Accepted Manuscript
Comparison of the Diagnostic Performance of Coronary Computed
Tomography Angiography Derived Fractional Flow Reserve in
Patients With Versus Without Diabetes Mellitus (From the MACHINE
Consortium)
Fay M.A. Nous MD , Adriaan Coenen MD ,
Eric Boersma MSc, PhD, FESC , Young-Hak Kim MD, PhD ,
Mariusz B.P. Kruk MD, PhD , Christian Tesche MD ,
Jakob de Geer MD, PhD , Dong Hyun Yang MD, PhD ,
Cezary Kepka MD, PhD , U. Joseph Schoepf MD ,
Anders Persson MD, PhD , Akira Kurata MD, PhD ,
Ricardo P.J. Budde MD, PhD , Koen Nieman MD, PhD
PII:
DOI:
Reference:
S0002-9149(18)32105-2
https://doi.org/10.1016/j.amjcard.2018.11.024
AJC 23636
To appear in:
The American Journal of Cardiology
Received date:
Revised date:
19 July 2018
28 October 2018
Please
cite
this
article
as:
Fay M.A. Nous MD ,
Adriaan Coenen MD ,
Eric Boersma MSc, PhD, FESC ,
Young-Hak Kim MD, PhD ,
Mariusz B.P. Kruk MD, PhD ,
Christian Tesche MD ,
Jakob de Geer MD, PhD ,
Dong Hyun Yang MD, PhD ,
Cezary Kepka MD, PhD ,
U. Joseph Schoepf MD ,
Anders Persson MD, PhD ,
Akira Kurata MD, PhD , Ricardo P.J. Budde MD, PhD , Koen Nieman MD, PhD , Comparison of
the Diagnostic Performance of Coronary Computed Tomography Angiography Derived Fractional
Flow Reserve in Patients With Versus Without Diabetes Mellitus (From the MACHINE Consortium),
The American Journal of Cardiology (2018), doi: https://doi.org/10.1016/j.amjcard.2018.11.024
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and
all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Comparison of the Diagnostic Performance of Coronary Computed Tomography Angiography
Derived Fractional Flow Reserve in Patients With Versus Without Diabetes Mellitus (From the
MACHINE Consortium)
IP
T
Fay M.A. Nous, MDa,b*, Adriaan Coenen, MDa,b, Eric Boersma, MSc, PhD, FESCb, Young-Hak Kim,
MD, PhDc, Mariusz B.P. Kruk, MD, PhDd, Christian Tesche, MDe, Jakob de Geer, MD, PhDf, Dong
US
CR
Hyun Yang, MD, PhDg, Cezary Kepka, MD, PhDd, U. Joseph Schoepf, MDe, Anders Persson, MD, PhDf,
Akira Kurata, MD, PhDa,h, Ricardo P.J. Budde, MD, PhDa,b, Koen Nieman, MD, PhDa,b,i
a Department
of Radiology & Nuclear Medicine, Erasmus University Medical Center, Doctor
Molewaterplein 40, 3015 GD, Rotterdam, The Netherlands.
Department of Cardiology, Erasmus University Medical Center, Doctor Molewaterplein 40, 3015 GD,
AN
b
Rotterdam, The Netherlands.
Department of Cardiology, Asan Medical Center, University of Ulsan College of Medicine, 88,
M
c
Olympic-ro 43-Gil, Songpa-gu, Seoul, 138-736, South Korea.
Coronary Disease and Structural Heart Diseases Department, Institute of Cardiology, ul. Alpejska 42,
PT
ED
d
04-628 Warsaw, Poland.
e
Department of Radiology and Radiological Science, Heart & Vascular Center, Medical University of
f
CE
South Carolina, Ashley River Tower, MSC 226, 25 Courtenay Drive, Charleston, SC 29425, USA.
Department of Radiology and Department of Medical and Health Sciences, Center for Medical Image
AC
Science and Visualization, CMIV, Linköping University, 58185 Linköping, Sweden.
g
Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of
Ulsan College of Medicine, 88, Olympic-ro 43-Gil, Songpa-gu, Seoul, 138-736, South Korea.
h
Department of Radiology, Ehime University Graduate School of Medicine, Shitsukawa, Toon, Ehime,
791-0295, Japan.
1
ACCEPTED MANUSCRIPT
i
Stanford University School of Medicine, Cardiovascular Institute, 300 Pasteur Drive, Stanford, CA
94305, USA.
*Corresponding author: Tel: +31 10 7032055; E-mail: f.nous@erasmusmc.nl; Postal address: Erasmus
IP
T
MC, Postbus 2040, 3000 CA, Rotterdam, room: Ne507k
US
CR
Running head:
Functional cardiac CT in diabetic patients
Disclosures:
AN
The authors of this manuscript declare relationships with the following companies: U. Joseph Schoepf:
Institutional research support to Medical University of South Carolina from Astellas, Bayer Healthcare,
M
GE Healthcare, Siemens Healthineers and speaker / consultancy fees from Bayer Healthcare, GE
Healthcare, HeartFlow Inc., Guerbet, and Siemens. Koen Nieman: Institutional research support to the
PT
ED
Erasmus MC from Siemens Healthineers, HeartFlow, GE Healthcare, Bayer Healthcare. Other authors
declare that they have no competing interests.
Funding:
CE
Fay M.A. Nous, Adriaan Coenen, Ricardo P.J. Budde and Koen Nieman were supported by grants from
AC
the Dutch Heart Foundation [NHS 2014T061 and NHS 2013T071].
2
ACCEPTED MANUSCRIPT
Abstract
Coronary computed tomography angiography derived fractional flow reserve (CT-FFR) is a noninvasive application to evaluate the hemodynamic impact of coronary artery disease by simulating
invasively measured FFR based on CT data. CT-FFR is based on the assumption of a normal coronary
IP
T
microvascular response. We assessed the diagnostic performance of a machine-learning based application
for on-site computation of CT-FFR in patients with and without diabetes mellitus. The population
US
CR
included 75 diabetic and 276 non-diabetic patients, who were enrolled in the MACHINE consortium. The
overall diagnostic performances of coronary CT angiography alone and in combination with CT-FFR
were analyzed in 110 vessels of the diabetic group and in 415 vessels of the non-diabetic group, with
direct invasive FFR comparison. Per-vessel discrimination of lesion-specific ischemia by CT-FFR was
AN
assessed by the area under receiver operating characteristic curves. The overall diagnostic accuracy of
CT-FFR in diabetic patients was 83% and 75% in non-diabetic patients (p=0.088), improved over the
M
diagnostic accuracy of coronary CT angiography, which was 58% and 65% (p=0.223), respectively.
Furthermore, the diagnostic accuracy of CT-FFR was similar between diabetic and non-diabetic patients
PT
ED
per stratified CT-FFR group (CT-FFR <0.6, 0.6-0.69, 0.7-0.79, 0.8-0.89, ≥0.9). The area under curves for
diabetic and non-diabetic patients were comparable, 0.88 and 0.82 (p=0.113), respectively. In conclusion,
on-site machine-learning CT-FFR analysis improved the diagnostic performance of coronary CT
CE
angiography and was able to accurately discriminate lesion-specific ischemia in both diabetic and nondiabetic patients suspected for coronary artery disease.
AC
Keywords: Coronary artery disease; diabetes mellitus; computed tomography; fractional flow reserve
Introduction
Diabetic patients are approximately two times more likely to develop coronary artery disease (CAD) than
non-diabetic patients, while the evaluation and management of CAD in these patients has proven to be
more challenging1-3. Due to chronic hyperglycemia, diabetic patients are at risk of developing a more
3
ACCEPTED MANUSCRIPT
diffuse atherosclerosis, a more profound microvascular dysfunction, an increased vascular resistance, and
a reduced vasodilation capacity4,5. The clinical reference standard to assess CAD is invasive coronary
angiography with invasively measured fractional flow reserve (FFR)6,7. To avoid unnecessary use of
invasive modalities, alternatives such as coronary computed tomography (CT) angiography with
IP
T
simulated FFR (CT-FFR) are under development8. Novel machine-learning based CT-FFR applications,
based on pattern recognition and computational learning, have been recently shown to be as accurate as
US
CR
the more established, on-site performed, computational fluid dynamics based CT-FFR, and furthermore,
they have been able to calculate CT-FFR much faster9,10. All CT-FFR applications, however, are based on
a computational simulation of normal coronary hyperemia, which is not the case in diabetic patients11,12.
This study assessed the diagnostic performance of an on-site machine-learning based CT-FFR application
AN
in diabetic patients compared to non-diabetic patients using invasive FFR as the reference standard.
Methods
M
The study population was part of the MACHINE consortium, which has been described in detail
PT
ED
previously10. In brief, the MACHINE consortium is a collaboration between five institutions in Europe,
USA, and Korea, that investigated the diagnostic performance of machine-learning based CT-FFR
software (clinicaltrials.gov identifier: 02805621). Patients were identified retrospectively at four sites and
prospectively in one site on the availability of coronary CT angiography and invasive FFR measurements.
CE
The studies were conducted in accordance with the Declaration of Helsinki and with research ethics
committee approval at each of the sites.
AC
All coronary CT angiograms were performed on first- and second-generation dual-source CT
scanners (Somatom Definition or Definition Flash; Siemens Healthineers, Forchheim, Germany) using a
retrospectively ECG-gated spiral (78%), prospectively ECG-triggered axial (20%) or prospectively-ECGtriggered high-pitch spiral (2%) scan mode. Patients received sublingual nitroglycerin before the
examination and β-blockers in case of high heart rates. Stenosis severity was classified per segment
4
ACCEPTED MANUSCRIPT
following the Society of Cardiovascular Computed Tomography criteria13: normal: 0% stenosis; minimal:
<25% stenosis; mild: 25-49% stenosis; moderate 50-69% stenosis; severe: 70-99%; and occluded: 100%
by observers with extensive previous experience in cardiac imaging. Stenosis ≥50% was considered
angiographically significant. Coronary CT angiography image quality was evaluated based on a 4-point
IP
T
Likert scale: 1 non-diagnostic; 2 impaired image quality, differentiation of coronary arty wall possible
with reduced confidence; 3 adequate, reduced image quality due to artifacts without limiting coronary
US
CR
artery wall differentiation; 4 excellent, no artifacts present and clear differentiation of the coronary artery
wall.
A machine-learning based CT-FFR software prototype (cFFR version 2.1, Siemens Healthineers,
Forchheim, Germany; not currently commercially available) was used, as previously described8. A 3D
AN
coronary artery tree was semi-automatically created and the left ventricular myocardial mass was
automatically determined based on the coronary CT angiography data. Each point on the coronary artery
M
tree was analyzed and CT-FFR was derived based on a combination of pattern recognition and
PT
ED
computational learning. CT-FFR ≤0.80 was considered hemodynamically significant.
Invasive coronary angiography was performed following local standards and invasive FFR
measurements were either performed for clinical reasons or for research purposes. An FFR pressure wire
was positioned distal to the stenosis of interest and an FFR was measured during hyperemia by
CE
intravenous infusion of adenosine at 140μg/kg/min. FFR ≤0.80 was considered hemodynamically
significant.
AC
Absolute variables are represented as totals and percentages and continuous variables as median
and 25th-75th percentiles. Chi-square, Fisher’s exact and Mann–Whitney U-tests were used to check for
differences in categorical and continuous variables between the diabetic and non-diabetic group. The
correlation between CT-FFR and invasive FFR in diabetic and non-diabetic patients was assessed with a
Bland-Altman plot. The diagnostic performances of coronary CT angiography and CT-FFR in diabetic
5
ACCEPTED MANUSCRIPT
and non-diabetic patients were analyzed with direct comparison of invasive FFR and were reported on a
per-vessel basis as sensitivity, specificity, positive predictive value, negative predictive value, and
accuracy with 95% limits of agreement. A multi-variable logistic regression analysis was performed to
determine whether the presence of diabetes mellitus affected the diagnostic performance after adjusting
IP
T
for known differences in patient characteristics. Additionally, the diagnostic accuracy of CT-FFR was
reported per stratified CT-FFR group (CT-FFR <0.6, 0.6-0.69, 0.7-0.79, 0.8-0.89, ≥0.9) with 95% limits
of agreement. Per-vessel discrimination of lesion-specific ischemia by CT-FFR was assessed by the area
US
CR
under curve with 95% limits of agreement using FFR ≤0.80 as the reference standard. C-statistics were
calculated for CT-FFR in diabetic and non-diabetic patients, and compared by using the method of
DeLong et al.14. Statistical analyses were performed using SPSS (version 21, IBM Corp, Armonk NY,
AN
USA). MedCalc (version 13.0; MedCalc Software, Ostend, Belgium) was used to compare the area under
curves.
M
Results
PT
ED
The study population consisted of 75 diabetic patients and 276 non-diabetic patients. Direct
invasive FFR comparison with CT-FFR was available in 110 vessels of diabetic patients and 415 vessels
of non-diabetic patients. A detailed overview of the patient characteristics is presented in Table 1. The
median age of the diabetic patients was 62 years (56-71 years) and 73% were men, which was similar in
CE
the non-diabetic group. Diabetic patients were more likely to smoke (44% vs. 32%, p=0.043) and to have
a myocardial infarction in their medical history (14% vs. 6%, p=0.035) than non-diabetic patients. Other
AC
patient characteristics were similar in both groups.
Hemodynamically significant disease (invasive FFR ≤0.8) was present in 41 vessels (37%) of the
diabetic patients and in 171 vessels (41%) of the non-diabetic patients (p=0.455). On coronary CT
angiography a ≥50% stenosis was present in 74 vessels (67%) of the diabetic patients and in 308 vessels
(74%) of the non-diabetic patients (p=0.146). CT-FFR classified 48 vessels (44%) as functionally
6
ACCEPTED MANUSCRIPT
obstructed in diabetic patients and 197 vessels (47%) of the non-diabetic patients (p=0.474) (Figure 1).
The Bland-Altman plot revealed a mean difference between CT-FFR and invasive FFR of -0.034±0.126
in diabetic patients and -0.034±0.116 in non-diabetic patients (p=0.95) (Figure 2).
Because of differences in patient characteristics between the diabetic patients and non-diabetic
IP
T
patients, a multivariable logistic regression model was used to evaluate the effect of diabetes mellitus on
the diagnostic performance of coronary CT angiography and CT-FFR adjusted for smoking and prior
US
CR
myocardial infarction. The diagnostic performance of CT-FFR was similar in diabetic and non-diabetic
patients and improved the diagnostic performance of coronary CT angiography (≥50% stenosis) in both
groups (Table 2). No differences were found in accuracy of CT-FFR per stratified group (CT-FFR 0.6,
0.6-0.69, 0.7-0.79, 0.8-0.89, ≥0.9) between diabetic and non-diabetic patients (Figure 3). Additionally, the
AN
area under curves of the diagnostic performance of CT-FFR were similar (p=0.113) (Figure 4).
M
Discussion
We demonstrated that CT-FFR improved the diagnostic performance of coronary CT
suspected for CAD.
PT
ED
angiography and discriminated ischemia with good accuracy in both diabetic and non-diabetic patients
The high prevalence of diabetes mellitus in the general population and its significant risk for
CE
developing cardiovascular disease emphasizes the importance of accurate anatomical and functional
assessment in patients suspected for CAD3. Coronary CT angiography has emerged as a valuable
AC
approach in the diagnosis of CAD15,16. However, the interpretation of coronary CT angiography results
differs in diabetic patients due to a more diffuse presentation of CAD and faster progression of the
disease5,17. An Agatston coronary calcium score of zero has a much shorter warranty period for all-cause
death in diabetic patients compared to a general asymptomatic population18. These findings might be
explained by a higher prevalence of non-calcified plaques in this population19. Moreover, diabetic patients
have greater plaque progression of especially the non-calcified plaques, assumed to increase their risk for
7
ACCEPTED MANUSCRIPT
plaque rupture17. Therefore, plaque morphology assessment by coronary CT angiography has an
additional value in cardiovascular risk stratification and could lead to a more focused medical therapy
especially in diabetic patients20,21. Nevertheless, no studies have been conducted that show if this will lead
to a reduction in cardiovascular events.
IP
T
CT-FFR has incremental value over coronary CT angiography by ruling out hemodynamic
significance of angiographic lesions, which is especially relevant in diabetic patients due to their less
US
CR
typical presentation of CAD22. However, CT-FFR algorithms are based on the assumption that the
hyperemic response to adenosine is predictable12. This could potentially make CT-FFR less reliable in
diabetic patients due to a higher prevalence of microvascular dysfunction and higher vascular
resistance4,5. Few studies are available on the interpretation of on-site CT-FFR analysis in diabetic
AN
patients. Coenen et al. suggested that the presence of diabetes mellitus resulted in an increased
discrepancy between computational fluid dynamics based CT-FFR and invasive FFR8. However, these
M
results were not sufficiently powered. A study in which the diagnostic performance of the CT-FFR
application commercialized by HeartFlow (HeartFlow Inc., Redwood City, California, USA) was
PT
ED
analyzed, showed no differences in diagnostic accuracy of CT-FFR between diabetic patients and a
control group23. In the present study, we confirm a similar performance of CT-FFR in diabetic and nondiabetic patients. Additionally, the diagnostic performance of CT-FFR per stratified group was
CE
comparable between diabetic and non-diabetic patients. The overall accuracy of CT-FFR did vary
depending on the CT-FFR outcome, in line with the systemic review of published studies by Cook et al.24.
AC
Overall, these findings suggest that the presence of diabetic mellitus does not negatively affect the
accuracy of CT-FFR.
Validation studies of invasively measured FFR have shown contradictive results on the diagnostic
accuracy in diabetes patients25-27. This might be explained by differences in study population, since
limited data was available on the coronary vasodilation capacity and blood glucose levels of these
patients. Clinical outcome studies of invasive FFR guided revascularization have shown higher event
8
ACCEPTED MANUSCRIPT
rates in diabetic patients than in non-diabetic patients28,29. Therefore, the diagnostic accuracy and
interpretation of invasively measured FFR in diabetic patients remains ambiguous.
This study has several limitations that should be acknowledged. Given the retrospective nature of
the study and the small sample size, the potential impact of other variables on the diagnostic performance
IP
T
cannot be excluded. Moreover, we assumed that the presence of microvascular dysfunction is based on
the presence of diabetes mellitus. Our study is limited by the unavailability of directly measured
US
CR
microvascular circulation and vascular resistance and therefore we do not know the impact of
microvascular dysfunction in our population. Vasodilatory dysfunction could also have been reversed
when patients were optimally treated30. Therefore, further studies are needed with extensive information
on long-term glucose control of diabetic patients and the extent of microvascular dysfunction to fully
AN
evaluate the diagnostic performance of CT-FFR and its effect on clinical outcome of diabetic patient with
CAD.
M
In conclusion, CT-FFR was able to accurately identify functionally significant CAD in both
References
PT
ED
diabetic and non-diabetic patients and improved the diagnostic performance of coronary CT angiography.
1. Hillegass WB, Patel MR, Klein LW, Gurm HS, Brennan JM, Anstrom KJ, Dai D, Eisenstein EL,
CE
Peterson ED, Messenger JC, Douglas PS. Long-term outcomes of older diabetic patients after
percutaneous coronary stenting in the United States: a report from the National Cardiovascular Data
AC
Registry, 2004 to 2008. J Am Coll Cardiol 2012;60:2280-2289.
2. Wit MA, de Mulder M, Jansen EK, Umans VA. Diabetes mellitus and its impact on long-term
outcomes after coronary artery bypass graft surgery. Acta Diabetol 2013;50:123-128.
3. Emerging Risk Factors C, Sarwar N, Gao P, Seshasai SR, Gobin R, Kaptoge S, Di Angelantonio E,
Ingelsson E, Lawlor DA, Selvin E, Stampfer M, Stehouwer CD, Lewington S, Pennells L, Thompson A,
9
ACCEPTED MANUSCRIPT
Sattar N, White IR, Ray KK, Danesh J. Diabetes mellitus, fasting blood glucose concentration, and risk of
vascular disease: a collaborative meta-analysis of 102 prospective studies. Lancet 2010;375:2215-2222.
4. Di Carli MF, Janisse J, Grunberger G, Ager J. Role of chronic hyperglycemia in the pathogenesis of
coronary microvascular dysfunction in diabetes. J Am Coll Cardiol 2003;41:1387-1393.
IP
T
5. Murthy VL, Naya M, Foster CR, Gaber M, Hainer J, Klein J, Dorbala S, Blankstein R, Di Carli MF.
Association between coronary vascular dysfunction and cardiac mortality in patients with and without
US
CR
diabetes mellitus. Circulation 2012;126:1858-1868.
6. Task Force M, Montalescot G, Sechtem U, Achenbach S, Andreotti F, Arden C, Budaj A, Bugiardini
R, Crea F, Cuisset T, Di Mario C, Ferreira JR, Gersh BJ, Gitt AK, Hulot JS, Marx N, Opie LH, Pfisterer
M, Prescott E, Ruschitzka F, Sabate M, Senior R, Taggart DP, van der Wall EE, Vrints CJ, Guidelines
AN
ESCCfP, Zamorano JL, Achenbach S, Baumgartner H, Bax JJ, Bueno H, Dean V, Deaton C, Erol C,
Fagard R, Ferrari R, Hasdai D, Hoes AW, Kirchhof P, Knuuti J, Kolh P, Lancellotti P, Linhart A,
M
Nihoyannopoulos P, Piepoli MF, Ponikowski P, Sirnes PA, Tamargo JL, Tendera M, Torbicki A, Wijns
W, Windecker S, Document R, Knuuti J, Valgimigli M, Bueno H, Claeys MJ, Donner-Banzhoff N, Erol
PT
ED
C, Frank H, Funck-Brentano C, Gaemperli O, Gonzalez-Juanatey JR, Hamilos M, Hasdai D, Husted S,
James SK, Kervinen K, Kolh P, Kristensen SD, Lancellotti P, Maggioni AP, Piepoli MF, Pries AR,
Romeo F, Ryden L, Simoons ML, Sirnes PA, Steg PG, Timmis A, Wijns W, Windecker S, Yildirir A,
Zamorano JL. 2013 ESC guidelines on the management of stable coronary artery disease: the Task Force
CE
on the management of stable coronary artery disease of the European Society of Cardiology. Eur Heart J
2013;34:2949-3003.
AC
7. Fihn SD, Gardin JM, Abrams J, Berra K, Blankenship JC, Dallas AP, Douglas PS, Foody JM, Gerber
TC, Hinderliter AL, King SB, 3rd, Kligfield PD, Krumholz HM, Kwong RY, Lim MJ, Linderbaum JA,
Mack MJ, Munger MA, Prager RL, Sabik JF, Shaw LJ, Sikkema JD, Smith CR, Jr., Smith SC, Jr.,
Spertus JA, Williams SV, American College of Cardiology F, American Heart Association Task Force on
Practice G, American College of P, American Association for Thoracic S, Preventive Cardiovascular
Nurses A, Society for Cardiovascular A, Interventions, Society of Thoracic S. 2012
10
ACCEPTED MANUSCRIPT
ACCF/AHA/ACP/AATS/PCNA/SCAI/STS Guideline for the diagnosis and management of patients with
stable ischemic heart disease: a report of the American College of Cardiology Foundation/American
Heart Association Task Force on Practice Guidelines, and the American College of Physicians, American
Association for Thoracic Surgery, Preventive Cardiovascular Nurses Association, Society for
IP
T
Cardiovascular Angiography and Interventions, and Society of Thoracic Surgeons. J Am Coll Cardiol
2012;60:e44-e164.
US
CR
8. Coenen A, Lubbers MM, Kurata A, Kono A, Dedic A, Chelu RG, Dijkshoorn ML, van Geuns RJ,
Schoebinger M, Itu L, Sharma P, Nieman K. Coronary CT angiography derived fractional flow reserve:
Methodology and evaluation of a point of care algorithm. J Cardiovasc Comput Tomogr 2016;10:105113.
AN
9. Tesche C, De Cecco CN, Baumann S, Renker M, McLaurin TW, Duguay TM, Bayer RR, 2nd,
Steinberg DH, Grant KL, Canstein C, Schwemmer C, Schoebinger M, Itu LM, Rapaka S, Sharma P,
M
Schoepf UJ. Coronary CT Angiography-derived Fractional Flow Reserve: Machine Learning Algorithm
versus Computational Fluid Dynamics Modeling. Radiology 2018:171291.
PT
ED
10. Coenen A, Kim YH, Kruk M, Tesche C, De Geer J, Kurata A, Lubbers ML, Daemen J, Itu L, Rapaka
S, Sharma P, Schwemmer C, Persson A, Schoepf UJ, Kepka C, Hyun Yang D, Nieman K. Diagnostic
Accuracy of a Machine-Learning Approach to Coronary Computed Tomographic Angiography-Based
Fractional Flow Reserve: Result From the MACHINE Consortium. Circ Cardiovasc Imaging
CE
2018;11:e007217.
11. Nahser PJ, Jr., Brown RE, Oskarsson H, Winniford MD, Rossen JD. Maximal coronary flow reserve
AC
and metabolic coronary vasodilation in patients with diabetes mellitus. Circulation 1995;91:635-640.
12. Sharma P, Itu L, Zheng X, Kamen A, Bernhardt D, Suciu C, Comaniciu D. A framework for
personalization of coronary flow computations during rest and hyperemia. Conf Proc IEEE Eng Med Biol
Soc 2012;2012:6665-6668.
13. Raff GL, Abidov A, Achenbach S, Berman DS, Boxt LM, Budoff MJ, Cheng V, DeFrance T,
Hellinger JC, Karlsberg RP, Society of Cardiovascular Computed T. SCCT guidelines for the
11
ACCEPTED MANUSCRIPT
interpretation and reporting of coronary computed tomographic angiography. J Cardiovasc Comput
Tomogr 2009;3:122-136.
14. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated
receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44:837-845.
IP
T
15. Budoff MJ, Dowe D, Jollis JG, Gitter M, Sutherland J, Halamert E, Scherer M, Bellinger R, Martin A,
Benton R, Delago A, Min JK. Diagnostic performance of 64-multidetector row coronary computed
US
CR
tomographic angiography for evaluation of coronary artery stenosis in individuals without known
coronary artery disease: results from the prospective multicenter ACCURACY (Assessment by Coronary
Computed Tomographic Angiography of Individuals Undergoing Invasive Coronary Angiography) trial. J
Am Coll Cardiol 2008;52:1724-1732.
AN
16. Miller JM, Rochitte CE, Dewey M, Arbab-Zadeh A, Niinuma H, Gottlieb I, Paul N, Clouse ME,
Shapiro EP, Hoe J, Lardo AC, Bush DE, de Roos A, Cox C, Brinker J, Lima JA. Diagnostic performance
M
of coronary angiography by 64-row CT. N Engl J Med 2008;359:2324-2336.
17. Nakanishi R, Ceponiene I, Osawa K, Luo Y, Kanisawa M, Megowan N, Nezarat N, Rahmani S,
PT
ED
Broersen A, Kitslaar PH, Dailing C, Budoff MJ. Plaque progression assessed by a novel semi-automated
quantitative plaque software on coronary computed tomography angiography between diabetes and nondiabetes patients: A propensity-score matching study. Atherosclerosis 2016;255:73-79.
18. Valenti V, Hartaigh BO, Cho I, Schulman-Marcus J, Gransar H, Heo R, Truong QA, Shaw LJ,
CE
Knapper J, Kelkar AA, Sciarretta S, Chang HJ, Callister TQ, Min JK. Absence of Coronary Artery
Calcium Identifies Asymptomatic Diabetic Individuals at Low Near-Term But Not Long-Term Risk of
AC
Mortality: A 15-Year Follow-Up Study of 9715 Patients. Circ Cardiovasc Imaging 2016;9:e003528.
19. Maffei E, Seitun S, Nieman K, Martini C, Guaricci AI, Tedeschi C, Weustink AC, Mollet NR, Berti
E, Grilli R, Messalli G, Cademartiri F. Assessment of coronary artery disease and calcified coronary
plaque burden by computed tomography in patients with and without diabetes mellitus. Eur Radiol
2011;21:944-953.
12
ACCEPTED MANUSCRIPT
20. Dimitriu-Leen AC, Scholte AJ, van Rosendael AR, van den Hoogen IJ, Kharagjitsingh AV,
Wolterbeek R, Knuuti J, Kroft LJ, Delgado V, Jukema JW, de Graaf MA, Bax JJ. Value of Coronary
Computed Tomography Angiography in Tailoring Aspirin Therapy for Primary Prevention of
Atherosclerotic Events in Patients at High Risk With Diabetes Mellitus. Am J Cardiol 2016;117:887-893.
IP
T
21. Inoue K, Motoyama S, Sarai M, Sato T, Harigaya H, Hara T, Sanda Y, Anno H, Kondo T, Wong ND,
Narula J, Ozaki Y. Serial coronary CT angiography-verified changes in plaque characteristics as an end
US
CR
point: evaluation of effect of statin intervention. JACC Cardiovasc Imaging 2010;3:691-698.
22. Khafaji HA, Suwaidi JM. Atypical presentation of acute and chronic coronary artery disease in
diabetics. World J Cardiol 2014;6:802-813.
23. Eftekhari A, Min J, Achenbach S, Marwan M, Budoff M, Leipsic J, Gaur S, Jensen JM, Ko BS,
AN
Christiansen EH, Kaltoft A, Botker HE, Jensen JF, Norgaard BL. Fractional flow reserve derived from
coronary computed tomography angiography: diagnostic performance in hypertensive and diabetic
M
patients. Eur Heart J Cardiovasc Imaging 2016.
24. Cook CM, Petraco R, Shun-Shin MJ, Ahmad Y, Nijjer S, Al-Lamee R, Kikuta Y, Shiono Y, Mayet J,
PT
ED
Francis DP, Sen S, Davies JE. Diagnostic Accuracy of Computed Tomography-Derived Fractional Flow
Reserve : A Systematic Review. JAMA Cardiol 2017;2:803-810.
25. Sahinarslan A, Kocaman SA, Olgun H, Kunak T, Kiziltunc E, Ozdemir M, Timurkaynak T. The
reliability of fractional flow reserve measurement in patients with diabetes mellitus. Coron Artery Dis
CE
2009;20:317-321.
26. Yanagisawa H, Chikamori T, Tanaka N, Usui Y, Takazawa K, Yamashina A. Application of pressure-
AC
derived myocardial fractional flow reserve in assessing the functional severity of coronary artery stenosis
in patients with diabetes mellitus. Circ J 2004;68:993-998.
27. Reith S, Battermann S, Hellmich M, Marx N, Burgmaier M. Impact of type 2 diabetes mellitus and
glucose control on fractional flow reserve measurements in intermediate grade coronary lesions. Clin Res
Cardiol 2014;103:191-201.
13
ACCEPTED MANUSCRIPT
28. Kennedy MW, Kaplan E, Hermanides RS, Fabris E, Hemradj V, Koopmans PC, Dambrink JH,
Marcel Gosselink AT, Van't Hof AW, Ottervanger JP, Roolvink V, Remkes WS, van der Sluis A,
Suryapranata H, Kedhi E. Clinical outcomes of deferred revascularisation using fractional flow reserve in
patients with and without diabetes mellitus. Cardiovasc Diabetol 2016;15:100.
IP
T
29. Liu Z, Matsuzawa Y, Herrmann J, Li J, Lennon RJ, Crusan DJ, Kwon TG, Zhang M, Sun T, Yang S,
Gulati R, Bell MR, Lerman LO, Lerman A. Relation between fractional flow reserve value of coronary
US
CR
lesions with deferred revascularization and cardiovascular outcomes in non-diabetic and diabetic patients.
Int J Cardiol 2016;219:56-62.
30. Takiuchi S, Rakugi H, Masuyama T, Ikegami H, Nishikage T, Shintani M, Komai N, Nagai M,
Kamide K, Higaki J, Ogihara T. Hypertension attenuates the efficacy of hypoglycemic therapy for
AC
CE
PT
ED
M
AN
preserving coronary flow reserve in patients with type 2 diabetes. Hypertens Res 2002;25:893-900.
14
AN
US
CR
IP
T
ACCEPTED MANUSCRIPT
Figure 1 Coronary CT angiography, invasive coronary angiography, and CT-FFR in a single patient. The
coronary CT angiography showed several moderate stenoses in the right (A) and left coronary artery (B),
M
but the evaluation was hampered by the amount of calcium in the coronary arteries. Invasive coronary
PT
ED
angiography showed no significant stenoses in right coronary artery (C) and multiple lesions in the left
anterior descending artery with a FFR of 0.90 (D). The CT-FFR analysis showed comparable results with
AC
CE
the invasively measured FFR (E).
15
PT
ED
M
AN
US
CR
IP
T
ACCEPTED MANUSCRIPT
Figure 2 A Bland-Altman plot of CT-FFR and invasive FFR in vessels of diabetic (red) and non-diabetic
CE
patients (blue). A horizontal line is placed at the mean difference between CT-FFR and invasive FFR. The
dotted lines are placed at the limits of agreement: mean ± 1.96 standard deviation. CT-FFR: computed
AC
tomography based fractional flow reserve.
16
AN
US
CR
IP
T
ACCEPTED MANUSCRIPT
M
Figure 3
PT
ED
Per-vessel diagnostic accuracy of CT-FFR per stratified group in diabetic (red) and non-diabetic patients
(blue). Error bars represent the standard error in each group. On the right y-axis the number of vessels
with an invasive FFR value corresponding to each interval of diabetic (red) and non-diabetic patients
AC
CE
(blue). CT-FFR: computed tomography based fractional flow reserve.
17
M
AN
US
CR
IP
T
ACCEPTED MANUSCRIPT
PT
ED
Figure 4
Per vessel receiver operating characteristics curves displaying the diagnostic performance of CT-FFR in
diabetic (red) and non-diabetic patients (blue). CT-FFR: computed tomography based fractional flow
AC
CE
reserve; AUC: area under curve; CI: confidence interval.
18
ACCEPTED MANUSCRIPT
Table 1: Patients characteristics
Diabetes Mellitus a
Yes (n=75)
p-value
Age (years)
63 (56-69)
62 (56-71)
0.374
Male gender (%)
203 (74%)
55 (73%)
0.970
Body mass index (kg/m2) *
27 (25-29)
26 (24-30)
0.988
181 (66%)
51 (68%)
0.695
163 (59%)
47 (63%)
0.572
90 (33%)
29 (39%)
0.336
Smoking within the last year
87 (32%)
33 (44%)
0.043
Body mass index ≥25 (kg/m ) *
194 (72%)
45 (63%)
0.124
Prior myocardial infarction (%) †
13 (6%)
9 (14%)
0.035
Prior percutaneous coronary intervention (%) †
43 (19%)
10 (15%)
0.526
Left ventricular mass (gram) †
161 (139-179)
165 (150-184)
0.061
Agatston coronary calcium score**
227 (39-661)
259 (32-712)
0.965
Coronary CT angiography image quality e
3 (3-4)
3 (3-4)
0.585
Dyslipidemia c
Family history of coronary artery disease
d
2
M
Hypertension b
US
CR
Cardiovascular risk factors (%)
IP
T
No (n=276)
AN
Variables
PT
ED
Values are reported as median and 25th-75th percentile or an absolute number n and percentage (%). Differences were tested
by Mann-Whitney-U-, Chi-square and Fisher’s t-tests. Data was not available * in 9 patients, ** in 37 patients and † in 53
patients.
a
defined as a fasting glucose level of 126 mg/dL (6.99 mmol/L) and higher, HbA1c >= 6.5% or use of antidiabetic
medication. b defined as blood pressure >140 mmHg systolic, >90 mmHg diastolic, or use of antihypertensive medication. c
defined as a total cholesterol of >200mg/dl or use of lipid-lowering therapy. d included known coronary artery disease, former
CE
myocardial infarction or any revascularizations in any first degree relative. e based on a 4-point Likert scale (1 non-diagnostic;
2 impaired image quality, differentiation of the coronary artery wall possible with reduced confidence; 3 adequate reduced
image quality due to artifacts without limiting coronary artery wall differentiation; 4 excellent, no artifacts present an clear
AC
differentiation of the coronary artery wall. CT: computed tomography
19
ACCEPTED MANUSCRIPT
Table 2: Per-vessel diagnostic performance
Sensitivity
Specificity
PPV
NPV
Accuracy
Non-diabetes
87% (82-93%)
37% (30-43%)
50% (44-56%)
80% (72-88%)
58% (53-63%)
Diabetes
90% (80-100%)
48% (35-61%)
54% (41-66%)
88% (76-100%)
65% (55-75%)
p-value
0.532
0.271
0.687
0.271
0.223
Non-diabetes
79% (73-86%)
72% (66-78%)
67% (60-74%)
83% (77-88%)
75% (71-80%)
Diabetes
88% (77-98%)
80% (70-90%)
74% (62-87%)
91% (82-99%)
83% (76-90%)
p-value
0.234
0.190
0.350
0.151
0.088
CT-FFR ≤0.80
US
CR
≥50%
IP
T
Coronary CT angiography
Values are reported as percentage with 95% confidence interval. Differences are conducted by a multivariable logistic regression model including
diabetes, smoking and prior myocardial infarction. CT: computed tomography; CT-FFR: computed tomography based fractional flow reserve;
AC
CE
PT
ED
M
AN
NPV: negative predictive value; PPV: positive predictive value.
20