Association between cardiovascular health and dementia risk in older adults

Data collection and ethical declarations

The data was collected from the Korea National Health Insurance Service (NHIS) Senior Database, which included data for 558,147 people recruited by a simple 10% random sampling method out of a total of 5.5 million subjects aged ≥ 60 years in the National Health Information Database. , which is an online database only for qualified analysis centers, with formal payment and strict regulations on data publication ( The NHIS integrates all health-related data covering the entire population of the Republic of Korea. The NHIS-Senior database covered the following parameters: socio-demographic and socio-economic information, insurance status, bi-annual health check-ups (including anthropometric measurements, laboratory surveys and self-administered health behaviors), and medical history records, including prescription records.11.12.

This study was approved by the Yonsei University Health System Institutional Review Board (4-2020-1387), and informed consent was removed. The study was conducted in accordance with the ethical guidelines of the Declaration of Helsinki. The NHIS-Senior database used in this study (NHIS-2016-2-171) was compiled by NHIS Korea.

Study population and design

From the Korean NHIS-Senior database, 312,736 participants who underwent health examinations between 2004 and 2012 were selected, and medical records were reviewed up to December 2014. Among them, a total of 278 003 participants aged 65 or over were screened. The following participants were excluded: (i) with missing or inadequate data to derive CVH measures in the baseline health examination (n=17,309); and (ii) having a history of any type of dementia (n=9,658), ischemic stroke or transient ischemic attack (n=32,182), hemorrhagic stroke (n=1,118). Finally, we included 217,736 participants without dementia or cerebrovascular disease in the study analysis (Fig. 1). An overview of the study design is shown in Fig. Additional S1. There were a total of 5 study enrollments with baseline health examinations at 2-year intervals for 8 years from 2004 to 2012.

Figure 1

Flowchart of the selection of the study population. Abbreviations: CVH cardiovascular health, SNIS Korea National Health Insurance Service, AITtransient ischemic attack.

CVH Metrics and CVH Status

The CVH measures used in this study were based on the AHA thresholds and criteria4. Dietary habits were not recorded during data collection and are therefore excluded from CVH measurements. Information on the other 6 measurements was obtained through routine health examinations and laboratory measurements. At each examination cycle, an interview and physical examination were conducted for each participant, and medical history information was collected.

We categorized the 6 metrics (total cholesterol, fasting blood glucose, BP, body mass index [BMI], smoking, and exercise) in 3 levels of poor, intermediate, and optimal, as recommended by the AHA (Supplementary Table S1). We used the sum of each optimal metric level to calculate the number of optimal CVH metrics, ranging from 0 to 6. Next, we stratified the participants into three groups based on the calculated number of optimal metrics as follows: low (0 –2 optimal metrics), CVH status moderate (3 to 4 optimal metrics) and high (5 to 6 optimal metrics). We also derived the continuous 12-point CVH score by calculating the sum of the scores assigned to each CVH metric level (poor = 0, intermediate = 1, and optimal = 2).

Comorbidities and other covariates

To adjust for potential confounders, except for diseases excluded at baseline according to the exclusion criteria, we selected basic sociodemographic factors, economic status based on income-related health insurance premiums, comorbidities and drugs that might be determinants of dementia incidence in the database available as covariates. for the main analysis. We obtained information on selected comorbidities from inpatient and outpatient diagnoses in the NHIS-Senior database using International Classification of Diseases-10th Revision (ICD-10) codes. Details of all ICD-10 definitions and codes used for clinical outcomes, covariates and comorbidities, including diseases in the exclusion criteria, are presented in Supplementary Table S2.

Some of the study participants (n=11,273) were screened for depressed mood, lower extremity function, and cognitive function at baseline. The Depressed Mood Screening involved selecting participants who needed an accurate diagnosis and counseling for depression through mood state questionnaires. Cognitive function was examined using the Activities of Daily Living Scale and the Korean Dementia Screening Questionnaire (KDSQ) which consists of questions on global memory function and instrumental life activities daily, including 5 items that can detect early changes in cognitive decline to diagnose dementia.13. The KDSQ is not influenced by age or education level and showed a sensitivity of 0.79 and a specificity of 0.80 in predicting dementia13.

Dementia Assessment

We defined a diagnosis of dementia based on the relevant ICD-10 codes (F00 or G30 for AD, F01 for VaD, F02 for dementia with other diseases classified elsewhere, and F03 or G31 for unspecified dementia) and the prescription drugs for dementia (rivastigmine, galantamine, memantine or donepezil) in records extracted from NHIS claims data (Supplementary Table S2). When codes for AD and VaD were recorded, we followed the primary diagnosis. The same participant could have one or more clinical events during the study period, but only the first event of each outcome was considered. To assess the accuracy of our definition of dementia, a validation study was conducted in 2 hospitals with 972 patients, using patient medical records and cognitive function test results, and the positive predictive value was 94.7%.14,15,16.

The time scale for study outcomes was defined as the time in months from the time of study enrollment to the date of first diagnosis with each type of dementia according to NHIS claims data. Participants without dementia were censored at the end of the observation period (end of study period or death).

statistical analyzes

Continuous variables are presented as mean ± standard deviation (SD) for normally distributed data or median (interquartile range [IQR]) for non-normally distributed data, and categorical variables are presented as a number (percentage). Incidence rates and differences in absolute rates of dementia by CVH status are presented as events per 100 person-years (PY) at risk. Differences in cumulative incidence between the three groups were assessed using the log-rank test. We estimated the hazard ratios (HR) of each type of dementia by CVH status, for 1 additional CVH metric at the optimal level, per 1 point increase in continuous CVH score and the levels of individual CVH components using time-varying Cox. proportional hazards models. Levels of individual CVH measurements, CVH status, number of optimal CVH measurements, and CVH score, which changes with each examination performed every 2 years during the study period, were used as time-varying variables for the dementia risk analysis. In sensitivity analyses, participants with intermediate stroke events were also censored on stroke date if they occurred before the end of the follow-up period. Unless otherwise stated, stroke censoring was performed in all results of the present study. We also assessed the risk of each type of dementia based on the number of optimal CVH measurements and the 12-point CVH score using the groups with the median of the number of optimal CVH measurements and the median of the CVH score as reference groups.

CVH measurements already contain multiple anthropometric (BMI, BP) and laboratory (fasting glucose, total cholesterol) factors that are affected by underlying comorbidities such as hypertension, diabetes, obesity, and hyperlipidemia . Therefore, we excluded the presence or absence of these underlying comorbidities from the fitted variables to avoid multi-collinearity or overfitting. All multivariate regression models were adjusted for age, sex, economic status, place of residence, comorbidities (atrial fibrillation, heart failure, myocardial infarction, coronary heart disease, peripheral arterial disease, anemia, kidney disease chronic, hyperthyroidism, hypothyroidism, osteoporosis, sleep apnea, chronic obstructive pulmonary disease, chronic liver disease, and cancer), medications (oral anticoagulants, antiplatelet agents), depressed mood, lower extremity function, and cognitive function at the time of initial examination . In the time-varying analysis of the association of dementia risk by individual CVH parameter levels, the CVH components were mutually adjusted for each other in addition to all other covariates. All analyzes were two-sided and P