Left Atrial Strain Impairment Precedes Geometric Remodeling as a Marker of Post-Myocardial Infarction Diastolic Dysfunction

Abstract

OBJECTIVES

The aims of this study were to test the magnitude of agreement between echocardiography- (echo) and cardiac magnetic resonance (CMR)-derived left atrial (LA) strain and to study their relative diagnostic performance in discriminating diastolic dysfunction (DD) and predicting atrial fibrillation (AF).

BACKGROUNDS

Peak atrial longitudinal strain (PALS) is a novel performance index. Utility of echo-quantified LA strain has yet to be prospectively tested in relation to current DD guidelines or compared to CMR.

METHODS

The study population comprised 257 post-myocardial infarction (MI) patients undergoing echo and CMR, including prospective derivation (n = 157) and clinical validation (n = 100) cohorts. DD was graded on echo using established consensus guidelines blinded to strain results.

RESULTS

PALS on both echo and CMR was nearly 2-fold lower among patients with versus no DD (p < 0.001) and was significantly different in those with mild versus no DD (p < 0.01). In contrast, LA geometric parameters including echo- and CMR-derived volumes were significantly different between advanced versus no DD groups (p < 0.001) but not between groups with mild versus no DD (all p > 0.05). Echo and CMR PALS yielded small differences irrespective of orientation and similar diagnostic performance for DD in the derivation (area under the curve: 0.70 to 0.78) and validation (area under the curve 0.75 to 0.78) cohorts. Impaired PALS on both modalities was independently associated with MI size (p < 0.001). During 4.4 ± 3.8 years of follow-up in the derivation cohort, 8% developed AF. Both 2-chamber echo- and CMR-derived PALS stratified arrhythmic risk (p = 0.004 and p = 0.02, respectively), including a 4-fold difference among patients in the Lowest versus remainder of quartiles of echo-derived PALS (24% vs. 6%). Similarly, echo and CMR PALS were Lower (both p < 0.05) among patients with subsequent heart failure hospitalizations.

CONCLUSIONS

Echo-derived PALS parallels results of CMR, yields incremental diagnostic utility versus LA geometry for stratifying presence and severity of DD, and improves prediction of AF and congestive heart failure after MI.

Left atrial (LA) mechanics are impacted myocardial infarction (MI). Myocyte necrosis augments left ventricular (LV) stiffness, resulting in augmented LA afterload due to altered LV diastolic filling. Echocardiography (echo) is widely used to assess both LA geometry and LV diastolic function: Diastolic dysfunction (DD) has been shown to be common after MI and to predict adverse clinical outcomes, including heart failure symptoms, atrial arrhythmias, and death (1-3). Whereas established guidelines exist for DD classification, application of current algorithms can be complex and entail incorporation of multiple parameters (including individual indices that yield indeterminate or conflicting results). Given that both presence and severity of DD have been strongly linked to prognosis, identification of novel approaches to adjudicate DD and predict clinical outcomes is of substantial importance (4).

Advances in echo technologies enable improved assessment of LA deformation (strain). LA strain impairment on echo has been reported in a range of pathologies, including atrial fibrillation (AF), hypertension, and heart failure (5-7), and has the potential to provide a singular index of LV diastolic performance (8). A key advantage of strain is that it can be quantified using data acquired in nearly all routine echo examinations, facilitating clinical implementation as well as application to large-scale pre-existing research datasets. However, echo-derived LA strain has yet to be studied in the post-MI setting or evaluated in relation to current diastolic function guidelines. More broadly, the relative utility of echo-derived LA strain has yet to be prospectively tested in relation to other imaging tests such as cardiac magnetic resonance (CMR), a modality that also can quantify strain and is well established as a volumetric reference standard for LA geometry (9).

This study assessed the pattern, predictors, and prognostic utility of LA strain among a prospectively accrued cohort of patients with post-MI who underwent a tailored multimodality imaging protocol. The aims of the study were to test the 1) magnitude of agreement between echo- and CMR-derived atrial strain, as well as structural markers of impaired strain by both modalities; 2) relative diagnostic performance of echo- and CMR-derived LA strain for DD categorized in accordance with updated consensus guidelines; and 3) incremental utility of LA strain compared to DD and conventional LA remodeling indices for prediction of new-onset AF.

MATERIAL AND METHODS

The data, analytic methods, and study materials will be made available to other researchers for purposes of reproducing the results or replicating the procedure, on request (contingent on approval of the Weill Cornell Institutional Review Board and assurance of data de-identification).

STUDY POPULATION.

The population comprised an initial derivation cohort of patients with ST-segment elevation MI enrolled in 2 prospective imaging studies focused on post-MI remodeling who underwent echo and CMR (10,11), as well as a subsequent validation cohort of patients with MI undergoing CMR and echo for whom data were retrieved from an established institutional registry (12,13).

In all patients, comprehensive demographic data were obtained, including cardiac risk factors, medications, and cardiovascular symptom status. AF, heart failure-related hospitalizations, and invasively quantified left ventricular end-diastolic pressure (LVEDP) at the time of cardiac catheterization were defined based on medical documentation as discerned by review of electronic medical records (blinded to results of imaging tests). All patients in the derivation cohort provided informed written consent at the time of imaging. This study was conducted with approval of the WCMC institutional review board.

IMAGE ACQUISITION AND ANALYTIC PROTOCOL.

Echo and CMR were performed within 3 days of each other using a standardized acquisition and analytic protocol.

Echocardiography.

Comprehensive transthoracic echoes were acquired using commercial equipment (Vivid 7, General Electric, Waukesha, Wisconsin; SC2000, Siemens Medical Solutions, Malvern, Pennsylvania; and iE33 and EPIQ7, Philips, Andover, Massachusetts). Echo included evaluation of the LA from the apical 4-chamber (4C) and 2-chamber (2C) views measured as specified in the American Society of Echocardiography guidelines, with method of disks used to quantify LA volume (14). Mitral regurgitation was quantified using regurgitant fraction as well as aggregate severity (5-point scale) based on additional parameters, including vena contracta, volumetric indices, jet depth, and mitral and pulmonary vein flow patterns (15). Pulmonary artery systolic pressure was calculated from tricuspid regurgitant (TR) velocity and inferior vena cava caliber. LV function, size, and mass were quantified via linear dimensions, consistent with methods validated in previous necropsy comparison and outcomes studies (16,17).

Tailored assessment for diastolic function was performed via transmitral pulsed-wave Doppler at the mitral tips in the 4C view, and early (E) and late (A) diastolic filling velocities, E/A ratio, and E-wave deceleration time were obtained. Tissue Doppler imaging of the mitral annulus was performed at the septal and lateral positions from which early (e′) velocities were measured. Presence and graded severity of DD were classified using current consensus guidelines (4). In brief, DD was identified in patients with preserved LV systolic function when more than one-half of the following cutoff values were met: average E/e′ >14; septal e′ velocity <7 cm/s or lateral e′ velocity <10 cm/s; TR velocity >2.8 m/s; and LA volume index >34 ml/m². All diastolic parameters were available in >95% patients. Further categorization was performed for those with preserved ejection fraction and DD, as well as for those with depressed LV ejection fraction as follows: grade I DD if E/A ≤0.8 and peak E velocity ≤50 cm/s; grade II DD if E/A ≤0.8 and peak E velocity >50 cm/s or E/A >0.8 and <2, with 2 of 3 or 3 of 3 of the following: average E/e′ >14, TR velocity >2.8 m/s, or LA volume index >34 ml/m²; and grade III DD if E/A ≥2.

Cardiac magnetic resonance.

CMR was performed using 1.5- and 3.0-T scanners (General Electric). Cine-CMR used a steady-state free precession pulse sequence. Images were acquired in standard short- and long-axis planes. LA size was quantified in 4C and 2C long-axis views, and method of disks was used to quantify LA volume. MI was evaluated using an inversion recovery pulse sequence acquired 10 to 30 min post-gadolinium (0.2 mmol/kg). LV infarction was colocalized via a standard American Heart Association/American College of Cardiology 17 segment model. MI size (percent myocardium) was measured based on regional transmural extent of enhancement (based on midpoint of each affected segment) in accordance with previous methods applied by our group and others (10,18).

LA STRAIN QUANTIFICATION.

For both echo and CMR, peak atrial longitudinal strain (PALS) was quantified in the LA in 2C and 4C views. For echo, images were acquired at a frame rate of 60 to 90 Hz. PALS was derived using commercial software (2D CPA, TomTec, Munich, Germany), for which automated border detection was manually adjusted to ensure optimal tracking throughout the cardiac cycle (Figure 1). For CMR, strain was measured using cine-CMR datasets for which automated border detection was initially performed using an analogous computational software program (2D CPA MR, TomTec). For both modalities, atrial endocardial reference land-marks were placed at the mitral annuli and posterior wall of the LA for automated border detection, excluding the appendage and pulmonary veins from the LA endocardium. Speckles were traced frame by frame, and LA longitudinal strain curves were generated throughout the cardiac cycle. LA reservoir strain was calculated with QRS onset as the reference point. Echo and CMR were interpreted by experienced physicians to respective modality, diastolic function status, and clinical indices (echo: J.K.; CMR: J.W.W.).

FIGURE 1. Representative Examples of Left Atrial Strain as Quantified Using Echo and CMR.

Note differential peak strain characteristics of normal PALS (A) and impaired PALS (B). (Top row) Echocardiography (echo). (Bottom row) Cardiac magnetic resonance (CMR). PALS = peak atrial longitudinal strain.

To test intra- and interobserver reproducibility of LA strain analyses on both echo and CMR, repeat measurements were performed in a subgroup of 25 randomly selected patients in the study cohort. Repeat analyses were performed at least 90 days after initial strain analyses by physicians blinded to initial results, as well as results of the other modality.

STATISTICAL ANALYSIS.

Continuous variables are summarized as mean ± SD and categorical variables as percent. Comparisons between groups were made using the Student’s t-test (expressed as mean ± SD) or, if multiple comparisons, by analysis of variance for continuous variables. Categorical variables were compared using the chi-square test or the Fisher exact test for fewer than 5 expected outcomes per cell. Bivariate Pearson correlation coefficients and regression analyses were used to evaluate univariable associations between continuous variables: Multivariate modeling was performed via logistic regression, for which CMR and echo 2C and 4C PALS were tested as continuous variables. Overall diagnostic test performance of PALS was evaluated in relation to DD using receiver-operating characteristics curves, which were first used to establish cutoffs in a derivation cohort and then to evaluate test performance in the validation cohort. Agreement between echo and CMR measurements was assessed using the Bland-Altman method. Bland-Altman plots and 95% limits of agreement (LOA) were constructed by fitting a linear regression of the differences on the means. Inter- and intraobserver agreement for each modality was also assessed using Bland-Altman methods, including mean difference and LOA (mean ± SD) between analyses. Inter- and intrarater reliability were calculated using intraclass correlation coefficient (ICC).

Statistical calculations were performed using SPSS 22.0 (SPSS Inc., Chicago, Illinois). Two-sided p < 0.05 was considered indicative of statistical significance.

RESULTS

POPULATION CHARACTERISTICS.

The derivation cohort comprised 157 patients with post-MI who underwent CMR and echo within a 72-hour interval (99% within 1 day), without interval change in medication regimen. Among this at-risk cohort, 73% had DD (grade 1: 23%; grade 2: 40%; grade 3: 10%). Table 1 details clinical and imaging characteristics of the population, including comparisons between patients with and without DD. Prevalence of DD increased in relation to MI severity as evidenced by larger MI size on CMR (p < 0.001), paralleling lower LV ejection fraction, higher wall-motion scores, and larger LV chamber size on both modalities (all p < 0.001).

TABLE 1.

Clinical and Imaging Characteristics

	Overall (n = 157)	LV Diastolic Dysfunction + (n = 115)	LV Diastolic Dysfunction - (n = 42)	p Value
Clinical
Age (yrs)	57 ± 12	59 ± 12	51 ± 10	<0.001
Male	88 (138)	86 (99)	93 (39)	0.41
Body surface area	2.0 ± 0.2	2.0 ± 0.2	2.0 ± 0.2	0.37
Coronary artery disease risk factors
Hypertension	48 (75)	56 (64)	26 (11)	0.001
Hypercholesterolemia	48 (76)	55 (63)	31 (13)	0.008
Diabetes mellitus	24 (38)	27 (31)	17 (7)	0.18
Tobacco use	27 (42)	27 (31)	26 (11)	0.92
Family history	29 (45)	26 (30)	36 (15)	0.24
Previous myocardial infarction	15 (23)	18 (21)	5 (2)	0.04
Previous coronary revascularization	16 (25)	20 (23)	5 (2)	0.03
Percutaneous intervention	15 (23)	18 (21)	5 (2)	0.04
Coronary artery bypass grafting	3 (4)	4 (4)	0 (0)	0.57
Cardiovascular medications
Beta-blocker	94 (148)	93 (107)	98 (41)	0.45
ACE inhibitor/ARB	59 (92)	61 (70)	52 (22)	0.36
Loop diuretic	10 (15)	12 (14)	2 (1)	0.07
HMG-CoA reductase inhibitor	93 (146)	93 (107)	93 (39)	1.00
Aspirin	98 (154)	98 (113)	98 (41)	1.00
Thienopyridine	87 (137)	85 (98)	93 (39)	0.28
Infarct-related artery
Left anterior descending	52 (82)	57 (65)	41 (17)	0.08
Left circumflex	8 (12)	8 (9)	7 (3)	1.00
Right coronary artery	41 (64)	37 (42)	52 (22)	0.07
NYHA functional class				0.007
I	82 (128)	77 (88)	95 (40)
II	14 (22)	17 (20)	5 (2)
II	4 (6)	5 (6)	0 (0)
IV	1 (1)	1 (1)	0 (0)
Cardiac morphology/function/tissue properties
Left ventricle
CMR
Ejection fraction (%)	52.6 ± 11.9	50.3 ± 12.2	59.0 ± 8.3	<0.001
End-diastolic diameter (cm)	5.5 ± 0.6	5.6 ± 0.6	5.4 ± 0.4	0.10
End-diastolic volume (ml)	157.3 ± 43.1	159.3 ± 45.6	151.8 ± 35.2	0.34
End-systolic volume (ml)	77.7 ± 38.2	82.7 ± 40.6	64.0 ± 26.4	0.001
LV myocardial mass (g)	137.6 ± 34.2	138.3 ± 36.4	135.6 ± 27.6	0.67
Myocardial infarct size (%LV)	16.1 ± 10.6	17.8 ± 10.9	11.2 ± 8.1	<0.001
Echo
Ejection fraction (%)	48.5 ± 10.5	46.8 ± 10.7	53.3 ± 8.0	<0.001
End-diastolic diameter (cm)	5.7 ± 0.5	5.7 ± 0.4	5.5 ± 0.4	0.004
End-diastolic volume (ml)	160.0 ± 29.4	164.3 ± 29.8	148.2 ± 26.8	0.002
End-systolic volume (ml)	83.4 ± 29.1	87.7 ± 28.8	71.8 ± 26.8	0.002
Diastolic parameters*
E/e′	13.2 ± 7.2	14.8 ± 7.8	8.9 ± 1.9	<0.001
Septal e′ (cm/s)	7.0 ± 1.8	6.2 ± 1.2	9.2 ± 1.1	<0.001
Lateral e′ (cm/s)	9.1 ± 2.8	8.0 ± 2.3	12.0 ± 1.9	<0.001
TR velocity (m/s)	2.5 ± 0.3	2.6 ± 0.3	2.3 ± 0.3	<0.001
Mitral regurgitation
Severity grade				0.03
I	26 (41)	29 (33)	19 (8)
II	9 (14)	9 (10)	10 (4)
III	6 (10)	9 (10)	0 (0)
IV	1 (1)	1 (1)	0 (0)
Pulmonary arterial pressure (mm Hg)	31.4 ± 6.8	32.1 ± 7.1	29.7 ± 5.9	0.09
Pulmonary hypertension†	24 (27)	26 (21)	19 (6)	0.40

Values are mean ± SD or % (n), unless otherwise indicated. Bold values indicate statistical significance.

^*Number of patients in whom parameter was measured: E/e′: n = 155 (99%); septal e′: n = 156 (99%); lateral e′: n = 155 (99%); TR velocity: n = 148 (94%). †Pulmonary hypertension defined as pulmonary artery systolic pressure >35 mm Hg.

ACE = angiotensin-converting enzyme; ARB = angiotensin receptor blocker; CMR = cardiac magnetic resonance; Echo = echocardiography; HMG-CoA = hydroxymethylglutaryl coenzyme A; LV = left ventricle; NYHA = New York Heart Association; TR = tricuspid regurgitation.

CMR- and echo-derived LA strain.

CMR-derived 4C and 2C PALS were quantifiable in nearly all cases (99%). Echo yielded physiological strain curves in a slightly lesser proportion of cases (98% 4C; 86% 2C); non-diagnostic results were primarily attributable to poor endocardial definition and/or off-axis imaging.

PALS on echo was generally lower than that by CMR, irrespective of whether quantified in 2C (27.8% ± 9.4% vs. 30.1% ± 9.0%) or 4C (25.9% ± 8.4% vs. 28.9% ± 9.2%; both p < 0.01) orientation. Figure 2 shows Bland-Altman and scatter plots for PALS as quantified by the 2 modalities. Figure 3 shows analyses of reproducibility of PALS measured by each modality. CMR and echo yielded similarly good reproducibility, as evidenced by intrareader data yielding near-equivalent ICCs (echo 2C ICC: 0.86 [0.71 to 0.94], 4C: 0.82 [0.63 to 0.92]; CMR 2C: 0.85 [0.69 to 0.93], 4C: 0.88 [0.72 to 0.95]), and LOA (echo 2C: Δ 1.39 [LOA −6.11, 8.88], 4C: Δ 0.82 [LOA −6.45, 8.10]; CMR 2C: Δ 1.21 [LOA −6.26, 8.68], 4C: Δ 1.77 [LOA, −5.58, 9.11]). Intrareader reproducibility also was similar for both echo (2C: ICC 0.91 [0.82 to 0.96], Δ 0.38 [LOA −6.25, 7.01]; 4C: ICC 0.83 [0.65 to 0.92], Δ −0.88 [LOA −8.27, 6.51]) and CMR (2C: ICC 0.82 [0.61 to 0.92], A 1.73 [LOA −5.97, 9.45]; 4C: ICC 0.87 [0.73 to 0.94], Δ 1.10 [LOA −7.25, 9.46]).

FIGURE 2. Scatter and BA Plots Comparing Echo- and CMR-Derived PALS in Equivalent 2C and 4C Orientations.

Note fair correlations (r = 0.57 to 0.58) and limit of agreements. Solid line indicates ordinary least squares (OLS) line of best fit. Dashed line indicates upper and lower 95% confidence limits. On OLS regression, Difference = 3.017 + .03 × Average (for 2C); Difference = 1.94 - 0.19 × Average (for 4C). 2C = 2-chamber; 4C = 4-chamber; BA = Bland-Altman; other abbreviations as in Figure 1.

FIGURE 3. Reproducibility of PALS as Quantified by Echo and CMR.

Echo-derived (top row) and CMR-derived (bottom row) reproducibility as assessed in both 2- and 4-chamber orientation (top: scatter plots; bottom: Bland-Altman plots). Note similar magnitude of good reproducibility yielded by both modalities, as evidenced by near-equivalent correlation coefficients and limits of agreement. Abbreviations as in Figure 1.

LA strain in relation to DD.

Table 2 lists associations between DD and both PALS as well as conventional indices of atrial geometry and function. LA performance was significantly impaired among those with DD whether measured as PALS or fractional area change (p ≤ 0.002), whereas there was no statistically significant difference in atrial area and volume on both echo and CMR. PALS decreased stepwise in relation to severity of DD, with mean PALS 2-fold lower among those with grade III DD versus no DD for both 2C and 4C PALS echo and CMR (CMR 2C: 20.0% vs. 35.1%; p < 0.001; 4C: 15.9% vs. 34.8%; p < 0.001; echo 2C: 16.3% vs. 34.2%; p < 0.001; 4C: 15.4% vs. 29.9%; p < 0.001). In contrast, Figure 4 shows that LA geometric parameters, including echo- and CMR-derived volumes, were significantly different between advanced DD versus no DD (both p ≤ 0.001) but not between groups with lesser degrees of DD (p = NS for all).

TABLE 2.

LA Geometric and Functional Indices

	Overall (n = 157)	LV Diastolic Dysfunction + (n = 115)	LV Diastolic Dysfunction - (n = 42)	p Value
Echo
Functional indices
LA 2C PALS (%)	27.9 ± 9.4	25.7 ± 8.5	34.2 ± 9.1	<0.001
LA 4C PALS (%)	25.9 ± 8.4	24.3 ± 8.0	29.9 ± 8.1	<0.001
LA 2C FAC (%)	38.3 ± 9.9	36.2 ± 9.0	44.3 ±10.0	<0.001
LA 4C FAC (%)	39.5 ± 10.4	37.7 ± 10.3	44.2 ± 9.2	<0.001
Geometric indices
LA 2C area
cm2	20.0 ± 4.3	20.3 ± 4.4	19.4 ± 4.0	0.25
cm2/m2	10.2 ± 2.1	10.3 ± 2.1	9.7 ± 2.2	0.13
LA 4C area
cm2	20.0 ± 4.6	20.4 ± 4.8	19.1 ± 4.0	0.12
cm2/m2	10.3 ± 2.9	10.6 ± 3.1	9.5 ± 2.2	0.05
LA volume
ml	63.3 ± 21.0	64.6 ± 21.9	59.6 ± 17.8	0.20
ml/m2	32.0 ± 10.1	32.7 ± 10.2	29.9 ± 9.6	0.12
CMR
Functional indices
LA 2C PALS (%)	30.3 ± 8.7	28.5 ± 8.5	35.1 ± 7.5	<0.001
LA 4C PALS (%)	28.8 ± 9.2	26.6 ± 8.8	34.8 ± 7.7	<0.001
LA 2C FAC (%)	37.4 ± 8.6	35.9 ± 8.7	41.5 ± 6.7	<0.001
LA 4C FAC (%)	38.1 ± 9.9	36.0 ± 10.1	43.7 ± 6.7	<0.001
Geometric indices
LA 2C area
cm2	22.0 ± 5.5	22.2 ± 5.5	21.5 ± 5.6	0.48
cm2/m2	11.2 ± 2.7	11.3 ± 2.6	10.8 ± 3.0	0.28
LA 4C area
cm2	22.7 ± 5.4	22.7 ± 5.7	22.4 ± 4.6	0.75
cm2/m2	11.5 ± 2.7	11.6 ± 2.8	11.2 ± 2.2	0.40
LA volume
ml	85.2 ± 29.2	85.9 ± 30.3	83.5 ± 26.3	0.66
ml/m2	43.0 ± 13.9	43.6 ± 14.2	41.6 ± 13.0	0.44

Values are mean ± SD. Bold values indicate statistical significance.

2C = 2-chamber; 4C = 4-chamber; FAC = fractional area change; LA = left atrium; CMR = cardiac magnetic resonance; PALS = peak atrial longitudinal strain; other abbreviations as in Table 1.

FIGURE 4. LA Geometric and LA/LV Strain Indices in Relation to DD Grade.

PALS decreased stepwise in relation to DD severity, whereas LA geometric indices significantly differed in only advanced versus no DD. PALS detected even mild versus no DD. DD = diastolic dysfunction; GLS = XXXXX; LA = left atrium; LV = left ventricle; other abbreviations as in Figures 1 and 2.

Invasively quantified hemodynamic data (available in n = 86 of derivation cohort) further demonstrated LA strain to be impacted by LV filling pressure. LVEDP at the time of cardiac catheterization was higher among patients with than in those without impaired 4C PALS on both CMR (20.8 ± 7.6 mm Hg vs. 17.9 ± 5.3 mm Hg; p = 0.046) and echo (21.4 ± 6.8 mm Hg vs. 17.4 ± 6.1 mm Hg; p = 0.006) based on strain cutoffs (Tables 3 and 4). LVEDP was equivalently higher among patients with impaired 2C PALS on CMR (21.6 ± 7.4 mm Hg vs. 16.7 ± 4.6 mm Hg; p < 0.001), with a similar trend when LVEDP was compared among patients grouped based on the presence or absence of impaired 2C PALS on echo (20.8 ± 7.1 mm Hg vs. 18.1 ± 6.2 mm Hg; p = 0.06).

TABLE 3.

Diagnostic Performance of LA Strain for Diastolic Dysfunction (Derivation Cohort)

	AUC (95% CI)	Cutoff	Sensitivity	Specificity	Accuracy	PPV	NPV
Echo PALS
2C	0.78 (0.69-0.87)	29.88	88	42	66	62	76
4C	0.70 (0.62-0.79)	27.01	85	41	65	63	69
CMR PALS
2C	0.75 (0.65-0.84)	32.96	83	42	69	69	62
4C	0.78 (0.69-0.87)	30.98	86	44	68	68	69

Values are %, unless otherwise indicated.

AUC = area under the curve; CI = confidence interval; NPV = negative predictive value; PPV = positive predictive value; other abbreviations as in Tables 1 and 2.

TABLE 4.

Diagnostic Performance of LA Strain for Diastolic Dysfunction (Validation Cohort)

	AUC (95% CI)	Cutoff	Sensitivity	Specificity	Accuracy	PPV	NPV
Echo PALS
2C	0.77 (0.66-0.88)	29.88	94	48	80	80	79
4C	0.78 (0.67-0.88)	27.01	86	58	77	82	64
CMR PALS
2C	0.75 (0.64-0.86)	32.96	94	23	72	73	64
4C	0.77 (0.67-0.88)	30.98	96	35	77	77	79

Values are %, unless otherwise indicated.

Abbreviations as in Tables 1, 2, and 3.

DIAGNOSTIC PERFORMANCE OF LA STRAIN FOR DD.

Tables 3 and 4 list the diagnostic performance of echo- and CMR-derived PALS in providing maximal sensitivity and specificity for DD. Overall diagnostic performance for each parameter was of similar magnitude (area under the curve [AUC]: 0.70 to 0.78) (Table 3). Application of derived cutoffs in a separate validation post-MI cohort (n = 100) yielded reasonable diagnostic performance for discriminating between patients with and those without DD (Table 4), which were similar for 2C echo and CMR (sensitivity 81% and 77%, respectively; specificity 73% and 60%, respectively) and for 4C echo and CMR (sensitivity 82% and 80%, respectively; specificity 54% and 75%, respectively). Positive predictive value for all derived cutoffs was high (≥80%) but negative predictive value low (33% to 52%). Of note, although similar with respect to sex (82 vs. 88% male; p = 0.19), patients in the validation cohort were older (67 ± 13 years vs. 57 ± 12 years; p < 0.001) and more likely to have a history of coronary revascularization (58% vs. 16%; p < 0.001).

MARKERS OF LA STRAIN.

Linear regression analyses were performed in the derivation cohort to evaluate the association of clinical and imaging indices most strongly associated with impaired atrial strain. The 2C and 4C PALS on echo were associated with CMR-quantified MI size independent of age, hypertension, and mitral regurgitation (p < 0.05 for all) (Tables 5 and 6). Substitution of CMR-derived 2C and 4C PALS yielded similar results, demonstrating similar magnitude of association between PALS and MI size.

TABLE 5.

Multivariate Regression for 2C and 4C PALS on Echo

Variable	Univariate Regression		Multivariate Regression
	Regression Coefficient (95% CI)	p Value	Regression Coefficient (95% CI)	p Value
2C PALS
Age*	−2.03 (−3.33 to −0.73)	0.002	−1.48 (−2.69 to −0.27)	0.017
Hypertension	−3.27 (−6.43 to −0.12)	0.04	−0.61 (−3.54 to 2.31)	0.68
Advanced MR†	−8.02 (−11.97 to −4.06)	<0.001	−4.80 (−8.67 to −0.93)	0.016
Myocardial infarct size (%LV)‡	−3.63 (−5.01 to −2.25)	<0.001	−2.95 (−4.34 to −1.55)	<0.001
4C PALS
Age*	−0.99 (−2.08 to 0.10)	0.07	−0.58 (−1.61 to 0.44)	0.26
Hypertension	−3.41 (−6.05 to −0.78)	0.01	−1.59 (−4.13 to 0.95)	0.22
Advanced MR†	−5.54 (−9.13 to −1.95)	0.003	−2.64 (−6.15 to 0.88)	0.14
Myocardial infarct size (%LV)‡	−3.26 (−4.42 to −2.10)	<0.001	−2.84 (−4.04 to −1.65)	<0.001

Bold values indicate statistical significance.

^*Per increments of 10 years.

^†Advanced MR defined as ≥2+.

^‡Per increments of 10%.

MR = mitral regurgitation; other abbreviations as in Table 1, 2, and 3.

TABLE 6.

Multivariate Regression for 2C and 4C PALS on CMR

Variable	Univariate Regression		Multivariate Regression
	Regression Coefficient (95% CI)	p Value	Regression Coefficient (95% CI)	p Value
2C PALS
Age*	−1.56 (−2.67 to −0.44)	0.006	−1.10 (−2.16 to −0.04)	0.04
Hypertension	−4.48 (−7.15 to −1.81)	0.001	−2.61 (−5.22 to 0.01)	0.05
Advanced MR†	−6.08 (−9.72 to −2.44)	0.001	−3.19 (−6.77 to −0.40)	0.08
Myocardial infarct size (%LV)‡	−2.84 (−4.06 to −1.62)	<0.001	−2.24 (−3.48 to −1.00)	<0.001
4C PALS
Age*	−2.04 (−3.20 to −0.89)	0.001	−1.50 (−2.55 to 0.45)	0.006
Hypertension	−4.86 (−7.69 to −2.03)	0.001	−2.25 (−4.86 to 0.36)	0.09
Advanced MR†	−8.54 (−12.30 to −4.78)	<0.001	−4.98 (−8.56 to 1.40)	0.007
Myocardial infarct size (%LV)‡	−3.65 (−4.91 to −2.40)	<0.001	−2.90 (−4.14 to −1.67)	<0.001

Bold values indicate statistical significance.

^*Per increments of 10 yrs.

^†Advanced MR defined as ≥2+.

^‡Per increments of 10%.

Abbreviations as Tables 1, 2, 3, and 5.

Associations between infarct size and impaired PALS paralleled differences with respect to infarct size and LV diastolic function. Global infarct size on CMR was >1.6-fold larger among patients with DD (17.8% vs. 11.2%; p < 0.001). MI size increased stepwise in relation to DD grade (11 ± 8, 14 ± 10, 18 ± 10, and 24 ± 11 %LV; p < 0.001) and was >2-fold larger in patients with advanced compared to those with no DD (p < 0.001).

Prognostic utility for atrial arrhythmias and heart failure hospitalization.

Clinical follow-up was available in 85% (n = 133) of the derivation cohort. Study participants with follow-up were similar to those without follow-up with respect to age and sex, as well as LA volume and PALS on cine-CMR and echo (n = NS).

During mean 4.4 ± 3.8 years after imaging, 8% (n = 13) of patients developed new-onset AF. Echo-derived 2C PALS stratified arrhythmic risk (p = 0.004), paralleling a similar trend for echo 4C PALS (p = 0.08) (Figure 5). Patients with the lowest quartile of echo-derived 2C PALS were at greatest risk for AF (24% vs. 6%; p = 0.004), with a similar trend for echo 4C PALS (18% vs. 7%; p = 0.09). Results were similar when groups were partitioned based on CMR-derived PALS results (20% vs. 5% and 21% vs. 5%; p = 0.02 and p = 0.01, respectively). Although impaired LA strain on both modalities conferred risk for atrial arrhythmias, no such association was found for LA volume on echo or CMR (p = NS) or for conventional assessment based on presence of DD (p = 0.51).

FIGURE 5. Kaplan-Meier Curve for Population-Based Quartiles Partitioned Based on 2C and 4C PALS on Echo Comparing Patients in the Lowest Quartile Versus Remainder.

Note increased atrial fibrillation risk among those with lowest quartile of echo-derived 2C PALS (p = 0.004) with similar trend for 4C PALS (p = 0.08). 2Ch = 2-chamber; 4Ch = 4-chamber; PALS = peak atrial longitudinal strain.

Heart failure-related hospitalization was also associated with impaired LA strain. Echo-derived 2C and 4C PALS were lower among patients with subsequent heart failure hospitalizations (2C: 22.9% ± 12.7% vs. 28.3% ± 8.9%; p = 0.14; 4C: 21.3% ± 8.1% vs. 26.1% ± 8.3%; p = 0.04), paralleling similar results based on CMR-derived 2C and 4C PALS (2C: 21.4% ± 9.5% vs. 31.3 ± 8.3%; 4C: 20.3% ± 9.6% vs. 29.8 ± 8.7%; both p < 0.001).

DISCUSSION

This is the first study to test echo-derived LA strain in relation to CMR with respect to its diagnostic performance for detecting DD. Key findings are as follows. First, LA strain on echo and CMR generally agreed with one another. Strain on both modalities was superior to chamber size for discriminating DD. Second, both echo- and CMR-derived LA strain decreased stepwise in relation to DD severity: Whereas strain significantly differed among all DD grades, atrial geometric indices discriminated only between those with advanced versus no DD. Finally, LA strain by echo and CMR stratified risk for subsequent AF, whereas DD by conventional criteria did not. In addition, echo-derived PALS was lower among patients with subsequent heart failure admissions, paralleling similar results for CMR-derived PALS. Taken together, our findings support the concept that, irrespective of modality, LA strain is a robust marker for both presence and severity of LV DD, which stratifies risk for heart failure and atrial arrhythmias.

Regarding mechanism, impaired LA strain after MI likely stems from several factors. Myocyte necrosis increases LV stiffness, resulting in increased afterload that would be expected to impede LA contractility. Consistent with this, our results demonstrated CMR-quantified MI size to be strongly associated with impaired strain (p < 0.001), irrespective of modality or orientation used for strain quantification. It also is possible that impaired LA strain reflects intrinsic alterations in atrial tissue properties that result from increased LA afterload and ultimately contribute to clinical events such as atrial arrhythmias. Although we are unaware of any study that has performed atrial tissue characterization in the post-MI setting, it should be noted that LA fibrosis has been linked to LA dilation (19) and to the extent of LV fibrosis in patients with nonischemic cardiomyopathies (20), supporting the general notion that LV and LA remodeling are tightly linked.

Our findings regarding strain as a marker of LV diastolic function extend findings of previous research. In a retrospective study of 90 patients with preserved LV function, LA strain via speckle tracking echo categorized DD with good to excellent diagnostic utility (AUC: 0.86 to 0.9) (8). LA strain has also been associated with LVEDP. In a study of 76 patients undergoing echo and left heart catheterization, speckle tracking echo-derived LA strain yielded high diagnostic performance for detecting LVEDP >5 mm Hg (AUC: 0.76) (21). However, although both studies assessed echo-derived strain in relation to diastolic function and LV filling parameters, neither tested its performance in relation to CMR, the established reference modality for atrial dimensions and volumes (9). Moreover, DD in those studies were graded in accordance with previous guidelines, thereby limiting generalizability. Our study, utilizing updated consensus recommendations, demonstrates that performance of CMR and echo for discriminating DD are similar (echo AUC: 0.70 to 0.78, CMR: 0.75 to 0.78) and that LA strain is closely related to LVEDP. These findings support the concept that the relationship between LA strain and DD is modality independent, and that despite known differences in endocardial delineation between echo and CMR, such differences do not substantially impact diagnostic accuracy of LA strain for discriminating DD and LV filling pressure.

Our study also demonstrates that although LA strain is significantly different between DD grades, LA geometric indices increase only with more advanced DD. Although LA remodeling is a known characteristic feature of DD, our findings suggest that LA enlargement represents a cumulative effect of elevated LV filling pressure over time, whereas LA strain may be more useful for detecting earlier diastolic alterations. In this context, we note that our cohort comprised patients with recent MI, and in this setting, DD may occur in the short term despite the absence of marked atrial geometric alterations. These findings are in accordance with previous studies demonstrating that in the setting of increased LV pressure, LA performance likely already is impaired before the LA begins to enlarge, and this impairment in atrial performance may be manifested as decrements in LA strain. For example, in a study of 329 patients with DD, 23% had impaired LA strain but normal LA geometry (22). Similarly, in a cohort of patients with aortic stenosis, LA performance was associated with AF even after controlling for LA dimension, and altered LA performance was associated in AF even among patients with normal LA geometry (23). Consistent with this, our results demonstrated strain to be superior to both atrial geometry and DD for predicting AF, thus adding to a growing body of literature indicating LA strain provides incremental insights beyond conventional parameters of geometric remodeling. Given our current data showing LA strain to vary in relation to graded severity of DD, future research is warranted to examine the manner in which LA strain impairment relates to individual components of diastolic performance, including pattern and magnitude of pulmonary vein inflow velocity.

It is important to recognize that although CMR- and echo-quantified LA strain demonstrated equivalently good reproducibility and similar relationships with echo-evidenced DD and clinical outcomes, substantial differences were evident with respect to absolute strain values, supporting the notion that CMR- and echo-derived strain results (and resultant normative cutoffs) are not equivalent between modalities. Although we speculate that differences between CMR- and echo-derived strain may stem from differences in temporal resolution and/or endocardial definition between modalities, further research is warranted to systematically test these concepts.

STUDY LIMITATIONS.

Regarding clinical application, although our data provide key validation of LA strain as a marker of DD, strain quantification requires dedicated off-line analysis, and the influence of vendor dependence may limit the applicability of thresholds validated in this study across vendors. Second, it is important to recognize that no patients in our cohort had AF at the time of baseline imaging. In this context, current findings are applicable to patients in sinus rhythm with intermediate to high probability of DD, such as those with previous MI and/or established coronary artery disease. It is also important to recognize that our study population underwent imaging via a dedicated research protocol that included pre-specified assessment of key indices necessary for diastolic function. Although the tailored nature of our dataset was well suited to elucidate physiological determinants and clinical manifestations of impaired LA strain, our ability to categorize diastolic performance among all patients in our study cohort likely is not reflective of echo performance in routine clinical practice and may explain discordance with previous research studies in which variable proportions of patients have been deemed indeterminate with respect to classification of diastolic function. Additionally, although our results found LA strain on both CMR and echo to be impaired among patients with elevated LVEDP, invasive catheterization was not uniformly performed in all patients in both our derivation and validation cohorts, thus prohibiting us from testing diagnostic utility of strain cutoffs as noninvasive markers of elevated LVEDP. Future research is warranted to test this issue, as well as the relative prognostic utility of LA strain, LA geometric indices, and graded severity of diastolic function in larger-scale epidemiological cohorts.

CONCULSIONS

The findings of this study demonstrate that echo-derived global longitudinal LA strain parallels the results of CMR and yields incremental diagnostic utility versus conventional LA geometric indices for stratifying presence and severity of DD, as well as improved prognostic utility for predicting AF after MI. Further studies are warranted to identify the mechanistic determinants of impaired strain and to determine whether temporal alterations in LA strain predict clinical symptoms and outcomes in broad at-risk populations.

CENTRAL ILLUSTRATION. LA Strain Is a Robust Marker for Presence and Severity of LV DD That Stratifies Atrial Arrhythmic Risk.

(Top) The study population. (Middle, top) Representative examples of LA strain quantified by echo and CMR. (Middle, bottom) LA geometric and strain indices in relation to diastolic dysfunction. (Bottom) Impaired LA strain as a risk marker for atrial fibrillation. 2C = 2-chamber; CMR = cardiac magnetic resonance; DD = diastolic dysfunction; ECHO = echocardiography; LA = left atrium; MI = myocardial infarction; PALS = peak atrial longitudinal strain.

PERSPECTIVES.

COMPETENCY IN MEDICAL KNOWLEDGE:

Left atrial strain derived from echocardiogram and magnetic resonance imaging yields incremental diagnostic utility compared to geometric indices alone for discrimination of diastolic dysfunction and improves prognostic stratification of atrial fibrillation.

TRANSLATIONAL OUTLOOK:

Future studies are needed to assess whether atrial strain can be used as a treatment marker and whether temporal changes in atrial strain predicts clinical symptoms and outcomes in a broad patient population.

ABBREVIATIONS AND ACRONYMS

2C

2-chamber

4C

4-chamber

AF

atrial fibrillation

AUC

area under the curve

CMR

cardiac magnetic resonance

DD

diastolic dysfunction

Echo

echocardiography

ICC

intraclass correlation coefficient

LA

left atrium

LOA

limits of agreement

LV

left ventricle

LVEDP

left ventricular end-diastolic pressure

MI

myocardial infarction

PALS

peak atrial longitudinal strain

TR

tricuspid regurgitation