Abstract
Background
Although depressive symptoms have been associated with subsequent cognitive outcomes, it remains unclear whether different long-term symptom patterns are linked to specific cognitive domains such as episodic memory and orientation.
Methods
We analyzed 7,211 participants from the China Health and Retirement Longitudinal Study (CHARLS). Based on a prespecified rule-based approach using CES-D-10 depressive symptom scores from 2011 to 2015, participants were categorized into four symptom pattern groups: never depressed, early-remission, late-onset, or persistent/recurrent. Linear regression was used to evaluate associations between these patterns and 2018 cognitive outcomes, adjusting for baseline cognition and potential confounders. Subgroup, sensitivity, and directional analyses were conducted to evaluate consistency and temporal direction. An exploratory pathway analysis examined whether grip strength and sleep disturbance might partly account for the association between the persistent/recurrent depressive symptom pattern and episodic memory. External validation was conducted using data from the Health and Retirement Study (HRS).
Results
Compared with the never depressed group, the late-onset and persistent/recurrent depressive symptom patterns were associated with lower 2018 episodic memory (adjusted coefficients: −0.932 [95% CI, − 1.227 to − 0.637] and − 0.842 [− 1.104 to − 0.580], respectively), whereas associations for the early-remission pattern were weaker. For orientation, significant associations were observed only for the late-onset and persistent/recurrent depressive symptom patterns. Associations were stronger among women and those with lower educational attainment. In an exploratory pathway analysis (n = 5,721), lower grip strength and greater sleep disturbance partly explained the association between the persistent/recurrent depressive symptom pattern and episodic memory (estimated indirect contribution: 16.9%). Directional analyses suggested potential bidirectionality between depressive symptoms and episodic memory. External validation in the HRS showed associations in the same direction.
Conclusions
Persistent/recurrent and late-onset depressive symptom patterns were associated with poorer subsequent cognitive outcomes, particularly among women and individuals with lower educational attainment. Lower grip strength and more severe sleep disturbance may partly account for the association between the persistent/recurrent depressive symptom pattern and episodic memory. Directional analyses also indicated a bidirectional interplay between depressive symptoms and cognitive outcomes. Our findings suggest that long-term monitoring of depressive symptom patterns, together with attention to physical function and sleep health, may help identify individuals at elevated risk of cognitive decline.
Supplementary Information
The online version contains supplementary material available at 10.1186/s13690-026-01966-4.
| Text box 1. Contributions to the literature | |
|---|---|
| • While late-life depression has been associated with cognitive outcomes, it remains unclear whether specific long-term depressive symptom patterns are differentially related to later cognitive performance. | |
| • Our findings suggest that persistent/recurrent and late-onset depressive symptom patterns were associated with lower episodic memory performance, and that lower grip strength and greater sleep disturbance may partly account for the association observed for the persistent/recurrent depressive symptom pattern. | |
| • These findings support the value of longitudinal monitoring of depressive symptom patterns and suggest that physical function and sleep disturbance may help identify individuals at higher risk of poorer cognitive outcomes with aging. |
Introduction
Cognitive decline is common in later life and is associated with loss of independence, reduced quality of life, and increased mortality [1, 2]. As populations age worldwide, long-term depressive symptom patterns have become an important topic in geriatric medicine and public health [3]. Depressive symptoms deserve particular attention because they are common in older adults, potentially modifiable, and often co-occur with chronic conditions and functional limitations [4].
Prior studies have reported associations between depressive symptoms and cognitive outcomes or dementia risk [5, 6], but the longitudinal course of depressive symptoms has received comparatively less attention in relation to cognitive aging [7]. In midlife and later life, depressive symptoms may persist, remit, or emerge later [8]. These long-term patterns may reflect distinct underlying processes and could be differentially associated with later cognitive outcomes [9]. In addition, depressive symptoms and cognitive function may influence each other over time, with poorer cognition potentially contributing to later depressive symptoms [10]. Episodic memory and orientation are important cognitive domains that may be sensitive to early change [11].
Biological mechanisms, including inflammation and neuroendocrine dysregulation, have been proposed to link depressive symptoms with cognitive outcomes, although findings remain inconsistent [12, 13]. However, such mechanisms are difficult to characterize comprehensively in large cohort studies. In this context, we focused on two functional factors that commonly co-occur with depressive symptoms and are also associated with cognitive outcomes: grip strength and sleep disturbance [14]. Sleep disturbance is common in middle-aged and older adults, has been linked to both depressive symptoms and later cognitive impairment, and may represent a potentially modifiable factor of public health relevance [15, 16]. Grip strength is a broad marker of physical reserve and functional status and may provide additional insight into the association between depressive symptoms and cognitive aging [17]. Examining these factors alongside long-term depressive symptom patterns may help clarify their relevance to cognitive outcomes.
Associations between depressive symptoms and cognitive outcomes may differ across population subgroups. Sex and educational attainment are strongly related to both depressive symptom burden and cognitive function [18]. Women generally experience a greater burden and persistence of depressive symptoms across the life course [19], whereas educational attainment is often considered a proxy for cognitive reserve [20]. Furthermore, age, baseline cognitive status, and somatic comorbidities may also shape vulnerability to depression-related cognitive decline [21].
Instead of relying on single-point assessments, we used repeated the China Health and Retirement Longitudinal Study (CHARLS) measurements (2011–2015) to identify long-term depressive symptom patterns [22]. We then examined prospective associations with episodic memory and orientation in 2018. We also examined potential heterogeneity by sex, educational attainment, age, baseline episodic memory, and chronic disease count. An exploratory pathway analysis was performed to assess whether grip strength and sleep disturbance might help explain the association between depressive symptom patterns and later cognitive outcomes. To address potential reverse causation between depressive symptoms and cognitive function, we conducted bidirectional analyses. Finally, the directional consistency of our main findings was evaluated in an independent sample from the Health and Retirement Study (HRS) [23].
Methods
Study population
We used data from CHARLS, an ongoing nationally representative cohort of middle-aged and older adults in China [22]. CHARLS used multistage probability sampling and collected information through face-to-face interviews and physical examinations. Baseline data were collected in 2011, with follow-up waves in 2013, 2015, and 2018. For the present analysis, we included participants aged 45 years or older at baseline who had valid CES-D-10 assessments in at least two of the 2011, 2013, and 2015 waves and completed the cognitive assessment in 2018. We excluded participants with missing grip strength data in 2015 or body mass index (BMI) ≥ 50 kg/m². The final analytic sample for the main analyses included 7,211 participants. The participant selection process is shown in Supplementary Figure S1.
Assessment of depressive symptoms and pattern classification
Depressive symptoms were assessed in 2011, 2013, and 2015 using the 10-item Center for Epidemiologic Studies Depression Scale (CES-D-10) [24]. Total scores range from 0 to 30, with higher scores indicating more depressive symptoms. Consistent with prior CHARLS studies, a score of 10 or higher was used to indicate elevated depressive symptoms [25].
Based on this cutoff, we used a prespecified rule-based classification to categorize participants into four interpretable patterns defined by clinically meaningful symptom transitions [8]. Participants were classified as follows:
never depressed: non-elevated in both 2011 and 2015, with 2013 also non-elevated or missing.
early-remission: elevated in 2011 and non-elevated in 2015, with 2013 being elevated, non-elevated, or missing.
late-onset: non-elevated in 2011 and elevated in 2015, with 2013 being elevated, non-elevated, or missing.
persistent/recurrent: elevated in both 2011 and 2015, with 2013 being elevated, non-elevated, or missing.
Participants were excluded if depressive symptom data were unavailable at either the 2011 or 2015 anchor waves, or if their observed symptom patterns did not conform to any of the four predefined groups (n = 839).
Cognitive outcomes
The primary outcome was episodic memory in 2018, assessed using immediate and delayed recall of a 10-word list [26]. The score was calculated as the sum of immediate and delayed recall, ranging from 0 to 20, with higher scores indicating better memory. The secondary outcome was orientation in 2018, assessed using items on date, month, year, and day of the week, with scores ranging from 0 to 4 and higher scores indicating better orientation. Baseline episodic memory and baseline orientation were measured using the corresponding assessments in the 2011 wave.
Grip strength and sleep disturbance
Grip strength was measured using a handheld dynamometer in 2011 (baseline) and 2015 [17]. The sum of the maximum values from the left and right hands was used for analyses. Sleep disturbance was derived from the CES-D-10 restlessness item, which was scored from 1 to 4, at both the 2011 (baseline) and 2015 waves, with higher scores indicating more frequent restlessness symptoms [15].
Covariates
Covariates were selected a priori based on their potential associations with depressive symptoms and cognitive outcomes. These included age, sex, educational attainment, marital status, urban residence, BMI, chronic disease count, baseline cognitive score in the corresponding domain, and baseline grip strength.
In the analyses, age, BMI, baseline cognitive scores, and baseline grip strength were treated as continuous variables. Categorical covariates were defined as follows: sex (male or female), educational attainment (illiterate/primary vs. secondary or above), marital status (married vs. not married), and urban residence (urban vs. rural, based on baseline household registration). Finally, the chronic disease count was calculated based on self-reported physician diagnoses of seven conditions at baseline: hypertension, diabetes, heart disease, stroke, chronic lung disease, arthritis, and cancer.
Statistical analysis
For the main analyses, we used linear regression models to examine associations between depressive symptom patterns and episodic memory in 2018, with the never depressed group as the reference category. Models were adjusted for the prespecified covariates. Parallel models were fit for orientation in 2018, using baseline orientation as the corresponding cognitive measure while adjusting for the same covariates. Unstandardized regression coefficients with 95% confidence intervals were reported. This baseline-adjusted follow-up modeling approach was chosen to estimate between-group differences in cognitive outcomes in 2018 while accounting for baseline cognitive performance in the same domain.
We performed subgroup analyses for episodic memory, stratified by sex, age, educational attainment, baseline episodic memory, and chronic disease status. Interaction terms were used to assess heterogeneity, and P values for interaction were reported.
To assess the robustness of our findings, we conducted several sensitivity analyses. First, to handle missing data, we used multiple imputation (MI) and inverse probability weighting (IPW), with MI estimates pooled across 20 imputed datasets using Rubin’s rules [27]. Second, we employed ordered logistic regression for orientation to account for its restricted score range, and median regression for episodic memory to assess robustness against potential distributional skewness. Finally, we formally assessed collinearity among covariates using variance inflation factors and condition indices.
We examined directional associations by comparing the forward association (from depressive symptom patterns to later episodic memory) assessed via linear regression, with the reverse association (from baseline episodic memory [2011] to subsequent depressive symptom pattern membership [2011–2015]) evaluated using multinomial logistic regression.
Exploratory pathway analysis
We conducted an exploratory pathway analysis (n = 5,721) to examine whether grip strength and sleep disturbance might partly account for the association between the persistent/recurrent depressive symptom pattern and episodic memory. The smaller sample size reflected the requirement for complete data on episodic memory, depressive symptom pattern, 2015 grip strength, 2015 sleep disturbance, and all covariates. Because grip strength and sleep disturbance were measured in the same wave (2015) as the final depressive symptom assessment, temporal precedence between the exposure and the pathway variables cannot be established. Standardized coefficients were estimated using generalized structural equation modeling for paths linking depressive symptom patterns, grip strength, sleep disturbance, and episodic memory. The pathway models were adjusted for age, sex, educational attainment, marital status, urban residence, body mass index, chronic disease count, baseline episodic memory, baseline grip strength, and baseline sleep disturbance. Indirect effects were calculated as the product of the corresponding path coefficients, and bootstrap percentile 95% confidence intervals were obtained using 5,000 replications. To assess the potential influence of missing data and selection into the pathway analytic sample, we conducted IPW and MI sensitivity analyses specifically for these models; these results are presented in Supplementary Table S4.
Additional analyses
We conducted several additional analyses. First, we performed targeted sensitivity analyses for the association between the persistent/recurrent depressive symptom pattern and episodic memory, including exclusion of participants with baseline mild cognitive impairment, exclusion of participants with instrumental activities of daily living (IADL) limitations, use of an alternative CES-D-10 cutoff of ≥ 12, and IPW. Second, we externally validated the association between depressive symptom patterns and episodic memory in the HRS. The validation sample included 13,339 participants aged 50 years or older with CES-D-8 data available across the 2010–2014 period and cognitive follow-up in 2016. Depressive symptom patterns were constructed using the same rule-based approach, with a CES-D-8 cutoff of ≥ 4 (scored 0–8). Directional consistency with the CHARLS findings was assessed by comparing the sign and statistical significance of regression coefficients for the association between depressive symptom patterns and episodic memory. Finally, we conducted exploratory analyses comparing C-reactive protein (CRP) levels measured in blood samples collected in 2015 across depressive symptom pattern groups. CRP values > 10 mg/L were excluded to reduce the influence of acute infection. Log transformation was applied prior to analysis. Differences across groups were assessed using the Kruskal–Wallis test for untransformed CRP and one-way ANOVA for log-transformed CRP. Multicollinearity diagnostics showed no concerning evidence of collinearity (Supplementary Table S8).
Software
Statistical analyses were conducted using Stata 18.0. Figures and tables were prepared using R version 4.5.3.
Results
Participant characteristics
The main analytic sample included 7,211 participants. According to the prespecified rule-based classification, 3,396 participants were categorized as never depressed, 1,194 as early-remission, 1,067 as late-onset, and 1,554 as persistent/recurrent. Baseline characteristics across the four depressive symptom pattern groups are shown in Table 1, and mean CES-D-10 scores across waves are shown in Fig. 1. The complete-case sample was reduced to 5,743 for episodic memory regression and 6,272 for orientation analyses because of missing outcome or covariate data. The pathway analytic sample was further reduced to 5,721 because complete data were required for depressive symptom pattern, grip strength, sleep disturbance, episodic memory, and covariates.
Table 1.
Baseline characteristics according to depressive symptom pattern groups
| Characteristic | Overall (N = 7,211) |
Never depressed (n = 3,396) |
Early remission (n = 1,194) |
Late onset (n = 1,067) |
Persistent/recurrent (n = 1,554) |
|---|---|---|---|---|---|
| Age, years | 57.2 (56.9–57.5) | 56.8 (56.5–57.1) | 57.8 (57.3–58.3) | 56.7 (56.2–57.2) | 57.9 (57.5–58.3) |
| BMI, kg/m² | 23.7 (23.6–23.8) | 24.0 (23.9–24.2) | 23.4 (23.1–23.7) | 23.8 (23.5–24.1) | 23.2 (22.9–23.5) |
| Chronic disease count | 0.6 (0.58–0.62) | 0.4 (0.38–0.42) | 0.7 (0.65–0.75) | 0.6 (0.55–0.65) | 0.9 (0.85–0.95) |
| Baseline episodic memory | 7.5 (7.4–7.6) | 8.2 (8.1–8.3) | 6.8 (6.6-7.0) | 7.4 (7.2–7.6) | 6.5 (6.3–6.7) |
| Baseline orientation | 3.0 (2.97–3.03) | 3.2 (3.17–3.23) | 2.8 (2.73–2.87) | 2.9 (2.83–2.97) | 2.6 (2.53–2.67) |
| Baseline grip strength, kg | 32.7 (32.4–33.0) | 35.5 (35.1–35.9) | 30.8 (30.2–31.4) | 31.8 (31.2–32.4) | 28.6 (28.1–29.1) |
| Baseline sleep disturbance | 2.0 (1.97–2.03) | 1.6 (1.55–1.65) | 2.6 (2.51–2.69) | 1.8 (1.71–1.89) | 2.8 (2.71–2.89) |
| Female sex, % | 52.6 (51.4–53.8) | 42.6 (41.0-44.2) | 56.3 (53.5–59.1) | 59.4 (56.5–62.3) | 67.1 (64.7–69.5) |
| Educational attainment, % | |||||
| Illiterate/primary | 88.5 (87.7–89.3) | 83.9 (82.6–85.2) | 90.8 (89.0-92.6) | 91.5 (89.7–93.3) | 94.9 (93.7–96.1) |
| Secondary or above | 11.5 (10.7–12.3) | 16.1 (14.9–17.3) | 9.2 (7.6–10.8) | 8.5 (6.9–10.1) | 5.1 (4.0-6.2) |
| Married, % | 86.4 (85.6–87.2) | 90.3 (89.3–91.3) | 81.9 (79.6–84.2) | 85.8 (83.6–88.0) | 81.7 (79.7–83.7) |
| Urban residence, % | 35.1 (34.0-36.2) | 41.5 (39.9–43.1) | 32.8 (30.1–35.5) | 31.6 (28.8–34.4) | 25.3 (23.2–27.4) |
Data are presented as mean (95% CI) or percentage (95% CI). Confidence intervals were calculated using normal approximation for continuous variables and Wilson score interval for proportions. Chronic disease count was based on baseline self-reported physician diagnosis of hypertension, diabetes, heart disease, stroke, chronic lung disease, arthritis, and cancer. Educational attainment was categorized as illiterate/primary versus secondary or above. BMI values exceeding 50 kg/m² were excluded as implausible. Sleep disturbance was derived from the CES-D-10 sleep/restlessness item (scored 1–4), with higher scores indicating more frequent symptoms. BMI, body mass index; CES-D-10, 10-item Center for Epidemiologic Studies Depression Scale
Fig. 1.
Mean CES-D-10 scores across waves by depressive symptom pattern group. Mean CES-D-10 scores are shown for the four prespecified depressive symptom pattern groups across 2011, 2013, and 2015 in CHARLS. The dashed horizontal line indicates the cutoff for elevated depressive symptoms (CES-D-10 ≥ 10). CES-D-10 scores range from 0 to 30. CES-D-10, 10-item Center for Epidemiologic Studies Depression Scale; CHARLS, China Health and Retirement Longitudinal Study
Compared with participants in the never depressed group, those in the persistent/recurrent group were older, had lower BMI, poorer baseline episodic memory and orientation, lower grip strength, and higher sleep disturbance scores. They were more likely to be female, less likely to have secondary education or above, less likely to be married, and less likely to live in urban areas (all P < 0.001).
Associations between depressive symptom patterns and episodic memory
In complete-case regression analyses for episodic memory (n = 5,743), the late-onset and persistent/recurrent groups were associated with lower episodic memory scores in 2018 compared with the never depressed group, whereas the association for the early-remission pattern was weaker (Fig. 2; Supplementary Table S1). In the fully adjusted model, the unstandardized coefficients were − 0.361 (95% CI, − 0.637 to − 0.086) for early-remission, − 0.932 (95% CI, − 1.227 to − 0.637) for late-onset, and − 0.842 (95% CI, − 1.104 to − 0.580) for persistent/recurrent.
Fig. 2.
Associations between depressive symptom patterns and cognitive outcomes in 2018. Adjusted associations between depressive symptom patterns and cognitive outcomes in 2018 are shown. A Episodic memory (N = 5,743). B Orientation (N = 6,272). Estimates are from linear regression models with the never depressed group as the reference category. The estimates shown are unstandardized coefficients. Error bars indicate 95% CI. CI, confidence interval
Missing-data sensitivity analyses showed broadly consistent findings. Inverse probability weighting yielded an estimate similar in direction and magnitude to the complete-case analysis for the persistent/recurrent versus never depressed comparison (β = −0.938, 95% CI, − 1.162 to − 0.714), whereas the multiple imputation estimate was larger in magnitude (β = −1.633, 95% CI, − 1.823 to − 1.442) (Supplementary Table S1). Additional sensitivity analyses for persistent/recurrent versus never depressed showed consistent findings after excluding participants with baseline mild cognitive impairment (β = −0.975, 95% CI, − 1.277 to − 0.673), excluding those with instrumental activities of daily living limitations (β = −0.689, 95% CI, − 0.986 to − 0.393), and using an alternative CES-D-10 cutoff of 12 or higher (β = −0.963, 95% CI, − 1.325 to − 0.601) (Supplementary Table S1).
Associations with orientation
In complete-case analyses for orientation in 2018 (n = 6,272), the late-onset and persistent/recurrent groups were also associated with lower orientation scores compared with the never depressed group, whereas the early-remission pattern was not associated with orientation after adjustment. The adjusted unstandardized coefficients were − 0.026 (95% CI, − 0.097 to 0.045) for early-remission, − 0.135 (95% CI, − 0.209 to − 0.060) for late-onset, and − 0.222 (95% CI, − 0.295 to − 0.150) for persistent/recurrent (Fig. 2; Supplementary Table S2). Alternative-model robustness analyses using ordered logistic regression for orientation and median regression for episodic memory yielded directionally consistent findings (Supplementary Table S1, Panel D).
Subgroup analyses
In subgroup analyses for episodic memory, the association for the persistent/recurrent group was evident across most strata, although its strength varied (Fig. 3; Supplementary Table S3). There was evidence of interaction by sex (P for interaction = 0.020) and educational attainment (P for interaction = 0.001). The association appeared stronger among women than men and was attenuated in the higher education subgroup, although estimates in some subgroups were imprecise. There was no clear evidence of interaction by age, baseline episodic memory, or chronic disease status.
Fig. 3.
Subgroup analyses for the association between depressive symptom patterns and episodic memory. Subgroup analyses were conducted for episodic memory according to sex, age, educational attainment, baseline episodic memory, and chronic disease status. Chronic disease count included hypertension, diabetes, heart disease, stroke, chronic lung disease, arthritis, and cancer. Estimates are shown for the persistent/recurrent and late-onset depressive symptom patterns, with never depressed as the reference group. The estimates shown are unstandardized coefficients. Error bars indicate 95% CI. CI, confidence interval; BMI, body mass index
Directional analyses
To explore directionality, we compared the association between depressive symptom patterns and later episodic memory with the reverse association between baseline episodic memory and subsequent depressive symptom patterns (Fig. 4). In the forward model, the persistent/recurrent depressive symptom pattern was associated with lower subsequent episodic memory (standardized β = −0.195, 95% CI, − 0.255 to − 0.134). In the reverse model, higher baseline episodic memory was associated with lower odds of subsequent membership in the persistent/recurrent group (log-odds coefficient = − 0.303, 95% CI, − 0.374 to − 0.233).
Fig. 4.
Directional analyses comparing forward and reverse associations between depressive symptom patterns and episodic memory. The forward model examined the association between depressive symptom patterns and subsequent episodic memory. The reverse model examined the association between baseline episodic memory and subsequent depressive symptom patterns. Estimates are shown for the persistent/recurrent depressive symptom pattern. The forward association is shown as a standardized coefficient, and the reverse association is shown as a log-odds coefficient. Error bars indicate 95% CI. CI, confidence interval
External validation in HRS
External validation was conducted in 13,339 participants from the HRS (Supplementary Figure S5; Supplementary Tables S5 and S6). The association patterns were directionally consistent with those observed in CHARLS, although effect sizes were smaller in HRS. Compared with the never depressed group, persistent/recurrent and late-onset depressive symptom patterns were associated with lower episodic memory scores in HRS, whereas the early-remission pattern was not significantly associated with episodic memory. The adjusted coefficients were − 0.467 (SE 0.073, P < 0.001) for persistent/recurrent, − 0.430 (SE 0.121, P < 0.001) for late-onset, and − 0.048 (SE 0.122, P = 0.694) for early-remission.
Exploratory pathway analysis
Standardized coefficients from the exploratory pathway analysis are shown in Table 2. This analysis was conducted in the pathway analytic sample (n = 5,721) with complete data on depressive symptom patterns, grip strength, sleep disturbance, episodic memory, and covariates. Participant characteristics for those included in and excluded from this sample, together with IPW and MI sensitivity analyses, are presented in Supplementary Table S4.
Table 2.
Exploratory pathway analysis of the association between the persistent/recurrent depressive symptom pattern and episodic memory
| Path / Effect | Coefficient | SE | 95% CI |
|---|---|---|---|
| Path a1: Depressive symptom patterns → Grip strength | −0.118 | 0.024 | −0.165, − 0.071 |
| Path b1: Grip strength → Episodic memory | 0.045 | 0.016 | 0.014, 0.077 |
| Indirect effect via grip strength | −0.005 | 0.002a | −0.010, − 0.001b |
| Path a2: Depressive symptom patterns → Sleep disturbance | 0.616 | 0.033 | 0.551, 0.681 |
| Path b2: Sleep disturbance → Episodic memory | −0.026 | 0.013 | −0.050, − 0.001 |
| Indirect effect via sleep disturbance | −0.016 | 0.008a | −0.032, − 0.001b |
| Direct effect (c′) | −0.100 | 0.030 | −0.159, − 0.041 |
| Total indirect effect | −0.021 | 0.008a | −0.038, − 0.006b |
| Total effect (c) | −0.121 | 0.028 | −0.181, − 0.071 |
All coefficients are fully standardized. The exploratory pathway analysis was conducted among participants with complete data on depressive symptom patterns, grip strength (2015), sleep disturbance (2015), episodic memory (2018), and all covariates (n = 5,721). This sample was smaller than the main analytic sample (n = 7,211) owing to additional requirements for complete data on the pathway variables. Because grip strength and sleep disturbance were measured in the same wave (2015) as the final depressive symptom assessment, temporal precedence between the exposure and these pathway variables cannot be established. Models were adjusted for age, sex, educational attainment, marital status, urban residence, body mass index, chronic disease count, baseline episodic memory, baseline grip strength, and baseline sleep disturbance. Indirect effects were calculated as the product of the a and b paths. The estimated indirect contribution of grip strength and sleep disturbance corresponded to 16.9% of the total effect. Bootstrap standard errors (based on 5,000 replications) are denoted bya, and bootstrap percentile 95% confidence intervals are denoted byb
The persistent/recurrent depressive symptom pattern was associated with lower grip strength (path a1: β = −0.118, 95% CI, − 0.165 to − 0.071) and greater sleep disturbance (path a2: β = 0.616, 95% CI, 0.551 to 0.681). In turn, lower grip strength (path b1: β = 0.045, 95% CI, 0.014 to 0.077) and greater sleep disturbance (path b2: β = −0.026, 95% CI, − 0.050 to − 0.001) were associated with lower episodic memory.
The indirect effect via grip strength was − 0.005 (bootstrap 95% CI, − 0.010 to − 0.001), and the indirect effect via sleep disturbance was − 0.016 (bootstrap 95% CI, − 0.032 to − 0.001). The total indirect effect was − 0.021 (bootstrap 95% CI, − 0.038 to − 0.006), corresponding to an estimated indirect contribution of 16.9% of the total effect. The direct effect remained negative after accounting for these factors (β = −0.100, 95% CI, − 0.159 to − 0.041). Sensitivity analyses for the exploratory pathway analysis showed that the overall pattern of path coefficients and indirect effects was broadly consistent in both IPW and MI analyses, although the sleep disturbance-to-memory path was attenuated (P = 0.081) in the MI analysis (Supplementary Table S4).
Exploratory analyses of CRP
In exploratory analyses of CRP measured in 2015, mean CRP levels were 2.46 mg/L in the never depressed group, 2.65 mg/L in the early-remission group, 2.63 mg/L in the late-onset group, and 2.81 mg/L in the persistent/recurrent group. Median CRP levels were 1.40, 1.30, 1.50, and 1.40 mg/L, respectively, and log-transformed CRP values were 0.35, 0.33, 0.40, and 0.37, respectively (Supplementary Table S7). Overall, CRP distributions were broadly similar across groups.
Discussion
This longitudinal study showed that distinct depressive symptom patterns were associated with later cognitive outcomes. Relative to the never depressed group, the late-onset and persistent/recurrent depressive symptom patterns were consistently associated with poorer episodic memory and orientation. In contrast, the early-remission pattern appeared to have smaller long-term cognitive implications. Subgroup analyses further indicated that these associations were stronger among women and participants with lower educational attainment. In exploratory analyses, lower grip strength and greater sleep disturbance may partly explain the association between the persistent/recurrent depressive symptom pattern and episodic memory. Directional analyses suggested potential bidirectionality between depressive symptoms and cognitive function, and the episodic memory findings were directionally consistent in the independent HRS cohort.
The stronger association observed for the persistent/recurrent depressive symptom pattern is consistent with previous studies suggesting that longer-term or recurrent depressive symptoms may be particularly detrimental to cognitive health in later life [28]. Prior longitudinal research has linked persistent/recurrent depressive symptoms to faster cognitive decline, a higher risk of incident cognitive impairment, and increased dementia risk. Although the absolute differences on the 0–20 episodic memory scale and 0–4 orientation scale were modest, they were consistent across sensitivity analyses and may still be relevant at the population level. Proposed mechanisms include prolonged activation of stress-related neuroendocrine pathways, increased vascular and behavioral risk, lower physical reserve, and reduced engagement in cognitively or socially stimulating activities [29, 30]. Because our data did not directly test these mechanisms, we interpret the persistent/recurrent pattern as a marker of elevated cognitive risk rather than as evidence of a specific causal pathway.
The late-onset pattern also warrants attention. In our analyses, this pattern was associated with lower episodic memory and poorer orientation. One possible explanation is that depressive symptoms emerging later in life may, in some individuals, reflect early neurodegenerative or cerebrovascular changes rather than only the long-term effects of depression [31]. This interpretation is speculative, but it is consistent with the possibility that late-onset symptoms may sometimes represent an early manifestation of cognitive decline rather than an entirely independent risk factor [32]. By contrast, the early-remission pattern showed only modest associations with episodic memory and no significant association with orientation after adjustment. These findings suggest that depressive symptoms that remit over time may have weaker long-term associations with cognitive function than symptoms that persist, recur, or emerge later. Previous studies have reported mixed findings on whether remission fully attenuates the depression–cognition association; our results are more consistent with weaker, rather than absent, long-term cognitive relevance [33].
These subgroup patterns suggest potential effect modification. The sex-related heterogeneity is consistent with prior research indicating that biological and psychosocial factors may contribute to heterogeneity in the depression–cognition association [19, 34]. For educational attainment, our findings are in line with the cognitive reserve hypothesis, which proposes that higher education may buffer against neuropathological insults through greater neural compensation and broader socioeconomic resources [20, 35]. In this context, long-term depressive symptoms may have stronger associations with poorer cognitive outcomes among individuals with lower cognitive reserve. These findings support the value of closer cognitive monitoring and depressive symptom management in subgroups at elevated risk.
The directional analyses provide additional context but do not establish causality. Because the forward and reverse models were estimated on different scales, their magnitudes should not be directly compared. The persistent/recurrent depressive symptom pattern was associated with lower subsequent episodic memory, while lower baseline episodic memory was associated with higher odds of subsequent membership in the persistent/recurrent group. Taken together, these findings are consistent with a possible bidirectional relationship over time [10].
The HRS analyses broadly supported the pattern observed in CHARLS, although effect sizes were smaller. This difference may reflect variation between the two cohorts in population characteristics, educational attainment, and depression measurement (CES-D-8 in HRS versus CES-D-10 in CHARLS). The HRS analysis should therefore be viewed as external validation of directional consistency rather than exact replication. Overall, the similar pattern across cohorts increases confidence that the observed associations are not entirely cohort-specific.
The exploratory pathway analysis suggested that lower grip strength and greater sleep disturbance may partly explain the association between the persistent/recurrent depressive symptom pattern and episodic memory. These findings are consistent with prior research linking reduced grip strength to frailty and poorer cognitive outcomes, and sleep disturbance to both depressive symptoms and memory decline [16, 36]. However, these factors accounted for only a modest proportion of the observed association. Sleep disturbance may be particularly relevant from a public health perspective because it is common in later life, associated with both depressive symptoms and cognitive outcomes, and potentially modifiable. By contrast, grip strength is better viewed as a marker of physical reserve and overall health status than as a direct intervention target.
Because grip strength and sleep disturbance were measured in the same wave as the final depressive symptom assessment, temporal ordering between the exposure and these potential pathway variables cannot be established. The pathway analysis was restricted to participants with complete data, which may have introduced selection bias. Additional IPW and MI analyses yielded broadly consistent results, although the sleep-disturbance-to-memory path was attenuated in the MI analysis. Accordingly, these analyses were exploratory and should not be interpreted as evidence of causal mediation. Other unmeasured factors may also contribute to the observed association, including vascular disease, inflammation, health behaviours, medication use, and social factors [37]. In exploratory CRP analyses, no clear differences were observed across depressive symptom pattern groups.
The handling of missing data also merits consideration. The complete-case and inverse probability weighting analyses yielded similar estimates, whereas the multiple imputation estimate was larger in magnitude. This pattern likely reflects the different assumptions and information structures underlying the two approaches. IPW reweights the observed sample but remains dependent on participants who remained in the study, whereas MI uses observed data to impute missing values and therefore incorporates information from participants with incomplete follow-up [38]. In this cohort, participants retained in the analysis were younger, better educated, and cognitively healthier than those lost to follow-up. Accordingly, the complete-case and IPW estimates may be conservative, whereas the MI estimate may reflect a less selective analytic sample. Importantly, the direction and statistical significance of the association were consistent across all three approaches.
Several features strengthen this study. A key strength was the repeated assessment of depressive symptoms across three waves. This allowed classification into four prespecified depressive symptom patterns using a rule-based approach, which went beyond a simple depressed-versus-non-depressed comparison. We used a large, nationally representative longitudinal cohort with prospective cognitive assessments across two cognitive domains. Furthermore, we conducted extensive sensitivity analyses addressing missing data and sample restrictions, and assessed directional consistency in an independent cohort.
Several limitations should be considered. First, depressive symptoms were assessed using the CES-D-10 rather than clinical diagnostic interviews; therefore, the resulting groups reflect depressive symptom patterns rather than clinically diagnosed depression [24]. In addition, depressive symptoms and sleep disturbance were self-reported and may be subject to recall error and social desirability bias [39]. Second, although the rule-based classification improves interpretability and preserves statistical power, it may still have introduced residual within-group heterogeneity. Third, residual confounding cannot be fully excluded despite extensive covariate adjustment. Finally, the pathway and CRP analyses were exploratory, and selective attrition and missing data may have influenced the findings, although sensitivity analyses were largely consistent.
In conclusion, long-term depressive symptom patterns were associated with poorer cognitive performance in this nationally representative cohort, with the strongest and most consistent associations observed for the persistent/recurrent and late-onset patterns. Lower grip strength and greater sleep disturbance may partly explain the association observed for the persistent/recurrent depressive symptom pattern. However, because these factors were measured concurrently with the final depressive symptom assessment, these analyses were exploratory and should not be interpreted as evidence of causal mediation. These findings underscore the importance of considering the longitudinal course of depressive symptoms, rather than relying on single time-point measures alone, when identifying older adults at elevated risk of cognitive decline.
Supplementary Information
Acknowledgements
We thank the CHARLS and the HRS research team and all participants for their contribution.
Abbreviations
- CES-D-10
10-item Center for Epidemiologic Studies Depression Scale
- CHARLS
China Health and Retirement Longitudinal Study
- HRS
Health and Retirement Study
- BMI
Body mass index
- CRP
C-reactive protein
- IADL
Instrumental activities of daily living
- MCI
Mild cognitive impairment
- MI
Multiple imputation
- IPW
Inverse probability weighting
- SD
Standard deviation
- SE
Standard error
- CI
Confidence interval
Authors’ contributions
CCC conceptualized the study, developed the methodology, and wrote the original draft. NZ performed data curation, formal analysis, and software implementation. XWW contributed to visualization, validation, and writing the original draft. YQ was responsible for investigation and provided resources. XBW supervised the study, managed project administration, acquired funding, and reviewed and edited the manuscript. All authors read and approved the final manuscript.
Funding
This work was supported by the Natural Science Foundation of Sichuan Province (Grant No. 2024NSFSC2096) and the Joint Project of Luzhou City and Southwest Medical University (Grant No. 2025LZXNYDYG05).
Data availability
The datasets generated and/or analysed during the current study are available in the CHARLS repository, (http:/charls.pku.edu.cn).
Declarations
Ethics approval and consent to participate
This study is a secondary analysis of the CHARLS and the HRS. The original CHARLS study was approved by the Biomedical Ethics Review Committee of Peking University (approval number: IRB00001052-11015). The study was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki. All participants provided written informed consent.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Cheng-Chuan Chen, Na Zhou and Xiao-Wen Wu contributed equally to this work.
References
- 1.Livingston G, Huntley J, Sommerlad A, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet. 2020;396(10248):413–46. 10.1016/S0140-6736(20)30367-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.World Health Organization. Risk reduction of cognitive decline and dementia: WHO guidelines. Geneva: World Health Organization; 2019. [PubMed] [Google Scholar]
- 3.Hu T, Zhao X, Wu M, et al. Prevalence of depression in older adults in China: A systematic review and meta-analysis. Psychiatry Res. 2022;311:114511. 10.1016/j.psychres.2022.114511. [DOI] [PubMed] [Google Scholar]
- 4.Zis P, Daskalaki A, Bountouni I, Sykioti P, Varrassi G, Paladini A. Depression and chronic pain in the elderly: links and management challenges. Clin Interv Aging. 2017;12:709–20. 10.2147/CIA.S113576. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Cantón-Habas V, Rich-Ruiz M, Romero-Saldaña M, Carrasco-High MJ. Depression as a Risk Factor for Dementia and Alzheimer’s Disease. Biomedicines. 2020;8(11):457. 10.3390/biomedicines8110457. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Dafsari FS, Jesskoen T. Depression and risk of Alzheimer disease: A systematic review and meta-analysis of recent evidence. Psychiatry Res. 2020;291:113204.32535512 [Google Scholar]
- 7.Kaup AR, Byers AL, Falvey C, et al. Trajectories of Depressive Symptoms in Older Adults and Risk of Dementia. JAMA Psychiatry. 2016;73(5):525–31. 10.1001/jamapsychiatry.2016.0004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Mirza SS, Wolters FJ, Swanson SA, et al. 10-year trajectories of depressive symptoms and risk of dementia: a population-based study. Lancet Psychiatry. 2016;3(7):628–35. [DOI] [PubMed] [Google Scholar]
- 9.Wu M, Zhao X, Chen Y, et al. Trajectories of depressive symptoms and subsequent cognitive decline: A longitudinal study of Chinese middle-aged and older adults. J Affect Disord. 2020;266:577–83. 10.1016/j.jad.2020.01.107. [Google Scholar]
- 10.Pan A, Keum N, Okereke OI, et al. Bidirectional association between depression and cognitive decline in middle-aged and older Chinese adults. J Affect Disord. 2018;227:14–20. 10.1016/j.jad.2017.10.010. [Google Scholar]
- 11.Murman DL. The Impact of Age on Cognition. Semin Hear. 2015;36(3):111–21. 10.1055/s-0035-1555115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Beurel E, Toups M, Nemeroff CB. The Bidirectional Relationship of Depression and Inflammation: Double Trouble. Neuron. 2020;107(2):234–56. 10.1016/j.neuron.2020.06.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Pariante CM. Why are depressed patients inflamed? A reflection on 20 years of research on depression, glucocorticoid receptor resistance and inflammation. Eur Neuropsychopharmacol. 2017;27(6):554–9. [DOI] [PubMed] [Google Scholar]
- 14.Shi L, Chen SJ, Ma MY, et al. Sleep disturbances increase the risk of dementia: A systematic review and meta-analysis. Sleep Med Rev. 2018;40:4–16. 10.1016/j.smrv.2017.06.010. [DOI] [PubMed] [Google Scholar]
- 15.Wang X, Ding X, Su S, et al. Sleep duration, sleep disturbance and cognitive decline in older Chinese adults. J Psychiatr Res. 2020;125:46–52. 10.1016/j.jpsychires.2020.03.004. [Google Scholar]
- 16.Yaffe K, Falvey CM, Hoang T. Connections between sleep and cognition in older adults. Lancet Neurol. 2014;13(10):1017–28. [DOI] [PubMed] [Google Scholar]
- 17.Zammit AR, Piccinin AM, Duggan EC, et al. Associations between handgrip strength and cognitive function in older adults: A systematic review and meta-analysis. J Am Med Dir Assoc. 2019;20(11):1454–60. 10.1016/j.jamda.2019.04.011. [Google Scholar]
- 18.Eid RS, Gobinath AR, Galea LA. Sex differences in depression: Insights from clinical and preclinical studies. Prog Neurobiol. 2019;176:86–102. 10.1016/j.pneurobio.2019.01.006. [DOI] [PubMed] [Google Scholar]
- 19.Vlassoff C. Gender Differences in Determinants and Consequences of Health and Illness. J Health Popul Nutr. 2007;25(1):47–61. [PMC free article] [PubMed] [Google Scholar]
- 20.Stern Y, Arenaza-Urquijo EM, Bartrés-Faz D, et al. Whitepaper: Defining and investigating cognitive reserve, brain reserve, and brain maintenance. Alzheimers Dement. 2020;16(9):1305–11. 10.1002/alz.12146. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Meng X, D’Arcy C. Education and dementia in the context of the cognitive reserve hypothesis: a systematic review with meta-analyses and qualitative analyses. PLoS ONE. 2012;7(6):e38268. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Zhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS). Int J Epidemiol. 2014;43(1):61–8. 10.1093/ije/dys203. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Sonnega A, Faul JD, Ofstedal MB, Langa KM, Phillips JW, Weir DR. Cohort Profile: the Health and Retirement Study (HRS). Int J Epidemiol. 2014;43(2):576–85. 10.1093/ije/dyu067. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Andresen EM, Malmgren JA, Carter WB, Patrick DL. Screening for depression in well older adults: evaluation of a short form of the CES-D. Am J Prev Med. 1994;10(2):77–84. [PubMed] [Google Scholar]
- 25.Boey KW. Cross-validation of a short form of the CES-D in Chinese elderly. Int J Geriatr Psychiatry. 1999;14(8):608–17. [DOI] [PubMed] [Google Scholar]
- 26.Lei X, Smith JP, Sun Y, Zhao Y. Gender differences in cognition in China and reasons for change over time: evidence from CHARLS. J Econ Ageing. 2014;4:46–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Rubin DB. Multiple Imputation for Nonresponse in Surveys. New York: Wiley; 1987. [Google Scholar]
- 28.Liang X, Zhao Z, Jiang R, et al. Distinct trajectories of depressive symptoms and their association with subsequent cognitive decline: A longitudinal study based on CHARLS. J Affect Disord. 2023;329:660–6. 10.1016/j.jad.2023.02.138. [Google Scholar]
- 29.Taylor WD, Aizenstein HJ, Alexopoulos GS. The vascular depression hypothesis: mechanisms linking vascular disease with depression. Mol Psychiatry. 2013;18(9):963–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Eyre HA, Lavretsky H, Forbes M, et al. Convergence of mechanistic pathways in vascular depression and Alzheimer’s disease: A systematic review. Psychiatry Res. 2019;281:112521. [Google Scholar]
- 31.Ma L. Depression in frail older adults: A review. Clin Interv Aging. 2020;15:1069–77. [Google Scholar]
- 32.Singh-Manoux A, Dugravot A, Fournier A, et al. Trajectories of depressive symptoms before diagnosis of dementia: a 28-year follow-up study. JAMA Psychiatry. 2017;74(7):712–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Diniz BS, Butters MA, Albert SM, Dew MA, Reynolds CF 3rd. Late-life depression and risk of vascular dementia and Alzheimer’s disease: systematic review and meta-analysis of community-based cohort studies. Br J Psychiatry. 2013;202(5):329–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Penninx BWJH, Milaneschi Y, Lamers F, Vogelzangs N. Understanding the somatic consequences of depression: biological mechanisms and the role of depression symptom profile. BMC Med. 2013;11:129. 10.1186/1741-7015-11-129. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Stern Y. Cognitive reserve in ageing and Alzheimer's disease. Lancet Neurol. 2012;11(11):1006–12. 10.1016/S1474-4422(12)70191-6. [DOI] [PMC free article] [PubMed]
- 36.Kobayashi-Cuya KE, Sakurai R, Suzuki H, et al. Observational Evidence of the Association Between Handgrip Strength, Hand Dexterity, and Cognitive Performance in Community-Dwelling Older Adults: A Systematic Review. J Epidemiol. 2018;28(9):373–81. 10.2188/jea.JE20170041. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Walker KA, Ficek BN, Westbrook R. Understanding the Role of Systemic Inflammation in Alzheimer’s Disease. ACS Chem Neurosci. 2019;10(8):3340–2. 10.1021/acschemneuro.9b00333. [DOI] [PubMed] [Google Scholar]
- 38.Sterne JAC, White IR, Carlin JB, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ. 2009;338:b2393. 10.1136/bmj.b2393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Radloff LS, The CES-D, Scale. A Self-Report Depression Scale for Research in the General Population. Appl Psychol Meas. 1977;1(3):385–401. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated and/or analysed during the current study are available in the CHARLS repository, (http:/charls.pku.edu.cn).