Author + information
- Received October 3, 2018
- Revision received January 5, 2019
- Accepted January 7, 2019
- Published online April 15, 2019.
- Brent D. Wilson, MD, PhDa,b,
- Stephen L. Wasmund, PhDa,b,
- Frank B. Sachse, PhDb,c,d,
- Gagandeep Kaur, BSb,
- Nassir F. Marrouche, MDa,b and
- Lisa A. Cannon-Albright, PhDb,e,f,g,∗ ()
- aDivision of Cardiovascular Medicine, Department of Internal Medicine, University of Utah School of Medicine, Salt Lake City, Utah
- bComprehensive Arrhythmia Research and Management Center, University of Utah Health, Salt Lake City, Utah
- cNora Eccles Harrison Cardiovascular Research and Training Institute, University of Utah, Salt Lake City, Utah
- dDepartment of Biomedical Engineering, University of Utah, Salt Lake City, Utah
- eGenetic Epidemiology, Department of Internal Medicine, University of Utah School of Medicine, Salt Lake City, Utah
- fHuntsman Cancer Institute, Salt Lake City, Utah
- gGeorge E. Wahlen Department of Veterans Affairs Medical Center, Salt Lake City, Utah
- ↵∗Address for correspondence:
Dr. Lisa A. Cannon-Albright, Genetic Epidemiology, University of Utah, 391 Chipeta Way, Suite D, Salt Lake City, Utah 84108.
Objectives The aim of this study was to define the population-based familial clustering of atrial fibrillation (AF) that is associated with fibrosis and describe evidence for a heritable predisposition.
Background Although a heritable contribution to AF is well-established and the association of fibrosis with AF is well-recognized, no studies have analyzed the genetic contribution to AF co-occurring with fibrosis.
Methods AF patients with magnetic resonance imaging–confirmed fibrosis were identified in a population-based health sciences center database linked to a Utah genealogy. Familial clustering of AF/fibrosis was defined by analysis of pairwise case relatedness, estimation of relative risk of AF/fibrosis in relatives, and identification of high-risk AF/fibrosis pedigrees.
Results The 694 individuals identified with AF/fibrosis who had at least 3 generations of genealogy data were found to have significantly elevated pairwise relatedness (p < 0.001), even when first- and second-degree relationships were ignored (p < 0.001). Significantly elevated risks for AF/fibrosis among first- (relative risk [RR]: 4.65), second- (RR: 3.14), and third-degree (RR: 2.70) relatives of individuals with AF/fibrosis were observed. We identified 157 extended Utah pedigrees with a significant excess of AF/fibrosis among descendants.
Conclusions There is a strong heritable contribution to predisposition to AF co-occurring with fibrosis. We suggest that this study provides a unique foundation for a search for predisposition genes, specifically for AF co-occurring with fibrosis.
Atrial fibrillation (AF) is the most common cardiac rhythm disorder, affecting 1% to 2% of the population and up to 6% of those over age 65 years. Although some individuals experience only mild symptoms, AF can be disabling or fatal because of complications such as heart failure and stroke. A significant heritable contribution to AF has been well-recognized for decades (1–9) with a population-based examination of familial clustering of AF in Iceland (10) and an investigation of AF and flutter in Sweden (11). There is strong evidence for a heritable contribution to AF, including the identification of rare highly penetrant variants (e.g., KCNQ1) (12), as well as the identification of common AF risk variants in large genome-wide association studies (13). However, despite multiple efforts, the rare and common variants identified to date do not explain the bulk of the observed heritability of AF. Here we consider evidence for a genetic contribution to a subset of AF (associated with fibrosis); the results suggest a focus on this more refined phenotype could provide an informative approach.
AF is a complex phenotype involving interplay of triggering and maintenance of arrhythmia by an arrhythmogenic substrate, both of which have been linked to atrial fibrosis. Atrial fibrosis, or excessive formation of connective tissue in the atria, is considered the hallmark of structural remodeling in AF (14,15). Historically, it was thought that atrial fibrotic remodeling was a consequence of AF, which then further perpetuates AF (16,17). However, recent progress in electro-anatomical mapping and noninvasive cardiac imaging suggests that these fibrotic alterations likely precede AF onset. An association between conditions linked with atrial fibrosis and AF is well-established, but we are only starting to understand how fibrosis contributes to AF (14). Proposed mechanisms for the contribution of fibrosis to AF account for the degree, composition, and organization of fibrotic remodeling of atrial tissue (18). One proposed mechanism involves discrete fibrotic areas anchoring sources of fibrillation in the surrounding electrically excitable cardiomyocytes. Another suggested mechanism is conduction abnormality arising in locations with a heterogeneous composition of fibrotic remodeling and excitable cardiomyocytes. We make no attempt to identify or distinguish among mechanisms of fibrosis in the cases analyzed in this study.
Previously it was only possible to assess left atrial fibrosis with invasive electro-anatomical mapping or with histology after biopsy. Histological studies have shown that fibrotic changes of the left atrial wall are associated with presence and persistence of AF (19). Invasive electro-anatomical mapping has shown that presence of low voltage or fibrosis predicts AF recurrence (20). More recently, it has been possible to detect the presence of left atrial fibrosis with high resolution, late gadolinium enhanced magnetic resonance imaging (LGE-MRI). The DECAAF (Delayed-Enhancement MRI [DE-MRI] Determinant of Successfully Radiofrequency Catheter Ablation of Atrial Fibrillation) I study, a large multicenter study involving 15 centers in the United States, Europe, and Australia, demonstrated that left atrial fibrosis as assessed by LGE-MRI predicts outcome after catheter ablation (21). The ability to detect left atrial fibrosis with LGE-MRI and its association with treatment outcome have been replicated by independent investigators (22). Additionally, areas of left atrial enhancement detected by LGE-MRI correlate with fibrosis seen by histological staining of biopsy samples obtained during cardiothoracic surgery procedures (23). In addition to predicting outcome after AF ablative therapy, presence of left atrial fibrosis as assessed by LGE-MRI is associated with stroke, heart failure, and cardiovascular mortality (24).
The University of Utah Comprehensive Arrhythmia Research and Management (CARMA) Center has amassed the largest collection in the world of AF patients with associated fibrosis confirmed by LGE-MRI. Using unique Utah genealogical resources, genetic relationships, both close and distant, among AF patients with associated fibrosis were identified and analyzed. Evidence for a strong genetic contribution to AF/fibrosis has been identified, and a uniquely powerful resource for predisposition gene identification has been described (Central Illustration).
The existence of a Utah resource linking genealogy with statewide hospital data has allowed many advances in genetic studies and the identification of disease predisposition genes. Decades of statewide hospital and clinic records have been linked to a Utah genealogy in the Utah Population Database (UPDB). The genealogy data in the UPDB represents the original Utah pioneer founders from the mid-1800s and extends to their descendants in modern day; it now includes over 3 million individuals who have from 3 to 16 generations of genealogy (25). The Utah population represented in the UPDB is representative of U.S. and Northern European populations, with a low rate of inbreeding (26–28); genes identified in Utah have been found to be representative of populations beyond the U.S. and European populations represented.
The genealogy data are linked to the electronic data warehouses of Utah health care providers. The University of Utah Health Sciences Center (UUHSC) is among the largest health care providers in Utah and has almost 900,000 patients linked to genealogy in the UPDB. The UPDB has been used successfully to define familial clustering and genetic influences in a variety of disorders including cancer (29,30), coronary heart disease (31), and diabetes (32), among others. The methods used to identify phenotypes, assess familial and genetic effects, and identify pedigrees using UPDB data have been described in detail in these studies.
A foundation for this study is a set of confirmed AF cases with LGE-MRI-identified fibrosis developed by the CARMA Center at UUHSC. UUHSC is estimated to serve 20% of the state of Utah and is the only Utah center using delayed-enhancement MRI to identify fibrosis. The CARMA registry includes over 4,087 AF patients: 1,908 with confirmed AF also have LGE-MRI data; 1,100 AF patients have significant atrial fibrosis confirmed by LGE-MRI examination with at least acceptable quality MRI data; 650 are lone AF cases. Of these 1,100 patients, 694 AF/fibrosis CARMA patients also have at least 3 generations of genealogy data in the UPDB; 399 of the AF/fibrosis cases (57%) analyzed were male. All patients with LGE-MRI data were assessed for degree of fibrosis (prior to ablative therapy). Of the 1,100 patients with AF and fibrosis, 61.5% had good-to-excellent quality fibrosis data, and 38.5% had moderate but acceptable quality. The 1,100 AF/fibrosis cases were classified by Utah stage according to the amount of left atrial fibrosis present compared with volume of left atrial tissue as follows: 346 were Utah stage 1 (<10% fibrosis); 451 were Utah stage 2 (10% to 20% fibrosis); 226 were Utah stage 3 (20% to 30% fibrosis); and 77 were Utah stage 4 (>30% fibrosis).
Genealogical Index of Familiality—analysis of excess relatedness
The combination of genealogy data with AF/fibrosis phenotype data allows consideration of the hypothesis of excess familial clustering, or excess relatedness, of individuals diagnosed with AF/fibrosis. This test was performed using the Genealogical Index of Familiality (GIF) method, developed specifically for the UPDB (25,33). The GIF measures the average pairwise relatedness of a set of cases and compares that measurement to the average pairwise relatedness expected for a similar group of individuals in the Utah population. The GIF test differs from relative risk (RR) estimates in that it includes analysis of both close and distant genetic relationships. The GIF uses the Malécot coefficient of kinship to measure pairwise relatedness. The coefficient is defined as the probability that randomly selected homologous genes from 2 individuals are identical by descent from a common ancestor (34). The contribution to the GIF statistic is smaller for pairs with greater genetic distance between them. The expected pairwise relatedness for a set of cases is estimated from the mean pairwise relatedness of 1,000 sets of matched control subjects who have genealogy data (matched on sex, 5-year birth year range, and birth state; each set of control subjects containing 1 matched control subject for each case). The empirical significance of the GIF test is measured by comparing the case GIF to the distribution of 1,000 control GIF values.
The GIF statistic measures familial clustering, which can be due to genetic, or to shared familial environmental effects, or to a combination of both. To better distinguish these effects, we also perform a GIF analysis while ignoring close (first- and second-degree) relationships. If this distant GIF test is significant, it provides additional strong evidence that there is significant distant excess relatedness that is unlikely to be due to shared environment.
The contribution to the GIF statistic can be quantified separately for the different genetic distances observed among pairs of cases and control subjects (Figure 1). The contribution to the GIF statistic is estimated as the kinship coefficient associated with the pairwise distance, multiplied by the number of such pairs among the set of individuals considered, adjusted for the total number of pairs considered, then multiplied by 105 to adjust scale. The kinship coefficient is smaller as genetic distance increases; the number of pairs of affected relatives observed depends on the specific set of affected individuals analyzed, but typically increases to the median relationship in the population, and then decreases. Pairwise genetic distance is shown on the x-axis for cases compared with control subjects. The genetic distance measure represents different pairwise relationships: 1 for parent/offspring; 2 for siblings or grandparent/grandchild; 3 for avunculars or great-grandparent/great-grandchild or similar; 4 for first cousins or similar; 6 for second cousins or similar, and so forth.
RR in relatives
The calculation of RR in relatives is a more traditional mechanism for identifying and quantifying genetic effects. A genetic contribution to a phenotype is supported when both close and distant relatives of probands who have the phenotype of interest have elevated rates of the phenotype. RR for the AF/fibrosis phenotype were estimated for first-, second-, and third-degree relatives of individuals with AF/fibrosis who were also hospital patients, as follows. All individuals in the UPDB with at least 3 generations of genealogy data and linked hospital data were assigned to a cohort based on birth year (in 5-year groups), sex, and birthplace (Utah or not). Cohort-specific rates of AF/fibrosis were estimated by dividing the total number of AF/fibrosis cases per cohort by the total number of individuals in each cohort. Expected numbers of first-degree relatives (for example) with AF/fibrosis were estimated by counting the number of first-degree relatives by cohort (without duplication), multiplying by the rate of AF/fibrosis in each cohort, and summing over all cohorts. Observed numbers of individuals with AF/fibrosis among first-degree relatives were counted, without duplication. RRs were estimated similarly for each degree of relationship as observed/expected number of individuals with AF/fibrosis; 95% confidence intervals for the RR were calculated using the method of Agresti (35). First-degree relatives include parents, siblings, and offspring; second-degree relatives are the first-degree relatives of the first-degree relatives (e.g., uncle or grandmother); third-degree relatives are the first-degree relatives of second-degree relatives (e.g., first cousin or great-grandchild).
To identify high-risk pedigrees, all relationships among all individuals diagnosed with AF/fibrosis were analyzed. Consideration of all ancestral vectors of all patients allowed identification of clusters of related cases. The nearest common ancestor was identified for each independent cluster of related cases. No completely overlapping clusters, or pedigrees, are considered, but individuals can appear in more than 1 cluster. To determine whether any observed cluster, or pedigree, is high risk, the observed number of individuals diagnosed with AF/fibrosis among the descendants is compared to the expected number. The expected number of AF/fibrosis cases among the descendants is calculated by counting all descendants by cohort, multiplying the number of descendants in each cohort by the cohort-specific rate of AF/fibrosis (estimated as described herein), and summing over all cohorts. An excess of the number of AF/fibrosis cases observed compared with the number expected among the descendants of p < 0.05 is used to classify a descending pedigree as high risk.
Evidence for genetic contribution to AF/fibrosis
The GIF analysis of all genetic relationships among the 694 AF/fibrosis patients who also have genealogy data shows significant excess relatedness of AF/fibrosis cases over the expected relatedness for a demographically matched group of Utah individuals (p < 0.001) (Table 1). A significant excess was also observed for the distant GIF test, where first- and second-degree relationships were ignored (p < 0.001).
Figure 1 shows the contribution to the average pairwise relatedness of all pairs of AF/fibrosis cases by pairwise genetic distance, compared with the average relatedness for the 1,000 sets of matched hospital control subjects, which estimates the expected relatedness in the UPDB (25). As an example of calculation of the contribution to the GIF statistic, there were 6 parent/offspring pairs of AF cases observed; the contribution is 6 × 0.25 / (694 × 693 / 2) × 105 = 0.62, where 0.25 is the coefficient for parent/offspring and 694 × 693 / 2 is the total number of case pairs. There is a clear excess of close and distant relationships (through first cousins, genetic distance = 4) (p < 0.001).
RR in relatives
Table 2 shows estimated RR in relatives of differing degree of relationship to the affected proband, including degree of relationship, number of relatives of the defined relationship, number of AF/fibrosis cases observed among the relatives, number of AF/fibrosis cases expected among the relatives using cohort-specific rates for AF/fibrosis estimated from the UPDB hospital-linked patients, RR, p value for the 2-tailed RR hypothesis test, and 95% confidence interval for the RR.
For more specific comparisons, Table 2 also shows RR estimates for first-degree relatives by relationship (parent, sibling, child), and by relative’s sex for each relationship. First-, second-, and third-degree relatives of AF/fibrosis cases all showed significantly elevated risk for AF/fibrosis; no significant differences in risk were observed by sex of the relative. These analyses of familial clustering among both close and distant relatives strongly support a genetic contribution to AF/fibrosis.
High risk AF/fibrosis pedigrees
In addition to confirming a heritable contribution to AF/fibrosis, the powerful Utah genealogical resource linked to AF/fibrosis cases at the CARMA Center allowed identification of AF/fibrosis pedigrees that are represented in the familial clustering analyses, as well as determination of which of these clusters (pedigrees) represent a significant excess of AF/fibrosis among the pedigree members. This is an important filtering step to identify high-risk pedigrees and identify the truly high-risk pedigrees that are most powerful for gene identification. Whereas many clusters of related affected individuals may be observed in a population, only those that have a significant excess of the phenotype of interest, over population rates, will be informative for identification of the inherited predisposition. A total of 702 clusters of from 2 to 104 related AF/fibrosis cases were identified in the UPDB; only 157 of these clusters of related cases (22%) had a statistical excess of AF/fibrosis cases; the others were chance clusters of related cases with no excess risk for AF/fibrosis. These high-risk AF/fibrosis pedigrees have multiple advantages over any other such pedigree collections: 1) cases are identified in a dedicated center with MRI confirmation of fibrosis; 2) studied pedigrees are selected for a statistically significant excess of AF/fibrosis cases; and 3) pedigrees are extensive, covering multiple generations. The 157 Utah high-risk pedigrees identified include from 2 to 12 related AF/fibrosis patients; 37 pedigrees include 4 or more related AF/fibrosis cases; this resource continues to grow.
As expected, given the low rate of rare, penetrant AF genes, the high-risk AF/fibrosis pedigrees identified do not appear to represent the small and dense clusters of cases that would be expected to be due to the rare syndromic, high-penetrance genes. Only 5 of the 157 high-risk AF/fibrosis pedigrees identified include a first-degree relative pair (parent and child, or at least 2 siblings affected).
Figure 2 shows 2 examples of high-risk AF/fibrosis pedigrees identified from the 694 AF/fibrosis patients with genealogy. As expected, since identification of AF/fibrosis cases from CARMA dates only from 2009 to present, there is a narrow window of view to known affected individuals, such that they only occur in the bottom few generations of a multigenerational high-risk pedigree; most individuals in the upper generations remain unknown for the AF/fibrosis phenotype. This does not affect the informativeness of the pedigrees; the power comes from the distant relationships represented among cases who are members of the high-risk pedigree.
AF is a major global health concern with rising incidence and prevalence (36) and increased morbidity and mortality (37). AF is associated with a variety of chronic conditions, including heart failure, ischemic heart disease, diabetes mellitus, hypertension, chronic obstructive pulmonary disease, and mitral valve disease (38–40). Previously, it was thought that atrial fibrosis occurred after onset of AF and was a consequence and perpetuator of the arrhythmia. However, more recently, the finding that fibrosis is already prevalent in patients with lone AF suggests that AF might be the first arrhythmogenic manifestation of fibrotic alteration (16,17,41). The ability to detect left atrial fibrosis noninvasively with LGE-MRI makes possible new diagnostic and therapeutic approaches. For example, left atrial fibrosis as detected by LGE-MRI predicts response to AF therapeutic intervention. It also correlates with stroke, heart failure, and increased cardiovascular mortality (24). Assessment of LA fibrosis allows identification of patients at highest risk for complications of AF and allows selection of most efficacious clinical therapies. Although the exact mechanism whereby fibrosis contributes to initiation of AF is debated, its association with presence and extent of AF is clear (19).
The heritable contribution to AF is widely recognized; however, the genes responsible for most AF remain unknown, as does the underlying etiology. As fibrosis is known to be associated with AF, we hypothesize that its presence is an important contributor to AF. Our analysis of the familial clustering (both close and distant) of the AF/fibrosis phenotype in Utah has clearly shown a strong heritable contribution to AF/fibrosis and has identified a large resource of extended high-risk pedigrees that will be powerful for gene identification for AF/fibrosis.
Whereas some groups have reported small, high-penetrance AF pedigrees (10,11,42), few extended high-risk AF pedigrees have been identified or studied. Extended high-risk pedigrees are critical to predisposition gene identification because they are likely to evidence a strong role for genetic factors and exhibit limited genetic heterogeneity. Although all pedigrees at high risk for AF are a valuable resource for predisposition gene identification, it is likely that focus on high-risk AF pedigrees with a more homogeneous phenotype (e.g., AF/fibrosis) would be extremely powerful for gene identification. The high-risk pedigrees identified here could potentially provide a better understanding of comorbid associations and their association with risk for AF.
This analysis of a population-based resource has some limitations that are primarily related to data censoring. Not all individuals with an AF/fibrosis diagnosis in Utah have been identified. Not all individuals with AF/fibrosis have genealogy data in the UPDB resource, or are record-linked appropriately to their genealogy data, if present; female subjects have lower record linkage rates than male subjects due to name changes. Genealogy data do not always represent biological relationships. Nevertheless, these limitations are assumed to be uniformly present across the resource and thus not to bias the observed results. Data for environmental factors that might also cluster in relatives was not available and could not be considered. It is possible that the AF/fibrosis phenotype could be secondary to another predisposition (e.g., obesity), which would confound the findings. Sample sizes for some RR estimates are small, resulting in wide confidence intervals for effect sizes. It is noted that the Utah population represented in the genealogic data analyzed here is primarily of Northern European origin and results should not be extrapolated further.
Our study provides the foundation for a search for AF/fibrosis predisposition genes that can be performed on related living individuals diagnosed with AF/fibrosis who are members of high-risk pedigrees. The value of these validated high-risk pedigrees for predisposition gene identification is the potential for a drastically increased rate of any rare, deleterious mutations that exist among these cases over what would be encountered in other designs such as genome-wide association studies or studies of nuclear families. This high-risk pedigree approach has been successfully applied in many Utah studies that have identified multiple disease predisposition genes (BRCA1 ; BRCA2 ; CDKN2A , GOLM1 ). Weeke et al. (42) previously sequenced related AF-affected individuals in 6 families to identify candidate predisposition variants and noted the importance of having large well-curated multigenerational pedigrees, such as have been identified here.
The AF/fibrosis resource, made possible by the powerful combination of resources from the UPDB and the CARMA registry, will enable recruitment and sampling of high-risk AF/fibrosis pedigrees and may enable identification of AF/fibrosis predisposition genes that have eluded other approaches. These new AF/fibrosis predisposition genes may account for the known AF heritability that has yet to be explained.
COMPETENCY IN MEDICAL KNOWLEDGE: This report suggests a genetic contribution to predisposition to AF/fibrosis, with estimated RR high enough that screening of family members of AF/fibrosis patients should be considered. These results may support expansion of standard clinical practice to include the collection of family history of AF/fibrosis followed by enhanced screening recommendations.
TRANSLATIONAL OUTLOOK: This report makes the important initial observation of evidence for a genetic contribution to AF/fibrosis and the existence of high-risk pedigrees that could provide a powerful resource for predisposition gene identification. Future studies should include validation in other populations, as well as in-depth analysis of sequence data from high-risk cases and pedigrees.
Partial support for Dr. Cannon-Albright and for all datasets within the Utah Population Database was provided by Huntsman Cancer Institute, Huntsman Cancer Foundation, University of Utah, and the Huntsman Cancer Institute's Cancer Center Support grant P30 CA42014 from National Cancer Institute. Dr. Sachse is partially supported by the Nora Eccles Treadwell Foundation. Dr. Marrouche has ownership interest in Marrek, Inc. and Cardiac Designs; has performed contracted research for Biosense Webster, Medtronic, St. Jude, Biotronik, and Boston Scientific; and has received consulting fees from Biotronik and Preventice. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
All authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the JACC: Clinical Electrophysiology author instructions page.
- Abbreviations and Acronyms
- atrial fibrillation
- Comprehensive Arrhythmia Research and Management
- Genealogical Index of Familiality
- late gadolinium enhanced magnetic resonance imaging
- relative risk
- Utah Population Database
- University of Utah Health Sciences Center
- Received October 3, 2018.
- Revision received January 5, 2019.
- Accepted January 7, 2019.
- 2019 American College of Cardiology Foundation
- Darbar D.,
- Herron K.J.,
- Ballew J.D.,
- et al.
- Christophersen I.E.,
- Ravn L.S.,
- Budtz-Joergensen E.,
- et al.
- Zoller B.,
- Ohlsson H.,
- Sundquist J.,
- Sundquist K.
- Chen Y.H.,
- Xu S.J.,
- Bendahhou S.,
- et al.
- Christopheresen I.E.,
- Ellinor P.T.
- Burstein B.,
- Nattel S.
- Platonov P.G.,
- Mitrofanova L.B.,
- Orshanskaya V.,
- Ho S.Y.
- Verma A.,
- Wazni O.M.,
- Marrouche N.F.,
- et al.
- Khurram I.M.,
- Habibi M.,
- Gucuk Ipek E.,
- et al.
- McGann C.,
- Akoum N.,
- Patel A.,
- et al.
- King J.B.,
- Azadani P.N.,
- Suksaranjit P.,
- et al.
- Cannon-Albright L.A.,
- Thomas A.,
- Goldgar D.E.,
- et al.
- Cannon L.,
- Bishop D.T.,
- Skolnick M.H.,
- Hunt S.,
- Lyon J.L.,
- Smart C.
- Hill J.R.
- Malecot G.
- Agresti A.
- Chugh S.S.,
- Havmoeller R.,
- Narayanan K.,
- et al.
- Chugh S.S.,
- Blackshear J.L.,
- Shen W.K.,
- Hammill S.C.,
- Gersh B.J.
- Wolf P.A.,
- Abbott R.D.,
- Kannel W.B.
- Fuster V.,
- Ryden L.E.,
- Cannom D.S.,
- et al.
- Camm A.J.,
- Lip G.Y.,
- De Caterina R.,
- et al.
- Miki Y.,
- Swensen J.,
- Shattuck-Eidens D.,
- et al.
- Teerlink C.C.,
- Huff C.,
- Stevens J.,
- et al.