INTRODUCTION
Depression is one of the most common geriatric psychiatric disorders [1]. Previous studies have shown that depression is a potential prodromal feature or a risk factor for dementia, including Alzheimer’s disease (AD) and vascular dementia (VaD) [2-4]. Although findings on the directionality of the causal relationship are mixed, making the nature of the association unclear, it is well known that dementia and depression share similar clinical symptoms, such as functional decline and mood changes [5,6]. Several neurobiological mechanisms that contribute to both dementia and depression could explain such linkage between these two disorders. These include vascular disease, elevated levels of glucocorticoid steroid, hippocampal atrophy, increased deposition of amyloid-β plaques, and chronic inflammation [7,8]. Since the underlying pathways associated with dementia and depression may not be mutually exclusive, interventions targeting both conditions may be beneficial in preventing and treating these disorders.
Recent studies have demonstrated that physical activity (PA) could target the neurobiological mechanisms that underlie depression and dementia, and therefore, may be an ideal treatment for both [9,10]. PA is known to be effective in alleviating mood symptoms in depression [11,12], and higher levels of PA are associated with a reduced risk of dementia [13]. In examining the effect of PA on dementia symptoms among clinically depressed populations, PA was associated with improved cognitive functioning, including memory, executive functioning, verbal fluency, and psychomotor speed [14,15]. Given that those with depression may be at a heightened risk of dementia, PA could play a promising role as a non-pharmaceutical treatment in preventing dementia, as well as enhancing mental health of the depressed population.
Although evidence from aforementioned studies suggests PA could be a useful tool in improving cognitive functioning among depressed subjects, the evidence is still limited and research literature focusing on whether PA could reduce the risk of dementia among the depressed subjects is sparse. Also, to our knowledge, only one recent study analyzed effects of changes in the status of PA on cognitive decline in older adults [16]. According to this study, older adults who regularly engaged in PA over 4 years were at a reduced risk of cognitive decline compared to those who maintained sedentary lifestyle or those who stopped engaging in PA at certain times. To address gap in literature, the current study sought to investigate the association between PA and dementia among the depressed subjects. Furthermore, we aimed to assess whether changes in PA status before and after the diagnosis of depression affect the risk of dementia.
METHODS
Data source
This study used the data from National Health Insurance Sharing Service (NHISS) of the National Health Insurance Service (NHIS) of South Korea. Managed by the South Korean government, NHIS is a single insurer that operates compulsory medical insurance, covering approximately 97% of the citizens. The NHISS provides eligibility database (e.g., sociodemographic information), medical service claims database (e.g., diagnosis statements, medical treatments, drug prescriptions), and health examination database (e.g., anthropometric measurements, self-questionnaire on health behaviors, laboratory test results). All subscribers are recommended to receive a regular health examination every 2 years.
This study was approved by the Institutional Review Board (IRB) of Samsung Medical Center (No. 2023-11-122). Because the data used in the analysis were anonymous and de-identifiable, informed consent was waived by the IRB.
Case identification
Data from NHISS database was extracted to identify 1,547,928 subjects who were newly diagnosed with depression (International Classification of Diseases, 10th Revision [ICD-10] codes: F32, F33) between January 1st, 2010 and December 31st, 2016 and received health examinations within 2 years before and after the diagnosis. Among these subjects, those who had missing data for one or more variables (n=35,902), who were under the age of 40 at baseline (n=162,617), and those who had been diagnosed with all-cause dementia (ICD-10 codes: F00, F01, F02, F03, G23.1, G30, G31; n=40,836) before index year were excluded. Additionally, to minimize the effects of reverse causality and confounding of preexisting conditions, we excluded those who were newly diagnosed with dementia during the first year of follow-up (n=16,873). As a result, a total of 1,291,706 subjects were included in the study. These participants were followed from their first health examination after a diagnosis of depression until the onset of dementia, death, or December 31, 2016, whichever occurred first (Figure 1). The mean follow-up duration was 4.27±1.95 years.
Changes in PA
Information on PA status was obtained during the medical examinations using a modified version of the International Physical Activity Questionnaire (IPAQ)-Short form [17]. The IPAQ has previously demonstrated acceptable validity (median ρ of about 0.30) in monitoring PA behaviors in diverse populations [17]. The questions ask the frequency of mild- (“In the past week, on how many days did you walk for at least 30 minutes per day, including at least 10 minutes at a time? [e.g., light exercise, including walking to and from work or leisure time]”), moderate- (“In the past week, on how many days did you engage in activities that made you breathe a little harder than usual for more than 30 minutes a day? [e.g., power walking, playing doubles tennis, and biking at normal speed]”), and vigorous-intensity (“In the past week, on how many days did you engage in strenuous activities that made you out of breath for more than 20 minutes a day? [e.g., running, aerobics, riding a bicycle at high speed, and mountain climbing]”) PA. We defined “regular physical activity” as engaging in moderate-intensity PA ≥5 or vigorous-intensity PA ≥3 per week, according to the guidelines for exercise testing and prescription [18]. Based on the status of regular PA obtained during the two medical examinations, before and after the diagnosis of depression, subjects were classified as “never” (no PA regular exercise → no PA regular exercise), “beginning” (no regular PA → regular PA), “discontinuation” (regular PA → no regular PA), “maintenance” (regular PA → regular PA).
Outcome
In this retrospective cohort study, the primary outcome was the occurrence of dementia, which was defined by at least two claims for prescription of antidementia with codes for AD (ICD-10 codes: F00 or G30), VaD (ICD-10 codes: F01), other dementia (ICD-10 codes: F02, F03, G23.1, or G31).
According to the National Health Reimbursement criteria, the prescription of acetylcholinesterase inhibitors or NMDA receptor antagonist (memantine) for dementia requires Korea-Mini Mental State Examination ≤26, along with either a Clinician Dementia Rating ≥1 or a Global Deterioration Scale ≥3.
Covariates
For sociodemographic factors, sex, age, and household income level were included. Household income levels were divided into quartiles depending on health insurance premium, which is determined based on wages and incomes of the insured. In addition, health behaviors, anthropometric and laboratory measurements were included as variables. Current smokers were defined as those who smoked >5 packs in a lifetime and continue to smoke and current drinkers were defined as those with intake alcohol >0 g per day. Obesity was defined by body mass index ≥25 kg/m2. Comorbid physical illnesses, such as type 2 diabetes mellitus (at least one claim for prescription of antidiabetic medication under ICD-10 codes of E11, E12, E13, or E14; or fasting glucose level ≥126 mg/dL), hypertension (at least one claim per year for the prescription of antihypertensive drugs under ICD-10 codes of I10-13 or I15; or a systolic/diastolic blood pressure ≥140/90 mm Hg), dyslipidemia (at least one claim per year for the prescription of anti-hyperlipidemic agents under ICD-10 code of E78; or total cholesterol ≥240 mg/dL), and chronic kidney disease (an estimated glomerular filtration rate <60 mL/min) were also included.
Statistical analysis
Baseline characteristics were presented as mean±standard deviation for continuous variables or numbers and percentages for categorical variables. To investigate the differences in baseline characteristics among PA groups, analysis of variance and chi-squared test were conducted. Using the Cox proportional-hazards regression analyses, hazard ratios (HRs) and 95% confidence intervals (CIs) were reported to examine the association between changes in PA before and after the diagnosis of depression and dementia, adjusting for other variables. Incidence rates of all-cause dementia, AD, and VaD were presented per 1,000 person-year. Kaplan-Meier curves were plotted to illustrate cumulative incidence of all-cause dementia, AD, and VaD. Moreover, we performed subgroup analyses for age and sex with “never” group as a reference group. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Inc.).
RESULTS
Baseline characteristics
Most of the subjects did not engage in regular PA both before and after the diagnosis of depression (never group; n=861,591, 66.70%). Among the total subjects, 164,379 (12.72%) started to engage in regular PA (beginning group) whereas 154,383 (11.95%) stopped engaging in regular PA (discontinuation group) after the diagnosis of depression. Only 111,353 (8.2%) continued to engage in regular PA (maintenance group) after the diagnosis of depression. Baseline characteristics of each group are summarized in Table 1. Those in the maintenance group were more likely to be males, have higher household income, be nonsmokers, and consume alcohol. Also, they were less likely to have obesity, hypertension, dyslipidemia, and chronic kidney disease.
Changes in regular PA and dementia
Among the 1,291,706 subjects, 58,934 (4.56%) were newly diagnosed with dementia during the follow-up period. The incidence rate of all types of dementia was 10.70 per 1,000 person-years. Specifically, the incidence rate for AD is 8.79 per 1,000 person-years, while VaD has an incidence rate of 1.02 per 1,000 person-years.
Table 2 presents the HRs and CIs for dementia based on changes in regular PA before and after the diagnosis of depression. Overall, compared to the never group, those who engaged in PA before and/or after the diagnosis of depression demonstrated a significantly decreased risk of all-cause dementia after adjusting for all compounding factors (Model 4), with the maintenance group having the lowest risk (adjusted HR [aHR]: 0.73; 95% CI: 0.70, 0.75). Similarly, the maintenance group had the lowest risk of AD (aHR=0.73; 95% CI: 0.70, 0.76) and VaD (aHR: 0.66; 95% CI: 0.58, 0.74) compared to never group in Model 4. Moreover, the aHRs of beginning group were higher than the aHRs of discontinuation group for all-cause dementia, and AD in Model 4. Figure 2 illustrates the incident of all-cause dementia, AD, and VaD according to changes in regular PA.
Subgroup analyses by age and sex
The subgroup analyses in Table 3 revealed that the effect of changes in regular PA before and after the diagnosis of depression on the risk for all-cause dementia was significantly stronger in females than men in Model 3 (p for interaction=0.01). Among males, discontinuation group did not show a significantly decreased risk of all-cause dementia compared to the never group. In assessing the effect of age on the association between changes in regular PA and all-cause dementia, the results were generally consistent with the main results in the age groups of 60s, 70s, and over 80s. However, among the age of 40s, there was no significant difference in the risk of all-dementia in maintenance, beginning, and discontinuation groups compared to never group. Additionally, among the age of 50s, only maintenance group had a significantly decreased risk of all-cause dementia compared to never group (aHR: 0.69; 95% CI: 0.59, 0.80) in Model 3. Supplementary Tables 1 and 2 further provides the effect of regular PA on the risk of AD and VaD.
DISCUSSION
The goal of this study was to explore whether the risk of dementia varies depending on changes in PA status before and after the onset of depression, using retrospective national cohort data. Approximately 4.56% of our sample developed dementia within few years after the onset of depression. Overall, those who consistently exercised before and after the diagnosis of depression were at the lowest risk of dementia. Specifically, this group had approximately 27% reduced risk of all-cause dementia and AD and 34% reduced risk of VaD compared to those who never engaged in PA before and after the diagnosis of depression. These results align with previous studies suggesting the potential preventive effect of PA on cognitive decline. According to a systematic review and meta-analysis of prospective cohort study, a high and a moderate amount of PA could decrease the risk of all-cause dementia, AD, and VaD among older adults [19]. A similar study concluded that even after reducing the effect of reverse causality, PA could play a role as a protective lifestyle factor for dementia [20]. Also, a previous cross-sectional study showed that regular PA is associated with better cognitive test performance in both depressed and non-depressed subjects compared to low level of PA, but the difference in the performance between regular PA and low PA groups were larger in the depressed subjects [14]. This result may imply that such preventive effect of PA may be more prominent among depressed patients. Furthermore, our findings may indirectly support previously proposed neurobiological mechanisms. Specifically, consistent PA could potentially mitigate cognitive decline through mechanisms such as reducing inflammation, improving cerebrovascular health, and promoting neurogenesis. However, given the observational nature of our study, we cannot definitively confirm these pathways. Future studies utilizing biomarker assessments or neuroimaging techniques could further elucidate these mechanistic relationships. Because our study also took account of changes in the status PA, the results from the current study further suggested that the consistency of PA may be a key factor in lowering the risk of dementia among depressed patients.
Subgroup analyses by sex revealed that the effect of changes in PA on the risk of dementia is stronger in women than in men. This finding concurs with previous studies that showed that exercise may elicit greater cognitive benefits, such as enhancing executive functioning, in women than men [21-23]. According to these studies, sex differences in neuroplasticity, glucometabolic and hypothalamic-pituitary-adrenal axis responses, sex steroid hormones, and brain-derived neurotrophic factor may explain such sex difference in exercise efficacy in improving cognition. However, a significant methodological limitation of our study is the use of identical PA criteria for both sexes. Given evidence suggesting that men may require higher PA intensity than women to achieve comparable cognitive benefits [24], our approach may underestimate the potential benefits of PA for men, thus influencing the interpretation of observed sex differences. Future research should consider applying sex-specific PA thresholds to more accurately evaluate these relationships. The subgroup analyses related to age groups revealed that the effect of PA in reducing the risk of dementia is less pronounced among relatively younger age groups of 40s and 50s. Considering that early-onset dementia is less common, a smaller number of dementia events may have contributed to these nonsignificant results in younger age groups.
This study has several limitations. First, PA status was assessed through self-report questionnaires limited to two oneweek recall periods, which may not accurately reflect long-term or habitual activity patterns relevant to dementia risk. Depressed individuals may be especially prone to recall bias or underreporting due to cognitive impairment or low motivation, potentially compromising the accuracy of these self-reports. Future studies should incorporate objective measures, such as accelerometers or activity trackers, to improve PA assessment precision. Second, the measurement of PA changes spanned up to 2 years before and after the depression diagnosis, creating ambiguity in the temporal relationship between changes in PA and depression onset. Specifically, this method cannot definitively clarify whether changes in PA preceded depression or occurred as a response to depressive symptoms. Prospective study designs with more frequent assessments could better clarify these temporal dynamics. Third, depression diagnoses were based solely on claims data, lacking detailed clinical information. Consequently, critical factors— including depression type, severity, duration, specific symptoms, comorbid psychiatric conditions, and treatment details (e.g., antidepressant use)—that could significantly influence PA engagement were not accounted for. These unmeasured factors may confound the observed association between PA and dementia risk, potentially biasing our estimates. Future studies should incorporate these detailed clinical data to better control for these potential confounding variables and enhance the robustness of findings. Fourth, the inclusion criteria requiring two regular health examinations may introduce selection bias. Participants included in this study might be healthier, more health-conscious, and more inclined toward health-promoting behaviors compared to the general depressed population. Additionally, individuals who consistently engage in PA may differ systematically from others in unmeasured health behaviors or access to healthcare resources, which could introduce residual confounding and potentially bias the observed associations. These biases may limit the generalizability of our findings, warranting caution when extrapolating these results broadly. Fifth, due to the retrospective nature of this cohort study, causal inference remains inherently limited. Although we excluded dementia diagnoses within the first year of follow-up to mitigate reverse causality, this approach may not fully account for reverse causation, given the typically prolonged prodromal phase of dementia. Future studies could employ longer exclusion periods or advanced statistical approaches, such as instrumental variable analyses, to enhance causal inference and better address this limitation. Furthermore, the specific mechanisms underlying the protective effect of PA against dementia warrant additional investigation. Finally, because our study utilized data from the Korean National Health Insurance Service, generalizability to other populations may be limited due to differences in healthcare systems, healthcare accessibility, and cultural attitudes toward PA. Future research conducted across diverse international settings is necessary to enhance the generalizability of our findings.
Despite these limitations, using a large national sample, this study contributes to our understanding of how changes in PA status before and after the onset of depression relate to dementia risk. To our knowledge, this is the first study to examine the changes in PA before and after the diagnosis of depression in relation to dementia. Our study highlights that consistent PA is associated with a significantly reduced risk of developing dementia, supporting the notion that PA may serve as a protective factor against cognitive decline. Future studies should aim to employ more robust and longitudinal methodologies to clarify the mechanisms through which PA influences dementia risk and to explore the differential effects across various demographic groups. Overall, promoting consistent PA, especially among individuals with depression, may offer a viable strategy for mitigating dementia risk.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
