Baseline hypertension, adiposity, depressive symptoms, and incident stroke in middle-aged and older Chinese adults: a longitudinal analysis of CHARLS, 2011–2020 Authors: Wanbin Li, Jiayun Shao, Yueqi Zhu Published: 2026 ABSTRACT BackgroundStroke is a major cause of disability and death among middle-aged and older adults in China. Hypertension, adiposity, and depressive symptoms often coexist, but their associations with incident stroke are often examined separately. We assessed baseline cardiometabolic, adiposity-related, and psychosocial factors associated with first reported incident stroke in a longitudinal China Health and Retirement Longitudinal Study (CHARLS) cohort.MethodsWe used CHARLS data. Participants aged 45 years or older who were free of stroke at baseline and had at least one follow-up assessment were eligible. A person-period dataset was constructed to incorporate repeated follow-up intervals and the timing of first reported incident stroke. Person-period modified Poisson regression models with participant-level cluster-robust standard errors were used to estimate risk ratios (RRs) and 95% confidence intervals (CIs).ResultsAmong 15,992 eligible participants, 1,246 first reported incident stroke events were identified during follow-up. The body mass index (BMI) and waist circumference complete-case models included 11,674 and 11,604 participants, with 949 and 941 incident stroke events, respectively. In the BMI model, hypertension (RR, 1.80; 95% CI, 1.56–2.07), diabetes (RR, 1.51; 95% CI, 1.21–1.88), heart problems (RR, 1.39; 95% CI, 1.17–1.64), BMI (RR per 1 kg/m², 1.05; 95% CI, 1.03–1.06), and 10-item Center for Epidemiologic Studies Depression Scale (CESD-10) score (RR per 1 point, 1.03; 95% CI, 1.02–1.04) were associated with increased stroke risk. In the waist circumference model, waist circumference was also positively associated with incident stroke (RR per 1 cm, 1.02; 95% CI, 1.01–1.03). Sensitivity analyses yielded broadly consistent results.ConclusionsHypertension, diabetes, heart problems, higher BMI, larger waist circumference, and higher CESD-10 score were associated with increased risk of first reported incident stroke among middle-aged and older Chinese adults. These observational findings suggest that vascular, adiposity-related, and psychosocial factors may help characterize stroke risk in this population, but they should be interpreted cautiously given anthropometric missingness, self-reported stroke ascertainment, and potential residual confounding.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28310-3. Background Stroke remains a major cause of death and long-term disability worldwide and continues to place a heavy burden on public health in China, especially among middle-aged and older adults [1, 2]. As the population ages and chronic non-communicable diseases become more common, there is growing interest in identifying modifiable factors associated with future stroke risk [1–3]. Hypertension is one of the best-established risk factors for stroke and remains a cornerstone of prevention [3, 4]. At the same time, stroke risk in later life is shaped by more than blood pressure alone. Adiposity has received increasing attention because both overall obesity and central obesity have been linked to adverse cardiovascular outcomes, including stroke [5–7]. Body mass index (BMI) is widely used as an indicator of general adiposity, whereas waist circumference may better reflect abdominal fat accumulation and the cardiometabolic risk that comes with it [7]. Since these measures capture related but not identical aspects of adiposity, it is useful to examine them separately in longitudinal population-based studies. Psychological factors may also matter. Depressive symptoms are common in middle-aged and older adults and have been linked to a higher risk of cardiovascular disease and stroke [8–10]. Even so, fewer studies have assessed whether depressive symptoms remain associated with incident stroke after conventional vascular factors and adiposity-related measures are considered together in a nationally representative Chinese cohort. Many previous analyses of large longitudinal datasets have focused on a single exposure domain, such as depressive symptoms, insulin resistance and sarcopenia, or obesity-related indices [10–12]. Those studies have been informative, but they offer only part of the picture. In everyday clinical and community settings, vascular risk factors, excess adiposity, and depressive symptoms often coexist. Looking at them within the same analytical framework may therefore be more useful for public health practice and risk stratification. Using data from the China Health and Retirement Longitudinal Study (CHARLS), a nationally representative longitudinal cohort [13], we examined whether baseline hypertension, adiposity, and depressive symptoms were associated with first reported incident stroke during follow-up from 2011 to 2020 among adults aged 45 years or older. To better characterize adiposity-related risk while minimizing collinearity, we fitted separate multivariable models including BMI or waist circumference. We also performed several sensitivity analyses to evaluate the robustness of the findings. Methods Study design and data source This study used data from the CHARLS, a nationally representative longitudinal survey of middle-aged and older adults in China [13]. CHARLS employed a multistage probability sampling design and collected detailed information on sociodemographic characteristics, health status, lifestyle factors, biomarkers, and anthropometric measurements. The present analysis used baseline information from 2011 and follow-up data through 2020. Study population and follow-up structure Participants aged 45 years or older at baseline who were free of stroke and had at least one follow-up assessment were eligible for the present analysis. Participants younger than 45 years at baseline were excluded so that the primary analytic sample corresponded to the target population of middle-aged and older adults. To better incorporate the longitudinal structure of CHARLS and the timing of first reported stroke, we constructed a person-period dataset. Each participant contributed one or more follow-up interval records until the first reported incident stroke event or the last available follow-up wave, whichever occurred first. The follow-up intervals were defined according to the observed CHARLS survey waves. Incident stroke was defined as the first self-reported physician-diagnosed stroke during follow-up. Baseline exposures and covariates The primary adiposity indicators were baseline body mass index (BMI, kg/m²) and waist circumference (cm). BMI was derived from measured height and weight when a direct BMI variable was unavailable. Waist circumference was obtained from baseline anthropometric measurements. Baseline covariates included age, sex, education, marital status, hukou type, smoking status, drinking status, hypertension, diabetes, heart problems, and depressive symptoms. Hukou type was categorized as agricultural hukou versus non-agricultural or other hukou. Smoking status was categorized as never versus ever smoking based on the baseline smoking question. Drinking status was categorized as current drinking versus no current drinking based on baseline alcohol-use frequency; participants reporting drinking more or less than once per month were classified as current drinkers. Depressive symptoms were assessed using the 10-item Center for Epidemiologic Studies Depression Scale (CESD-10) [14, 15]. Data cleaning and cohort construction Baseline variables were extracted from the 2011 demographic background, health status and functioning, biomarkers, and weight files. Follow-up stroke information was extracted from the corresponding health status and functioning files for 2013, 2015, 2018, and 2020. Individual identifiers were harmonized across waves and raw CHARLS files before linkage. Baseline anthropometric variables and behavioral variables were checked against the original CHARLS files and then linked to the person-period dataset. Implausible anthropometric values were set to missing before analysis. The cohort construction, missingness assessment, and model-specific complete-case analytical samples are summarized in Fig. 1. Fig. 1Flow diagram of cohort construction, missingness assessment, and model-specific complete-case analytical samples. BMI, body mass index; CHARLS, China Health and Retirement Longitudinal Study Flow diagram of cohort construction, missingness assessment, and model-specific complete-case analytical samples. BMI, body mass index; CHARLS, China Health and Retirement Longitudinal Study Statistical analysis Baseline characteristics were summarized among participants included in at least one of the primary complete-case models. Continuous variables are presented as mean and standard deviation (SD), and categorical variables are presented as numbers and percentages. Baseline characteristics were compared between participants with and without incident stroke using t tests for continuous variables and chi-square or Fisher’s exact tests for categorical variables, as appropriate. Because exact stroke onset dates were unavailable, conventional Cox time-to-event models based on exact event times were not used. Instead, we used the follow-up wave of first reported stroke to construct a person-period discrete-time dataset. For the primary analysis, we fitted person-period modified Poisson regression models with a log link to estimate risk ratios (RRs) and 95% confidence intervals (CIs) for incident stroke [16]. Robust standard errors were clustered at the participant level to account for within-participant correlation across repeated person-period records. Follow-up interval was included as a categorical covariate to account for the discrete-time structure of the data and unequal follow-up intervals. Two primary models were fitted separately. One model included BMI, and the other included waist circumference. These adiposity measures were modeled separately because they capture related but distinct dimensions of adiposity and may be correlated if included simultaneously. Both models were adjusted for age, sex, education, marital status, hukou type, smoking status, drinking status, hypertension, diabetes, heart problems, and CESD-10 score. Complete-case analysis was used to ensure that all covariates in each multivariable model were observed and consistently defined within the person-period framework. To assess potential selection bias due to missing data, we summarized variable-level missingness, described the complete-case flow, and compared baseline characteristics between participants included in and excluded from the primary analyses. Multiple imputation was not performed because the main missingness arose from baseline anthropometric measurements in the biomarker/physical examination component rather than from isolated item nonresponse. We therefore prioritized transparent complete-case estimates with detailed missingness diagnostics and interpreted the findings cautiously. Several sensitivity analyses were performed to evaluate the robustness of the findings. Logistic regression models with cluster-robust standard errors were fitted to facilitate comparison with the original modeling strategy. The modified Poisson models were then repeated after excluding smoking and drinking variables and after using more detailed behavioral covariates, including three-category smoking status and drinking frequency. Generalized estimating equation (GEE) Poisson models with an exchangeable working correlation structure were also used as an alternative approach to within-participant correlation [17]. To examine whether non-standard observation patterns affected the estimates, we additionally restricted the analysis to standard adjacent follow-up intervals. All analyses were conducted using R. A two-sided P value < 0.05 was considered statistically significant. Results Study population and baseline characteristics Among 15,992 eligible participants, 1,246 first reported incident stroke events were identified during follow-up. After complete-case restriction, the BMI and waist circumference models included 11,674 and 11,604 participants, with 949 and 941 incident stroke events, respectively. A total of 11,810 participants were included in at least one of the two primary complete-case models, among whom 959 experienced incident stroke. Table 1 shows the baseline characteristics of participants included in the primary analyses. Participants who developed incident stroke were older than those without incident stroke. They also had higher BMI, larger waist circumference, and higher CESD-10 scores. The prevalences of hypertension, diabetes, and heart problems were higher among participants who developed incident stroke. Smoking and drinking status did not differ significantly between participants with and without incident stroke. Table 1Baseline characteristics by incident stroke statusCharacteristicTotalNo incident strokeIncident strokeP valueN11,81010,851959Age, years59.24 (9.46)59.07 (9.50)61.12 (8.78)< 0.001Sex, n (%)0.835 Male5512 (46.7%)5068 (46.7%)444 (46.3%) Female6298 (53.3%)5783 (53.3%)515 (53.7%)Education, n (%)0.03 Primary school or below8443 (71.5%)7722 (71.2%)721 (75.2%) Middle school2510 (21.3%)2334 (21.5%)176 (18.4%) High school or above857 (7.3%)795 (7.3%)62 (6.5%)Marital status, n (%)0.105 Not married1508 (12.8%)1369 (12.6%)139 (14.5%) Married/partnered10,302 (87.2%)9482 (87.4%)820 (85.5%)Hukou type, n (%)0.813 Agricultural hukou9890 (83.7%)9090 (83.8%)800 (83.4%) Non-agricultural or other hukou1920 (16.3%)1761 (16.2%)159 (16.6%)Smoking status, n (%)0.316 Never smoker7131 (60.4%)6567 (60.5%)564 (58.8%) Ever smoker4679 (39.6%)4284 (39.5%)395 (41.2%)Drinking status, n (%)0.551 No current drinking7958 (67.4%)7303 (67.3%)655 (68.3%) Current drinking3852 (32.6%)3548 (32.7%)304 (31.7%)Hypertension, n (%)< 0.001 No9093 (77.0%)8517 (78.5%)576 (60.1%) Yes2717 (23.0%)2334 (21.5%)383 (39.9%)Diabetes, n (%)< 0.001 No11,174 (94.6%)10,311 (95.0%)863 (90.0%) Yes636 (5.4%)540 (5.0%)96 (10.0%)Heart problems, n (%)< 0.001 No10,491 (88.8%)9711 (89.5%)780 (81.3%) Yes1319 (11.2%)1140 (10.5%)179 (18.7%)BMI, kg/m²23.42 (3.84)23.33 (3.81)24.40 (4.06)< 0.001Waist circumference, cm85.13 (10.18)84.86 (10.11)88.14 (10.51)< 0.001CESD-10 score8.55 (6.34)8.45 (6.30)9.68 (6.62)< 0.001Note: Values are mean (standard deviation [SD]) or n (%)P values compare participants with and without incident stroke. Sex was coded as male and female; education was grouped as primary school or below, middle school, and high school or above; marital status was grouped as not married and married/partnered; hukou type was grouped as agricultural hukou and non-agricultural or other hukou. Smoking was categorized as never versus ever smoking, and drinking as no current drinking versus current drinking Table 1Baseline characteristics by incident stroke statusCharacteristicTotalNo incident strokeIncident strokeP valueN11,81010,851959Age, years59.24 (9.46)59.07 (9.50)61.12 (8.78)< 0.001Sex, n (%)0.835 Male5512 (46.7%)5068 (46.7%)444 (46.3%) Female6298 (53.3%)5783 (53.3%)515 (53.7%)Education, n (%)0.03 Primary school or below8443 (71.5%)7722 (71.2%)721 (75.2%) Middle school2510 (21.3%)2334 (21.5%)176 (18.4%) High school or above857 (7.3%)795 (7.3%)62 (6.5%)Marital status, n (%)0.105 Not married1508 (12.8%)1369 (12.6%)139 (14.5%) Married/partnered10,302 (87.2%)9482 (87.4%)820 (85.5%)Hukou type, n (%)0.813 Agricultural hukou9890 (83.7%)9090 (83.8%)800 (83.4%) Non-agricultural or other hukou1920 (16.3%)1761 (16.2%)159 (16.6%)Smoking status, n (%)0.316 Never smoker7131 (60.4%)6567 (60.5%)564 (58.8%) Ever smoker4679 (39.6%)4284 (39.5%)395 (41.2%)Drinking status, n (%)0.551 No current drinking7958 (67.4%)7303 (67.3%)655 (68.3%) Current drinking3852 (32.6%)3548 (32.7%)304 (31.7%)Hypertension, n (%)< 0.001 No9093 (77.0%)8517 (78.5%)576 (60.1%) Yes2717 (23.0%)2334 (21.5%)383 (39.9%)Diabetes, n (%)< 0.001 No11,174 (94.6%)10,311 (95.0%)863 (90.0%) Yes636 (5.4%)540 (5.0%)96 (10.0%)Heart problems, n (%)< 0.001 No10,491 (88.8%)9711 (89.5%)780 (81.3%) Yes1319 (11.2%)1140 (10.5%)179 (18.7%)BMI, kg/m²23.42 (3.84)23.33 (3.81)24.40 (4.06)< 0.001Waist circumference, cm85.13 (10.18)84.86 (10.11)88.14 (10.51)< 0.001CESD-10 score8.55 (6.34)8.45 (6.30)9.68 (6.62)< 0.001Note: Values are mean (standard deviation [SD]) or n (%)P values compare participants with and without incident stroke. Sex was coded as male and female; education was grouped as primary school or below, middle school, and high school or above; marital status was grouped as not married and married/partnered; hukou type was grouped as agricultural hukou and non-agricultural or other hukou. Smoking was categorized as never versus ever smoking, and drinking as no current drinking versus current drinking Baseline characteristics by incident stroke status Note: Values are mean (standard deviation [SD]) or n (%) P values compare participants with and without incident stroke. Sex was coded as male and female; education was grouped as primary school or below, middle school, and high school or above; marital status was grouped as not married and married/partnered; hukou type was grouped as agricultural hukou and non-agricultural or other hukou. Smoking was categorized as never versus ever smoking, and drinking as no current drinking versus current drinking Missingness and complete-case assessment Missingness patterns and the complete-case flow are summarized in Supplementary Table S1 (Additional file 1). The main sources of missingness were waist circumference and BMI, which were missing for 3,454 (21.60%) and 3,378 (21.12%) eligible participants, respectively. Smoking and drinking covariates were nearly complete. Among the 15,992 eligible participants, 11,674 were included in the BMI complete-case model and 11,604 were included in the waist circumference complete-case model. The corresponding numbers of included incident stroke events were 949 and 941, respectively. Baseline characteristics of participants included in and excluded from the primary analyses are shown in Supplementary Table S2 (Additional file 2). Compared with excluded participants, those included in the primary analyses had a higher proportion of incident stroke and differed in several baseline characteristics, including sex, education, hukou type, drinking status, hypertension, heart problems, CESD-10 score, and waist circumference. Primary person-period modified Poisson models Table 2 presents the primary person-period modified Poisson models. In the BMI model, older age was associated with a higher risk of incident stroke (RR per 1 year, 1.03; 95% CI, 1.02–1.04; P < 0.001). Hypertension was strongly associated with incident stroke (RR, 1.80; 95% CI, 1.56–2.07; P < 0.001). Diabetes (RR, 1.51; 95% CI, 1.21–1.88; P < 0.001) and heart problems (RR, 1.39; 95% CI, 1.17–1.64; P < 0.001) were also associated with increased stroke risk. BMI was positively associated with incident stroke, with an RR of 1.05 per 1 kg/m² increase (95% CI, 1.03–1.06; P < 0.001). Ever smoking was modestly associated with higher stroke risk (RR, 1.20; 95% CI, 1.01–1.43; P = 0.043), whereas current drinking was not. A higher CESD-10 score was also associated with increased stroke risk (RR per 1 point, 1.03; 95% CI, 1.02–1.04; P < 0.001). In the waist circumference model, the results were similar. Hypertension (RR, 1.76; 95% CI, 1.52–2.04; P < 0.001), diabetes (RR, 1.51; 95% CI, 1.21–1.87; P < 0.001), and heart problems (RR, 1.35; 95% CI, 1.13–1.60; P < 0.001) remained significantly associated with incident stroke. Waist circumference was positively associated with stroke risk (RR per 1 cm, 1.02; 95% CI, 1.01–1.03; P < 0.001), and the CESD-10 score remained significant (RR per 1 point, 1.03; 95% CI, 1.02–1.04; P < 0.001). Ever smoking was also modestly associated with incident stroke (RR, 1.21; 95% CI, 1.02–1.45; P = 0.031), whereas current drinking was not significantly associated with incident stroke. Table 2Associations of baseline characteristics with incident stroke in person-period modified Poisson modelsCharacteristicBMI model RR (95% CI)BMI model P valueWaist model RR (95% CI)Waist model P valueAge, per 1 year1.03 (1.02–1.04)< 0.0011.03 (1.02–1.03)< 0.001Sex: Female vs. male0.92 (0.76–1.11)0.4030.95 (0.78–1.14)0.572Education: Middle school vs. primary school or below0.95 (0.80–1.13)0.5940.95 (0.80–1.13)0.58Education: High school or above vs. primary school or below1.06 (0.81–1.39)0.6721.05 (0.80–1.38)0.708Marital status: Married/partnered vs. not married0.97 (0.80–1.18)0.7930.97 (0.79–1.17)0.723Hukou type: Non-agricultural or other hukou vs. agricultural hukou0.97 (0.81–1.16)0.7220.94 (0.79–1.13)0.526Smoking: Ever smoker vs. Never smoker1.20 (1.01–1.43)0.0431.21 (1.02–1.45)0.031Drinking: Current drinking vs. No current drinking1.02 (0.87–1.20)0.7721.01 (0.86–1.18)0.903Hypertension: Yes vs. No1.80 (1.56–2.07)< 0.0011.76 (1.52–2.04)< 0.001Diabetes: Yes vs. No1.51 (1.21–1.88)< 0.0011.51 (1.21–1.87)< 0.001Heart problems: Yes vs. No1.39 (1.17–1.64)< 0.0011.35 (1.13–1.60)< 0.001BMI, per 1 kg/m²1.05 (1.03–1.06)< 0.001Waist circumference, per 1 cm1.02 (1.01–1.03)< 0.001CESD-10 score, per 1 point1.03 (1.02–1.04)< 0.0011.03 (1.02–1.04)< 0.001Note: Estimates are risk ratios (RRs) with 95% confidence intervalsModels were adjusted for follow-up interval and listed covariates. Reference categories were male sex, primary school or below education, not married, agricultural hukou, never smoking, no current drinking, and absence of the corresponding chronic condition. Continuous variables were modeled per unit shown in the row label Table 2Associations of baseline characteristics with incident stroke in person-period modified Poisson modelsCharacteristicBMI model RR (95% CI)BMI model P valueWaist model RR (95% CI)Waist model P valueAge, per 1 year1.03 (1.02–1.04)< 0.0011.03 (1.02–1.03)< 0.001Sex: Female vs. male0.92 (0.76–1.11)0.4030.95 (0.78–1.14)0.572Education: Middle school vs. primary school or below0.95 (0.80–1.13)0.5940.95 (0.80–1.13)0.58Education: High school or above vs. primary school or below1.06 (0.81–1.39)0.6721.05 (0.80–1.38)0.708Marital status: Married/partnered vs. not married0.97 (0.80–1.18)0.7930.97 (0.79–1.17)0.723Hukou type: Non-agricultural or other hukou vs. agricultural hukou0.97 (0.81–1.16)0.7220.94 (0.79–1.13)0.526Smoking: Ever smoker vs. Never smoker1.20 (1.01–1.43)0.0431.21 (1.02–1.45)0.031Drinking: Current drinking vs. No current drinking1.02 (0.87–1.20)0.7721.01 (0.86–1.18)0.903Hypertension: Yes vs. No1.80 (1.56–2.07)< 0.0011.76 (1.52–2.04)< 0.001Diabetes: Yes vs. No1.51 (1.21–1.88)< 0.0011.51 (1.21–1.87)< 0.001Heart problems: Yes vs. No1.39 (1.17–1.64)< 0.0011.35 (1.13–1.60)< 0.001BMI, per 1 kg/m²1.05 (1.03–1.06)< 0.001Waist circumference, per 1 cm1.02 (1.01–1.03)< 0.001CESD-10 score, per 1 point1.03 (1.02–1.04)< 0.0011.03 (1.02–1.04)< 0.001Note: Estimates are risk ratios (RRs) with 95% confidence intervalsModels were adjusted for follow-up interval and listed covariates. Reference categories were male sex, primary school or below education, not married, agricultural hukou, never smoking, no current drinking, and absence of the corresponding chronic condition. Continuous variables were modeled per unit shown in the row label Associations of baseline characteristics with incident stroke in person-period modified Poisson models Note: Estimates are risk ratios (RRs) with 95% confidence intervals Models were adjusted for follow-up interval and listed covariates. Reference categories were male sex, primary school or below education, not married, agricultural hukou, never smoking, no current drinking, and absence of the corresponding chronic condition. Continuous variables were modeled per unit shown in the row label Sensitivity analyses The sensitivity analyses did not materially change the interpretation of the primary findings (Supplementary Table S3 and Supplementary Table S3b; Additional files 3 and 4). Similar patterns were observed in logistic regression models with cluster-robust standard errors, modified Poisson models with alternative behavioral adjustment, GEE Poisson models, and analyses restricted to standard adjacent follow-up intervals. Across these specifications, hypertension, diabetes, heart problems, BMI or waist circumference, and CESD-10 score remained positively associated with incident stroke. Discussion In this longitudinal analysis of CHARLS, several baseline cardiometabolic, adiposity-related, and psychosocial characteristics were associated with first reported incident stroke among middle-aged and older Chinese adults. In person-period modified Poisson models with participant-level cluster-robust standard errors, hypertension, diabetes, heart problems, higher BMI, larger waist circumference, and higher CESD-10 score were associated with higher stroke risk. Ever smoking also showed a modest association after multivariable adjustment. These findings were broadly stable across alternative model specifications, including logistic regression, models with different behavioral adjustment, GEE Poisson models, and analyses restricted to standard adjacent follow-up intervals. The results suggest that stroke risk in later life is not well described by a single exposure domain; rather, vascular disease, adiposity, metabolic dysfunction, behavioral factors, and depressive symptoms may together characterize a group at higher risk. This interpretation is consistent with large international evidence showing that most stroke burden is attributable to potentially modifiable vascular, metabolic, behavioral, and psychosocial risk factors [18, 19]. Hypertension showed the strongest and most consistent association with incident stroke. This finding is biologically plausible given the central role of elevated blood pressure in vascular injury, arterial stiffness, endothelial dysfunction, small-vessel damage, and cerebrovascular remodeling [3, 18, 19]. Hypertension may also interact with other cardiometabolic abnormalities by accelerating atherosclerosis and increasing susceptibility to both ischemic and hemorrhagic cerebrovascular events. Diabetes was likewise associated with a higher risk of incident stroke, which is compatible with the known vascular consequences of chronic metabolic dysfunction. Hyperglycemia and diabetes can promote endothelial dysfunction, oxidative stress, inflammation, vascular stiffness, and prothrombotic changes; large collaborative evidence also supports diabetes as a major risk factor for vascular outcomes, including stroke [20]. Self-reported heart problems showed a positive association with incident stroke in the person-period models. Given the broad and self-reported nature of this variable, the finding should be interpreted cautiously and should not be taken to indicate a specific cardiac mechanism. It may instead reflect heterogeneous cardiovascular histories, including arrhythmia, heart failure, or other cardiac conditions that are clinically relevant to subsequent cerebrovascular risk [19]. Both BMI and waist circumference were positively associated with incident stroke when modeled separately. In the BMI model, each 1 kg/m² higher BMI corresponded to a 5% higher risk of incident stroke, while each 1 cm higher waist circumference corresponded to a 2% higher risk in the waist circumference model. These findings are consistent with the view that both general and central adiposity may contribute to cerebrovascular risk. Excess adiposity is closely linked to hypertension, insulin resistance, dyslipidemia, chronic low-grade inflammation, sleep-disordered breathing, and vascular dysfunction [5–7]. Waist circumference may additionally reflect visceral fat accumulation, which is metabolically active and more strongly related to insulin resistance and inflammatory signaling than body weight alone [7]. The similar direction of the BMI and waist circumference models suggests that adiposity-related risk was not dependent on a single anthropometric definition. Because BMI and waist circumference capture related but not identical aspects of adiposity, we modeled them separately to reduce potential collinearity and to allow interpretation of each indicator. A higher CESD-10 score was consistently associated with increased stroke risk across the primary and sensitivity analyses. This finding is consistent with meta-analytic and cohort evidence linking depressive symptoms to subsequent cardiovascular disease and stroke [8, 9]. Several pathways may underlie this association. Depressive symptoms may be associated with autonomic dysregulation, hypothalamic-pituitary-adrenal axis activation, inflammation, platelet activation, sleep disturbance, physical inactivity, smoking, unhealthy diet, and poorer adherence to cardiometabolic treatment. Depression may also be a marker of social vulnerability or subclinical vascular disease. Therefore, the CESD-10 association should not be interpreted as evidence that depressive symptoms directly cause stroke. Rather, it suggests that psychological distress may help identify individuals with an adverse vascular and behavioral risk profile who may benefit from more integrated prevention strategies. These findings have practical implications for stroke prevention in middle-aged and older adults. Hypertension remained the most stable factor, underscoring the continuing importance of early detection, sustained blood pressure control, and long-term treatment adherence. The associations of BMI and waist circumference also suggest that prevention strategies may need to address adiposity before it is expressed only as diagnosed hypertension or diabetes, for example through weight management, diet, physical activity, and screening for central obesity. The modest association of ever smoking supports the continued relevance of tobacco-control and smoking-cessation efforts in older adults. The CESD-10 finding further suggests that psychological distress may be worth considering within chronic disease management, especially because depressive symptoms can make adherence to antihypertensive, antidiabetic, and lifestyle interventions more difficult. In community settings, stroke prevention may therefore benefit from a more integrated assessment of vascular, metabolic, anthropometric, behavioral, and psychosocial risk. Limitations Several limitations should be considered. Stroke was based on self-reported physician diagnosis, and information on stroke subtype, severity, exact onset date, and medical validation was not available. Outcome misclassification therefore cannot be excluded, and mechanisms may differ between ischemic and hemorrhagic stroke. Although the person-period structure allowed us to use the follow-up wave in which stroke was first reported, the exact timing of onset within each interval remained unknown. Death before stroke could not be formally modeled as a competing risk because exact stroke onset dates and the temporal ordering of stroke and death within survey intervals were unavailable. The primary models used complete-case analysis, with missingness concentrated in baseline anthropometric measurements from the biomarker/physical examination component. Participants included in the primary analyses differed from excluded participants in several baseline characteristics, so selection bias due to missing data could not be fully ruled out. Several covariates were also self-reported or broadly defined. For example, heart problems may include heterogeneous cardiac conditions, and self-reported hypertension or diabetes may not fully capture measured blood pressure, glycemic control, disease duration, or treatment status. Residual confounding may remain because of unmeasured or imperfectly measured factors, such as measured blood pressure levels, antihypertensive or antidiabetic medication use, lipid profiles, lipid-lowering treatment, diet, physical activity, sleep, healthcare access, and stroke subtype. The observational design also precludes causal inference. Therefore, the estimated RRs should be interpreted as associations within the observed complete-case analytic samples rather than as causal effects or population-wide risk estimates. Conclusions In conclusion, hypertension, diabetes, heart problems, higher BMI, larger waist circumference, and higher CESD-10 score were associated with increased risk of first reported incident stroke among middle-aged and older Chinese adults. These associations were generally robust across alternative model specifications. These findings suggest that cardiometabolic, adiposity-related, behavioral, and psychosocial characteristics may help identify individuals at higher risk of first reported incident stroke. From a public health perspective, they are compatible with the value of integrated risk assessment in aging populations. However, because this was an observational analysis with anthropometric missingness, self-reported stroke ascertainment, lack of exact event dates, and potential residual confounding, the results should not be interpreted as evidence of causality. Supplementary Information Additional file 1: Table S1. Missingness and complete-case flow. Description: Variable-level missingness, model-specific complete-case counts, and sequential exclusion summaries. Additional file 1: Table S1. Missingness and complete-case flow. Description: Variable-level missingness, model-specific complete-case counts, and sequential exclusion summaries. Additional file 2: Table S2. Comparison of participants included in and excluded from the primary analyses. Description: Baseline characteristics comparing participants included in at least one primary model with those excluded from both primary models. Additional file 2: Table S2. Comparison of participants included in and excluded from the primary analyses. Description: Baseline characteristics comparing participants included in at least one primary model with those excluded from both primary models. Additional file 3: Table S3. Sensitivity analyses for associations with incident stroke. Description: Key sensitivity-analysis results using alternative model specifications and covariate coding. Additional file 3: Table S3. Sensitivity analyses for associations with incident stroke. Description: Key sensitivity-analysis results using alternative model specifications and covariate coding. Additional file 4: Table S3b. Robustness summary. Description: Summary of the direction, statistical significance, and estimate ranges across sensitivity analyses. Additional file 4: Table S3b. Robustness summary. Description: Summary of the direction, statistical significance, and estimate ranges across sensitivity analyses.