Author + information
- Received February 24, 2012
- Revision received August 13, 2012
- Accepted August 21, 2012
- Published online January 1, 2013.
- Zhijing Zhao, MD⁎,
- Bernhard Witzenbichler, MD†,
- Gary S. Mintz, MD‡,
- Markus Jaster, MD†,
- So-Yeon Choi, MD, PhD⁎,
- Xiaofan Wu, MD, PhD⁎,
- Yong He, MD⁎,
- M. Pauliina Margolis, MD, PhD§,
- Ovidiu Dressler, MD⁎,
- Ecaterina Cristea, MD‖,
- Helen Parise, ScD⁎,
- Roxana Mehran, MD¶,
- Gregg W. Stone, MD⁎ and
- Akiko Maehara, MD⁎,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Akiko Maehara, Cardiovascular Research Foundation, 111 East 59th Street, New York, New York 10022
Objectives The authors sought to report the temporal stability of an untreated, nonculprit lesion phenotype in patients presenting with ST-segment elevation myocardial infarction (STEMI).
Background The temporal stability of the untreated, nonculprit lesion phenotype has been studied using intravascular ultrasound-virtual histology (IVUS) in patients with stable ischemic heart disease, but not in STEMI patients.
Methods As part of a formal substudy of the HORIZONS-AMI (Harmonizing Outcomes With Revascularization and Stents in Acute Myocardial Infarction) trial, baseline and 13-month follow-up IVUS was performed in 99 untreated nonculprit lesions in 63 STEMI patients. Lesions were classified as pathological intimal thickening (PIT), IVUS–derived thin-cap fibroatheroma (TCFA), thick-cap fibroatheroma (ThCFA), fibrotic plaque, or fibrocalcific plaque.
Results The frequency of TCFA increased from 41% at baseline to 54% at follow-up, whereas ThCFAs decreased from 41% to 34% and PIT decreased from 16% to 8%. Among the 41 lesions classified at baseline as TCFA, at follow-up, 32 (78%) were still classified as TCFA, whereas 9 (22%) were classified as ThCFAs or fibrotic plaques. An additional 21 lesions at follow-up were newly classified as TCFA, developing from either PIT or ThCFA. TCFA at baseline that evolved into non-TCFAs trended toward a more distal location than TCFA that did not change (p = 0.12). In lesions classified as TCFA, the minimum lumen area (MLA) decreased from 8.1 (interquartile range [IQR]: 7.4 to 8.8) mm2 at baseline to 7.8 (IQR: 7.2 to 8.4) mm2 at follow-up, p < 0.05; this was associated with an increase in percent necrotic core at the MLA site (14% [IQR: 12 to 16] to 19% [IQR: 17 to 22], p < 0.0001) and over the entire length of the lesion (14% [IQR: 12 to 16] to 18% [IQR: 17 to 20], p < 0.0001).
Conclusions Untreated nonculprit lesions in STEMI patients frequently have TCFA morphology that does not change during 13-month follow-up and is accompanied by a decrease in MLA and an increase in necrotic core. (Harmonizing Outcomes With Revascularization and Stents in Acute Myocardial Infarction [HORIZONS-AMI]; NCT00433966)
Autopsy data suggest that rupture of an atheromatous plaque with superimposed thrombosis is the cause of most acute coronary syndromes (ACS) and sudden cardiac death (1–3). A thin-cap fibroatheroma (TCFA) with a large lipid-rich necrotic core (NC) is particularly prone to rupture and result in coronary artery thrombotic occlusion (4–7). Furthermore, 3-vessel imaging studies in patients with ACS have shown that the incidence of secondary plaque rupture in nonculprit lesions is fairly common (8). Intravascular ultrasound-virtual histology (IVUS) provides quantitative and qualitative information about plaque composition and lesion phenotype, including plaques prone to cause events (9–12), and may therefore be useful to assess serial plaque composition evolution. A previous IVUS study in patients, mostly with stable ischemic heart disease, demonstrated that the IVUS lesion phenotype can evolve and, specifically, that virtual histology–derived thin-cap fibroatheromas (TCFA) can develop and/or heal within 1 year (13). No such studies have been performed in patients with ACS. In the present study, therefore, we used serial (baseline and follow-up) IVUS to assess the dynamic nature of nonculprit coronary artery lesion morphology in patients with ST-segment elevation myocardial infarction (STEMI), in particular to evaluate the development and evolution of TCFA.
HORIZONS-AMI (Harmonizing Outcomes With Revascularization and Stents in Acute Myocardial Infarction) was a prospective, open-label, multicenter, dual-arm, 2 × 2 factorial randomized trial in patients with STEMI. The 2 randomization arms consisted of: 1) the direct thrombin inhibitor bivalirudin alone versus heparin plus a glycoprotein IIb/IIIa inhibitor (1:1 randomization); and 2) TAXUS EXPRESS paclitaxel-eluting stents versus otherwise equivalent EXPRESS bare-mental stents (3:1 randomization) (Boston Scientific, Natick, Massachusetts). The present study cohort consists of 63 patients (99 lesions) with serial IVUS examinations of at least 1 untreated nonculprit lesion in a native coronary artery at baseline and at follow-up that were studied at Charitė University Medicine Berlin (Campus Benjamin Franklin, Berlin, Germany) where the ethics committee approved the protocol, and written informed consent was obtained from all patients.
Clinical follow-up was performed at 30 days, 6 months, 1 year, 2 years, and 3 years. Routine angiographic (with or without IVUS) follow-up was performed at 13 months or earlier if a clinical event occurred. Pre-specified endpoints included ischemia-driven revascularization and the composite of major adverse cardiac events (death, reinfarction, stroke, or stent thrombosis). An independent clinical events committee (Cardiovascular Research Foundation [CRF], New York, New York) that was masked to treatment assignment adjudicated all adverse events from original source documents and procedural angiograms.
IVUS image acquisition
After intracoronary nitroglycerin, a phased-array, 20-MHz, 3.2-F IVUS catheter (Eagle Eye, Volcano Corporation, Rancho Cordova, California) was placed into the distal coronary artery and pulled back to the aorto-ostial junction using an R-100 motorized catheter pullback device (Volcano Corporation) at a speed of 0.5 mm/s. During pull back, grayscale IVUS was recorded, raw radiofrequency data were captured at the top of the R-wave, and reconstruction of the color-coded map by an IVUS data recorder was performed (In-Vision Gold, Volcano Corporation). The grayscale IVUS and captured radiofrequency data were written onto a CD-R or DVD-R and sent to an independent IVUS core laboratory (CRF) for quantitative and qualitative analyses.
Grayscale and IVUS analyses
Off-line grayscale and IVUS analyses were performed using: 1) QCU-CMS (Medis, Leiden, the Netherlands) for contouring; 2) pcVH 2.1 software (Volcano Corporation) for contouring and data output; and 3) proprietary qVH software (qVH 2.5, developed and validated at CRF) for qualitative segmental assessment and quantitative data output. Corresponding baseline and follow-up images were identified by the distance from 2 fiduciary landmarks such as side branches and stent edges. Quantitative grayscale IVUS measurements of lumen, external elastic membrane (EEM), and plaque and media (defined as EEM minus lumen) cross-sectional area (CSA) and plaque burden (defined as plaque and media divided by EEM) were performed for every recorded frame. IVUS analysis was also performed for every recorded frame. Volumes were calculated using Simpson's rule and normalized for analysis length. The 4 IVUS plaque components were color-coded as white (dense calcium [DC]), red (NC), light green (fibrofatty [FF]), and dark green (fibrotic tissue [FT]), and reported as CSA and percentage of total plaque CSA and volume. Volumetric grayscale and IVUS analyses were performed only for lesions in which motorized pull back was consistent and reliable.
Definition of lesion types
Lesions were classified by 2 experienced, independent observers (Z.Z. and A.M.) on the basis of plaque composition. Interobserver variability was good (κ = 0.87), and disagreements were resolved by consensus discussion. Fibroatheromas (TCFA and thick-cap fibroatheroma [ThCFA]) contained >10% confluent NC in at least 3 consecutive frames. Because the window size currently applied for the selection of the region of interest and the eventual tissue map or IVUS image reconstruction using this system and catheter is approximately 246 μm in the radial direction, it was not possible to detect fibrous cap thickness <65 μm (the typical pathological definition of a thin fibrous cap) (9). Therefore, TCFA was defined as a fibroatheroma without virtual histology evidence of a fibrous cap, with >30° NC abutting the lumen in at least 3 consecutive frames (12,14). ThCFA was defined as a fibroatheroma with a definable fibrous cap. Fibrotic plaque had mainly FT with <10% confluent NC, <15% FF tissue, and <10% confluent DC. Fibrocalcific plaque had mainly FT with >10% confluent DC, but <10% confluent NC. All other plaques were classified as pathological intimal thickening (PIT), which was mainly a mixture of FT and FF (>15%) with <10% confluent NC and <10% confluent DC. Fibroatheromas were considered multiple and distinct if they were separated by at least 3 consecutive image slices containing a different fibroatheroma subtype (TCFA and/or ThCFA) or a nonfibroatheroma phenotype.
IVUS phenotypes and their evolution were assessed at 2 levels: lesion level and fibroatheroma level. For lesion-level IVUS classification, the entire lesion was evaluated. We identified untreated, nonculprit lesions as having a plaque burden ≥40% in at least 3 consecutive frames (≥1.5 mm in length) and >5 mm from the stent edge. Lesions were considered separate if there was a ≥5-mm-long segment with <40% plaque burden between them. A lesion could contain more than 1 fibroatheroma. Fibroatheromas were considered multiple and distinct if they were separated by at least 3 consecutive frames containing a different lesion phenotype (14). In lesions that contained more than 1 fibroatheroma subtype, the following hierarchy was used: a TCFA took precedence over ThCFA, and any fibroatheroma took precedence over any nonfibroatheroma. If there was no fibroatheroma in the entire lesion, the virtual histology classification at the minimum lumen area (MLA) site was evaluated and defined as fibrotic plaque, fibrocalcific plaque, or PIT. Because a lesion could contain more than 1 fibroatheroma, the evolution of each fibroatheroma within a lesion was also assessed.
Statistical analysis was performed using SAS software, version 9.1 (SAS Institute, Cary, North Carolina). Categorical variables are presented as frequencies and were compared with the chi-square or Fisher exact test. Continuous variables were presented as medians and interquartile ranges, and were compared using the Wilcoxon signed rank test. For lesion- and fibroatheroma-level data, a model with a generalized estimating equation approach was used to compensate for any potential cluster effect of multiple fibroatheromas or lesions in the same patient using an exchangeable working correlation structure; data were presented as least square means with 95% confidential interval (15). A p value <0.05 was considered statistically significant.
Patient and lesion characteristics
Overall, 99 nonculprit, untreated lesions in 73 target vessels in 63 patients were identified and analyzed by grayscale and IVUS at baseline and follow-up. A total of 59 nonculprit lesions were located in the right coronary artery, 19 nonculprit lesions were located in the left anterior descending coronary artery, 9 nonculprit lesions were located in the left circumflex coronary artery, and 12 nonculprit lesions were located in the left main coronary artery. There were 3 predominant IVUS phenotypes at baseline: TCFA (n = 41), ThCFA (n = 41), or PIT (n = 16), whereas 1 additional lesion was classified as fibrocalcific. At follow-up, there were 3 new lesions; 1 was a ThCFA and 2 were PITs.
Baseline clinical characteristics are listed in Table 1. From the baseline to follow-up imaging study, the MLA decreased significantly (from 8.1 [7.4, 8.8] mm2 to 7.8 [7.2, 8.4] mm2, p = 0.01); this was associated with an increase in the percent NC, whether assessed at the MLA site (14% [12, 16] to 19% [17, 22], p < 0.0001), over the entire length of the lesion (14% [12, 16] to 18% [17, 20], p < 0.0001), or at the maximum NC site (22% [20, 24] to 27% [25, 29], p < 0.0001) (Table 2). The change of NC volume was negatively correlated to the change of fibrofatty plaque volume (R = −0.783, p < 0.0001) The frequency of lesions that were classified as TCFA increased from 41% at baseline to 54% at 13-month follow-up, whereas the proportion of ThCFAs and PIT lesions decreased from 41% to 34% and from 16% to 8%, respectively (Table 3). Among the TCFA lesions, 12.5% (5 of 40) of lesions at baseline and 7.8% (4 of 51) of lesions at follow-up had an MLA ≤4 mm2; and 7.5% (3 of 40) of lesions at baseline and 13.7% of lesions (7 of 51) at follow-up had a plaque burden of >70% at the MLA site.
Evolution of lesion-level IVUS phenotype
The temporal progression of nonculprit lesion-level IVUS phenotype is shown in Table 3. At baseline, 41 lesions were classified as TCFA. At follow-up, 32 lesions (78%) were still classified as TCFA, whereas 9 lesions (22%) evolved into ThCFAs or fibrous plaques. An additional 21 lesions were newly classified as TCFA at 13-month follow-up, developing from either PIT (n = 3, 19% of baseline PITs) or ThCFA (n = 18, 44% of baseline ThCFAs) phenotypes. Among the 32 TCFA that did not change their overall phenotypical classification during follow-up, 37% had an evolution in appearance suggesting an increase in instability: 1) progression (n = 12, 37%), an increase in arc or length of NC abutting to the lumen (implying an increase in the size of the thin cap of the fibroatheroma), or an increase in the length of the TCFA; or 2) decrease and increase of fibrous cap thickness (n = 4, 13%), a heterogeneous response such that part of the TCFA had a decrease in NC abutting the lumen (increase of fibrous cap thickness) or a reduction in TCFA length, while part of the TCFA had a decrease in fibrous cap thickness. Conversely, there was a partial increase of the fibrous cap thickness without any progression in 6 TCFA (19%) and no change in 10 TCFA (31%). This is summarized in Figure 1.
At baseline, there were 41 nonculprit lesions that were classified as ThCFAs. At follow-up, 18 of these ThCFAs evolved into TCFA (Fig. 2) with 2 different patterns: 1) new TCFA within the lesion, either circumferentially (n = 7, 39%) or longitudinally (n = 6, 33%), but at a site unrelated to the baseline ThCFA; or 2) disappearance of fibrous cap at the site of the baseline ThCFA, converting the lesion into a TCFA (n = 5, 28%). Conversely, 20 (49%) baseline ThCFAs were still classified as ThCFA at follow-up, and 3 (7%) evolved into other stable phenotypes.
At baseline there were 16 lesions that were classified as PIT. Of these lesions, 3 (19%) evolved into a TCFA, 76 (43%) evolved into a ThCFA, and 6 (38%) were still classified as a PIT at follow-up.
TCFA distribution at the fibroatheroma level
As shown in Figure 3, approximately 50% of TCFA were located in the proximal 20 mm of the coronary artery both at baseline and at follow-up. Using the distance from the coronary ostium to the maximum NC site, TCFA that healed during follow-up tended to be more distally located than those that did not change (36.2 [24.2, 48.2] mm vs. 25.7 [18.9, 32.6] mm, p = 0.12). The distance between newly developed TCFA at follow-up compared with persistent TCFA from baseline was not significantly different: 27.8 [18.9, 36.8] mm versus 24.7 [17.7, 31.7] mm, p = 0.50.
IVUS findings of patients with late clinical events
During 3 years of follow-up, 4 unplanned revascularizations were required for progressive or unstable angina arising from these 99 nonculprit lesions. In 2 cases, the phenotypes progressed from ThCFAs to TCFA at the time of event; in 1 case, the phenotype progressed from PIT to ThCFA; and 1 case was a TCFA both at baseline and at the time of the event. The grayscale and IVUS findings in these 4 lesions are shown in Table 4.
This is the first report, to our knowledge, in which serial (baseline and follow-up) IVUS was used to assess the dynamic nature of nonculprit coronary artery lesion morphology in patients presenting with STEMI. The major findings of our analysis are: 1) more than 82% of untreated, nonculprit lesions in patients with STEMI were fibroatheromas at baseline (half of which were TCFAs), with the majority demonstrating a lack of healing and/or an increase in unstable phenotype during 13 months of follow-up; and 2) fibroatheromas that became more stable in appearance over time tended to be located more distally, whereas fibroatheromas that did not change or that became more unstable tended to be located more proximally.
Characterization of the nonculprit lesion phenotype
In the current prospective study in patients with STEMI, 82% of secondary, nonculprit lesions were fibroatheromas, with half of these (41%) having a TCFA phenotype at baseline. The high rate of fibroatheromas are consistent with an earlier study by Asakura et al. (16), who reported that in 20 noninfarcted vessels, yellow, nondisrupted plaques as seen by angioscopy were present in most patients 1 month after a myocardial infarction (MI). By contrast, in the IVUS study by Kubo et al. (13) in patients mainly with stable ischemic heart disease, only 43% and 9% of secondary lesions had a ThCFA and TCFA phenotype at baseline, respectively. Thus, these data suggest that patients with STEMI compared with those without acute coronary syndromes are significantly more likely to have fibroatheromas and TCFAs at untreated, nonculprit lesion sites. These studies are also consistent with reports of simultaneous plaque ruptures in some patients with STEMI (8,17).
Other intravascular imaging studies have compared secondary lesions in stable versus unstable patients. Nakamura et al. (18) showed that TCFA were more common in ACS patients compared with stable patients (64.6% vs 35.7%, p = 0.006). Hong et al. (8) performed a 3-vessel grayscale IVUS study of 235 patients (122 acute MI, 113 stable angina); secondary plaque ruptures were present in 21 MI patients (17%) versus 6 stable patients (5%, p = 0.008). In a 3-vessel IVUS study, Hong et al. (19) reported a higher frequency of TCFA and plaque ruptures in secondary lesions in patients with ACS compared with stable angina (69% vs. 45%, and 8% vs. 4%, respectively, both p < 0.001). Using integrated backscatter IVUS (another radiofrequency IVUS technique), Ando et al. (20) showed a larger percent lipid area (p = 0.03) and a smaller percent fibrotic plaque area (p = 0.04) in nonculprit lesions in ACS patients compared with stable angina patients. Finally, Kubo et al. (21), using optical coherence tomography (OCT), reported that fibrous cap thickness (111 ± 65 vs. 181 ± 70 μm, p = 0.002) was significantly thinner and the frequency of OCT-derived TCFA (38% vs. 6%, p = 0.030) was significantly greater in acute MI than in stable angina.
Serial intravascular imaging assessment of nonculprit lesion phenotype
There have been few serial intravascular imaging studies assessing changes in lesion phenotype of unruptured, secondary plaques in stable or unstable patients (13,22–25). Using serial (baseline and follow-up) IVUS in patients with mostly stable ischemic heart disease, Kubo et al. (13) reported that most nonculprit TCFA stabilized or healed during 13 months of follow-up. By contrast, in the current study in patients with STEMI, nonculprit TCFA remained unchanged or became more unstable over time, while more stable lesions evolved into more unstable phenotypes. Of note, both of these studies were analyzed at the same core laboratory (CRF) using the same analytic techniques and definitions. A comparison of these 2 studies showing the difference between stable and unstable patients is shown in Figure 4.
The current study is also consistent with 2 small serial angioscopy studies. In 1 study in 48 patients with ACS and previous MI, the 12-month healing rate of nonculprit, ruptured plaques was only 30% (22). In the second study in 13 consecutive acute MI patients, the number of nonculprit yellow plaques did not change from post-reperfusion to 6-month follow-up (24).
TCFA location and “increase of fibrous cap.”
In the current study of patients with STEMI, approximately 50% of TCFA were in the proximal 20 mm of a coronary artery at baseline. Numerous pathological, angiographic, and imaging studies have shown a similar, proximal predisposition of TCFAs, acute occlusions, or plaque ruptures. In a human autopsy study, Cheruvu et al. (5) reported that 50% of TCFAs were present within the first 22 mm of the left anterior descending and left circumflex arteries and the first 31 mm of the right coronary artery. In another ex vivo study using histology, IVUS, and OCT, Kume et al. (26) reported that 70% of TCFAs clustered in the first 30 mm of a coronary artery. Wang et al. (27) analyzed 208 consecutive STEMI patients and found that angiographic occlusions tended to cluster within the proximal third of each coronary artery. In a 3-vessel grayscale IVUS study, Hong et al. (28) evaluated the axial location of plaque rupture in 392 patients; plaque ruptures occurred mainly in proximal segments of the left anterior descending coronary artery, the proximal half of the left circumflex, and the proximal and distal segments of right coronary artery. In a subsequent IVUS analysis, Hong et al. (20) also reported that 83% of TCFA were located mainly within 40 mm of the coronary ostium. The current study extends these findings by showing that proximal TCFA tended to progress to more unstable phenotypes, whereas distal TCFA did not.
In the current study, IVUS phenotype evolution was assessed at 2 levels: lesion level and fibroatheroma level because a single lesion may contain more than 1 discrete fibroatheroma. If there was more than 1 fibroatheroma subtype within a given lesion, a TCFA took precedence over ThCFA; thus, for example, even if a lesion contained multiple ThCFAs at baseline, the evolution of only 1 ThCFA into a TCFA changed the overall lesion phenotype classification into a TCFA. Such heterogeneous changes at the fibroatheroma level increased the frequency of TCFA classification significantly, from 41% at baseline to 54% at 13-month follow-up at the lesion level.
Several limitations of the present study should be noted. Despite the fact that the HORIZONS-AMI is one of the largest IVUS multicenter trials, this substudy is of modest size and from a single center. Potential selection bias cannot be excluded. Most of the patients had IVUS in only 1 vessel, and none had 3-vessel imaging. The present results should therefore be considered exploratory and hypothesis generating. Details on lipid levels and corresponding pharmacotherapy are not available. IVUS generates only 1 image per every heart beat, thereby severely limiting coregistration and potentially causing misalignment between baseline and follow-up images. There were too few clinical events to permit a well-powered exploration of the relationship between recurrent events and plaque morphology as disclosed by IVUS. The anatomic locations and distributions of the lesions in this study vary from the typical distribution of culprit lesions as well as from the distribution of culprit lesions in the HORIZONS trial. As a result, lesions in this substudy may be subjected to different shear stress and other atherogenic factors that may influence progression compared with culprits There was no control group, and the number of new TCFA was too small to identify predictors of the transition from stable to vulnerable phenotype.
Untreated, nonculprit lesions in patients with STEMI are frequently fibroatheromas and typically become increasingly unstable during 13-month follow-up, with a decrease in MLA, increase in NC, and overall devolution to a TCFA morphology.
Dr. Zhao has received research grants from Boston Scientific China. Dr. Witzenbichler has received speaker's fees (modest) from Boston Scientific and The Medicines Company. Dr. Mintz has received grant support from Volcano Corporation and Boston Scientific; and consulting fees from Volcano Corporation and Boston Scientific. Drs. Wu and He have received research grants from Boston Scientific China. Dr. Margolis was an employee of Volcano Corporation at the time the manuscript of this paper was submitted and holds stock in the company. Dr. Mehran has received research grants from BMS/Sanofi, The Medicines Company, and Lilly/Daiichi Sankyo; serves on the advisory board for Regado Biosciences; and is a consultant for AstraZeneca, Abbott, Johnson & Johnson, Merk Sharp & Dohme, and Maya Medical. Dr. Stone is a consultant for Volcano Corporation. Dr. Maehara has received speaker's fees from St. Jude Medical; and grant support from Boston Scientific. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- acute coronary syndromes
- Cardiovascular Research Foundation
- cross-sectional area
- dense calcium
- external elastic membrane
- fibrotic tissue
- intravascular ultrasound-virtual histology
- myocardial infarction
- minimum lumen area
- necrotic core
- optical coherence tomography
- pathological intimal thickening
- ST-segment elevation myocardial infarction
- thin-cap fibroatheroma
- thick-cap fibroatheroma
- Received February 24, 2012.
- Revision received August 13, 2012.
- Accepted August 21, 2012.
- American College of Cardiology Foundation
- Virmani R.,
- Burke A.P.,
- Farb A.,
- Kolodgie F.D.
- Virmani R.,
- Kolodgie F.D.,
- Burke A.P.,
- Farb A.,
- Schwartz S.M.
- Cheruvu P.K.,
- Finn A.V.,
- Gardner C.,
- et al.
- Finn A.V.,
- Nakano M.,
- Narula J.,
- Kolodgie F.D.,
- Virmani R.
- Kolodgie F.D.,
- Virmani R.,
- Burke A.P.,
- et al.
- Hong M.K.,
- Mintz G.S.,
- Lee C.W.,
- et al.
- Rodriguez-Granillo G.A.,
- García-García H.M.,
- McFadden E.P.,
- et al.
- Nasu K.,
- Tsuchikane E.,
- Katoh O.,
- et al.
- Kubo T.,
- Maehara A.,
- Mintz G.S.,
- et al.
- Maehara A.,
- Cristea E.,
- Mintz G.S.,
- et al.
- Lian K.Y.,
- Zeger S.L.
- Asakura M.,
- Ueda Y.,
- Yamaguchi O.,
- et al.
- Takano M.,
- Inami S.,
- Ishibashi F.,
- et al.
- Wang J.C.,
- Normand S.L.,
- Mauri L Kuntz R.E.
- Hong M.K.,
- Mintz G.S.,
- Lee C.W.,
- et al.