- Email: [email protected]
Prediction of Ventricular Arrhythmias With Left Ventricular Mechanical Dispersion
JACC: CARDIOVASCULAR IMAGING
VOL.
-, NO. -, 2019
ª 2019 BY THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION PUBLISHED BY ELSEVIER
ORIGINAL RESEARCH
Prediction of Ventricular Arrhythmias With Left Ventricular Mechanical Dispersion A Systematic Review and Meta-Analysis Hiroshi Kawakami, MD, PHD,a Nitesh Nerlekar, MBBS, MPH,a Kristina H. Haugaa, MD, PHD,b Thor Edvardsen, MD, PHD,b Thomas H. Marwick, MBBS, PHD, MPHa
ABSTRACT OBJECTIVES The aim of this study was to assess the association between left ventricular mechanical dispersion (LVMD) and the incidence of ventricular arrhythmias (VAs). BACKGROUND Recent, mainly single-center, studies have demonstrated that LVMD assessed using speckle tracking might be a powerful marker in risk stratification for VA. A systematic review and meta-analysis provides a means of understanding the prognostic value of this parameter, relative to other parameters, the most appropriate cutoff for designating risk. METHODS A systemic review of studies reporting the predictive value of LVMD for VA was undertaken from a search of MEDLINE and Embase. VA events were defined as sudden cardiac death, cardiac arrest, documented ventricular tachyarrhythmia, and appropriate implantable cardioverter-defibrillator (ICD) therapy. Hazard ratios were extracted from univariate and multivariate models reporting on the association of LVMD and VA and described as pooled estimates with 95% confidence intervals. In a meta-analysis, the predictive value of LVMD was compared with that of left ventricular ejection fraction and global longitudinal strain. RESULTS Among 3,198 patients in 12 published studies, 387 (12%) had VA events over follow-up ranging from 17 to 70 months. Patients with VAs had greater LVMD than those without (weighted mean difference 20.3 ms; 95% confidence interval: 27.3 to 13.2; p < 0.01). Each 10 ms increment of LVMD was significantly and independently associated with VA events (hazard ratio: 1.19; 95% confidence interval: 1.09 to 1.29; p < 0.01). The predictive value of LVMD was superior to that of left ventricular ejection fraction or global longitudinal strain. CONCLUSIONS LVMD assessed using speckle tracking provides important predictive value for VA in patients with a number of cardiac diseases and appears to have superior predictive value over left ventricular ejection fraction and global longitudinal strain for risk stratification. (J Am Coll Cardiol Img 2019;-:-–-) © 2019 by the American College of Cardiology Foundation.
From the aBaker Heart and Diabetes Institute, Melbourne, Australia; and the bCentre of Cardiological Innovation, Department of Cardiology, Oslo University Hospital, Oslo, Norway. Dr. Marwick has received research support from the National Health and Medical Research Council (grants 1119955, 1080582, 1059738, and 1149692) and GE Medical Systems for an ongoing research study on the use of strain for the assessment of cardiotoxicity. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Allan L. Klein, MD, served as the Guest Editor for this paper. Manuscript received December 10, 2018; revised manuscript received February 12, 2019, accepted March 13, 2019.
ISSN 1936-878X/$36.00
https://doi.org/10.1016/j.jcmg.2019.03.025
2
Kawakami et al.
JACC: CARDIOVASCULAR IMAGING, VOL.
C
ABBREVIATIONS AND ACRONYMS CI = confidence interval GLS = global longitudinal strain
ardiac imaging is widely used for
English, published up to July 17, 2018. The study was
assessing
prospectively registered with the Prospero database
sudden
cardiac
death
(SCD) risk (1,2), but the prediction
ICD = implantable cardioverter-defibrillator
STUDY SELECTION. Studies involving prediction of
nonsustained
VA events were included in this analysis on the basis
and
sustained
ventricular
of
the
availability
of
LVMD
assessed
using
SCD remains a challenge. Severe left ventric-
2-dimensional speckle-tracking transthoracic echo-
ular (LV) dysfunction (LV ejection fraction
cardiography. LVMD was defined as the SD of time
[LVEF] <35%) is a useful predictor of VA in
LV = left ventricular
(CRD42018104240).
of ventricular arrhythmias (VAs) including tachycardia and ventricular fibrillation and
HR = hazard ratio
-, NO. -, 2019 - 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
from Q/R on electrocardiography to peak negative
patients with structural heart disease (3–6)
strain from each LV segment (Figure 1). LVMD was
and is the most widely used marker for pre-
evaluated as a predictor of VA events during follow-
dicting VA events. However, LVEF has
up. Studies in which mechanical dispersion was
mechanical dispersion
several limitations, including the influence
assessed using tissue Doppler imaging, 3-dimensional
SCD = sudden cardiac death
of heart rate and translational motion, geo-
strain imaging, or nonechocardiographic imaging,
VA = ventricular arrhythmia
metric
reproducibility
including cardiac magnetic resonance imaging, were
(7,8). In addition, most cases of SCD occur
excluded because of technical differences. We also
LVEF = left ventricular ejection fraction
LVMD = left ventricular
assumptions,
and
in patients with LVEF better than 35% (9,10).
excluded studies in which strain analysis was
Two-dimensional speckle tracking uses myocardial
restricted to the right ventricle. Abstracts without
acoustic reflections and interference patterns to
complete published papers, case reports, review pa-
measure myocardial deformation and has emerged as
pers, editorials, and letters were also excluded.
a powerful tool for detecting LV dysfunction (7,8,11). In 2010, Haugaa et al. (12,13) reported that LV mechanical dispersion (LVMD) assessed using speckle tracking strain was a useful marker for predicting VA in a variety of cardiovascular diseases (12,13). Subsequently, several studies have assessed the association
ENDPOINTS. The primary endpoint was VA events,
defined as SCD, cardiac arrest, documented ventricular fibrillation, documented ventricular tachycardia (sustained and/or nonsustained), and appropriate ICD therapy such as antitachycardia pacing and/or shock.
between LVMD and VA, but these studies were mostly
DATA
single-center analyses with relatively small sample
extracted and reviewed by 2 investigators. Discrep-
EXTRACTION. Data
sizes. The aim of this study was to clarify the associ-
ancies were manually reviewed and resolved by
ation between LVMD and the incidence of VA. We
consensus. Among the included publications, we
performed a systematic review and combined the
extracted the following data for systemic review:
data using meta-analytic techniques to strengthen
study design, study population, demographic char-
the level of evidence and provide deeper insight into
acteristics, follow-up period, outcomes, and echo-
the issue of LVMD in the incidence of VA in the pa-
cardiographic parameters including LVEF, GLS, and
tients with cardiac disease. We also assessed whether
LVMD. We contacted the investigators if study data
LVMD provided superior predictive information on
were incomplete.
were
independently
the incidence of VA compared with LVEF and global
STATISTICAL ANALYSIS. Data for continuous vari-
longitudinal strain (GLS).
ables were extracted as a weighted mean of the studies reporting continuous variable data. The
METHODS
weighted mean difference for LVMD between patients with and without VA events was calculated.
SEARCH STRATEGY. A systematic electronic search
Pooled hazard ratios (HRs) and 95% confidence in-
of published research was conducted using the
tervals (CIs) were computed using random-effect
MEDLINE and Embase databases in adherence with
models of the association of mechanical dispersion
the Preferred Reporting Items for Systematic Reviews
with VA events, adjusted for clinical differences be-
and Meta-Analyses guidelines (14). The search terms
tween the populations. Some studies reported HRs in
used the Medical Subject Headings and key words
LVMD relative to a 1-ms increase in LVMD; hence, we
“strain,”
“dyssynchrony,”
rescaled the HRs to a 10-ms increase in all studies.
“dispersion,” “ventricular arrhythmia,” “ventricular
Furthermore, we compared the pooled HRs for LVMD,
tachycardia,” “ventricular fibrillation,” “sudden car-
LVEF, and GLS to compare their predictive value of
diac death,” and “implantable cardioverter defibril-
VA events. To allow direct comparison of their
lation.” Additional strategies involved reference
impact, all HRs were rescaled to a 1-SD increase in
searches to identify further relevant studies. We
LVMD, LVEF, and GLS. Forest plots were constructed
limited the search to studies of adult humans, in
to show the overall effect of each parameter.
“speckle
tracking,”
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019
Kawakami et al.
- 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
F I G U R E 1 Calculation of Left Ventricular Mechanical Dispersion Using Speckle-Tracking Technique
Standard Deviation of Time to Peak Negative Strain
GS=-18.5%
Time to Peak Negative Strain
White arrows indicate peak negative strain. Left ventricular mechanical dispersion (LVMD) was defined as the SD of time from Q/R on electrocardiography to peak negative strain from each left ventricular segment. AVC ¼ aortic valve closure; GS ¼ global strain.
Heterogeneity was described using the I2 statistic,
Figure 2. From the initial 244 papers identified from
which was quantified as low (<25%), moderate
the search strategy and additional papers, 38 studies
(25% to 75%), or high (>75%) (15). Publication bias was
were considered to be potentially eligible, following
assessed using funnel plots and the Egger and Begg
the exclusion of duplicates and screening by title and
test. Study quality was assessed using the Newcastle-
abstract. After full review, 12 studies were included in
Ottawa scale (0 to 9 points) using the methodology
the final analysis (13,18–28) (Table 1). The reasons for
described by Downs and Black (16). Furthermore, we
exclusion are provided in Supplemental Table 1.
also identified the documentation of blinded perfor-
The characteristics of the 12 studies included in the
mance of data collection and analysis, description of
systemic review and meta-analysis are presented in
strain calculation technique, and interobserver and
Table 1. All studies were observational (6 prospective
intraobserver variability as important quality metrics
and 6 retrospective), and half were multicenter.
(17). Statistical analysis was performed using RevMan
Among 3,198 patients (weighted mean age 63 years;
5.3 (Cochrane Information Management System,
weighted mean 30% women), the most prevalent
Oxford, United Kingdom) and Stata (StataCorp, Col-
cardiac
lege Station, Texas), with 2-tailed p values <0.05
(n ¼ 2,634).
considered to indicate statistical significance.
condition
was
ischemic
heart
disease
Echocardiographic data at baseline are presented in Table 1. Most studies solely used GE echocardio-
RESULTS
graphic platforms (n ¼ 11) and proprietary software (n ¼ 10). Among the 9 papers reporting frame rates for
STUDY SELECTION AND PATIENT CHARACTERISTICS.
speckle-tracking imaging, most were >50 frames/s.
The process of study selection is presented in
Eleven studies presented data on LVEF (weighted
3
4
Kawakami et al.
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019 - 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
F I G U R E 2 Flowchart of the Study Selection Process
Records identified through database search: Medline (56), EMBASE (167) (n = 223)
Additional records identified through other sources (n = 21)
Records after duplicates removed (n = 172)
Records screened (n = 172)
Excluded by title and abstract (n = 134)
Full-text studies assessed for eligibility (n = 38)
Full-text articles excluded (please see supplemental table 1) (n = 26)
Studies included in quantitative synthesis (Meta-analysis) (n = 12)
The reasons for exclusion are provided in Supplemental Table 1.
mean
(weighted
continuous variable (per 10-ms increase) for VA
mean 14.1%). The reported means of LVMD ranged
46%),
and
11
reported
GLS
events was available from 9 studies and was 1.26
from 43 to 112 ms (weighted mean 60 ms). Repro-
(95% CI: 1.14 to 1.39; p < 0.01; I 2 ¼ 84%). In the
ducibility of strain imaging was demonstrated in
multivariate analysis, the pooled adjusted HR of
11 studies, in 9 of which expressed as an intraclass
LVMD as a continuous variable (per 10-ms increase)
correlation coefficient (Table 2). The range of intra-
for VA events was available from 6 studies and was
observer and interobserver variability was from
1.19 (95% CI: 1.09 to 1.29; p < 0.01; I 2 ¼ 48%). LVMD
0.92 to 0.99 and from 0.90 to 0.99, respectively, for
was significantly and independently associated with
GLS and from 0.86 to 0.99 and from 0.78 to 0.96,
VA events. The association of LVEF and GLS with VA
respectively, for LVMD.
events in univariable and multivariable models is
RELATIONSHIP BETWEEN LVMD AND VA. The defi-
nitions of VA events are provided in Table 1. Over 17 to 70 months of follow-up, 387 patients (12%) had VA events. In patients with VA events, LVMD was greater than in those without VA events (Figure 3A). LVMD in patients with VAs was 20.3 ms shorter than in those without (95% CI: 27.3 to 13.2 ms; p < 0.01; I 2 ¼ 76%) (Figure 3B).
shown in Supplemental Figures 1 and 2. Figure 5 shows the results of comparison of the predictive value for VA events between LVMD and LVEF or GLS. In these analyses, we selected studies that reported multivariate HRs for both LVMD and LVEF or GLS. In addition, to facilitate comparison, we used rescaled HRs per 1-SD increase. On the basis of 4 studies, LVMD (HR: 1.59; 95% CI: 1.28 to 1.97; p < 0.01; I 2 ¼ 34%) and LVEF (HR: 0.81; 95% CI: 0.68 to 0.96;
PREDICTIVE VALUE OF LVMD FOR VA EVENTS.
p ¼ 0.01; I2 ¼ 0%) were independently associated with
All studies provided risk variables of LVMD for VA
VA events, and the predictive value of LVMD was
events; 9 studies provided HRs and 3 studies pro-
greater than that of LVEF (Figure 5). Likewise, the
vided odds ratios. Figure 4 shows the association of
comparison between LVMD and GLS on the basis of
LVMD with VA events in univariate and multivariate
5 studies revealed that LVMD was independently
models. The pooled unadjusted HR of LVMD as a
associated with VA events (HR: 1.46; 95% CI: 1.18 to
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019
Kawakami et al.
- 2019:-–-
5
Mechanical Dispersion and Ventricular Arrhythmias
T A B L E 1 Study design, patient characteristics, details of ventricular arrhythmic events, and echocardiographic parameters
Year
n
Design
Age, years
Female, %
Population
IHD, %
Definition of VA events
Follow-up period, months
Haugaa et al. (13)
Publication
2010
85
Prosp
63 10
15
Post-MI
100
Appropriate ICD therapy
28 (range; 07-66)
Haugaa et al. (18)
2012
94
Prosp
49 12
19
NICM
0
Appropriate ICD therapy/VT/VF/ SCD
22 (range; 1-46)
Ersbøll et al. (19)
2013
988
Prosp
63 12
28
Post-MI
100
Appropriate ICD therapy/VT/VF/ SCD
30 (23.5-32.7)
Haugaa et al. (20)
2013
569
Prosp
61 11
34
Post-MI
100
VT/VF/SCD
30 (18)
Kosiuk et al. (21)
2015
20
Retro
62 11
25
NICM
0
Appropriate ICD therapy /VT/VF
70 40
Leong et al. (22)
2015
206
Retro
67 (57-73)
13
Post-MI
100
Appropriate ICD therapy
24 (7.8-47)
Nguyen et al. (23)
2015
467
Retro
68 14
39
Post-MI
100
VT
25 (range; 6-43)
Negishi et al. (24)
2016
124
Retro
56 13
46
NICM
0
Appropriate ICD therapy
46 (26-72)
Matsuzoe et al. (25)
2016
72
Retro
58 15
18
Unselected
32
Appropriate ICD therapy
17 (range; 0.2–72.5)
Hasselberg et al. (26)
2016
170
Prosp
66 10
24
HF with CRT
48
Appropriate ICD therapy/VT/VF/ Cardiac arrest
23 4
Mornos et al. (27)
2017
340
Prosp
63 12
33
Unselected
63
VT/VF/SCD
36 9
Candan et al. (28)
2017
63
Retro
49 14
22
HCM
0
Appropriate ICD therapy
22 7
TABLE 1 Continued VA events
QRS duration, ms
LVEF, %
GLS, %
LVMD, ms
Vendor
Software (Version)
FR, frames/s
Haugaa et al. (13)
Publication
38 (45%)
100 19
34.4 10.1
-10.7 3.9
69.0 20.2
GE
EchoPAC (NR)
63 23
Haugaa et al. (18)
12 (13%)
118 34
36.7 12.6
-11.5 5.0
61.4 21.2
GE
EchoPAC (NR)
>70
Ersbøll et al. (19)
34 (3%)
98 19
50.7 10.4
-13.6 3.5
56.8 15.0
GE
EchoPAC (BT11.1.0)
>60
Haugaa et al. (20)
15 (3%)
95 16
54.8 11.2
-18.1 3.7
42.6 17.2
GE
EchoPAC (NR)
NR
Kosiuk et al. (21)
9 (45%)
102 14
32 6
NR
70 29
GE
MatLab
>60
Leong et al. (22)
75 (36%)
121 27
38.8 10.5
-11.4*
87.7 36.5
GE
EchoPAC (11.0.0)
NR
Nguyen et al. (23)
51 (11%)
133 35
43.1 5.8
-14.4 3.7
47.2 13.5
GE
EchoPAC (BT11)
60-100
Negishi et al. (24)
36 (29%)
NR
31.4 9.9
-9.1 3.5
103 43
GE
EchoPAC (11.0.0)
50 20
Matsuzoe et al. (25)
34 (47%)
113 27
52.2 12.0
-11.2 3.4
83.1 28.6
GE/Toshiba
Ultra Extend
45
Hasselberg et al. (26)
18 (11%)
165 22
26 9
-8.2 3.9
112*
GE
EchoPAC (NR)
>50
Mornos et al. (27)
48 (14%)
NR
41.7 12.3
-17.1 6.5
44.3 32.3
GE
EchoPAC (NR)
NR
Candan et al. (28)
17 (27%)
NR
NR
-12.1 3.4
66.4 19.4
GE
EchoPAC (NR)
50-70
Data are expressed as mean SD, median (interquartile range), median (range), or number (%). *Mean calculated from their original figure or table. CRT ¼ cardiac resynchronization therapy; FR ¼ frame rate; GE ¼ general electric; GLS ¼ global longitudinal strain; HF ¼ heart failure; HCM ¼ hypertrophic cardiomyopathy; ICD ¼ implantable cardioverter defibrillator; IHD ¼ ischemic heart disease; LVEF ¼ left ventricular ejection fraction; LVMD ¼ left ventricular mechanical dispersion; MI ¼ myocardial Infarction; NICM ¼ non-ischemic cardiomyopathy; NR ¼ not reported; Prosp ¼ prospective; Retro ¼ retrospective; SCD ¼ sudden cardiac death; VA ¼ ventricular arrhythmia; VF ¼ ventricular fibrillation; VT ¼ ventricular tachycardia.
1.82; p < 0.01; I 2 ¼ 50%) but not GLS (HR: 1.35; 95% CI:
T A B L E 2 Reproducibility of GLS and Mechanical Dispersion
0.84 to 2.14; p ¼ 0.21; I 2 ¼ 69%) (Figure 5).
GLS
Supplemental Figures 1 and 2 repeat these comparisons against each 1% change of EF and GLS.
First Author (Ref. #)
Method
Mechanical Dispersion
Intraobserver Interobserver Intraobserver Interobserver Variability Variability Variability Variability
The Central Illustration shows the ability of LVMD
Haugaa et al. (13)
ICC
0.98
0.98
0.86
0.81
to identify VA events. Seven studies described the
Haugaa et al. (18)
ICC
0.98
0.95
0.86
0.78
results of receiver-operating characteristic curve an-
Ersbøll et al. (19)
Bland-Altman analysis (ms)
-0.7 2.5*
-0.05 1.3*
1.4 7.5*
1.3 1.08*
Haugaa et al. (20)
ICC
0.92
0.90
0.89
0.82
Kosiuk et al. (21)
NR
NR
NR
NR
NR
Leong et al. (22)
ICC
NR
0.94
NR
0.96 6.1 3.5†
alyses and the accuracy of mechanical dispersion. The range of the optimal cutoff value, area under the curve, sensitivity, specificity, and accuracy were 47 to 101 ms, 0.69 to 0.84, 38% to 91%, 55% to 92%%, and
Nguyen et al. (23) Mean difference
3.5 2.1†
5.5 3.4†
4.2 2.3†
65% to 79%, respectively (Central Illustration).
Negishi et al. (24)
ICC
0.99
0.99
0.99
0.93
ICC
NR
NR
0.940
0.917
QUALITY ASSESSMENT AND PUBLICATION BIAS
Matsuzoe et al. (25) Hasselberg et al. (26)
ICC
0.94
0.92
NR
NR
ANALYSIS. Study quality was assessed using the
Newcastle-Ottawa scale (Supplemental Table 2). All included studies had high scores on the Newcastle-
Mornos et al. (27)
ICC
0.93
0.92
0.91
0.90
Candan et al. (28)
ICC
0.92
0.94
0.96
0.93
Ottawa scale (all studies scored 8 points or more). In addition, all studies defined the study objective, described the outcomes and confounders, and outlined the main findings (Supplemental Table 3). As for
Values are % or mean SD. *Mean differences in agreement. †Determined as the difference between the 2 sets of observations divided by the mean of the observations and expressed as a percentage. ICC ¼ intraclass correlation coefficient; NR ¼ not reported.
Kawakami et al.
6
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019 - 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
F I G U R E 3 Difference in Left Ventricular Mechanical Dispersion Between Patients With and Those Without Ventricular Arrhythmias
A (ms) 140
51.5 ± 18.7
78.6 ± 27.5
VA (–) N = 2526
VA (+) N = 296
120 100 80 60 40 20 0
B
VA (+)
VA (–) Study
Mean
SD
N
Mean
SD
N
Candan et al28 Ersbøll et al19 Haugaa et al13 Haugaa et al18 Haugaa et al20 Kosiuk et al21 Matsuzoe et al25 Mornos et al27 Negishi et al24 Nguyen et al23
62.4 56.3 56 56 42 50 73.9 39.7 106 45.3
17.1 15.5 13 18 17 16 22.6 33.1 46 13.1
46 954 47 82 554 9 38 292 88 416
77.1 70.7 85 98 63 84 93.3 72.3 95 62.7
21.8 29.7 29 43 25 31 31.3 27.6 33 16.6
2526
Total (95% CI)
Weight
Mean Difference, 95% CI
17 34 38 12 15 11 34 48 36 51
10.6% 11.4% 11.4% 5.3% 10.0% 6.4% 10.0% 12.1% 9.1% 13.9%
–14.70 [–26.18, –3.22] –14.40 [–24.43, –4.37] –29.00 [–38.94, –19.06] –42.00 [–66.64, –17.36] –21.00 [–33.73, –8.27] –34.00 [–55.09, –12.91] –19.40 [–32.14, –6.66] –32.60 [–41.28, –23.92] 11.00 [–3.44, 25.44] –17.40 [–22.13, –12.67]
296
100.0%
–20.25 [–27.30, –13.20]
Heterogeneity: Tau2 = 87.57; Chi2 = 36.87, df = 9 (P < 0.0001); I2 = 76% Test for overall effect: Z = 5.63 (P < 0.00001)
Mean Difference, 95% CI
–50
–25 VA (–)
0
25
50
VA (+)
(A) Mean left ventricular mechanical dispersion (LVMD) in patients with and those without ventricular arrhythmias (VAs). The colored bars show the weighted mean LVMD, and gray dots show the mean LVMD in the original studies in the 2 groups. (B) The forest plot displays the weighted mean differences and 95% confidence intervals (CIs) for difference between patients with and without VAs.
echocardiographic quality, all studies described the
(Supplemental Figure 3). This might be because
strain imaging protocol. In 10 studies, echocardio-
baseline LVMD varies depending on patients’ cardiac
graphic parameters were calculated by blinded re-
conditions.
searchers.
All
but
1
study
evaluated
the
reproducibility of GLS and/or mechanical dispersion
DISCUSSION
(Table 2). Publication bias was explored using a funnel plot
There are 3 main findings of the present study. First,
and the Egger test for all studies (Supplemental
LVMD was significantly and independently associated
Figures 3 to 7). No evidence of significant publication
with VA events. Second, a 1-SD change in LVMD was a
bias was identified in the forest plots displaying the
stronger predictor of VA events compared with the
summary HRs for increasing association of each
same changes in LVEF and GLS. Third, patients with
parameter with VA (Supplemental Figures 4 to 7).
VA events had significantly greater mechanical
However, there was evidence of publication bias in
dispersion compared with those without VA events,
the comparison of weighted mean difference of LVMD
with 60 ms being an acceptable cutoff LVMD value for
between the patients with and without VA events
predicting VA events.
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019
Kawakami et al.
- 2019:-–-
7
Mechanical Dispersion and Ventricular Arrhythmias
F I G U R E 4 Left Ventricular Mechanical Dispersion as a Predictor of Ventricular Arrhythmia
A Study Ersbøll et al19 Hasselberg et al26 Haugaa et al13 Haugaa et al18 Haugaa et al20 Leong et al22 Matsuzoe et al25 Mornos et al27 Negishi et al24
Log [Hazard Ratio]
SE
0.3221 0.006 0.2624 0.3293 0.5878 0.131 0.0953 0.1989 –5.2983
0.0546 0.0456 0.059 0.0708 0.1024 0.0229 0.1001 0.0528 4.457
Total (95% CI)
Hazard Ratio, 95% CI
Weight 13.3% 13.9% 12.9% 12.0% 9.5% 15.2% 9.7% 13.4% 0.0%
1.38 [1.24, 1.54] 1.01 [0.92, 1.10] 1.30 [1.16, 1.46] 1.39 [1.21, 1.60] 1.80 [1.47, 2.20] 1.14 [1.09, 1.19] 1.10 [0.90, 1.34] 1.22 [1.10, 1.35] 0.01 [0.00, 31.10]
100.0%
1.26 [1.14, 1.39]
Heterogeneity: Tau2 = 0.02; Chi2 = 50.12, df = 8 (P < 0.00001); I2 = 84% Test for overall effect: Z = 4.55 (P < 0.00001)
B
Log [Hazard Ratio]
SE
Ersbøll et al19 Haugaa et al13 Haugaa et al18 Haugaa et al20 Leong et al22
0.1398 0.2546 0.1823 0.5306 0.1133
0.0662 0.0691 0.0779 0.1777 0.0281
19.9% 19.0% 16.7% 4.8% 33.9%
1.15 [1.01, 1.31] 1.29 [1.13, 1.48] 1.20 [1.03, 1.40] 1.70 [1.20, 2.41] 1.12 [1.06, 1.18]
Mornos et al27
0
0.1606
5.7%
1.00 [0.73, 1.37]
100.0%
1.19 [1.09, 1.29]
Study
Total (95% CI)
Hazard Ratio, 95% CI
Hazard Ratio, 95% CI
Weight
0.5
0.7
1
Covariates in Multivariable Model
1.5
2
Hazard Ratio, 95% CI
Age, QRS, LVEDV, GLS Age, Sex, LVEF, GLS Age, QRS, LVEF, GLS Age, LVEF, GLS Age, QRS, Cr, Revascurized infarct-related artery Time from MI, LVESVI, LV scare, VT inducible, LVEF, GLS Age, LBBB, E/e’, LVEF, GLS
Heterogeneity: Tau2 = 0.00; Chi2 = 9.53, df = 5 (P = 0.09); I2 = 48% Test for overall effect: Z = 4.10 (P < 0.0001)
0.5
0.7
1
1.5
2
(A) Univariate analysis for prediction of ventricular arrhythmia (VA). (B) Multivariate analysis for prediction of VA. The forest plots display the summary hazard ratios per 10-ms increase and 95% confidence intervals (CIs) for increasing association of left ventricular (LV) mechanical dispersion with VA. Cr ¼ creatinine; GLS ¼ global longitudinal strain; LBBB ¼ left bundle branch block; LVEDV ¼ left ventricular end-diastolic volume; LVEF ¼ left ventricular ejection fraction; LVESVI ¼ left ventricular end-systolic volume index; MI ¼ myocardial infarction; VT ¼ ventricular tachycardia.
ROLE OF ECHOCARDIOGRAPHY IN RISK STRATIFICATION
deaths in patients with heart failure with preserved
FOR VA EVENTS. About one-quarter of patients among
ejection fraction (32). These observations suggest that
emergency medical services–treated out-of-hospital
better parameters than LVEF are necessary for risk
cardiac arrests have an initial rhythm of ventricular
stratification for VA.
fibrillation or ventricular tachycardia (29). Although the genesis of these arrhythmias is multifactorial,
LVMD AS A RISK MARKER OF VA. Over the past 15
heterogeneous
by
years, GLS has emerged as a robust tool for the
myocardial fibrosis has been thought to be the main
assessment of global LV function and marker of
substrate for VA in patients with structural heart
subclinical LV dysfunction (7,8). The time course of
disease (30,31). Cardiac imaging has played an
segmental strain has also permitted the assessment
important role in assessing the structural substrate
of LVMD, and several studies have demonstrated
for arrhythmogenesis (1,2), especially echocardiogra-
the value of LVMD for the prediction of VA in a wide
phy (because it is noninvasive, highly accessible, and
range of conditions, from long-QT syndrome to
portable), and usually through measurement of LVEF
cardiac resynchronization therapy (12,33,34). The
(3–6). However, LVEF is not a reliable predictor, as
performance of meta-analysis is useful when the
many patients with VA have preserved LVEFs (9,10).
evidence is still relatively underdeveloped, as in
Indeed, SCD accounts for 30% to 40% of cardiac
the case of LVMD, and permits confirmation of
ventricular
activation
caused
8
Kawakami et al.
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019 - 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
F I G U R E 5 Comparison of Predictive Value for Ventricular Arrhythmia Between Left Ventricular Mechanical Dispersion and Left Ventricular Ejection Fraction
or Global Longitudinal Strain
Log [Hazard Ratio]
SE
Haugaa et al13
–0.0101
0.2809
9.5%
0.99 [0.57, 1.72]
Haugaa et al20
–0.3527
0.3027
8.2%
0.70 [0.39, 1.27]
Leong et al22
–0.2143
0.1093
62.7%
0.81 [0.65, 1.00]
Mornos et al27
–0.2485
0.1951
19.7%
0.78 [0.53, 1.14]
100.0%
0.81 [0.68, 0.96]
Study
Weight
Hazard Ratio, 95% CI
Hazard Ratio, 95% CI
LVEF
Total (95% CI) LVEF vs LVMD (per 1-SD Increase)
Heterogeneity: Tau2 = 0.00; Chi2 = 0.77, df = 3 (P = 0.86); I2 = 0% Test for overall effect: Z = 2.46 (P = 0.01) Mechanical Dispersion Haugaa et al13
0.5185
0.1416
33.5%
1.68 [1.27, 2.22]
Haugaa et al20
0.9127
0.3057
11.1%
2.49 [1.37, 4.54]
Leong et al22
0.4136
0.1025
45.4%
1.51 [1.24, 1.85]
0
0.3263
10.0%
1.00 [0.53, 1.90]
100.0%
1.59 [1.28, 1.97]
Mornos et al27 Total (95% CI)
Heterogeneity: Tau2 = 0.02; Chi2 = 4.55, df = 3 (P = 0.21); I2 = 34% Test for overall effect: Z = 4.18 (P < 0.0001) Log [Hazard Ratio]
Study
SE
Weight
0.2
Hazard Ratio, 95% CI
0.5
1
2
5
Hazard Ratio, 95% CI
GLS Ersbøll et al19
0.7529
0.214
24.5%
2.12 [1.40, 3.23]
Haugaa et al13
–0.2083
0.2549
22.6%
0.81 [0.49, 1.34]
Haugaa et al18
1.1555
0.5142
12.6%
3.18 [1.16, 8.70]
Haugaa et al20
0
0.3442
18.6%
1.00 [0.51, 1.96]
Mornos et al27
0.0647
0.2737
21.7%
1.07 [0.62, 1.82]
100.0%
1.35 [0.84, 2.14]
Total (95% CI) GLS vs LVMD (per 1-SD Increase)
Heterogeneity: Tau2 = 0.18; Chi2 = 12.72, df = 4 (P = 0.01); I2 = 69% Test for overall effect: Z = 1.25 (P = 0.21) Mechanical Dispersion Ersbøll et al19
0.2096
0.0993
32.5%
Haugaa et al13
0.5185
0.1416
25.7%
1.68 [1.27, 2.22]
Haugaa et al18
0.3865
0.1652
22.4%
1.47 [1.06, 2.03]
Haugaa et al20
0.9127
0.3057
10.2%
2.49 [1.37, 4.54]
Mornos et al27
0
0.3263
9.3%
1.00 [0.53, 1.90]
100.0%
1.46 [1.18, 1.82]
Total (95% CI)
1.23 [1.02, 1.50]
Heterogeneity: Tau2 = 0.03; Chi2 = 7.99, df = 4 (P = 0.09); I2 = 50% Test for overall effect: Z = 3.41 (P = 0.0007)
0.2
0.5
1
2
5
The top forest plot shows the result of the comparison between left ventricular mechanical dispersion (LVMD) and left ventricular ejection fraction (LVEF), and the bottom forest plot shows the result of the comparison between LVMD and global longitudinal strain (GLS). Both forest plots display the summary hazard ratios per 1-SD increase and the 95% confidence intervals (CIs) in the multivariate model for increasing association of each parameter with ventricular arrhythmia.
efficacy in a large group, development of cutoffs,
This meta-analysis revealed that patients with VAs
and comparison with existing markers. It seems
had greater LVMD than those without (Figure 3).
likely that in addition to measuring contraction
However, setting a clear optimal cutoff of LVMD is
heterogeneity, LVMD is a marker of the forms
difficult because of differences among studies and
of
heterogeneity—including
differences in cardiac diseases. In patients with long-
nonuniform anisotropy, electric uncoupling, and
QT syndrome, LVMD is less pronounced compared
conduction
with those with structural disease. Furthermore,
electrophysiological block
contribute to VA.
and
slow
conduction—that
bundle branch block will increase LVMD with an
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019
Kawakami et al.
- 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
C ENTR AL I LL U STRA T I O N Predictive Value and Accuracy of Left Ventricular Mechanical Dispersion in Risk Stratification for Ventricular Arrhythmia
100 Specificity (%)
1 ROC AUC
0.8 0.6 0.4 0.2 0
60 40 20 0
0
20 40 60 80 100 Cut Off of Mechanical Dispersion (ms)
100
100
80
80
Accuracy (%)
Sensitivity (%)
80
60 40 20 0
0
20 40 60 80 100 Cut Off of Mechanical Dispersion (ms)
0
20 40 60 80 100 Cut Off of Mechanical Dispersion (ms)
60 40 20 0
0
20 40 60 80 100 Cut Off of Mechanical Dispersion (ms)
Publication
Optimal Cut Off (ms)
ROC AUC
Sensitivity (%)
Specificity (%)
Accuracy (%)
Haugaa et al13
70
0.84
65
92
78.5
Haugaa et
al18
72
0.8
67
89
78
Haugaa et
al20
47
0.75
80
62
71
50
0.81
91
55
73
61
0.84
85
73
79
101.2
0.685
38
92
65
63.5
0.71
70.6
63
66.8
Kosiuk et
al21
Nguyen et
al23
Matsuzoe et Candan et
al25
al28
Kawakami, H. et al. J Am Coll Cardiol Img. 2019;-(-):-–-. The sizes of the bubbles indicate the number of patients. AUC ¼ area under the curve; ROC ¼ receiver-operating characteristic.
unknown effect of cardiac risk. We therefore recom-
65% to 79%. In addition, Rodriguez-Zanella et al. (35)
mend using different cutoff values for patients with
recently calculated normal LVMD in 303 healthy sub-
long-QT syndrome and patients with cardiomyopa-
jects assessed using strain echocardiography and re-
thies and bundle branch blocks. Nonetheless, the
ported the normal LVMD to be 34 10 ms, with an
optimal cutoff value in 5 of 7 studies was >60 ms,
upper limit of normal of 56 ms. On the basis of these
which was also the difference in LVMD between pa-
findings, a LVMD of 60 ms seems reasonable as a cutoff
tients with and without VA events (62.7 16.6 ms)
value for risk stratification for VA in these patients.
(Figure 3A). When a cutoff value >60 ms was used for
Importantly, LVMD is unable to predict VA in pa-
predicting VA events in patients with ischemic or
tients receiving cardiac resynchronization therapy.
nonischemic cardiomyopathies, the range of area
Cardiac
under the curve was 0.69 to 0.84, with sensitivity of
changes regional timing to improve myocardial syn-
38% to 85%, specificity of 63% to 92%, and accuracy of
chrony and modulate arrhythmic risk (36).
resynchronization
therapy
profoundly
9
10
Kawakami et al.
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019 - 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
STUDY LIMITATIONS. First, as with any meta-analysis of
published after 2010 (37). Nonetheless, further vali-
observational studies, variations in the inclusion
dation studies are needed to clarify intervendor and
criteria and endpoints are all potential sources of het-
intersoftware variability.
erogeneity among studies. We could not access patientlevel data to allow adjustment for other covariates that might influence the incidence of VA, including medication, laboratory data, and other imaging parameters. Second, a group of researchers (including the authors), whose work was seminal in defining this test, were responsible for 4 of 12 studies in our systematic review (13,18,20,26), including the 2 largest studies (19,20), accounting for close to one-half of the patients. Although, reproducibility analyses have been provided, prospective multicenter studies are needed to confirm the relationship between LVMD and VAs and to integrate LVMD and decision making regarding ICD therapy. Third, an overestimation of the pooled effect sizes is possible, as the small number of studies limits the assessment of publication bias. Fourth, strain imaging is dependent on highquality echocardiographic imaging and appropriate
CONCLUSIONS LVMD assessed using speckle tracking provides important predictive value for VA in patients with a number of cardiac diseases. The predictive value of LVMD appears to be superior to that of LVEF or GLS. Hence, the clinical use of LVMD assessed by speckle tracking
should
be
considered
for
noninvasive
assessment of patients with cardiac disease. Prospective multicenter studies are warranted to confirm the external validity of these findings and translate LVMD into clinical decision making regarding ICD therapy. ADDRESS FOR CORRESPONDENCE: Dr. Thomas H.
Marwick, Baker Heart and Diabetes Institute, 75 Commercial Road, Melbourne 3004, Australia. E-mail: [email protected].
imaging settings (e.g. a frame rate of 50 to 70 frames/s), and the performance of LVMD in settings with less experienced users than the investigators of these studies in unclear. Nonetheless, the reproducibility of
PERSPECTIVES COMPETENCY IN MEDICAL KNOWLEDGE:
LVMD seemed to be acceptable, with interclass cor-
Prediction of VA remains a challenge. Recent studies
relation coefficients of LVMD within and between
have demonstrated that LVMD assessed using speckle
observers being similar to those for GLS (Table 2).
tracking might be a powerful marker in risk
Finally, lack of intervendor and intersoftware
stratification for VA. This meta-analysis confirmed the
standardization is often posed as an important limi-
association between LVMD and the incidence of VA in
tation of speckle tracking. However, 11 of 12 studies in
a number of cardiac diseases, and LVMD appears to
this meta-analysis solely used the same vendor (GE),
have superior predictive value over LVEF and GLS for
and the remaining study also used the same vendor in
risk stratification.
several patients (Table 1). In addition, 10 of 12 studies in this meta-analysis used EchoPAC to calculate LV strain. Although only 4 of 10 studies reported the version of the software, we believe that the effect of differences among EchoPAC versions was limited,
TRANSLATIONAL OUTLOOK: Further prospective studies are needed to integrate LVMD and decision making regarding ICD therapy in a variety of cardiovascular diseases.
because all studies in this meta-analysis were
REFERENCES 1. Bertini M, Schalij MJ, Bax JJ, Delgado V.
Arrhythmia Rate in High-risk MI patients (SEARCH-
of the Marburg Cardiomyopathy Study. Circulation
Emerging role of multimodality imaging to evaluate patients at risk for sudden cardiac death. Circ Cardiovasc Imaging 2012;5:525–35.
MI). Europace 2009;11:476–82.
2003;108:2883–91.
4. Moss AJ, Zareba W, Hall WJ, et al. Prophylactic
7. Nesbitt GC, Mankad S, Oh JK. Strain imaging in
implantation of a defibrillator in patients with myocardial infarction and reduced ejection fraction. N Engl J Med 2002;346:877–83.
echocardiography: methods and clinical applications. Int J Cardiovasc Imaging 2009;25:9–22.
2. Macatangay C, Viles-Gonzalez JF, Goldberger JJ. Role of cardiac imaging in evaluating risk for sudden cardiac death. Card Electrophysiol Clin 2017;9: 639–50. 3. Santini M, Russo M, Botto G, et al. Clinical and arrhythmic outcomes after implantation of a defibrillator for primary prevention of sudden death in patients with post-myocardial infarction cardiomyopathy: the Survey to Evaluate
5. Kadish A, Dyer A, Daubert JP, et al. Prophylactic defibrillator implantation in patients with nonischemic dilated cardiomyopathy. N Engl J Med 2004;350:2151–8.
8. Marwick TH. Methods used for the assessment of LV systolic function: common currency or tower of Babel? Heart 2013;99:1078–86.
6. Grimm W, Christ M, Bach J, Muller HH, Maisch B. Noninvasive arrhythmia risk stratifica-
9. Stecker EC, Vickers C, Waltz J, et al. Populationbased analysis of sudden cardiac death with and without left ventricular systolic dysfunction: twoyear findings from the Oregon Sudden Unexpected
tion in idiopathic dilated cardiomyopathy: results
Death Study. J Am Coll Cardiol 2006;47:1161–6.
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019
- 2019:-–-
10. Goldberger JJ, Subacius H, Patel T, Cunnane R, Kadish AH. Sudden cardiac death risk stratification in patients with nonischemic dilated cardiomyopathy. J Am Coll Cardiol 2014;63:1879–89. 11. Helle-Valle T, Crosby J, Edvardsen T, et al. New noninvasive method for assessment of left ventricular rotation: speckle tracking echocardiography. Circulation 2005;112:3149–56.
ventricular arrhythmias after myocardial infarction. J Am Coll Cardiol Img 2013;6:841–50.
report from the American Heart Association. Circulation 2015;131:e29–322.
21. Kosiuk J, Dinov B, Bollmann A, et al. Association between ventricular arrhythmias and myocardial mechanical dispersion assessed by strain analysis in patients with nonischemic cardiomyopathy. Clin Res Cardiol 2015;104:1072–7.
30. Stevenson WG, Brugada P, Waldecker B, Zehender M, Wellens HJ. Clinical, angiographic, and electrophysiologic findings in patients with aborted sudden death as compared with patients
12. Haugaa KH, Amlie JP, Berge KE, Leren TP, Smiseth OA, Edvardsen T. Transmural differences in myocardial contraction in long-QT syndrome: mechanical consequences of ion channel dysfunction. Circulation 2010;122:1355–63.
22. Leong DP, Hoogslag GE, Piers SR, et al. The relationship between time from myocardial infarction, left ventricular dyssynchrony, and the risk for ventricular arrhythmia: speckle-tracking echocardiographic analysis. J Am Soc Echocardiogr 2015;28:470–7.
13. Haugaa KH, Smedsrud MK, Steen T, et al. Mechanical dispersion assessed by myocardial strain in patients after myocardial infarction for risk prediction of ventricular arrhythmia. J Am Coll
23. Nguyen BL, Capotosto L, Persi A, et al. Global and regional left ventricular strain indices in postmyocardial infarction patients with ventricular arrhythmias and moderately abnormal ejection
Cardiol Img 2010;3:247–56.
fraction. Ultrasound Med Biol 2015;41:407–17.
14. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate
24. Negishi K, Negishi T, Zardkoohi O, et al. Left atrial booster pump function is an independent predictor of subsequent life-threatening ventricular arrhythmias in non-ischaemic cardiomyopa-
healthcare interventions: explanation and elaboration. BMJ 2009;339:b2700. 15. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-an-
thy. Eur Heart J Cardiovasc Imaging 2016;17: 1153–60.
alyses. BMJ 2003;327:557–60.
25. Matsuzoe H, Tanaka H, Matsumoto K, et al. Left ventricular dyssynergy and dispersion as
16. Downs SH, Black N. The feasibility of creating a checklist for the assessment of the methodological quality both of randomised and nonrandomised studies of health care interventions.
determinant factors of fatal ventricular arrhythmias in patients with mildly reduced ejection fraction. Eur Heart J Cardiovasc Imaging 2016;17: 334–42.
J Epidemiol Commun Health 1998;52:377–84.
26. Hasselberg NE, Haugaa KH, Bernard A, et al. Left ventricular markers of mortality and ventricular arrhythmias in heart failure patients with cardiac resynchronization therapy. Eur Heart J Cardiovasc Imaging 2016;17:343–50.
17. Levy PT, Sanchez Mejia AA, Machefsky A, Fowler S, Holland MR, Singh GK. Normal ranges of right ventricular systolic and diastolic strain measures in children: a systematic review and metaanalysis. J Am Soc Echocardiogr 2014;27:549–60. e3. 18. Haugaa KH, Goebel B, Dahlslett T, et al. Risk assessment of ventricular arrhythmias in patients with nonischemic dilated cardiomyopathy by strain echocardiography. J Am Soc Echocardiogr 2012;25:667–73. 19. Ersboll M, Valeur N, Andersen MJ, et al. Early echocardiographic deformation analysis for the prediction of sudden cardiac death and lifethreatening arrhythmias after myocardial infarction. J Am Coll Cardiol Img 2013;6:851–60. 20. Haugaa KH, Grenne BL, Eek CH, et al. Strain echocardiography improves risk prediction of
Kawakami et al. Mechanical Dispersion and Ventricular Arrhythmias
27. Mornos C, Muntean D, Mornos A, et al. Risk stratification in patients with heart failure: the value of considering both global longitudinal left ventricular strain and mechanical dispersion. Can J Physiol Pharmacol 2017;95:1360–8. 28. Candan O, Gecmen C, Bayam E, Guner A, Celik M, Dogan C. Mechanical dispersion and global longitudinal strain by speckle tracking echocardiography: predictors of appropriate implantable cardioverter defibrillator therapy in hypertrophic cardiomyopathy. Echocardiography 2017;34:835–42. 29. Mozaffarian D, Benjamin EJ, Go AS, et al. Heart disease and stroke statistics—2015 update: a
with sustained ventricular tachycardia after myocardial infarction. Circulation 1985;71: 1146–52. 31. Hsia HH, Marchlinski FE. Characterization of the electroanatomic substrate for monomorphic ventricular tachycardia in patients with nonischemic cardiomyopathy. Pacing Clin Electrophysiol 2002;25:1114–27. 32. Vaduganathan M, Patel RB, Michel A, et al. Mode of death in heart failure with preserved ejection fraction. J Am Coll Cardiol 2017;69: 556–69. 33. Kutyifa V, Pouleur AC, Knappe D, et al. Dyssynchrony and the risk of ventricular arrhythmias. J Am Coll Cardiol Img 2013;6:432–44. 34. Leren IS, Hasselberg NE, Saberniak J, et al. Cardiac mechanical alterations and genotype specific differences in subjects with long QT syndrome. J Am Coll Cardiol Img 2015;8:501–10. 35. Rodriguez-Zanella H, Haugaa K, Boccalini F, et al. Physiological determinants of left ventricular mechanical dispersion: a 2-dimensional speckle tracking echocardiographic study in healthy volunteers. J Am Coll Cardiol Img 2018;11:650–1. 36. Gorcsan J, Haugaa Kristina H. Ventricular arrhythmias and reduced echocardiographic inferior wall strain. Circ Cardiovasc Imaging 2017;10: e005900. 37. Lang RM, Badano LP, Mor-Avi V, et al. Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging. J Am Soc Echocardiogr 2015;28:1–39. e14.
KEY WORDS mechanical dispersion, meta-analysis, strain, ventricular arrhythmia
A PP END IX For supplemental tables and figures, please see the online version of this paper.
11
VOL.
-, NO. -, 2019
ª 2019 BY THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION PUBLISHED BY ELSEVIER
ORIGINAL RESEARCH
Prediction of Ventricular Arrhythmias With Left Ventricular Mechanical Dispersion A Systematic Review and Meta-Analysis Hiroshi Kawakami, MD, PHD,a Nitesh Nerlekar, MBBS, MPH,a Kristina H. Haugaa, MD, PHD,b Thor Edvardsen, MD, PHD,b Thomas H. Marwick, MBBS, PHD, MPHa
ABSTRACT OBJECTIVES The aim of this study was to assess the association between left ventricular mechanical dispersion (LVMD) and the incidence of ventricular arrhythmias (VAs). BACKGROUND Recent, mainly single-center, studies have demonstrated that LVMD assessed using speckle tracking might be a powerful marker in risk stratification for VA. A systematic review and meta-analysis provides a means of understanding the prognostic value of this parameter, relative to other parameters, the most appropriate cutoff for designating risk. METHODS A systemic review of studies reporting the predictive value of LVMD for VA was undertaken from a search of MEDLINE and Embase. VA events were defined as sudden cardiac death, cardiac arrest, documented ventricular tachyarrhythmia, and appropriate implantable cardioverter-defibrillator (ICD) therapy. Hazard ratios were extracted from univariate and multivariate models reporting on the association of LVMD and VA and described as pooled estimates with 95% confidence intervals. In a meta-analysis, the predictive value of LVMD was compared with that of left ventricular ejection fraction and global longitudinal strain. RESULTS Among 3,198 patients in 12 published studies, 387 (12%) had VA events over follow-up ranging from 17 to 70 months. Patients with VAs had greater LVMD than those without (weighted mean difference 20.3 ms; 95% confidence interval: 27.3 to 13.2; p < 0.01). Each 10 ms increment of LVMD was significantly and independently associated with VA events (hazard ratio: 1.19; 95% confidence interval: 1.09 to 1.29; p < 0.01). The predictive value of LVMD was superior to that of left ventricular ejection fraction or global longitudinal strain. CONCLUSIONS LVMD assessed using speckle tracking provides important predictive value for VA in patients with a number of cardiac diseases and appears to have superior predictive value over left ventricular ejection fraction and global longitudinal strain for risk stratification. (J Am Coll Cardiol Img 2019;-:-–-) © 2019 by the American College of Cardiology Foundation.
From the aBaker Heart and Diabetes Institute, Melbourne, Australia; and the bCentre of Cardiological Innovation, Department of Cardiology, Oslo University Hospital, Oslo, Norway. Dr. Marwick has received research support from the National Health and Medical Research Council (grants 1119955, 1080582, 1059738, and 1149692) and GE Medical Systems for an ongoing research study on the use of strain for the assessment of cardiotoxicity. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Allan L. Klein, MD, served as the Guest Editor for this paper. Manuscript received December 10, 2018; revised manuscript received February 12, 2019, accepted March 13, 2019.
ISSN 1936-878X/$36.00
https://doi.org/10.1016/j.jcmg.2019.03.025
2
Kawakami et al.
JACC: CARDIOVASCULAR IMAGING, VOL.
C
ABBREVIATIONS AND ACRONYMS CI = confidence interval GLS = global longitudinal strain
ardiac imaging is widely used for
English, published up to July 17, 2018. The study was
assessing
prospectively registered with the Prospero database
sudden
cardiac
death
(SCD) risk (1,2), but the prediction
ICD = implantable cardioverter-defibrillator
STUDY SELECTION. Studies involving prediction of
nonsustained
VA events were included in this analysis on the basis
and
sustained
ventricular
of
the
availability
of
LVMD
assessed
using
SCD remains a challenge. Severe left ventric-
2-dimensional speckle-tracking transthoracic echo-
ular (LV) dysfunction (LV ejection fraction
cardiography. LVMD was defined as the SD of time
[LVEF] <35%) is a useful predictor of VA in
LV = left ventricular
(CRD42018104240).
of ventricular arrhythmias (VAs) including tachycardia and ventricular fibrillation and
HR = hazard ratio
-, NO. -, 2019 - 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
from Q/R on electrocardiography to peak negative
patients with structural heart disease (3–6)
strain from each LV segment (Figure 1). LVMD was
and is the most widely used marker for pre-
evaluated as a predictor of VA events during follow-
dicting VA events. However, LVEF has
up. Studies in which mechanical dispersion was
mechanical dispersion
several limitations, including the influence
assessed using tissue Doppler imaging, 3-dimensional
SCD = sudden cardiac death
of heart rate and translational motion, geo-
strain imaging, or nonechocardiographic imaging,
VA = ventricular arrhythmia
metric
reproducibility
including cardiac magnetic resonance imaging, were
(7,8). In addition, most cases of SCD occur
excluded because of technical differences. We also
LVEF = left ventricular ejection fraction
LVMD = left ventricular
assumptions,
and
in patients with LVEF better than 35% (9,10).
excluded studies in which strain analysis was
Two-dimensional speckle tracking uses myocardial
restricted to the right ventricle. Abstracts without
acoustic reflections and interference patterns to
complete published papers, case reports, review pa-
measure myocardial deformation and has emerged as
pers, editorials, and letters were also excluded.
a powerful tool for detecting LV dysfunction (7,8,11). In 2010, Haugaa et al. (12,13) reported that LV mechanical dispersion (LVMD) assessed using speckle tracking strain was a useful marker for predicting VA in a variety of cardiovascular diseases (12,13). Subsequently, several studies have assessed the association
ENDPOINTS. The primary endpoint was VA events,
defined as SCD, cardiac arrest, documented ventricular fibrillation, documented ventricular tachycardia (sustained and/or nonsustained), and appropriate ICD therapy such as antitachycardia pacing and/or shock.
between LVMD and VA, but these studies were mostly
DATA
single-center analyses with relatively small sample
extracted and reviewed by 2 investigators. Discrep-
EXTRACTION. Data
sizes. The aim of this study was to clarify the associ-
ancies were manually reviewed and resolved by
ation between LVMD and the incidence of VA. We
consensus. Among the included publications, we
performed a systematic review and combined the
extracted the following data for systemic review:
data using meta-analytic techniques to strengthen
study design, study population, demographic char-
the level of evidence and provide deeper insight into
acteristics, follow-up period, outcomes, and echo-
the issue of LVMD in the incidence of VA in the pa-
cardiographic parameters including LVEF, GLS, and
tients with cardiac disease. We also assessed whether
LVMD. We contacted the investigators if study data
LVMD provided superior predictive information on
were incomplete.
were
independently
the incidence of VA compared with LVEF and global
STATISTICAL ANALYSIS. Data for continuous vari-
longitudinal strain (GLS).
ables were extracted as a weighted mean of the studies reporting continuous variable data. The
METHODS
weighted mean difference for LVMD between patients with and without VA events was calculated.
SEARCH STRATEGY. A systematic electronic search
Pooled hazard ratios (HRs) and 95% confidence in-
of published research was conducted using the
tervals (CIs) were computed using random-effect
MEDLINE and Embase databases in adherence with
models of the association of mechanical dispersion
the Preferred Reporting Items for Systematic Reviews
with VA events, adjusted for clinical differences be-
and Meta-Analyses guidelines (14). The search terms
tween the populations. Some studies reported HRs in
used the Medical Subject Headings and key words
LVMD relative to a 1-ms increase in LVMD; hence, we
“strain,”
“dyssynchrony,”
rescaled the HRs to a 10-ms increase in all studies.
“dispersion,” “ventricular arrhythmia,” “ventricular
Furthermore, we compared the pooled HRs for LVMD,
tachycardia,” “ventricular fibrillation,” “sudden car-
LVEF, and GLS to compare their predictive value of
diac death,” and “implantable cardioverter defibril-
VA events. To allow direct comparison of their
lation.” Additional strategies involved reference
impact, all HRs were rescaled to a 1-SD increase in
searches to identify further relevant studies. We
LVMD, LVEF, and GLS. Forest plots were constructed
limited the search to studies of adult humans, in
to show the overall effect of each parameter.
“speckle
tracking,”
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019
Kawakami et al.
- 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
F I G U R E 1 Calculation of Left Ventricular Mechanical Dispersion Using Speckle-Tracking Technique
Standard Deviation of Time to Peak Negative Strain
GS=-18.5%
Time to Peak Negative Strain
White arrows indicate peak negative strain. Left ventricular mechanical dispersion (LVMD) was defined as the SD of time from Q/R on electrocardiography to peak negative strain from each left ventricular segment. AVC ¼ aortic valve closure; GS ¼ global strain.
Heterogeneity was described using the I2 statistic,
Figure 2. From the initial 244 papers identified from
which was quantified as low (<25%), moderate
the search strategy and additional papers, 38 studies
(25% to 75%), or high (>75%) (15). Publication bias was
were considered to be potentially eligible, following
assessed using funnel plots and the Egger and Begg
the exclusion of duplicates and screening by title and
test. Study quality was assessed using the Newcastle-
abstract. After full review, 12 studies were included in
Ottawa scale (0 to 9 points) using the methodology
the final analysis (13,18–28) (Table 1). The reasons for
described by Downs and Black (16). Furthermore, we
exclusion are provided in Supplemental Table 1.
also identified the documentation of blinded perfor-
The characteristics of the 12 studies included in the
mance of data collection and analysis, description of
systemic review and meta-analysis are presented in
strain calculation technique, and interobserver and
Table 1. All studies were observational (6 prospective
intraobserver variability as important quality metrics
and 6 retrospective), and half were multicenter.
(17). Statistical analysis was performed using RevMan
Among 3,198 patients (weighted mean age 63 years;
5.3 (Cochrane Information Management System,
weighted mean 30% women), the most prevalent
Oxford, United Kingdom) and Stata (StataCorp, Col-
cardiac
lege Station, Texas), with 2-tailed p values <0.05
(n ¼ 2,634).
considered to indicate statistical significance.
condition
was
ischemic
heart
disease
Echocardiographic data at baseline are presented in Table 1. Most studies solely used GE echocardio-
RESULTS
graphic platforms (n ¼ 11) and proprietary software (n ¼ 10). Among the 9 papers reporting frame rates for
STUDY SELECTION AND PATIENT CHARACTERISTICS.
speckle-tracking imaging, most were >50 frames/s.
The process of study selection is presented in
Eleven studies presented data on LVEF (weighted
3
4
Kawakami et al.
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019 - 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
F I G U R E 2 Flowchart of the Study Selection Process
Records identified through database search: Medline (56), EMBASE (167) (n = 223)
Additional records identified through other sources (n = 21)
Records after duplicates removed (n = 172)
Records screened (n = 172)
Excluded by title and abstract (n = 134)
Full-text studies assessed for eligibility (n = 38)
Full-text articles excluded (please see supplemental table 1) (n = 26)
Studies included in quantitative synthesis (Meta-analysis) (n = 12)
The reasons for exclusion are provided in Supplemental Table 1.
mean
(weighted
continuous variable (per 10-ms increase) for VA
mean 14.1%). The reported means of LVMD ranged
46%),
and
11
reported
GLS
events was available from 9 studies and was 1.26
from 43 to 112 ms (weighted mean 60 ms). Repro-
(95% CI: 1.14 to 1.39; p < 0.01; I 2 ¼ 84%). In the
ducibility of strain imaging was demonstrated in
multivariate analysis, the pooled adjusted HR of
11 studies, in 9 of which expressed as an intraclass
LVMD as a continuous variable (per 10-ms increase)
correlation coefficient (Table 2). The range of intra-
for VA events was available from 6 studies and was
observer and interobserver variability was from
1.19 (95% CI: 1.09 to 1.29; p < 0.01; I 2 ¼ 48%). LVMD
0.92 to 0.99 and from 0.90 to 0.99, respectively, for
was significantly and independently associated with
GLS and from 0.86 to 0.99 and from 0.78 to 0.96,
VA events. The association of LVEF and GLS with VA
respectively, for LVMD.
events in univariable and multivariable models is
RELATIONSHIP BETWEEN LVMD AND VA. The defi-
nitions of VA events are provided in Table 1. Over 17 to 70 months of follow-up, 387 patients (12%) had VA events. In patients with VA events, LVMD was greater than in those without VA events (Figure 3A). LVMD in patients with VAs was 20.3 ms shorter than in those without (95% CI: 27.3 to 13.2 ms; p < 0.01; I 2 ¼ 76%) (Figure 3B).
shown in Supplemental Figures 1 and 2. Figure 5 shows the results of comparison of the predictive value for VA events between LVMD and LVEF or GLS. In these analyses, we selected studies that reported multivariate HRs for both LVMD and LVEF or GLS. In addition, to facilitate comparison, we used rescaled HRs per 1-SD increase. On the basis of 4 studies, LVMD (HR: 1.59; 95% CI: 1.28 to 1.97; p < 0.01; I 2 ¼ 34%) and LVEF (HR: 0.81; 95% CI: 0.68 to 0.96;
PREDICTIVE VALUE OF LVMD FOR VA EVENTS.
p ¼ 0.01; I2 ¼ 0%) were independently associated with
All studies provided risk variables of LVMD for VA
VA events, and the predictive value of LVMD was
events; 9 studies provided HRs and 3 studies pro-
greater than that of LVEF (Figure 5). Likewise, the
vided odds ratios. Figure 4 shows the association of
comparison between LVMD and GLS on the basis of
LVMD with VA events in univariate and multivariate
5 studies revealed that LVMD was independently
models. The pooled unadjusted HR of LVMD as a
associated with VA events (HR: 1.46; 95% CI: 1.18 to
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019
Kawakami et al.
- 2019:-–-
5
Mechanical Dispersion and Ventricular Arrhythmias
T A B L E 1 Study design, patient characteristics, details of ventricular arrhythmic events, and echocardiographic parameters
Year
n
Design
Age, years
Female, %
Population
IHD, %
Definition of VA events
Follow-up period, months
Haugaa et al. (13)
Publication
2010
85
Prosp
63 10
15
Post-MI
100
Appropriate ICD therapy
28 (range; 07-66)
Haugaa et al. (18)
2012
94
Prosp
49 12
19
NICM
0
Appropriate ICD therapy/VT/VF/ SCD
22 (range; 1-46)
Ersbøll et al. (19)
2013
988
Prosp
63 12
28
Post-MI
100
Appropriate ICD therapy/VT/VF/ SCD
30 (23.5-32.7)
Haugaa et al. (20)
2013
569
Prosp
61 11
34
Post-MI
100
VT/VF/SCD
30 (18)
Kosiuk et al. (21)
2015
20
Retro
62 11
25
NICM
0
Appropriate ICD therapy /VT/VF
70 40
Leong et al. (22)
2015
206
Retro
67 (57-73)
13
Post-MI
100
Appropriate ICD therapy
24 (7.8-47)
Nguyen et al. (23)
2015
467
Retro
68 14
39
Post-MI
100
VT
25 (range; 6-43)
Negishi et al. (24)
2016
124
Retro
56 13
46
NICM
0
Appropriate ICD therapy
46 (26-72)
Matsuzoe et al. (25)
2016
72
Retro
58 15
18
Unselected
32
Appropriate ICD therapy
17 (range; 0.2–72.5)
Hasselberg et al. (26)
2016
170
Prosp
66 10
24
HF with CRT
48
Appropriate ICD therapy/VT/VF/ Cardiac arrest
23 4
Mornos et al. (27)
2017
340
Prosp
63 12
33
Unselected
63
VT/VF/SCD
36 9
Candan et al. (28)
2017
63
Retro
49 14
22
HCM
0
Appropriate ICD therapy
22 7
TABLE 1 Continued VA events
QRS duration, ms
LVEF, %
GLS, %
LVMD, ms
Vendor
Software (Version)
FR, frames/s
Haugaa et al. (13)
Publication
38 (45%)
100 19
34.4 10.1
-10.7 3.9
69.0 20.2
GE
EchoPAC (NR)
63 23
Haugaa et al. (18)
12 (13%)
118 34
36.7 12.6
-11.5 5.0
61.4 21.2
GE
EchoPAC (NR)
>70
Ersbøll et al. (19)
34 (3%)
98 19
50.7 10.4
-13.6 3.5
56.8 15.0
GE
EchoPAC (BT11.1.0)
>60
Haugaa et al. (20)
15 (3%)
95 16
54.8 11.2
-18.1 3.7
42.6 17.2
GE
EchoPAC (NR)
NR
Kosiuk et al. (21)
9 (45%)
102 14
32 6
NR
70 29
GE
MatLab
>60
Leong et al. (22)
75 (36%)
121 27
38.8 10.5
-11.4*
87.7 36.5
GE
EchoPAC (11.0.0)
NR
Nguyen et al. (23)
51 (11%)
133 35
43.1 5.8
-14.4 3.7
47.2 13.5
GE
EchoPAC (BT11)
60-100
Negishi et al. (24)
36 (29%)
NR
31.4 9.9
-9.1 3.5
103 43
GE
EchoPAC (11.0.0)
50 20
Matsuzoe et al. (25)
34 (47%)
113 27
52.2 12.0
-11.2 3.4
83.1 28.6
GE/Toshiba
Ultra Extend
45
Hasselberg et al. (26)
18 (11%)
165 22
26 9
-8.2 3.9
112*
GE
EchoPAC (NR)
>50
Mornos et al. (27)
48 (14%)
NR
41.7 12.3
-17.1 6.5
44.3 32.3
GE
EchoPAC (NR)
NR
Candan et al. (28)
17 (27%)
NR
NR
-12.1 3.4
66.4 19.4
GE
EchoPAC (NR)
50-70
Data are expressed as mean SD, median (interquartile range), median (range), or number (%). *Mean calculated from their original figure or table. CRT ¼ cardiac resynchronization therapy; FR ¼ frame rate; GE ¼ general electric; GLS ¼ global longitudinal strain; HF ¼ heart failure; HCM ¼ hypertrophic cardiomyopathy; ICD ¼ implantable cardioverter defibrillator; IHD ¼ ischemic heart disease; LVEF ¼ left ventricular ejection fraction; LVMD ¼ left ventricular mechanical dispersion; MI ¼ myocardial Infarction; NICM ¼ non-ischemic cardiomyopathy; NR ¼ not reported; Prosp ¼ prospective; Retro ¼ retrospective; SCD ¼ sudden cardiac death; VA ¼ ventricular arrhythmia; VF ¼ ventricular fibrillation; VT ¼ ventricular tachycardia.
1.82; p < 0.01; I 2 ¼ 50%) but not GLS (HR: 1.35; 95% CI:
T A B L E 2 Reproducibility of GLS and Mechanical Dispersion
0.84 to 2.14; p ¼ 0.21; I 2 ¼ 69%) (Figure 5).
GLS
Supplemental Figures 1 and 2 repeat these comparisons against each 1% change of EF and GLS.
First Author (Ref. #)
Method
Mechanical Dispersion
Intraobserver Interobserver Intraobserver Interobserver Variability Variability Variability Variability
The Central Illustration shows the ability of LVMD
Haugaa et al. (13)
ICC
0.98
0.98
0.86
0.81
to identify VA events. Seven studies described the
Haugaa et al. (18)
ICC
0.98
0.95
0.86
0.78
results of receiver-operating characteristic curve an-
Ersbøll et al. (19)
Bland-Altman analysis (ms)
-0.7 2.5*
-0.05 1.3*
1.4 7.5*
1.3 1.08*
Haugaa et al. (20)
ICC
0.92
0.90
0.89
0.82
Kosiuk et al. (21)
NR
NR
NR
NR
NR
Leong et al. (22)
ICC
NR
0.94
NR
0.96 6.1 3.5†
alyses and the accuracy of mechanical dispersion. The range of the optimal cutoff value, area under the curve, sensitivity, specificity, and accuracy were 47 to 101 ms, 0.69 to 0.84, 38% to 91%, 55% to 92%%, and
Nguyen et al. (23) Mean difference
3.5 2.1†
5.5 3.4†
4.2 2.3†
65% to 79%, respectively (Central Illustration).
Negishi et al. (24)
ICC
0.99
0.99
0.99
0.93
ICC
NR
NR
0.940
0.917
QUALITY ASSESSMENT AND PUBLICATION BIAS
Matsuzoe et al. (25) Hasselberg et al. (26)
ICC
0.94
0.92
NR
NR
ANALYSIS. Study quality was assessed using the
Newcastle-Ottawa scale (Supplemental Table 2). All included studies had high scores on the Newcastle-
Mornos et al. (27)
ICC
0.93
0.92
0.91
0.90
Candan et al. (28)
ICC
0.92
0.94
0.96
0.93
Ottawa scale (all studies scored 8 points or more). In addition, all studies defined the study objective, described the outcomes and confounders, and outlined the main findings (Supplemental Table 3). As for
Values are % or mean SD. *Mean differences in agreement. †Determined as the difference between the 2 sets of observations divided by the mean of the observations and expressed as a percentage. ICC ¼ intraclass correlation coefficient; NR ¼ not reported.
Kawakami et al.
6
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019 - 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
F I G U R E 3 Difference in Left Ventricular Mechanical Dispersion Between Patients With and Those Without Ventricular Arrhythmias
A (ms) 140
51.5 ± 18.7
78.6 ± 27.5
VA (–) N = 2526
VA (+) N = 296
120 100 80 60 40 20 0
B
VA (+)
VA (–) Study
Mean
SD
N
Mean
SD
N
Candan et al28 Ersbøll et al19 Haugaa et al13 Haugaa et al18 Haugaa et al20 Kosiuk et al21 Matsuzoe et al25 Mornos et al27 Negishi et al24 Nguyen et al23
62.4 56.3 56 56 42 50 73.9 39.7 106 45.3
17.1 15.5 13 18 17 16 22.6 33.1 46 13.1
46 954 47 82 554 9 38 292 88 416
77.1 70.7 85 98 63 84 93.3 72.3 95 62.7
21.8 29.7 29 43 25 31 31.3 27.6 33 16.6
2526
Total (95% CI)
Weight
Mean Difference, 95% CI
17 34 38 12 15 11 34 48 36 51
10.6% 11.4% 11.4% 5.3% 10.0% 6.4% 10.0% 12.1% 9.1% 13.9%
–14.70 [–26.18, –3.22] –14.40 [–24.43, –4.37] –29.00 [–38.94, –19.06] –42.00 [–66.64, –17.36] –21.00 [–33.73, –8.27] –34.00 [–55.09, –12.91] –19.40 [–32.14, –6.66] –32.60 [–41.28, –23.92] 11.00 [–3.44, 25.44] –17.40 [–22.13, –12.67]
296
100.0%
–20.25 [–27.30, –13.20]
Heterogeneity: Tau2 = 87.57; Chi2 = 36.87, df = 9 (P < 0.0001); I2 = 76% Test for overall effect: Z = 5.63 (P < 0.00001)
Mean Difference, 95% CI
–50
–25 VA (–)
0
25
50
VA (+)
(A) Mean left ventricular mechanical dispersion (LVMD) in patients with and those without ventricular arrhythmias (VAs). The colored bars show the weighted mean LVMD, and gray dots show the mean LVMD in the original studies in the 2 groups. (B) The forest plot displays the weighted mean differences and 95% confidence intervals (CIs) for difference between patients with and without VAs.
echocardiographic quality, all studies described the
(Supplemental Figure 3). This might be because
strain imaging protocol. In 10 studies, echocardio-
baseline LVMD varies depending on patients’ cardiac
graphic parameters were calculated by blinded re-
conditions.
searchers.
All
but
1
study
evaluated
the
reproducibility of GLS and/or mechanical dispersion
DISCUSSION
(Table 2). Publication bias was explored using a funnel plot
There are 3 main findings of the present study. First,
and the Egger test for all studies (Supplemental
LVMD was significantly and independently associated
Figures 3 to 7). No evidence of significant publication
with VA events. Second, a 1-SD change in LVMD was a
bias was identified in the forest plots displaying the
stronger predictor of VA events compared with the
summary HRs for increasing association of each
same changes in LVEF and GLS. Third, patients with
parameter with VA (Supplemental Figures 4 to 7).
VA events had significantly greater mechanical
However, there was evidence of publication bias in
dispersion compared with those without VA events,
the comparison of weighted mean difference of LVMD
with 60 ms being an acceptable cutoff LVMD value for
between the patients with and without VA events
predicting VA events.
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019
Kawakami et al.
- 2019:-–-
7
Mechanical Dispersion and Ventricular Arrhythmias
F I G U R E 4 Left Ventricular Mechanical Dispersion as a Predictor of Ventricular Arrhythmia
A Study Ersbøll et al19 Hasselberg et al26 Haugaa et al13 Haugaa et al18 Haugaa et al20 Leong et al22 Matsuzoe et al25 Mornos et al27 Negishi et al24
Log [Hazard Ratio]
SE
0.3221 0.006 0.2624 0.3293 0.5878 0.131 0.0953 0.1989 –5.2983
0.0546 0.0456 0.059 0.0708 0.1024 0.0229 0.1001 0.0528 4.457
Total (95% CI)
Hazard Ratio, 95% CI
Weight 13.3% 13.9% 12.9% 12.0% 9.5% 15.2% 9.7% 13.4% 0.0%
1.38 [1.24, 1.54] 1.01 [0.92, 1.10] 1.30 [1.16, 1.46] 1.39 [1.21, 1.60] 1.80 [1.47, 2.20] 1.14 [1.09, 1.19] 1.10 [0.90, 1.34] 1.22 [1.10, 1.35] 0.01 [0.00, 31.10]
100.0%
1.26 [1.14, 1.39]
Heterogeneity: Tau2 = 0.02; Chi2 = 50.12, df = 8 (P < 0.00001); I2 = 84% Test for overall effect: Z = 4.55 (P < 0.00001)
B
Log [Hazard Ratio]
SE
Ersbøll et al19 Haugaa et al13 Haugaa et al18 Haugaa et al20 Leong et al22
0.1398 0.2546 0.1823 0.5306 0.1133
0.0662 0.0691 0.0779 0.1777 0.0281
19.9% 19.0% 16.7% 4.8% 33.9%
1.15 [1.01, 1.31] 1.29 [1.13, 1.48] 1.20 [1.03, 1.40] 1.70 [1.20, 2.41] 1.12 [1.06, 1.18]
Mornos et al27
0
0.1606
5.7%
1.00 [0.73, 1.37]
100.0%
1.19 [1.09, 1.29]
Study
Total (95% CI)
Hazard Ratio, 95% CI
Hazard Ratio, 95% CI
Weight
0.5
0.7
1
Covariates in Multivariable Model
1.5
2
Hazard Ratio, 95% CI
Age, QRS, LVEDV, GLS Age, Sex, LVEF, GLS Age, QRS, LVEF, GLS Age, LVEF, GLS Age, QRS, Cr, Revascurized infarct-related artery Time from MI, LVESVI, LV scare, VT inducible, LVEF, GLS Age, LBBB, E/e’, LVEF, GLS
Heterogeneity: Tau2 = 0.00; Chi2 = 9.53, df = 5 (P = 0.09); I2 = 48% Test for overall effect: Z = 4.10 (P < 0.0001)
0.5
0.7
1
1.5
2
(A) Univariate analysis for prediction of ventricular arrhythmia (VA). (B) Multivariate analysis for prediction of VA. The forest plots display the summary hazard ratios per 10-ms increase and 95% confidence intervals (CIs) for increasing association of left ventricular (LV) mechanical dispersion with VA. Cr ¼ creatinine; GLS ¼ global longitudinal strain; LBBB ¼ left bundle branch block; LVEDV ¼ left ventricular end-diastolic volume; LVEF ¼ left ventricular ejection fraction; LVESVI ¼ left ventricular end-systolic volume index; MI ¼ myocardial infarction; VT ¼ ventricular tachycardia.
ROLE OF ECHOCARDIOGRAPHY IN RISK STRATIFICATION
deaths in patients with heart failure with preserved
FOR VA EVENTS. About one-quarter of patients among
ejection fraction (32). These observations suggest that
emergency medical services–treated out-of-hospital
better parameters than LVEF are necessary for risk
cardiac arrests have an initial rhythm of ventricular
stratification for VA.
fibrillation or ventricular tachycardia (29). Although the genesis of these arrhythmias is multifactorial,
LVMD AS A RISK MARKER OF VA. Over the past 15
heterogeneous
by
years, GLS has emerged as a robust tool for the
myocardial fibrosis has been thought to be the main
assessment of global LV function and marker of
substrate for VA in patients with structural heart
subclinical LV dysfunction (7,8). The time course of
disease (30,31). Cardiac imaging has played an
segmental strain has also permitted the assessment
important role in assessing the structural substrate
of LVMD, and several studies have demonstrated
for arrhythmogenesis (1,2), especially echocardiogra-
the value of LVMD for the prediction of VA in a wide
phy (because it is noninvasive, highly accessible, and
range of conditions, from long-QT syndrome to
portable), and usually through measurement of LVEF
cardiac resynchronization therapy (12,33,34). The
(3–6). However, LVEF is not a reliable predictor, as
performance of meta-analysis is useful when the
many patients with VA have preserved LVEFs (9,10).
evidence is still relatively underdeveloped, as in
Indeed, SCD accounts for 30% to 40% of cardiac
the case of LVMD, and permits confirmation of
ventricular
activation
caused
8
Kawakami et al.
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019 - 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
F I G U R E 5 Comparison of Predictive Value for Ventricular Arrhythmia Between Left Ventricular Mechanical Dispersion and Left Ventricular Ejection Fraction
or Global Longitudinal Strain
Log [Hazard Ratio]
SE
Haugaa et al13
–0.0101
0.2809
9.5%
0.99 [0.57, 1.72]
Haugaa et al20
–0.3527
0.3027
8.2%
0.70 [0.39, 1.27]
Leong et al22
–0.2143
0.1093
62.7%
0.81 [0.65, 1.00]
Mornos et al27
–0.2485
0.1951
19.7%
0.78 [0.53, 1.14]
100.0%
0.81 [0.68, 0.96]
Study
Weight
Hazard Ratio, 95% CI
Hazard Ratio, 95% CI
LVEF
Total (95% CI) LVEF vs LVMD (per 1-SD Increase)
Heterogeneity: Tau2 = 0.00; Chi2 = 0.77, df = 3 (P = 0.86); I2 = 0% Test for overall effect: Z = 2.46 (P = 0.01) Mechanical Dispersion Haugaa et al13
0.5185
0.1416
33.5%
1.68 [1.27, 2.22]
Haugaa et al20
0.9127
0.3057
11.1%
2.49 [1.37, 4.54]
Leong et al22
0.4136
0.1025
45.4%
1.51 [1.24, 1.85]
0
0.3263
10.0%
1.00 [0.53, 1.90]
100.0%
1.59 [1.28, 1.97]
Mornos et al27 Total (95% CI)
Heterogeneity: Tau2 = 0.02; Chi2 = 4.55, df = 3 (P = 0.21); I2 = 34% Test for overall effect: Z = 4.18 (P < 0.0001) Log [Hazard Ratio]
Study
SE
Weight
0.2
Hazard Ratio, 95% CI
0.5
1
2
5
Hazard Ratio, 95% CI
GLS Ersbøll et al19
0.7529
0.214
24.5%
2.12 [1.40, 3.23]
Haugaa et al13
–0.2083
0.2549
22.6%
0.81 [0.49, 1.34]
Haugaa et al18
1.1555
0.5142
12.6%
3.18 [1.16, 8.70]
Haugaa et al20
0
0.3442
18.6%
1.00 [0.51, 1.96]
Mornos et al27
0.0647
0.2737
21.7%
1.07 [0.62, 1.82]
100.0%
1.35 [0.84, 2.14]
Total (95% CI) GLS vs LVMD (per 1-SD Increase)
Heterogeneity: Tau2 = 0.18; Chi2 = 12.72, df = 4 (P = 0.01); I2 = 69% Test for overall effect: Z = 1.25 (P = 0.21) Mechanical Dispersion Ersbøll et al19
0.2096
0.0993
32.5%
Haugaa et al13
0.5185
0.1416
25.7%
1.68 [1.27, 2.22]
Haugaa et al18
0.3865
0.1652
22.4%
1.47 [1.06, 2.03]
Haugaa et al20
0.9127
0.3057
10.2%
2.49 [1.37, 4.54]
Mornos et al27
0
0.3263
9.3%
1.00 [0.53, 1.90]
100.0%
1.46 [1.18, 1.82]
Total (95% CI)
1.23 [1.02, 1.50]
Heterogeneity: Tau2 = 0.03; Chi2 = 7.99, df = 4 (P = 0.09); I2 = 50% Test for overall effect: Z = 3.41 (P = 0.0007)
0.2
0.5
1
2
5
The top forest plot shows the result of the comparison between left ventricular mechanical dispersion (LVMD) and left ventricular ejection fraction (LVEF), and the bottom forest plot shows the result of the comparison between LVMD and global longitudinal strain (GLS). Both forest plots display the summary hazard ratios per 1-SD increase and the 95% confidence intervals (CIs) in the multivariate model for increasing association of each parameter with ventricular arrhythmia.
efficacy in a large group, development of cutoffs,
This meta-analysis revealed that patients with VAs
and comparison with existing markers. It seems
had greater LVMD than those without (Figure 3).
likely that in addition to measuring contraction
However, setting a clear optimal cutoff of LVMD is
heterogeneity, LVMD is a marker of the forms
difficult because of differences among studies and
of
heterogeneity—including
differences in cardiac diseases. In patients with long-
nonuniform anisotropy, electric uncoupling, and
QT syndrome, LVMD is less pronounced compared
conduction
with those with structural disease. Furthermore,
electrophysiological block
contribute to VA.
and
slow
conduction—that
bundle branch block will increase LVMD with an
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019
Kawakami et al.
- 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
C ENTR AL I LL U STRA T I O N Predictive Value and Accuracy of Left Ventricular Mechanical Dispersion in Risk Stratification for Ventricular Arrhythmia
100 Specificity (%)
1 ROC AUC
0.8 0.6 0.4 0.2 0
60 40 20 0
0
20 40 60 80 100 Cut Off of Mechanical Dispersion (ms)
100
100
80
80
Accuracy (%)
Sensitivity (%)
80
60 40 20 0
0
20 40 60 80 100 Cut Off of Mechanical Dispersion (ms)
0
20 40 60 80 100 Cut Off of Mechanical Dispersion (ms)
60 40 20 0
0
20 40 60 80 100 Cut Off of Mechanical Dispersion (ms)
Publication
Optimal Cut Off (ms)
ROC AUC
Sensitivity (%)
Specificity (%)
Accuracy (%)
Haugaa et al13
70
0.84
65
92
78.5
Haugaa et
al18
72
0.8
67
89
78
Haugaa et
al20
47
0.75
80
62
71
50
0.81
91
55
73
61
0.84
85
73
79
101.2
0.685
38
92
65
63.5
0.71
70.6
63
66.8
Kosiuk et
al21
Nguyen et
al23
Matsuzoe et Candan et
al25
al28
Kawakami, H. et al. J Am Coll Cardiol Img. 2019;-(-):-–-. The sizes of the bubbles indicate the number of patients. AUC ¼ area under the curve; ROC ¼ receiver-operating characteristic.
unknown effect of cardiac risk. We therefore recom-
65% to 79%. In addition, Rodriguez-Zanella et al. (35)
mend using different cutoff values for patients with
recently calculated normal LVMD in 303 healthy sub-
long-QT syndrome and patients with cardiomyopa-
jects assessed using strain echocardiography and re-
thies and bundle branch blocks. Nonetheless, the
ported the normal LVMD to be 34 10 ms, with an
optimal cutoff value in 5 of 7 studies was >60 ms,
upper limit of normal of 56 ms. On the basis of these
which was also the difference in LVMD between pa-
findings, a LVMD of 60 ms seems reasonable as a cutoff
tients with and without VA events (62.7 16.6 ms)
value for risk stratification for VA in these patients.
(Figure 3A). When a cutoff value >60 ms was used for
Importantly, LVMD is unable to predict VA in pa-
predicting VA events in patients with ischemic or
tients receiving cardiac resynchronization therapy.
nonischemic cardiomyopathies, the range of area
Cardiac
under the curve was 0.69 to 0.84, with sensitivity of
changes regional timing to improve myocardial syn-
38% to 85%, specificity of 63% to 92%, and accuracy of
chrony and modulate arrhythmic risk (36).
resynchronization
therapy
profoundly
9
10
Kawakami et al.
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019 - 2019:-–-
Mechanical Dispersion and Ventricular Arrhythmias
STUDY LIMITATIONS. First, as with any meta-analysis of
published after 2010 (37). Nonetheless, further vali-
observational studies, variations in the inclusion
dation studies are needed to clarify intervendor and
criteria and endpoints are all potential sources of het-
intersoftware variability.
erogeneity among studies. We could not access patientlevel data to allow adjustment for other covariates that might influence the incidence of VA, including medication, laboratory data, and other imaging parameters. Second, a group of researchers (including the authors), whose work was seminal in defining this test, were responsible for 4 of 12 studies in our systematic review (13,18,20,26), including the 2 largest studies (19,20), accounting for close to one-half of the patients. Although, reproducibility analyses have been provided, prospective multicenter studies are needed to confirm the relationship between LVMD and VAs and to integrate LVMD and decision making regarding ICD therapy. Third, an overestimation of the pooled effect sizes is possible, as the small number of studies limits the assessment of publication bias. Fourth, strain imaging is dependent on highquality echocardiographic imaging and appropriate
CONCLUSIONS LVMD assessed using speckle tracking provides important predictive value for VA in patients with a number of cardiac diseases. The predictive value of LVMD appears to be superior to that of LVEF or GLS. Hence, the clinical use of LVMD assessed by speckle tracking
should
be
considered
for
noninvasive
assessment of patients with cardiac disease. Prospective multicenter studies are warranted to confirm the external validity of these findings and translate LVMD into clinical decision making regarding ICD therapy. ADDRESS FOR CORRESPONDENCE: Dr. Thomas H.
Marwick, Baker Heart and Diabetes Institute, 75 Commercial Road, Melbourne 3004, Australia. E-mail: [email protected].
imaging settings (e.g. a frame rate of 50 to 70 frames/s), and the performance of LVMD in settings with less experienced users than the investigators of these studies in unclear. Nonetheless, the reproducibility of
PERSPECTIVES COMPETENCY IN MEDICAL KNOWLEDGE:
LVMD seemed to be acceptable, with interclass cor-
Prediction of VA remains a challenge. Recent studies
relation coefficients of LVMD within and between
have demonstrated that LVMD assessed using speckle
observers being similar to those for GLS (Table 2).
tracking might be a powerful marker in risk
Finally, lack of intervendor and intersoftware
stratification for VA. This meta-analysis confirmed the
standardization is often posed as an important limi-
association between LVMD and the incidence of VA in
tation of speckle tracking. However, 11 of 12 studies in
a number of cardiac diseases, and LVMD appears to
this meta-analysis solely used the same vendor (GE),
have superior predictive value over LVEF and GLS for
and the remaining study also used the same vendor in
risk stratification.
several patients (Table 1). In addition, 10 of 12 studies in this meta-analysis used EchoPAC to calculate LV strain. Although only 4 of 10 studies reported the version of the software, we believe that the effect of differences among EchoPAC versions was limited,
TRANSLATIONAL OUTLOOK: Further prospective studies are needed to integrate LVMD and decision making regarding ICD therapy in a variety of cardiovascular diseases.
because all studies in this meta-analysis were
REFERENCES 1. Bertini M, Schalij MJ, Bax JJ, Delgado V.
Arrhythmia Rate in High-risk MI patients (SEARCH-
of the Marburg Cardiomyopathy Study. Circulation
Emerging role of multimodality imaging to evaluate patients at risk for sudden cardiac death. Circ Cardiovasc Imaging 2012;5:525–35.
MI). Europace 2009;11:476–82.
2003;108:2883–91.
4. Moss AJ, Zareba W, Hall WJ, et al. Prophylactic
7. Nesbitt GC, Mankad S, Oh JK. Strain imaging in
implantation of a defibrillator in patients with myocardial infarction and reduced ejection fraction. N Engl J Med 2002;346:877–83.
echocardiography: methods and clinical applications. Int J Cardiovasc Imaging 2009;25:9–22.
2. Macatangay C, Viles-Gonzalez JF, Goldberger JJ. Role of cardiac imaging in evaluating risk for sudden cardiac death. Card Electrophysiol Clin 2017;9: 639–50. 3. Santini M, Russo M, Botto G, et al. Clinical and arrhythmic outcomes after implantation of a defibrillator for primary prevention of sudden death in patients with post-myocardial infarction cardiomyopathy: the Survey to Evaluate
5. Kadish A, Dyer A, Daubert JP, et al. Prophylactic defibrillator implantation in patients with nonischemic dilated cardiomyopathy. N Engl J Med 2004;350:2151–8.
8. Marwick TH. Methods used for the assessment of LV systolic function: common currency or tower of Babel? Heart 2013;99:1078–86.
6. Grimm W, Christ M, Bach J, Muller HH, Maisch B. Noninvasive arrhythmia risk stratifica-
9. Stecker EC, Vickers C, Waltz J, et al. Populationbased analysis of sudden cardiac death with and without left ventricular systolic dysfunction: twoyear findings from the Oregon Sudden Unexpected
tion in idiopathic dilated cardiomyopathy: results
Death Study. J Am Coll Cardiol 2006;47:1161–6.
JACC: CARDIOVASCULAR IMAGING, VOL.
-, NO. -, 2019
- 2019:-–-
10. Goldberger JJ, Subacius H, Patel T, Cunnane R, Kadish AH. Sudden cardiac death risk stratification in patients with nonischemic dilated cardiomyopathy. J Am Coll Cardiol 2014;63:1879–89. 11. Helle-Valle T, Crosby J, Edvardsen T, et al. New noninvasive method for assessment of left ventricular rotation: speckle tracking echocardiography. Circulation 2005;112:3149–56.
ventricular arrhythmias after myocardial infarction. J Am Coll Cardiol Img 2013;6:841–50.
report from the American Heart Association. Circulation 2015;131:e29–322.
21. Kosiuk J, Dinov B, Bollmann A, et al. Association between ventricular arrhythmias and myocardial mechanical dispersion assessed by strain analysis in patients with nonischemic cardiomyopathy. Clin Res Cardiol 2015;104:1072–7.
30. Stevenson WG, Brugada P, Waldecker B, Zehender M, Wellens HJ. Clinical, angiographic, and electrophysiologic findings in patients with aborted sudden death as compared with patients
12. Haugaa KH, Amlie JP, Berge KE, Leren TP, Smiseth OA, Edvardsen T. Transmural differences in myocardial contraction in long-QT syndrome: mechanical consequences of ion channel dysfunction. Circulation 2010;122:1355–63.
22. Leong DP, Hoogslag GE, Piers SR, et al. The relationship between time from myocardial infarction, left ventricular dyssynchrony, and the risk for ventricular arrhythmia: speckle-tracking echocardiographic analysis. J Am Soc Echocardiogr 2015;28:470–7.
13. Haugaa KH, Smedsrud MK, Steen T, et al. Mechanical dispersion assessed by myocardial strain in patients after myocardial infarction for risk prediction of ventricular arrhythmia. J Am Coll
23. Nguyen BL, Capotosto L, Persi A, et al. Global and regional left ventricular strain indices in postmyocardial infarction patients with ventricular arrhythmias and moderately abnormal ejection
Cardiol Img 2010;3:247–56.
fraction. Ultrasound Med Biol 2015;41:407–17.
14. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate
24. Negishi K, Negishi T, Zardkoohi O, et al. Left atrial booster pump function is an independent predictor of subsequent life-threatening ventricular arrhythmias in non-ischaemic cardiomyopa-
healthcare interventions: explanation and elaboration. BMJ 2009;339:b2700. 15. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-an-
thy. Eur Heart J Cardiovasc Imaging 2016;17: 1153–60.
alyses. BMJ 2003;327:557–60.
25. Matsuzoe H, Tanaka H, Matsumoto K, et al. Left ventricular dyssynergy and dispersion as
16. Downs SH, Black N. The feasibility of creating a checklist for the assessment of the methodological quality both of randomised and nonrandomised studies of health care interventions.
determinant factors of fatal ventricular arrhythmias in patients with mildly reduced ejection fraction. Eur Heart J Cardiovasc Imaging 2016;17: 334–42.
J Epidemiol Commun Health 1998;52:377–84.
26. Hasselberg NE, Haugaa KH, Bernard A, et al. Left ventricular markers of mortality and ventricular arrhythmias in heart failure patients with cardiac resynchronization therapy. Eur Heart J Cardiovasc Imaging 2016;17:343–50.
17. Levy PT, Sanchez Mejia AA, Machefsky A, Fowler S, Holland MR, Singh GK. Normal ranges of right ventricular systolic and diastolic strain measures in children: a systematic review and metaanalysis. J Am Soc Echocardiogr 2014;27:549–60. e3. 18. Haugaa KH, Goebel B, Dahlslett T, et al. Risk assessment of ventricular arrhythmias in patients with nonischemic dilated cardiomyopathy by strain echocardiography. J Am Soc Echocardiogr 2012;25:667–73. 19. Ersboll M, Valeur N, Andersen MJ, et al. Early echocardiographic deformation analysis for the prediction of sudden cardiac death and lifethreatening arrhythmias after myocardial infarction. J Am Coll Cardiol Img 2013;6:851–60. 20. Haugaa KH, Grenne BL, Eek CH, et al. Strain echocardiography improves risk prediction of
Kawakami et al. Mechanical Dispersion and Ventricular Arrhythmias
27. Mornos C, Muntean D, Mornos A, et al. Risk stratification in patients with heart failure: the value of considering both global longitudinal left ventricular strain and mechanical dispersion. Can J Physiol Pharmacol 2017;95:1360–8. 28. Candan O, Gecmen C, Bayam E, Guner A, Celik M, Dogan C. Mechanical dispersion and global longitudinal strain by speckle tracking echocardiography: predictors of appropriate implantable cardioverter defibrillator therapy in hypertrophic cardiomyopathy. Echocardiography 2017;34:835–42. 29. Mozaffarian D, Benjamin EJ, Go AS, et al. Heart disease and stroke statistics—2015 update: a
with sustained ventricular tachycardia after myocardial infarction. Circulation 1985;71: 1146–52. 31. Hsia HH, Marchlinski FE. Characterization of the electroanatomic substrate for monomorphic ventricular tachycardia in patients with nonischemic cardiomyopathy. Pacing Clin Electrophysiol 2002;25:1114–27. 32. Vaduganathan M, Patel RB, Michel A, et al. Mode of death in heart failure with preserved ejection fraction. J Am Coll Cardiol 2017;69: 556–69. 33. Kutyifa V, Pouleur AC, Knappe D, et al. Dyssynchrony and the risk of ventricular arrhythmias. J Am Coll Cardiol Img 2013;6:432–44. 34. Leren IS, Hasselberg NE, Saberniak J, et al. Cardiac mechanical alterations and genotype specific differences in subjects with long QT syndrome. J Am Coll Cardiol Img 2015;8:501–10. 35. Rodriguez-Zanella H, Haugaa K, Boccalini F, et al. Physiological determinants of left ventricular mechanical dispersion: a 2-dimensional speckle tracking echocardiographic study in healthy volunteers. J Am Coll Cardiol Img 2018;11:650–1. 36. Gorcsan J, Haugaa Kristina H. Ventricular arrhythmias and reduced echocardiographic inferior wall strain. Circ Cardiovasc Imaging 2017;10: e005900. 37. Lang RM, Badano LP, Mor-Avi V, et al. Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging. J Am Soc Echocardiogr 2015;28:1–39. e14.
KEY WORDS mechanical dispersion, meta-analysis, strain, ventricular arrhythmia
A PP END IX For supplemental tables and figures, please see the online version of this paper.
11