Incremental prognostic value of left ventricular function analysis over non-invasive coronary angiography with multidetector computed tomography.

Abstract

The purpose of this study was to determine the prognostic value of computed tomography coronary angiography (CTA)-derived left ventricular (LV) function analysis and to assess its incremental prognostic value over the detection of significant stenosis using CTA.
In 728 patients (400 males, mean age 55 ± 12 years) with known or suspected CAD, the presence of significant stenosis (≥ 50% stenosis) and LV function were assessed using CTA. LV end-systolic volume (LVESV), LV end-diastolic volume (LVEDV), and LV ejection fraction (LVEF) were calculated. LV function was assessed as a continuous variable and using cutoff values (LVEDV > 215 mL, LVESV > 90 mL, LVEF < 49%). The following events were combined in a composite end-point: all-cause mortality, non-fatal myocardial infarction, and unstable angina pectoris requiring hospitalization.
On CTA, a significant stenosis was observed in 221 patients (30%). During follow-up [median 765 days, 25-75th percentile: 493-978] an event occurred in 45 patients (6.2%). After multivariate correction for clinical risk factors and CTA, LVEF < 49% and LVESV > 90 mL were independent predictors of events with an incremental prognostic value over clinical risk factors and CTA.
The present results suggest that LV function analysis provides independent and incremental prognostic information beyond anatomic assessment of CAD using CTA.

Full text

Page 1

ORIGINAL ARTICLE
Incremental prognostic value of left ventricular
function analysis over non-invasive coronary
angiography with multidetector computed
tomography
Fleur R. de Graaf, MD,aJacob M. van Werkhoven, MSc,a,bJoe ¨lla E. van Velzen,
MD,a,bM. Louisa Antoni, MD,aMark J. Boogers, MD,a,bLucia J. Kroft, MD, PhD,c
Albert de Roos, MD, PhD,cMartin J. Schalij, MD, PhD,aJ. Wouter Jukema, MD,
PhD,a,bErnst E. van der Wall, MD, PhD,a,bJoanne D. Schuijf, PhD,a
and Jeroen J. Bax, MD, PhDa
Background. The purpose of this study was to determine the prognostic value of computed
tomography coronary angiography (CTA)-derived left ventricular (LV) function analysis and
to assess its incremental prognostic value over the detection of significant stenosis using CTA.
Methods. In 728 patients (400 males, mean age 55 ± 12 years) with known or suspected
CAD, the presence of significant stenosis (‡ 50% stenosis) and LV function were assessed using
CTA. LV end-systolic volume (LVESV), LV end-diastolic volume (LVEDV), and LV ejection
fraction (LVEF) were calculated. LV function was assessed as a continuous variable and using
cutoff values (LVEDV > 215 mL, LVESV > 90 mL, LVEF < 49%). The following events were
combined in a composite end-point: all-cause mortality, non-fatal myocardial infarction, and
unstable angina pectoris requiring hospitalization.
Results. On CTA, a significant stenosis was observed in 221 patients (30%). During follow-
up [median 765 days, 25-75th percentile: 493-978] an event occurred in 45 patients (6.2%).
After multivariate correction for clinical risk factors and CTA, LVEF < 49% and
LVESV > 90 mL were independent predictors of events with an incremental prognostic value
over clinical risk factors and CTA.
Conclusions. The present results suggest that LV function analysis provides indepen-
dent and incremental prognostic information beyond anatomic assessment of CAD using
CTA. (J Nucl Cardiol 2010;17:1034–40.)
Key Words: Left ventricular function Æ Æ computed tomography (CT) Æ Æ prognosis
INTRODUCTION
Multidetector
angiography (CTA) has emerged as an important non-
computedtomography coronary
invasive imaging modality by providing direct anatomic
assessmentofcoronary
Recently, several studies have shown that, in addition to
its value for the diagnosis of CAD, stenosis detection
with CTA may also be useful for risk stratification.4-7
Furthermore, besides the assessment of coronary anat-
omy, left ventricular (LV) function may be evaluated
using information derived from the same CTA data-
set.8-10LV function is an established prognostic marker,
as has been demonstrated using several imaging modal-
ities including left ventriculography, echocardiography,
magnetic resonance imaging (MRI), and single-photon
emission computed tomography (SPECT).11-14However,
no data are currently available concerning the prognostic
significance of CTA-derived LV function assessment.
The purpose of this study was therefore to determine the
arterydisease(CAD).1-3
From the Department of Cardiology,aLeiden University Medical
Center, Leiden, The Netherlands; The Interuniversity Cardiology
Institute of the Netherlands,bUtrecht, The Netherlands; Department
of Radiology,cLeiden University Medical Center, Leiden, The
Netherlands.
Received for publication Mar 2, 2010; final revision accepted Jul 18,
2010.
Reprint requests: Jeroen J. Bax, MD, PhD, Department of Cardiology,
Leiden University Medical Center, Albinusdreef 2, 2333 ZA Leiden,
The Netherlands; j.j.bax@lumc.nl.
1071-3581/$34.00
Copyright ? 2010 The Author(s). This article is published with open
access at Springerlink.com
doi:10.1007/s12350-010-9277-4
1034

Page 2

prognostic value of CTA-derived LV function analysis
and to assess its incremental prognostic value over the
detection of significant coronary artery stenosis using
CTA.
METHODS
The study group consisted of consecutive patients with
suspected or known CAD who were clinically referred for
CTA because of chest pain, a positive or inconclusive exercise
electrocardiogram (ECG) test, or an elevated risk profile for
cardiovascular disease. The study population is part of a large
ongoing registry exploring the prognostic value of CTA.15For
the current analysis, only patients with a CTA examination of
diagnostic image quality and with additional LV function data
were included. Exclusion criteria for CTA examination were
(1) (supra)ventricular arrhythmias, (2) renal failure (glomerular
filtration rate\30 mL/minutes), (3) known allergy to iodine
contrast material, (4) severe claustrophobia, or (5) pregnancy.
CTA Data Acquisition
CTA studies were performed using a 64-row (N = 647)
or 320-row (N = 81) multidetector scanner (Aquilion 64, and
Aquilion ONE, respectively, Toshiba Medical Systems,
Otawara, Japan). One hour prior to the examination, a single
dose of oral beta-blocker medication was administered to
patients with a heart rate C 65 bpm, unless contraindicated.
Based on a pre-defined beta-blocking medication administra-
tion protocol, patients with a heart rate between 65 and 75 bpm
received 50 mg metoprolol, whereas patients with a heart
rate C 75 bpm received 100 mg metoprolol. The total amount
of non-ionic contrast media (Iomeron 400; Bracco, Milan,
Italy) injected into the antecubital vein was 60-100 mL
(depending on body weight and scanner type) at a flow rate of
5.0 or 6.0 mL/second, followed by a saline flush of 25-50 mL.
In order to synchronize the arrival of the contrast media, bolus
arrival was detected using a real-time bolus tracking tech-
nique.16All images were acquired during a single inspiratory
breath-hold of maximally 12 seconds. For 64-row CTA, a
helical-scanning technique was used as previously descri-
bed.17,18In brief, during the scan, the ECG was registered
simultaneously for retrospective gating of the data. A colli-
mation of 64 9 .5 mm was used. Additional scan parameters
were 400 or 500 ms gantry rotation time depending on cardiac
frequency, 120 kV tube voltage, and 300-350 mA (depending
on body mass index, (BMI) and thoracic geometry). During
320-row CTA, the ECG was registered simultaneously for
prospective triggering of the data. In order to perform LV
function analysis, ECG-triggered tube modulation was used, as
previously described.19
The entire heart was imaged in a single heart beat,
attaining maximal tube current during 75% of R–R interval (in
patients with stable heart rate\60 bpm), during 65-85% of
R–R interval (in patients with a heart rate 60-65 bpm), or
during 30-80% of R–R interval (in patients with a heart
rate[65 bpm). Outside the pre-defined interval, tube current
was 25% of the maximal tube current. In patients with a heart
rate[65 bpm images were acquired during multiple heart
beats (typically two). A collimation of 320 9 .5 mm was
used. Additional scan parameters were 350 ms gantry rotation
time, 120-135 kV tube voltage, and 400-580 mA (depending
on BMI and thoracic geometry). For the 64-row CTA scanner,
the estimated average radiation dose for CTA was 17.6 ±
5.6 mSv. For the 320-row CTA scanner, the estimated average
radiation dose for single heart beat CTA’s using ECG-trig-
gered tube modulation was 10.7 ± 3.6 mSv. In patients,
in whom CTA image acquisition was performed during mul-
tiple heart beats, average estimated radiation dose was
16.7 ± 6.3 mSv.
Coronary arteries were evaluated using the reconstruction
dataset with the least motion artifacts, typically acquired dur-
ing an end-diastolic phase. To assess LV function and LV
volumes, ten series of 2.0-mm slices were reconstructed from
the same dataset at every 10% throughout the cardiac cycle,
starting at early systole (0% of cardiac cycle) to end-diastole
(90% of cardiac cycle).
Data Analysis
CTA and LV function reconstructions were transferred
to a remote workstation with dedicated analysis software (for
64-row CTA reconstructions: Vitrea 2, Vital Images, Minne-
tonka, MN, USA; for 320-row CTA reconstructions: Vitrea FX
1.0, Vital Images). The presence of coronary stenosis was
assessed by scrolling through axial images, combined with
visual assessment of curved multiplanar reconstructions in C 2
orthogonal planes. CTA examinations were evaluated on a
patient basis for the presence of a significant stenosis (C 50%
luminal narrowing) by two experienced observers. Discrep-
ancies in interpretation were resolved by consensus.
For the purpose of LV function analysis, appropriate
phases for LV end-systolic volume (LVESV) and LV end-
diastolic volume (LVEDV) were acquired, by selecting the
smallest and largest cross-sectional LV cavity areas, respec-
tively. Upper limit of the LV was determined at the basal level
of the mitral valve and the start of the LV outflow tract, as
previously described.20Endocardial borders were outlined
using a semi-automated method from the base to the apex on
the short-axis cine images by an independent observer. Papil-
lary muscles were excluded from the ventricular cavity. The
LVEDV and LVESV volumes were calculated and the LVEF
was derived by subtracting the LVESV from the LVEDV and
dividing the result by the LVEDV.9Observers for LV function
analysis were blinded to CTA data.
End-Points
Patient follow-up data were gathered by a single observer,
blinded to the baseline CTA results, using clinical visits and/or
standardized telephone interviews. A composite end-point was
constructed using the following events: all-cause mortality,
non-fatal myocardial infarction, and unstable angina pectoris
requiring hospitalization. Non-fatal infarction was defined
based on criteria of typical chest pain, elevated cardiac enzyme
Journal of Nuclear Cardiology
Volume 17, Number 6;1034–40
de Graaf et al 1035
Incremental prognostic value of LV function

Page 3

levels, and typical changes on the ECG.21Unstable angina was
defined according to the European Society of Cardiology
guidelines as acute chest pain with or without the presence of
ECG abnormalities, and negative cardiac enzyme levels.22
Statistical Analysis
Statistical analysis was performed using SPSS software
(version 16.0, SPSS Inc., Chicago, IL, USA). Quantitative data
were expressed as mean ± SD. Categorical variables were
described as numbers and percentages. Cox regression analysis
was used to determine the prognostic value of the presence of
significant stenosis and LV function parameters on CTA. First,
univariate analysis of baseline characteristics, the presence of
significant stenosis, and LV function parameters was performed
using a composite end-point of all-cause mortality, non-fatal
infarction, and unstable angina. For each variable, a hazard
ratio (HR) with a 95% confidence interval (CI) was calcu-
lated. Increased LVEDV and LVESV were defined as
LVEDV[215 mL, LVESV[90 mL and reduced LVEF was
defined as LVEF\49%, as described previously.23These
values were determined using the 95% CI boundaries for 3D
LV dimensions and function as determined by CTA.23After
univariate analysis, several multivariate models were created to
correct LV function parameters for clinical risk factors and the
presence of significant stenosis on CTA. In case variables
showed strong interrelation (Pearson’s correlation coeffi-
cient[.8), variables were excluded from the same multivariate
model. Finally, the incremental value of LV function variables
over clinical risk factors (age, gender, smoking, and known
CAD) and the presence of significant stenosis on CTA were
assessed by calculating the global v2values. A P-value\.05
was considered statistically significant.
RESULTS
From the registry, additional LV function mea-
surements (derived from CTA) were available in 813
patients. In 29 patients (3.6%) the CTA examination was
of non-diagnostic image quality and these patients were
excluded from the study. Furthermore, a total of 56
patients (6.9%) were lost to follow-up. As a result, the
patient group consisted of 728 individuals. The main
clinical characteristics of the population are listed in
Table 1. In summary, 55% were male and the mean age
was 55 ± 12 years.
Baseline CTA and LV Function
In the study population of 728 patients, a significant
stenosis (luminal narrowing C 50%) was identified
on CTA in 221 patients (30%). When evaluating LV
volumes, an average LVEDV of 138 ± 38 mL was
observed. For LVESV an average value of 52 ± 29 mL
was observed. As a result, average LVEF was 64 ± 10%.
Increased LVEDV ([215 mL) and LVESV ([90 mL)
were present in 31 patients (4.3%) and 50 patients
(6.9%), respectively. A reduced LVEF (\49%) was
present in 43 patients (7.3%).
Follow-up
Themedian follow-upperiodwas765 days
(25-75th percentile: 493-978); an event occurred in 45
patients (6.2%). All-cause death was reported in 23
patients (3.2%), non-fatal myocardial infarction occur-
red in 7 patients (1.0%), and 15 patients (2.0%) were
admitted to the hospital due to unstable angina.
Survival Analysis
Univariate analysis of clinical risk factors, signifi-
cant stenosis on CTA, and LV function analysis
parameters is depicted in Table 2. Of the clinical risk
factors, age, smoking, and known CAD were significant
predictors of events. Also, the presence of a significant
Table 1. Patient characteristics
Total:
N 5 728
Gender (male/female)
Age (years)
Reason for referral
Typical chest pain
Atypical chest pain
Non-anginal chest pain
Elevated risk profile
Positive or inconclusive exercise ECG
Clinical risk factors
Diabetes
Hypercholesterolemia*
Hypertension?
Family history of coronary artery
disease?
Smoking
Obesity (BMI C 30 kg/m2)
Known CAD
400/328
55 ± 12
83 (12%)
189 (26%)
132 (18%)
213 (29%)
111 (15%)
222 (31%)
261 (36%)
317 (44%)
309 (42%)
137 (19%)
143 (20%)
96 (13%)
*Serum total cholesterol C 230 mg/dL and/or serum tri-
glycerides C 200 mg/dL or treatment with lipid lowering
drugs.
?Defined as systolic blood pressure C 140 mmHg and/or
diastolic blood pressure C 90 mmHg and/or the use of
antihypertensive medication.
?Defined as presence of coronary artery disease in first
degree family members at\55 years in men and\65 years
in women.
BMI, Body mass index; CAD, coronary artery disease.
1036 de Graaf et al
Incremental prognostic value of LV function
Journal of Nuclear Cardiology
November/December 2010

Page 4

stenosis on CTA was identified as a significant predictor
of events. Regarding LV function, all three parameters
(LVEDV, LVESV, and LVEF) were significant univar-
iate predictors when assessed as continuous variables
and categorical variables, using predefined cutoff values.
Multivariate models were created correcting for
clinical risk factors (age, gender, smoking, and known
CAD) and significant stenosis on CTA. The prognostic
values of LV volumes and LVEF are shown in Table 3.
A total of six models were created. In the first three
models, LVEDV, LVESV and LVEF were assessed as
continuous variables, whereas in the last three models
the predictive value of LV function was determined
using the pre-defined cutoff values to indicate reduced
LV function. Although none of the continuous LV
function variables remained independent predictors of
events, using the pre-defined cutoff values as markers
for reduced LV function, increased LVESV and reduced
LVEF provided additional prognostic information over
clinical risk factors and significant stenosis on CTA.
Figure 1 illustrates the incremental prognostic value,
depicted by v2value, of LVESV[90 mL over age,
gender, smoking, known CAD, and significant stenosis
on CTA (P\.05). Similarly, Figure 2 shows that
LVEF\49% has incremental value (P\.05), and thus
enhanced risk stratification beyond the detection of
significant stenosis using CTA.
DISCUSSION
The main finding of this study is that CTA-derived
LV function may provide important prognostic infor-
mation beyond the detection of significant stenosis on
CTA.
LV volumes and LVEF are used extensively as
clinical markers of cardiac function. In patients with
known CAD, LV function analysis provides important
information for risk assessment and further manage-
ment.11,24Post-myocardial infarction patients with a
normal LVEF have a relatively good prognosis, whereas
the risk of future events has been shown to increase with
deteriorating LVEF.25LV function analysis may also
be useful in patients with suspected CAD. Using 2D
Table 2. Univariate predictors of events
HR
(95% CI)P-value
Baseline clinical characteristic
Age
Male gender
Diabetes
Hypercholesterolemia*
Hypertension?
Family history of
coronary artery
disease?
Smoking
Obesity
(BMI C 30 kg/m2)
Known CAD
CTA
Significant stenosis
LV function (continuous)
LVEDV (per
10 mL increase)
LVESV (per
10 mL increase)
LVEF (per % increase)
LV function (categorical)
LVEDV[215 mL
LVESV[90 mL
LVEF\49%
1.04 (1.01–1.07)
1.36 (.75–2.49)
1.03 (.55–1.93)
.73 (.38–1.39)
.89 (.53–1.73)
.61 (.32–1.15)
.005
.315
.939
.333
.958
.125
2.31 (1.25–4.25)
.55 (.22–1.39)
.007
.205
2.74 (1.46–5.13).002
3.52 (1.59–7.83).002
1.07 (1.00–1.14).048
1.10 (1.03–1.18).006
.97 (.94–.99) .006
2.89 (1.22–6.86)
3.95 (1.99–7.84)
3.82 (1.93–7.57)
.016
\.001
\.001
*Serum total cholesterol C 230 mg/dL and/or serum tri-
glycerides C 200 mg/dL or treatment with lipid lowering
drugs.
?Defined as systolic blood pressure C 140 mmHg and/or
diastolic blood pressure C 90 mmHg and/or the use of
antihypertensive medication.
?Defined as presence of coronary artery disease in first
degree family members at\55 years in men and\65 years
in women.
BMI, Body mass index; CAD, coronary artery disease; CI,
confidence interval; CTA, computed tomography coronary
angiography; HR, hazard ratio; LVEDV, left ventricular end-
diastolic volume; LVEF, left ventricular ejection fraction;
LVESV, left ventricular end-systolic volume.
Table 3. Multivariate models for the prediction
of events
HR (95% CI)P-value
LV function (continuous)
LVEDV* (per 10 ml
increase)
LVESV* (per 10 ml
increase)
LVEF* (per % increase)
LV function (categorical)
LVEDV[215 mL*
LVESV[90 mL*
LVEF\49%*
1.05 (.99–1.13) .123
1.06 (.98–1.15).119
.98 (.96–1.01).177
1.98 (.79–4.95)
3.11 (1.45–6.67)
2.61 (1.22–5.60)
.143
.004
.014
*Corrected for baseline clinical variables (age, gender,
smoking, known CAD) and significant stenosis on CTA.
CAD, Coronary artery disease; CI, confidence interval; HR,
hazard ratio; LVEDV, left ventricular end-diastolic volume;
LVEF, left ventricular ejection fraction; LVESV, left ventricular
end-systolic volume.
Journal of Nuclear Cardiology
Volume 17, Number 6;1034–40
de Graaf et al1037
Incremental prognostic value of LV function

Page 5

echocardiography in 2,948 patients without prevalent
CAD, Devereux et al12showed that LVEF\40% was a
strong predictor of cardiovascular death (RR 6.9, CI
3.0-15.9, P\.001) and all-cause mortality (RR 4.8, CI
2.8-8.1, P\.001), independent of clinical risk fac-
tors. In a study by Sharir et al,14evaluating 1,680
patients referred for gated SPECT, LVEF\45% and
ESV[70 mL were identified as optimal thresholds to
accurately stratify patients into high-risk and low-risk
groups. Importantly, patients with a normal LVEF or
LVESV had a substantially lower annual mortality rate
(regardless of the degree of perfusion abnormalities on
SPECT) than patients with reduced LVEF or increased
LVESV.Inthisstudy,theprognosticvalueofLVfunction
measurements on CTA was evaluated. LV volumes and
LVEF provided prognostic information independent of
baseline clinical variables. These findings illustrate that
measurements of LV function (and volumes) from CTA
acquisitionscanbebeneficialindefiningpatientriskandif
available should be incorporated into the clinical report.
Importantly, patients undergoing CTA are primarily
referred for assessment of CAD. As this anatomic infor-
mation has been established to provide prognostic
information,4-6animportantobjectiveofthisstudywasto
determine the additional value of LV function analysis
beyond the detection of significant CAD. Limited infor-
mation is available regarding the additional prognostic
value of LV function over the non-invasive angiographic
assessment of CAD using CTA. In a recent study by Min
et al,26evaluating 5,330 consecutive patients without
known CAD with a meanfollow-upof 2.3 ± .6 years, the
addition of LV function measured by CTA significantly
increased risk correlation for death. Annualizedmortality
rates in patients with significant CAD and LVEF B 50%
were significantly higher (3.79%) than in patients with
significant CAD and LVEF[50% (1.76%). Further-
more, Chow et al27recently determined the incremental
prognostic value of LV function measured with CTA in
2,076 consecutive patients with a mean follow-up of
16 ± 8 months. It was shown that LVEF had incremental
prognostic value over CAD severity (HR 1.47, 95% CI
1.17-1.86). These data indicate that LV function is an
important predictor of survival with incremental value to
CTA. In line with these results, this study suggests that,
although the assessment of significant CAD on CTA is a
powerful predictor, LV function parameters provide
independent incremental value in predicting adverse
events.Thus,theadditionofLVfunctionanalysistoCTA
may further improve risk stratification of patients with
known of suspected CAD referred for stenosis detection.
Limitations
Several limitations of this study merit consideration.
First, 6.9% of patients were lost to follow-up. Second,
regional wall motion abnormalities were not assessed in
this study. Third, this study is limited by a relatively
small patient number. Fourth, CTA is inherently asso-
ciated with radiation exposure.28As a result, prospective
ECG triggering was recently introduced for the purpose
of radiation reduction.29,30This technique allows image
Figure 1. Bar graph illustrating the incremental prognostic
value (depicted by v2values on the y-axis) of LVESV[90 mL
over age, gender, smoking, known CAD, and significant
stenosis. The presence of significant stenosis on CTA has a
significant incremental prognostic value over age, gender,
smoking, and known CAD (asterisk). A further incremental
prognostic value over clinical risk factors and significant
stenosis on CTA is observed with the addition of LVESV[
90 mL (#). CAD, Coronary artery disease; LVESV, left
ventricular end-systolic volume.
Figure 2. Bar graph illustrating the incremental prognostic
value (depicted by v2values on the y-axis) of LVEF\49%
over age, gender, smoking, known CAD, and significant
stenosis on CTA. The presence of significant stenosis on CTA
has a significant incremental prognostic value over age,
gender, smoking, and known CAD (asterisk). A further
incremental prognostic value over clinical risk factors and
significant stenosis on CTA is observed with the addition of
LVEF\49% (#). CAD, Coronary artery disease; LVEF, left
ventricular ejection fraction.
1038de Graaf et al
Incremental prognostic value of LV function
Journal of Nuclear Cardiology
November/December 2010