed/divorced/separated/unmarried participants (77.79%). Significantly higher prevalences of diabetes (32.50%) and depressive symptoms (42.67%) were also observed in this group relative to the non-CoI group. Overall, CoI in the Chinese cohort was characterized by lower educational attainment, a higher proportion of unmarried or widowed status, and a notable co-occurrence of diabetes and depressive symptoms.
Association between DM and CoI
As shown in Figure 2, after adjusting for potential confounders such as age, education, marital status, smoking, and hypertension, analyses in both the NHANES cohort (assessed using the PHQ-9) and the CHARLS cohort (assessed using the CES-D-10) showed that diabetic participants had a signif-icantly higher odds of clinically significant CoI than non-diabetic participants. Specifically, the odds ratio (OR) for major depressive symptoms was 1.61 (95% Cl: 1.14 to 2.27, p<0.001) in the NHANES cohort, and 1.93 (95% CI: 1.67 to 2.22, p< 0.001) in the CHARLS cohort.
Multivariable logistic regression of DM and depressive symptoms in participants with cognitive impairment (NHANES and CHARLS). Model 1: unadjusted; Model 2: Model 1+age and education; Model 3: Model 2+marital status, smoking, and hypertension. CHARLS, China Health and Retirement Longitudinal Study; NHANES, National Health and Nutrition Examination Survey; DM, diabetes mellitus; OR, odds ratio; CI, confidence interval.
Association between depressive symptoms and CoI
As shown in Figure 2, after adjusting for potential confounders such as age, education, marital status, smoking, and hypertension, analyses in both the NHANES cohort and the CHARLS cohort indicated that participants with depressive symptoms had a significantly higher odds of clinically significant CoI than those without depressive symptoms. Specifically, the data showed that the OR for CoI was 1.94 (95% CI: 1.19 to 3.16, p<0.001) in the NHANES cohort and 1.71 (95% CI: 1.50 to 1.94, p<0.001) in the CHARLS cohort for those with depressive symptoms.
Effect of the interaction between diabetes and depressive symptoms on CoI
In both cohorts, diabetes or depressive symptoms (either alone or in combination) were significantly associated with CoI. In the CHARLS cohort, the OR was 3.32 (95% CI: 2.67 to 4.14, p<0.001) (Table 3). In the NHANES cohort, the OR was 3.73 (95% CI: 1.81 to 7.68, p<0.001) (Table 4). No significant interaction was observed in the NHANES cohort. In the CHARLS cohort, the RERI (0.92–1.1), AP (0.28–0.31), and S (1.66–1.74) all indicated the presence of additive interactions (Figure 3).
Analysis of the interactive effect of DM and depressive symptoms (CHARLS)
Analysis of the interactive effect of DM and depressive symptoms (NHANES)
Interaction of depressive symptoms, diabetes exposure, and CoI risk visualized across NHANES (A) and CHARLS (B) cohorts. CHARLS, China Health and Retirement Longitudinal Study; NHANES, National Health and Nutrition Examination Survey; DM, diabetes mellitus.
Mediation analysis
After adjusting for all covariates, when DM served as the independent variable and depressive symptom severity as the mediator, both the NHANES cohort (Figure 4A) and the CHARLS cohort (Figure 4C) exhibited significant direct effects (NHANES: DE=0.0465, p=0.018; CHARLS: DE=0.1158, p<0.001) and indirect effects (NHANES: IE=0.0031, p=0.028; CHARLS: IE=0.0058, p=0.002), with mediation ratios of 6.26% and 4.78%, respectively. When depressive symptom severity was the independent variable and DM the mediator, only the CHARLS cohort (Figure 4D) showed significant direct (DE=0.091, p<0.001) and indirect (IE=0.0068, p=0.002) effects (mediation proportion=6.98%), whereas the NHANES cohort (Figure 4B) had a nonsignificant indirect effect (IE=0.001, p=0.424).
Bidirectional mediating effects of depressive symptom severity and DM on CoI in the NHANES (A and B) and CHARLS (C and D) cohorts. CHARLS, China Health and Retirement Longitudinal Study; NHANES, National Health and Nutrition Examination Survey; DM, diabetes mellitus; IE, indirect effect; DE, direct effect; CoI, cognitive impairment.
Generative additive model
Generative additive model using fully adjusted models revealed a significant association pattern among DM, depressive symptoms scores, and CoI. In the NHANES cohort, DM exerted a significant effect on CoI (β=0.891, z=3.422, p=0.001), with a significant nonlinear interaction observed between DM and PHQ-9 scores: the linear term interaction estimate was -0.355 (z=-2.466, p=0.014), and the quadratic term interaction estimate was 0.748 (z=2.670, p=0.008). Of clinical relevance, the RCS analysis revealed that the effect of diabetes on CoI began to increase notably once PHQ-9 scores exceeded approximately 4 points—a threshold well below the standard diagnostic cutoff of 10 for clinically significant depressive symptoms (Supplementary Figure 1 and Supplementary Table 1). In the CHARLS cohort, DM significantly influenced CoI (β=0.503, z=2.902, p=0.004), whereas CES-D-10 scores exhibited only a weak linear trend with CoI (β=0.080, z=1.729, p=0.084) and showed no significant interaction with DM (all p>0.05).
DISCUSSION
This study is the first to examine the role of diabetes and depressive symptoms on CoI in older adults using a highly representative prospective cohort database in East Asia and North America. The results showed a significant synergistic effect of diabetes and depressive symptoms on CoI in older adults.
Evidence has demonstrated that individuals diagnosed with diabetes exhibit substantial damage to both white and grey matter, with these alterations exhibiting a robust correlation with a deterioration in cognitive function [30,31]. Although some studies have explored the effects of obesity, medication, and other factors on cognitive function in diabetic patients, the exact mechanisms are unclear [32]. After controlling for confounders, this study found that diabetes was significantly associated with CoI.
In fully adjusted models, a significant correlation was identified between depressive symptoms and CoI. Several studies have indicated that individuals may show CoI during the initial episode of depressive symptoms. Furthermore, patients with recurrent depressive symptoms may have a higher likelihood of CoI compared to those experiencing a single episode of depressive symptoms [33]. Despite the administration of effective antidepressant medication, CoI remains a persistent issue among patients [34]. Depressive symptoms have been identified as a significant factor associated with CoI in the older adult population [35]. In older persons suffering from moderate to severe depressive symptoms, CoI may be a reliable indicator of dementia [13]. Structural magnetic resonance imaging studies have demonstrated volume reductions in the cerebellar vermis and the inferior temporal gyrus in patients with depressive symptoms. Furthermore, a comprehensive investigation of the brain’s structure and function has revealed significant atrophy in additional regions, including the temporal lobe, superior temporal gyrus, inferior frontal gyrus, amygdala, hippocampus, thalamus, and cingulate gyrus. The atrophy of these brain regions may indicate the presence of information-processing deficits in patients, reducing their ability to participate in mental and social activities. These activities are considered risk factors for COI and MDD [36,37]. Furthermore, the presence of white matter hyperintensity has been implicated as a potential causative factor contributing to the increased prevalence of neuropsychiatric conditions, including depressive symptoms [38]. In diabetic patients, Moran et al. [39] reported grey mat-ter loss in multiple brain regions, including the medial temporal lobe, anterior cingulate gyrus, and medial frontal lobe, in a large cross-sectional study. Additionally, the study observed white matter loss in frontal and temporal lobe regions. The study suggests that these changes in brain volume are strongly associated with poorer performance on cognitive tests in people with type 2 diabetes. Overall, both depressive symptoms and diabetes showed negative effects on brain structure and showed similar brain region changes in imaging studies.
In addition to CoI, diabetes is associated with mood changes that often co-occur with depressive symptoms in diabetics. Factors such as dysregulation of the hypothalamic-pituitaryadrenal (HPA) axis and overactivity of the sympathetic nervous system are closely associated with the development of diabetes and depressive symptoms. Pathological mechanisms such as chronic inflammation, atherosclerosis, and abnormal cytokine secretion may play an important role in the interrelationship between these two conditions [21,40]. These physiological mechanisms may be linked to mood changes in diabetes, which may in turn correlate with depressive symptoms. In older adults, diabetes and depressive symptoms are risk factors for CoI in earlier research. The underlying mechanisms are intricate, though. According to the present study, having diabetes and depressive symptoms together is associated with a significantly higher odds of CoI when compared to having either condition alone. The mechanisms underlying this association are not fully understood but may be complex. Given the observed association with cognition, this highlights the need for healthcare providers to work together to treat older adults with depressive symptoms and diabetes. For older individuals with diabetes or depressive symptoms, screening for additional illnesses may be advised to stop future CoIs.
This study revealed a potential interaction between diabetes and depressive symptoms that differed across cohorts. In both cohorts, diabetes or depressive symptoms (alone or in combination) were associated with a significantly higher risk of CoI, but an additive interaction was observed only in the CHARLS cohort. These findings share both similarities and differences with the results of previous studies. For instance, while Downer et al. [41] investigated the impacts of diabetes and depressive symptoms on CoI, they did not detect a similar interaction effect. In contrast, Liu et al. [42] demonstrated that diabetic patients have a significantly elevated risk of MDD but failed to evaluate the synergistic effect of these two conditions on CoI.
The results of the mediation analyses suggest that depressive symptom severity plays only a limited bridging role in the observed relationship between diabetes and CoI and that there is heterogeneity across populations. The bidirectional patterns of the mediators in the CHARLS cohort may be related to the assessment methods of depressive symptoms in Chinese older adults, differences in diabetes management, cultural factors, or environmental stress levels [43,44]. Several factors may contribute to the observed heterogeneity between the two cohorts. First, healthcare system differences may play a role: in China, diabetes management often relies on fragmented primary care services, whereas the U.S. has more structured chronic disease management programs, which may influence the timing and effectiveness of metabolic and mental health interventions [45,46]. Second, social support structures differ substantially between the two populations. Chinese older adults traditionally rely on family-based care, whereas Western populations often depend on formal social services and community support networks; variations in social support availability may modify the impact of comorbid conditions on cognitive health [47,48]. Third, cultural factors may influence both the expression and reporting of depressive symptoms. In Chinese culture, mental health symptoms are often somatized or underreported due to stigma, which may affect the detection and management of depression in older adults [49]. Fourth, differences in health behaviors, including dietary patterns, physical activity levels, and adherence to medical recommendations, may also contribute to the divergent interaction patterns observed [50,51]. Finally, genetic susceptibility and biological mechanisms—such as variations in inflammatory markers or HPA axis regulation across populations—may further explain the differential synergistic effects [52]. Although the small proportion of mediating effects suggests that additional biological (e.g., inflammatory processes, vascular lesions) and psychosocial mechanisms remain to be explored, this study emphasizes the importance of simultaneously monitoring depressive symptom severity in diabetes management. Future studies should adopt longitudinal designs, incorporating neuroimaging and biomarkers, to validate the potential protective effect of depressive symptoms-focused interventions on cognitive outcomes in patients with diabetes. The significant nonlinear interaction between PHQ-9 scores and diabetes in the NHANES cohort suggests that the impact of diabetes on CoI is partially attenuated at lower levels of depressive symptoms but becomes markedly potentiated once PHQ-9 scores exceed approximately four points. This threshold is clinically significant, as it falls well below the standard diagnostic cutoff of 10, suggesting that even subthreshold depressive symptoms may amplify the adverse cognitive effects of diabetes in older adults. This nonlinear pattern may reflect threshold effects in neuroinflammatory pathways or hyperactivation of the hypothalamic–pituitary–adrenal axis [53,54]. By contrast, no significant interaction between CES-D-10 scores and diabetes was observed in the CHARLS cohort, possibly owing to population differences in genetic susceptibility, social support structures, or diabetes management practices between Eastern and Western settings [55,56].
This study has two core strengths. First, the cross-cohort consistency of our findings strengthens the external validity of our conclusions by leveraging large-scale databases (NHANES and CHARLS) from both China and the United States, with samples spanning both Eastern and Western populations that differ significantly in ethnicity, culture, and lifestyle. Second, NHANES employs a multistage probability sampling design to ensure broad representativeness of the U.S. population, while CHARLS relies on a nationwide random sample from China, collectively enhancing the generalizability of the results.
This study has the following limitations. First, its cross-sectional design cannot establish causality, necessitating longitudinal studies to further validate these associations. Second, limited data timeliness may affect interpretability: the NHANES dataset covers only 2011–2014, with cognitive assessments restricted to samples from that cycle; CHARLS has a relatively short follow-up period, potentially impairing the detection of long-term associations. Third, reliance on self-reported data for core variables introduces susceptibility to recall bias or social desirability bias. Additionally, divergences between the two cohorts in racial composition, cultural background, and healthcare systems warrant caution when generalizing results across populations.
Conclusion
This study, using large-scale cohorts from East Asia and North America, found that the coexistence of diabetes and depressive symptoms is associated with a significantly higher odds of CoI in older adults, with significant additive interaction effects observed in the Chinese older adult population (CHARLS). Population heterogeneity is evident: depressive symptoms partially mediate the diabetes-CoI association in NHANES, while bidirectional mediation (diabetes ↔ depressive symptoms ↔ CoI) is suggested in CHARLS, potentially linked to cultural, socioeconomic, or healthcare differences. These conditions may contribute to CoI through shared biological pathways (e.g., inflammation, HPA axis dysregulation) and structural brain changes (e.g., hippocampal atrophy). Clinically, integrated screening and joint management of diabetes and depressive symptoms in older adults are warranted, with a focus on high-risk subgroups (e.g., low education, specific BMI-age profiles). Future longitudinal studies incorporating neuroimaging and biomarkers are needed to further investigate causal relationships and inform cross-cultural cognitive health strategies.
Supplementary Materials
The Supplement is available with this article at https://doi.org/10.30773/pi.2025.0446.
Key results of RCS regression analysis on the interaction between depression scores and diabetes in NHANES and CHARLS cohorts
RCS curve-fitting analysis of the interaction of PHQ-9/CES-D-10 with DM and cognitive impairment in the NHANES (A) and CHARLS (B) cohorts. RCS, restricted cubic spline; NHANES, National Health and Nutrition Examination Survey; CHARLS, China Health and Retirement Longitudinal Study; DM, diabetes mellitus; PHQ-9, Patient Health Questionnaire-9; CES-D-10, 10- item Centre for Epidemiological Studies Depression Scale.
Notes
Availability of Data and Material
The National Health and Nutrition Examination Survey (NHANES). (NHANES Questionnaires, Datasets, and Related Documentation) provided the data for this investigation.
Conflicts of Interest
The authors have no potential conflicts of interest to disclose.
Author Contributions
Each of the authors made significant contributions to this paper. Qilong Wang, Dongrong Zhao, Min Cai, and Huaning Wang formulated the research hypotheses. Siheng Ma analyzed the data, and Qilong Wang and Siheng Ma drafted the manuscript. Min Cai critically revised the manuscript for important intellectual content and supervised the writing of the manuscript. All authors approved the final version of the manuscript.
Funding Statement
This research was funded by the Clinical Research Program of the Xijing Hospital (XJZT24LY35) and the Fourth Military Medical University (No. 2023LC2306), Shaanxi Province Science and Technology New Star Project (No. 2023KJXX-024).
Acknowledgments
The author sincerely thanks all participants in NHANES and CHARLS.