Association Between Markers of Atrial Cardiopathy and Cognitive Impairment Risk Among Adults with No Suggestive History of Atrial Fibrillation

Abstract

Background

Clinical atrial fibrillation (AF) is often preceded by substantive structural and electrical remodelling, termed atrial cardiopathy. This cohort study aims to investigate associations between markers of atrial cardiopathy and the risk of cognitive impairment among aging adults without a suggestive history or documentation of AF.

Methods

This analysis included 21,856 participants from the REasons for Geographic and Racial Differences in Stroke (REGARDS) study who were free of AF and cognitive impairment (2003-2007). Markers of atrial cardiopathy included atrial premature complexes (APCs) and N-terminal pro-brain natriuretic peptide (NT-proBNP). Cognitive status was measured using the 6-item screener. Cox proportional hazards regression was used to examine the association between markers of atrial cardiopathy and risk of cognitive impairment (primary outcome) and stroke during the follow-up period (secondary outcomes), separately.

Results

During a mean follow-up of 9.9 years, 6058 participants had incident cognitive impairment, and 876 had a stroke. APCs increased the hazard of cognitive impairment by 8% (hazard ratio [HR], 1.08, 95% confidence interval [CI]: 1.01-1.15) and stroke by 10% (HR, 1.10, 95% CI: 1.04-1.17). The highest NTpro-BNP-level quintile ( > 235.2 pg/mL) increased the hazard of cognitive impairment by 53% (HR, 1.53, 95% CI: 1.36-1.72) and stroke by 54% (HR, 1.54, 95% CI: 1.39-1.71).

Conclusions

Markers of left atrial cardiopathy, including presence of APCs and a high level of NT-proBNP, may be associated with an increased risk of cognitive impairment and stroke with no suggestive history or documentation of AF.

Atrial fibrillation (AF) is the most common cardiac arrhythmia and a leading cause of cardiovascular disease and death globally, independently increasing the risk of stroke and dementia.1, 2, 3, 4 If AF is detected, the risk of AF-related stroke and dementia can be reduced significantly with oral anticoagulation, a treatment with established efficacy for stroke and dementia prevention in AF.⁵ Despite the importance of detecting AF for early treatment initiation, AF is often paroxysmal or clinically silent, limiting the utility of current risk-prediction algorithms that are based on a documented diagnosis of clinical AF for many individuals.^6,7

Clinical AF often is preceded by structural abnormality and electrical left atrial remodelling, termed left atrial cardiopathy (AC).⁸ The prevalence of atrial cardiopathy is substantial in patients with embolic stroke of unknown sources.⁹ Previous evidence demonstrated that markers of AC, including increased P-wave terminal in lead V1, left atrial enlargement, and excessive atrial ectopy, are associated with an increased risk of AF.10, 11, 12 In addition, growing evidence indicates that AC in the absence of AF is itself a direct risk factor for stroke, and potentially is associated with dementia, and these markers improve the predictive utility of existing stroke prediction tools (Congestive Heart Failure, Hypertension, Age ≥ 75 years, Diabetes, Stroke/Transient Ischemic Attack, Vascular Diseases, Age 65 to 74 years, Sex Category [CHA2DS2-VASc]) in those without AF.13, 14, 15, 16 Under this new broader model of AF-related stroke, both AF and stroke are potential results of an underlying AC,¹⁷ suggesting that AC might represent an important risk marker to identify individuals at risk of AF, stroke, and/or cognitive decline.

Currently, evidence on whether markers of AC are independently associated with cognitive decline in the absence of AF—and what these markers are—remains limited. Due to the data availability, this study focused on only 2 markers of AC—atrial premature complexes (APCs) and N-terminal pro-brain natriuretic peptide (NT-proBNP). Recently, emerging evidence demonstrated that APCs and NT-proBNP are also important markers to identify high-risk individuals with AC.10, 11, 12^,18 APCs are a common type of arrhythmia characterized by premature heartbeats within the atria, and they have been associated with incident AF and other adverse outcomes, including stroke and all-cause mortality.^19,20 Previous systematic reviews found that premature atrial contractions were associated with a 2-to-4-fold increased risk of incident AF, depending on different exposure measurements and confounders adjusted.^21,22 Previous studies^21,22 have reported a 2-fold to almost 3-fold increased risk of stroke among individuals with APCs, suggesting that presence of APC may be an alternative indicator of AC and that investigations of APC may provide valuable insights into the prevention of adverse outcomes. NT-proBNP is a serum biomarker released by the myocardium, and an elevated level of NT-proBNP has been found in heart failure, structural heart disease, AF, and ventricular strain.^18,23 NT-proBNP is not only a ventricular marker but also an important indicator of atrial dysfunction,^24,25 and it can predict AF,^26,27 subclinical cerebrovascular disease and stroke,^28,29 and adverse anatomic left atrial remodelling and dysfunction.^24,25 Population studies reported that every 1000 pg/mL increase in NT-proBNP was associated a 16% increase in the risk of AF in individuals with heart failure, and this marker had moderate prediction performance for incident AF, with a C-statistic of 0.67.³⁰ The purpose of this study was to investigate associations between the 2 markers of AC (presence of APCs and NT-proBNP level) and the risk of cognitive decline in adults with no suggestive history or documentation of AF.

Methods

Study cohort

This study utilized data from the population-based prospective REGARDS (REasons for Geographical and Racial Differences in Stroke) cohort. REGARDS enrolled 30,239 adults aged ≥ 45 years between January 2003 and October 2007 and follows-up with them every 6 months by computer-assisted telephone interview (CATI) to examine regional and racial differences in incident stroke.^2,31,32 Demographic information and medical histories were obtained via baseline CATI, and medication history, blood samples, and a resting electrocardiogram (ECG) were collected by trained staff during the subsequent in-home visit.³³ From 2013-2016, participants who were still active completed Exam 2 (a follow-up CATI and/or in-home examination) to collect incident risk factors and clinical outcomes.³³ All participants provided written informed consent, and the study was approved by all participating institutional review boards. To abide by its obligations with the National Institutes of Health / National Institute of Neurological Disorders and Stroke and the Institutional Review Board of the University of Alabama at Birmingham, REGARDS facilitates data sharing through formal data use agreements. Any investigator is welcome to access the REGARDS data through this process. Requests for data access may be sent to regardsadmin@uab.edu.

For the present study, we identified a subcohort of adults with no prior AF or screening risk of cognitive impairment (defined as a 6-item screener (SIS) score < 5³⁴) who had baseline APC measurements, further excluding those with complete atrioventricular block (defined using Minnesota code 6.1), or ventricular ectopic beats (defined using Minnesota code 8.2), and participants with < 2 cognitive outcome measurements. This process resulted in 21,856 participants (Fig. 1). Among the 21,856, a total of 6335 had an NT-proBNP measurement from their stored baseline blood sample. Applying the same inclusion and exclusion criteria, we conducted a secondary analysis of 4492 participants who had the NT-proBNP measurements.

Figure 1.

Cohort description. APC, atrial premature complex; NT-proBNP, N-terminal pro-brain natriuretic peptide.

Study exposures

Markers of AC included APCs and NT-proBNP level. Baseline APCs were detected on 12-lead ECG and defined by Minnesota code criteria (Minnesota codes 8.1.1, 8.1.3).^33,35 More details on NT-proBNP measurement can be found in previous literature.³⁶ APCs were dichotomized as either present or absent. Data on continuous APCs were not available. NT-proBNP level was measured using an electrochemiluminescence immunoassay.³⁶ NT-proBNP was a continuous variable, and an elevated NT-proBNP level was defined as that measured in those ranked in the top NT-proBNP quintile.¹⁸

Outcomes

Individuals at high risk of cognitive impairment were identified using an SIS score over a follow-up period from 2003-2019.³⁴ The SIS is a measurement of global cognitive function and consists of a 3-item recall and 3-item temporal orientation evaluation.³⁴ The SIS score has been shown previously to be reliable for identifying subjects with cognitive impairment. Compared to the gold-standard clinical assessment—defined as cognitive assessments and neurologic examination—the SIS showed a sensitivity of 74.2% and a specificity of 80.2% in predicting cognitive impairment in the community setting, and a sensitivity of 84% and specificity of 85.3% in the clinical setting.³⁴ Additionally, an observational study evaluating the performance of SIS in the clinical setting assessed the score of SIS and cognitive impairment recorded in a patient’s chart, and it demonstrated that individuals with poor SIS scores were more likely to have cognitive impairment (sensitivity 86%, specificity 77%).³⁷ The SIS was administered by telephone. Starting in December 2003, participants’ cognitive function was measured annually using the SIS (score range, 0-6).³⁴ Individuals with a shift from an intact cognitive screening status (score of 5 or 6) at baseline to an impaired cognitive screening status (a score of ≤ 4) were considered to be at high risk of cognitive impairment. The secondary outcome was a diagnosis of stroke during follow-up (either incident or recurrent stroke). Medical records were requested and examined for all suspected strokes, including those identified via interviews with proxies after a report of death. A team of stroke experts reviewed the records to validate and classify potential strokes, using published guidelines.^32,38

Covariates

Age, sex, race, exercise, smoking status, and alcohol use were self-reported. Exercise was defined based on participants’ response to the following question: How many times per week do you engage in intense physical activity?³⁹ Smoking and alcohol use were coded as ever (ie, current and former) or never. Comorbidities included diabetes, hypertension, hyperlipidemia, left ventricular hypertrophy, and a history of myocardial infarction, stroke, or heart diseases. Diabetes was defined as a fasting glucose level ≥ 126 mg/dL, a nonfasting glucose level ≥ 200 mg/dL, or self-reported diabetes medication use.^32,36 Hypertension was defined as systolic blood pressure ≥ 140 mm Hg and diastolic blood pressure ≥ 90 mm Hg or self-reported antihypertensive medication use, in accordance with previous REGARDS literature.^32,36 Hyperlipidemia was defined as a total cholesterol level of ≥ 240 mg/dL, a low-density lipoprotein cholesterol level of ≥ 160 mg/dL, a high-density lipoprotein cholesterol level of ≤ 40 mg/dL, or self-reported antihyperlipidemic medication use.^32,36 Use of statin, antiplatelets, or antidepressant medications were self-reported. Myocardial infarction and coronary artery disease were self-reported or were diagnosed using ECG data. Left ventricular hypertrophy was measured using ECG data and was defined using the Sokolow-Lyon criteria.³³

Incident AF and stroke during the follow-up period

Incident AF and stroke during the follow-up period are considered potential intermediators. Incident AF was identified based on results of an ECG and self-reported history at the Exam 2.³³

Statistical analysis

Descriptive statistics were used to characterize the study cohort. All demographic and clinical variables were compared with χ² tests or t-tests between exposure categories. Cumulative incidence curves were generated to estimate the adjusted incidence rate of cognitive impairment for those with APCs present vs absent and who were in the fifth NT-proBNP quintile ( > 235.2 pg/mL) vs other NT-proBNP quintiles.

Cox proportional hazards models were used to estimate cause-specific hazard for primary (incident cognitive impairment) and secondary outcome (stroke during follow-up) with the presence of APCs, and to estimate associations of elevated NT-proBNP level with the outcomes for the stratified random cohort.^40,41 All models were adjusted for death as a competing risk and for demographic and clinical comorbidities, including age, sex, race, exercise, smoking status, alcohol use, history of hypertension, diabetes, coronary artery diseases, myocardial infarction, stroke, left ventricular hypertrophy, heart failure, and medication status, including baseline use of antiplatelet, statin, or antidepressant drugs. A similar model was performed to estimate cause-specific hazards for the primary and secondary outcome for each 1 unit increase in log NT-proBNP level. To determine whether the magnitude of the association of markers with risk of cognitive impairment differs by sex or race, stratified models were conducted to examine effect modification by sex (male and female) and race (White and Black). To examine the impacts of unmeasured confounders on the estimated effect, the E-value was examined. The E-value has been used to quantify the strength of association between unmeasured confounding, the exposure, and the outcome of interest that would explain away the observed association in observational studies.⁴² The E-value was calculated using the following equation: E-value = risk ratio + √risk ratio (risk ratio – 1); a larger E-value suggests a robust association between the exposure and the outcome.⁴²

Sensitivity analysis

In a sensitivity analysis, participants with incident AF or stroke during follow-up prior to cognitive impairment were excluded, as incident AF and stroke may be potential mediators. Sex and age were entered as an interaction term to determine if an interaction of sex and age with the exposure variables for the outcomes of cognitive impairment and stroke during the follow-up period. Similarly, a sensitivity analysis also was performed, with presence of APCs and NT-proBNP level entered as an interaction term to examine whether an interaction between these markers was present that affected the outcomes. Further, a sensitivity analysis was conducted to test the association between markers of AC and outcome defined as an SIS score of ≤ 3.

Statistical significance for all analyses was 2-sided, with P < 0.05. All analyses used R software, version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Study cohort and characteristics

Demographic and clinical characteristics for those in each exposure category for APCs and NT-proBNP are presented in Table 1. Individuals with APCs and an NT-proBNP level in the highest quintile ( > 235.2 pg/mL) were significantly older than those without; they also had significantly more comorbidities, including myocardial infarction, coronary diseases, diabetes, hypertension, and heart failure, and they had more use of antiplatelets (Table 1). Demographic and clinical characteristics for those across different NT-proBNP quintiles are presented in Supplemental Table S1.

Table 1.

Cohort characteristics

Variables	APCs absent (N = 20,048)	APCs present (N = 1808)	P	NT-proBNP 1st–4th quintile (N = 3593)	Highest NT-proBNP-level quintile (N = 899)	P
Age, y	63.5 ± 9.0	68.4 ± 9.5	< 0.001	64.6 ± 10.5	72.4 ± 9.8	< 0.001
Females	11,459 (57.2)	882 (48.8)	< 0.001	1797 (50.0)	480 (53.4)	0.08
Race, White	12,348 (61.6)	1042 (57.6)	< 0.001	2003 (55.7)	554 (61.6)	0.002
Exercise, times per wk
1–3	7477 (37.3)	663 (36.7)	0.62	1302 (36.2)	270 (30.0)	< 0.001
≥ 4	5920 (29.5)	540 (29.9)	0.78	1092 (30.4)	229 (25.5)	0.004
Smoke, current or past	10,617 (53.0)	994 (55.0)	0.11	1967 (54.7)	471 (52.4)	0.22
Alcohol, current or past	14,176 (70.7)	1250 (69.1)	0.17	2527 (70.3)	562 (62.5)	< 0.001
Medical history at baseline
Myocardial infarction	1983 (9.9)	218 (12.1)	0.004	311 (8.7)	242 (26.9)	< 0.001
Coronary artery disease	2830 (14.1)	318 (17.6)	< 0.001	463 (12.9)	321 (35.7)	< 0.001
Diabetes	3807 (19.0)	404 (22.3)	< 0.001	794 (22.1)	300 (33.4)	< 0.001
Hypertension	11,195 (55.8)	1172 (64.8)	< 0.001	2125 (59.1)	681 (75.8)	< 0.001
Dyslipidemia	11,289 (56.3)	1006 (55.6)	0.60	2083 (58.0)	553 (61.5)	0.06
Stroke	949 (4.7)	112 (6.2)	0.006	192 (5.3)	97 (10.8)	< 0.001
Heart failure	2830 (14.1%)	318 (17.6%)	< 0.001	463 (12.9)	321 (35.6%)	< 0.001
Medication use
Statin	6029 (30.1)	582 (32.2)	0.07	1,025 (28.5)	387 (43.0)	< 0.001
Antiplatelets	798 (4.0)	103 (5.7)	< 0.001	135 (3.8)	105 (11.7)	< 0.001
Antidepressant	2474 (12.3)	172 (9.5)	< 0.001	408 (11.4)	124 (13.8)	0.05
Left ventricular hypertrophy	582 (2.9)	59 (3.3)	0.42	87 (2.4)	65 (7.2)	< 0.001

Values are mean ± standard deviation, or n (%), unless otherwise indicated.

APC, atrial premature complex; NT-proBNP, N-terminal pro-brain natriuretic peptide.

Markers of left AC, cognitive impairment, and stroke during the follow-up period

During a mean follow-up period of 9.9 years (median follow-up period: 11.3 years), 6058 participants had incident cognitive impairment, and 876 had a stroke (either a first or a recurrent stroke). A Kaplan-Meier estimator of the cumulative incidence function showed that the cumulative incidence was increased for those with APCs and an elevated NT-proBNP level (Fig. 2). Individuals with APCs had significantly higher cumulative incidence rates for cognitive impairment (45.1, 95% confidence interval [CI]: 41.5-48.8 vs 32.9, 95% CI: 32.0-33.8; P < 0.001) and stroke during follow-up (6.0, 95% CI: 4.9-7.3 vs 3.9, 95% CI, 3.7-4.2; P < 0.001; Table 2). Similarly, individuals in the highest NT-proBNP level quintile had significantly greater incidence rates of cognitive impairment (68.0, 95% CI: 60.9-75.7 vs 44.2, 95% CI: 41.7-46.8; P < 0.001) and stroke during follow-up (22.1, 95% CI: 18.4-26.2 vs 13.7, 95% CI: 12.5-15.1; P < 0.001; Table 2). Incident rates of cognitive impairment and stroke during follow-up across different NT-proBNP quintiles are presented in Supplemental Table S2.

Figure 2.

Adjusted cumulative incidence (95% confidence intervals) for cognitive impairment for (A) atrial premature complexes (APCs) present vs absent, and (B) elevated N-terminal pro-brain natriuretic peptide (bnp) level and non-elevated N-terminal pro-brain natriuretic peptide level.

Table 2.

Incidence rates (per 1000 person-years [PY]) of cognitive impairment and stroke during follow-up period

Outcomes	APCs absent (N = 20,048)	APCs present (N = 1808)	P	NT-proBNP 1st–4th quintile (N = 3593)	Highest NT-proBNP quintile (N = 899)	P
Cognitive impairment
Rate 1000 PY (95% CI)	32.9 (32.1–33.8)	45.1 (41.5–48.8)	< 0.001	44.2 (41.7–46.8)	68.0 (60.9–75.7)	< 0.001
Total event count	5464	596		1195	334
Total PY	166,032.37	13,222.88		27,931.69	4914.52
Stroke during follow-up period
Rate 1000 PY (95% CI)	3.9 (3.7–4.2)	6.0 (4.9–7.3)	< 0.001	13.7 (12.5–15.1)	22.1 (18.4–26.2)	< 0.001
Total event count	778	97		441	130
Total PY	197,682.09	16,172.34		32,142.07	5890.05

APCs, atrial premature complexes; CI, confidence interval. NT-proBNP, N-terminal pro-brain natriuretic peptide.

Competing risks Cox proportional hazards analyses indicated that, after adjustment for age, sex, race, exercise, smoking status, alcohol use, hypertension, diabetes, a history of stroke at baseline, hyperlipidemia, myocardial infarction, coronary artery diseases, left ventricular hypertrophy, and baseline medication status (statin, antiplatelet, antidepressant), and adjusting for all-cause death as a competing risk, presence of APCs increased the hazard of cognitive impairment by 8% (hazard ratio [HR], 1.08, 95% CI: 1.01-1.15; Table 3). Having an NT-proBNP level in the highest quintile increased the hazard of cognitive impairment by 53% (HR, 1.53, 95% CI: 1.36-1.72). Compared with individuals with an NT-proBNP level in the lower quintiles, individuals in the highest quintile showed increased hazards of cognitive impairment (HR, 1.64; 95% CI: 1.41-1.91). Each log NT-proBNP increase increased the cause-specific hazard by 19% (HR, 1.19; 95% CI: 1.14-1.24). When models were stratified by sex and race, the effects of APCs on cognitive impairment were only significant in White participants. Effects of NT-proBNP level on cognitive impairment remained stable in stratified models. The E-values are 1.37 for the observed association between APCs and cognitive impairment, and 2.43 for the observed association between NT-proBNP level and cognitive impairment, suggesting a robust link between these exposures and cognitive impairment.

Table 3.

Adjusted hazard ratios (95% confidence intervals) for incident cognitive impairment (primary outcome)

	All participants (N = 21,856)	Female participants (N = 12,341)	Male participants (N = 9515)	Whites (N = 13,390)	Blacks (N = 8466)
APCs present vs absent	1.08 (1.01–1.15)	1.06 (0.97–1.16)	1.08 (0.97–1.20)	1.12 (1.03 –1.22)	1.02 (0.92–1.14)

	All participants (N = 4492)	Female participants (N = 2277)	Male participants (N = 2215)	Whites (N = 2557)	Blacks (N = 1935)
NT-proBNPhighest quintilevs1st–4th quintile	1.53 (1.36–1.72)	1.32 (1.14–1.54)	2.03 (1.66–2.49)	1.43 (1.23–1.67)	1.69 (1.38–2.06)
NTpro-BNP quintile
2nd vs 1st	0.98 (0.86–1.12)	1.01 (0.84–1.21)	0.94 (0.77–1.13)	1.03 (0.87–1.23)	0.93 (0.76–1.13)
3rd vs 1st	1.08 (0.94–1.23)	1.25 (1.04–1.50)	0.85 (0.68–1.05)	1.08 (0.91–1.29)	1.09 (0.89–1.35)
4th vs 1st	1.16 (1.00–1.33)	1.22 (1.01–1.48)	1.10 (0.87–1.39)	1.04 (0.86–1.26)	1.38 (1.10–1.73)
5th vs 1st	1.64 (1.41–1.91)	1.52 (1.24–1.87)	1.97 (1.55–2.50)	1.50 (1.22–1.83)	1.82 (1.43–2.32)
Log NTpro-BNP	1.19 (1.14–1.24)	1.17 (1.11–1.24)	1.21 (1.13–1.30)	1.14 (1.07–1.21)	1.23 (1.16–1.31)

The model was adjusted for age, race, sex, exercise, smoking, alcohol use, baseline medication use, diabetes, hypertension, and history of myocardial infarction, heart failure, hyperlipidemia, stroke, and left ventricular hypertrophy.

APC, atrial premature complex; NT-proBNP, N-terminal pro-brain natriuretic peptide.

Results for secondary outcomes are shown in Table 4. After adjusting for baseline covariates and all-cause death as a competing risk, having APCs and an NT-proBNP level in the highest quintile increased the hazard of stroke during follow-up by 10% (HR, 1.10, 95% CI: 1.04-1.17) and 54% (HR, 1.54, 95% CI: 1.39-1.71), respectively. Compared with individuals with anNT-proBNP level in the lowest quintiles, individuals with a level in the highest NT-proBNP quintile had a 64% increased hazard of stroke (HR, 1.64, 95% CI: 1.44-1.88). Each log NT-proBNP level increase increased the cause-specific hazard of stroke by 17% (HR, 1.17; 95% CI: 1.13-1.22). Stratified models showed that the effects of these markers on stroke were stable, except in White participants.

Table 4.

Adjusted hazard ratios (95% confidence intervals) for stroke during the follow-up period (secondary outcome)

	All participants (N = 21,856)	Female participants (N = 12,341)	Male participants (N = 9515)	Whites (N = 13,390)	Blacks (N = 8466)
APCs present vs absent	1.10 (1.04–1.17)	1.10 (1.01–1.18)	1.09 (1.00–1.20)	1.15 (91.06–1.24)	1.05 (0.96–1.15)

	All participants (N = 4492)	Female participants (N = 2277)	Male participants (N = 2215)	Whites (N = 2557)	Blacks (N = 1935)
NT-proBNP highest quintile vs1st–4th quintile	1.54 (1.39–1.71)	1.35 (1.19–1.55)	1.98 (1.66–2.35)	1.52 (1.33–1.75)	1.58 (1.34–1.87)
NT-proBNP quintile
2nd vs 1st	0.97 (0.86–1.09)	0.90 (0.76–1.06)	0.98 (0.83–1.16)	1.00 (0.84–1.17)	0.94 (0.79–1.12)
3rd vs 1st	1.09 (0.97–1.23)	1.10 (0.93–1.30)	0.99 (0.82–1.20)	1.18 (1.00–1.40)	1.01 (0.84–1.21)
4th vs 1st	1.14 (1.01–1.30)	1.09 (0.92–1.30)	1.14 (0.93–1.40)	1.09 (0.91–1.30)	1.25 (1.03–1.51)
5th vs 1st	1.64 (1.44–1.88)	1.40 (1.17–1.69)	2.04 (1.65–2.53)	1.65 (1.37–2.00)	1.65 (1.34–2.02)
Log NTpro-BNP	1.17 (1.13–1.22)	1.15 (1.09–1.21)	1.21 (1.14–1.28)	1.17 (1.10–1.24)	1.18 (1.12–1.25)

The models were adjusted for age, race, sex, exercise, smoking, alcohol use, baseline medication use, diabetes, hypertension, and history of myocardial infarction, heart failure, hyperlipidemia, stroke, and left ventricular hypertrophy.

APC, atrial premature complex; NT-proBNP, N-terminal pro-brain natriuretic peptide.

In a sensitivity analysis in which incident AF and stroke during follow-up were excluded, the effects of APCs on cognitive impairment remained significant (HR, 1.14; 95% CI: 1.02-1.27; Table 5). Sensitivity analyses showed no significant interaction between sex and age or between APCs and NT-proBNP level for any of the models. Results remained consistent when the outcome was defined as an SIS score of ≤ 3 (Table 6).

Table 5.

Adjusted hazard ratios (95% confidence intervals) for incident cognitive impairment in sensitivity analysis in which participants with incident atrial fibrillation or stroke during the follow-up period were excluded

	All participants (N = 10,231)	Female participants (N = 5875)	Male participants (N = 4356)	Whites (N = 6531)	Blacks (N = 3700)
APCs present vs absent	1.14 (1.02-1.27)	1.09 (0.95-1.25)	1.25 (1.04-1.51)	1.14 (0.99-1.32)	1.10 (0.92-1.31)

	All participants (N = 1700)	Female participants (N = 877)	Male participants (N = 823)	Whites (N = 1008)	Blacks (N = 692)
NT-proBNP highest quintile vs 1st–4th quintile	1.04 (0.79–1.35)	1.06 (0.77–1.46)	1.07 (0.61–1.89)	1.03 (0.75–1.42)	0.92 (0.52–1.62)

The models adjusted for age, race, sex, exercise, smoke, alcohol use, baseline medication use, diabetes, hypertension, and history of myocardial infarction, heart failure, hyperlipidemia, stroke, and left ventricular hypertrophy.

APC, atrial premature complex; NT-proBNP, N-terminal pro-brain natriuretic peptide.

Table 6.

Adjusted hazard ratios (95% confidence intervals) in sensitivity analysis in which the outcome was defined as 6-item screener score of ≤ 3

	All participants (N = 10,231)	Female participants (N = 5875)	Male participants (N = 4356)	Whites (N = 6531)	Blacks (N = 3700)
APCs present vs absent	1.13 (1.03–1.25)	1.12 (0.99–1.26)	1.15 (0.98–1.35)	1.13 (0.98–1.32)	1.12 (0.99–1.27)

	All participants (N = 1700)	Female participants (N = 877)	Male participants (N = 823)	Whites (N = 1008)	Blacks (N = 692)
NT-proBNP highest quintile vs 1st–4th quintile	1.07 (0.81–1.33)	1.16 (0.81–1.50)	1.09 (0.79–1.68)	1.08 (0.73–1.34)	0.96 (0.63–1.59)

The models were adjusted for age, race, sex, exercise, smoking, alcohol use, baseline medication use, diabetes, hypertension, and history of myocardial infarction, heart failure, hyperlipidemia, stroke, and left ventricular hypertrophy.

APC, atrial premature complex; NT-proBNP, N-terminal pro-brain natriuretic peptide.

Discussion

This study examined the associations between markers of AC (presence of APCs and NT-proBNP level) and risk of cognitive impairment in adults aged ≥ 45 years with no suggestive history or documentation of AF. Results of this study provide novel evidence that individuals with APCs and an elevated NT-proBNP level had significantly higher incidence rates and increased hazards of cognitive impairment and stroke during the follow-up period, adjusting for all-cause death as a competing risk. Although several prior studies have shown that markers of AC are associated with adverse cognitive outcomes, the design of these studies was either cross-sectional^19,43 or case-control,⁴⁴ and some did not exclude participants who had documented AF.^44,45 Thus, previous studies cannot establish temporal associations between markers and outcomes. The present study is unique in using data from a longitudinal cohort study to investigate associations between markers of left AC (presence of APCs and an elevated NT-proBNP level) and risk of cognitive impairment in those without known AF. Notably, in the sensitivity analysis in which participants with documented incident AF or stroke during follow-up were excluded, associations between APCs and cognitive impairment remained significant.

Findings of the present study are consistent with increasing evidence for a new mechanistic model of AC, AF, and outcomes.^46,47 Strong evidence from prospective and retrospective studies indicates that AF is associated with cognitive impairment.1, 2, 3^,48,49 Cross-sectional analyses of 316 dementia-free participants from the Atherosclerosis Risk in Communities (ARIC) study reported an association between AC (defined as left atrial volume index ≥ 34 mL/m²) and elevated brain amyloid, after adjusting for AF.⁵⁰ The analyses did not find a similar association in participants with AF,⁵¹ indicating that AC may be an important impetus for thrombi and subsequent embolism without the development of AF.⁵⁰ Similar to evidence that AF may not be a necessary step in the pathogenesis of ischemic stroke,^46,51 findings of the present study suggest that left AC, as a state of structural abnormality and left atrial remodelling,⁵⁰ contributes to the development of cognitive impairment independent of AF. Mechanisms linking left AC and cognitive impairment may be similar to mechanisms of AF-associated cognitive impairment (ie, microembolism, hypoperfusion, inflammation, breakdown of the blood-brain barrier, and subsequent clinical stroke).⁴⁷

Results of this study indicate that APCs and NT-proBNP may be important biomarkers for the early identification (ie, screening) of individuals at higher risk of cognitive impairment, even without clinical manifestations of AF. Given the increasing evidence for associations between AC and cognitive outcomes, markers of AC may represent a potential therapeutic target for treatment (ie, anticoagulation) before the onset of AF.¹⁵ Notably, the ongoing ARCADIA (Atrial Cardiopathy and Antithrombotic Drugs in Prevention After Cryptogenic Stroke) trial is the first to test oral anticoagulation for secondary risk reduction in patients with cryptogenic stroke and no prior AF, using AC as a selection criterion.⁵² Additionally, results of this study have potential implications for the identification of target candidates for preventive therapies to mitigate the risk of cognitive impairment, which is an important cause of disability and is significantly associated with reduced quality of life.^53,54 Although clinical guidelines make identical recommendations for anticoagulation based on AF type (paroxysmal vs non-paroxysmal), evidence suggests that decision-making on diagnosis and/or treatment needs to consider indications beyond paroxysmal or non-paroxysmal AF.⁵⁵ Current treatment relies on the detection of AF, but AF is intermittent or clinically silent, highlighting the fact that effective widespread screening for clinical AF remains elusive and that new detectable screening targets for atrial disease are needed. Further, commonly used decision algorithms, such as the CHA2DS2-VASc for stroke risk, rely on the confirmation of AF for anticoagulation indication. Future studies are required to confirm the incremental predictive utility that the inclusion of these markers into existing risk tools provides for stroke and dementia risk. Additional trials testing the efficacy of screening and treatment in individuals with markers of AC are also required.

This study has several limitations. First, only a subgroup of participants had NT-proBNP measurement, limiting the statistical power to examine the association between NT-proBNP level and outcomes. Second, this study considered only clinically apparent incident AF and stroke during the follow-up period. Other important time-varying covariates, such as medication status and comorbidities, were not available for inclusion in the analytic models. AF was measured based on self-report or a baseline ECG, and thus, some participants may have undiagnosed AF, underestimating the prevalence of AF in the cohort. In addition, this study may be subject to selection bias, as incident AF was available only for participants who completed the second in-home exam, despite the fact that incident AF was considered in only the sensitivity analysis.⁵⁶ Further, this study considered only 2 markers of AC (APCs and NT-proBNP). Other markers of AC—including left atrial volume index, P-wave terminal force at V1, and excessive atrial ectopy–that have been demonstrated to increase stroke risk57, 58, 59 and may be associated with the risk of cognitive impairment, are not available in the REGARDS dataset. Markers of AC were measured only at baseline in the REGARDS data. Thus, the results may not reflect the associations between time-varying values and outcomes. On the other hand, due to the limited data availability, this study used only the SIS score, a brief cognitive function test, to assess cognitive impairment. Although a brief cognitive test can help determine the degree of cognitive status, a diagnosis of cognitive impairment should be based on other clinical tests also, such as neurologic examinations and brain imaging. To reduce the misclassification bias, a sensitivity analysis was conducted in which the primary outcome was defined using a different cutoff point, and the results were consistent with the primary results. Additionally, data on other cardiac information, such as valvular disease, ECG results, and left ventricular ejection fraction, were not available. Without these important cardiac factors, the study cannot exclude the possibility that other cardiac conditions, which correlate with AC, led to the occurrence of cognitive impairment. Finally, although this study provides insights into associations between markers of AC and cognitive impairment, the results do not imply causality and should be interpreted with caution. Further studies are needed to explore the mechanisms underlying AC and cognitive impairment.

Conclusions

APCs and N-terminal pro-brain natriuretic peptide, both promising prognostic predictors of AC, significantly increase the risk of cognitive impairment in individuals without a documentation of AF. These findings have implications for future research on use of markers of AC to identify individuals with a high risk for cognitive impairment.