Factors Associated With Cognitive Impairment in Heart Failure With Preserved Ejection Fraction

Abstract

Background:

Cognitive impairment is prevalent in heart failure and is associated with higher mortality rates. The mechanism behind cognitive impairment in heart failure with preserved ejection fraction (HFpEF) has not been established.

Objective:

The aim of this study was to evaluate associations between abnormal cardiac hemodynamics and cognitive impairment in individuals with HFpEF.

Methods:

A secondary analysis of Atherosclerosis Risk in Communities (Atherosclerosis Risk in Communities) study data was performed. Participants free of stroke or dementia who completed in-person assessments at visit 5 were included. Neurocognitive test scores among participants with HFpEF, heart failure with reduced ejection fraction (HFrEF), and no heart failure were compared. Sociodemographics, comorbid illnesses, medications, and echocardiographic measures of cardiac function that demonstrated significant (P < .10) bivariate associations with neurocognitive test scores were included in multivariate models to identify predictors of neurocognitive test scores among those with HFpEF. Multiple imputation by chained equations was used to account for missing values.

Results:

Scores on tests of attention, language, executive function, and global cognitive function were worse among individuals with HFpEF than those with no heart failure. Neurocognitive test scores were not significantly different among participants with HFpEF and HFrEF. Worse diastolic function was weakly associated with worse performance in memory, attention, and language. Higher cardiac index was associated with worse performance on 1 test of attention.

Conclusions:

Cognitive impairment is prevalent in HFpEF and affects several cognitive domains. The current study supports the importance of cognitive screening in patients with heart failure. An association between abnormal cardiac hemodynamics and cognitive impairment was observed, but other factors are likely involved.

Heart failure (HF) is a highly prevalent condition that affects nearly 6.2 million Americans older than 20 years.¹ Heart failure with preserved ejection fraction (HFpEF) is a subclass of HF that accounts for up to 55% of those with HF and is characterized by clinical signs of HF, an ejection fraction greater than 50%, and evidence of left ventricular diastolic dysfunction on echocardiogram or cardiac catheterization.² Compared with individuals who have HF with reduced ejection fraction (HFrEF), those with HFpEF are more likely to be older and female, and have multiple chronic comorbid illnesses (hypertension, diabetes mellitus, atrial fibrillation, anemia, and obesity).³ Although survival rates are improving for patients with HF, mortality remains high.^1,4 Cognitive impairment, which is prevalent in HF, may play a role.^5–7

Approximately 40% of patients with HF have cognitive impairment, which is defined as an objective deficit in 1 or more cognitive domains that falls between the extremes of normal cognitive function and dementia.^8–10 Persons with HF most often experience deficits in memory, attention, and executive function.^6,11 A deficit in any one of these domains can impair HF self-care, including the ability to monitor symptoms and respond to changes in symptoms appropriately.⁵ Poor HF self-care can contribute to adverse events, including higher mortality rates.^5,12–16 Although HFpEF accounts for half of all cases of HF, research on cognitive impairment in HFpEF is profoundly limited. Authors of the limited research on cognitive impairment in HFpEF suggest that the prevalence of cognitive impairment in HFpEF and HFrEF is similar and that individuals with HFpEF experience deficits in the same cognitive domains as individuals with HFrEF.^7,17–21 However, authors of most previous research compared the severity of cognitive impairment in patients with HFpEF with those with HFrEF,^7,17–22 so little is known about how cognitive function among those with HFpEF compares with those without HF. In a study of participants in the Atherosclerosis Risk in Communities study cohort, authors revealed that the risk of mild cognitive impairment was 36% higher among participants with HFpEF than those with no HF.²² That article did not include consideration of domain-specific cognitive performance, however, and did not compare neurocognitive performance in participants with HFpEF with those with no HF.²² Furthermore, the mechanisms behind cognitive impairment in HFpEF have not been established, so the factors that contribute to a higher risk of cognitive impairment in HFpEF remain unclear.^22,23

The Conceptual Model of Cognitive Deficits in Heart Failure suggests that impaired cardiac function in HF leads to circulatory insufficiency, cerebral hypoperfusion, and neuronal damage.²³ Left ventricular ejection fraction is proposed as an indicator of cardiac function; however, as left ventricular ejection fraction is normal in HFpEF by definition, this model does not adequately explain how cognitive impairment develops in HFpEF. Several other abnormalities in cardiac function, including diastolic dysfunction and impaired ventricular-vascular coupling, are prevalent in HFpEF and could contribute to cognitive impairment through impaired hemodynamics,²⁴ but mechanistic studies linking these potential etiologies to cognitive impairment are lacking. It also is likely that left ventricular ejection fraction is not a reliable indicator of cardiac performance in HFpEF. Alternative measures of cardiac hemodynamics, such as cardiac index, may be better indicators of cardiac performance, but to our knowledge, there are no previous studies evaluating the association between alternative measures of cardiac performance and cognitive impairment in HFpEF.

Objectives

For the large number of patients with HFpEF, cognitive impairment has the potential to contribute to high morbidity and mortality rates through poor self-care. Understanding the factors and mechanisms that contribute to cognitive impairment in HFpEF will help clinicians identify patients at risk for cognitive impairment and potentially poor self-care. Therefore, the aims of this study were to

1. compare neurocognitive test scores among individuals with HFpEF, with HFrEF, and with no HF;

2. identify associations between neurocognitive test scores and echocardiographic measures of diastolic function in individuals with HFpEF;

3. identify associations between neurocognitive test scores and several echocardiographic measures of cardiac hemodynamics (left ventricular ejection fraction, stroke volume index, cardiac index) in individuals with HFpEF;

4. identify associations between neurocognitive test scores and echocardiographic measures of ventricular-vascular coupling in individuals with HFpEF.

Methods

A secondary analysis of data from visit 5 of the Atherosclerosis Risk in Communities study was performed. The design of the Atherosclerosis Risk in Communities study has been described previously.²⁵ Briefly, participants between the ages of 45 and 64 years in 4 different communities were recruited beginning in 1986 and returned for routine assessments and interviews every few years.^25–27 At visit 5 (2011–2013), an echocardiogram and neurocognitive assessment also was completed on all participants.^26,27 The parent study received approval from the institutional review board at each participating site. This secondary analysis was approved by the institutional review board at New York University Langone Medical Center.

Study Sample

The sample for this analysis was derived from those individuals who completed in-person neurocognitive assessments at Atherosclerosis Risk in Communities visit 5 (n = 6471).²⁶ Individuals with a history of stroke or dementia as determined by expert adjudicated diagnosis (n = 870) were excluded from the current analysis because both conditions have the potential to influence cognitive function. The remaining individuals (n = 5601) in whom ejection fraction could be quantified were categorized as having prevalent HFpEF, HFrEF, or no history of HF at visit 5. Participants were categorized as having HFpEF if they had an adjudicated history of HF based on the presence of HF symptoms (dyspnea, fatigue, exercise intolerance, fluid retention) and an ejection fraction of 50% or greater. Individuals with an adjudicated history of HF and an ejection fraction less than 50% were classified as having HFrEF.^2,28 Individuals who did not have an adjudicated history of HF were classified as “no HF.” Fifty-five participants who had an adjudicated diagnosis of HF but no data on ejection fraction were excluded from further analysis because they could not be classified as having HFpEF or HFrEF. Thirteen participants identified as a race other than white or black and were excluded from the final sample because of low numbers. The final sample for this analysis included 5533 participants.

Sample Size Determination

An estimate of the sample size necessary to support multivariate models was calculated using G*Power version 3.1 and was based on the formula by Cohen.²⁹ To obtain the most conservative estimate, a Bonferroni-adjusted significance level of P = .017 was applied to ensure that the likelihood of a type I error was held to .05 across all related analyses (among tests that evaluated the same cognitive domain, eg, attention). Given the potential scenario in which all 29 predictors would be included in a multiple regression model, sample size calculations were performed assuming all 29 predictors would be included. Using an effect size calculated from previous research (f² = 0.056),³⁰ it was determined that a sample of n = 558 with HFpEF would be necessary to obtain a power of 0.80 in multiple linear regression models with 29 predictors.

Sociodemographics and Clinical Characteristics

Participant-reported sex, race, education, date of birth, and smoking status were collected at Atherosclerosis Risk in Communities visit 1 (1987–1989). Updated smoking status as well as history of depression diagnosis and medications taken in the 4 weeks before neurocognitive assessment (including antianxiety medications, central nervous system–altering drugs, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, β-blockers, loop diuretics, digoxin) were obtained from the participant at the time of the evaluation at Atherosclerosis Risk in Communities visit 5. Vital signs and blood specimens also were evaluated at the time of assessment and were used to define presence of hypertension, diabetes mellitus, and anemia. Blood pressure was measured 3 times in a seated position after a 5-minute rest with a 30-second break between measurements. The average of the 3 measurements was recorded. Hypertension was defined as a systolic blood pressure of 140 mm Hg or greater, a diastolic blood pressure of 90 mm Hg or greater, or currently taking antihypertensive medication.³¹ Blood specimens, including blood chemistry and hematology, were collected and processed following a predefined protocol. History of diabetes was defined as a fasting glucose level of 126 mg/dL or greater, a glycosylated hemoglobin of 6.5% or greater, or currently taking medications for diabetes.³² Presence of anemia was defined as a hemoglobin concentration less than 13 g/dL for men and less than 12 g/dL for women.³³ Score on the Center for Epidemiologic Studies Depression Scale was used to measure presence of depressive symptoms at the time of assessment.^34,35

Echocardiographic Assessment

Details regarding echocardiographic assessment in the Atherosclerosis Risk in Communities parent study, including training and quality assurance protocols, have been described elsewhere.²⁷ In summary, echocardiograms were performed on study participants at the 4 Atherosclerosis Risk in Communities centers using dedicated equipment and predetermined protocols at visit 5.²⁷ Two-dimensional, 3-dimensional, Doppler, and speckle-tracking–based techniques were included.²⁷ Images were stored in digital format on local computers and were transferred to the Echocardiography Reading Center at Brigham and Women’s Hospital in Boston, Massachusetts, for evaluation and interpretation.²⁷ Echocardiographic data on left ventricular geometry, left atrial structure, left ventricular diastolic function, left ventricular systolic function, cardiac hemodynamics, and ventricular-vascular coupling were evaluated in the current analysis.²⁷

Neurocognitive Evaluation

This analysis used the results of the neurocognitive assessment conducted at Atherosclerosis Risk in Communities visit 5. As described elsewhere, participants who attended visit 5 completed a battery of neurocognitive tests.²⁶ The tests and the cognitive domains they evaluate are listed in Table 1. On the Trail Making Test Part A and the Trail Making Test Part B, higher scores reflect worse performance. For all other tests, higher scores reflect better performance. Interviewers who administered in-person neurocognitive tests were trained in advance, with careful quality control measures.^36–38 Raw scores were evaluated in this analysis.

TABLE 1.

Cognitive Domains and Neurocognitive Tests Used in This Secondary Analysis

Cognitive Domain	Definition	Neurocognitive Test(s)
Recent memory	Recall of information immediately following storage	Delayed Word Recall Test
		Logical Memory Test Part I
		Incidental Learning Test
Delayed memory	Recall of stored information following a delay	Logical Memory Test Part II
Attention and processing speed	Ability to focus on a particular stimulus	Digit Span Backwards Test
		Digit Symbol Substitution Test
		Trail Making Test Part A
Language and verbal fluency	Communication using symbols and words	Animal Naming Test
		Boston Naming Test
		Word Fluency Test
Executive function	Goal-directed behavior	Trail Making Test Part B
Global cognitive function	Overall cognitive ability	Mini-Mental State Examination

References 5, 7, 55–58.

Statistical Analysis

Statistical analysis was conducted only on individuals who were able to be classified as having HFpEF, HFrEF, or no HF. Descriptive statistics were calculated to describe sample characteristics. Analysis of variance was conducted on continuous variables to identify statistically significant differences between those with HFpEF, those with HFrEF, and those who are free of HF. χ² Analysis was completed to identify differences between the 3 groups with regard to categorical variables. Hedges’ g was calculated to determine the effect of HFpEF on neurocognitive test scores.³⁹

Consistent with the aim of this study, the following bivariate analyses were conducted on only the sample with HFpEF. Pearson correlations were calculated to identify associations between continuous independent variables (age, echocardiographic measures, etc) and scores on the neurocognitive tests. φ Coefficients were calculated to identify associations between pairs of dichotomous variables. Point-biserial correlations were calculated to identify associations between binary independent variables (comorbid conditions, medications, etc) and scores on neurocognitive tests. Spearman rank correlation was calculated to evaluate associations between education and scores on neurocognitive tests, because education is an ordinal variable.

Multiple linear regression analysis was performed to quantify the magnitude of the effect of cardiac function on scores on neurocognitive tests in the sample with HFpEF. Separate regression models were generated for each neurocognitive test, but the specific variables in each model varied. To create a parsimonious model and maintain statistical power, predictors in multivariate models were selected based on significance of bivariate associations with scores on individual neurocognitive tests. Echocardiographic measures (left ventricular mass index, left atrial volume index, E/A (early transmitral filling velocity/peak atrial transmitral filling velocity) ratio, E/E’ (early transmitral filling velocity/tissue velocity of mitral annulus) ratio, deceleration time, left ventricular ejection fraction, stroke volume index, cardiac index, longitudinal strain, circumferential strain, arterial elastance/end-systolic elastance [Ea/Ees] ratio) that demonstrated significant (P < .10) associations with scores on neurocognitive tests were included in multivariate models. Sociodemographics, comorbid illnesses, and medication classes that demonstrated significant associations with scores on neurocognitive tests were included in multivariate models as potential confounders, because all of these factors have demonstrated an association with cognitive impairment in previous research.^{23,25,35,40–43} The Atherosclerosis Risk in Communities field center from which the participant was recruited was included in the initial analyses. However, because of the high degree of collinearity between Atherosclerosis Risk in Communities field center and race, and because race demonstrated a stronger bivariate association with scores on neurocognitive tests, Atherosclerosis Risk in Communities field center was dropped from the analysis. Between 5% and 10% of the data on echocardiographic measures were missing in the current sample. Therefore, multiple imputation by chained equations was performed to mitigate the effect of missing values on sample power and potential bias. Ten imputations were performed creating 10 complete data sets with imputed values replacing missing values.^44,45 Pooled regression estimates were generated from the imputed data sets by mathematical averaging of the regression estimates from each of the 10 imputed data sets following Rubin’s rules.⁴⁶ To minimize the likelihood of false discovery due to multiple comparisons, a Bonferroni-corrected P value of .017 was used to determine statistical significance. All statistical analyses were conducted using Stata version 15.⁴⁷

Results

Sample Characteristics

Characteristics of the sample (n = 5533) are presented in Table 2. Compared with individuals with no HF (n = 5268), individuals with HFpEF (n = 205) were older (mean age, 77.2 vs 75.4 years; P < .001), less likely to have completed high school or have a college education (P < .001), and more likely to report a history of smoking (P = .001). Participants with HFrEF (n = 60) were older than those with no HF (mean age, 77.6 vs 75.4 years; P = .002) and were more likely to be male (P < .001). When compared with the sample with HFrEF, participants with HFpEF were more likely to be female (P < .001), but no other significant differences were observed between the 2 groups. Chronic comorbid illnesses (hypertension, diabetes mellitus, anemia, atrial fibrillation, depressive symptoms) were more prevalent among those with HFpEF than among those with no HF (P < .001). Anemia (P = .015) and atrial fibrillation (P < .001) were more prevalent among those with HFrEF than those with no HF, but no other significant differences were observed. As expected, history of medications used to treat HF (β-blockers, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, loop diuretics, digoxin) was more prevalent among participants with HF than among those with no history of HF (P < .001).

TABLE 2.

Sociodemographic Characteristics, Comorbid Conditions, Medication History, and Echocardiographic Measures of Sample at Atherosclerosis Risk in Communities Visit 5 (N = 5533)

Characteristic	Classifications	No HF (n = 5268)	HFpEF (n = 205)	HFrEF (n = 60)
Age, y		75.4 (5.1)	77.2 (5.6)a	77.6 (5.4)b
Female		3105 (58.9%)	115 (56.1%)c	13 (21.7%)b
Raced	White	4231 (80.3%)	153 (74.6%)	45 (75.0%)
	Black	1037 (19.7%)	52 (25.4%)	15 (25.0%)
Educationd	Less than high school	647 (12.3%)	49 (23.9%)a	11 (18.3%)
	High school graduate	1791 (34.0%)	76 (37.1%)a	16 (26.7%)
	College/graduate/professional school	2379 (45.3%)	66 (32.2%)a	30 (50.0%)
	Vocational school	442 (8.4%)	14 (6.8%)a	3 (5.0%)
Smoking historyd	Current smoker	297 (5.8%)	8 (3.9%)a	3 (5.0%)
	Former smoker	2430 (47.3%)	121 (59.6%)a	31 (51.7%)
	Never smoked	1995 (38.8%)	53 (26.1%)a	19 (31.7%)
	Unknown/unable to be categorized	416 (8.1%)	21 (10.3%)a	7 (11.7%)
Comorbid illness	Hypertension	3826 (73.3%)	178 (88.6%)a	46 (79.3%)
	Diabetes	1854 (36.1%)	107 (52.5%)a	29 (48.3%)
	Atrial fibrillation	359 (7.2%)	75 (36.8%)a	23 (39.0%)b
	Anemia	1005 (19.6%)	81 (40.5%)a	19 (32.2%)b
	Depressiond	298 (13.9%)	17 (18.3%)	3 (13.0%)
	CES-D score	3.1 (3.2)	4.1 (3.7)a,c	2.9 (2.7)
Medication history	Antianxiety medicationd	369 (7.0%)	29 (14.2%)a	3 (5.0%)
	Antipsychotic medicationd	20 (0.38%)	1 (0.49%)	0
	CNS altering medicationd	379 (7.22%)	29 (14.2%)a	4 (6.7%)
	β-blocker	1634 (31.2%)	148 (72.2%)a	51 (85.0%)b
	ACE inhibitor	1173 (22.4%)	81 (39.5%)a	24 (40.0%)b
	ARB	538 (10.3%)	39 (19.0%)a	12 (20.0%)b
	Loop diuretic	358 (6.8%)	114 (55.6%)a	34 (56.7%)b
	Digoxin	53 (1.0%)	19 (9.3%)a	10 (16.7%)b
Measures of diastolic function	LVMI, g/m2	78.6 (19.1)	94.7 (26.7)a,c	127.4 (42.8)b
	LAVI, mL/m2	25.7 (8.8)	33.7 (12.6)a,c	38.0 (14.1)b
	E/A ratio	0.86 (0.28)	0.94 (0.44)a	1.0 (0.68)b
	E/E’ ratio	10.1 (3.7)	12.8 (6.5)a	13.7 (6.4)b
	E wave deceleration time, ms	204.9 (45.6)	209.0 (58.9)d	188.8 (60.7)b
Measures of cardiac hemodynamics	LVEF, %	65.6 (6.1)	62.7 (6.3)a,c	39.9 (6.7)b
	Stroke volume index, mL/m2 per beat	28.4 (6.2)	30.5 (7.9)a	29.9 (8.0)
	Cardiac index, L/min per m2	1.817 (410)	1.963 (525)a	1.910 (559)
	Longitudinal strain, %	−18.1 (2.4)	−16.5 (2.7)a,c	−11.7 (3.0)b
	Circumferential strain, %	−28.0 (3.7)	−26.2 (4.3)a,c	−16.7 (4.9)b
Ventricular/vascular coupling	Ea/Ees	0.57 (0.15)	0.63 (0.21)a,c	1.0 (0.38)b

Numbers represent mean (SD) for continuous variables and count (%) for categorical variables.

Abbreviations: A, peak atrial transmitral filling velocity; ACE, angiotensin-converting enzyme; ARB, angiotensin receptor blocker; CES-D, Center for Epidemiologic Studies Depression Scale; CNS, central nervous system; E, early transmitral filling velocity; E’, tissue velocity of mitral annulus; Ea, arterial elastance; Ees, end-systolic elastance; HF, heart failure; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; LAVI, left atrial volume index; LVEF, left ventricular ejection fraction; LVMI, left ventricular mass index.

^aGroup with HFpEF significantly different than group with no heart failure (P < .01).

^bGroup with HFrEF significantly different than group with no heart failure (P < .01).

^cGroup with HFpEF significantly different than group with HFrEF (P < .01).

^dFisher exact test used to compare frequencies because several expected cell counts were <5.

On average, diastolic function was worse among participants with either HFpEF or HFrEF than it was among participants free of HF (all Ps < .05), suggesting that both groups of participants with HF had some degree of diastolic dysfunction. Mean E/A ratio for the sample with HFpEF was less than 1, and mean E wave deceleration time for the sample with HFpEF was greater than 200 milliseconds, suggesting that, on average, the sample with HFpEF had mild diastolic dysfunction and early HFpEF.^48,49 Left ventricular ejection fraction, longitudinal strain, and circumferential strain were worse among individuals with HFrEF and HFpEF than among individuals free of HF (P < .001), suggesting that both groups with HF had abnormalities with systolic function and/or contractility. Mean Ea/Ees ratio, an indicator of ventricular-vascular coupling, was less than 1 in the 3 groups, suggesting that ventricular-vascular coupling was not impaired.⁵⁰

Cognitive Performance in Participants With Heart Failure With Preserved Ejection Fraction, Heart Failure With Reduced Ejection Fraction, and No Heart Failure

On average, participants with HFpEF performed significantly worse than individuals with no history of HF on tests in the domains of attention/processing speed (Digit Symbol Substitution Test, Trail Making Test Part A), language/verbal fluency (Animal Naming Test, Word Fluency Test), executive function (Trail Making Test Part B), and global cognitive function (Mini Mental State Examination) (Table 3). Effect sizes revealed that HFpEF had the greatest effect on scores in the domains of attention/processing speed, language/verbal fluency, and executive function, with moderate effects on the Digit Symbol Substitution Test, Trail Making Test Parts A and B, and Word Fluency Test. Participants with HFrEF demonstrated significantly worse scores than participants with no HF on tests of attention/processing speed and memory. Although other differences were nonsignificant, scores on all other cognitive tests were nominally worse among the 2 groups with HF than among those with no HF. Neurocognitive test scores were not significantly different between those with HFpEF and those with HFrEF.

TABLE 3.

Mean Scores on Neurocognitive Tests by Heart Failure Type and Effect Sizes Compared With Those With Heart Failure With Preserved Ejection Fraction

		No HF (n=5,281)	HFpEF (n=205)	HFrEF (n=60)	HFpEF vs No HF	HFpEF vs HFrEF
Cognitive Domain	Neurocognitive Test	Mean (SD)	Mean (SD)	Mean (SD)	Hedges’ g	Hedges’ g
Memory	DWRT	5.4 (1.7)	5.1 (1.8)	4.9 (1.8)a	0.17	0.15
	LMT Part I	22.3 (7.1)	21.8 (7.0)	21.2 (6.3)	0.08	0.08
	Incidental Learning Test	3.5 (2.3)	3.1 (2.3)	3.4 (2.1)	0.15	0.13
	LMT Part II	17.4 (7.7)	16.8 (7.2)	16.8 (6.7)	0.08	0.01
Attention/processing speed	DSB	5.7 (2.0)	5.4 (1.9)	5.4 (1.7)	0.14	0.03
	DSST	39.3 (11.5)	34.3 (10.1)b	34.3 (10.9)a	0.44	0.004
	TMT Part A	46.0 (24.1)	55.2 (26.7)b	55.7 (31.6)a	0.38	0.02
Language/verbal fluency	Animal naming	16.6 (4.9)	15.5 (4.8)b	15.6 (4.3)	0.24	0.02
	BNT	25.3 (4.8)	24.5 (5.3)	24.7 (5.1)	0.17	0.04
	WFT (fas)	34.0 (12.0)	30.4 (12.2)b	30.4 (10.0)	0.30	0.002
Executive function	TMT Part B	122.5 (58.2)	145.0 (59.3)b	135.1 (53.4)	0.39	0.17
Global cognitive function	MMSE	27.6 (2.5)	27.1 (2.8)b	27.1 (2.7)	0.22	0.01

Abbreviations: BNT, Boston Naming Test; DSB, Digit Span Backwards; DSST, Digit Symbol Substitution Test; DWRT, Delayed Word Recall Test; HF, heart failure; HFrEF, heart failure with reduced ejection fraction; LMT, Logical Memory Test; MMSE, Mini-Mental State Examination; TMT, Trail Making Test; WFT (fas), Word Fluency Test using letters f, a, and s.

^aGroup with HFrEF significantly different than group with no heart failure.

^bGroup with HFpEF significantly different than group with no heart failure.

Echocardiographic Measures of Diastolic Function and Neurocognitive Performance Among Participants With Heart Failure With Preserved Ejection Fraction

All 5 echocardiographic measures of diastolic function demonstrated a significant but weak bivariate association with at least 1 neurocognitive test (Table 4). Consistently, worse diastolic function was associated with worse performance on neurocognitive tests; however, the echocardiographic measures and the cognitive domains affected were not consistent. The left ventricular mass index and E/A ratio demonstrated the greatest number of bivariate associations with neurocognitive test scores. Worse diastolic function was most often associated with worse performance on neurocognitive tests evaluating the domains of attention/processing speed and language/verbal fluency.

TABLE 4.

Pairwise Correlations between Echocardiographic Measures and Scores on Neurocognitive Tests

Neurocognitive Test	LVMI	LAVI	E/A Ratio	E/E’ Ratio	DT	LVEF	Stroke Volume Index	Cardiac Index	Longitudinal Strain	Circumferential Strain	Ea/Ees Ratio
DWRT	−0.24 (P = .001)	−0.06 (P = .43)	0.01 (P = .88)	0.001 (P = .99)	−0.15 (P = .04)	0.09 (P = .32)	−0.20 (P = .004)	−0.09 (P = .23)	−0.03 (P = .72)	0.06 (P = .52)	0.06 (P = .39)
LMT Part 1	−0.10 (P = .17)	−0.04 (P = .55)	0.13 (P = .11)	0.03 (P = .66)	−0.08 (P = .26)	0.03 (P = .73)	−0.09 (P = .19)	−0.08 (P = .28)	0.04 (P = .62)	−0.06 (P = .50)	−0.09 (P = .19)
Incidental Learning	−0.07 (P = .37)	−0.22 (P = .003)	0.10 (P = .21)	−0.04 (P = .56)	−0.03 (P = .71)	−0.02 (P = .78)	−0.18 (P = .01)	−0.06 (P = .39)	0.07 (P = .35)	0.07 (P = .44)	0.07 (P = .31)
LMT Part II	−0.06 (P = .41)	−0.08 (P = .29)	0.09 (P = .23)	0.05 (P = .50)	−0.03 (P = .69)	0.01 (P = .87)	−0.05 (P = .46)	−0.03 (P = .66)	0.04 (P = .59)	−0.04 (P = .69)	−0.05 (P = .52)
DSB	−0.03 (P = .70)	−0.08 (P = .25)	0.21 (P = .01)	0.11 (P = .15)	−0.04 (P = .58)	−0.11 (P = .13)	−0.06 (P = .41)	−0.09 (P = .22)	0.12 (P = .16)	−0.03 (P = .77)	0.06 (P = .44)
DSST	−0.06 (P = .38)	−0.01 (P = .86)	0.09 (P = .27)	−0.10 (P = .18)	−0.05 (P = .53)	−0.12 (P = .10)	−0.20 (P = .01)	−0.15 (P = .04)	0.07 (P = .37)	0.01 (P = .93)	0.11 (P = .12)
TMT Part A	0.13 (P = .07)	0.04 (P = .63)	−0.16 (P = .05)	0.06 (P = .44)	0.13 (P = .08)	0.16 (P = .03)	0.16 (P = .03)	0.20 (P = .01)	0.08 (P = .27)	−0.05 (P = .55)	−0.07 (P = .34)
Animal Naming	−0.07 (P = .35)	−0.11 (P = .12)	0.05 (P = .48)	−0.14 (P = .04)	−0.04 (P = .53)	−0.05 (P = .47)	0.001 (P = .96)	−0.003 (P = .97)	−0.11 (P = .13)	0.09 (P = .29)	0.02 (P = .78)
BNT	−0.06 (P = .40)	−0.03 (P = .69)	0.16 (P = .04)	−0.03 (P = .70)	−0.11 (P = .12)	−0.10 (P = .17)	−0.07 (P = .36)	−0.09 (P = .21)	−0.10 (P = .18)	−0.12 (P = .17)	−0.09 (P = .21)
WFT (fas)	−0.21 (P = .004)	−0.12 (P = .09)	0.03 (P = .72)	−0.05 (P = .47)	−0.02 (P = .78)	0.08 (P = .27)	−0.08 (P = .29)	−0.08 (P = .23)	−0.11 (P = .12)	−0.02 (P = .86)	0.01 (P = .93)
TMT Part B	−0.03 (P = .72)	0.06 (P = .45)	−0.03 (P = .68)	0.10 (P = .18)	0.10 (P = .21)	0.07 (P = .38)	0.05 (P = .48)	0.002 (P = .98)	−0.12 (P = .14)	−0.10 (P = .28)	−0.04 (P = .63)
MMSE	−0.16 (P = .02)	−0.07 (P = .35)	0.12 (P = .13)	0.03 (P = .71)	−0.07 (P = .32)	0.09 (P = .18)	−0.08 (P = .26)	−0.12 (P = .101)	−0.18 (P = .01)	−0.09 (P = .29)	−0.14 (P = .05)

Significant associations are in bold.

Abbreviations: A, peak atrial transmitral filling velocity; BNT, Boston Naming Test; DSB, Digit Span Backwards; DSST, Digit Symbol Substitution Test; DT, E wave deceleration time; DWRT, Delayed Word Recall Test; E, early transmitral filling velocity; E’, tissue velocity of mitral annulus; Ea, arterial elastance; Ees, end-systolic elastance; LAVI, left atrial volume index; LMT, Logical Memory Test; LVEF, left ventricular ejection fraction; LVMI, left ventricular mass index; MMSE, Mini-Mental State Examination; TMT, Trail Making Test; WFT (fas), Word Fluency Test using the letters f, a, and s.

After controlling for the confounding effects of sociodemographics, chronic comorbid conditions, and medication history, 3 associations remained significant in multivariate analysis indicating that worse diastolic function was independently associated with worse neurocognitive test performance (Table 5). Higher E wave deceleration time remained independently associated with worse performance on the Delayed Word Recall Test (a test of memory), an increased left ventricular mass index was associated with worse performance on the Trail Making Test Part A (a test of attention/processing speed), and a higher E/E’ ratio was associated with a lower score on the animal naming test (a test of language/verbal fluency). These diastology-related echocardiographic measures demonstrated a small but similar influence across all 3 neurocognitive tests, with standardized β coefficients ranging from 0.125 to 0.17 after controlling for confounders in their respective multivariate models. Older age, black race, and higher score on the Center for Epidemiologic Studies Depression Scale demonstrated the greatest influence on these 3 neurocognitive tests.

TABLE 5.

Pooled Regression Coefficients from Multivariate Models Predicting Scores on Neurocognitive Tests

	Neurocognitive Tests
	DWRT			TMT Part A			Animal Naming			MMSE
Predictors	B (SE)	95% CI (B)	β	B (SE)	95% CI (B)	β	B (SE)	95% CI (B)	β	B (SE)	95% CI (B)	β
Sociodemographics and smoking history
Age, y	−0.13a (0.02)	[−0.16 to −0.09]	−0.40a	1.01b (0.29)	[0.43 to 1.59]	0.21b	−0.25a (0.056)	[−0.36 to −0.14]	−0.29a	–	–	–
Male	−1.10a (0.23)	[−1.56 to −0.64]	−0.31a	–	–	–	–	–	–	−1.70a (0.37)	[−2.44 to −0.96]	−0.30a
Black race	–	–	–	23.88a (4.02)	[15.94 to 31.82]	0.38a	−2.96a (0.74)	[−4.41 to −1.50]	−0.27a	−2.26a (0.38)	[−3.01 to −1.51]	−0.35a
High school graduate	–	–	–	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Less than high school	–	–	–	10.28c (4.28)	[1.84 to 18.72]	0.16c	−1.48 (0.83)	[−3.12 to 0.17]	−0.13	−1.52a (0.42)	[−2.35 to −0.70]	−0.23a
Vocational school	–	–	–	−4.20 (6.38)	[−16.75 to 8.35]	−0.04	−0.04 (1.26)	[−2.52 to 2.44]	−0.002	0.50 (0.66)	[−0.81 to 1.81]	0.05
College/graduate school/professional school	–	–	–	−6.60 (3.89)	[−14.26 to 1.12]	−0.12	1.59c (0.75)	[0.11 to 3.07]	0.16c	0.92c (0.38)	[0.17 to 1.68]	0.15c
Never smoked	–	–	–	–	–	–	–	–	–	Ref	Ref	Ref
Former smoker	–	–	–	–	–	–	–	–	–	0.45 (0.38)	[−0.29 to 1.19]	0.08
Current smoker	–	–	–	–	–	–	–	–	–	−0.42 (0.83)	[−2.07 to 1.22]	−0.03
Unknown/unable to categorize smoking history	–	–	–	–	–	–	–	–	–	−1.32c (0.57)	[−2.44 to −0.19]	−0.14c
Comorbid illnesses and medications
Hypertension	−0.69c (0.34)	[−1.36 to −0.03]	−0.13c	–	–	–	–	–	–	−0.03 (0.52)	[−1.05 to 1.00]	−0.003
Anemia	–	–	–	–	–	–	–	–	–	−0.21 (0.34)	[−0.88 to 0.46]	−0.04
CES-D	–	–	–	1.61b (0.47)	[0.68 to 2.54]	0.21b	−0.08 (0.09)	[−0.25 to 0.09]	−0.06	−0.13b (0.04)	[−0.22 to −0.05]	−0.18b
β-blocker	−0.25 (0.24)	[−0.72 to 0.23]	−0.06	–	–	–	–	–	–	–	–	–
Measures of diastolic function
LVMI, g/m2	−0.006 (0.004)	[−0.016 to 0.003]	−0.095	0.17c (0.07)	[0.02 to 0.31]	0.17c	–	–	–	−0.01 (0.01)	[−0.02 to 0.004]	−0.08
E/A ratio	–	–	–	1.86 (4.31)	[−6.65 to 10.37]	0.03	–	–	–	–	–	–
E wave deceleration time, ms	−0.004c (0.002)	[−0.007 to −0.0002]	−0.125c	0.05 (0.03)	[−0.004 to 0.112]	0.12	–	–	–	–	–	–
E/E’, cm/s	–	–	–	–	–	–	−0.12c (0.05)	[−0.21 to −0.02]	−0.16c	–	–	–
Measures of cardiac hemodynamics
LVEF, %	–	–	–	0.40 (0.26)	[−0.12 to 0.91]	0.09	–	–	–	–	–	–
SVI, mL/m2 per beat	0.004 (0.016)	[−0.028 to 0.035]	0.016	−0.47 (0.39)	[−1.25 to 0.30]	−0.14	–	–	–	–	–	–
Cardiac Index, L/min	–	–	–	12.30 (5.7)	[1.06 to 23.53]	0.25c	–	–	–	0.0002 (0.0002)	[−0.0001 to 0.0005]	0.09
Longitudinal strain, %	–	–	–	–	–	–	–	–	–	−0.06 (0.07)	[−0.19 to 0.07]	−0.06
Ea/Ees	–	–	–	–	–	–	–	–	–	−0.80 (0.84)	[−2.44 to 0.85]	−0.06
Constant	17.47a (1.56)			122.29a (30.57)			40.28a (4.60)			30.75a (1.77)
Observations	200			190			201			204
R 2	0.344			0.402			0.231			0.459
Adjusted R2	0.321			0.362			0.203			0.416

Multivariate models included predictors that demonstrated significant bivariate associations with scores on the given neurocognitive test. Hyphens indicate that the variable was not included in the multivariate model for the given neurocognitive test because of lack of a significant bivariate association. Significant associations in multivariate models are in bold.

Abbreviations: A, peak atrial transmitral filling velocity; B, unstandardized regression coefficient; β, standardized regression coefficient; CES-D, Center for Epidemiologic Studies Depression Scale; CI, confidence interval; DWRT, delayed word recall test; E, early transmitral filling velocity; E’, tissue velocity of mitral annulus; Ea, arterial elastance; Ees, end-systolic elastance; LV, left ventricular; MMSE, Mini-Mental State Examination; Ref, reference category; SE, standard error; SVI, stroke volume index; TMT, trail making test.

^aP < .001.

^bP < .01.

^cP < .05.

Echocardiographic Measures of Cardiac Hemodynamics and Neurocognitive Performance Among Participants With Heart Failure With Preserved Ejection Fraction

Several measures of cardiac hemodynamics also demonstrated significant weak bivariate associations with neurocognitive test scores (Table 4). Interestingly, better cardiac function (higher ejection fraction, stroke volume index, cardiac index, better longitudinal strain) was associated with worse performance on neurocognitive tests. Stroke volume index demonstrated the greatest number of significant bivariate associations with neurocognitive test scores. Better cardiac function was most commonly associated with worse performance on tests of attention/psychomotor speed, but associations between better cardiac function and worse performance on tests of memory and global cognitive function also were observed.

Only 1 association remained significant in multivariate models. After controlling for age, race, education, depressive symptoms, diastolic function, left ventricular ejection fraction, and stroke volume index, higher cardiac index demonstrated a significant independent association with the Trail Making Test Part A (Table 5). On average, a 1-L/m² increase in cardiac index was associated with an additional 12.3 seconds needed to complete the Trail Making Test Part A. This finding suggests that better cardiac index is associated with worse attention in this sample of participants with an ejection fraction greater than 50%. After controlling for the confounders listed previously, cardiac index demonstrated a moderate influence on Trail Making Test Part A score (β = 0.25). Only black race demonstrated a greater influence on Trail Making Test Part A score (β = 0.38). Older age and a higher score on the Center for Epidemiologic Studies Depression Scale demonstrated a similar but slightly smaller influence on Trail Making Test Part A score than cardiac index did (β = 0.21 for both).

Echocardiographic Measures of Ventricular-Vascular Coupling and Neurocognitive Performance Among Participants With Heart Failure With Preserved Ejection Fraction

Mean Ea/Ees ratio, a measure of ventricular-vascular coupling, only demonstrated 1 significant bivariate association (Table 4). Higher Ea/Ees ratio, indicating impaired ventricular-vascular coupling, was associated with worse performance on the Mini-Mental State Examination, a test of global cognitive function (r = −0.14, P = .05). The association was not significant in the presence of covariates in multivariate analysis, however (B = −0.80; 95% confidence interval, −2.44 to 0.85; Table 5).

Discussion

In this study, cognitive function was worse among individuals with HF than among those with no HF. Cognitive function among participants with HFpEF and those with HFrEF was not statistically different, however. These findings are similar to those reported by Witt and colleagues²² in a previous study of participants in the Atherosclerosis Risk in Communities cohort. In their study, the risk of mild cognitive impairment was higher among participants with HFpEF and with HFrEF than among those with no HF.²² These researchers did not explore the specific cognitive deficits observed in participants with HF, however.²² This secondary analysis builds on previous research by providing a detailed analysis of the cognitive deficits observed in participants with HFpEF. In the current analysis, participants with HFpEF demonstrated deficits in the domains of attention/processing speed, language/verbal fluency, and executive function at Atherosclerosis Risk in Communities visit 5. These patterns of cognitive impairment are similar to those reported in previous studies of individuals with HFpEF in which mean scores were compared with published norms.^17–19 As half of patients with HF have HFpEF,⁵¹ clinicians need to be alert for cognitive impairment in their patients with HFpEF. Cognitive impairment has been associated with poor self-care, including poor medication adherence and worse disease management in response to symptoms.⁵ To mitigate the effects of cognitive impairment on self-care, clinicians should consider modifying plans of care to accommodate for deficits in cognitive function.

This analysis of Atherosclerosis Risk in Communities data also adds to the existing literature by describing a potential association between abnormal cardiac function and cognitive impairment in HFpEF. Diastolic dysfunction was independently associated with abnormalities in memory, attention/psychomotor speed, and language/verbal fluency in the current study. Diastolic dysfunction may contribute to reduced ventricular filling and, ultimately, reduced cerebral blood flow; however, additional research on cerebral blood flow in HFpEF is necessary to evaluate this potential association. Associations between diastolic dysfunction and cognitive impairment were weak, which may be because of low disease severity in this sample. Participants at Atherosclerosis Risk in Communities visit 5 were required to complete a full-day evaluation, which may have been too burdensome for patients with high disease severity. As a result, the sample likely is biased toward healthier participants with better diastolic function. Future research on samples with advanced HFpEF are necessary to determine the strength of the association between diastolic dysfunction and cognitive impairment.

Surprisingly, higher cardiac index was associated with worse cognitive impairment in the current study. This finding is counterintuitive and contradicts the Conceptual Model of Cognitive Impairment in Heart Failure, but the findings are consistent with previous research.^52–54 High ejection fraction was associated with worse performance on tests of memory, language, and executive function in a sample of adults (n = 1114) enrolled in the Framingham study.⁵³ The findings of these studies suggest that, although left ventricular output is necessary for cognitive function, left ventricular output may be detrimental to cognitive function if it rises above a certain threshold. The association between high cardiac index and cognitive impairment observed in this study may be incidental, however. Catecholamine levels often rise in HF in an attempt to improve cardiac function and may be responsible for the higher cardiac index.⁵⁵ Furthermore, there is evidence that abnormal catecholamine levels may be one of the many factors that contribute to cognitive impairment in HF.^56,57 Additional research is needed to understand this association, however. Survival and visit attendance bias also may have contributed to the findings observed in this secondary analysis. Individuals with a low cardiac index may have been too ill to participate in this study, leaving only participants with high cardiac index, some of whom had cognitive impairment, to participate. The mechanisms behind this association have not yet been explored, however.

In this secondary analysis of Atherosclerosis Risk in Communities data, elevated Ea/Ees ratio, indicative of impaired ventricular-vascular coupling, was associated with lower scores on the Mini-Mental State Examination in bivariate analysis, but the association was lost after controlling for potential confounders. Mean Ea/Ees was within the range of optimal functioning in the current sample.⁵⁰ As individuals in the current study did not display significant impairment in ventricular-vascular coupling, the lack of association in the current study is not surprising. To determine whether an association between ventricular-vascular coupling and cognitive impairment exists in HF, authors of future research should evaluate the association in a sample of individuals with advanced HFpEF, in which impaired ventricular-vascular coupling is expected to be more prevalent.⁵⁰

Limitations of the Current Secondary Analysis

Although the sample with HFpEF in our analysis was larger than that in previous research,^17–21 the sample was smaller than anticipated. To account for the smaller sample and to maintain statistical power, variables had to be selected carefully to create a parsimonious model. Despite the smaller sample, power was maintained at greater than 0.90 using a corrected P value to minimize the possibility of false discovery.

Echocardiographic data suggested that the participants in this secondary analysis had mild diastolic dysfunction and early HFpEF.^48,49 Because diastolic dysfunction was mild, hemodynamics may not have been abnormal. Despite the low severity of HFpEF, 4 significant associations were identified in multivariate models. If HFpEF had been more severe, it is possible that hemodynamics may have been altered and that associations may have been stronger.

Because only 1 time point was used in this analysis, measurements of cognitive function and cardiac function are only reflective of cognitive function and cardiac function at the time of measurement. Measurement at several time points would provide a more accurate reflection of average cognitive function and cardiac function of the participants.

Implications for Further Research

Because this secondary analysis was cross-sectional, the findings are only correlational and do not imply that diastolic dysfunction and high cardiac index cause cognitive impairment in HFpEF. However, the findings from the current study provide important insight into potential mechanisms behind cognitive impairment in HFpEF. Prospective studies are needed to determine whether diastolic dysfunction and high cardiac index contribute to the development of cognitive impairment in HFpEF and whether changes in cardiac function over time correlate with changes in cognitive function.

There remains an important need to identify the prevalence of global cognitive impairment more accurately and to measure cognitive function in a consistent manner. A standardized battery of sensitive measures of cognitive function in HF should be established and should be used consistently in future research. Use of a common measure facilitates comparison across studies and meta-analysis of findings.

Despite the inconsistencies in measurement, there is ample evidence that cognitive impairment is prevalent in HF. However, research on interventions to mitigate cognitive impairment is scarce. Some researchers have begun to explore interventions to improve memory and attention in patients with cardiovascular disease,^58–61 but more intervention research is needed.

The findings reported here suggest that abnormal hemodynamics may contribute to cognitive impairment in HFpEF through reduced cerebral perfusion, but other mechanisms are likely involved. Subclinical microemboli and the neurotoxic effects of chronic inflammation also have been hypothesized to contribute to cognitive impairment,^8,23,62–65 but associations between these potential mechanisms and cognitive impairment in HFpEF have not been established. Authors of future research should explore other factors that may contribute to cognitive impairment in HFpEF.

Implications for Practice

The study findings underscore the importance of screening individuals with HF, including those with HFpEF, for cognitive impairment. Because researchers demonstrate that cognitive impairment can influence self-care and can contribute to worse outcomes, including higher mortality rates,^5,12–16 screening for cognitive impairment should be included in HF assessment. Early identification of cognitive impairment and appropriate intervention have the potential to improve self-care, reduce mortality, and reduce healthcare costs. The findings also suggest that individuals with HFpEF have more depressive symptoms compared with individuals with no history of HF. Given the high prevalence of depressive symptoms in HFpEF^66–68 and the association between depressive symptoms and cognitive impairment,^69–71 screening for depressive symptoms should be incorporated into clinical practice.

Conclusion

Individuals with HFpEF demonstrated deficits in the cognitive domains of attention/psychomotor speed, language/verbal fluency, executive function, and global cognitive function that exceeded those observed in individuals with no HF. Although abnormal cardiac function was associated with abnormalities in these cognitive domains, the mechanism behind cognitive impairment in HFpEF is likely multifactorial and needs to be explored in greater detail. Clinicians should be aware of the possibility of cognitive impairment in their patients, and researchers should continue to study cognitive impairment in HFpEF to determine whether interventions can mitigate the negative effect of cognitive impairment on outcomes.

What’s New and Important.

• The effects of HFpEF on cognitive impairment were small to moderate, particularly in the domains of attention, language, and executive function.

• Measures of diastolic dysfunction, which are common in HFpEF, were associated with impaired memory, attention, and language. These associations support the theory that impaired cardiac function contributes to cognitive impairment.

• The findings of this study provide evidence that factors other than left ventricular ejection fraction contribute to cognitive impairment in HF.