Associations of echocardiographic features with stroke in those without atrial fibrillation

Abstract

Objective

To determine the associations between transthoracic echocardiogram (TTE) cardiac structure/function measures and cardioembolic stroke (CES) and new-onset atrial fibrillation (AF) in patients without known AF.

Methods

Inpatients at a single institution (2013–2015) with imaging-confirmed ischemic stroke, no AF, and TTE within the 1st week were included. TTE structure/function variables were abstracted. Stroke subtype (CES vs other) was defined according to Trial of Org 10172 in Acute Stroke Treatment, blinded to TTE results. New AF was defined as any duration of AF on ECG, telemetry, or event monitor. Separate multivariable logistic regression models defined associations between CES or new-onset AF and TTE measures, adjusting for demographic and vascular risk factors.

Results

Of 322 participants (mean age 60 years), 55% were male and 56% African American. In adjusted models (odds ratio, 95% confidence interval), odds of CES increased per 0.1 cm increase in left atrial (LA) systolic diameter (1.06, 1.02–1.11), 1 cm/s in mitral E point velocity (1.03, 1.02–1.05), with presence of mitral valve dysfunction (3.78, 1.42–10.02), and with wall motion abnormality (2.00, 1.13–3.55). As ejection fraction increased (per 10%), odds of CES decreased (0.65, 0.53–0.79). New-onset AF was also associated with increasing LA systolic diameter (1.13, 1.04–1.22).

Conclusions

Cardiac structural changes independent of AF and detectable on TTE may be on the CES causal pathway. Confirming these results could have implications for future use of TTE and decisions about antithrombotic vs anticoagulant treatment.

Routine inpatient transthoracic echocardiogram (TTE) has been primarily used in acute stroke care to rule out a cardiac thrombus or septic vegetation among those in whom cardioembolic stroke (CES) is suspected. However, recent stroke guidelines¹ now question prolonged monitoring for atrial fibrillation (AF) and routine use of TTE due to lack of evidence. There remains a paucity of literature regarding TTE structural features and how they should be interpreted, specifically with regards to stroke etiology, if AF is not detected or subsequently diagnosed.

As the benefits of poststroke anticoagulation for AF² are known, research has focused on duration of monitoring and choice of anticoagulant. Etiology of embolic-appearing strokes, termed embolic strokes of undetermined source (ESUS), has recently been the focus of more extensive evaluation, with several studies identifying markers that might indicate a cardiac etiology for some ESUS such as left atrial (LA) enlargement,³ p-wave terminal force,⁴ and pro-BNP.⁵

The objective of this study was to identify cardiac structure and function measures, obtained on prior to guideline routine stroke care TTE, that are associated with CES in those without AF, and to determine if these associations are independent of other traditional stroke risk factors. We hypothesize that (1) LA size and indices of valve dysfunction will be most strongly associated with CES, even among participants without AF; and (2) TTE measures will be associated with new AF, as certain echo findings are often believed to be the result of longstanding AF and may represent a mechanism by which CES may occur.

Methods

Standard protocol approvals, registrations, and patient consents

The JHH institutional review board approved the study and the protocol met criteria for waiver of consent.

Study population

The study was conducted at a single center, the Johns Hopkins Hospital (JHH), which is a comprehensive, tertiary referral center for stroke. Inclusion criteria were adult patients (≥18 years of age) admitted to JHH with an ICD-9 code of ischemic stroke (433, 434, 434.01, 434.11, 434.91, 436) during a 2-year period (July 1, 2013–July 1, 2015) and a TTE during the first 7 days of admission. The primary exclusion criterion was a known diagnosis of AF at time of admission.

Echocardiogram measures

TTE was obtained at the discretion of the inpatient treatment team as had been standard practice at the time of study enrollment. All echocardiograms were performed by accredited technicians and reports were generated by echocardiography board certified cardiologists at JHH. TTE was performed using the Philips (Best, the Netherlands) IE33 ultrasound in the standard 2D analysis during each acute inpatient admission. The anatomical and hemodynamic abnormalities were observed by following the sequential segmental approach. Standard views were used in all cases including parasternal long-axis and short-axis views of the ventricles, parasternal view of the right ventricular inflow tract, and apical 4-chamber views. The color Doppler images were utilized to identify blood flow and valvular regurgitation. An atrial septal defect was determined by administration of agitated saline injection.

The following continuous variables were recorded for each participant: LA systolic diameter (cm), mitral A point velocity (cm/s), mitral E point velocity (cm/s), left ventricle (LV) posterior wall diastolic thickness (cm), LV septal wall diastolic thickness (cm), and LV ejection fraction (EF) (%). Mitral E and A point velocities are components of the mitral inflow signal and are used to assess the diastolic function of the heart, representing the peak of passive filling of the LV (peak E wave) and late filling as a result of the active contraction of the LA (peak A wave) and represent a pressure difference between the atria and the ventricle. The following dichotomous variables were recorded for each participant: LA diameter enlargement, mitral valve (MV) dysfunction, aortic valve (AV) dysfunction, septal defect, atrial septal aneurysm, mitral annular calcification, LV aneurysm, and regional wall motion abnormality. LA diameter was dichotomized into enlarged vs not based on established norms and was not run in any model with continuous LA systolic diameter.⁶ Valve dysfunction was defined as cardiologist-classified severe regurgitation. Regional wall motion abnormality was defined as evidence of focal hypokinesis.

Covariates

Age, sex, race, body mass index (BMI; kg/m²), hemoglobin A1C (HgA1C; %), low-density lipoprotein (LDL; mg/dL), international normalized ratio (INR), admission NIH Stroke Scale (NIHSS), antihypertensive medication use, antithrombotic medication use, anticholesterol medication use, and anticoagulant use were covariates of interest. Medication use was defined as medications prescribed and verified as taking with the patient at the time of admission. Antihypertensive medication use functioned as a surrogate for hypertension since actual elevated blood pressure values at the time of stroke might not be indicative of preexisting hypertension.

Outcome adjudication

Stroke subtype was classified according to Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria by a cerebrovascular neurologist who was masked to the echo features. A subset of the cases (n = 60) were also reviewed by a second cerebrovascular neurologist to assess interrater agreement. The TOAST classification, which has demonstrated high interrater reliability,⁷ denotes 5 subtypes of ischemic stroke: (1) large-artery atherosclerosis, (2) cardioembolism or cardioembolic (CES), (3) small-vessel occlusion, (4) stroke of other determined etiology, and (5) stroke of undetermined etiology.

A patient-reported history of AF or any prior documentation of a diagnosis of AF at the time of admission was used as diagnostic of prior AF (and was, as stated above, the primary exclusionary criterion). A new diagnosis of AF was defined as evidence of AF on ECG or inpatient telemetry evidence of AF at any point during the duration of the admission, to include the emergency department if it immediately preceded the inpatient stay, without a prior diagnosis of AF. If a cardiac event monitor (either 30 days Holter or implantable loop recorder) was ordered after discharge and demonstrated AF, this was also included. The window for time of diagnosis for new AF was defined as 30 days after placement of the cardiac event monitor in order to make the devices comparable. The decision about placing an implantable loop recorder or 30-day Holter was made by the clinical team, who was not aware of the study hypotheses.

Statistical analysis

The primary dependent variables in separate analyses were presence of CES (vs other) and new AF. Echocardiographic variables, which were the primary independent variables, were evaluated descriptively and in univariate analyses in association with the outcomes of interest using crude logistic regression. Subsequent multivariable models then explored the following sequential levels of adjustment, for both the CES analyses and the new AF analysis.

Model 1 considered participant age, sex, and race (black vs other). Model 2 further accounted for model 1 and antihypertensive medication use, BMI, HgA1C, and LDL. Model 3 accounted for model 1 and additionally adjusted for INR, antithrombotic medication use, and anticholesterol medication use. Model 4 included model 2 plus INR, anticoagulant use, and anticholesterol medication use. The same models were used for adjustment when considering potential confounders in the association of a diagnosis of new AF and TTE measures.

Models considered each echo variable separately as the primary analysis. After verifying that TTE variables were not collinear, they were also included together in a sensitivity analysis for the outcome of CES. We tested whether the associations between CES and TTE findings were modified by age (≥ or <75 years) or race (black vs other) with an interaction term and, when significant, in stratified models. An additional sensitivity analysis was a multivariable logistic regression with new-onset AF as an additional covariate along with those listed in model 3 in the evaluation of CES. All statistical analyses were performed using Stata v14.1.⁸ Two-sided p < 0.05 or an interaction p value < 0.1 were considered statistically significant.

Data availability

The datasets generated during analysis of this study are available from the corresponding author on reasonable request.

Results

Demographics

Baseline patient characteristics (number and % or mean ± SD) are provided for all included individuals (table 1). During the study duration, 550 patients were admitted to JHH with an acute ischemic stroke. Of that group, 7 people lacked TTE within the first 7 days of admission and 48 participants had prevalent AF and were excluded. In addition, 38 patients without complete BMI, HgA1c, LDL, or INR data, necessary for multivariable models, were excluded. Of the remaining patients (n = 332), the majority were African American, male, with a history of hypertension, with a moderate NIHSS and normal EF (table 1). Mean LA diameter was in the range of normal for both male and female participants.⁶ Seventy-six of these patients were adjudicated as having a CES, 98 with small vessel disease and 90 with large artery atherosclerosis. Of the 20% of the cases adjudicated by the second cerebrovascular neurologist, there was agreement in TOAST subclassification in 72% of the cases. During the hospital stay or on follow-up monitoring, 23 patients were diagnosed with incident AF.

Table 1.

Demographic characteristics and transthoracic echocardiogram (TTE) features of patients

Associations with CES

Increasing LA systolic diameter was significantly associated with CES and remained significant after adjustment for potential confounders (table 2). In the fully adjusted model, for every 0.1 cm increase in LA systolic diameter, the odds of CES increased 9% (odds ratio [OR] 1.09, 95% confidence interval [CI] 1.04–1.15). Those with increased (dichotomized) LA systolic diameter had more than 3 times the odds of a CES than those without (OR 3.31, 95% CI 1.75–6.27). MV dysfunction (OR 3.70, 95% CI 1.35–10.15), but not AV dysfunction, was associated with increased odds of CES. Participants with focal wall motion abnormality on TTE had approximately 2 times the odds of CES (OR 2.18, 95% CI 1.19–3.97).

Table 2.

Multivariable logistic regression demonstrating the adjusted odds ratios (OR) of cardioembolic stroke with echocardiography variables

As participant EF increased per 10%, the odds of CES decreased (OR 0.65, 95% CI 0.53–0.79). When considering mitral inflow signal used to evaluate cardiac diastolic function, increasing E point velocity (per cm/s), which represents the peak of passive filling of the LV, was associated with a higher odds of CES (OR 1.04, 95% CI 1.02–1.05). Increases in mitral A point velocity (per cm/s), which represents late filling as a result of active contraction of the LA, was associated with a lower odds of CES (OR 0.99, 95% CI 0.98–1.00). However, given the small effect estimates and the overlapping CIs, we interpret these findings cautiously. The presence of mitral annular calcification was also associated with decreased odds of CES (OR 0.47, 95% CI 0.21–1.07).

When considering all echo variables together in the same model (table 3), the direction of association with CES did not change for any of the TTE markers and EF and mitral annular calcification remained significantly associated with CES (model 3).

Table 3.

Multivariable logistic regression demonstrating the adjusted odds ratios (OR) of cardioembolic stroke with echocardiography variables in one model

As associations may have been affected differentially by antithrombotic use (antiplatelet or anticoagulant) vs anticoagulant use alone, a sensitivity analysis was performed adjusting for age, sex, black race, INR, and anticholesterol medication use as per prior adjustment models but also included anticoagulant use (model 4). The direction of any significant associations with CES remained unchanged with similar effect estimates (results not shown).

Effect modification by age and race

Age, dichotomized at 75 years, significantly modified the relationships between several TTE markers and CES (figure 1). Increasing LV posterior wall diastolic thickness was nonsignificantly associated with an increased odds of CES in those older than 75 years (OR 1.07, 95% CI 0.71–1.61) but a nonsignificantly decreased odds in those younger than 75 years (OR 0.90, 95% CI 0.77–1.05) (p interaction = 0.044). Older participants with wall motion abnormality had a nonsignificantly decreased odds of CES (OR 0.64, 95% CI 0.11–3.68) while those who were <75 years of age had an increased odds (OR 2.84, 95% CI 1.41–5.73) (p interaction = 0.066).

Figure 1. Odds ratios for cardioembolic stroke by various echocardiogram variables, stratified by age.

There was a significant interaction between black race and valve (mitral and aortic) dysfunction for CES that was not observed for any other echo measures (figure 2). There were no black participants with CES who had AV dysfunction, but the number of observations was small (n = 9). MV dysfunction (p interaction 0.07) was associated with an increased odds of CES in all race groups, but the effect estimate was much higher, although imprecise, among nonblack participants (black OR 1.63, 95% CI 0.36–7.32; nonblack OR 14.50, 95% CI 2.50–84.13).

Figure 2. Odds ratios for cardioembolic stroke by various echocardiogram variables, stratified by race.

Mediation analysis

To evaluate whether the effect of the TTE markers on the outcome of CES might be mediated through AF, new diagnosis of AF was included in a multivariable logistic regression with model 3 covariates and each of the echo measures. With new AF added to the models, there was no change in the direction of the association with any of the TTE markers and CES, nor were there large changes in effect estimate size (table 4).

Table 4.

Multivariable logistic regression demonstrating the odds ratios (OR) of cardioembolic stroke including new atrial fibrillation (AF)

Associations with new diagnosis of AF

Increasing LA systolic diameter (per 0.1 cm) was significantly associated with an increased odds of a new diagnosis of AF, irrespective of adjustment model (table 5). LA systolic diameter dichotomized, however, was not significantly associated with new-onset AF. None of the 18 patients with MV dysfunction were diagnosed with new AF, so MV dysfunction was not analyzed as a predictor of new AF due to model instability.

Table 5.

Multivariable logistic regression demonstrating the adjusted odds ratios (OR) of new atrial fibrillation with each echocardiography variable

Discussion

In this study of acute ischemic stroke patients without a preexisting diagnosis of AF, odds of CES is associated with changes in structural and functional measurements demonstrated on routine stroke care TTE. Specifically, increases in LA systolic diameter were significantly associated with increased odds of CES among participants without a diagnosis of AF, as well as with new-onset AF, even after adjustment for potential demographic and vascular risk factor confounders. We also show that increasing EF was robustly associated with decreased odds of CES, and this association remained even when considering changes in other echo variables. We also reported significant differences in some of the described relationships among distinct race or age groups; AV and MV dysfunction were stronger predictors of CES among nonblack participants than among black participants, and odds of CES differed in younger vs older patients with wall motion abnormalities.

Given the recent change in guidelines coupled with the established causal association and treatment pathway between AF and CES, and the need for better management algorithms for CES patients without AF, we sought to characterize changes on TTE among those without AF. We recognize that it may be that our participants actually had AF, either an undocumented history or new-onset AF not captured in this cross-sectional analysis, and changes seen on TTE were secondary to this mechanism. However, it may also be that there are independent, cardiac structural changes irrespective of AF that are associated with CES that were captured with small changes on routine TTE. This hypothesis is supported by our sensitivity analysis, which included a new diagnosis of AF as a covariate, and suggested that some of the TTE findings still were significantly associated with CES.

Current practice dictates anticoagulation after ischemic stroke in the setting of AF, but clinicians and patients may experience frustration when the stroke appears embolic but AF is not present and never subsequently captured. Ongoing studies are investigating if non–vitamin K antagonist oral anticoagulant administration is superior to aspirin among those with ESUS^9,10; however, in one recently published study of ESUS evaluating rivaroxaban, the study was terminated due to lack of benefit and increased bleeding risk.¹¹ This may suggest anticoagulation for all those with embolic-appearing strokes, in whom no source is identified, is not an effective strategy. Further understanding of the relationship between cardiac structural and functional measures and CES may help bridge this gap, as it is possible (and perhaps even likely) that a subset of these patients (ESUS but without AF) exists in whom anticoagulation would be the ideal treatment.

The association between LA enlargement and AF is well-known and our findings support this prior work.^12,13 There has been some research that also supports the found association between LA enlargement and stroke recurrence even when controlling for AF.¹⁴ In one prior study, LA enlargement on TTE combined with LA mechanical dysfunction, assessed with strain analysis, reflected traditional transesophageal high-risk findings and was diagnostically useful in CES patient evaluation.¹⁵ Our findings would support that even subtle TTE changes may be reflective of a dynamic process in the structure of the atria and are independently associated with a diagnosis of CES, and not simply a marker of preexisting AF.¹⁶

We also found a decreased odds of CES among those with higher EF. While heart failure is considered a risk factor for stroke, there is no evidence of benefit in treating patients with warfarin (anticoagulation) over aspirin (antiplatelet) when there is no evidence of AF¹⁷ and as a result, current secondary prevention of CES patients with reduced EF is currently the same as any other ischemic stroke subgroup. Perhaps the relationship with EF is more nuanced and outside of frank heart failure, future research building upon our findings may find even subtle changes in EF play a role in predisposing to CES.

Differences in association between younger and older participants as well as black and nonblack participants were also considered and our data suggest that while there are markers of CES that are similar across age/race, there are also some differences. The strength of association with valve dysfunction was much stronger among nonblack participants, which would support prior work suggesting racial differences in valve degenerative etiologies and morphology.¹⁸ We also acknowledge that while attempting to capture hypertension through prior medication use, it may be that this association among black patients, who are known to have a higher percentage of uncontrolled hypertension,¹⁹ may have been a result of residual confounding.

We recognize that there are limitations inherent in the longitudinal retrospective cohort design, such as the inability to determine causation. As previously discussed, defining new AF is difficult with a retrospective cohort. The data were rigorously reviewed, with direct re-review of ECG tracings/telemetry data to decrease ascertainment bias, but selection bias (who received an event monitor) cannot be ruled out, although notably none of the participants who wore an event monitor during the defined study window had AF. It is possible that additional participants who were not given an event monitor or implantable device on discharge had AF that was not ultimately diagnosed. We recognize that different echocardiogram readers may have introduced some heterogeneity, but this methodology increases generalizability to real-world practice and ensured that the readers were blinded to stroke etiology. We also recognize that the majority of participants in this study self-identified as black or white and as a result these findings may not be generalizable to other race–ethnic groups.

Our findings support the idea that TTE can be clinically meaningful and that even mild changes on TTE, such as increases in LA systolic diameter, are associated with an increased odds of CES as well as new AF. Such dynamic changes, often thought to be a result of prevalent AF, may in fact represent more subtle structural change that either precede or are independent of AF. Future research is needed in this area to facilitate the development of clear evidence-based guidelines regarding cardiac evaluation in acute ischemic stroke patients. Recent initiation of clinical trials, such as Atrial Cardiopathy and Antithrombotic Drugs in Prevention After Cryptogenic Stroke (ARCADIA),²⁰ that use a biomarker-driven approach to include left atrial size index ≥3.0 cm/m² on echocardiogram to define atrial cardiopathy are an important first step. While our findings should be confirmed prospectively, we anticipate a time when antithrombotic therapy decisions are influenced by a comprehensive assessment of markers of cardiac structure and function. Further definition of such markers might pinpoint even earlier those who would benefit from anticoagulation in embolic-appearing strokes when there is no AF, or help steer a clinician towards more aggressive monitoring due to increased suspicion for AF.

Glossary

AF

atrial fibrillation

AV

aortic valve

BMI

body mass index

CES

cardioembolic stroke

CI

confidence interval

EF

ejection fraction

ESUS

embolic strokes of undetermined source

HgA1C

hemoglobin A1C

ICD-9

International Classification of Diseases–9

INR

international normalized ratio

JHH

Johns Hopkins Hospital

LA

left atrial

LDL

low-density lipoprotein

LV

left ventricle

MV

mitral valve

NIHSS

NIH Stroke Scale

OR

odds ratio

TOAST

Trial of Org 10172 in Acute Stroke Treatment

TTE

transthoracic echocardiogram

Author contributions

Michelle Johansen: drafting/revising the manuscript, data acquisition, study concept or design, analysis or interpretation of data, accepts responsibility for conduct of research and final approval, acquisition of data, statistical analysis, study supervision, obtaining funding. Michelle Lin: data acquisition, accepts responsibility for conduct of research and final approval, acquisition of data. Saman Nazarian: drafting/revising the manuscript, analysis or interpretation of data, accepts responsibility for conduct of research and final approval, acquisition of data. Rebecca F. Gottesman: drafting/revising the manuscript, data acquisition, study concept or design, analysis or interpretation of data, accepts responsibility for conduct of research and final approval.

Study funding

M. Johansen: American Heart Association Mentored Clinical and Population Research Award #16MCPRP30350000, NIH/ICTR (KL2). R. Gottesman: National Institute on Aging (K24 AG052573).

Disclosure

M. Johansen and M. Lin report no disclosures relevant to the manuscript. S. Nazarian: consultant to CardioSolv, Siemens, and Biosense Webster, and research grants from Siemens, Biosense Webster, and ImriCor. R. Gottesman: Associate Editor of Neurology®. Go to Neurology.org/N for full disclosures.