JACC: Cardiovascular Imaging
CT Angiography for the Detection of Coronary Artery Stenoses in Patients Referred for Cardiac Valve SurgerySystematic Review and Meta-Analysis
Author + information
- Received September 8, 2015
- Accepted September 24, 2015
- Published online September 1, 2016.
Author Information
- Maksymilian P. Opolski, MDa,∗ (opolski.mp{at}gmail.com),
- Adam D. Staruch, MDa,
- Michal Jakubczyk, PhDb,
- James K. Min, MDc,
- Heidi Gransar, MSd,
- Michal Staruch, MSe,
- Adam Witkowski, MDa,
- Cezary Kepka, MDf,
- Won-Keun Kim, MDg,
- Christian W. Hamm, MDg,h,
- Helge Möllmann, MDg and
- Stephan Achenbach, MDi
- aDepartment of Interventional Cardiology and Angiology, Institute of Cardiology, Warsaw, Poland
- bInstitute of Econometrics, Warsaw School of Economics, Warsaw, Poland
- cDepartment of Radiology, The New York-Presbyterian Hospital and Weill Cornell Medical College, New York, New York
- dDepartment of Imaging and Division of Cardiology, Department of Medicine, Cedars-Sinai Heart Institute, Cedars-Sinai Medical Center, Los Angeles, California
- eMedical University of Warsaw, Warsaw, Poland
- fDepartment of Coronary and Structural Heart Diseases, Institute of Cardiology, Warsaw, Poland
- gDepartment of Cardiology, Kerckhoff Heart Center, Bad Nauheim, Germany
- hDepartment of Cardiology and Angiology, Justus-Liebig University of Giessen, Giessen, Germany
- iDepartment of Internal Medicine 2 (Cardiology), University of Erlangen, Erlangen, Germany
- ↵∗Reprint requests and correspondence:
Dr. Maksymilian P. Opolski, Department of Interventional Cardiology and Angiology, Institute of Cardiology, Alpejska 42, Warsaw 04-628, Poland.
Abstract
Objectives This study aimed to evaluate the diagnostic accuracy of coronary computed tomography angiography (CTA) for detecting coronary artery stenoses in patients with valvular heart disease undergoing valve surgery.
Background Coronary CTA is currently not routinely recommended for detecting coronary artery stenoses before cardiac valve surgery. However, recent improvements in computed tomography technology may enable the identification of the most appropriate candidates for coronary CTA before valve surgery.
Methods A systematic review was performed of PubMed, EMBASE, and the Cochrane databases for all studies that used ≥16-detector row computed tomography scanning to perform coronary CTA in patients with valvular heart disease scheduled for valve surgery and validated the results against invasive angiography. Summary diagnostic accuracies were calculated by using a bivariate random effects model, and a generalized linear mixed model was applied for heterogeneity analysis.
Results Seventeen studies analyzing 1,107 patients and 12,851 coronary segments were included. Patient-based analysis revealed a pooled sensitivity of 93% (95% confidence interval [CI]: 86 to 97), specificity of 89% (95% CI: 86 to 91), a negative likelihood ratio (LR) of 0.07 (95% CI: 0.04 to 0.16), and a positive LR of 8.44 (95% CI: 6.49 to 10.99) for coronary CTA to identify individuals with stenosis ≥50%. Specificity and positive LR were higher in patients without aortic stenosis (AS) versus those with AS (96% vs. 87% and 21.2 vs. 7.4, respectively), as well as with ≥64 detectors versus <64 detectors (90% vs. 86% and 9.5 vs. 6.9). Heterogeneity analysis revealed a significant impact of AS and the number of detectors on specificity of CTA.
Conclusions Coronary CTA using currently available technology is a reliable imaging alternative to invasive angiography with excellent sensitivity and negative LR for the detection of significant coronary stenoses in patients undergoing cardiac valve surgery. The specificity of coronary CTA may be decreased against the background of AS (Computed Tomography Angiography for the Detection of Coronary Artery Disease in Patients Referred for Cardiac Valve Surgery: A Meta-Analysis; CRD42015016213)
Valvular heart disease (VHD) that requires surgery is common in industrialized countries (1). In the majority of such patients, pre-operative evaluation for coronary artery disease (CAD) with invasive coronary angiography (ICA) is recommended (1,2), although most patients are found to have no significant coronary stenoses (3). In recent years, several studies have tested the diagnostic performance of coronary computed tomography angiography (CTA) in patients undergoing cardiac valvular surgical treatment (4–20). However, all of these studies were limited by small sample sizes in single centers, which may introduce bias that obfuscates the actual diagnostic performance of coronary CTA compared with ICA. We thus performed a comprehensive meta-analysis of all available studies comparing coronary CTA with ICA for the detection of significant coronary stenoses in patients scheduled for elective valve surgery.
Methods
Search strategy
This study followed the protocol specified in the Preferred Reporting Items for Systematic Reviews and Meta-analyses statement (21). Database searches for papers reporting on patients with VHD, pre-operatively examined by using coronary CTA and invasive angiography, were performed by 2 independent reviewers (A.D.S. and M.S.). The PubMed, EMBASE, and Cochrane Library databases were used. We used the search terms and corresponding Medical Subject Headings for “computed tomography” AND “heart valve disease.” The search was limited to English-language studies of humans and adults (>18 years of age) in peer-reviewed journals up to July 2015. The references of all papers retrieved were also searched. All studies were carefully examined to exclude potential duplicates or overlapping data. Trials in abstract form without a published manuscript were excluded.
Eligibility criteria
We included a study if: 1) it used at least 16-slice computed tomography (CT) scan as a diagnostic test for the detection of coronary stenoses in the pre-operative assessment of patients referred for heart valve surgery; 2) it selected 50% diameter stenosis as the cut-off criterion for significant CAD using ICA as the standard of reference; and 3) absolute numbers of true-positive, true-negative, false-positive, and false-negative results were provided or could be derived at the per-patient or per-segment level. Studies were excluded if they were performed in patients with aortic valve stenosis (AS) referred for transcatheter aortic valve replacement.
Data extraction and quality assessment
Data extraction of the selected studies was performed by 2 independent investigators (M.P.O. and A.D.S) using a standardized data extraction form. The form included the following characteristics of each trial: first author, year of publication, and journal; study population characteristics, including sample size (number of subjects screened and evaluated with both tests, and number of subjects excluded), sex, mean age, mean body mass index, type of VHD, and mean heart rate during CTA scanning; prevalence of CAD; and technical characteristics of the intervention (number of detector rows, effective radiation dose, beta-blocker or ivabradine use, nitroglycerin use, and amount of contrast agent). Differences between reviewers were resolved by consensus whenever necessary. The methodological quality of included studies was assessed by 2 independent reviewers (M.P.O. and A.D.S.) using the modified Quality Assessment of Studies of Diagnostic Accuracy Included in Systematic Reviews–2 criteria (22), and discrepancies were resolved by consensus.
Statistical analysis
Categorical variables from individual studies are presented as frequency (%), and continuous variables are presented as mean ± SD. Measures of diagnostic accuracy are reported as point estimates with 95% confidence intervals (CIs). The primary analysis was performed at the per-patient level, with secondary analyses evaluating the available per-segment and per-vessel data. Sensitivity, specificity, and positive and negative likelihood ratios (LRs) were computed based on the true-positive, true-negative, false-positive, and false-negative rates for each study. Both LRs are independent from prevalence rates, and there is consensus that a positive LR >10 and a negative LR <0.1 provide reliable evidence of satisfactory diagnostic performance (23). The ratio of positive LR to negative LR was combined in a single global accuracy measure, the diagnostic odds ratio (24).
Summary sensitivity and specificity, positive and negative LRs, and diagnostic odds ratios were estimated by using a bivariate random effects model. This approach assumes bivariate normal distributions for the logit transformations of sensitivity and specificity from individual studies (25,26). In addition, the hierarchical summary receiver-operating characteristic (ROC) curve presenting the point estimates for each study, the joint ROC curve, and the pooled characteristics, including the 95% confidence region and the 95% prediction region, was constructed (27).
Sources of clinical and statistical heterogeneity were explored at the patient level by using the bivariate generalized linear mixed model (2 outcomes, sensitivity and specificity, modeled jointly) as previously described (28,29). We assessed the following covariates, which were selected a priori: sample size (divided by a median of 50 patients), average age of patients (divided by a median of 67 years), percentage of male subjects (divided by a median of 60%), number of CT detector rows (defined as either <64 detectors or ≥64 detectors), and the presence of any patient with AS. Publication bias was assessed for each analysis by using the Deeks test (30). The κ statistic was calculated for the inter-rater reliability between 2 investigators for quality assessment. Statistical computations were performed using Stata version 14 (Stata Corporation, College Station, Texas) and R version 3.1.0 statistical software (R Foundation for Statistical Computing, Vienna, Austria). Significance testing was at the 2-tailed 0.05 level.
Results
Study selection
After exclusion of duplicates, the systematic review retrieved 9,379 potentially relevant papers that were screened according to title or abstract for possible inclusion (Figure 1). Thirty-four full-text reports were selected for further evaluation, from which 17 studies were excluded for the following reasons: no systematic ICA was performed (9 papers)—it was either completely absent (31,32) or was used only to confirm abnormal and/or nondiagnostic CTA findings (33–39); data to calculate diagnostic accuracies were not provided or could not be derived (4 papers) (40–43); no reference to valve surgery was present (1 paper) (43); and a heterogeneous patient population inclusive of patients without VHD was included (4 papers) (44–47) or no values for true-positive and false-negative observations were reported (i.e., variance is infinite and CIs as well as sensitivity could not be computed) (2 papers) (44,45). Seventeen studies were included in the final analysis (Table 1).
Flow Diagram of the Review Process
Seventeen articles were excluded from the final analysis due to incomplete or unsystematic data.
Baseline Characteristics of Included Studies
Baseline characteristics
Overall, a total of 1,153 patients (mean age 65.0 ± 6.4 years; 60% male) were included. The mean heart rate at the time of coronary CTA was 65 beats/min (60 to 68 beats/min), and 13 studies used beta-blockers for reduction of heart rate. The use of contrast media ranged from 80 to 125 ml, and the radiation dose ranged from 2.55 to 26.0 mSv. Of the 17 studies, 4 studies analyzing 321 patients (mean age 70.0 ± 1.4 years; 57% male) evaluated only patients with AS (6,9,10,12). Of the remaining 13 studies, 9 studies analyzed 693 patients (mean age 66.0 ± 5.8 years; 59% male) with various types of VHD (the percentage of patients with AS ranged from 32% to 68%) (4,5,8,11,13,15–17,19). Four studies included 148 patients (mean age 59.0 ± 6.7 years; 65% male) with a single type of valvular disease that was not AS (2 studies with aortic regurgitation, 1 study with mitral regurgitation, and 1 study with tricuspid regurgitation) (7,14,18,20). In 10 studies analyzing 761 patients (mean age 63.0 ± 7.2 years; 61% male) at least a 64-slice CTA was used (4,7,8,10,11,13,14,18–20), whereas 7 studies evaluated 16- or 40-slice CTA in a total of 392 patients (mean age 68.0 ± 3.1 years; 59% male) (5,6,9,12,15–17). All studies reported a ≥50% stenosis cutoff to determine the presence of significant CAD (in 1 study that originally used a ≥70% stenosis cutoff, the corresponding author was approached to provide supplementary accuracy data for ≥50% stenosis cutoff) (10).
Quality assessment
Overall, the selected studies showed excellent quality in terms of applicability and satisfactory quality scores in terms of the risk of bias (Figure 2), with an inter-rater agreement of κ = 0.91. Risk of bias was high in 1 study regarding blinding the reference standard result from the assessor of the index test (11); however, because this study failed to provide diagnostic accuracy data on the per-patient basis, it was not included in the primary endpoint analysis. Four studies manifested an unclear risk of bias and did not explain whether the reference standard (3 papers) (7,14,20) or index test (2 papers) (5,20) was interpreted in a blinded fashion. Similarly, risk of bias regarding flow and timing was unclear in 6 studies because not all recruited patients were analyzed, and timing between the index test and the reference standard was unknown (5,7,10,11,19,20).
Assessment of Methodological Quality of Included Studies Using the QUADAS-2 Criteria
Stacked bars represent the proportion of studies with a low (green), high (red), or unclear (blue) risk of bias or concerns of applicability. QUADAS-2 = Quality Assessment of Studies of Diagnostic Accuracy Included in Systematic Reviews–2.
Publication bias
Using the Deeks test, there was no indication of publication bias for any of the analyses (p = 0.28 for the per-patient evaluation, p = 0.87 for the per-vessel evaluation, and p = 0.24 for the per-segment evaluation).
Patient-based analysis
In the per-patient analysis, results from 16 studies with a total of 1,107 patients, after exclusion of 24 patients (2%), were pooled. Overall, 313 patients (28%) had evidence of at least 1 significant stenosis in ICA. The pooled sensitivity, specificity, positive and negative LRs, and a diagnostic odds ratio for CTA were 93% (95% confidence interval [CI]: 86 to 97), 89% (95% CI: 86 to 91), 8.44 (95% CI: 6.49 to 10.99), 0.07 (95% CI: 0.04 to 0.16), and 113.74 (95% CI: 47.98 to 269.64), respectively (Figures 3, 4, and 5). Analyses of diagnostic accuracy, stratified according to the presence of AS, found more favorable specificity (96%) and positive LR (21.2) in studies without AS compared with specificity (87%) and positive LR (7.4) in studies with AS, with equally high values of sensitivity and negative LR in both groups. Similarly, an analysis stratified according to the number of CT detectors demonstrated more favorable specificity (90%) and positive LR (9.5) for ≥64 detectors compared with the specificity (86%) and positive LR (6.9) for <64 detectors, with equally high values of sensitivity and negative LR in both groups (Table 2). Statistical heterogeneity was found for sensitivity (I2 = 58.8%; p = 0.001), negative LR (I2 = 59.0%; p = 0.001), and diagnostic odds ratio (I2 = 100.0%; p = 0.001). Vessel-based and segment-based analyses are presented in the Online Appendix.
Forest Plots of Per-Patient Sensitivity and Specificity of CTA Compared With Invasive Coronary Angiography
The size of the square-plotting symbol is proportional to the size of the study. Horizontal lines are the 95% confidence intervals (CIs). CTA = computed tomography angiography.
Forest Plots of Per-Patient Positive and Negative LRs of CTA Compared With Invasive Coronary Angiography
LR = likelihood ratio; other abbreviations as in Figure 3.
Hierarchical SROC Plot Showing Average Sensitivity and Specificity Estimate of the Study Results at the Per-Patient Level With 95% Confidence Region
The 95% prediction region represents the confidence region for a forecast of the true sensitivity (SENS) and specificity (SPEC) in a future study. The red diamond indicates per-patient level. AUC = area under the curve; SROC = summary receiver-operating characteristic.
Summary Estimates for Pooled Measures of Diagnostic Accuracy of Coronary CTA for Evaluating the Presence of Significant Coronary Stenoses Compared With Invasive Coronary Angiography on a Per-Patient Basis
Heterogeneity analyses
The generalized linear mixed model showed significant interactions between the presence of ≥64 detectors, absence of patients with AS, fewer male subjects, and lower age, as well as smaller sample size and higher specificity of coronary CTA for the detection of significant coronary stenoses. Conversely, male sex was the only covariate with a negative effect on sensitivity (Figure 6). Statistical heterogeneity was present for AS (I2 = 69.0%; p = 0.04) and age (I2 = 71.0%; p = 0.03) only.
Graphical Presentation of the Generalized Linear Mixed Model Exploring the Impact of Selected Variables on Sensitivity and Specificity of CTA
Continuous variables were dichotomized according to its median values. as = aortic stenosis; det = detectors; ss = sample size; other abbreviations as in Figures 3 and 5.
Discussion
This meta-analysis investigated the diagnostic accuracy of coronary CTA with ≥16 slices for the detection of significant coronary stenoses in patients undergoing cardiac valve surgery. Our main finding was that coronary CTA using currently available technology can effectively rule out significant coronary stenoses in patients scheduled for valve surgery compared with the clinical standard of ICA, and it may thus serve as a gatekeeper for ICA before cardiac valve surgery. Importantly, the ability for coronary CTA to detect significant coronary stenoses in patients with VHD is not universal and was lower for patients with AS, a finding possibly reflective of the higher prevalence of atherosclerosis.
The presence of concomitant CAD in patients undergoing cardiac valvular surgery is related to worse clinical outcomes (48), and various clinical studies have suggested that combined valve and bypass surgery reduces early and late mortality (48,49). Consequently, pre-surgical assessment of CAD is essential for risk stratification and determination if concomitant coronary revascularization is indicated. According to current guidelines for the management of valvular heart disease, ICA remains the gold standard for the identification of coronary stenoses in all male patients >40 years of age, post-menopausal women, and pre-menopausal women with any risk factors for CAD (1,2). However, ICA imparts a non-negligible risk for catheter-related complications and can be time-consuming (1,13). Furthermore, ICA should be avoided when its risk outweighs potential benefits (e.g., in acute aortic dissection or aortic vegetation) (2).
On the basis of large multicenter clinical trials and meta-analyses, coronary CTA can accurately rule out coronary stenoses in the general population with a low or intermediate pre-test probability of CAD (50–52). However, to date, only a limited number of single-center studies have tested the accuracy of coronary CTA to detect coronary stenoses in patients before cardiac valve surgery, and definitive evidence advocating for the use of coronary CTA in VHD requires further validation (4–20). In this regard, the present meta-analysis provides a stronger basis for clinical application of coronary CTA in patients with VHD. Using ICA as a reference standard, coronary CTA exhibited an excellent pooled negative LR (0.07) for exclusion of significant coronary stenoses in patients with VHD. Based on our data, ICA could have been avoided in 64% (703 of 1,107) of patients with a true-negative CTA result, whereas in 3% (34 of 1,107) of patients, the correct diagnosis of obstructive CAD would have been missed on the basis of a false-negative CTA finding. These results are clinically relevant because they indicate that coronary CTA may be a reasonable gatekeeper to ICA before valve surgery. Consequently, our findings support the potential for reduction of unnecessary downstream testing and catheter-related complications and costs. By relevance, a normal CTA examination confers an excellent prognosis, thus supporting its clinical utility if used in patients with suspected CAD (53). In contrast to the excellent ability to exclude obstructive CAD, the pooled positive LR (8.44) of coronary CTA to identify significant coronary stenoses was modest, further substantiating the additional need for ICA to confirm or refute a positive CTA test result in patients with VHD.
Given that the risk factors for CAD vary among different valve diseases (3), a subanalysis of diagnostic performance of coronary CTA for specific VHD is desirable. Particularly germane to this concept, patients with AS are older and have a higher prevalence of CAD compared with other valve diseases (48). Indeed, this scenario has been consistently emphasized in several studies that have observed a limited specificity of coronary CTA for the detection of significant stenoses in patients with AS (6,9,10,12). These findings are in line with the present meta-analysis demonstrating decreased specificity (87%) and positive LR (7.4) of coronary CTA against the background of AS. Conversely, the presence of AS did not affect the ability of CTA to reliably exclude significant CAD as confirmed by the high negative LR (0.08) and heterogeneity analysis. Of note, coronary CTA showed excellent diagnostic performance both to detect and rule out significant coronary stenoses in patients without AS, highlighting the need for the assessment of the pre-test probability of CAD.
Although coronary CTA with ≥64 detectors was associated with improved specificity compared with <64 detectors in the generalized linear model analysis, sensitivity remained high for both ≥64 and <64 detectors. These results are clinically important because they indicate that coronary CTA could be used effectively to rule out significant CAD in patients with VHD across a wide range of technical scanners. In contrast, the use of CTA with <64 detectors was inadequate for the detection of significant coronary stenoses, likely as a result of worse spatial and temporal resolution.
Beyond its use in the identification of CAD, coronary CTA offers information incremental to that of ICA for the assessment of cardiac valve morphology. Indeed, coronary CTA may contribute to the evaluation of the severity of valve disease, either by quantifying valvular calcification or through the measurement of valve planimetry (11). It can also provide practical knowledge to the cardiac surgeon about the intrathoracic anatomy, including aortic arch calcification and dimensions, course of arteries, or presence of congenital anomalies (2).
In the present meta-analysis, the reported dose of radiation was 2.55 to 26.0 mSv, and the dose of contrast agent ranged from 80 to 125 ml, which are somewhat higher compared with ICA. However, with the growing use of CTA imaging, newer techniques to lower radiation exposure are constantly being developed. Recently, ultra-low-dose protocols with mean estimated radiation doses <1 mSv and preserved image quality have been documented (54). Moreover, the use of low concentration contrast medium protocols might effectively reduce the risk of contrast nephropathy by obtaining the same quality of CTA (55).
Study limitations
The results of our meta-analysis are limited to studies with both a retrospective and prospective design. Nonetheless, this study was based on the largest currently available pooled cohort of patients with VHD, and it can thus recognize the most optimal candidates for the identification of coronary stenoses with CTA. Furthermore, patients with severely impaired renal function, known contrast allergy, cardiac arrhythmias, and hemodynamic instability were excluded from the majority of the analyzed studies. Thus, the possibility of selection bias that is present in all meta-analyses cannot be discounted. Also, none of the included studies reported details concerning the clustered nature of the per-segment and per-vessel data. Finally, not all studies provided data regarding patient, vessel, and segment levels.
Conclusions
The present meta-analysis showed that coronary CTA provides, across a wide range of diverse clinical centers, high diagnostic performance for the exclusion of significant coronary stenoses in patients scheduled for valve surgery. Considering the differences in the pre-test probabilities of CAD among various valvular diseases, the diagnostic accuracy of coronary CTA to detect significant CAD decreased against the background of AS.
COMPETENCY IN MEDICAL KNOWLEDGE: Coronary CTA using currently available technology has excellent diagnostic performance for the exclusion of significant coronary stenoses in patients scheduled for valve surgery, and it can serve as a gatekeeper for ICA before cardiac valve surgery. The diagnostic ability of coronary CTA to detect significant CAD may be decreased against the background of AS.
TRANSLATIONAL OUTLOOK: Future studies are needed to test whether the routine performance of coronary CTA in broad populations of patients scheduled for cardiac valve surgery may help reduce downstream testing as well as catheter-related complications and costs. Furthermore, whereas the image acquisition and reconstruction protocols of coronary CTA are constantly being developed, large-scale prospective clinical trials comparing radiation and contrast exposure between ICA and coronary CTA among patients with valvular heart disease now seem warranted.
Appendix
Appendix
For supplemental results and figures, please see the online version of this article.
Footnotes
Dr. Opolski has received a scholarship from the Foundation for Polish Science. Dr. Min has served on the medical advisory boards of GE Healthcare, Arineta, AstraZeneca, and Bristol-Myers Squibb; serves on the Speakers Bureau of GE Healthcare; received research support from GE Healthcare, Vital Images, and Phillips Healthcare; and received grants from the National Institutes of Health/National Heart, Lung and Blood Institute (R01HL111141, R01HL115150, R01HL118019, and U01HL105907) and from the Qatar National Research Foundation (NPRP09-370-3-089). Dr. Kim has served as a proctor for Symetis SA and St. Jude Medical. Dr. Achenbach has received grant support and serves on the Speakers Bureau of Siemens Healthcare. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. João Lima, MD, served as Guest Editor for this paper.
- Abbreviations and Acronyms
- AS
- aortic valve stenosis
- CAD
- coronary artery disease
- CI
- confidence interval
- CTA
- computed tomography angiography
- ICA
- invasive coronary angiography
- LR
- likelihood ratio
- ROC
- receiver-operating characteristic
- VHD
- valvular heart disease
- Received September 8, 2015.
- Accepted September 24, 2015.
- 2016 American College of Cardiology Foundation
References
- ↵
- Nishimura R.A.,
- Otto C.M.,
- Bonow R.O.,
- et al.
- ↵
- ↵
- ↵
- Bettencourt N.,
- Rocha J.,
- Carvalho M.,
- et al.
- ↵
- ↵
- Gilard M.,
- Cornily J.C.,
- Pennec P.Y.,
- et al.
- ↵
- ↵
- ↵
- Laissy J.P.,
- Messika-Zeitoun D.,
- Serfaty J.M.,
- et al.
- ↵
- ↵
- Lee S.C.,
- Ko S.M.,
- Song M.G.,
- et al.
- ↵
- ↵
- Meijboom W.B.,
- Mollet N.R.,
- Van Mieghem C.A.,
- et al.
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- Takaoka H.,
- Funabashi N.,
- Kataoka A.,
- et al.
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- Harbord R.M.,
- Whiting P.
- ↵
- ↵
- ↵
- Diab A.Y.,
- Gonsorcik J.,
- Dobrovicova A.,
- Gibarti C.,
- Javorsky M.,
- Stancak B.
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- Russo V.,
- Gostoli V.,
- Lovato L.,
- et al.
- ↵
- ↵
- ↵
- ↵
- Tanaka H.,
- Shimada K.,
- Yoshida K.,
- Jissho S.,
- Yoshikawa J.,
- Yoshiyama M.
- ↵
- ↵
- ↵
- ↵
- ↵
- Mullany C.J.,
- Elveback L.R.,
- Frye R.L.,
- et al.
- ↵
- ↵
- Meijboom W.B.,
- Meijs M.F.,
- Schuijf J.D.,
- et al.
- ↵
- ↵
- ↵
- Otaki Y.,
- Arsanjani R.,
- Gransar H.,
- et al.
- ↵
- ↵
Toolbox
Citation Manager Formats
Figures in Article
Supplementary Materials
- Online Appendix and Online Figures 1–3[S1936878X16303114_mmc1.docx]