resting electrocardiogram and stress myocardial perfusion ... · medicine,asdasdasd 1. department...

Original Article

www.nuclmed.gr Hellenic Journal of Nuclear Medicine May - August 2010 118

Resting electrocardiogram and stress myocardial imaging in the determination of left ventricular s

perfusion ystolic function:

an assessment enhancing the performance of gated SPET

tratios Moralidis 1, MD

, MD,

3 D,

, MD, PhD,

oulos

, PhD

ent of Nuclear asd

ar sdasdasdasd

f Nuclear ,asdasdasdasd

Department of Nuclear Medicine,asdasdasdasd

Department of Nuclear

d estimate resting left ollected during routine ry heart disease. This emission tomography rdial perfusion imaging A) were separated into tic regression analysis

al, electrocardiographic an ERNA-LVEF<0.50. RNA-LVEF from those

201Tl gated SPET. Our m was the best simple models including ECG, vation and validation for the assessment of

quilibrium radionuclide G-Scintigraphic model

.711, P=0.000) and the 00). The Bland-Altman 76, 0.0710.196 and

The average LVEF was a better discriminator of receiver operating characteristic

(ROC) analysis and identified more patients (89%) with a 10% difference from VEF than gated SPET (65%, P=0.000). In conclusion, resting left lar systolic dysfunction can be determined effectively from simple resting

s. This model provides ted SPET, and can

: 22-6-2010

tion, with or without ent can improve the

t disease (CHD) is by ystolic function

ectrocardiographic

to virtually exclude the 12-lead ECG is line investigation in

owever, the clinical hereas clinical and

ictive information in etting [15-17]. More derived from gated yocardial perfusion

een validated successfully against other accepted techniques [18-20]. However, various limitations with this type of assessment also have been recognized [21-25], while the clinical relevance of non-gated myocardial perfusion images in predicting the level of left ventricular function has not been determined.

Based on earlier observations and the pathophysiology that underlies depression of left ventricular systolic function, we hypothesized that it is possible to predict reliably resting systolic function impairment in patients with suspected or known CHD from simple and readily available demographic, clinical, resting ECG and stress myocardial perfusion imaging data. Moreover, an estimation of resting LVEF would be feasible with a high degree of certainty. There-

,

Abstract This study aimed to determine systolic dysfunction anventricular ejection fraction (LVEF) from information cevaluation of patients with suspected or known coronaapproach was then compared to gated single photon(SPET). Patients having undergone stress 201Tl myocafollowed by equilibrium radionuclide angiography (ERNderivation (n=954) and validation (n=309) groups. Logiswas used to develop scoring systems, containing clinic(ECG) and scintigraphic data, for the discrimination of Linear regression analysis provided equations predicting Escores. In 373 patients LVEF was also assessed with results showed that an ECG-Scinigraphic scoring systepredictor of an ERNA-LVEF<0.50 in comparison to otherclinical and scintigraphic variables in both the derisubpopulations. A simple linear equation was derived alsoresting LVEF from the ECG-Scintigraphic model. Eangiography-LVEF had a good correlation with the ECLVEF (r=0.716, P=0.000), 201Tl gated SPET LVEF (r=0average LVEF from those assessments (r=0.796, P=0.0statistic (mean2SD) provided values of 0.0010.10.0400.152, respectively.

PhD,

Tryfon Spyridonidis 2,

Georgios Arsos , MD, Ph

Vassilios Skeberis 4

Constantinos Anagnostop5, MD, PhD,

Stavros Gavrielidis 6 , MD

1. Department of Nuclear asd Medicine,asdasdasd

1. DepartmMedicine,asdasdasd

1. Department of NucleMedicine,a

1. Department of Nuclear Medicine,asdasdasdasd

1. Department of Nuclear Medicine,asdasdasdasd

1. Department oMedicine

1.

systolic dysfunction than gated SPET-LVEF in

ERNA-LventricuECG and stress myocardial perfusion imaging variablereliable LVEF estimations, comparable to those from 201Tl gaenhance the clinical performance of the latter.

Hell J Nucl Med 2010; 13(2):118-126 Published on line

Introduction Early detection of left ventricular systolic dysfuncsymptoms, is of great importance as medical treatmoutcome and the quality of life [1-3]. Coronary hearfar the most common cause of left ventricular simpairment [4] and is frequently accompanied by el(ECG) abnormalities

Previous studies supported a normal ECG resting left ventricular systolic dysfunction and advocated by international guidelines as first patients suspected of having the disorder [5-10]. Hvalue of a resting ECG has been disputed [11-14], wradiographic criteria were found to offer limited predsituations other than the acute or primary care srecently, left ventricular ejection fraction (LVEF) single photon emission tomography (SPET) mscanning, with either 201Tl or 99mTc compounds, has b

hD, 2, D,

D, D, ulos 5

PhD

r Medicine,niversityi,

edicine, as,

dicine

ternal ital,

ical School,

lear Medicine,ool of

University om

diology, AHEPersity Medica

ece

phy sion imaging

SPET

pondence address: ant Professor

Department of Nuclear Medicine AHEPA Hospital Aristotle University Medical School 1 Stilponos Kyriakidi Str 54124 Thessaloniki, Greece Tel. +30 2310994688 email: [email protected] Received: 23 March 2010 Accepted: 25 May 2010

Efstratios Moralidis 1 MD, PTryfon Spyridonidis MGeorgios Arsos 3, MD PhVassilios Skeberis 4 MD PhConstantinos AnagnostopoMD, PhD, Stav

ros Gavrielidis 6 MD, 1. De ment of NucleapartAHEPA Hospital, Aristotle UMedical School, ThessalonikGreece 2. Department of Nuclear M Rio University Hospital, PatrGreece 3. Department of Nuclear Meand 4. Cardiology Unit, Second Propedeutic Department of In Medicine, Hippokration HospAristotle University Med

Thessaloniki, Greece 5. Department of Nuc

on SchBarts and the LondMedicine and Dentistry,

, United Kingdof London6. Department of Car A

l Hospital, Aristotle UnivSchool, Thessaloniki, Gre *** Keywords:

ction-Left ventricular fun-Electrocardiogra-Myocardial perfu-Gated CorresE. Moralidis, Assist

Original Article


fore, the aim of this study was twofold: a) to ithe above hypothesis and develop an optimasystem for the assessment of left ventriculafunction, and b) to determine the clinical valuapproach in relation to gated SPET, using eq

nvestl scr sye ouilib

radionuclide angiography (ERNA) as the refer

ethods

ectne Sd, t

ons yoca

sessme ef

aluaoses

h a higheprobability of CHD [26]

faf t

eek, al fotion

byts n g firsdetertion ere the

patidy pe

andavaiograedic

d by thress. Exclusion criteri

re a history of an acute myocaeous coronary intervention in

las

myoputter. e pocoro

sympt

Electrocardiography

All patients had a standard resting 12-lead ECG obtained before stress testing. Electrocardiograms were interpreted prospectively by the physician performing the stress without knowledge of imaging data. The observer described ECG findings and categorized them as normal or abnormal according to his best judgment, in a way similar to that employed in routine clinical practice. Patients having borderline abnormalities (minor ST-segment and T-wave changes, left ventricular hypertrophy and so forth) were placed in the abnormal

ecutive patients was y by two physicians (E.M. and

.) to assess inter-observer agreement.

dynamic exercise, ng and an activity of body weight, was cessation of stress ctive medication was

f treating physicians, but it was for at least 12h and stain from caffeine

containing beverages and smoking for 24h.

rocessing and

s performed with a w γ-camera (APEX-sing an acquisition

previously in detail 4h later to assess of patients with no

minute, delayed 201Tl he ECG signal for the mmercially available

yocardium was cumulation in each

t scale, and the sum t scores (SRS) were least two contiguous

ely reduced or absent sidered myocardial was assessed semi aw data and the Two different 3-point elative size of the

al; 2, equivocal; 3, lative width of the acent normal wall (1,

equal size; 3, cavity wider than the wall). When both criteria were scored with 3, dilatation

otherwise as absent. ndently assessed by , T.S.), unaware of were resolved by was performed ft ventricular ejection ardial imaging was observers, and the

.

um radionuclide angiography

Resting ERNA was performed immediately after the completion of myocardial perfusion imaging with 740MBq 99mTc labeled red blood cells, in the best septal view, using the same imaging system and a previously described methodology [23]. Left ventricular ejection fraction was calculated in duplicate, separately by two observers (E.M., T.S.), in the standard manner; the mean value of these measurements was incorporated in patients’ report and entered in analysis. The lower limit of normality in our site is 0.50.

Statistical analysis

igate oring

last month, coronary artery bypass grafting in the months, significant valve disease, congenital disease, paced rhythm, hypertrophic cardiodigitalis therapy and atrial fibrillation or flduration of disease was estimated from the timthe index event in the following order: a) intervention, b) acute coronary syndrome, c) d) first presentation to medical office.

category. The ECG from 100 consinterpreted independentl

stolic f this

V.S

rium ence

Stress testing

Participants were submitted to dipyridamole or dobutamine stressi90-130MBq 201Tl, depending onintravenously administered prior to[23]. The discontinuation of cardio-aleft at the discretion o

standard.

Patients and m

Patient’s recruitment From our database, among all patients with suspknown coronary artery disease referred for routimyocardial perfusion imaging over a 2-year periohaving undergone the following examinatiidentified: a stress-redistribution 201Tl mperfusion study followed by a resting ERNA asDuring the period of collection of those data wperforming this combination of studies in patients rfor risk assessment and myocardial viability evAmong patients referred for diagnostic purpventricular function was assessed in those witperceived clinical

ed or PET

ensured that nitrates were withheld all patients were instructed to ab

hose were rdial ent.

were erred tion.

, left

201Tl SPET acquisition, panalysis

Post-stress SPET acquisition wasingle-headed, large field of vieSPX4, Elscint, Haifa, Israel), umethodology that has been described

r or ilure

hose 201Tl r the

from

not were roup t 18 mine and

[23]. Imaging was repeated 3-redistribution. In a random samplemore than 6 extrasystolic beats perSPET acquisition was triggered to tassessment of LVEF using a cosoftware (QGS, Cedars-Sinai) [18].

From the reconstructed slices the mdivided into 20 segments, tracer acsegment was graded using a 5-poinof the stress scores (SSS) and rescalculated [27]. Fixed defects in at segments with moderately or severtracer accumulation were coninfarction. Left ventricular dilatation

manifestations suspicious of heart (breathlessness, fatigue). In a subgroup opatients, assessed in a predefined day of the wSPET acquisition was triggered to the ECG signsimultaneous assessment of left ventricular funcmyocardial perfusion data.

In order to minimize the bias introducedenrolling consecutive patients, participanseparated into two groups. The derivatioconsisted of patients presenting during themonths of the study and was used to independent predictors of systolic dysfuncdevelop scoring systems. These findings wtested in the validation cohort includingpresenting during the last 6 months of the stuDuplicate nuclear examinations were removed one assessment per patient, at the earliest date, was entered in the study database.Demcharacteristics, medical history, symptoms, mand the reason for testing were recordephysi ian supervising the st

n ents riod.

only lable phic

ation

quantitatively from both the rreconstructed mid-ventricular slices.scales were used to grade the rventricle in the thoracic cage (1, normdefinite enlargement) and the reventricular cavity to the thickest adjcavity smaller; 2,

e c in

a for was classified as present and

entry the study weinfarction or percutan

rdial the t six

heart athy, The

int of nary oms,

Myocardial images were indepetwo experienced observers (E.M.patients’ data; discrepancies consensus. Scan interpretationprospectively to provide a report. Lefraction from gated SPET myoccalculated twice, separately by twomean value was entered in analysis

Equilibri

Original Article


Continuous variables are expressed as mstandard deviation (SD) and categorical varinumbers or proportions. Mann-Whitney statiused for the comparison of two independent sapatients. The chi-square statistic, including continuity correction, and Fischer’s exact test wfor categorical data comparisons. Inteagreement of ECG interpretations was eva

eanablestic mpl

Yere

r-obseluaterror ongs

n anaf a realue< cateen adel, w redellesogra

facyocaulascom

ed byin logistiere ced aspo

es inhe intf one an Nom

also with zero if absent and oed by th

smal w

r equssoc

d mme

anatory performan

various models in the prediction of a resting E0 [29]. Both the area under the curve (A

re reported. Linear regression anavide equations for the prediction o

andetholatio

greetati

Results

Patients’ details

One thousand two hundred and sixty-three patients fulfilled the entry criteria. Among them 514(41%) were referred for diagnostic purposes, 536(42%) for risk assessment, 186(15%) for evaluation after a coronary intervention and 27(2%) for myocardial viability determination. At the time-point of the examination, 86(7%) patients had angiographically documented CHD and were treated medically, whereas 58(5%) patients had been submitted to percutaneous coronary interventions

ypass grafting on the 54 patients having

, 82(7%) had a 6%) had undergone cs of the 954(76%)

group and 309(24%) patients in able 1, allocated in

A LVEF. st was

myocardial perfusion and 54 female, and d validation subsets, 6%) had suspected y documented CHD

05(28%) gave a n the past. Eighteen

percutaneous coronary surgery. An ERNA

f these patients.

clear testing

nd stress data are recordings the inter-irely normal versus

ood (κ=0.949, SEE 98% of cases). In changes (with the wave > poor R wave erted T waves > ST-

normalities > normal) ood (κ=0.715, SEE

77% of cases). arized in Table 1.

dged normal or near atients (359 in the

e validation cohort). re also presented in found in 152 (12%)

) patients.

tion of systolic

in logistic regression f continuous variables story time >8 months,

ic blood pressure 3. The independent by logistic regression e determination of a as scoring systems, prising resting ECG

ing system combining also included in that vailable and reliable

Receiver operating characteristics analysis showed no significant difference between the Clinical-ECG-Scinti-graphic and the ECG-Scintigraphic scoring systems in both the derivation (AUC 0.914, SEE 0.011 versus AUC 0.910, SEE 0.001, respectively, P=0.803) and validation subpopulation (AUC 0.866, SEE 0.028 versus AUC 0.870, SEE 0.028, respectively, P=0.920) in the discrimination of resting left ventricular dysfunction. Therefore, only the latter was entered in further analyses, as it contains fewer categories of data. In the derivation group, the ECG-Scintigraphic model was a better predictor of a resting ERNA LVEF<0.50 than both the Clinical-ECG scoring (AUC 0.847, SEE 0.014, P=0.000)

1 s as was

es of ates’ used

ERNA derived LVEF from the scoring systemsto estimate the correlation between various mLVEF calculation. Pearson’s coefficient of correis quoted together with the 95% limits of aaccording to Bland-Altman statistic [30]. Ssignificance was accepted for P values0.05.

and 152(12%) to coronary artery bgrounds of chronic CHD. Among 4suffered a myocardial infarctionsubsequent angioplasty and 75(bypass surgery. The characteristipatients in the derivation

rver d by f the

with

lysis sting 0.10 gory

the validation cohort are listed in Tsubgroups according to resting ERN

Left ventricular ejection fraction at recalculated by 201Tl gated SPET imaging in 373 patients (319 male286 versus 87 in the derivation anrespectively). Among them, 209(5CHD, 21(6%) had angiographicalland were on medical treatment and 1

Cohen’s kappa statistic (κ); both the standard eestimate (SEE) and the percentage of recordiidentical classification are also reported. In each category of data sets, logistic regressiowas used to identify independent predictors oERNA LVEF<0.50 among variables with a P vin univariate analysis. In variables from eachconsidered to be significant (P<0.05) were ththose of the next category to create a new modcontained only the variables that continued toindependent predictors and so forth. A final mocreated after the addition of all significant variaball categories in a hierarchical order (demcharacteristics, prior cardiac history, risksymptoms, resting electrocardiogram, mperfusion results). Since logistic regression formthe form of exponentials and require calculations, a simple linear score was creatanalysis. To accomplish this, before entered regression analysis all continuous variables win a binary fashion using a cut-off point definlevel below which the number of false positive reexceeded the number of true positive responsdetermination of a resting LVEF<0.50 [28]. Twith the largest values was assigned a value othat with the lowest values was given zero.variables were coded

de hich

history of myocardial infarction ipatients (5%) had undergone

d

main was from phic tors, rdial

take plex the

interventions and 45(12%) bypassLVEF<0.50 was found in 77(21%) o

Electrocardiography and nu

All resting ECG interpretations apresented in Table 1. In 100 ECGobserver agreement of an entabnormal interpretation was very g

c oded t the nses the

erval

0.035; absolute agreement in identifying particular resting ECGpriority left bundle branch block > Qprogression in precordial leads > invsegment deviation > other minor abinter-observer agreement was g0.051; absolute agreement ind

inal ne if

Scintigraphic results are summMyocardial perfusion images were ju

present. The coefficients of all variables providregression equation were divided by the coefficient and adjusted to their proportioninteger weights. This resulted in a simple lineain which each variable code (zero or one) was awith the probability of systolic dysfunction anmultiplied by its respective weight and suproduce a composite score.

Receiver operating characteristic (ROC) was used to compare the discrimina

e allest hole ation iated was

d to

lysis ce of RNA UC

normal (SSS 0-4) in 479(38%) pderivation group and 120 in thMeasurements LVEF with ERNA aTable 1. A resting LVEF<0.40 was patients and a LVEF<0.30 in 71 (6%

Models in the determinafunction

Acording to the criterion usedanalysis, the dichotomous points o

LVEF<0.5 ) lysis f the also ds of n (r) ment stical

were set as follows: age >60 yrs, hiresting heart rate >70bpm, systol>135mmHg, SSS>10 and SRS>variables of the models provided analysis and their coefficients in thresting ERNA LVEF<0.50, presentedare listed in Table 2. A model comreadings alone and a different scorECG and scintigraphic variables are Table. These data are more readily athan clinical information.

and the SEE awas used to pro

Original Article


Table 1. all study participants allocated in sub s according to ERNA LVEF

oup

Characteristics of group

Derivation gr

F EF<0 0 P LVE 0.50 LV .5

rs) 10.2 62.5 0.000

(n=702) ( n=252)

Age (y 59.2 9.5

Male 561 (80%) 215 (85%) 0.060

Diabetes 117 (17%) 66 (26%) 0.001

dem 2 (3 0.038

rtension 374 (53%) 113 (45%) 0.023

(3 (2 0.078

ly his 1 0 ( 0.183

Dyspnoea on 72 (10%) 52 (21%) 0.000

1 (1 0.437

en

(4 2 (2 0.001

symptom 3 (3 0.262

0.000

phCHD

0.905

0.000

82 (12%) 13 (5%) 0.002

CABG 118 (17%) 63 (25%) 0.006

y time 38.1 42.0 0.000

Antiplatelet 494 (70%) 203 (81%) 0.002

g nitrate

363 (52%) 135 (54%) 0.659

a-blocker 374 (53%) 152 (60%) 0.055

CC inhibitor 133 (19%) 53 (21%) 0.532

Diuretic 143 (20%) 62 (25%) 0.189

ACE-i / ARB 325 (46%) 147 (58%) 0.001

Entirely normal ECG

359 (51%) 21 (8%) 0.000

Fascicular block

83 (12%) 44 (18%) 0.031

mellitus

Dyslipi ia 197 ( 8%) 89 5%)

Hype

Smoking 220 1%) 64 5%)

Famiof CHD

tory 78 (1 %) 2 8%)

exertion

Angina 125 ( 8%) 39 6%)

Atypical chpai

st 286 1%) 7 9%)

A atic 219 ( 1%) 89 5%)

Suspected CHD

325 (46%) 63 (25%)

Angiogra ic 48 (7%) 16 (6%)

MI

PCI

192 (27%) 145 (58%)

Histor(mo)

24.8 48.2

Long actin

Bet

Deri on group vati

LVEF0.50 LVEF<0.50 P

(n=702) ( n=252)

mchanges

(1 44 (18%) 0.050 ST-seg ent 86 2%)

Inverted T 94 (13%) 70 (28%) 0.000

si53 (21%) 0.000

wa (1 109 (43%) 0.000

r Qwave

2 ( 74 (29%) 0.000

Q 6 ( 36 (14%) 0.000

Lateral Q 4 (0.5%) 18 (7%) 0.000

1 ( 32 (13%) 0.000

Dynamic 446 (64 125 (50%) 0.000

e ( 123 (49%) 0.000

mstress

(1 4 (2%) 0.303

(bpm) 14.6 0.001

Resting SBP )

141.522.9 138.022.0 0.060

) .0 87.39.9 0.558

Sc6.56.3 17.910.1 0.000

o 10.89.1 0.000

Anteroseptal MI

25 (4%) 102 (41%) 0.000

Inferior MI 55 (8%) 91 (36%) 0.000

Lateral MI 11 (2%) 40 (16%) 0.000

Dilated left ventricle

8 (1%) 59 (23%) 0.000

ERNA LVEF 0.620.07 0.38 0.10 0.000

wave

Poor R progres on

71 (10%)

Any Q ve 68 0%)

Anterio 3 5%)

Inferior 3wave

5%)

wave

LBBB 2 3%)

exercise %)

Dipyridastr ss

mole 250 36%)

Dobuta ine 6 %)

Resting HR 72.6 13.9 75.7

(mmHg

Resting DBP(mmHg

88 9.9

Summed Stress ore

Summed Rest Sc re

2.4 4.0

Original Article


Table 1. (C

a ou ontinued)

V lidation gr p

F0LVE .50 LVEF<0.50 P

Age

(n= 2 (n

(yrs) 0. 0.824

28) =81)

59.9 9.9 6 311.7

Male 173 (76%) 67 (83%) 0.219

(18 9 0.415

Dyslipidemia 77 (34%) 28 (35%) 1.000

on (50 1 1.000

0.086

y history D

17 (8%) 9 (11%) 0.433

spnoea on 27 (12%) 13 (16%) 0.438

24 4 0.276

pical chest 86 (38%) 21 (26%) 0.058

mptoma (27 3 0.027

spected

(46 1 0.002

phi (7 7 0.712

32 4 0.001

37 (16%) 8 (10%) 0.201

0.375

me 38.6 39.1 0.163

et (7 6 0.225

acting (47 49 0.039

ocke (5 4 0.067

CC inhibitor 54 (24%) 12 (15 ) 0.115

0

Diabetes mellitus

42 %) 1 (24%)

Hypertensi 114 %) 4 (51%)

Smoking

Famil

72 (32%) 17 (21%)

of CH

Dyexertion

Angina

Aty

54 ( %) 1 (17%)

pain

Asy tic 61 %) 3 (41%)

SuCHD

104 %) 2 (26%)

Angiogra c 15 CHD

%) (9%)

MI

PCI

73 ( %) 4 (54%)

CABG 31 (14%) 15 (19%)

History ti 26.7(mo)

52.0

Antiplatel 169 4%) 6 (82%)

Longnitrate

107 %) (61%)

Beta-bl r 124 4%) 5 (67%)

%

Diuretic 47 (21%) 25 (31%) 0.085

ACE-i / ARB 108 (47%) 47 (58%) 0.121

Entirely normal ECG

103 (45%) 6 (7%) 0.00

Fascicular block

36 (16%) 19 (24%) 0.167

on group

EF LVEF<0.50 P

Validati

LV 0.50

= (n=81)

enchanges

( 11 (14%) 0.721

(n 228)

ST-segm t 36 16%)

Inverted T wave

27 (12%) 22 (27%) 0.002

Poor R sion

34 (15%) 12 (15%) 1.000

ave 29 (36%) 0.000

wa

( 21 (26%) 0.000

wave 3 8 (10%) 0.305

8 (10%) 0.003

LBBB 4 (2%) 14 (17%) 0.000

Dynamic 134 (59%) 43 (53%) 0.449

Dipyridamole 90 (40%) 36 (44%) 0.515

Dobutamine 4 (2%) 2 (3%) 0.621

Resting HR (bpm)

73.312.7 74.212.8 0.600

SBP

1.5 138.223.8 0.356

Resting DBP 88.29.8 87.710.3 0.842

Summed co

6.96.1 15.610.4 0.000

re

5 9.79.3 0.000

%) 0.000

Inferior MI 24 (11%) 24 (30%) 0.000

0 (0%) 13 (16%) 0.000

5%) 0.000

LVEF 0.610.06 0.380.10 0.000

progres

Any Q w 23 (10%)

Anterior Qve

8 4%)

Inferior Q 1 (6%)

Lateral Q wave

4 (2%)

exercise

stress

stress

Resting (mmHg)

14 22.3

(mmHg)

Stress S re

SummedRest Sco

2. 3.5

Anteroseptal MI

9 (4%) 30 (37

Lateral MI

Dilated left 0 (0%) 20 (2ventricle

ERNA

MI,myocardial infarction; PCI, percutaneous coronary intervention; CABG, coronary artery bypass grafting; CC, calcium channel; ACE-i, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; LBBB, left bundle branch block; HR, heart rate; SBP, systolic blood pressure; DBP, diastolic blood pressure. Rest of abbreviations as in the text.

and ECG scoring (AUC 0.816, SEE 0.016, P=0.000) (Fig. 1a). Similarly, in the validation sample the ECG-Scintigraphic scoring system had a better discriminatory probability for systolic function impairment than both the clinical-ECG model (AUC 0.789, SEE 0.030, P=0.001) and the ECG score (AUC 0.795, SEE 0.029, P=0.003) (Fig. 1b). The Clinical-ECG model was a better discriminator of resting left ventricular dysfunction than

ECG scoring in the derivation cohort (P=0.001), but not in the validation sample (P=0.757).

Regression analysis provided the following equation for the prediction of resting ERNA LVEF from the ECG-Scintigraphic scoring system:

“ERNA LVEF = 66 - 3 x ECG-Scintigraphic score” (r=0.731, P=0.000) (1).

Original Article


Table 2. Models for the prediction of a resting ERNALVEF<0.50.

a ori e partifeature is present*

cular Models nd sc ng if th

Variables Cl-

Clinical-ECG-Scintigraphic

ESr

linicaECG

ECG CG-cintigaphic

Age > 60 1 - 1 -

Dyspnoea on

Histor

exertio 1 -

y of myocardial 2 - -

eart rate 2 2 -

ECG 2 1

es 1 1

ave 1 1

ression 1 1

2 - 2 -

r Q-wave 2 2 1 1

ave - 2 2

4 3 3 2

ted left ventricl 4

1

3

2

- 3 - 2

0

n 2 -

-

1

2

-

-

-

infarction

Resting h>70bpm

Abnormal 3

ST-chang -

Inverted T w -

Poor R wave prog

-

Any Q-wave

Anterio

Lateral Q-w -

LBBB

Dila e - 7 -

SSS >10 - 2 -

Anteroseptal MI - 5 -

Lateral MI - 4 -

Inferior MI

Sum 0 - 19 0 - 34 0 – 12 - 20

*For example: “ECG-scintigraphic score = 1 (if resting HR2 (if ECG is abnormal) + 1 (if anterior Q-wave present) + 2 (if lateral Q-wave present) + 2 (if LBBB present) + 4 (if left ventricle is dilated) + 1 (if SSS10) + 3 (if anteroseptal MI presscan) + 2 (if lateral MI present in scan) + 2 (if inferior MI prin scan)”. If the particular characteristic is not presenrespective value is 0

Apparently, LVEF values with equation (1)provided in 3-point increments.

70) +

ent in esent t the

are

Figure 1. Receiver operating characteristic curves of scoring systems in the prediction of a resting ERNA LVEF<0.50 in the derivation (a) and validation (b) groups.

SPET

e gated SPET, ROC stic efficacy between stem predicted LVEF LVEF calculated by ng (AUC 0.920, SEE

n of a resting ERNA average of LVEF

nd gated SPET (AUC significantly better

either calculation of respectively) (Fig. 2).

LVEF versus gated model LVEF and the aches are plotted in limits of agreement 010.176 versus

gh as opposed to ssion analysis does tests, for reasons

omparison with we have also performed such analysis

in the current study. An ERNA LVEF<0.50 was predicted correctly in 63/77 patients (82%) with gated SPET (adjusted for the systematic underestimation of 7.1% in Bland-Altman analysis) and in 71/77 patients (92%, P=0.000) with the average LVEF.

Scoring systems and gated

In 373 patients who had undergonanalysis showed a similar diagnothe ECG-Scintigraphic scoring sy(AUC 0.912, SEE 0.020) and the201Tl gated SPET myocardial imagi0.022, P=0.7642) in the discriminatioLVEF<0.50 (Fig. 2). Moreover, thevalues from that scoring system a0.961, SEE 0.016) had a discriminatory performance than LVEF alone (P=0.004 and P=0.001,

The associations between ERNASPET LVEF, the ECG-Scintigraphic average LVEF by both these approFigure 3. The corresponding 95% ranged 0.0710.196 versus 0.00.0400.152, respectively. AlthouBland-Altman statistic, linear regrenot express agreement betweenof consistency and for facilitating cprevious studies,

stic curves of the ECG-

aphic scoring system LVEF, Tl Gated SPET LVEF and VEF, derived from both those assessments, in the

crimination of a resting ERNA LVEF<0.50.

VEF 60%, a 10% d gated SPET LVEF n) was found in

VEF of the ECG-T, 170 patients (89%,

P=0.000) were found with no more than 10% difference

from the ERNA LVEF.

Discussion Left ventricular ejection fraction has the attraction of being a simple numerical parameter that characterizes systolic function and is an established determinant of prognosis in patients with suspected or known CHD [31, 32]. This study developed models, based on readily available variables collected during routine clinical evaluation of those patients, for the discrimination of

Figure 2. Receiver operating charateriscintigr 201

average Ldis

In the 191 patients with an ERNA Ldifference between ERNA LVEF an(adjusted for the underestimatio125(65%) cases. In using the average LScintigraphic model and gated SPE

Original Article


resting left ventricular systolic impairment and aestimation of LVEF. It was demonstrated that tachieved by a simple algebraic summation of cweighted ECG and scintigraphic data, whinformation comparable to that provided by 20

lso fohis isoded

ich o1Tl g

S nce o

t andardio

correldaliti How aveus, ited v

value the hes

onfidhe ernd-Alr of

ng sily on wide

eF an

issendatal dosET ere

Tc g ERNA LVEF has also been obse

p [21, thTc g

r con. Tooent ques

l boe on echocardiography, ERN

ssmen[10, of g

01Tl gre tremed

phivelyes

gh i

We have also reported earlier an exceinter-observer agreement in the calculation of LVEF by both ERNA and 201Tl gated SPET [23]. In our internal control we have found a good and consistent over time agreement between experienced observers in the evaluation of both myocardial ischaemia and left ventricular dilatation by the above set criteria. Although inter-institutional differences may exist in grading myocardial perfusion, imaging systems usually offer the option of perfusion quantification in a polar map format, thus minimizing interpretative discrepancies. Similarly, many ECG recorders identify abnormal patterns automatically.

ediction of left ventricular

r systolic dysfunction G is entirely normal reement with earlier ts with an abnormal

tolic function. Clinical r limited predictive e discrimination of a Other authors have

usefulness of clinical n with ECG and

rmination of impaired 38, 38]. Extending

ohort of patients we of a resting ERNA

ntly improved with the data (Fig. 1). In this g system estimated ance comparable to

LVEF (Fig. 2). More importantly, roposed model and

bability was further increased (Fig. 2) and significantly more cases with an

tly identified (92%)

ricular ejection

d and weighted ECG quation (1) provided a

y to ERNA LVEF reement comparable

201Tl gated SPET

F values from the d SPET provided a ned variability and

(Fig. 3). Furthermore, ERNA LVEF ranged 15.2% with the

was reflected in in which an accurate important in clinical ation average LVEF cases (89%) with no the ERNA LVEF, e (65%). arisons of 201Tl and

nt of LVEF, a similar limits of agreement

d either first-pass nce imaging has been

reported [19, 20]. However, 201Tl is considered somewhat inferior to 99mTc-compounds in terms of image quality, study repeatability and interpretive reproducibility [40, 41]. Despite the fact that our gated myocardial data were of suboptimal count statistics, the association between 201Tl gated SPET and ERNA LVEF measurements in our population is in agreement with previous works [18, 23, 25, 34, 35]. Moreover, the uncertainty in the prediction of ERNA LVEF from 201Tl gated SPET in our study is comparable with the disparity found between ERNA, contrast ventriculography or MRI measured and 99mTc gated SPET calculated LVEF values [21, 22, 25, 36].

r the best and

Pr

ffers ated f the

has logy. ation es in ever, rage n an alue (e.g. 95%

e are ence ror of tman the milar

dysfunction This study shows that left ventriculaat rest is highly unlikely if the EC(Table 1), which is in substantial agseries [5-9]. However, many patienECG would also have normal sysvariables seemed to add no oinformation over ECG readings in thresting ERNA LVEF<0.50 (Fig. 1). reported conflicting results on the features alone or in combinatioradiographic information in the deteleft ventricular function [15, 16,previous observations, in a large cdemonstrated that the prediction LVEF<0.50 by the ECG is significaassimilation of myocardial perfusionsetting the ECG-scintigraphic scorinLVEF had a discriminatory performthat of 201Tl gated SPET

PET and can enhance the clinical performalatter.

Gated SPET has been a major achievemenbecome standard practice in nuclear cNumerous studies showed a high degree of between gated SPET and other accepted mothe assessment of left ventricular function [18].in clinical practice individual values, rather thangroup tendencies need to be determined. Thindividual patient the variability of the predic(e.g. ERNA LVEF) from a given independent gated SPET LVEF) can be assessed with prediction intervals of the regression equation. Tdifferent from the commonly quoted 95% cintervals of the regression line, which provide tthe comparisons between groups [33]. The Blastatistic can also be used as an indicatoagreement of paired measurements. In usiimaging techniques, we have reported previous95% prediction intervals and 95% limits of agreemcomparing 201Tl gated SPET with ERNA LVEhave found a 10% absolute difference in LVEFless than 25% of cases [23]. Our figures were ereplicated when we extracted and re-analyzed other published series administering higher 201Trest [24, 34]. Similar findings with 201Tl gated SPbeen reported by other investigators [35], whcomparable degree of inaccuracy between 99m

SPET LVEF and

nt in d we

o

by averaging LVEF values from the pgated SPET the discriminatory pro

n ntially

from es at have as a

ERNA LVEF<0.50 were correccompared to gated SPET (82%).

Estimation of left vent

ated rved

36]. at a ated trast ls for and

are

dies

fraction The assessment of LVEF from codeand scintigraphic information with eproportion of unexplained variabilitmeasurements and 95% limits of agto those of the association between and ERNA LVEF (Fig. 3).

In this setting, average LVEproposed scoring system and gatedecreased proportion of unexplai

reviously by our team and other authors Moreover, meta-analyses also concurred insubstantial disparity exists in LVEF between 99m

SPET and cardiac magnetic resonance oventriculography on a per-patient basis [22, 25]improving precision of LVEF measuremagreement amongst different imaging technitherefore desirable.

It should be emphasized that internationarecommend relianc A or

t of 37], ated ated ated left asise , left were

nter-nd in llent

narrower 95% predictions intervals the 95% limits of agreement to 19.6% with gated SPET versusaverage LVEF. This improvement patients with an ERNA LVEF60%,LVEF assessment is particularly decision making. In that subpopulvalues predicted significantly more more than 10% deviation fromcompared to gated SPET LVEF alon

In studies of head-to-head comp99mTc gated SPET in the assessmecoefficient of correlation and 95%between those methods anangiography or magnetic resona

cardiac magnetic resonance for the assesystolic function in clinical decision making whereas no clinical guideline endorses the useSPET for this purpose. Thus, in this study 2

SPET and the developed scoring systems weas modalities identifying patients with presuventricular dysfunction. It is also important to emthat although patients were enrolled retrospectventricular systolic function related estimatperformed prospectively.

In keeping with previous data [7], a hiobserver agreement in ECG interpretations was fouour study.

Original Article


1. Wang TJ, Evans JC, Benjamin EJ Figure 3. Linear regression analysis, with the reg(solid line) and the 95% prediction intervals (doillustrated, and the Blan

ressiotted

d-Altman statistic between ERNA 201 LVEF (a, b), the ECG-scintigraphic

average LVEF from the ECG-Scintrigr

wouldSPETthis i

form inexpla

LVEF, 36]

ation of LVEF from the Estem. As this model contains applicable with either exe

Thus,trol to

T, aleads

er, in sites in which gated SPET is not performed or in situations it cannot be applied for various reasons (e.g. tachyarrhythmia), the proposed scoring system would offer a competent alternative.

Limitations of the study

It is possible that selection bias skews our patients. In an effort to minimize this, the results of a derivation group were verified in a second cohort. Moreover, since study participants were evaluated with myocardial perfusion imaging, referral bias is also a problem. It should be

selected population, ere disease would be n that subpopulation, ove the threshold of

with gated SPET e important from the clinical standpoint,

ll almost invariably

on or flutter were e rhythm on systolic

scintigraphic features his subpopulation. clinically significant

oted with heart rate ent evidence

e left ventricular heart failure [43]. In tion would not have

suspected or known ction at rest can be

ively from a simple model containing resting ECG and stress myocardial perfusion imaging

bles. This scoring system can provide an estimation ich is comparable to s and can enhance

r.

et al. Natural history of ic dysfunction in the 2. Group on ACE Inhibitor angiotensin-converting rbidity in patients with

ittees. The Cardiac -II): a randomised trial.

3. e A et al. Symptomatic

ysfunction in an 3.

et al. Value of the t failure due to left

312: 222. ons RJ. Value of normal

t ventricular nd suspected coronary

-62. et al. A normal eft ventriculography in Heart 1998; 79: 262-

wee P et al. Prevalence ction with normal rest 79: 1295-8. ardiogram as a predictor lation with gated SPECT

imaging. Mount Sinai J Med 2003; 70: 306-9. 10. Swedberg K, Cleland J, Dargie H et al. Guidelines for the diagnosis and treatment of chronic heart failure: full text (update 2005). The Task Force for the diagnosis and treatment of CHF of the European Society of Cardiology. Available : www.escardio.org/knowledge/guidelines/Chronic_Heart_Failure.htm 11. Khunti K, Squire I, Abrams KR, Sutton AJ. Accuracy of a 12-lead electrocardiogram in screening patients with suspected heart failure for open access echocardiography: a systematic review and meta-analysis. Eur J Heart Failure 2004; 6: 571-6. 12. Galasko GIW, Barnes SC, Collinson P et al. What is the most cost-effective strategy to screen for left ventricular systolic dysfunction: natriuretic peptides, the electrocardiogram, hand-

n line lines) LVEF

asymptomatic left ventricular systolcommunity. Circulation 2003; 108: 977-82. Garg R, Yusuf S, for the Collaborative

and Tl gated SPETLVEF (c, d) and the

score aphic

Trials. Overview of randomized trials onenzyme inhibitors on mortality and mo

score and gated SPET (e,f).

Thus, it would be expected that our results have been significantly affected if 99mTc gated used instead, although firm conclusions on require a direct evaluation.

Clinical relevance

The ECG-scintigraphic scoring LVEF offers inequivalent to that of gated SPET LVEFdetermination of systolic function. This can be by the uncertainty in the prediction of ERNA gated SPET on a per-patient basis [21-25, 35the independent estim

not

heart failure. JAMA 1995; 273: 1450-6. 3.CIBIS-II Investigators and CommInsufficiency Bisoprolol Study II (CIBISLancet 1999; 353: 9-1 was

ssue 4. McDonagh TA, Morrison CE, Lawrencand asymptomatic left-ventricular systolic durban population. Lancet 1997; 350: 829-3

ation the

n. BMJ 1996; 6. O’Keefe JH Jr, Zinsmeister AR, Gibbelectrocardiographic findings in predicting resting lef

ined with and GC-

function in patients with chest pain aartery disease Am J Med 1989; 86: 6587. Khan MA, Sinha S, Hayton Selectrocardiogram precludes the need for lthe assessment of coronary artery disease

Scintigraphic scoring systress ECG variables, it i

s no rcise the

7. 8. Christian TF, Miller TD, Chareonthaitaof normal resting left ventricular funof pharmacologic myocardial scintigraphy.

proposed scoring system can serve as a conresting LVEF assessments with gated SPEaveraging LVEF values from those approaches more accurate prediction of ERNA LVEF. Moreov

added, however, that in a less patients with suspected or less sevrepresented in larger proportions. ILVEF would very likely be well ababnormality. Hence deviations in LVEF would not b

ol of s in to a

electrocardiograms. Am J Cardiol 1997; 9. Henzlova MJ, Croft LB. The electrocof left ventricular systolic function: corre

because predicted values would fawithin the normal range.

Patients with atrial fibrillatiexcluded. Apart from the effect of thfunction, however, other ECG and probably would also be important in t

Regarding medication, no changes in resting LVEF have been nlimiting drugs [42], whereas more recshowed that beta-blockers can increassystolic function in patients with addition, heart rate limiting medicamasked extensive ischemia.

In conclusion, in patients withCHD left ventricular systolic dysfundetermined effect

variaof resting LVEF with a reliability whthat of 201Tl gated SPET calculationthe clinical performance of the latte

Bibliography

5.Davie AP, Francis CM, Love MPelectrocardiogram in identifying hearventricular systolic dysfunctio

Original Article


held echocardiography, traditional echocardiography, or

J. S to diogra

hnique relev

of pheart fa

tilizatilar sy

9: 394-8. f previcommrdiogr

antitatapproa

ventrl per

ith: 412ssme

Tompa

. J Am

on fra usefuysiol

ent tative

resong 200

ted sl per

t ventrd it be favoured over equilibrium radion

. Indiencieol Im

ardiogram-tomograph versus

cardiac magnetic resonance imaging for the assessment of left ventricular volumes and ejection fraction. J Am Coll Cardiol 2002; 39: 2059-68. 26. Morise AP, Haddad J, Beckner D. Development and validation of a clinical score to estimate the probability of coronary artery disease in men and women presenting with suspected coronary disease. Am J Med.1997; 102: 350-6. 27. Hansen CL, Goldstein RA, Akinboboye OO, et al. ASNC imaging guidelines for nuclear cardiology procedures. Myocardial perfusion and function: Single photon emission computed tomography. J Nucl Cardiol 2007;14:e39-60.

thallium-201 scintigraphy y artery disease. Ann

aring the areas curves derived from the 3.

for assessing clinical measurement.

al. Influence of ejection n a broad spectrum of 112: 3738-44. Research Group. Risk

rdial infarction. N Engl J

for Medical Research. 324.

parative value of ECG-yocardial perfusion

stolic left ventricular 03; 30: 859-67. et al. Quantitative gated ing with 201Tl: an

Med Commun 2000;

G et al. Comparison of with 99mTc-RBC

left ventricular volume th ischemic heart failure

(suppl): S234. et al. ACC/AHA 2005 anagement of chronic American College of

Task Force on Practice te the 2001 Guidelines

Heart Failure). Available e//index.pdf.

il-Smoller S et al. The nship between left ventricular systolic function and

d by clinical criteria.

ersh BJ. The utility of raphic variables in

. Am J Cardiol 1995; 75:

et al. Comparison of Tc-rfusion SPECT. J Nucl

41. Hyun IY, Kwan J, Park KS, Lee WH. Reproducibility of Tl-201 and Tc-99m sestamibi gated myocardial perfusion SPECT measurement of myocardial function. J Nucl Cardiol 2001;8: 182-7. 42. Nestico PF, Hakki AH, Iskandrian AS. Effects of cardiac medication on ventricular performance: emphasis on evaluation with radionuclide angiography. Am Heart J 1985; 109: 1070-84. 43. Fisher ML, Gottlieb SS, Plotnick GD et al. Beneficial effects of metoprolol in heart failure associated with coronary artery disease: a randomised trial. J Am Coll Cardiol 1994; 23: 943-50.

their 28. Kotler TS, Diamond GA. Exercise in the diagncombination? Eur Heart J 2006; 27: 193-200.

13. Houghton AR, Sparrow NJ, Toms E, Cowley Ageneral practitioners use the electrocardiogram

hould select

osis and prognosis of coronarIntern Med 1990; 113: 684-702. 29. Hanley JA, McNeil BJ. A method of comp

patients with suspected heart failure for echocarIntern J Cardiol 1997; 62: 31-6. 14. Franklin K, Marwick TH. Non-invasive

phy?

s for

under receiver operating characteristic same cases. Radiology 1983; 148: 839-430.Bland Jtec

assessing cardiovascular function: which, if any, areclinical practice? Coronary Artery Disease 2002; 13: 391-7.

ant to M, Altman DG. Statistical methods

agreement between two methods ofLancet 1986;1 (8476):307-10.

15. Stevenson LW, Perloff JK. The limited reliabilitysigns for estimating hemodynamics in chronic h

ysical ilure.

31. Solomon SD, Anavekar N, Skali H etfraction on cardiovascular outcomes i

JAMA 1989; 261: 884-8. 16. Talreja D, Gruver C, Sklenar J et al. Efficient uechocardiography for the assessment of left ventricufunction. Am Heart J 2000; 13

on of heart failure patients. Circulation. 2005; 32. The Multicenter Postinfarction

stolic

ously

stratification and survival after myocaMed 1983; 309: 331-6. 33. Altman DG. Practical Statistics 17. Ng LL, Loke I, Davies JE et al. Identification o

undiagnosed left ventricular systolic dysfunction: unity aphy.

ion in

Chapman & Hall, London 1994, pp 299-34. Daou D, Vilain D, Colin P et al. Comgated blood pool SPET and ECG-gated m

screening using natriuretic peptides and electrocaEur J Heart Failure 2003; 5: 775-82. 18. Germano G, Kavanagh PB, Slomka PJ et al. Qugated perfusion SPECT imaging: The Cedars-Sinai Nucl Cardiol 2007; 14: 433-54. 19. He ZX, Cwajg E, Preslar JS et al. Accuracy of leftejection fraction determined by gated myocardiaSPECT with Tl-201

ch. J

icular fusion

SPET in the assessment of global syfunction. Eur J Nucl Med Mol Imaging 2035. Wright G A, McDade M, Keeble W SPECT myocardial perfusion imagassessment of the limitations. Nucl

and Tc-99m sestamibi: comparison wpass radionuclide angiography. J Nucl Cardiol 1999; 620. Tadamura E, Kudoh T, Motooka M et al. Asseregional and global left ventricular function by reinjection

first--7. nt of l-201 rison Coll

21:1147-51. 36. Tsorbatzoglou K, Karatzas N, Arsos99mTc tetrofosmin Gated SPECT (GSPECT) radionuclide ventriculography (RNV) inand function assessment in patients wi(abstr.). Eur J Nucl Med Mol I

and rest Tc-99m sestamibi ECG-Gated SPECT. Cwith three-dimensional magnetic resonance imagingCardiol 1999; 33: 991-7. 21. Wright GA, McDade M, Keeble W et al. Are ejecti ctions

l? A

maging 2006; 33 37. Hunt SA, Abraham WT, Chin MH guideline update for the diagnosis and mheart failure in the adult: a report of the from gated SPECT perfusion studies clinically

comparison with radionuclide ventriculography. Ph2001; 22: 413-22. 22. Kondo C, Fukushima K, Kusakabe K. Measurem

Meas

of left

Cardiology/American Heart AssociationGuidelines (Writing Committee to Updafor the Evaluation and Management of

ventricular volumes and ejection fraction by quantiSPECT, contrast ventriculography and magnetic imaging: a meta-analysis. Eur J Nucl Med Mol Imagin851-8.

gated ance

3; 30:

at www.acc.org/clinical/guidelines/failur38. Marantz PR, Tobin JN, Wassertherelatio

23. Moralidis E, Spyridonidis T, Arsos G et al. 201Tl gaphoton emission computed tomographic myocardiaimaging in the assessment of global and regional leffunction. Woul

ingle-fusion

congestive heart failure diagnoseCirculation 1988; 77: 607-12. 39. Rihal CS, Davis KG, Kennedy JW, G

icular uclide

clinical, electrocardiographic, and roentgenogthe prediction of left ventricular function220-3. angiography? Nucl Med Commun 2004; 25: 665-73.

24. Moralidis E, Arsos G, Boundas D, Karakatsanis Kmeasurements may not closely follow average tend

vidual s: the

ing

40. DePuey EG, Parmett S, Ghesani M 99m sestamibi and Tl-201 gated peCardiol 1999; 6: 278-85. paradigm of nuclear cardiology. Eur J Nucl Med M

2003; 30: 1598-9. 25. Ioannidis JPA, Trikalinos TA, Danias PG. Electrocgated single-photon emission computed

ag

y

resting electrocardiogram and stress myocardial perfusion ... · medicine,asdasdasd 1. department...

Documents