JACC: CARDIOVASCULAR IMAGING VOL. 14, NO.

7, 2021

ª 2021 BY THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION

PUBLISHED BY ELSEVIER

ORIGINAL RESEARCH

CT Angiography Followed by Invasive

Angiography in Patients With Moderate
or Severe Ischemia-Insights From the
ISCHEMIA Trial
G.B. John Mancini, MD,a Jonathan Leipsic, MD,a Matthew J. Budoff, MD,b Cameron J. Hague, MD,a,c
James K. Min, MD,d Susanna R. Stevens, MS,e Harmony R. Reynolds, MD,f Sean M. O’Brien, PHD,e
Leslee J. Shaw, PHD,g Cholenahally N. Manjunath, MD,h Kreton Mavromatis, MD,i Marcin Demkow, MD,j
Jose Luis Lopez-Sendon, MD,k Alexander M. Chernavskiy, MD, PHD,l Gilbert Gosselin, MD,m
Herwig Schuchlenz, MD,n Gerard P. Devlin, MD,o Anoop Chauhan, MD,p Sripal Bangalore, MD, MHA,f
Judith S. Hochman, MD,f David J. Maron, MDq

ABSTRACT

OBJECTIVES This study aimed to examine the concordance of coronary computed tomographic angiography (CCTA)
assessment of coronary anatomy and invasive coronary angiography (ICA) as the reference standard in patients enrolled
in the ISCHEMIA (International Study of Comparative Health Effectiveness with Medical and Invasive Approaches).

BACKGROUND Performance of CCTA compared with ICA has not been assessed in patients with very high burdens of
stress-induced ischemia and a high likelihood of anatomically significant coronary artery disease (CAD). A blinded CCTA
was performed after enrollment to exclude patients with left main (LM) disease or no obstructive CAD before random-
ization to an initial conservative or invasive strategy, the latter guided by ICA and optimal revascularization.

METHODS Rates of concordance were calculated on a per-patient basis in patients randomized to the invasive strategy.
Anatomic significance was defined as $50% diameter stenosis (DS) for both modalities. Sensitivity analyses using a
threshold of $70% DS for CCTA or considering only CCTA images of good-to-excellent quality were performed.

RESULTS In 1,728 patients identified by CCTA as having no LM disease $50% and at least single-vessel CAD, ICA
confirmed 97.1% without LM disease $50%, 92.2% with at least single-vessel CAD and no LM disease $50%, and only
4.9% without anatomically significant CAD. Results using a $70% DS threshold or only CCTA of good-to-excellent
quality showed similar overall performance.

CONCLUSIONS CCTA before randomization in ISCHEMIA demonstrated high concordance with subsequent ICA for
identification of patients with angiographically significant disease without LM disease.
(J Am Coll Cardiol Img 2021;14:1384–93) © 2021 by the American College of Cardiology Foundation.

From the aCenter for Cardiovascular Innovation, University of British Columbia, Vancouver, British Columbia, Canada; bLundquist
Institute, Torrance, California, USA; cSt. Paul’s Hospital Department of Radiology, Vancouver, British Columbia, Canada; dCleerly,
Inc., New York, New York, USA; eDuke Clinical Research Institute, Durham, North Carolina, USA; fNew York University Grossman
School of Medicine, New York, New York, USA; gWeill Cornell Medicine, New York, New York, USA; hSri Jayadeva Institute of
Cardiovascular Sciences and Research, Bangalore, India; iEmory University School of Medicine, Atlanta, Georgia, USA; jInstitute
of Cardiology, Warsaw, Poland; kHospital Universitario La Paz-IdiPaz- CIBER-CV, Madrid, Spain; lE.Meshalkin National Medical
m
Research Center of the Ministry of Health of the Russian Federation, Novosibirsk, Russia; Montréal Heart Institute, Montréal,
Québec, Canada; nLKH Graz II, Department fuer Kardiologie und Intensivmedizin, Graz, Austria; oGisborne Hospital, Gisborne,
p q
New Zealand; Blackpool Teaching Hospitals, Lancashire, United Kingdom; and the Department of Medicine, Stanford
University School of Medicine, Stanford, California, USA.
Jagat Narula, MD, served as Guest Editor for this paper.

ISSN 1936-878X/$36.00 https://doi.org/10.1016/j.jcmg.2020.11.012

JACC: CARDIOVASCULAR IMAGING, VOL. 14, NO. 7, 2021 Mancini et al. 1385
JULY 2021:1384–93 CCTA in the ISCHEMIA Trial

T he ISCHEMIA (International Study of stenosis (DS) LM disease on CCTA; 4) absence ABBREVIATIONS

Comparative Health Effectiveness With Med- of $50% DS on CCTA for participants enrolled AND ACRONYMS

ical and Invasive Approaches) assessed after stress imaging tests; 5) absence of $70%
CAD = coronary artery disease
whether patients with moderate or severe ischemia DS on CCTA for participants enrolled after
CCTA = coronary computed
on functional testing would have improved outcome nonimaging exercise tolerance tests; 6) tomographic angiogram
if treated with an initial invasive strategy that >6 months between pre-randomization CCTA
CX = circumflex
included optimal medical therapy (OMT) compared and ICA; and 7) post-randomization revas-
DS = diameter stenosis
with OMT alone with cardiac catheterization reserved cularization procedure before the date of the
ICA = invasive coronary
for failure of medical therapy (1,2). The protocol baseline diagnostic angiogram (Supplemental angiography
included a pre-randomization blinded coronary Figure 1, Supplemental Table 1). A total of
LAD = left anterior descending
computed tomographic angiogram (CCTA) in the ma- 1,757 participants were available for analyses,
LM = left main
jority of participants to address 3 issues. First, it representing 67.9% of the total number ran-
OMT = optimal medical therapy
was a pragmatic way to address safety and commen- domized to the invasive arm (n ¼ 2,588) and
RCA = right coronary artery
surately to facilitate physician willingness to 91.9% of the scans received by the core CCTA
randomize patients by excluding those with impor- laboratory for participants randomized to the inva-
tant left main (LM) disease. Second, CCTA helped sive strategy (n ¼ 1,913). There were 1,728 patients
avoid randomization of patients with no significant with complete information to make overall patient
obstructive coronary artery disease (CAD) who would level assessment for the primary per-patient analysis
not benefit from revascularization and who would and the remaining (n ¼ 29) were suitable only for
dilute statistical power of the trial. Finally, it over- inclusion in various vessel and segment-level ana-
came concerns that randomization at the time of lyses. There were 1,296 patients with complete in-
invasive coronary angiography (ICA) might dissuade formation for analysis of agreement between CCTA
physicians from randomizing patients in the catheter- and ICA for the number of vessels diseased; a case
ization laboratory with knowledge of the presence of was considered not evaluable for the number of ves-
high anatomic burden of disease (2,3). The use of sels diseased if certain designated segments could not
CCTA to address these issues was based upon the be interpreted for the presence of $50% DS: for
strong relationship between CCTA and ICA and the example, the mid-right coronary artery (RCA),
rapidly emerging role of CCTA as a tool to improve because of cardiac motion or other artifact.
use of angiography suites (4–13). The primary aim of The primary analysis was based upon use of
this analysis was to identify the per-patient concor- the $50% DS threshold in all suitable patients and
dance between CCTA and ICA for identification of applied to both ICA and CCTA. Secondary analyses
obstructive CAD and absence of LM disease. We also (shown in the Supplemental Appendix and
explored discordance for absence of significant LM Supplemental Tables 2 to 6) used: 1) only those
and concordance for burden of disease as reflected participants enrolled after use of stress imaging to
by 1-, 2- or 3-vessel disease, location of disease per assess ischemic burden, and the $50% DS threshold
major vessel, and for assessment of proximal disease (n ¼ 1,260, excluding those enrolled only after
in the left anterior descending (LAD) artery. nonimaging exercise tolerance testing because
randomization of such patients was based upon
METHODS a $70% DS threshold on CCTA); 2) $70% DS
threshold applied to all patients for both CCTA and
The study population consisted of the participants ICA; and 3) a $70% DS threshold applied to CCTA
randomized to the invasive group who had core compared with a $50% threshold applied to ICA.
laboratory-interpreted, pre-randomization CCTAs Finally, sensitivity analyses of the primary and sec-
and core laboratory-interpreted baseline ICA within ondary aims were performed using only good- or
6 months of CCTA. Exclusion criteria for this analysis excellent-quality CCTA. CCTA image quality was
were: 1) previous coronary artery bypass grafting; 2) recorded based on assessment of overall image
noninterpretable CCTA or ICA; 3) $50% diameter noise, presence of motion artifacts, poor contrast,

The authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’
institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information,
visit the Author Center.

Manuscript received March 5, 2020; revised manuscript received November 2, 2020, accepted November 5, 2020.
1386 Mancini et al. JACC: CARDIOVASCULAR IMAGING, VOL. 14, NO. 7, 2021

CCTA in the ISCHEMIA Trial JULY 2021:1384–93

C E N T R A L IL L U ST R A T I O N Overall Percent Agreement Between ICA and CCTA

100% 2.9% (50 ICA / 1,728 CCTA)

4.9% (84 ICA / 1,728 CCTA)
90%

80%

70%

60%

50%
92.2% (1,594 ICA / 1,728 CTA) 97.1% agreement (no LM ≥50%)

40%

30%

20%

10%

0%
ICA LM ≥50% ICA CAD <50%, LM <50% ICA CAD ≥50%, LM <50%
Mancini, G.B.J. et al. J Am Coll Cardiol Img. 2021;14(7):1384–93.

There was 97.1% concordance for the absence of LM disease and 92.2% concordance for the designation of at least single-vessel coronary
disease without LM. There were 50 cases of the 1,728 patients in total (2.9%) who had ICA evidence of LM disease $50% DS despite CCTA
reports of <50% DS. CCTA ¼ cardiac computed tomographic angiography; ICA ¼ invasive coronary angiography; LM ¼ left main. The 1,728
denominator comes from 1,757 patients eligible for the overall study minus 21 patients with nonevaluable CCTAs and 8 with missing ICA
information for $50% CAD.

misregistration, adequacy of field of view, calcium York University Grossman School of Medicine (the
affecting ease of segmental analysis or difficult to clinical coordinating center) and by the institutional
assess stents, among other factors. Each scan was review board and ethics committee at each partici-
also graded as either excellent, good, fair, or poor. pating site (Supplemental Appendix).
All CCTA segmental assessments were determined Statistical analysis focused on estimating the
by consensus of at least 2 independent readers, and probability that patients who were classified as
cases of LM disease were also reviewed and finalized having significant CAD and no significant LM disease
by a third reader. Readers assessed 17 segments (14), on CCTA would be classified as the same by ICA. We
with additional identification of whether lesions also examined the frequency of CCTA and ICA
were in the ostium of the LAD artery or the agreeing on the number of diseased vessels and the
circumflex (CX). The lesions were graded categori- presence or absence of stenosis in specific vessels.
cally as 0%, 1% to 24%, 25% to 49%, 50% to 69%, The study’s main concordance measures were the
and 70% or greater. Quantitative coronary angiog- percent of participants with CCTA-defined CAD $50%
raphy was performed as previously described (15). and no LM disease $50% whose ICA result was
All patients provided informed consent for partici- concordant in the sense of having ICA-defined
pation in the ISCHEMIA trial. The protocol was CAD $50% and no ICA LM disease $50%, the
approved by the institutional review board at New percent who were discordant because of ICA
JACC: CARDIOVASCULAR IMAGING, VOL. 14, NO. 7, 2021 Mancini et al. 1387
JULY 2021:1384–93 CCTA in the ISCHEMIA Trial

CAD <50%, and the percent who were discordant

F I G U R E 1 ICA Detection of LM $ 50% DS as a Function of CCTA Classification
because of ICA LM disease $50%. Rates of discor-
dance due to ICA LM disease $50% were estimated CCTA Classification
overall and across subgroups based on CCTA-defined LM = 0 0.5%
n = 376 (0.1%, 1.9%)
number of diseased vessels and degree of stenosis in
specific vessels. All concordance measures in this
LM = 1-24% 1.8%
study were conditional on having CCTA-defined n = 921 (1.1%, 2.9%)

CAD $50% and no LM disease $50%. Patients who

Ostial CX ≥50% 5.4%
failed screening because of CCTA-defined LM n = 149 (3.3%, 6.5%)

disease $50% or no CAD $50% did not receive study

LM = 25-49% 7.0%
ICAs in ISCHEMIA and were therefore excluded. n = 442 (4.8%, 9.8%)

Traditional accuracy measures treating ICA as the

Ostial LAD ≥50% 8.9%
reference standard could not be calculated because of n = 271 (5.8%, 12.9%)

missing ICA results for the patients who failed

0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%
screening. The exclusion of such patients prevented LM ≥50% by ICA
us from estimating sensitivity and specificity but did
not invalidate estimation of concordance probabili-
The CCTA classification of LM (0%, 1% to 24% and 25% to 49% DS) or the classification
ties as defined here for patients meeting inclusion of ostial CX or ostial LAD is shown on the left, with the number of individual reports. The
and exclusion criteria for ISCHEMIA. length of the horizontal bars represents the percentage of cases (with 95% exact
binomial confidence intervals) having LM $50% DS by ICA for each of the CCTA clas-
RESULTS sifications. Note that 18 CCTAs were not interpretable for the presence of LM $50%
DS, leaving a sample size of 1,739. CCTA ¼ cardiac computed tomographic angiography;
CX ¼ circumflex; DS ¼ diameter stenosis; ICA ¼ invasive coronary angiography; LAD ¼ left
The median (25th, 75th percentile) time between
anterior descending; LM ¼ left main. Note that 18 CCTAs were not interpretable for the
CCTA and ICA was 29 (18, 46) days with a mean and
presence of LM $50% DS, leaving a sample size of 1,739.
standard deviation of 35.7  27.1 days.
PER-PATIENT ANALYSIS. ICA and CCTA were
concordant for the identification of at least single-
vessel CAD and absence of LM $50% in 92.2% of ASSESSMENT OF NUMBER OF DISEASED VESSELS.
cases (95% confidence interval [CI]: 90.9% to The agreement between CCTA and ICA for the desig-
93.5%). In 4.9% (95% CI: 3.9% to 6.0%) of cases, nation of presence of single-, double-, or triple-vessel
ICA did not confirm presence of CAD $50% DS. And disease without LM disease was 54.5% (Figure 3).
in 2.9% (95% CI: 2.2 to 3.8%) of cases, CCTA missed Overestimation of disease by CCTA occurred in 25.3%
ICA identified LM $50% DS (Central Illustration). of cases and underestimation—including underesti-
Thus, ICA and CCTA were concordant in 97.1% mation of LM disease—in 20.2% of cases.
(1,678 of 1,728) for the absence of significant LM ASSESSMENT OF LOCATION OF DISEASE. Per-vessel
disease $50%. analysis. Figure 4 demonstrates concordance between
ASSESSMENT OF LM DISEASE DISCORDANCE. ICA and CCTA for the presence of $50% DS in the
When CCTA identified LM ¼ 0% DS (n ¼ 376), it was LAD, CX, and RCA (including their major branches;
rare (n ¼ 2, 0.5%) for ICA to report LM $50% DS. Of see Definitions in Supplemental Appendix) of 84.1%,
921 patients with CCTA findings of 1% to 24% DS, 1.8% 81.3%, and 83.3% of patients, respectively.
(n ¼ 17) had LM $50% DS on ICA. Of 442 patients with Conversely, concordance for <50% DS was 56.9%,
CCTA reports of LM 25% to 49% DS, 7% (n ¼ 31) had 70.2%, and 74.3%, respectively.
LM disease $50% on ICA. The average percent of DS Assessment of LAD artery segments. Because of the
of the significant LM stenosis detected by ICA was importance attached to LAD artery disease, and
62  3%, 62  14% and 64  10%, respectively, for because of variable visual cues for segmentation be-
these 3 CCTA categories of LM assessment (0, 1% to tween ICA and CCTA (e.g., septals are infrequently
24%, 25% to 49%). When CCTA identified $50% DS in used to define proximal LAD on CCTA, whereas this is
the ostial CX (n ¼ 149) or ostial LAD (n ¼ 271), more feasible with ICA), we analyzed concordance
but <50% DS in the LM, there were 5.4% (n ¼ 8) and between CCTA and ICA for the LAD artery as a whole,
8.9% (n ¼ 24) with ICA showing $50% DS in the LM, for an “extended” proximal LAD or mid-LAD
respectively (Figure 1). Examples of discrepancies in 3 segment, and for the isolated proximal LAD artery
patients are shown in Figure 2. (Figure 5). This shows an expected decrease in
1388 Mancini et al. JACC: CARDIOVASCULAR IMAGING, VOL. 14, NO. 7, 2021

CCTA in the ISCHEMIA Trial JULY 2021:1384–93

F I G U R E 2 Left Main ICA and CCTA Discordance

A B C

D E F

G H I

Each row represents 1 of 3 patients with ICA detection of LM $50% and CCTA report of LM <50% DS. A, B, D, E, G, and H are 2 ICA frames showing a potential stenosis
in the LM $50% for each patient. A single corresponding CCTA reconstruction for each patient is shown at the end of each row in C, F, and I. Abbreviations as in
Figure 1.

concordance for presence of significant stenosis as (56.9%) to the proximal or mid-LAD vessel (70.8%)
one progresses from considering the entire LAD and the proximal LAD vessel itself (79.3%).
vessel (84.1%), the proximal or mid-LAD vessel SENSITIVITY ANALYSES. Other secondary and
(76.7%), and only the proximal LAD vessel itself sensitivity analyses for the main per-patient results
(54.0%). Conversely, exclusion of significant disease are provided in the Supplemental Appendix and
improved progressively from LAD vessel overall showed similar results. In particular, use of a $70%
JACC: CARDIOVASCULAR IMAGING, VOL. 14, NO. 7, 2021 Mancini et al. 1389
JULY 2021:1384–93 CCTA in the ISCHEMIA Trial

DS threshold applied to CCTA and a $50% DS

F I G U R E 3 Concordance Matrix Between CCTA and ICA for Detection of
threshold applied to ICA did not improve per-patient Single-, Double-, or Triple-Vessel Disease
performance. In addition, a focus on the subset
(n ¼ 830 of 1,757, 47%) with CCTA quality rated as Assessment of Number of Diseased Vessels
good or excellent did not change performance (Sup- ICA
plemental Tables 5 and 6). CCTA
0VD 1VD 2VD 3VD LM

DISCUSSION
3VD 0.8% 2.9% 12.3% 28.2% 1.7%
(10) (37) (159) (366) (22)
In this cohort with a very high a priori likelihood of
obstructive CAD, CCTA was highly concordant with
2VD 1.4% 5.6% 12.5% 9.4% 1.0%
ICA in patients randomized to the invasive arm of (18) (72) (162) (122) (13)
ISCHEMIA (97.1% for excluding LM disease $50% and
92.2% for identifying patients with at least 1-vessel 2.5% 13.7% 5.8% 1.9% 0.4%
1VD
CAD and no LM disease). Concordance for burden of (32) (178) (75) (25) (5)
disease based on numbers of diseased vessels (1, 2, or
3 and without LM disease) was modest (54.5%). Total n = 1,296
Overestimation of disease burden by CCTA occurred Concordance 54.5% Overestimation 25.3% Underestimation 20.2%
in 1 in 4 and underestimation in 1 in 5 cases compared
with ICA. Thus, CCTA successfully ensured that Shading in green identifies concordance between the 2 imaging methods. Shading in red
randomization of patients with noninvasive evidence identifies underestimation, and, in yellow, overestimation. Designation of LM <50% by
of moderate-to-severe ischemia would be limited as CCTA but $50% by ICA was considered an underestimation by CCTA. The n per cell is

much as possible to those without LM disease and provided and expressed as a percent of the total N ¼ 1,296. The overall concordance
was 54.5% (95% confidence interval: 51.7% to 57.2%). Note that, in 426 CCTAs and 35
avoided as much as possible those patients without
ICAs, certain designated segments could not be interpreted for the presence of $ 50%
significant CAD. It also enabled avoidance of the pit- DS, leaving a sample size of 1,296 for which the number of diseased vessels is known.
falls of study enrolment at the time of ICA when Abbreviations as in Figure 1.
biases might have compromised randomization to
OMT of patients with higher burdens of disease and
disease in locations such as the ostial or proximal LAD
distal LM stenosis and an ostial LAD artery or ostial
artery.
CX stenosis compared with CCTA (17). The sample
STUDY LIMITATIONS. The analysis of LM disease in images in Figure 2 may be taken to support this hy-
this report is limited by necessity: first, to only those pothesis. Regardless, exclusion of LM disease by
patients thought to be eligible for randomization after CCTA in 97.1% of cases proceeding to ICA was
excluding LM disease through use of CCTA and, excellent.
second, to those randomized to the invasive strategy. Although concordance between CCTA and ICA for
It has already been reported that of the 5,757 patients identifying severe disease in specific major vessels
undergoing a study CCTA, 434 were not eligible for (LAD vs. CX vs. RCA) was high (81.3% to 84.1%),
randomization, based upon a CCTA core laboratory excluding disease in each specific vessel was only
report of LM disease $50% (7.5% of subjects) (16). But modest (56.9% to 74.3%) (Figure 4). Such analyses are
ICA core laboratory corroboration is not available in affected by the binary categorization based upon the
those patients. It is conceivable that the percentage of 50% threshold. We did not see material differences
patients having ICA-confirmed LM disease might with secondary analyses using only good to excellent
be <7.5%, based on the potential for overestimation scans or using a higher stenosis threshold ($70%).
by CCTA and the imperative for CCTA readers to Our findings are in line with the findings of the
maximize patient safety, possibly leading to “over- recent SYNTAX (TAXUS Drug-Eluting Stent Versus
calling” of LM disease. But of those who proceeded to Coronary Artery Bypass Surgery for the Treatment of
ICA, detailed analysis of the small number of discor- Narrowed Arteries) II trial, in which participants un-
dant cases allowed us to identify some CCTA findings derwent CCTA and ICA with SYNTAX scores calcu-
associated with a higher likelihood of finding lated from both modalities (18). The SYNTAX scores
LM $50% DS on ICA, most notably CCTA reporting of from CCTA overestimated the disease burden
ostial LAD vessel or CX DS $50% or LM 25% to 49% compared with the invasive gold standard. Thus,
DS. One may argue that ICA, a 2-dimensional imaging CCTA is a good tool to help enrich the population
modality, may not be an appropriate gold standard referred for ICA compared with stress testing alone
for determining the precise difference between a but should not be considered adequate for precise
1390 Mancini et al. JACC: CARDIOVASCULAR IMAGING, VOL. 14, NO. 7, 2021

CCTA in the ISCHEMIA Trial JULY 2021:1384–93

F I G U R E 4 Concordance Between CCTA and ICA for Location of Significant CAD in Major Vessels

100%

90% 84.1% 83.3%

(82.0%, 85.9%) 81.3% (80.8%, 85.5%)
(78.8%, 83.6%)
80% 74.3%
1,186 ICA / 865 ICA / (69.7%, 78.4%)
1,411 CCTA 1039 CCTA 70.2%
858 ICA /
(65.2%, 74.9%)
70% 1,056 CCTA
303 ICA /
250 ICA / 408 CCTA
Concordance with ICA

56.9%
60% 356 CCTA
(49.8%, 63.8%)

50% 116 ICA /

204 CCTA

40%

30%

20%

10%

0%
LAD CX RCA LAD CX RCA
≥50% <50%
CCTA

The x-axis shows CCTA reports of disease $50% and <50% DS in the LAD, the CX, or the RCA. The y-axis shows the percent concordance
with ICA as derived from the numbers shown within the bars (numerator ¼ cases that agree with ICA; denominator ¼ number of cases with
the CCTA result shown); 95% confidence intervals are provided in parentheses. CAD ¼ coronary artery disease; RCA ¼ right coronary artery;
other abbreviations as in Figure 1.

planning of any specific revascularization strategy of moderate or severe stress-induced ischemia and
without ICA corroboration. Others have also high- history of angina in 89%. This feature, however, is
lighted these types of discrepancies (19–22). also a unique strength of the analyses because the
Elements of this study may limit the application to majority of previous studies comparing the perfor-
routine clinical CCTA or to more broad populations, mance of CCTA with ICA have been undertaken in
particularly those with much lower a priori likelihood populations with a much lower risk and extent of
of CAD. The analyses were based upon core- underlying CAD, accounting for the well-known high
laboratory assessment and were all determined by negative predictive value of CCTA (4–10). It is worth
consensus (23). This element of quality control does emphasizing that calcium remains an impediment to
not occur in routine practice. The analyses are the facile reading of CCTA scans, and calcium suffi-
based solely on those patients eligible by CCTA and cient to potentially impair segmental analysis was
subsequently undergoing core-laboratory analysis of present in 31% of scans (Supplemental Table 5)
ICA. We do not have ICA comparisons for those (24,25). Accordingly, only 75% of CCTA studies could
patients deemed by CCTA to have either no signifi- be included in the analysis of number of vessels
cant CAD or LM $50% DS, thereby precluding valid diseased because key segments were not evaluable
calculations of sensitivity, specificity, and positive for $50% DS in the remaining 25%, because of calci-
and negative predictive values. The highly selected fication, motion, or other artifacts. Stents (noted in
cohort had a very high probability of significant un- 17% of patients in this study) provide similar chal-
derlying CAD based on selection by previous evidence lenges to CCTA reading. All readings of calcified or
JACC: CARDIOVASCULAR IMAGING, VOL. 14, NO. 7, 2021 Mancini et al. 1391
JULY 2021:1384–93 CCTA in the ISCHEMIA Trial

F I G U R E 5 Assessment of LAD Segments

100%

90% 84.1%
(82.0%, 85.9%)
79.3%
76.7% (76.5%, 81.9%)
80% (74.2%, 79.0%)
70.8%
1,186 ICA / (66.0%, 75.2%)
1,411 CCTA 735 ICA /
70%
970 ICA / 927 CCTA
1,265 CCTA
56.9% 276 ICA /
60%
Concordance with ICA

54.0% (49.8%, 63.8%) 390 CCTA

(50.4%, 57.6%)

50% 116 ICA /

412 ICA / 204 CCTA
763 CCTA
40%

30%

20%

10%

0%
LAD p / mLAD pLAD LAD p / mLAD pLAD
≥50% <50%
CCTA

The x-axis shows CCTA reports of disease $50% and <50% DS in the entire LAD vessel, in an extended segment consisting of both the
proximal (p) and mid (m) LAD, and solely in the pLAD. The y-axis shows the concordance with ICA in percent as derived from the numbers
shown within the bars (numerator ¼ cases that agree with ICA; denominator ¼ number of cases with the CCTA result shown); 95% confidence
intervals are provided in parentheses. m ¼ mid; p ¼ proximal; other abbreviations as in Figure 1.

stented segments were performed using the principle (27). It has been shown in the CONSERVE (Coronary
of “best effort;” use of appropriate views, re- Computed Tomographic Angiography for Selective
constructions, filters, windows, and levels by the Cardiac Catheterization) trial that use of CCTA in pa-
individual readers; and then reaffirmed by the tients with a lower probability of CAD compared with
consensus process. Although exact concordance for the ISCHEMIA trial helps to avoid unnecessary ICA,
segments within major vessels was suboptimal, this diminish ICA costs, and improve overall use (28).
segmentation problem is recognized from studies After stress testing and before randomization, 21% of
attempting to correlate lesion-specific fractional flow subjects undergoing CCTA were excluded because of
reserve measured invasively with values derived the absence of significant CAD (16). Thus, the total
from CCTA (26). Finally, CCTA-derived fractional flow experience of using CCTA in this trial and the unique,
reserve was not calculated so correlation with frac- high-risk cohort suggest that an expanded role for
tional flow reserve performed in some patients at the CCTA before ICA may be warranted even when the
time of ICA was not possible. ischemic burden is moderate or severe (11).
The utility of CCTA in this research protocol of
high-risk patients may have important implications CONCLUSIONS
for clinical practice and for expanding appropriate-
use criteria for CCTA in patients with moderate or We demonstrated that, in patients first shown to have
severe ischemia, despite the limitations noted here moderate or severe ischemia on stress testing, CCTA
1392 Mancini et al. JACC: CARDIOVASCULAR IMAGING, VOL. 14, NO. 7, 2021

CCTA in the ISCHEMIA Trial JULY 2021:1384–93

was an effective method to identify patients with PERSPECTIVES

significant CAD and without LM disease as deter-
mined by ICA.
COMPETENCY IN MEDICAL KNOWLEDGE: Car-
diac computed tomographic angiography was used in
FUNDING SUPPORT AND AUTHOR DISCLOSURES
a novel fashion to determine coronary anatomy eligi-
This project was funded by National Institutes of Health (NIH) bility for a randomized trial that compared a conser-
grants U01HL105907, U01HL105462, and U01HL105561 vative strategy with an invasive strategy for the
(NCT01471522). Dr. Mancini has received grants from National management of patients with chronic coronary artery
Heart, Lung and Blood Institute. Dr. Leipsic has served as
disease. Patients had very high a priori likelihood of
consultant and has reported stock options in HeartFlow and CIR-
CLE CVI; has received a research grant from GE Healthcare. Dr. underlying coronary disease by virtue of moderate or
Budoff has received grant support from General Electric. Dr. Hague severe ischemia documented by stress testing. Sub-
has received grants from National Heart, Lung and Blood Institute.
sequent evaluation by cardiac computed tomographic
Dr. Min has received grants from National Heart, Lung, and Blood
angiography before invasive angiography served well
Institute, support from Cleerly Inc., grants and other support from
GE Healthcare, and support from Arineta. Ms. Stevens has received to ensure presence of angiographically significant
grants from National Heart, Lung and Blood Institute. Dr. Reynolds coronary artery disease and to exclude left main dis-
has received grants from National Heart, Lung and Blood Institute
ease and nonobstructive disease.
and nonfinancial support from Abbott Vascular, Siemens, and
BioTelemetry. Drs. O’Brien, Shaw, Manjunath, Mavromatis, Dem-
kow, Lopez-Sendon, Chernyavskiy, Gosselin,. Schuelenz, Devlin, COMPETENCY IN PATIENT CARE: This analysis
and Chauhan have received grants from National Heart, Lung and demonstrates that patients suspected of coronary ar-
Blood Institute. Dr. Bangalore has received grants from National tery disease with stress tests showing moderate or
Heart, Lung and Blood Institute; and has served on advisory
severe ischemia can be effectively stratified further
boards for Abbott Vascular, Pfizer, Amgen, Biotronik, Meril, Reata
Pharmaceuticals, and Abbott Vascular. Dr. Hochman has served as using cardiac computed tomography to assess left
Principal Investigator for the International Study of Comparative main stenosis and to ensure that when
Health Effectiveness with Medical and Invasive Approaches intervention may be warranted, procession to invasive
(ISCHEMIA) trial, for which—in addition to support by a National
angiography is reserved for patients with verified
Heart, Lung, and Blood Institute grant—devices and medications
were provided by Abbott Vascular, Medtronic Inc., St. Jude Medical anatomic burden of disease.
Inc., Volcano Corporation, Arbor Pharmaceuticals LLC, AstraZeneca,
Merck Sharp and Dohme Corp., Omron Healthcare Inc.; and has TRANSLATIONAL OUTLOOK: The results of this
received financial support from Arbor Pharmaceuticals LLC and
study potentially expand the appropriate use of
AstraZeneca. Dr. Maron has received grants from National Heart,
Lung, and Blood Institute. computed tomographic angiography before referral to
invasive angiography when there is evidence of
moderate or severe ischemia. This sequence more
ADDRESS FOR CORRESPONDENCE: Dr. G.B. John
precisely identifies patients who truly have anatomi-
Mancini, University of British Columbia, Diamond
cally significant underlying coronary atheroma when
Centre, Room 9111, 2775 Laurel Street, Vancouver,
intervention is being considered.
British Columbia V57 1M9, Canada. E-mail: mancini@
mail.ubc.ca.

REFERENCES

1. Maron DJ, Hochman JS, Reynolds HR, et al. Initial without known coronary artery disease: results the non-invasive evaluation of significant coronary
invasive or conservative strategy for stable coro- from the prospective multicenter ACCURACY artery disease. Eur Heart J 2010;31:1908–15.
nary disease. N Engl J Med 2020;382:1395–407. (Assessment by Coronary Computed Tomographic
8. Abdulla J, Abildstrom SZ, Gotzsche O,
Angiography of Individuals Undergoing Invasive
2. Maron DJ, Hochman JS, O’Brien SM, et al. In- Christensen E, Kober L, Torp-Pedersen C. 64-
Coronary Angiography) trial. J Am Coll Cardiol
ternational Study of Comparative Health Effec- multislice detector computed tomography coro-
2008;52:1724–32.
tiveness with Medical and Invasive Approaches nary angiography as potential alternative to con-
(ISCHEMIA) trial: rationale and design. Am Heart J 5. Meijboom WB, Meijs MFL, Schuijf JD, et al. ventional coronary angiography: a systematic
2018;201:124–35. Diagnostic accuracy of 64-slice computed to- review and meta-analysis. Eur Heart J 2007;28:
mography coronary angiography: a prospective, 3042–50.
3. Maron DJ, Stone GW, Berman DS, et al. Is car-
multicenter, multivendor study. J Am Coll Cardiol
diac catheterization necessary before initial man- 9. Ollendorf DA, Kuba M, Pearson SD. The diag-
2008;52:2135–44.
agement of patients with stable ischemic heart nostic performance of multi-slice coronary
disease? Results from a Web-based survey of 6. Miller JM, Rochitte CE, Dewey M, et al. Diag- computed tomographic angiography: a systematic
cardiologists. Am Heart J 2011;162:1034–43.e13. nostic performance of coronary angiography by review. J Gen Intern Med 2011;26:307–16.
64-row CT. N Engl J Med 2008;359:2324–36.
4. Budoff MJ, Dowe D, Jollis JG, et al. Diagnostic 10. Paech DC, Weston AR. A systematic review of
performance of 64-multidetector row coronary 7. de Graaf FR, Schuijf JD, van Velzen JE, et al. the clinical effectiveness of 64-slice or higher
computed tomographic angiography for evalua- Diagnostic accuracy of 320-row multidetector computed tomography angiography as an alter-
tion of coronary artery stenosis in individuals computed tomography coronary angiography in native to invasive coronary angiography in the
JACC: CARDIOVASCULAR IMAGING, VOL. 14, NO. 7, 2021 Mancini et al. 1393
JULY 2021:1384–93 CCTA in the ISCHEMIA Trial

investigation of suspected coronary artery disease. ticipants in the ISCHEMIA randomized clinical computed tomographic angiography? A compari-
BMC Cardiovasc Disord 2011;11:32. trial baseline characteristics of participants in son of alternate reader combinations. J Cardiovasc
the ISCHEMIA study. JAMA Cardiol 2019;4: Comput Tomogr 2010;4:384–90.
11. Taylor AJ, Cerqueira M, Hodgson JM, et al.
273–86.
ACCF/SCCT/ACR/AHA/ASE/ASNC/NASCI/SCAI/ 24. Qi L, Tang LJ, Xu Y, et al. The diagnostic
SCMR 2010 appropriate use criteria for cardiac 17. Kocka V, Theriault-Lauzier P, Xiong T-Y, et al. performance of coronary CT angiography for the
computed tomography: a report of the American Optimal fluoroscopic projections of coronary ostia assessment of coronary stenosis in calcified pla-
College of Cardiology Foundation Appropriate Use and bifurcations defined by computed tomography que. PLoS One 2016;11:e0154852.
Criteria Task Force, the Society of Cardiovascular coronary angiography. J Am Coll Cardiol Intv
25. Muzzarelli S, Suerder D, Murzilli R, et al. Pre-
Computed Tomography, the American College of 2020;13:2560–70.
dictors of disagreement between prospectively
Radiology, the American Heart Association, the 18. Collet C, Miyazaki Y, Ryan N, et al. Fractional ECG-triggered dual-source coronary computed
American Society of Echocardiography, the flow reserve derived from computed tomographic tomography angiography and conventional
American Society of Nuclear Cardiology, the North angiography in patients with multivessel CAD. coronary angiography. Eur J Radiol 2016;85:
American Society for Cardiovascular Imaging, the J Am Coll Cardiol 2018;71:2756–69. 1138–46.
Society for Cardiovascular Angiography and In-
terventions, and the Society for Cardiovascular 19. Arbab-Zadeh A, Miller JM, Rochitte CE, et al. 26. Min JK, Leipsic J, Pencina MJ, et al. Diag-
Magnetic Resonance. J Am Coll Cardiol 2010;56: Diagnostic accuracy of computed tomography nostic accuracy of fractional flow reserve from
1864–94. coronary angiography according to pre-test anatomic CT angiography. JAMA 2012;308:
probability of coronary artery disease and 1237–45.
12. National Institute for Health and Clinical severity of coronary arterial calcification: the
Excellence. Chest pain of recent onset: assessment 27. Meinel FG, Schoepf UJ, Townsend JC,
CORE-64 (Coronary Artery Evaluation Using 64-
and diagnosis of recent onset chest pain or et al. Diagnostic yield and accuracy of coro-
Row Multidetector Computed Tomography Angi-
discomfort of suspected cardiac origin (update). ography) international multicenter study. J Am nary CT angiography after abnormal nuclear
CG95. London, UK: National Institute for Health Coll Cardiol 2012;59:379–87. myocardial perfusion imaging. Sci Rep 2018;
and Clinical Excellence; 2016. 8:9228.
20. Kerl JM, Schoepf UJ, Bauer RW, et al. 64-slice
13. Moss AJ, Williams MC, Newby DE, Nicol ED. multidetector-row computed tomography in the 28. Chang HJ, Lin FY, Gebow D, et al. Selective
The updated NICE guidelines: cardiac CT as the diagnosis of coronary artery disease: interobserver referral using CCTA versus direct referral for in-
first-line test for coronary artery disease. Curr agreement among radiologists with varied levels dividuals referred to invasive coronary angiog-
Cardiovasc Imaging Rep 2017;10:15. of experience on a per-patient and per-segment raphy for suspected CAD: a randomized,
basis. J Thorac Imaging 2012;27:29–35. controlled, open-label trial. J Am Coll Cardiol Img
14. Leipsic J, Abbara S, Achenbach S, et al. SCCT
2019;12:1303–12.
guidelines for the interpretation and reporting of 21. Song YB, Arbab-Zadeh A, Matheson MB, et al.
coronary CT angiography: a report of the Society Contemporary discrepancies of stenosis assess-
of Cardiovascular Computed Tomography Guide- ment by computed tomography and invasive cor-
lines Committee. J Cardiovasc Comput Tomogr KEY WORDS cardiac computed
onary angiography. Circ Cardiovasc Imaging 2019;
2014;8:342–58. tomographic angiography, cardiac
12:e007720.
catheterization, invasive coronary
15. Chuang ML, Almonacid A, Popma JJ. Qualita- 22. Knaapen PMDP. Computed tomography to angiography, ischemia, left main coronary
tive and quantitative coronary angiography. In: replace invasive coronary angiography? Close, but artery disease
Topol EJ, Teirstein PS, editors. Textbook of not close enough. Circ Cardiovasc Imaging 2019;
Interventional Cardiology. 8th ed. Philadelphia, 12:e008710.
PA: Elsevier, 2020; p 999–1017.
23. LaBounty TM, Leipsic J, Srichai MB, et al. What A PP END IX For supplemental figures and
16. Hochman JS, Reynolds HR, Bangalore S, et al. is the optimal number of readers needed to ach- tables, please see the online version of this
Baseline characteristics and risk profiles of par- ieve high diagnostic accuracy in coronary paper.