The Atherosclerosis Risk in Communities Study (ARIC) study began in 1987, and four field centers in ARIC (Washington County, Maryland, Forsyth County, North Carolina, Jackson City, Missouri, and Minneapolis) were randomly selected and recruited approximately 4,000 people between the ages of 45 and 64 years from specific populations in their communities as cohort samples (No authors listed, 1989). A total of 15,792 participants were comprehensively examined, including medical, social, and demographic data. The participants were re-examined every 3 years, and the first screening (baseline) was conducted in 1987–1989. Participants were asked to respond to phone calls every six months and attend ARIC exam visits periodically. The study program was approved by the institutional review committees of all study sites, and all participants provided written informed consent.

In this study, we used Visit 2 (1991–1993) as our baseline. In Visit 4, we excluded individuals without cognition assessment data at Visit 2 (n = 418), prevalent CVD (n = 1,524), and participants from the Minnesota and Maryland centers (n = 51). In the hs-CTnT group, we excluded individuals whose hs-CTnT levels were ≥14 ng/L or were missing hs-CTnT data at Visit 2 (n = 311) or Visit 4 (n = 136). Finally, 9,216 participants were included in the hs-CTnT group (Figure 1). For the NT-proBNP group, we excluded individuals whose NT-proBNP levels were ≥300 pg/mL or were missing NT-proBNP data at Visit 2 (n = 432) or Visit 4 (n = 134). Thus, 9,097 participants were included in the NT-proBNP group (Figure 1).

High-sensitivity cardiac troponin T was measured using a sandwich immunoassay on the Roche Elecsys 2010 Analyzer (Roche Diagnostics, Indianapolis, IN, United States) from plasma samples collected at Visit 2 and stored at −70°C. The detection limit of hs-cTnT was 6 ng/L and the limit of blank was 3 ng/L. For individuals with undetectable hs-cTnT levels, a value of 1.5 ng/L was imputed. The interassay coefficient of variation for hs-cTnT was 6.4% (at a mean control of 29 ng/L) (Matsushita et al., 2018). An elevated hs-cTnT level was set as ≥14 ng/L, according to the 99th percentile value reported by the manufacturer, and defined as subclinical myocardial damage (Thygesen et al., 2018). NT-proBNP was also measured from plasma samples on the Roche Elecsys 2010 Analyzer. The lower limit of detection was 5 pg/mL. For individuals with undetectable NT-proBNP levels, a value of 2.5 ng/L was imputed. The interassay coefficient of variation for NT-proBNP was 7.4% (at a mean control of 134 pg/mL) (Nambi et al., 2013). An elevated NT-proBNP level was set as ≥300 ng/L according to previous pre-specified levels (Blaha et al., 2016; Lazo et al., 2016; Hussain et al., 2021), yielding a 98% negative predictive value to exclude heart failure (Januzzi et al., 2006). The predictive values of these cut-off points for CVD were also validated by previous ARIC studies (Saunders et al., 2011; Hussain et al., 2021).

In Visits 2, 4, and 5, professional researchers conducted three cognitive tests on each participant in a standard order in a quiet room. These tests included a delayed word recall test (DWRT), digit symbol substitution test (DSST), and word fluency test (WFT), which measured memory, executive function, and language function, respectively (Kim et al., 2017; Li et al., 2020). In the DWRT, participants needed to learn a word and apply the word to one or two sentences. After 5 min, participants had 1 min to recall ten words, and their score was the correct number of words (Knopman and Ryberg, 1989). In the DSST, participants used a keyboard to translate symbols into numbers, and their score was the number of correct translations within 90 s (Jaeger, 2018). In the WFT, participants were asked to generate words starting with the letters F, A, and S at 1-min intervals, excluding proper nouns, and their score was the number of words generated (Pendleton et al., 1982). The global cognition z-score was calculated as an average of the three individual z-scores at each study visit and standardized using the global z-score mean and standard deviation (SD) of Visit 2. In addition, one of the magnetic resonance imaging (MRI) sub-studies, five other neurocognitive tests were performed: logical memory test, incidental learning, animal naming score, and trail making tests A and B. In the factor analysis, the ARIC cohort was treated as a single group, and the global cognition, language domain, memory domain, and executive functioning domain factor scores were obtained.

Major adverse cardiac events were defined as a definite or probable stroke, myocardial infarction, or definite coronary heart disease (CHD) death that occurred after cognitive testing.

Baseline demographic, lifestyle, and clinical characteristics were obtained in Visit 2. Education was divided into basic education (less than high school), secondary education (high school), and higher education (college). Self-reported income was divided into high (>$35,000), middle ($16,000–35,000), and low (<$16,000) incomes. Self-reported drinking was divided into current, former, and never. Self-reported smoking was divided into current, former, and never smokers. Body mass index was defined as weight in kilograms divided by the height squared. The sitting blood pressure was measured twice with a sphygmomanometer, and the average value of the two measurements was recorded. Hypertension was defined as self-reported antihypertensive drug use or blood pressure ≥140/90 mmHg. Diabetes was defined as fasting blood glucose ≥126 mg/dL, non-fasting blood glucose ≥200 mg/dL, or self-reported antidiabetic drugs. Fasting blood glucose was collected at baseline and measured using the modified hexokinase or glucose-6-phosphate dehydrogenase method. History of CVD was defined as myocardial infarction (self-reported or diagnosed by electrocardiographic changes), revascularization, or hospitalization for heart failure at or before Visit 4. The definition of chronic obstructive pulmonary disorder (COPD) was based on the doctor’s self-reported diagnosis or obstructive vital capacity measurement model, which was defined as the ratio of forced expiratory volume in one second (FEV1)/FVC <0.70. Total cholesterol was determined using an enzymatic method, and high-density lipoprotein cholesterol levels were measured after dextran-magnesium precipitation of non-high-density lipoprotein particles.

The participants were categorized into quartiles (Q1, Q2, Q3, and Q4) based on the global cognition z-score and factor score. Parametric and nonparametric continuous variables are reported as the mean ± SD and median (25th and 75th percentiles) and compared using analysis of variance and Mann–Whitney U test, respectively. Categorical variables are reported as frequencies and percentages, and compared using a Chi-squared test.

For the prospective analyses, the main outcomes were increased hs-CTnT (≥14 ng/L) and NT-proBNP (≥300 pg/mL) levels at Visit 4. The logistic regression model was used to assess the relationship between the categorical baseline cognitive function and incidence of elevated hs-CTnT and NT-proBNP. To further determine whether these relationships were independent risk factors, the model was adjusted for age, sex, center-race, education (<high school, high school, or >high school), annual household income (<$16,000, $16,000–25,000, $25,000–35,000, $35,000–50,000, or >$50,000), smoking (never, former, and current), drinking (never, former, and current), body mass index, systolic blood pressure, heart rate, total cholesterol, triglycerides, high-density lipoprotein, hypertension, diabetes, and chronic obstructive pulmonary disease. In addition, subgroup analyses were stratified by age (<54 versus ≥54 years), sex, race (black versus white), hypertension (yes versus no), and diabetes mellitus (yes versus no).

To explore the indirect effect of subclinical CVD determined by the global cognition z-score on MACE, path analysis was established by structural equation modeling (Stein et al., 2017). The results of the path analysis were calculated using standardized regression coefficients (β) to show the direct and indirect effects on MACE.

A two-tailed P < 0.05 was considered significant for all tests. All statistical analyses were performed using SPSS version 26.0 (IBM Corp, Armonk, NY, United States), and R software 3.5.0 (Vienna, Austria).

ARIC data are available through NIH NHLBI-sponsored Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC) at https://biolincc.nhlbi.nih.gov.

Among the 9,216 and 9,097 subjects included in our prospective, there were 558 (6.1%) and 447 (5.0%) participants with incident elevated hs-CTnT (≥14 ng/L) and NT-proBNP (≥300 pg/mL), respectively, at Visit 4. Baseline (1991–1993) characteristics are described and compared in Table 1. Compared to the population with lower cardiac biomarkers, participants with elevated hs-CTnT (≥14 ng/L) and NT-proBNP (≥300 pg/mL) were older, more likely to be female, African American, and smoker, complicated with hypertension and diabetes, have lower education, and higher concentrations of cardiovascular risk markers in midlife (P < 0.05 for all comparisons). Similarly, patients with lower global cognition z-scores had more cardiovascular risk factors described above (P < 0.05 for all, Supplementary Table 1).

Additionally, participants with elevated hs-CTnT (≥14 ng/L) had significant lower global cognition z-scores; DSST, WFT, and DWRT scores; global cognition factor score; and language, memory, and executive functioning domain factor scores (P < 0.001 for all comparisons). As for participants with elevated NT-proBNP (≥300 pg/mL), they only had significant lower global cognition z-scores, DSST score, global cognition factor score, and executive functioning domains factor score (P < 0.001 for all comparisons).

For individual tests of cognitive function, risk-factor adjusted odds ratio (OR) for elevated hs-CTnT (≥14 ng/L) and NT-proBNP (≥300 pg/mL) levels are presented in Table 2. Comparing Q1 and Q4 of global cognition z-scores and factor scores, a higher incidence of elevated hs-CTnT [OR = 1.511, 95% confidence interval (CI): 1.093–2.088, P = 0.013; OR = 1.564, 95% CI: 1.123–2.177, P = 0.008, respectively) was observed. Impaired cognitive function was also associated with NT-proBNP levels ≥300 pg/mL (global cognition z-score: OR = 1.929, 95% CI: 1.350–2.755, P < 0.001; global cognition factor score: OR = 1.506, 95% CI: 1.050–2.158, P = 0.026) but not with NT-proBNP levels ≥100 pg/mL (data not shown). Complete risk-factor-adjusted logistic models for each cognitive test are presented in Supplementary Table 2. Significant associations of DWRT, DSST, and memory and executive functioning domain factor scores with elevated hs-CTnT were observed. Significant associations of DSST score and executive functioning domain factor score with elevated NT-proBNP were observed.

We found similar results of risk-factor adjusted ORs when we stratified individuals by age (<54 versus ≥54 years), sex, hypertension (yes versus no), and diabetes mellitus (yes versus no). There were no significant interactions between these risk factors, cognition scores, and elevated cardiac biomarkers (Tables 3, 4). Furthermore, impaired cognitive function was not associated with elevated cardiac biomarkers for African Americans, and a significant interaction between race, cognition scores, and elevated cardiac biomarkers was observed (P < 0.05).

The global cognition z-score was significantly associated with both elevated cardiac biomarkers and long-term MACE by path analysis (P < 0.001). In structural equation modeling, the indirect effects of global cognition z-scores on myocardial injury (hs-cTnT) and cardiac strain or dysfunction (NT-proBNP) were 30.0% (β coefficient, 0.125) and 12.1% (β coefficient, 0.087), respectively (Figure 2).