Predictors of Response to Cardiac Resynchronization Therapy in the Multicenter

Automatic Defibrillator Implantation Trial With Cardiac Resynchronization Therapy

(MADIT-CRT)Clinical Perspective
Ilan Goldenberg, Arthur J. Moss, W. Jackson Hall, Elyse Foster, Jeffrey J. Goldberger, Peter
Santucci, Timothy Shinn, Scott Solomon, Jonathan S. Steinberg, David Wilber, Alon
Barsheshet, Scott McNitt, Wojciech Zareba and Helmut Klein
Circulation. 2011;124:1527-1536; originally published online September 6, 2011;
doi: 10.1161/CIRCULATIONAHA.110.014324
Circulation is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright 2011 American Heart Association, Inc. All rights reserved.
Print ISSN: 0009-7322. Online ISSN: 1524-4539

The online version of this article, along with updated information and services, is located on the
World Wide Web at:
http://circ.ahajournals.org/content/124/14/1527

Data Supplement (unedited) at:

http://circ.ahajournals.org/content/suppl/2011/09/08/CIRCULATIONAHA.110.014324.DC1.html

Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published
in Circulation can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial
Office. Once the online version of the published article for which permission is being requested is located,
click Request Permissions in the middle column of the Web page under Services. Further information about
this process is available in the Permissions and Rights Question and Answer document.
Reprints: Information about reprints can be found online at:
http://www.lww.com/reprints
Subscriptions: Information about subscribing to Circulation is online at:
http://circ.ahajournals.org//subscriptions/

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

Heart Failure
Predictors of Response to Cardiac Resynchronization
Therapy in the Multicenter Automatic Defibrillator
Implantation Trial With Cardiac Resynchronization
Therapy (MADIT-CRT)
Ilan Goldenberg, MD; Arthur J. Moss, MD; W. Jackson Hall, PhD; Elyse Foster, MD;
Jeffrey J. Goldberger, MD; Peter Santucci, MD; Timothy Shinn, MD; Scott Solomon, MD;
Jonathan S. Steinberg, MD; David Wilber, MD; Alon Barsheshet, MD; Scott McNitt, MA;
Wojciech Zareba, MD; Helmut Klein, MD; on behalf of the MADIT-CRT Executive Committee
BackgroundWe hypothesized that combined assessment of factors that are associated with favorable reverse remodeling
after cardiac resynchronization-defibrillator therapy (CRT-D) can be used to predict clinical response to the device.
Methods and ResultsThe study population comprised 1761 patients enrolled in the Multicenter Automatic Defibrillator
Implantation Trial With Cardiac Resynchronization Therapy (MADIT-CRT). Best-subset regression analysis was
performed to identify factors associated with echocardiographic response (defined as percent reduction in left ventricular
end-diastolic volume 1 year after CRT-D implantation) and to create a response score. Cox proportional hazards
regression analysis was used to evaluate the CRT-D versus defibrillator-only reduction in the risk of heart failure or
death by the response score. Seven factors were identified as associated with echocardiographic response to CRT-D and
made up the response score (female sex, nonischemic origin, left bundle-branch block, QRS 150 milliseconds, prior
hospitalization for heart failure, left ventricular end-diastolic volume 125 mL/m2, and left atrial volume 40 mL/m2).
Multivariate analysis showed a 13% (P0.001) increase in the clinical benefit of CRT-D per 1-point increment in the
response score (range, 0 14) and a significant direct correlation between risk reduction associated with CRT-D and
response score quartiles: Patients in the first quartile did not derive a significant reduction in the risk of heart failure or
death with CRT-D (hazard ratio0.87; P0.52); patients in the second and third quartiles derived 33% (P0.04) and
36% (P0.03) risk reductions, respectively; and patients in the upper quartile experienced a 69% (P0.001) risk
reduction (P for trend0.005).
ConclusionCombined assessment of factors associated with reverse remodeling can be used for improved selection of
patients for cardiac resynchronization therapy.
Clinical Trial RegistrationURL: http://www.clinicaltrials.gov. Unique identifier: NCT00180271.
(Circulation. 2011;124:1527-1536.)
Key Words: cardiac pacing, artificial heart failure prevention and control ventricular remodeling

uring the past 10 years, several randomized clinical

trials have shown that cardiac resynchronization therapy
(CRT) is associated with reduced morbidity and mortality in
patients with left ventricular dysfunction and advanced heart
failure (HF) symptoms who have a wide QRS duration.15
Recently, the Resynchronization Reverses Remodeling in
Systolic Left Ventricular Dysfunction (REVERSE) trail and

the Multicenter Automatic Defibrillator Implantation Trial

With Cardiac Resynchronization Therapy (MADIT-CRT)
have extended the observed benefit to patients with less
advanced (New York Heart Association [NYHA] class I/II)
signs and symptoms.6 8 Although most treated patients respond to CRT, approximately one third are considered
nonresponders in clinical trials using a variety of measures of

Received July 21, 2010; accepted July 29, 2011.

From the Heart Research Follow-Up Program (I.G., A.J.M., W.J.H., S.M., A.B., W.Z., H.K.) and Department of Biostatistics (W.J.H.), University of
Rochester Medical Center, Rochester, NY; University of California Medical Center, San Francisco (E.F.); Northwestern University Feinberg School of
Medicine, Chicago, IL (J.J.G.); Loyola University Medical Center, Maywood, IL (D.W., P.S.); Michigan Heart PC, Ypsilanti (T.S.); Cardiovascular
Division, Brigham and Womens Hospital, Boston, MA (S.D.S.); and St. Lukes and Roosevelt Hospitals, New York, NY (J.S.S.).
Members of the MADIT-CRT Executive Committee are listed in the Acknowledgments.
Guest Editor for this article was Salvador Borges-Neto, MD.
The online-only Data Supplement is available with this article at http://circ.ahajournals.org/lookup/suppl/doi:10.1161/CIRCULATIONAHA.
110.014324/-/DC1.
Correspondence to Ilan Goldenberg, MD, Heart Research Follow-Up Program, Box 653, Cardiology Division, University of Rochester Medical Center,
Rochester, NY 14642. E-mail Ilan.Goldenberg@heart.rochester.edu
2011 American Heart Association, Inc.
Circulation is available at http://circ.ahajournals.org

DOI: 10.1161/CIRCULATIONAHA.110.014324

Downloaded from http://circ.ahajournals.org/

by guest on July 6, 2013
1527

1528

Circulation

October 4, 2011

Figure 1. Flow diagram of study design

showing the number of patients and the
corresponding number of clinical events
(comprising the first occurrence of heart
failure or death) in each step of the analysis. In the derivation analyses, 371
patients of the 1089 in the cardiac
resynchronization-defibrillator therapy
(CRT-D) group (34%) were omitted
because of missing paired echocardiographic data; in the validation, 59 patients
of the 1820 study population (3%) were
omitted owing to missing information on
at least 1 of the 7 chosen covariates. ICD
indicates implantable cardioverter-defibrillator.

clinical responsiveness.9 12 Currently, however, there are

limited data on factors that can distinguish responders from
nonresponders among eligible patients.

Clinical Perspective on p 1536

Cardiac resynchronization therapy is associated with reverse remodeling, a process characterized by a reduction in
left ventricular volumes leading to improved systolic and
diastolic function.1 8 Recent data suggest that the magnitude
of reverse remodeling can be used to predict the subsequent
outcome of CRT-treated patients.1315 We hypothesized that
combined assessment of factors that are associated with
favorable reverse remodeling can be used to predict clinical
response to CRT.

Methods

and at 1 year. Paired echocardiograms from baseline and at 12

months with the device turned on were available in 718 of 1089
CRT-Dtreated patients (66%).
Echocardiograms were sent on videotape or digital storage to the
echocardiographic core laboratory at Brigham and Womens Hospital, where they were screened for quality and left ventricular, right
ventricular, and left atrial measurements were made. Echocardiographic parameters were measured according to established American Society of Echocardiography protocols.17 Left ventricular and
atrial volumes were measured by the Simpson method of disks in the
apical 4- and 2-chamber views and averaged. Left ventricular
ejection fractions were calculated according to standard methods.
Reproducibility of the primary volumetric measures was assessed by
having the primary observer reanalyze 101 random studies. The
coefficients of variation for left ventricular end-diastolic volume
(LVEDV), end-systolic volume (LVESV), and ejection fraction were
5.2%, 6.2%, and 5.5%, respectively.

Outcome Measures

Study Population

Remodeling Effects

The design and primary results of MADIT-CRT have been published

recently.8 Briefly, MADIT-CRT was designed to determine whether
CRT with a defibrillator (CRT-D) would reduce the risk of death or
HF events in patients with mild cardiac symptoms, a reduced
ejection fraction, and wide QRS complex compared with implantable
cardioverter-defibrillator (ICD) therapy. The patients were randomly
assigned in a 3:2 ratio to receive either CRT-D or ICD. From
December 22, 2004, through April 23, 2008, a total of 1820 patients
were enrolled at 110 hospital centers. The protocol was approved by
the institutional review board at each of the participating centers.
Patients of either sex who were at least 21 years of age were enrolled
in the study if they had ischemic cardiomyopathy (NYHA class I or
II) or nonischemic cardiomyopathy (NYHA class II only), sinus
rhythm, an ejection fraction of 0.30, and prolonged intraventricular
conduction with a QRS duration of 130 milliseconds. All eligible
subjects met guideline indications for ICD therapy.16 Patients were
excluded from enrollment for a variety of reasons, as previously
reported.8 The present study population comprised 1761 patients
(97%) for whom complete information on baseline clinical, echocardiographic, and laboratory variables was available.

An echocardiographic response was defined as percent reduction in

LVEDV (assessed as a continuous measure) between enrollment and
1 year (calculated as the difference between 1-year cardiac volumes
and baseline cardiac volumes divided by baseline cardiac volumes).
It was alternatively assessed as percent reduction in LVESV and by
dichotomizing percentage reduction in cardiac volumes (prespecified
as 10% and 15% reductions in LVEDV and LVESV, respectively,
based on prior studies that assessed response to CRT17).

Echocardiographic Studies
Echocardiograms were obtained according to a study specific protocol at baseline, which was before device implantation (n1809),

Clinical Response
The primary end point for clinical response was defined as a first HF
event or death, whichever came first, during follow-up. The secondary end point was defined as the first occurrence of HF, death, or
advancement to NYHA class II during follow-up. The change in
the 6-minute walk test at 1 year after enrollment (among the 1500
patients with available data) and the separate occurrence of all-cause
mortality were also assessed as secondary end points.

Study Design
The present study was carried out in 3 steps (Figure 1). In the first
step, we identified factors associated with a favorable echocardiographic response to CRT-D therapy among the 718 CRT-Dallocated patients (66%) with active CRT-D use for 1 year and paired
baseline and 1-year echocardiograms. In the second step, we con-

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

Goldenberg et al
structed a response score to CRT-D by combined assessment of the
relative contribution of the factors that were identified to be
associated with a favorable echocardiographic response to CRT-D.
In the third step, we assessed the clinical benefit of CRT-D versus
ICD-only therapy by response score quartile in MADIT-CRT patients with complete baseline information variables (n1761 [97%])
and by assessing the interaction with treatment of the total response
score; the benefit was likewise assessed with individual response
scores.
The sample sizes available for each step in the analysis and the
corresponding end point events are shown in Figure 1. The methods
used in each of the 3 steps of the analysis are described in the
Statistical Analysis section.

Statistical Analysis
Covariates Associated With Echocardiographic Response
to Cardiac Resynchronization-Defibrillator Therapy
We included 31 potential clinical, electrocardiographic, and laboratory binary risk factors (listed in Table I in the online-only Data
Supplement) for the echocardiographic response model. Numeric
variables were made binary by the use of cut points with the goal of
finding a simple, easily implemented scoring method to be derived
from them. Thresholds for categorization of numeric variables were
prespecified using clinical and laboratory-accepted criteria. Univariate relationships between candidate covariates and an echocardiographic response (as defined above) in the CRT-D arm of the trial
were assessed by t tests (2 for binary responses). The covariates
with values of P0.20 were further evaluated by carrying out a
best-subset regression analysis in the CRT-D arm, examining the
models created from all possible combinations of predictor variables,
and using a penalty of 3.84 on the likelihood ratio 2 value for any
additional factor included (corresponds to a P of 5% for a 1-df 2
test). Model selection was repeated after unselected factors were
dropped, one at a time, to minimize the effects of missing data.
However, because of the relatively small number of patients without
complete baseline information (n59 [3.2%]), no further adjustment
was made for missing data.

Response Score
After selection of binary covariates, each was assigned a numeric
value based on the relative value of its regression coefficient in the
multivariate regression model (specifically, the response factor with
the lowest regression coefficient among the 7 factors in the model
was assigned a numeric value of 1, and the other 6 factors were
assigned numeric values based on the relative values of their
regression coefficients to that of the lowest value). A response score
was constructed in each patient by adding the assigned numeric
values of the factors identified in each patient, and the study
population was categorized into approximate quartiles based on the
distribution of the response scores (assessed as an ordinal measure)
among patients. The clinical characteristics of study patients were
compared across treatment groups by response score quartiles using
the 2 test for categorical variables and the Wilcoxon rank-sum test
for continuous variables.

Clinical Benefit of Cardiac ResynchronizationDefibrillator Therapy Versus Implantable

Cardioverter-DefibrillatorOnly Therapy
KaplanMeier estimates for HF or death in each response quartile,
stratified by treatment arm, were determined and statistically evaluated with the log-rank test. Cox proportional hazards regression
analyses were carried out in the total population for the assessment
of the primary (first occurrence of HF or death) and secondary (first
occurrence of HF, death, or advancement to NYHA class II) end
points. The following covariates were included in the regression
models: response score quartiles, treatment arm, and their interactions (CRT-D effect in each response score quartile). The response
score was assessed for its linearity in relationship with the primary
end point by analyzing smaller response score subgroups than were
used in the quartile analysis. The hazard ratios for each of these

Response to CRT

1529

subgroups were plotted across the response score subgroups. In an

additional analysis, the response score was assessed as a continuous
measure (by replacing the response score quartile covariate in the
Cox models with a single response score numeric covariate), again
with assessment for linearity. Checks were made for validity of the
proportional hazards assumption.
Multivariate regression analysis was used to evaluate the change
in 6-minute walk test at 1 year compared with baseline among
CRT-D versus ICD-only patients by response score quartiles after
assessment of linearity in smaller response score subgroups of
CRT-D patients. Covariates in the regression models were the same
as in the Cox proportional hazards regression analysis.
All P values were 2 sided, and values of P0.05 were considered
significant. Analyses were performed with SAS software (version
9.20).

Sensitivity and Bootstrap Validation Studies

Several analyses were carried out (described in more detail in the
Appendix in the online-only Data Supplement). First, the response
score was developed in a subgroup of the CRT-D patients, whereas
the clinical benefit was assessed using all patients, including this
subgroup, subgroup A. As a sensitivity analysis, we partitioned the
remaining CRT-D group into 2 subgroups, subgroups B and C, with
subgroup B similar to subgroup A in that all patients had active
CRT-Ds at baseline and at 12 months but did not have paired
echocardiographic studies. Subgroup C consisted of the remainder of
the CRT-D group, including those dying or withdrawing within 12
months and others not having an active CRT-D for 12 months. We
then assessed the clinical benefit in each of the 3 subgroups relative
to ICD-only patients. Other analyses included consideration of other
choices of baseline factors for the echocardiographic response score
and by bootstrapping the selection of factors, bootstrapping the
scoring derived from the selected factors, and bootstrapping the use
of the response score in distinguishing risk for the CRT-D versus
ICD-only groups.

Results
Predictors of Echocardiographic Response to
Cardiac Resynchronization-Defibrillator Therapy
A best-subset regression analysis in the CRT-D arm of the
trial identified 7 factors (from the 31 candidate covariates
listed in Table II in the online-only Data Supplement) as
being jointly associated with a favorable echocardiographic
response to CRT-D therapy (Table 1). These factors were
female sex, nonischemic origin of cardiomyopathy, QRS
150 milliseconds, the presence of left bundle-branch block
on the baseline ECG, hospitalization for HF at any time
before enrollment, baseline LVEDV 125 mL/m2 (above the
first quartile), and baseline left atrial volume 40 mL/m2
(below the first quartile). The same 7 factors were identified
when a favorable echocardiographic response was defined as
percent reduction in LVESV (Table 2) and when percent
reductions for LVEDV and LVESV were dichotomized at
10% and 15%, respectively (Tables 1 and 2).

Categorization of Response Score Groups

After selection of the 7 covariates, each was assigned a numeric
value based on its relative effect in the regression model (as
reflected by the point estimate of each covariate in the 7-variable
model). The regression model showed that the contribution was
lowest for prior hospitalization for HF; intermediate for female
sex, nonischemic cardiomyopathy, left bundle-branch block,
QRS 150 milliseconds, and LVEDV; and highest for left atrial
volume. Accordingly, prior HF hospitalization was assigned a
numeric value of 1; the intermediate factors, a value of 2 each;

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

1530

Circulation

October 4, 2011

Table 1. Factors Identified as Jointly Contributing to a Favorable Echocardiographic Response in the Cardiac
Resynchronization-Defibrillation Therapy Arm of the Trial: Reduction in Left Ventricular End-Diastolic Volume
High Response
Risk Factor
(Covariate)

Definition

Low Response

Reduction
10%
(SD), % Reduction, %

Definition

High vs Low*

Reduction
10%
Difference in
(SD), % Reduction, % Reduction (SE), %*

Score

Women

275

24 (11)

91

Men

814

19 (11)

83

2.9 (1.0)

0.003

CMP origin

Nonischemic

491

24 (12)

90

Ischemic

598

18 (10)

82

4.2 (0.9)

0.001

QRS

150 ms

688

22 (12)

88

383

18 (9)

79

2.7 (0.9)

0.003

LBBB

750

22 (11)

88

Non-LBBB

321

12 (10)

79

3.4 (1.0)

0.001

Yes

493

22 (12)

87

No

578

19 (11)

82

1.9 (0.8)

0.02

125 mL/m2 803

21 (11)

88

125 mL/m2 267

18 (11)

83

4.2 (1.1)

0.001

40 mL/m2 258

23 (12)

87

40 mL/m2 811

20 (11)

78

5.6 (1.0)

0.001

Sex

QRS pattern
Prior HF hospitalization
Baseline LVEDV
Baseline LAV

150 ms

CMP indicates cardiomyopathy; LBBB, left bundle-branch block; HF, heart failure; LVEDV, left ventricular end-diastolic volume; and LAV, left atrial volume.
*Denotes the difference in percent reduction in LVEDV at 1 year between high and low responders of each covariate; these descriptive findings were derived from
the covariate-selection regression model for the end point of percent reduction in LVEDV (each with a P0.025); see the Statistical Analysis section.
Regression coefficients for each covariate were as follows: sex, 2.92; CMP origin, 4.16; QRS, 2.67; prior HF hospitalization, 1.88; baseline LVEDV, 4.18; and
baseline LAV, 5.57.

and left atrial volume, a value of 3 (Table 1). A response score

was constructed by adding the numeric values of the factors
identified in each patient. The response score ranged from 0 to
14. Subsequently, the study population was categorized into
approximate quartiles based on the distribution of the response
scores. Group 1 comprised 391 patients in the lowest response
score quartile (with a score of 0 4); group 2 comprised 401
patients in the second response score quartile (with a score of
5 6); group 3 comprised 469 patients in the third response score
quartile (with a score of 7 8); and group 4 comprised 500
patients in the upper response score quartile (with a score of
9 14).
The baseline clinical characteristics of study patients by
response quartiles are shown in Table 3. Patients with higher
response scores showed somewhat more favorable clinical
characteristics, including a younger age and a lower serum
creatinine, but had higher use of digitalis and aldosterone
antagonists (Table 3). Response score quartiles showed a
direct correlation with percent reduction in cardiac volumes
among patients randomized to CRT-D therapy (Figure 2A
and 2B). In contrast, in the ICD-only arm of the trial, percent

reductions in cardiac volumes were significantly lower (7%

mean percent reduction of both LVEDV and LVESV) and
related only to the origin of cardiomyopathy (Tables IIA and
IIB in the online-only Data Supplement).

Relation of Response Score to Clinical Benefit of

Cardiac Resynchronization-Defibrillator Therapy
During follow-up, 367 patients (20%) experienced the primary
clinical end point of the study (the first occurrence of HF or
death during follow-up), and 629 patients (35%) experienced the
secondary end point (the first occurrence of HF, death, or
advancement to NYHA class II during follow-up).
The benefits of CRT-D versus ICD-only therapy for the
reduction in the primary and secondary end points are shown
in Tables 4 and 5, respectively. Cox proportional hazards
regression modeling showed that among patients in the
lowest response score quartile, CRT-D therapy was not
associated with a statistically significant reduction in the risk
of HF or death compared with ICD-only therapy (hazard
ratio0.87; P0.52); among patients in the second and third
response score quartiles, CRT-D therapy was associated with

Table 2. Factors Identified as Jointly Contributing to a Favorable Echocardiographic Response in the Cardiac
Resynchronization-Defibrillation Therapy Arm of the Trial: Reduction in Left Ventricular End-Systolic Volume
High Response

Definition

Low Response

Reduction
15%
(SD), % Reduction, %

Definition

High vs Low*

Reduction
15%
Difference in
(SD), % Reduction, % Reduction (SE), %*

Score
2

Women

275

38 (15)

94

Men

814

31 (15)

88

4.4 (1.3)

0.01

CMP origin

Nonischemic

491

37 (16)

93

Ischemic

598

29 (14)

87

4.6 (1.2)

0.001

QRS

150 ms

688

34 (15)

92

383

28 (14)

85

3.6 (1.1)

0.004

LBBB

750

35 (15)

92

Non-LBBB

321

26 (15)

85

5.3 (1.4)

0.001

Yes

No

Sex

QRS pattern
Prior HF hospitalization
Baseline LVEDV
Baseline LAV

150 ms

493

32 (15)

91

578

27 (14)

87

2.9 (1.2)

0.02

125 mL/m2 803

33 (15)

91

125 mL/m2 267

29 (16)

84

3.9 (1.4)

0.006

40 mL/m2 258

35 (16)

93

40 mL/m2 811

30 (15)

87

6.5 (1.2)

0.001

CMP indicates cardiomyopathy; LBBB, left bundle-branch block; HF, heart failure; LVEDV, left ventricular end diastolic volume; and LAV, left atrial volume.
*Denotes the difference in percent reduction in left ventricular end-systolic volume at 1 year between high and low responders of each covariate; these descriptive
findings were derived from the covariate-selection regression model for the end point of percent reduction in left ventricular end diastolic volume (each with a P-value
0.025); see the Statistical Analysis section.

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

Goldenberg et al
Table 3.

Response to CRT

1531

Baseline Characteristics of Study Patients by Response Score Quartile

Response Score Quartile

Characteristic

Score

04

5 6

7 8

391

401

469

500

66 (10)

66 (10)

64 (11)

62 (11)

0.001

58

62

60

59

0.64

Age, mean (SD), y

CRT-D, %
Nonischemic CMP, %

22

53

85

BUN, mean (SD), mg/dL

22 (9)

22 (9)

21 (8)

21 (9)

0.05

NA

Creatinine, mean (SD), mg/dL

1.2 (0.3)

1.3 (0.5)

1.2 (0.3)

1.1 (0.3)

0.001

12

21

55

NA*

NYHA class II, %

72

79

88

99

0.001

Hypertension, %

68

66

60

61

0.06

Diabetes mellitus, %

33

37

29

24

0.001

Past CABG, %

54

39

23

0.001

Past PCI, %

43

36

24

11

0.001

QRS duration 150 ms, %

18

57

81

90

NA*

LBBB, %

22

63

88

97

NA*

Prior HF hospitalization, %

36

45

48

57

NA*

Atrial fibrillation 1 mo before enrollment, %

10

15

15

0.001

10

0.01

15

12

10

12

0.32

Female sex, %

Prior ventricular arrhythmias, %

Cigarette smoking, %
SBP, mean (SD), mm Hg

123 (18)

123 (17)

122 (17)

122 (17)

0.31

DBP, mean (SD), mm Hg

72 (11)

71 (10)

72 (11)

71 (10)

0.79

BMI, mean (SD), kg/m2

29 (5)

29 (5)

29 (5)

28 (6)

0.06

EF, mean (SD), %

29 (3)

29 (3)

20 (3)

30 (4)

0.001

LVESV, mean (SD), mL/m2

82 (18)

85 (22)

92 (24)

92 (25)

NA*

115 (23)

120 (28)

128 (30)

129 (31)

NA*

48 (9)

47 (10)

48 (10)

45 (11)

NA*

LVEDV, mean (SD), mL/m

LAV, mean (SD), mL/m2

Apical position of LV lead, %

12

14

16

13

0.86

ACE inhibitors, %

76

77

77

80

0.89

ARBs, %

20

19

20

21

0.94

Digitalis, %

24

22

26

30

0.05

Diuretics, %

73

76

73

77

0.45

-blockers, %

93

91

93

96

0.02

Aldosterone antibody, %

28

28

33

38

0.006

CRT-D indicates cardiac resynchronization-defibrillator therapy; CMP, cardiomyopathy; BUN, blood urea nitrogen; NYHA, New York
Heart Association; CABG, coronary artery bypass surgery; PCI, percutaneous coronary revascularization; LBBB, left bundle-branch
block; HF, heart failure; SBP, systolic blood pressure; DBP, diastolic blood pressure; BMI, body mass index; EF, ejection fraction;
LVESV, left ventricular end-systolic volume; LVEDV, left ventricular end-diastolic volume; LAV, left atrial volume; LV, left ventricular;
ACE, angiotensin-converting enzyme inhibitors; and ARBs, angiotensin receptor blockers.
*Statistical comparison not applicable because response groups are based on the factor.
Assessed among 776 CRT-Dtreated patients with available data.

33% (P0.04) and 36% (P0.03) risk reductions, respectively; and among patients in the upper response score
quartile, CRT-D was associated with a pronounced 69%
(P0.001) reduction in the risk of HF or death. Accordingly,
there was a significant direct correlation between the clinical
benefit of CRT-D therapy and response score quartiles (P for
trend0.005). Furthermore, CRT-D therapy was shown to be
associated with a significantly greater clinical benefit among
patients in the upper response score quartile than among
patients in each of the lower 3 response score quartiles (P for all

treatment-by-response quartile interactions 0.05; Table 4).

Assessing the response score as a continuous measure showed
that the benefit of CRT-D therapy for the reduction in the risk of
HF or death was increased by 13% (95% confidence interval,
4 19; P0.001) for each 1-point increment in the response
score (range, 0 14); likewise, reductions were increased by an
estimated 24%, 50%, and 75%, for 2-, 5-, and 10-point increments in the response score, respectively.
A similar pattern showing a significant direct correlation
between risk reduction associated with CRT-D and response

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

1532

Circulation

October 4, 2011

Figure 2. Average percent reduction in left ventricular end-diastolic volume (LVEDV; A) and end-systolic volume (LVESV; B), among
CRT-D-allocated patients in a given response category.

score quartiles was also shown when the secondary end point
of HF, death, or advancement to NYHA II was assessed (P
for trend0.002; Table 5). The risk of this end point was
reduced by 10% (95% confidence interval, 515; P0.001)
for each 1-point increment in the response score. Furthermore, regression analysis showed that, compared with ICDonly therapy, treatment with CRT-D was associated with an
average of a 16-m increment (standard error, 6 m) in the
6-minute walk test at 1 year per each response score quartile
(P for trend0.01).
Notably, when the response score was created with replacement of left atrial volume and LVEDV with the respecTable 4. Multivariate Analysis: Clinical Benefit of Cardiac
Resynchronization-Defibrillation Therapy Versus Implantable
Cardioverter-DefibrillatorOnly Therapy by Response Group: End
Point of First Occurrence of Heart Failure or Death*

tive atrial and ventricular diameters (left atrial diameter 2

cm/m2 [lower quartile]; left ventricular diastolic diameter 3
cm/m2 [upper 3 quartiles], respectively), the CRT-D versus
ICD-only therapy for the reduction of HF or death was also
shown to have a significant direct correlation with increasing
response score quartiles (P for trend0.003; Table III in the
online-only Data Supplement).
Consistent with these findings, KaplanMeier analysis
showed that, among patients in the lowest response score
Table 5. Multivariate Analysis: Clinical Benefit of Cardiac
Resynchronization-Defibrillation Therapy Versus Implantable
Cardioverter-DefibrillatorOnly Therapy by Response Group:
End Point of First Occurrence of Heart Failure, Death, or
Advancement to New York Heart Association Class >II*
CRT-D vs ICD-Only Risk of
HF, Death, or Advancement
to New York Heart Association
Class II

CRT-D vs ICD-Only Risk

of HF or Death
Response Groups

Score

HR

95% CI

P for
Trend

All patients (n1761)

0 14

0.62

0.51 0.77

0.001

NA

By response score
quartile

Response Groups

Score

HR

95% CI

All patients (n1761)

0 14

0.62

0.50 0.76

0.001

NA

0.002

By response score
quartile

1 (n391)

04

0.87

0.581.32

0.52

1 (n391)

04

0.97

0.711.32

0.83

2 (n401)

56

0.67

0.460.98

0.04

2 (n401)

56

0.74

0.541.01

0.06

3 (n469)

78

0.64

0.430.97

0.03

3 (n469)

78

0.68

0.500.93

0.02

4 (n500)

0.31

0.200.53

0.001

4 (n500)

0.47

0.340.65

0.001

0.87

0.810.96

0.001

0.90

0.850.95

0.001

By individual response
scores (per unit
increment)

0.005

CRT-D indicates cardiac resynchronization-defibrillator therapy; ICD, implantable cardioverter-defibrillator; HF, heart failure; HR, hazard ratio; and CI,
confidence interval.
*Results obtained from Cox proportional hazards regression models adjusted
for treatment arm and response group. Treatment effect in each response
group was obtained by including treatment-by-response group interaction
terms in the multivariate models.
P for trend obtained by including an interaction term between treatment
arm and the response group assessed as a single ordinal categorical variable
in the multivariate model.
P for treatment-by-response quartile 1 through 3 interactions with response quartile 4 (ie, each P effectively reflects a comparison between the
CRT-D vs ICD-only HR in quartile 4 with the CRT-D vs ICD-only HR in another
quartile): quartile 10.002, quartile 20.02, and quartile 30.03.
Hazard ratio decrease per unit increase in response score based on the
interaction between treatment group and score.

By individual response
scores (per unit
increment)

CRT-D indicates cardiac resynchronization-defibrillator therapy; ICD, implantable cardioverter-defibrillator; HF, heart failure; HR, hazard ratio; and CI,
confidence interval.
*Results obtained from Cox proportional hazards regression models adjusted
for treatment arm and response group. Treatment effect in each response
group was obtained by including treatment-by-response group interaction
terms in the multivariate models.
P for trend obtained by including an interaction term between treatment
arm and the response group assessed as a single ordinal categorical variable
in the multivariate model.
P for treatment-by-response quartile 1 through 3 interactions with response quartile 4 (ie, each P effectively reflects a comparison between the
CRT-D vs ICD-only HR in quartile 4, with the CRT-D vs ICD-only HR in another
quartile): quartile 10.002, quartile 20.05, and quartile 30.11.
Hazard ratio decrease per unit increase in response score based on the
interaction between treatment group and score.

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

Goldenberg et al

Response to CRT

1533

Figure 3. Cumulative probability of heart

failure (HF) or death by treatment arm in
response score quartiles 1 through 4
(AD, respectively). CRT-D indicates
cardiac resynchronization therapydefibrillator; ICD, implantable
cardioverter-defibrillator.

quartile, the cumulative probability of HF or death at 3 years

of follow-up was virtually identical between the 2 treatment
arms (Figure 3A); patients in the intermediate response score
quartiles (quartiles 2 and 3) experienced 26% and 36%
reductions in cumulative event rates, respectively, at 3 years
with CRT-D therapy (Figure 3B and 3C, respectively),
whereas patients in the upper response score quartile experienced a pronounced 65% reduction in the cumulative rate of
HF or death with CRT-D therapy at 3 years of follow-up
(Figure 3D).

Relation of Response Score to Mortality

Reduction Associated With Cardiac
Resynchronization-Defibrillator Therapy
Among patients in the upper response score quartile, CRT-D
therapy was associated with a statistically significant reduction in the risk of all-cause mortality compared with ICDonly therapy (hazard ratio0.39; 95% confidence interval,
0.16 0.98; P0.04). Accordingly, 3-year mortality rates in
the upper response score quartile were significantly higher in
the ICD-only arm (8%) compared with the CRT-D arm (5%;
log-rank P0.039 for the overall mortality difference during
follow-up). In contrast, among patients with a lower response
score (score 9), treatment with CRT-D was not associated
with a significant survival benefit (hazard ratio1.07; 95%
confidence interval, 0.721.59; P for treatment-by-responsescore interaction0.05).

Model Stability and Sensitivity Analyses

Several methods were used to validate the stability of model
selection for echocardiographic response to CRT-D and its
relationship to the clinical response to the device in the trial.
Results of some of these analyses are described in the
Appendix in the online-only Data Supplement. Briefly, assessing the clinical benefit in each of the 3 subgroups
(subgroups AC) led to very similar hazard rates for the

individual response score in the 3 subgroups (Table IV in the

online-only Data Supplement), providing justification for
pooling the 3 subgroups for the primary analyses. Second,
substitutions of individual identified covariates with other
choices resulted in weaker conclusions, and bootstrapping the
selection of covariates showed moderate stability associated
with echocardiographic response. Third, bootstrapping the
scoring derived from the selected factors showed high correlations among the derived response scores. Finally, bootstrapping also demonstrated considerable stability in the effects of
scores on evaluation of the relative risk for the primary end
point.

Discussion
Our findings from the MADIT-CRT population have several
important implications in terms of selection of patients for
treatment with CRT. We have identified 7 simple baseline
clinical and echocardiographic factors that were associated
with a favorable reverse remodeling effect in MADIT-CRT.
Combined assessment of the identified factors (through a
response score) showed a direct correlation with the clinical
benefit of CRT-D therapy in the trial. Accordingly, study
patients in the upper response score quartile (score 9)
derived significantly greater clinical benefit from CRT-D
therapy compared with patients with a lower response score,
including a 69% reduction in the risk of HF or death and a
61% reduction in the risk of all-cause mortality. In contrast,
CRT-D therapy was not associated with a statistically significant clinical benefit among patients with a low response
score (score, 0 4).
In patients with NYHA class III and ambulatory class IV
systolic HF and ECG evidence of ventricular dyssynchrony,
CRT (with and without a defibrillator) improves quality of
life and functional status, reduces HF-related hospitalizations,
and prolongs survival.15 Thus, there is a strong clinical
mandate for the use of CRT in eligible patients that is

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

1534

Circulation

October 4, 2011

supported by current guidelines.16 However, the echocardiographic and clinical benefit of CRT is not uniform, and
approximately one third of eligible patients have been considered nonresponders in clinical trials using a variety of
measures of clinical responsiveness.9 12 Similar to prior
studies among patients with more advanced HF symptoms,
the benefit of CRT-D therapy in the MADIT-CRT population
was not uniform, and treatment with the device in the study
was shown to be associated with a significant differential
effect when related to baseline factors, including sex, QRS
duration, and left bundle-branch block status.8,18 These consistent findings suggest that more appropriate selection of
patients for treatment with CRT is needed. Improved patient
selection may also reduce the relatively high costs associated
with treatment with the device in eligible patients.19,20
MADIT-CRT showed that CRT is associated with both a
reverse remodeling effect (including significant reductions in
LVESV and LVEDV and improvement in ejection fraction at
1 year of follow-up) and a clinical benefit (resulting in a
significant 34% reduction in the risk of HF or death with
CRT-D compared with ICD-only therapy) in mildly symptomatic patients with left ventricular dysfunction.8 Furthermore, the echocardiographic effects of CRT in the trial were
shown to be concordant with the subsequent clinical outcome
of study patients.13 Our findings extend these observations
and show that combined assessment of factors that were
associated with a favorable echocardiographic response to
CRT-D therapy during the trial can be used to distinguish
between patients in whom CRT-D was associated with
pronounced HF and mortality reductions and those in whom
treatment with the device was not associated with a significant clinical benefit.
Of the 7 identified baseline clinical and echocardiographic
covariates that made up the response score in the present
study, 4 clinical variables were also previously reported to
predict reverse remodeling in patients with more advanced
HF symptoms: a nonischemic origin of cardiomyopathy,15,21,22 female sex,22,23 QRS width,2,22,24 and the presence
of left bundle-branch block on the baseline ECG.25,26 In
contrast, prior studies did not identify baseline cardiac volumes and a history of HF hospitalization as independent
predictors of LV reverse remodeling in patients with advanced HF symptoms.22 Thus, it is possible that the relationships between the last 3 predictors of reverse remodeling in
MADIT-CRT are specific to the study population, or our
sample from it, that was made up of patients with less
advanced HF symptoms.
Notably, the presence of prior revascularization displayed
an inverse correlation with response score quartiles (Table 3)
but was not identified as independently related to the echocardiographic and clinical response to CRT-D therapy. These
findings may be due to the fact that the presence of prior
revascularization is correlated with factors that provide a
more significant and independent contribution to the response
score (including a history of prior hospitalization for HF, a
prolonged QRS duration, the presence of left bundle-branch
block, and left ventricular and left atrial volumes).
It should be noted that the lack of an individual factor in a
patient does not indicate a lack of clinical response to CRT.

Rather, the present results suggest that the combined presence

of factors that are associated with more favorable echocardiographic response to CRT-D is directly related to the
degree of clinical response to the device. Thus, individual
unfavorable factors were present in various degrees in patients in response score quartiles 2 through 4 in whom CRT
therapy was associated with increasing clinical benefit. These
findings stress the importance of a comprehensive approach
to risk assessment in candidates for CRT-D therapy.

Limitations
Patients in MADIT-CRT were followed up over an average
of 2.4 years. Therefore, further studies are needed to validate
the longer-term consistency of the present results in terms of
risk assessment for CRT-D therapy. It should also be stressed
that the present results concerning factors associated with
echocardiographic response (and their relative contribution to
clinical response) to CRT-D therapy are applicable only to
the MADIT-CRT population, who were patients with mild
HF symptoms.
The present response score comprises 2 components that rely
on echocardiographic assessment (including measurements of
atrial and ventricular volumes). It should be noted, however, that
the echocardiographic assessment in MADIT-CRT was carried
out in a core laboratory with excellent reproducibility,
whereas the repeatability of echocardiographic measurements
in individual laboratories may be less consistent. The fact that
similar results were obtained when left atrial and ventricular
diameters replaced the corresponding volumes as covariates
in the response score suggests that these more common and
easily derived measurements can also be used as part of a
response score for the prediction of response to CRT-D
therapy.
An apical position of the LV lead was recently shown to be
associated with attenuated CRT-D benefit in the MADIT-CRT
population.26a However, the present study shows that lead
position did not correlate with either echocardiographic
response to CRT-D therapy (0.22% reduction in LVEDV
for apical versus nonapical lead position; P0.88) or the
response score (Table 3), suggesting that the combined
assessment of simple clinical and echocardiographic variables provides prognostic information among CRT-D recipients incremental to sole assessment of LV lead position.
Nevertheless, it is possible that the relation between LV lead
position and scar location also relates to response to CRT-D
therapy. This information, however, was not available in the
present study. Furthermore, recent studies have shown that
additional clinical parameters, including scar burden and
areas of delayed activation, affect response to CRT-D therapy.2729 These parameters, however, were not assessed in the
present study. Thus, unmeasured parameters may also affect
CRT response beyond the 7 covariates that were identified in
the present study.
Our findings are derived from a single population study.
Thus, despite the fact that results of the present study were
consistent in several sensitivity analyses, the present findings
should be considered largely descriptive of what occurred in
the MADIT-CRT study rather than predictive of what might
occur in other or future settings.30 Thus, there is a great need

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

Goldenberg et al
to carry out similar analyses in other sets of clinical trial data.
These analyses would provide important confirmatory data
for the proposed response score.

Conclusions and Clinical Implications

Prior studies failed to demonstrate improved response rates to
CRT when adding echocardiographic measures of dyssynchrony.21,31,32 Thus, QRS width currently is the sole criterion
for treatment with the device in eligible patients, resulting in
a nonuniform effect of this costly technology in implanted
patients. Our findings suggest that baseline factors that are
associated with a favorable echocardiographic response to
CRT-D therapy can be used to identify patients who derive
clinical benefit from the device. Furthermore, we have shown
that the degree of clinical response to CRT-D therapy is
directly related to the number of factors associated with
echocardiographic response that are present in an individual
patient. These findings, if appropriately validated in similar
populations of CRT-D recipients, may be used for improved
selection of patients for treatment with CRT for the prevention of HF or death.

Acknowledgments
We want to thank the members of the Executive Committee of
MADIT-CRT: Mary W. Brown, MS; David S. Cannom, MD; James
P. Daubert, MD; Steven L. Higgins, MD; Mark Estes III, MD; Mark
A. Pfeffer, and Henry Greenberg, MD.

Sources of Funding
MADIT-CRT was supported by a research grant from Boston
Scientific Corp, St. Paul, MN, to the University of Rochester School
of Medicine and Dentistry.

Disclosures
Drs Moss, Solomon, Klein, Foster, Hall, and Zareba have received
research support for the conduct of the MADIT-CRT trial from
Boston Scientific through a grant to the University of Rochester. The
other authors report no conflicts.

References
1. Bristow MR, Saxon LA, Boehmer J, Krueger S, Kass DA, De Marco T,
Carson P, DiCarlo L, DeMets D, White BG, DeVries DW, Feldman AM;
Comparison of Medical Therapy, Pacing, and Defibrillation in Heart
Failure (COMPANION) Investigators. Cardiac-resynchronization therapy
with or without an implantable defibrillator in advanced chronic heart
failure. N Engl J Med. 2004;350:2140 2150.
2. Cleland JG, Daubert JC, Erdmann E, Freemantle N, Gras D, Kappenberger L, Tavazzi L; Cardiac Resynchronization-Heart Failure
(CARE-HF) Study Investigators. The effect of cardiac resynchronization
on morbidity and mortality in heart failure. N Engl J Med. 2005;352:
1539 1549.
3. Cleland JG, Daubert JC, Erdmann E, Freemantle N, Gras D, Kappenberger
L, Tavazzi L. Longer-term effects of cardiac resynchronization therapy on
mortality in heart failure [the CArdiac REsynchronization-Heart Failure
(CARE-HF) trial extension phase]. Eur Heart J. 2006;27:19281932.
4. Bradley DJ, Bradley EA, Baughman KL, Berger RD, Calkins H,
Goodman SN, Kass DA, Powe NR. Cardiac resynchronization and death
from progressive heart failure: a meta-analysis of randomized controlled
trials. JAMA. 2003;289:730 740.
5. Abraham WT, Young JB, Leon AR, Adler S, Bank AJ, Hall SA,
Lieberman R, Liem LB, OConnell JB, Schroeder JS, Wheelan KR;
Multicenter InSync ICD II Study Group. Effects of cardiac resynchronization on disease progression in patients with left ventricular systolic
dysfunction, an indication for an implantable cardioverter-defibrillator,
and mildly symptomatic chronic heart failure. Circulation. 2004;110:
2864 2868.

Response to CRT

1535

6. Linde C, Abraham WT, Gold MR, St John Sutton M, Ghio S, Daubert C.

Randomized trial of cardiac resynchronization in mildly symptomatic
heart failure patients and in asymptomatic patients with left ventricular
dysfunction and previous heart failure symptoms. J Am Coll Cardiol.
2008;52:1834 1843.
7. Daubert C, Gold MR, Abraham WT, Ghio S, Hassager C, Goode G,
Szili-T[iml]or[iml]ok T, Linde C; REVERSE Study Group. Prevention of
disease progression by cardiac resynchronization therapy in patients with
asymptomatic or mildly symptomatic left ventricular dysfunction:
insights from the European cohort of the REVERSE (Resynchronization
Reverses Remodeling in Systolic Left Ventricular Dysfunction) trial.
J Am Coll Cardiol. 2009;54:18371846.
8. Moss AJ, Hall WJ, Cannom DS, Klein H, Brown MW, Daubert JP, Estes
NA 3rd, Foster E, Greenberg H, Higgins SL, Pfeffer MA, Solomon SD,
Wilber D, Zareba W; MADIT-CRT Trial Investigators. Cardiacresynchronization therapy for the prevention of heart-failure events.
N Engl J Med. 2009;361:1329 1338.
9. Abraham WT, Fisher WG, Smith AL, Delurgio DB, Leon AR, Loh E,
Kocovic DZ, Packer M, Clavell AL, Hayes DL, Ellestad M, Trupp RJ,
Underwood J, Pickering F, Truex C, McAtee P, Messenger J; MIRACLE
Study Group, Multicenter InSync Randomized Clinical Evaluation.
Cardiac resynchronization in chronic heart failure. N Engl J Med. 2002;
346:18451853.
10. Young JB, Abraham WT, Smith AL, Leon AR, Lieberman R, Wilkoff B,
Canby RC, Schroeder JS, Liem LB, Hall S, Wheelan K; Multicenter
InSync ICD Randomized Clinical Evaluation (MIRACLE ICD) Trial
Investigators. Combined cardiac resynchronization and implantable cardioversion defibrillation in advanced chronic heart failure: the MIRACLE
ICD Trial. JAMA. 2003;289: 26852694.
11. Saxon LA, Boehmer JP, Hummel J, Kacet S, De Marco T, Naccarelli G,
Daoud E; VIGOR CHF and VENTAK CHF Investigators. Biventricular
pacing patients with congestive heart failure: two prospective randomized
trials. Am J Cardiol. 1999;83:120D123D.
12. Packer M. Proposal for a new clinical end point to evaluate the efficacy
of drugs and devices in the treatment of chronic heart failure. J Card Fail.
2001;7:176 182.
13. Solomon SD, Foster E, Bourgoun M, Shah A, Viloria E, Brown MW,
Hall WJ, Pfeffer MA, Moss AJ; MADIT-CRT Investigators. Effect of
cardiac resynchronization therapy on reverse remodeling and relation to
outcome: multicenter automatic defibrillator implantation trial: cardiac
resynchronization therapy. Circulation. 2010;122:985992.
14. Yu CM, Bleeker GB, Fung JW, Schalij MJ, Zhang Q, van der Wall EE,
Chan YS, Kong SL, Bax JJ. Left ventricular reverse remodeling but not
clinical improvement predicts long-term survival after cardiac resynchronization therapy. Circulation. 2005;112:1580 1586.
15. Ypenberg C, van Bommel RJ, Borleffs CJW, Bleeker GB, Boersma E,
Schalij MJ, Bax JJ. Long-term prognosis after cardiac resynchronization
therapy is related to the extent of left ventricular reverse remodeling at
midterm follow-up. J Am Coll Cardiol. 2009;53:483 490.
16. ACC/AHA/HRS 2008 guidelines for device-based therapy of cardiac
rhythm abnormalities: a report of the American College of Cardiology/
American Heart Association Task Force on Practice Guidelines (Writing
Committee to Revise the ACC/AHA/NASPE 2002 Guideline Update for
Implantation of Cardiac Pacemakers and Antiarrhythmia Devices).
Circulation. 2008;117:e350 e408.
17. Lang RM, Bierig M, Devereux RB, Flachskampf FA, Foster E, Pellikka
PA, Picard MH, Roman MJ, Seward J, Shanewise JS, Solomon SD,
Spencer KT, Sutton MS, Stewart WJ; Chamber Quantification Writing
Group; American Society of Echocardiographys Guidelines and
Standards Committee; European Association of Echocardiography. Recommendations for chamber quantification: a report from the American
Society of Echocardiographys Guidelines and Standards Committee and
the Chamber Quantification Writing Group, developed in conjunction
with the European Association of Echocardiography, a branch of the
European Society of Cardiology. J Am Soc Echocardiogr. 2005;18:
1440 1463.
18. Zareba W, Klein H, Cygankiewicz I, Hall WJ, McNitt S, Brown M,
Cannom D, Daubert JP, Eldar M, Gold MR, Goldberger JJ, Goldenberg
I, Lichstein E, Pitschner H, Rashtian M, Solomon S, Viskin S, Wang P,
Moss AJ; MADIT-CRT Investigators. Effectiveness of cardiac resynchronization therapy by QRS morphology in the Multicenter Automatic Defibrillator Implantation TrialCardiac Resynchronization Therapy
(MADIT-CRT). Circulation. 2011;123:10611072.
19. Calvert MJ, Freemantle N, Yao G, Cleland JGF, Billingham L, Daubert
JC, Bryan S; CARE-HF Investigators. Cost-effectiveness of cardiac

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

1536

20.

21.

22.

23.

24.

25.

26.

Circulation

October 4, 2011

resynchronization therapy: results from the CARE-HF trial. Eur Heart J.

2005;26:26812888.
Nichol G, Kaul P, Huszti E, Bridges JFP. Cost-effectiveness of cardiac
resynchronization therapy in patients with symptomatic heart failure. Ann
Intern Med. 2004;141:343351.
Sutton MG, Plappert T, Hilpisch KE, Abraham WT, Hayes DL, Chinchoy
E. Sustained reverse left ventricular structural remodeling with cardiac
resynchronization at one year is a function of etiology: quantitative
Doppler echocardiographic evidence from the Multicenter InSync Randomized Clinical Evaluation (MIRACLE). Circulation. 2006;113:
266 272.
Verhaert D, Grimm RA, Puntawangkoon C, Wolski K, De S, Wilkoff BL,
Starling RC, Tang WH, Thomas JD, Popovic ZB. Long-term reverse
remodeling with cardiac resynchronization therapy: results of extended
echocardiographic follow-up. J Am Coll Cardiol. 2010;55:1788 1795.
Lilli A, Ricciardi G, Porciani MC, Perini AP, Pieragnoli P, Musilli N, Colella
A, Del Pace S, Michelucci A, Turreni F, Sassara M, Achilli A, Barold SS,
Padeletti L. Cardiac resynchronization therapy: gender related differences in
left ventricular reverse remodeling. Pacing Clin Electrophysiol. 2007;30:
13491355.
St John Sutton MG, Plappert T, Abraham WT, Smith AL, DeLurgio DB,
Leon AR, Loh E, Kocovic DZ, Fisher WG, Ellestad M, Messenger J,
Kruger K, Hilpisch KE, Hill MR; Multicenter InSync Randomized
Clinical Evaluation (MIRACLE) Study Group. Effect of cardiac resynchronization therapy on left ventricular size and function in chronic heart
failure. Circulation. 2003;107:19851990.
Adelstein EC, Saba S. Usefulness of baseline electrocardiographic QRS
complex pattern to predict response to cardiac resynchronization. Am J
Cardiol. 2009;103:238 242.
Wokhlu A, Rea RF, Asirvatham SJ, Webster T, Brooke K, Hodge DO,
Wiste HJ, Dong Y, Hayes DL, Cha YM. Upgrade and de novo cardiac

resynchronization therapy: impact of paced or intrinsic QRS morphology

on outcomes and survival. Heart Rhythm. 2009;6:1439 1447.
26a.Singh JP, Klein HU, Huang DT, Reek S, Kuniss M, Quesada A,
Barsheshet A, Cannom D, Goldenberg I, McNitt S, Daubert JP, Zareba
W, Moss AJ. Left ventricular lead position and clinical outcome in the
multicenter automatic defibrillator implantation trial-cardiac resynchronization therapy (MADIT-CRT) trial. Circulation. 2011;123:1159 1166.
27. Mele D, Agricola E, Galderisi M, Rigo F, Citro R, Dal Monte A, Della
Valentina P, Calabrese A, Ferrari R; Study Group of Echocardiography,
Italian Society of Cardiology. Echocardiographic myocardial scar burden
predicts response to cardiac resynchronization therapy in ischemic heart
failure. J Am Soc Echocardiogr. 2009;22:702708.
28. Adelstein EC, Saba S. Scar burden by myocardial perfusion imaging
predicts echocardiographic response to cardiac resynchronization therapy
in ischemic cardiomyopathy. Am Heart J. 2007;153:105112.
29. Michael O. Sweeney, Sweeney MO, van Bommel RJ, Schalij MJ, Borleffs
CJ, Hellkamp AS, Bax JJ. Analysis of ventricular activation using surface
electrocardiography to predict left ventricular reverse volumetric remodeling
during cardiac resynchronization therapy. Circulation. 2010;121:626634.
30. Pepe MS, Janes H, Longton G, Leisenring W, Newcomb P. Limitations
of the odds ratio in gauging the performance of a diagnostic, prognostic,
or screening marker. Am J Epidemiol. 2004;159:882 890.
31. Beshai JF, Grimm RA, Nagueh SF, Baker JH II, Beau SL, Greenberg SM,
Pires LA, Tchou PJ; RethinQ Study Investigators. Cardiacresynchronization therapy in heart failure with narrow QRS complexes.
N Engl J Med. 2007;357:24612471.
32. Chung ES, Leon AR, Tavazzi L, Sun JP, Nihoyannopoulos P, Merlino J,
Abraham WT, Ghio S, Leclercq C, Bax JJ, Yu CM, Gorcsan J 3rd, St
John Sutton M, De Sutter J, Murillo J. Results of the Predictors of
Response to CRT (PROSPECT) trial. Circulation. 2008;117:2608 2616.

CLINICAL PERSPECTIVE
The Multicenter Automatic Defibrillator Implantation Trial With Cardiac Resynchronization Therapy (MADIT-CRT) trial
showed that cardiac resynchronization-defibrillator therapy is associated with a significant reduction in the risk of heart
failure or death compared with defibrillator-only therapy. Currently, however, there is limited information on the factors
that can be used to distinguish between responders and nonresponders in this population. In the present study, we developed
a response score that was based on 7 factors we identified as being associated with favorable reverse remodeling in
MADIT-CRT: female sex, a nonischemic origin of cardiomyopathy, left bundle-branch block, QRS 150 milliseconds,
prior hospitalization for heart failure, left ventricular end-diastolic volume 125 mL/m2, and left atrial volume 40
mL/m2. The score was then used to assess the clinical benefit of cardiac resynchronization-defibrillator therapy during the
trial. We show a direct correlation between an increasing response score and the magnitude of the reduction in the risk of
heart failure or death with cardiac resynchronization therapy. There was a significant 13% increase in the clinical benefit
of cardiac resynchronization-defibrillator therapy per 1-point increment in the response score and a significant direct
correlation between risk reduction associated with cardiac resynchronization-defibrillator therapy and response score
quartiles. Thus, our findings suggest that combined assessment of factors associated with reverse remodeling can be used
for improved selection of patients for cardiac resynchronization therapy.

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

SUPPLEMENTAL MATERIAL

Supplementary Appendix to: Predictors of Response to Cardiac

Resynchronization Therapy in MADIT-CRT
By Goldenberg I, Moss AJ, Hall WJ, et al.
CIRCULATIONAHA/2010/ 014324

The first part of this Appendix describes the methods and results of
the sensitivity and bootstrap validation analyses that were employed to
assess the robustness of the present results.
Appendix Table 1 presents the list of baseline covariates that were used
as candidates in the derivation models for the identification of the
echocardiographic response factors. Appendix Tables 2A and B present
changes in cardiac volumes in ICD-only patients by the individual response
factors (in parallel with Tables 1A and B in the main report). Appendix
Table 3 presents the clinical effect of CRT-D by the response score, with
diameters replacing volumes in the 2 echocardiographic predictors (in
parallel with Table 3A in the main report). Appendix Table 4 presents some
of the results of the sensitivity analyses.

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

Sensitivity and Bootstrap Validation Studies

There were several steps in the development of the echocardiographic response scores
and their usage in predicting clinical response in the two device groups:
1) Selection of response factors associated with 12-month changes in LVEDV and
LVESV from 718 CRT-D patients, together with the resulting fitted linear
regression equation, relating percentage reductions in LVEDV, and similarly in
LVESV, to the selected response factors.
2) Use of the relative sizes of the regression coefficients in these two models to code
the response factors as 1, 2 or 3, to enable, by summing, the creation of a simple
response score, ranging from 0 to 14.
3) Use of the scores to assess the relative effects of CRT-D and ICD-only on the risk
of the HF/death endpoint event, using all 1761 patients. First, the patient group
was partitioned into response-score quartiles and proportional hazards regression
carried out within each quartile. Second, the response scores were used as a
single linear predictor, along with treatment group and score x treatment
interaction, in a proportional-hazards regression. The regression coefficient for
interaction, and the corresponding hazard ratio, quantify the effect of the response
score in reducing risk of the endpoint for CRT-D patients (last rows of Tables 3A
and B).
Steps 1 and 2 were carried out in a subset of the CRT-D patient group while step 3 used
all CRT-D patients along with the ICD-only group. Several sensitivity and bootstrap
validation analyses were performed to assess the repeatability and reliability of the above
steps.

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

First, with regard to this last point that a subgroup was used both in development of the
response score and in assessing its clinical value, we note that the remaining CRT-D group
patients are of two basic types, leading us to partition the CRT-D group into three subgoups:
Subgroup (A) consists of the 718 patients (709 with complete data) used to derive the response
score; subgroup (B) consists of the 222 additional patients also with CRT-D usage at baseline
and 12 months but without the echocardiologic data needed for derivation of the response
score; and subgroup (C) consists of the remaining 124 CRT-D group patients not having
active CRT-D usage for the full 12 months, including some who died or dropped out.
Groups (A) and (B) are fairly similar but group (C) is conceptually different. Omission of
any of these groups in assessing clinical benefit of the response score may bias the overall
results, possibly more so than would the inclusion of the group on which the response score
was derived; inclusion of all three subgroups reflects the composition of the MADIT-CRT
population. We repeated assessment of the individual response score by fitting a
proportional hazards regression analysis including the response score, identification of four
treatment groups subgroups (A)-(C) and the ICD-only group and response score by
treatment group interactions. Results are presented in Appendix Table 4. There it is seen
that the hazard ratios, CRT-D:ICD-only, for the clinical effect of the response score in these
three subgroups are very similar. (P-values vary, in part due to differences in sample sizes.)
To further validate step (1), we carried out two analyses. In the first, we removed one
selected response factor at a time and replaced it with a potential contender, thereby creating 7
alternative sets of response factors. When replacing LVEDV by ejection fraction, prior heart
failure hospitalization by diuretic usage, LBBB by non-RBBB, or any of the response factors
with prior revascularization, similar patterns were found for the CRT-D vs. ICD only effect on

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

the risk of HF or death. However, the findings in these analyses were no longer statistically
significant (p-value for trend for CRT-D vs. ICD-only benefit across quartiles >0.05 in all
analyses).
Secondly, 10 bootstrap samples were selected (each a sample of size 718, with
replacement, from the 718 patients in subgroup (A)) and used to repeat the response
factor selection (1) (LVEDV only). Factors LAV and LBBB were selected every time,
ischemia was selected 9 of 10 times, and QRS 7 times; the other 3 were selected 3 to 5
times each. Other factors selected, with frequencies from 5 down to 1, were heart rate,
race, age, LVEF, creatinine, smoking history, no prior MI and history of hypertension.
Hence, the response factor selection step was moderately stable, suggesting that
additional factors may play a role in predicting echocardiographic response to CRT-D.
The regression models for each of the 10 bootstrap-specific sets of response
factors, when fit to all 1820 patients, yielded linear scores which correlated reasonably
well with the linear score from the regression using the original set, with correlation
coefficients averaging 0.82. Hence, even with other choices of factors, the resulting
regression fit did not differ greatly.
To validate (2), we re-fit a linear regression model for LVEDV changes on the
originally selected set of 7 response factors using each of the 10 bootstrap samples of
data, and then coded each of the 10 re-fits, just as described earlier, thereby creating 10
sets of integer-valued patient-specific scores. Correlation coefficients of these with the
original scores averaged 0.98, demonstrating great stability in the scoring once the
response factors were chosen.

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

To validate (3), we carried out 10 proportional-hazards regression analyses using

the scores derived from the 10 bootstrap samples. The resulting coefficient of variation
was very low, namely 0.54%. The variance of the interaction regression coefficients was
36% of the variance of the original interaction coefficient, suggesting relatively minor
unaccounted for additional variation.
These several evaluations lead us to conclude that the use of the seven selected
response factors in evaluating relative risk of HF/death is well-validated.

Reference: Efron B, Tibshirani RJ (1993). An Introduction to the Bootstrap. New York:

Chapman & Hall.

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

Supplemental Table 1. Prespecified candidate baseline clinical and echocardiographic

factors for the assessment of response to CRT-D therapy
FACTOR
Etiology of cardiomyopathy
Age, yrs
Diabetes
BUN, mg/dL
BMI, kg/m2
Race
Gender
Ejection fraction
Prior CABG
Prior CVA
Prior PCI
Prior atrial arrhythmia
Prior ventricular arrhythmia
Creatinine, mg/dL
Treated hyertension
Smoking
Heart rate, bpm
NYHA class (ischemic)
ECG morphology

CATEGORIZATION
Ischemic/Nonischemic
65/<65
Yes/No
>25/25
30/30
Caucasian/AA/Other
Male/female
<25%/25%
Yes/No
Yes/No
Yes/No
Yes/No
Yes/No
1.4/<1.4
Yes/No
Current/past/never
80/<80
I/II
LBBB/RBBB/IVCD (considered as 3
separate candidate covariates)
Prior MI
Yes/No
Prior hospitalization for heart failure
Yes/No
QRS duration, msec
150/<150
Baseline EF
<25%/25%
Baseline LVEDV*
Q1-3/Q4
Baseline LAV*
Q1/Q2-4
Apical position of LV lead
Yes/No
Baseline diuretic usage
Yes/No
Systolic blood pressure, mm Hg
110/>110
Diastolic blood pressure, mm Hg
<80/80
*In an alternative set of analyses, baseline left atrial volume was replaced with left atrial
diameter and left ventricular end diastolic volume was replaced with left ventricular
diastolic diameter.
BMI = body mass index; BUN = blood urea nitrogen; CABG = coronary artery bypass
surgery; EF = ejection fraction; NYHA = New York Heart Association class; LAV =
left atrial volume; LV = left ventricular; LVEDV = left ventricular end diastolic volume;
LVESV = left ventricular end systolic volume; PCI = percutaneous coronary
revascularization; SBP = systolic blood pressure

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

Supplemental Table 2. Echocardiographic response in the ICD-only group

A. Reduction in left ventricular end diastolic volume
HIGHER CRT-D RESPONSE

LOWER CRT-D RESPONSE

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

Definition

Percent
reduction
(SD)

Women

154

-7 (5)

25%

Men

469

-6 (6)

17%

-1 (1)

0.20

Non-isch.

289

-7 (5)

22%

Ischemic

334

-5 (5)

15%

-2 (0.5)

0.003

QRS (msec)

150

412

-6 (6)

18%

<150

211

-6 (6)

19%

-0.5 (0.6)

0.32

QRS pattern

LBBB

443

-6 (6)

19%

Non-BBB

179

-6 (6)

18%

-0.5 (0.6)

0.44

Yes

294

-6 (6)

17%

No

321

-6 (6)

21%

0.2 (0.6)

0.64

125ml/m2

465

-6 (6)

18%

<125ml/m2 154

-6 (6)

20%

-0.3 (0.6)

0.55

-6 (6)

18%

-0.03 (0.6)

0.96

Risk factor
(covariate)
Gender
CMP etiology

Prior HF hosp.
Baseline LVEDV
Baseline LAV

<40 ml/m

140

-6 (6)

>10%
reduction

Definition

21%

40 ml/m

Percent
reduction
(SD)

>10%
reduction

HIGH VS. LOW

Adjusted
pincrement
value
reduction
(SE)*

478

*Denotes the difference in percent reduction left ventricular end diastolic volume at 1-year between high- and low- responders
of each covariate; findings were obtained from a regression model for the end point of percent reduction in left ventricular
end diastolic volume, based on the seven risk factors listed.
CMP = cardiomyopathy; HF = heart failure; LAV = left atrial volume; LVEDV = left ventricular end diastolic volume.

B. Reduction in left ventricular end systolic volume

HIGHER CRT-D RESPONSE

LOWER CRT-D RESPONSE

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

Definition

Percent
reduction
(SD)

Women

154

-11 (9)

32%

Men

469

-10 (9)

23%

-1 (0.9)

0.17

Non-isch.

289

-11 (9)

29%

Ischemic

334

-9 (9)

23%

-2 (0.8)

0.009

QRS (msec)

150

412

-10 (9)

25%

<150

211

-11 (9)

27%

0.9 (0.9)

0.32

QRS pattern

LBBB

443

-10 (9)

26%

Non-LBBB

179

-10 (9)

26%

-0.7 (0.9)

0.46

Yes

294

-10 (9)

24%

No

321

-10 (9%)

27%

0.2 (0.9)

0.78

125ml/m2

465

-10 (9)

26%

<125 ml/m2

154

-10% (9)

26%

-0.3 (0.9)

0.61

478

-10 (9)

27%

-0.1 (0.9)

0.16

Risk factor
(covariate)
Gender
CMP Etiology

Prior HF hosp.
Baseline LVEDV
Baseline LAV

<40ml/m

140

-10 (9)

>10%
reduction

Definition

23%

40 ml/m

Percent
reduction
(SD)

>10%
reduction

HIGH VS. LOW

Adjusted
pincrement
value
reduction
(SE)*

*Denotes the difference in percent reduction left ventricular end systolic volume at 1-year between high- and low- responders of
each covariate; findings were obtained from a regression model for the end point of percent reduction in left ventricular end
diastolic volume, based on the seven risk factors listed.
CMP = cardiomyopathy; HF = heart failure; LAV = left atrial volume; LVEDV = left ventricular end diastolic volume.

Supplemental Table 3. Multivariate analysis: CRT-D vs. ICD-only risk of HF or death by

response group when left atrial volume and left ventricular end diastolic volume were replaced
by respective diameters in the response score*
RESPONSE GROUPS

CRT-D VS. ICD-ONLY RISK OF

HF OR DEATH

SCORE

HR

95%CI

P-value

P for trend

0 - 14

0.62

0.50 0.76

<0.001

NA

By response score
quartile

SCORE

HR

95%CI

P-value

Q1 (n = 383)

0-4

1.11

0.71 1.73

0.65

Q2 (n = 383)

5-6

0.62

0.41 0.92

0.02

Q3 (n = 494)

7-8

0.56

0.38 0.80

0.002

0.39

0.24 0.62

<0.001

All Patients (n=1237)

Q4 (n = 504)

9
(per unit
increment )

0.003

By individual
0.800.95
<0.001
0.86
response scores
*A favorable left atrial diameter response was categorized as 2 cm/ml, and a favorable left
ventricular diastolic diameter
response was categorized as 3 cm/ml; results obtained from Cox proportional hazards
regression models adjusted for
treatment arm and response group; treatment effect in each response group was obtained by
including treatment-by-response
group interaction terms in the multivariate models.

P-value for trend obtained by including an interaction-term between treatment arm and the
response group, assessed as a
single ordinal categorical variable in the multivariate model.

Hazard ratio decrease, per unit increase in response score, based upon the interaction between
treatment group and score.

9
Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

Supplemental Table 4. Sensitivity analysis: Individual response score effects in CRT-D

subgroups*
End point: first occurrence of heart failure or death**
n

HR

95%CI

P-value

706

1.01

0.96 1.06

0.69

1055

0.88

0.84 0.93

<0.001

Subgroup A

709

0.88

0.81 0.94

<0.001

Subgroup B

222

0.92

0.83 1.02

0.12

Subgroup C

124

0.84

0.76 0.94

0.002

Main effect in ICD-only patients

CRT-D relative to ICD-only
All CRT-D patients

*CRT-D subgroups:
Group A: Patients with baseline and 12-month ECHO studies while on CRT-D
therapy, the group from which the response score was developed..
Group B: Other patients with baseline and 12-month CRT-D usage but without
needed echocardiographic studies.
Group C: Remaining CRT-D patients without baseline and 12-month ECHO
studies, including those dying or dropping out within 12 months and those who
never had a CRT-D implanted.
**Results obtained from Cox proportional hazards regression models of time to first
heart failure or death. Models included individual response score (0-14), treatment
groups and interaction between response score and treatment groups.

Hazard ratio (HR) reflects decrease in risk per unit increase in the individual response
score.

Hazard ratios (HR) reflect decrease in risk per unit increase in the individual response
score in CRT-D group relative to risk in the ICD-only group. There are no significant
differences among these treatment-effect HRs. The main effect in a CRT-D group is
this HR times the ICD-only HR.

Downloaded from http://circ.ahajournals.org/ by guest on July 6, 2013

10