Author + information
- Received January 24, 2019
- Revision received February 13, 2019
- Accepted February 13, 2019
- Published online April 29, 2019.
- Suzanne V. Arnold, MD, MHAa,
- Khaja M. Chinnakondepalli, MSa,
- John A. Spertus, MD, MPHa,
- Elizabeth A. Magnuson, ScDa,
- Suzanne J. Baron, MD, MSca,
- Saibal Kar, MDb,
- D. Scott Lim, MDc,
- Jacob M. Mishell, MDd,
- William T. Abraham, MDe,
- JoAnn A. Lindenfeld, MDf,
- Michael J. Mack, MDg,
- Gregg W. Stone, MDh,
- David J. Cohen, MD, MSca,∗ (, )@djc795@MidAmericaHeart,
- on behalf of the COAPT Investigators
- aSaint Luke’s Mid America Heart Institute and University of Missouri-Kansas City, Kansas City, Missouri
- bCedars-Sinai Medical Center, Los Angeles, California
- cUniversity of Virginia, Charlottesville, Virginia
- dKaiser Permanente-San Francisco Hospital, San Francisco, California
- eOhio State University, Columbus, Ohio
- fVanderbilt Heart and Vascular Institute, Nashville, Tennessee
- gBaylor Scott and White Health, Plano, Texas
- hNew York-Presbyterian Hospital and Cardiovascular Research Foundation, New York, New York
- ↵∗Address for correspondence:
Dr. David J. Cohen, Saint Luke’s Mid America Heart Institute, University of Missouri-Kansas City, 4401 Wornall Road, Kansas City, Missouri 64111.
Background In the COAPT (Cardiovascular Outcomes Assessment of the MitraClip Percutaneous Therapy for Heart Failure Patients with Functional Mitral Regurgitation) trial, transcatheter mitral valve repair (TMVr) led to reduced heart failure (HF) hospitalizations and improved survival in patients with symptomatic HF and 3+ to 4+ secondary mitral regurgitation (MR) on maximally-tolerated medical therapy. Given the advanced age and comorbidities of these patients, improvement in health status is also an important treatment goal.
Objectives The purpose of this study was to understand the health status outcomes of patients with HF and 3+ to 4+ secondary MR treated with TMVr versus standard care.
Methods The COAPT trial randomized patients with HF and 3+ to 4+ secondary MR to TMVr (n = 302) or standard care (n = 312). Health status was assessed at baseline and at 1, 6, 12, and 24 months with the Kansas City Cardiomyopathy Questionnaire (KCCQ) and the SF-36 health status survey. The primary health status endpoint was the KCCQ overall summary score (KCCQ-OS; range 0 to 100; higher = better; minimum clinically important difference = 5 points).
Results At baseline, patients had substantially impaired health status (mean KCCQ-OS 52.4 ± 23.0). While health status was unchanged over time in the standard care arm, patients randomized to TMVr demonstrated substantial improvement in the KCCQ-OS at 1 month (mean between-group difference 15.9 points; 95% confidence interval [CI]: 12.3 to 19.5 points), with only slight attenuation of this benefit through 24 months (mean between-group difference 12.8 points; 95% CI: 7.5 to 18.2 points). At 24 months, 36.4% of TMVr patients were alive with a moderately large (≥10-point) improvement versus 16.6% of standard care patients (p < 0.001), for a number needed to treat of 5.1 patients (95% CI: 3.6 to 8.7 patients). TMVr patients also reported better generic health status at each timepoint (24-month mean difference in SF-36 summary scores: physical 3.6 points; 95% CI: 1.4 to 5.8 points; mental 3.6 points; 95% CI: 0.8 to 6.4 points).
Conclusions Among patients with symptomatic HF and 3+ to 4+ secondary MR receiving maximally-tolerated medical therapy, edge-to-edge TMVr resulted in substantial early and sustained health status improvement compared with medical therapy alone. (Cardiovascular Outcomes Assessment of the MitraClip Percutaneous Therapy for Heart Failure Patients With Functional Mitral Regurgitation [The COAPT Trial] [COAPT]; NCT01626079)
Edge-to-edge transcatheter mitral valve repair (TMVr) with the MitraClip device (Abbott, Santa Clara, California) effectively reduces mitral regurgitation (MR) with low risk for periprocedural complications (1,2). Although the device was originally approved in the United States exclusively for patients with degenerative (primary) MR at extreme surgical risk, the COAPT (Cardiovascular Outcomes Assessment of the MitraClip Percutaneous Therapy for Heart Failure Patients with Functional Mitral Regurgitation) trial recently demonstrated that the benefit of TMVr extends to patients with heart failure (HF) and functional (secondary) MR as well (3). In this trial, TMVr reduced HF hospitalizations and all-cause mortality over 24 months compared with maximally-tolerated medical therapy alone.
Beyond prolonging survival and reducing hospitalizations, improving patients’ health status (i.e., symptoms, functional status, quality of life) is a key treatment goal of TMVr. In fact, among older patients with comorbidities and high symptom burden, health status improvement may be of greater importance to patients than improved survival (4,5). Prior uncontrolled studies among patients with symptomatic degenerative MR showed rapid and marked improvement in health status with TMVr (6). However, the effect of TMVr on the health status of patients with HF and secondary MR is not known. To address this critical gap in knowledge and to better define the full clinical value of TMVr, we performed a prospective health status substudy in the COAPT trial.
The design (7) and primary results (3) of the COAPT trial have been published (NCT01626079). Briefly, COAPT was a multicenter, randomized, open-label trial of TMVr with the MitraClip device in HF patients with left ventricular ejection fraction between 20% and 50% and 3+ to 4+ secondary MR who remained symptomatic despite maximally-tolerated guideline-directed therapy (including use of cardiac resynchronization therapy when indicated). Patients were randomized 1:1 to TMVr or standard care and followed through a minimum of 1 year (and up to 2 years) for clinical events and health status. The study was approved by the institutional review board at each participating site, and all patients provided written informed consent.
Health status outcomes
Health status was evaluated at baseline and at 1, 6, 12, and 24 months from baseline with the Kansas City Cardiomyopathy Questionnaire (KCCQ) (8) and the Medical Outcomes Study Short-Form 36 (SF-36) Health Survey (9). The KCCQ is an HF-specific health status measure that consists of 23 questions and 5 domains: physical limitation, symptoms, quality of life, social limitation, and self-efficacy. The first 4 domains are combined into an overall summary score (KCCQ-OS), which was the primary health status outcome of COAPT. Scores for domains and the summary score range 0 to 100 with higher scores indicating better health status. KCCQ-OS scores correlate roughly with New York Heart Association functional class as follows: class I ∼KCCQ-OS 75 to 100; class II ∼KCCQ-OS 60 to 74; class III ∼KCCQ-OS 45 to 59; and class IV ∼KCCQ-OS 0 to 44 (10), and changes in KCCQ-OS of 5, 10, and 20 points correspond with small, moderate, or large clinical changes, respectively (11). The SF-36 is a generic health status measure that provides physical summary scores (SF-36 PCS) and mental summary scores (SF-36 MCS), which are scaled to an overall population mean of 50 and SD of 10; higher scores indicate better health status, and the minimal clinically important change is ∼2.5 points (12).
All analyses were performed from the time of randomization using an intention-to-treat approach. Within each treatment group, mean scores for each of the health status measures at each follow-up time point were compared with baseline using paired Student’s t-tests. For the primary analysis, between-group differences of health status scores of over time were estimated from piecewise linear regression models with a knot at 1 month. Models included time (linear and spline), treatment, and the interactions between treatment and time, age, sex, and severe lung disease. Subgroup analyses explored potential heterogeneity in health status differences at 1 year of follow-up by introducing interaction terms between treatment and patient factors: age (<74 years vs. ≥74 years), sex, lung disease, etiology of cardiomyopathy (ischemic vs. nonischemic), left ventricular end-diastolic volume index (<94 ml/m2 vs. ≥94 ml/m2), severity of mitral regurgitation (effective regurgitant orifice <0.4 cm2 vs. ≥0.4 cm2), gait speed on 5-m walk test (<0.8 m/s vs. ≥0.8 m/s), and dependency in activities of daily living.
To aid in clinical interpretability, we performed a series of categorical analyses among surviving patients and among all eligible patients (including those who died). At each time point, we calculated the proportion of patients in each treatment group who were alive with a moderately large health status improvement (change ≥10 points from baseline), alive with a large health status improvement (change ≥20 points from baseline), and “alive and well” as previously defined (KCCQ-OS ≥60 and no decline ≥10 points from baseline) (13). Proportions were compared between groups at each time point using chi-square tests, and absolute risk differences (with 95% confidence interval [CI]) and numbers needed to treat were estimated. In addition, an ordinal analysis was performed that calculated the proportion of patients at each time point who were dead, had worse health status (change ≤−5), no change in health status (change >−5 to <5), or improved health status (change ≥5).
Finally, as a sensitivity analysis, survival and health status were jointly modeled using a Bayesian approach (14). An important limitation to health status analysis is that health status can only be assessed in surviving patients. Because patients with worse health status are more likely to die (15–17), ignoring these deaths may bias the estimates of health status upward, as the sickest patients are systematically removed from the analyses due to death. In this novel approach, survival and health status were jointly modeled to allow the deaths to inform the health status estimates. A population level piecewise linear model was fit with a knot at 1-month and patient-specific trajectories modeled by 3 random effects: baseline intercept, 1-month intercept, and post–1-month slope; the latter 2 parameters were jointly modeled with the survival data. Joint modeling was done in a fully Bayesian framework, and priors were selected to be weakly informative, which stabilized the model while allowing the data to inform the posterior as much as possible (18). Point estimates and credible intervals (CrI) were generated for the KCCQ-OS, SF-36 PCS, and SF-36 MCS scores over time and the treatment effect of TMVr. The mean treatment effects from these joint analyses can be conceptualized as the expected benefit of TMVr if the patient survives to the time point of interest. All analyses were performed with SAS version 9.4 (SAS Institute, Cary, North Carolina) and R (R Foundation, Vienna Austria), and statistical significance was defined as a 2-sided p value <0.05. The primary outcome was the between-group comparison of the KCCQ-OS over time with all other comparisons considered secondary; as such, there was no correction for multiple comparisons (19).
Between December 2012 and June 2017, 614 patients were enrolled in COAPT at 78 centers in the United States and Canada: 302 were randomized to TMVr and 312 to standard care. Three patients in the standard care arm did not complete baseline health status measures and were excluded. Baseline characteristics were well balanced between treatment groups (Table 1). Mean age of the analytic cohort was 72 ± 11 years, 64% were men, and mean left ventricular ejection fraction was 31.3 ± 9.3%.
Baseline health status and within-group comparisons
Compliance was high for health status measures over time, with ≥88% of eligible patients having KCCQ-OS data at each time point (Online Table 1). Mean KCCQ-OS score at baseline was 52.4 ± 23.0, with the lowest domain score being quality of life at 44.9 ± 25.7. Mean SF-36 PCS was 32.8 ± 9.6, and mean SF-36 MCS was 46.0 ± 12.9.
Among patients randomized to TMVr, KCCQ-OS increased by an average of 16.9 points by 1 month (95% CI: 14.2 to 19.6 points), with similar within-group changes at later time points (Table 2). All KCCQ domains improved significantly by 1 month, with the largest change in the quality of life domain (mean change 23.2 points; 95% CI: 20.0 to 26.4 points). Scores on the SF-36 PCS and MCS both also increased significantly at 1 month, with changes of 6.0 points (95% CI: 5.0 to 7.1 points) and 4.2 points (95% CI: 2.8 to 5.6 points), respectively. Among patients randomized to standard care, the KCCQ-OS increased, on average, by 2.1 points in the first month (95% CI: −0.1 to 4.3 points) with small but significant changes at later time points of 5 to 6 points and no significant changes in the SF-36 PCS or MCS scores.
Between-group comparisons and subgroups
Among surviving patients at 1 month, the KCCQ-OS increased to a greater extent in the TMVr arm compared with standard care (mean difference 15.9 points; 95% CI: 12.3 to 19.5 points; p < 0.001) (Central Illustration), with only slight attenuation of the treatment effect over time (mean difference among surviving patients of 14.5 points [95% CI: 10.9 to 18.1 points] and 12.8 points [95% CI: 7.5 to 18.2 points] at 12 and 24 months, respectively; p < 0.001 for both comparisons). Results were similar for each of the KCCQ domains (Online Table 2). A similar pattern was observed for both the SF-36 PCS and MCS, which improved significantly within 1 month in the TMVr arm compared with standard care, with mean between group differences of 5.3 points (95% CI: 3.8 to 6.8 points) and 5.2 points (95% CI: 3.3 to 7.1 points), respectively, and only slight attenuation over time (Figures 1A and 1B). The health status benefit of TMVr compared with standard therapy was consistent across all subgroups (Table 3) (all interaction p values >0.2), except cause of cardiomyopathy; patients with an ischemic cardiomyopathy appeared to derive greater health status benefit from TMVr compared with those with a nonischemic cardiomyopathy (p value for interaction = 0.02). Notably, patients with chronic lung disease, slow gait speed, or dependency in an activity of daily living tended to have lower KCCQ-OS scores both at baseline and 1-year follow-up.
Ordinal and categorical outcomes
Figure 2 shows the proportion of patients by treatment group at each time point who were dead, worse, unchanged, and improved from baseline according to the KCCQ-OS. At each time point, more patients who were randomized to TMVr were improved, fewer patients had significantly worsened from baseline, and fewer had died (with the exception of 1 month) compared with standard therapy. At 24 months, 39.3% of patients in the TMVr arm were alive and improved compared with 20.8% in the standard care arm. Examining different thresholds of improvement at 24 months, 36.4% of TMVr patients were alive with a moderately large improvement in KCCQ-OS versus 16.6% of standard care patients (number needed to treat [NNT] 5.1; 95% CI: 3.6 to 8.7), and 29.1% of TMVr patients were alive with a large health status improvement versus 11.7% of standard care patients (NNT 5.7; 95% CI: 4.0 to 10.2) (Table 4).
Joint model outcomes
Figure 3 shows the estimated health status trajectories for the KCCQ-OS, SF-36 PCS, and SF-36 MCS with TMVr versus standard care as derived from piecewise linear regression (primary analysis) and the joint model that allowed the survival data to inform the health status estimates. Given the different mortality rates between the 2 groups, the treatment benefit of TMVr in the joint Bayesian model was greater than was estimated in the primary analysis at each time point. At 1 month, the mean treatment difference between TMVr and standard care in the KCCQ-OS was 18.5 points (95% CrI: 14.3 to 22.7 points) using the joint approach, compared with 15.9 points in the original model (Online Table 3). Furthermore, after accounting for the competing risk of mortality in the joint models, there was no attenuation in the treatment benefit of TMVr over time, with a mean difference in the KCCQ-OS of 18.7 points (95% CrI: 14.1 to 23.3 points) at 12 months and 18.9 points (95% CrI: 11.4 to 26.0 points) at 24 months. Similar patterns were observed for the SF-36 PCS and MCS.
In the COAPT trial, TMVr led to a substantial reduction in mortality and HF hospitalizations over 24 months compared with standard care in patients with HF and 3+ to 4+ secondary MR (3). In this integrated substudy, we found that TMVr also provided substantial benefits in terms of symptoms, functional status, and quality of life. The health status benefit of TMVr was moderately large and fully evident by 1 month. Among surviving patients, there was a slight attenuation of the health status effect over 24 months of follow-up; nonetheless, there was substantial sustained health status benefit at 12- and 24-month follow-up. In secondary analyses that accounted for bias due to exclusion of patients who died, the expected health status benefit with TMVr over standard therapy alone was stable over the full 24-month follow-up period. Although deaths were common in both treatment groups due to advanced age, comorbidities, and underlying cardiomyopathy, a higher proportion of patients who were randomized to TMVr were alive with clinically meaningful improvement in health status at every follow-up time point. Indeed, our study suggests that only 5 patients would need to be treated with TMVr for 1 additional patient to be alive, with a moderately large health status improvement at 24 months. Finally, the health status benefits of TMVr compared with standard care were consistent across all pre-specified subgroups except cause of cardiomyopathy, for which patients with ischemic cardiomyopathy appeared to derive greater health status benefit from TMVr compared with patients with nonischemic cardiomyopathy.
Because COAPT was the first randomized trial to collect detailed disease-specific and generic health status data for patients with secondary MR randomized to TMVr versus standard care, there are limited data to serve as a basis for comparison. Two previous observational studies that included patients with both degenerative and functional MR (60% to 70% functional) found that TMVr was associated with mean reductions (improvements) of ∼13 points in Minnesota Living With Heart Failure Questionnaire score compared with baseline (20,21)—roughly equivalent to a 14-point increase in the KCCQ-OS (22). In a recent analysis from the ACC/STS Transcatheter Valve Therapy Registry, patients treated with TMVr (89% of whom had degenerative MR) had a mean change in KCCQ-OS of 25 points by 1 month, which remained stable among surviving patients through 1 year. Patients with degenerative MR appear to have greater improvements in health status after TMVr compared with those with functional MR, likely due to the greater relative contribution of the MR to the health status decrement, as opposed to any underlying cardiomyopathy. Importantly, in contrast to these prior studies, COAPT was restricted to patients with functional MR, provides longer-term health status follow-up, and is the only health status study to compare TMVr with a randomized standard care group.
In a study where mortality is not only high but differs between treatments, it is important to recognize the challenges of interpreting health status outcomes, which can only be observed in surviving patients. The 1-year health status analysis described in the main COAPT paper approached this challenge by imputing missing KCCQ-OS scores to patients who died from HF as the lowest KCCQ-OS value reported in that time period (3). This resulted in a worst-case scenario, but ignores the potential variability of health status scores of patients who died, had they survived. A second strategy (used in the current study) is to jointly model survival and health status, so that the informative missing data due to deaths can be integrated into the health status estimates. The resulting estimates thus denote the expected health status outcomes of the average COAPT patient had he or she survived. Finally, a third strategy (also employed in this study) is to examine combined outcomes (e.g., being alive with a clinically meaningful change in health status), which integrates the 2 most important considerations of patients and provides enhanced clinical interpretability.
First, COAPT was a nonblinded trial, which may introduce bias. Although lack of blinding should have minimal impact on outcomes such as mortality, there is a possibility for a placebo effect when examining patient-reported measures. Given the magnitude of sustained health status benefit and the concordance of the results with other less-subjective outcomes (including death and rehospitalization), it is unlikely that the health status benefit of TMVr is simply a placebo effect. Second, as discussed in the previous text, health status can only be measured in surviving patients. Although we performed analyses that integrated mortality with health status and that used deaths to inform the health status outcomes, the true health status of patients who died, had they survived, is not knowable. Third, the durability of the health status benefit of TMVr beyond 24 months is not known, which is an important consideration in patients with underlying cardiomyopathy and comorbidities. Given the design of COAPT (which permits crossover to TMVr among surviving patients in the standard care group after 24 months), reliable longer-term health status data will be limited to the TMVr group. Fourth, the interaction noted between cause of cardiomyopathy and health status should be considered exploratory as it may be a spurious finding, particularly given the number of secondary outcomes investigated.
Finally, the health status results observed in COAPT may not be generalizable to patients outside of the inclusion/exclusion criteria of the trial. Given the conflicting clinical results of the COAPT and MitraFR trials and the limited health status data collected in MitraFR, it will be important to investigate the health status outcomes of patients with HF and secondary MR treated with TMVr both in the real world and among patients who would have been ineligible for COAPT.
In the COAPT trial, edge-to-edge TMVr resulted in substantial health status improvement compared with standard care in symptomatic HF patients and 3+ to 4+ secondary MR. This benefit emerged early, was consistent across key subgroups, and was sustained through 24 months of follow-up. Considering the previously reported benefits of TMVr on survival and HF hospitalization, these health status findings further support TMVr as a valuable treatment option for HF patients with severe secondary MR who remain symptomatic despite maximally tolerated guideline-directed medical therapy.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: In patients with HF and secondary MR, edge-to-edge TMVr results in rapid and sustained improvement in health status as assessed in terms of symptoms, functional status, and quality of life.
TRANSLATIONAL OUTLOOK: Further research is needed to understand the generalizability of these outcomes outside the setting of a clinical trial when deployed in less experienced centers.
The COAPT trial was sponsored by Abbott and designed collaboratively by the principal investigators and the sponsor. The present analysis was conducted by academic investigators at Saint Luke’s Mid America Heart Institute. Dr. Arnold is supported by a Career Development Grant Award (K23 HL116799) from the National Heart, Lung, and Blood Institute. Dr. Spertus owns the copyright for the KCCQ; has equity interest in Health Outcomes Sciences; has received consulting income from Novartis, Bayer, AstraZeneca, V-wave, Corvia, and Janssen; has served on the Advisory Board for United Healthcare; and has served on the Board of Directors for Blue Cross Blue Shield of Kansas City. Dr. Magnuson has received research grant support from Edwards Lifesciences, Medtronic, Boston Scientific, Abbott, and CSI. Dr. Baron has received consulting income from Edwards Lifesciences; and has received research grant support from Boston Scientific. Dr. Kar has received research grant support from Abbott, Boston Scientific, Edwards Lifesciences, and Mitralign; and has received consulting income from Abbott and Boston Scientific. Drs. Lim and Abraham have received research grant support and consulting income from Abbott. Dr. Lindenfeld has received research grant support from AstraZeneca; and has received consulting income from Abbott, Edwards Lifesciences, Boston Scientific, RESMED, Relypsa, Boehringer Ingelheim, and V-Wave. Dr. Mack has served as co-primary investigator for the Partner Trial for Edwards Lifesciences and COAPT trial for Abbott; and has served as Study Chair for the APOLLO trial for Medtronic. Dr. Stone has received consulting income from Claret, Medical Development Technologies, Backbeat, Sirtex, Matrizyme, Miracor, Neovasc, V-wave, Shockwave, Valfix, TherOx, Reva, Vascular Dynamics, Robocath, HeartFlow, Gore, Ablative Solutions, and Ancora; has received speaker honoraria from Terumo and Amaranth; has received Advisory Board fees from QOOL Therapeutics and SpectraWAVE; has equity/options in Ancora, Cagent, Qool Therapeutics, Aria, Caliber, MedFocus family of funds, Biostar family of funds, Applied Therapeutics, and SpectraWAVE; has served as director of SpectraWave; and his employer, Columbia University, receives royalties for sale of the MitraClip from Abbott. Dr. Cohen has received research grant support from Edwards Lifesciences, Medtronic, Boston Scientific, and Abbott; and has received consulting fees from Medtronic and Edwards Lifesciences. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
Listen to this manuscript's audio summary by Editor-in-Chief Dr. Valentin Fuster on JACC.org.
- Abbreviations and Acronyms
- confidence interval
- credible interval
- heart failure
- Kansas City Cardiomyopathy Questionnaire
- Kansas City Cardiomyopathy Questionnaire-overall summary score
- mitral regurgitation
- number needed to treat
- Short-Form 36 Health Survey
- SF-36 PCS
- Short-Form 36 Health Survey physical summary score
- SF-36 MCS
- Short-Form 36 Health Survey mental summary score
- transcatheter mitral valve repair
- Received January 24, 2019.
- Revision received February 13, 2019.
- Accepted February 13, 2019.
- 2019 American College of Cardiology Foundation
- Vakil K.,
- Roukoz H.,
- Sarraf M.,
- et al.
- Deuschl F.,
- Schofer N.,
- Lubos E.,
- Blankenberg S.,
- Schafer U.
- Stone G.W.,
- Lindenfeld J.,
- Abraham W.T.,
- et al.
- Stevenson L.W.,
- Hellkamp A.S.,
- Leier C.V.,
- et al.
- Arnold S.V.,
- Li Z.,
- Vemulapalli S.,
- et al.
- Mack M.J.,
- Abraham W.T.,
- Lindenfeld J.,
- et al.
- Green C.P.,
- Porter C.B.,
- Bresnahan D.R.,
- Spertus J.A.
- Heidenreich P.A.,
- Spertus J.A.,
- Jones P.G.,
- et al.
- Ware J.,
- Kosinski M.,
- Bjorner J.B.,
- Turner-Bowkes D.M.,
- Gandek B.,
- Maruish M.E.
- Arnold S.V.,
- Spertus J.A.,
- Lei Y.,
- et al.
- ↵Spertus JV, Hatfield LA, Cohen DJ, et al. Integrating Quality of Life and Survival Outcomes Cardiovascular Clinical Trials: Results from the PARTNER Trial. Available at: https://www.biorxiv.org/content/early/2019/01/12/513895. Accessed January 12, 2019.
- Heidenreich P.A.,
- Spertus J.A.,
- Jones P.G.,
- et al.
- Kosiborod M.,
- Soto G.E.,
- Jones P.G.,
- et al.
- Soto G.E.,
- Jones P.,
- Weintraub W.S.,
- Krumholz H.M.,
- Spertus J.A.
- Gelman A.,
- Carlin J.B.,
- Stern H.S.,
- Dunson D.B.,
- Vehtari A.,
- Rubin D.B.
- Metze C.,
- Matzik A.S.,
- Scherner M.,
- et al.
- Maisano F.,
- Franzen O.,
- Baldus S.,
- et al.
- Nassif M.E.,
- Tang Y.,
- Cleland J.G.,
- et al.