Author + information
- Received April 12, 2016
- Revision received July 22, 2016
- Accepted September 8, 2016
- Published online December 19, 2016.
- Christoph Lutter, MDa,
- Hiroyoshi Mori, MDa,
- Kazuyuki Yahagi, MDa,
- Elena Ladich, MDa,
- Michael Joner, MDa,
- Robert Kutys, MSa,
- David Fowler, MDb,
- Maria Romero, MDa,
- Jagat Narula, MDc,
- Renu Virmani, MDa and
- Aloke V. Finn, MDd,∗ ()
- aCVPath Institute, Gaithersburg, Maryland
- bOffice of the Chief Medical Examiner, Baltimore, Maryland
- cIcahn School of Medicine at Mount Sinai, New York, New York
- dUniversity of Maryland, School of Medicine, Baltimore, Maryland
- ↵∗Reprint requests and correspondence:
Dr. Aloke V. Finn, CVPath Institute, 19 Firstfield Road, Gaithersburg, Maryland 20878.
Objectives The aim of this study was to identify histological features that correlate with terms commonly used to describe optical coherence tomographic (OCT) and optical frequency-domain imaging (OFDI) images of stented vessels, by means of a histopathological validation study using stented human coronary arteries.
Background OCT imaging and OFDI are used to evaluate vascular responses to stent implantation. Descriptive terms such as “peristrut low attenuation” and “heterogeneous” have been used to describe neointimal characteristics that may have clinical relevance. However, only limited histopathological correlations are available.
Methods Using the CVPath stent registry, 19 cases were identified in whom implantation duration was >30 days and OCT imaging or OFDI and histological findings were available. Consecutive OCT or OFDI frames (n = 1,063) of stented coronary arteries were categorized according to their predominant imaging features in 1-mm intervals. Coregistration of OCT or OFDI frames and histopathological cross sections was performed in 111 frames.
Results Seven distinct OCT or OFDI patterns were found: homogenous (45%), layered (15%), high intensity with high attenuation (14%), intraluminal protruding masses (8%), peristrut low attenuation (7%), heterogeneous (2%), and honeycomb (1%). Histopathologically, the homogenous pattern correlated most often with smooth muscle cells within collagenous/proteoglycan matrix and less often with organized thrombus. The layered pattern correlated with healed neointimal rupture or erosion, peristrut neovascularization, or smooth muscle cells within collagen/proteoglycan matrix. High intensity with high attenuation correlated with superficial macrophage accumulation in the majority of cases, but with other histological findings in 30% of cases. The diagnostic accuracy was greater in restenotic lesions. The only OCT or OFDI finding that had a single histological feature was the honeycomb pattern.
Conclusions This study suggests a lack of correlation between OCT image patterns and distinct histological tissue characteristics.
Optical coherence tomography (OCT) imaging has been introduced as an innovative tool capable of providing high-resolution images of anatomic structures at micrometer-level resolution (1–3). Because of its ability to provide near histology-level images, intravascular OCT imaging has been used to describe vascular responses following percutaneous coronary intervention. Previous histopathological studies of human coronary bare-metal stents (BMS) and drug-eluting stents (DES) have shown increases in collagen and decreases in proteoglycan-rich matrix occurring over time, with these changes relatively delayed in DES compared with BMS (4). In an era in which a majority of devices used are DES, it also remains possible that neointimal responses may differ among patients and that this might have clinical significance. Observational OCT studies of patients receiving DES have described the appearance of neointimal characteristics such as heterogeneous, homogeneous, and layered neointimal patterns, as well as peristrut low attenuation (PLIA), and have linked some of the features with adverse clinical outcomes (5,6). However, validation studies correlating these descriptions with histopathological findings are lacking. Consequently, clinical use of OCT imaging to identify patients at risk for adverse events is on the basis of association only rather than the actual tissue responses that underlie such images. Such correlations are impossible in the clinical setting because histological samples are impossible to obtain. Using our unique pathological database of stented human implants, we conducted the largest to date ex vivo study investigating the correlation between OCT images generated from our samples and histopathological findings in stents implanted for >30 days.
All available cases from the CVPath stent registry (2002 to 2015), in which OCT or optical frequency-domain imaging (OFDI) data were available were included in the study; 22 autopsy cases with a total of 36 lesions and 42 implanted stents (17 BMS, 11 first-generation DES, and 14 second-generation DES) were reviewed. Stent duration was either obtained from clinical records (15 cases) or, if no clinical records were available, estimated independently by 2 experienced pathologists (R.V. and E.L., 7 cases). Cause of death and the indication for stent implantation are shown in Table 1.
Angiography and intravascular imaging procedure
Hearts were received from the medical examiner and consecutively included in the present study, but only stents implanted for >30 days were included. Following gross inspection of the heart and review of the clinical history, angiography was performed. Frequency-domain OCT imaging or OFDI without simultaneous occlusion of the vessel was then performed following pressure fixation (10% neutral-buffered formalin) to maintain physiological vessel dimensions. OCT imaging or OFDI was not attempted in the case of occlusive thrombus. To avoid disruption of nonocclusive thrombus, a guidewire and an OCT catheter were advanced in small increments under fluoroscopic guidance. OCT catheters (2.7-F, St. Jude Medical, St. Paul, Minnesota) or OFDI catheters (2.4-F, Terumo Corporation, Tokyo, Japan) were introduced after a 0.014-inch guidewire was placed in the coronary arteries. An imaging run was performed from the distal to the proximal arterial segments to include the stented portion of the artery at a pull-back speed of 15 mm/s (120 frames/s). For improved image quality, vessels were simultaneously flushed with contrast at low pressure (7). The data were transferred to an offline station for further investigation and coregistration.
Coronary arteries and myocardial sections were processed and sectioned as previously described (4).
Classification of lesions
Stents implanted with a gap of at least 5 mm were considered separate lesions, whereas overlapping or consecutively implanted stents were treated as 1 lesion. For the determination of the cause of death, clinical and medical examiner or hospital records were reviewed, and gross heart as well as myocardial sections were analyzed.
OCT and OFDI analysis
All available OCT and OFDI pull-backs were reviewed by 2 investigators (C.L. and H.M.) and interpreted completely independently of the histological findings. Image frames were systematically evaluated in 1-mm increments starting at the first frame in which stent struts were visible along the full vascular circumference. Frames were stratified into distinct categories on the basis of the predominant qualitative imaging features (Table 2).
Image frames showing more than 1 image feature were counted consecutively. Cases were further categorized according to previously published algorithms on the basis of their typical OCT or OFDI tissue-scattering characteristics as homogenous pattern, layered neointimal pattern, heterogeneous pattern, high-intensity and high-attenuation pattern, PLIA pattern, honeycomb (lotus root) pattern, or tissue coverage with irregular surface or intraluminal protruding mass.
Homogenous backscattering was defined as uniform reflection of light without circumscribed areas of stronger or weaker backscattering properties. Images showing focal variation of the backscattering pattern were considered to be heterogeneous. The presence of a low-intensity backscattering signal layer close to stent struts with a higher intensity backscattering layer above was considered to be of layered character (5,6,8,9). The PLIA pattern was defined as a circumscribed low-intensity area surrounding struts. Frames showing a high-intensity luminal margin of the neointima with strong light attenuation were grouped as high-attenuation surface (8). Tissue coverage with irregular surface or intraluminal protruding mass was defined as irregular convex or concave interruption of the neointimal surface (other than side branches). Honeycomb or lotus root OCT or OFDI appearance was defined as restenotic lumen with (multiple) septa dividing the residual lumen into 2 or more cavities (10). Insufficient stent strut coverage was defined by the absence of neointimal layer above the stent struts. Uncovered struts with a clear separation from the underlying vessel wall (exceeding the visually projected strut thickness) were defined as malapposed struts, as previously published (7). Neovascularization was defined by the presence of round or tubular structures spanning the neointimal tissue in at least 3 consecutive frames, as previously described (8).
Coregistration of OCT and OFDI images with histological findings
OCT and OFDI frames were coregistered with histological cross sections by visual alignment of anatomic landmarks such as areas of calcification, orientation of stent struts, side branches, and eccentricity of neointimal growth. Two observers (C.L. and H.M.) coregistered OCT and OFDI frames with histological cross sections, considering pull-back speed and frame rate.
Following diagnostic assignment of OCT and OFDI frames and successful coregistration (n = 111 frames) (Table 3), when necessary, serial sectioning was performed for the assessment of dominant imaging features.
Neoatherosclerosis was defined by the presence of any lipid deposition in the peristrut or interstrut regions or on the luminal surface (foamy macrophages), necrotic core formation, or calcification (11). Neovascularization was defined as the formation of microvascular networks. Hypersensitivity was diagnosed when there were severe transmural inflammatory infiltrates, including eosinophils and T-lymphocytes (12–14). Foreign material was identified by its characteristic appearance either as basophilic material on hematoxylin and eosin–stained sections or as foreign material with crystalline characteristics.
Strut-level analysis of tissue components
Strut-level analysis (n = 924 struts) of tissue components was performed on all available coregistered frames showing a minimum of 100-μm strut coverage (n = 111 coregistered images) to correlate relevant pathological features (Tables 4, 5, and 6).
Therefore, only struts that caused a distinct predominant OCT or OFDI appearance (i.e., homogenous) were assessed on the basis of the dominant histological characteristics. Struts with a secondary OCT or OFDI pattern or with insufficient coverage were excluded. Layered OCT or OFDI appearance is reported histologically as a combination of findings, whereas strut-level findings by OCT imaging or OFDI is histopathologically assessed and scored according to the predominant tissue component (Tables 4 and 6) (layered OCT or OFDI pattern, Table 5). All findings are reported as total number and percentage.
Numeric parameters are expressed as mean ± SD. Calculations were performed using JMP version 9.0 (SAS Institute, Cary, North Carolina). The degree of agreement between 2 reviewers was quantified using Cohen’s κ test for concordance.
Twenty-two patients were identified in the group of cases with implantation duration >30 days, of whom 19 (Table 1) could be adequately assessed for OCT or OFDI characteristics and form the basis of this study. The study group consisted of 4 women and 15 men, with a mean age of 59.3 ± 15.1 years. Only 2 patients were treated in the setting of acute coronary syndromes, whereas all other patients underwent revascularization in the setting of stable coronary artery disease. Thirty-seven stents were implanted in 31 lesions (9 in the right coronary artery, 17 in the left anterior descending coronary artery, 1 from the left main to the left anterior descending coronary artery, 2 from the left anterior descending coronary artery to the first left diagonal branch, 1 in the first left obtuse marginal branch, and 1 in the second left obtuse marginal branch). In-stent restenosis was observed in 5 lesions (16.1%) from 4 patients. Five stents in 2 patients had nonocclusive stent thrombosis (13.2%). All other stents (13 patients) did not show any thrombi and had varying degrees of luminal narrowing. Stent-related death was observed in 4 patients, 12 patients died of noncardiac causes, and non–stent-related cardiac death occurred in 3 patients.
Analysis of neointimal OCT and OFDI features
In total, 1,063 image frames with implantation duration ≥30 days were analyzed. Sixty-four frames (6%) could not be categorized, because of insufficient image quality due to post-mortem clots. A homogenous pattern was the most frequent primary finding, observed in 45% (477 of 1,063); a layered neointimal backscattering pattern was predominant in 15% of frames (164 of 1,063). High-intensity and high-attenuation surface was observed in 14% of frames (144 of 1,063), and tissue coverage with irregular surface or intraluminal protruding masses was observed in 8% (89 of 1,063). PLIA was seen in 7% (77 of 1,063), and heterogeneous backscattering was found in 2% of frames (16 of 1,063). A honeycomb (lotus root) appearance was detected in 1% (8 of 1,063); 24 frames showed well-apposed struts without sufficient tissue coverage (2% of all frames) as the dominant finding and therefore could not be classified into a predominant imaging pattern (Table 2).
The overall intraobserver κ coefficient for detecting strut-based tissue components was 0.882, and the overall interobserver κ coefficient was 0.730.
Identification of OCT and OFDI features with differential diagnosis
In 19 cases, a total of 111 pairs of matching coregistered frames and cross sections were analyzed. For each predominant imaging feature category, at least 3 frames (minimum 3 frames, maximum 46 frames) with an average of 15.9 ± 13.5 frames were assessed (Table 3). Following histopathological verification of the underlying pathology, we identified differential diagnoses for 6 of the commonly described OCT and OFDI features (Table 3). For the rarely seen OCT or OFDI feature of honeycomb (lotus root) pattern, only 1 histopathological diagnosis was present.
Histological differential diagnosis of OCT and OFDI features
In 6 of 7 categories of OCT imaging descriptions, we found at least 2 different underlying histopathological conditions, which revealed similar OCT and OFDI features (Table 3). The OCT or OFDI finding of a homogenous pattern revealed the presence of smooth muscle cell (SMC)–rich neointimal tissue by histology in 39% of coregistered frames, collagen-rich tissue with rare SMCs in 39% of coregistered frames or sections, SMCs in proteoglycan- and collagen-rich tissue in 15%, and organized thrombus in 7% (Table 3, Figure 1). Therefore, the informative value of OCT imaging or OFDI was relatively high in this pattern, as histological correlations consisted of different combinations of SMCs in collagen- and/or proteoglycan-rich tissue, overall representing maturing neointimal tissue and organizing thrombus. The OCT or OFDI finding of layered neointimal backscattering revealed a multitude of differential matching histopathological components, including healed neointimal tissue from rupture or erosion (59% of coregistered layered image frames), peristrut neovascularization (24%), SMCs in proteoglycan- and collagen-rich tissue (12%), and neointimal calcification (6%) (Table 3, Figure 2). The feature of high-attenuation surface with or without invisible stent struts was observed in frames with superficial macrophage accumulation (70%) or superficial elastic fibers (10%), neointimal calcification at the luminal margin (10%), calcifying neointimal lipid pool (5%), and neoatherosclerosis with necrotic core (5%) (Table 3, Figure 3) (8). Tissue coverage with irregular surface or “intraluminal protruding” masses correlated with histological cross sections exhibiting stent struts with adherent organized thrombus in 63% of frames and platelet-rich thrombus in 37% (Table 3). PLIA areas showed a number of different matching histopathological components, including peristrut calcification (30%), peristrut neovascularization (30%), stent-induced hypersensitivity vasculitis (30%), and foreign material (10%).
Strut tissue characteristics
Strut-based imaging findings and tissue composition can vary from strut to strut, even within the same cross section, and might be missed using a frame-by-frame analysis. We therefore performed a strut-level analysis examining the correlation between strut-based imaging findings and strut-based tissue composition (n = 944 coregistered struts). Individual struts were grouped according to the predominant imaging feature, and histological correlations were performed as described previously.
A total of 444 struts (46 coregistered frames) were detected in which OCT imaging or OFDI showed a homogenous pattern. Within frames showing this pattern, the dominant histopathological finding in the proximity of struts and in the above-strut regions close to the lumen was SMC-rich tissue (39%). But in 36% of coregistered struts, collagen-rich tissue with rare SMCs was detected, and 14% showed an equal distribution of SMCs, proteoglycan, and collagen. Organized thrombus was present in 9% of all struts, and proteoglycan-rich tissue with interspersed SMCs was scored in 2% (Figure 1, Table 4).
A total of 104 struts (17 coregistered frames) were identified as showing layered neointimal backscattering. Strut-based analysis of tissue composition revealed the presence of different compositions of collagen- or SMC-rich tissue (e.g., healed rupture or erosion) in 59%, a combination of neovascularization and SMC-rich neointima in 25%, a combination of collagen-rich neointimal tissue and calcification at the luminal border in 10%, and a combination of neovascularization and collagen-rich neointima in 6% (Figure 2, Table 5).
High-intensity, high-attenuation lesions
Investigations of strut-level tissue components (n = 122 struts in 20 coregistered frames), which caused high-intensity high signal attenuation beyond the endoluminal tissue, revealed a predominance of foam cell accumulation (68%), superficial elastic fibers without foamy macrophages (12%), and neointimal calcification (11%). Cholesterol clefts (1%), healthy neointima (2%), and organized thrombus (1%) were seen less frequently (Figure 3, Table 6).
Tissue coverage with irregular surface or intraluminal protruding mass
Eighty-two struts were observed in 8 coregistered frames. The majority of these struts showed neointimal coverage with organized thrombus (72%) or platelet-rich thrombus (23%). In 5% of struts, there was fibrin coverage (Figure 4, Table 6).
Peristrut low attenuation
A total of 39 struts (10 coregistered frames) were identified in which OCT imaging or OFDI showed PLIA. Histological assessment showed a predominance of peristrut neovascularization (36%) or inflammatory reaction, characterized by peristrut giant cell accumulation in 23% or peristrut leukocyte accumulation (13%). Eighteen percent of the struts were surrounded by calcification, whereas cholesterol clefts (5%), fibrin accumulation (3%), and foreign material (3%) were seen less frequently. A sharp border of PLIA was caused mainly by peristrut calcification (7 struts) (Figure 5, Table 6).
Within 7 coregistered frames, a total of 94 struts could be included in the histopathological analysis, and a variety of different findings were detected: inflammation with neointimal giant cell accumulation (34%) or leukocyte accumulation (29%), neoatherosclerotic foam cell accumulation (12%) or cholesterol clefts (11%), inconspicuous healthy neointima (12%), and fibrin accumulation (3%) (Figure 1, Table 6).
Honeycomb or lotus root pattern
The overall intraobserver κ coefficient for detecting strut-based tissue components was 0.818, and the overall interobserver κ coefficient was 0.763.
A subanalysis of 5 restenotic lesions in 4 patients for which 166 OCT or OFDI frames were analyzed in 1-mm interval showed layered appearance in 35%, high-intensity and -attenuation surface in 31%, and homogenous pattern in 22%. PLIA was seen in 2%, and heterogeneous pattern or surface coverage with irregular outline was observed in only 1%. A total of 8% of frames could not be categorized because of poor image quality. Of the 166 frames, 23 could be coregistered, and these images were distributed as follows: layered appearance in 48%, homogenous pattern in 26%, high-intensity and -attenuation surface in 22%, and PLIA in 4%. The homogenous pattern correlated with SMC-rich neointimal tissue with varying proportions of collagen- and/or proteoglycan-rich matrix. Layered appearance was caused mainly by healed neointimal tissue from rupture or erosion (90%) or peristrut neovascularization (10%). High-intensity and high-attenuation surface showed the presence of superficial macrophage accumulation (80%) and neoatherosclerosis with necrotic core (20%). PLIA was caused by peristrut calcifications.
The present study is the first of its kind aimed at understanding histological correlations between commonly used OCT and OFDI patterns after stenting with the pathological findings using a large number of human post-mortem samples. We were able to identify specific commonly used OCT and OFDI descriptions for neointimal and peristrut tissue responses and defined the histological findings underpinning these terms. Homogenous pattern (45%), layered neointimal appearance (15%), and High-intensity and high-attenuation pattern (14%) were the most common OCT and OFDI features in our study population, whereas other terms, such as tissue coverage with irregular surface or intraluminal protruding mass (8%), peristrut low-intensity pattern (7%), heterogeneous pattern (2%), and honeycomb or lotus root pattern (1%), were less frequently observed. Histopathological assessment of neointimal tissue revealed a variety of underlying conditions within all of the recognized OCT and OFDI features, with the exception of honeycomb (lotus root) pattern, for which only 1 diagnosis could be established. Detailed strut-level analysis of neointimal tissue responses revealed substantial heterogeneity of imaging descriptions and various histological findings.
Consecutive imaging studies of stented segment in humans have revealed dynamic neointimal tissue maturation over time (15). It has been documented that homogenous tissue backscattering can turn into heterogeneous and vice versa in patients receiving DES. Heterogeneous tissue backscattering was also more often observed in focal restenotic lesions and at earlier time points (9), whereas layered tissue backscattering was more frequent in diffuse stent restenosis (6). Heterogeneous backscattering of neointimal tissue has been associated with increased major adverse cardiac events in patients treated with DES at a median follow-up of 22 months (16).
In this study, many commonly used OCT and OFDI patterns correlated with more than 1 histopathological finding, whereas for others, the differential histopathological diagnosis was more limited. Histopathological correlation of the homogenous imaging pattern was relatively limited (SMCs, collagen, and proteoglycans in different proportions), although in a few samples (7%), organized thrombus was also observed as homogenous. Layered appearance was detected not only in cases with excessive neovascularization, but also in cases with healed neointimal rupture or erosion, neointimal calcification, and SMC-rich tissue within proteoglycan and collagen matrix. High-intensity and high-attenuation surface was interpreted as a signal for multiple histological findings, including neoatherosclerosis with or without necrotic core, superficial elastic fibers, or calcification at the luminal margin of neointimal tissue as well as in lipid pools. Peristrut low intensity was caused by calcifications, neovascularization, and hypersensitivity vasculitis with massive accumulation of inflammatory cells as well by foreign material. Stent-induced hypersensitivity vasculitis was also the underlying pathology in some cases with heterogeneous imaging pattern, whereas others showed neointimal SMC-rich tissue with proteoglycans and collagen. In cases of restenosis, the diagnostic accuracy for specific histological findings of homogenous, layered, and high-attenuation surface patterns was substantially improved. The latter finding raises the question of whether some of the nonspecificity of OCT for determining neointimal composition might be due to the relatively thin layer of neointima that develops after DES implantation and whether substantially greater tissue response is required to improve the diagnostic accuracy of OCT imaging.
Only recently, Shibuya et al. (8) investigated OCT imaging findings from 6 autopsy hearts relative to coregistered histological cross sections and found that the features of heterogeneous and layered tissue backscattering were elicited by a number of different underlying pathologies, including markers of incomplete healing and in-stent neoatherosclerosis. In contrast, the investigators described homogeneous pattern as caused mainly by SMC-rich tissue with collagen. In another interesting study, Phipps et al. (17) reported that a variety of different histopathological findings can cause bright spots in optical coherence tomography images from 14 coronary arteries. Histopathological analysis revealed the presence of macrophages only, cellular fibrous tissue, calcifications, and lipid accumulation. The findings of both studies are in clear agreement with the present findings.
OCT and OFDI findings derived from autopsy studies may not be entirely comparable with clinical findings, as several factors (e.g., missing heartbeat during OCT pull-back, high-pressure flushing, or residual blood during image acquisition) are known to influence the interpretation of results. As in every autopsy study, presentation of cases is selective. However, we investigated the largest sample size of post-mortem autopsy hearts to date, applying coregistration of histopathology and OCT imaging or OFDI, and therefore we believe our findings hold important value.
The present study represents the largest compilation of matched OCT and OFDI findings with histopathological verification to understand important sources of misinterpretation during OCT and OFDI analysis. Clinically relevant differential diagnosis was elaborated for 6 of 7 characteristic OCT and OFDI features. Histological tissue characterization in stents with implantation duration >30 days exhibiting specific imaging features revealed great variability of neointimal tissue components, except in the case of restenosis tissues, for which the specific imaging patterns seemed to correlate well with specific histological findings. In nonrestenotic tissues, it is essential to be aware of the differential diagnosis of distinct OCT and OFDI findings to avoid inappropriate image interpretation.
WHAT IS KNOWN? OCT imaging and OFDI are often used to evaluate vascular responses after stent implantation. Descriptive terms such as “heterogeneous,” “homogenous,” and “peristrut low intensity” have been used to describe neointimal characteristics, even though only limited histopathological correlations have been conducted.
WHAT IS NEW? Our study demonstrates that a wide variety of histological differential diagnoses underlie many commonly used neointimal OCT and OFDI patterns, which necessitates careful interpretation.
WHAT IS NEXT? Further understanding of which histological findings OCT imaging can reliably predict is needed.
The authors thank all members at CVPath Institute for their technical support.
CVPath Institute, a private, nonprofit research organization in Gaithersburg, Maryland, provided full support for this study. CVPath Institute has research grants from Abbott Vascular, Atrium Medical, Boston Scientific, Biosensors International, Cordis/Johnson & Johnson, Medtronic CardioVascular, OrbusNeich Medical, and Terumo. Dr. Joner is a consultant for Biotronik; receives grant support from the European Commission; and has received speaking honoraria from Abbott Vascular, Biotronik, Boston Scientific, and OrbusNeiche. Dr. Virmani has speaking engagements with Merck; receives honoraria from Abbott Vascular, Boston Scientific, Lutonix, Medtronic, and Terumo; and is a consultant for 480 Biomedical, Abbott Vascular, Medtronic, and W.L. Gore. Dr. Finn has sponsored research agreements with Boston Scientific and Medtronic CardioVascular; and is an advisory board member for Medtronic CardioVascular. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- bare-metal stent(s)
- drug-eluting stent(s)
- optical coherence tomographic
- optical frequency-domain imaging
- peristrut low attenuation
- smooth muscle cell
- Received April 12, 2016.
- Revision received July 22, 2016.
- Accepted September 8, 2016.
- American College of Cardiology Foundation
- Kawasaki M.,
- Bouma B.E.,
- Bressner J.,
- et al.
- Yabushita H.,
- Bouma B.E.,
- Houser S.L.,
- et al.
- Nakano M.,
- Otsuka F.,
- Yahagi K.,
- et al.
- Habara M.,
- Terashima M.,
- Nasu K.,
- et al.
- Prati F.,
- Guagliumi G.,
- Mintz G.S.,
- et al.
- Shibuya M.,
- Fujii K.,
- Hao H.,
- et al.
- Kim J.S.,
- Afari M.E.,
- Ha J.,
- et al.
- Kang S.J.,
- Nakano M.,
- Virmani R.,
- et al.
- Otsuka F.,
- Byrne R.A.,
- Yahagi K.,
- et al.
- Otsuka F.,
- Yahagi K.,
- Ladich E.,
- et al.
- Virmani R.,
- Guagliumi G.,
- Farb A.,
- et al.
- Lee S.Y.,
- Hong M.K.,
- Mintz G.S.,
- et al.
- Kim J.S.,
- Lee J.H.,
- Shin D.H.,
- et al.
- Phipps J.E.,
- Vela D.,
- Hoyt T.,
- et al.