Author + information
- Received September 10, 2014
- Revision received April 14, 2015
- Accepted April 15, 2015
- Published online June 23, 2015.
- Nirmalatiban Parthiban∗,†,
- Adrian Esterman, PhD‡,
- Rajiv Mahajan, MD, PhD∗,
- Darragh J. Twomey, MBBS∗,
- Rajeev K. Pathak, MBBS∗,
- Dennis H. Lau, MBBS, PhD∗,
- Kurt C. Roberts-Thomson, MBBS, PhD∗,
- Glenn D. Young, MBBS∗,
- Prashanthan Sanders, MBBS, PhD∗ and
- Anand N. Ganesan, MBBS PhD∗∗ ()
- ∗Centre for Heart Rhythm Disorders, South Australian Health and Medical Research Institute, University of Adelaide and Royal Adelaide Hospital, Adelaide, Australia
- †Perdana University-Royal College of Surgeons in Ireland, Serdang, Malaysia
- ‡University of South Australia, Adelaide, Australia
- ↵∗Reprint requests and correspondence:
Dr. Anand N. Ganesan, Centre for Heart Rhythm Disorders, University of Adelaide, Adelaide SA 5000, Australia.
Background Remote monitoring (RM) of implantable cardioverter-defibrillators (ICD) is an established technology integrated into clinical practice. One recent randomized controlled trial (RCT) and several large device database studies have demonstrated a powerful survival advantage for ICD patients undergoing RM compared with those receiving conventional in-office (IO) follow-up.
Objectives This study sought to conduct a systematic published data review and meta-analysis of RCTs comparing RM with IO follow-up.
Methods Electronic databases and reference lists were searched for RCTs reporting clinical outcomes in ICD patients who did or did not undergo RM. Data were extracted from 9 RCTs, including 6,469 patients, 3,496 of whom were randomized to RM and 2,973 to IO follow-up.
Results In the RCT setting, RM demonstrated clinical outcomes comparable with office follow-up in terms of all-cause mortality (odds ratio [OR]: 0.83; p = 0.285), cardiovascular mortality (OR: 0.66; p = 0.103), and hospitalization (OR: 0.83; p = 0.196). However, a reduction in all-cause mortality was noted in the 3 trials using home monitoring (OR: 0.65; p = 0.021) with daily verification of transmission. Although the odds of receiving any ICD shock were similar in RM and IO patients (OR: 1.05; p = 0.86), the odds of inappropriate shock were reduced in RM patients (OR: 0.55; p = 0.002).
Conclusions Meta-analysis of RCTs demonstrates that RM and IO follow-up showed comparable overall outcomes related to patient safety and survival, with a potential survival benefit in RCTs using daily transmission verification. RM benefits include more rapid clinical event detection and a reduction in inappropriate shocks.
Implantable cardioverter-defibrillators (ICD) have become the standard of care therapy for primary and secondary prevention of sudden cardiac death (1). Conventionally, in-office (IO) follow-up on at least an every 3- to 6-month basis has been recommended for ICD interrogation, review of device data, and programmed parameters, as well as assessment of system function (2–4).
Remote monitoring (RM) of ICD devices has been proposed as an alternative strategy to reduce the need for routine device follow-up visits while providing continuous surveillance and immediate problem notification (5). With this technology, ICDs can be interrogated automatically using wireless data transfer to the remote monitor. Patient diagnostic information is then transmitted to a central server that may be accessed by treating clinicians through an Internet-based interface or provide automatically generated clinician alerts (6).
Recently, clinical data presented in the IN-TIME (Influence of Home Monitoring on Mortality and Morbidity in Heart Failure Patients with Impaired Left Ventricular Function) trial suggested that RM could potentially lead to a decisive survival advantage in ICD patients (7). This powerful survival benefit is supported by data from large-scale national device registries, showing that RM may lead to a significant survival advantage over patients not using RM (8–10). The ALTITUDE study, for example, followed a nonrandomized cohort of 69,556 patients implanted with ICD or cardiac resynchronization therapy (CRT) devices with defibrillator capability (CRT-D) (Boston Scientific Corporation, Natick, Massachusetts), and identified a striking 50% reduction in mortality (ICD hazard ratio [HR]: 0.56; CRT-D HR: 0.45; p < 0.0001) in remote networked patients, compared with non-networked device recipients (10). Similar mortality reductions with RM use have been seen in 2 other national device databases collectively enrolling more than 100,000 ICD patients (8,9).
Until recently, there has been insufficient randomized controlled trial (RCT) evidence to evaluate the overall impact of RM on clinical outcomes in ICD patients. In the current study, we conducted a systematic review and meta-analysis of RCTs comparing clinical outcomes in ICD patients undergoing RM with those receiving conventional IO follow-up. We specifically sought to evaluate the impact of RM on all-cause and cardiovascular (CV) mortality, hospitalization, unscheduled clinic visits, atrial arrhythmia detection, device shocks, and the time taken to clinical decision or clinical event detection.
We conducted a systematic search of PubMed, Embase, Scopus, Web of Science, and the Cochrane databases to identify RCTs comparing RM with conventional IO follow-up in ICD patients. The search was conducted with the assistance of a research librarian, and the details of the search grid are outlined in the Online Appendix. Databases were last accessed on July 30, 2014 and results were updated after the publication of the IN-TIME trial on August 16, 2014.
Two authors (A.N.G., N.P.) reviewed titles and abstracts retrieved from our search strategy and selected RCTs reporting on clinical outcomes of home monitoring (treatment) of ICDs compared with conventional IO follow-up (control). RCTs were included if results were published in peer-reviewed journal articles or as published abstracts with extractable data. Studies were excluded if they provided outcome data only from nonrandomized cohorts or case series, evaluated ICDs but not RM, or evaluated RM in contexts other than ICD patients. Figure 1 shows the number and reasons for exclusion of publications extracted from the search strategy. For included trials, all-cause mortality, hospitalizations, unscheduled visits, shock delivery, and atrial fibrillation detections were extracted by 2 authors (N.P., A.N.G.). Study quality was assessed on the basis of adherence to the Consolidated Standards of Reporting Trials (CONSORT) statement.
Statistical analysis was performed with Comprehensive Meta-Analysis, version 2 (Biostat, Inc., Englewood, New Jersey). Odds ratios (OR) were used for dichotomous variables. The I2 statistic was used as a measure of variability in observed effect estimates attributable to between-study heterogeneity (11). For variables exhibiting mild heterogeneity (I2 ≤25%), pooled estimates were derived with fixed-effects models. For variables exhibiting more than moderate heterogeneity (I2 >25%), pooled estimates were derived with random-effects models, according to the method of DerSimonian and Laird (12). For time to detection of clinical event/clinical decision data, variances were imputed from interquartile ranges and 95% confidence intervals (CIs) according to the method of Deeks et al. (13).
A total of 4,376 citations were retrieved after exclusion of duplicates, then 3,491 citations were excluded after initial screening of abstracts and titles on general criteria, as related to topics other than home monitoring of ICDs (Figure 1). Of 885 citations selected for a secondary review, we identified 20 journal articles referencing 8 published RCTs. An additional completed RCT, the EVATEL (Evaluation of Tele Follow-up) trial, was identified as a peer-reviewed published abstract with extractable data (14).
Baseline characteristics of RCTs
A total of 6,469 patients were included in the 9 RCTs, all published in the years 2010 through 2014 (Tables 1 and 2⇓) (7,14–21). The sample size for each study varied, from small (Perl et al. , n = 36) to much larger studies (Crossley et al. , n = 1,997). Except for the studies of Al-Khatib et al. (15) and Perl et al. (19), most RCTs (78%) had been conducted at multiple centers. The mean or median duration of follow-up varied from 12 to 26 months. Inclusion criteria for the studies are listed in Table 1. The mean age of patients in the trials varied from 59 to 68 years. The proportion of male patients in the studies ranged from 43% to 88%. The mean left ventricular ejection fraction varied from 26% to 35% and the mean proportion of patients with underlying ischemic cardiomyopathy ranged from 46% to 69%. Study quality overall was strong based on adherence to the reporting principles of the CONSORT statement (Online Appendix).
All-cause and CV mortality
All-cause mortality was reported in 7 trials enrolling 4,932 patients (7,15–18,20,21). The OR for mortality with RM was not statistically significant from IO follow-up (OR: 0.83; 95% CI: 0.58 to 1.17; p = 0.285; I2 = 34.8%) (Figure 2). CV mortality was reported in 4 trials (Central Illustration). The OR for CV death showed a nonsignificant trend toward a reduction in CV mortality with RM (OR: 0.66; 95% CI: 0.41 to 1.09; p = 0.103; I2 = 35.1%).
To examine the stability of these findings, a series of sensitivity analyses was undertaken. The only trial to show a statistically significant reduction in mortality was IN-TIME (OR: 0.35; 95% CI: 0.17 to 0.73; p = 0.005) (7). Exclusion of IN-TIME eliminated the trend toward improvement in all-cause mortality (OR: 0.96; 95% CI: 0.74 to 1.24; p = 0.737), and eliminated between-study heterogeneity (I2 = 0.0%). Exclusion of IN-TIME also reduced the trend toward decreased CV mortality with RM (OR: 0.92; 95% CI: 0.50 to 1.70; p = 0.800) and eliminated between-study heterogeneity (I2 = 0.0%). To examine the effect of trial size, we compared all-cause mortality in the 3 trials with >500 patients with 4 trials with <500 patients. No difference in all-cause mortality was identified between trials with >500 patients and trials with <500 patients (p = 0.939).
We also examined the effect of competing technologies. The 3 trials using home monitoring from Biotronik SE & Co. KG (Berlin, Germany), which incorporated daily verification of home monitoring transmission, showed a reduction in mortality with RM (OR: 0.65; 95% CI: 0.45 to 0.94; p = 0.021) (7,16,21), but the 4 trials using CareLink (Medtronic, Inc., Tempe, Arizona) did not (OR: 1.07; 95% CI: 0.77 to 1.49; p = 0.767) (15,17,18,20).
Impact of RM
Hospitalization data were reported as an endpoint in 7 studies enrolling 5,372 patients. Pooled data from all 7 studies showed no significant reduction in the odds of hospitalization with RM (OR: 0.83; 95% CI: 0.63 to 1.10; p = 0.196; I2 = 72.2%) (Figure 3A). The only trial reporting hospitalization data to show a significant reduction in hospitalizations with RM was the TRUST (Lumos-T Safely Reduces Routine Office Device Follow-up) trial (OR: 0.41; 95% CI: 0.30 to 0.56; p < 0.001) (16). Exclusion of TRUST reduced the trend toward a reduction in hospitalization (OR: 0.95: 95% CI: 0.84 to 1.08; p = 0.442) and eliminated between-study heterogeneity (I2 = 0.00%).
The impact of RM on unscheduled office visit burden was reported in 5 studies. Overall, RM showed a nonsignificant trend toward an increase in unscheduled visits (OR: 1.29; 95% CI: 0.99 to 1.67; p = 0.061; I2 = 90.7%) (Figure 3B). Only 1 trial, EVOLO (Evolution of Management Strategies of Heart Failure Patients with Implantable Defibrillators) (18), showed a reduction in unscheduled visits with RM (OR: 0.64; 95% CI: 0.48 to 0.86; p = 0.003). Exclusion of EVOLVO showed that RM use increased unscheduled office visits (OR: 1.48; 95% CI: 1.16 to 1.89; p = 0.001).
The prevalence of patients receiving 1 or more ICD shocks was reported in 4 studies. The odds of receiving any ICD shock were similar in RM patients to IO patients (OR: 1.05; 95% CI: 0.62 to 1.78; p = 0.861; I2 = 52.5%) (Figure 4). However, 3 studies including 2,085 patients reporting on inappropriate shock recipients found a statistically significant reduction with RM when compared with IO follow-up (OR: 0.55; 95% CI: 0.38 to 0.80; p = 0.002; I2 = 0.002) (Central Illustration).
Atrial arrhythmia detection was assessed in 6 RCTs enrolling 4,268 patients. There was no statistically significant change in the prevalence of atrial arrhythmia detection between RM and IO follow-up patients (OR: 1.24; 95% CI: 0.89 to 1.71; p = 0.203; I2 = 39.8%) (Central Illustration).
An anticipated benefit of RM is that it may facilitate early detection of clinical events by ICD alerts, leading to the capacity to make early clinical decisions. Time to clinical decision and/or event detection was reported in 4 trials containing 3,176 patients (16–18,20). A significant decrease in the time to clinical decision/event detection was noted in each of these trials with RM compared with IO follow-up, with a mean difference in days to clinical decision/event detection of -27.1 days (95% CI: -40.1 to -13.3 days; p < 0.001) (Central Illustration).
A potentially important consideration in the effect of RM is transmission rate and efficacy. The RCTs in the current study predominantly used 2 proprietary technologies. The CareLink system (Medtronic, Inc.) involves automated transmissions of pre-specified clinician alerts for specific clinic events. The earliest CareLink study (Al-Khatib et al. ) used a version of RM requiring patient interaction that was superseded and replaced with automated transmission in the later studies. Three RCTs used Biotronik RM technology. This technology also generates automated pre-specified clinician alerts. However, a specific feature with this technology is that it involves daily transmission from the device to the home monitor, with failure to transmit generating an alert to the clinician. Transmission efficacy and transmission delay data were presented in a variety of formats, preventing meta-analysis of these data. The results are presented in Table 3. Transmission failure data were reported in 8 of 9 trials. The lowest rates of failed transmission were seen in the IN-TIME trial, with failure to transmit noted in 3 of 333 (0.9%) of RM patients. The study with the highest rates of failed transmission was the CONNECT (Clinical Evaluation of Remote Notification to Reduce Time to Clinical Decision) trial, in which up to 45% of alerts in the RM group were not transmitted.
RM has in recent years become an established technology that is increasingly integrated into the routine clinical care of ICD patients. Several clinical studies have demonstrated that ICD RM may offer a variety of potential benefits (Central Illustration) including earlier notification of clinical (17) and device (16) events, reductions in hospital and resource use, and improvements in follow-up adherence (9). On the basis of these data, RM has been endorsed in current guidelines and consensus documents as a safe alternative to IO follow-up of ICD patients (2–4).
Despite these potential advantages, RM is not available to large numbers of patients around the world. In North America, where RM is reimbursed (22), it is estimated that RM is not used in as many as 50% of eligible patients (5). Significant variation exists in the uptake of RM between geographic regions and health care systems (23). Possible reasons include clinician reluctance to adopt RM because of lack of familiarity with the technology, concerns about data management and legal liability, and cost (5). RM availability is even more constrained in many regions outside North America, where reimbursement is frequently unavailable to clinicians undertaking RM (3,6,21).
Impact of RM on survival
One possible reason for the incomplete adoption of RM may be uncertainty about whether RM has a significant effect on patient clinical outcomes. An intriguing possibility, raised by the recently published IN-TIME trial, is that RM could drive a substantial improvement in survival in ICD patients (7). In IN-TIME, RM led to an improved heart failure composite score, a finding driven by a sharp decrease in all-cause mortality in the RM group (10 deaths in the RM arm vs. 27 deaths in the IO follow-up arm) (7). Although a single specific mechanism responsible for the mortality benefit was not clear, the authors postulated that the improved outcomes with RM may have occurred via a combination of early ventricular and atrial arrhythmia detection, early recognition of suboptimal biventricular pacing on CRT-D devices, and increased telephone contact with patients prompted by telemonitoring alerts (7). No statistical difference between RM and IO-follow-up was identified for other heart failure measures including heart failure hospital admission, New York Heart Association functional class, and global self-assessment (7).
The findings of IN-TIME have been supported by nonrandomized data from large-scale device database registries. The ALTITUDE study, which followed 69,556 patients, showed a statistically significant 50% reduction in mortality (ICD HR: 0.56; CRT-D HR: 0.45; p < 0.001) in networked patients (10). The findings presented in ALTITUDE were recently confirmed by PREDICT RM (Patient Related Determinants of ICD Remote Monitoring Utilization and Outcomes), a subgroup analysis in which 37,742 patients from the ALTITUDE database were studied with matched outcome data from the National Cardiovascular Data Registry. In PREDICT RM, RM was associated with reduced mortality (ICD HR: 0.60; CRT-D HR: 0.71) (8,9). Similar data also have been presented for 148,976 patients on the Merlin patient care network database (St. Jude Medical, Inc., Sylmar, California), with RM use associated with increased survival (ICD HR: 2.51; CRT-D HR: 2.44; both p < 0.001) (9). However, none of the analyses of the large RM databases were randomized. The mortality of the group undergoing RM was compared with patients who chose not to participate in RM or where the service was not available. This may have potentially resulted in selection bias and this should be considered when interpreting outcomes of these studies.
In our study, we were unable to identify a significant overall all-cause mortality benefit in meta-analysis of the 7 RCTs reporting mortality data from 4,932 patients, with a nonsignificant reduction in the odds of death with RM (OR: 0.83; 95% CI: 0.58 to 1.24; p = 0.74). A nonsignificant trend toward a reduction in the odds of CV death with RM was observed (OR: 0.69; 95% CI: 0.37 to 1.31; p = 0.26), although this outcome was reported in only 4 trials, so there may have been insufficient power to definitively assess this outcome.
However, a highly significant mortality benefit was seen in the subset of trials using the Biotronik RM system (OR: 0.65; 95% CI: 0.45 to 0.94; p = 0.021) (7,16,21). This benefit was driven by the significant reduction in the rate of death in the IN-TIME trial in RM patients (OR: 0.35; p = 0.005). A possible mechanism to explain this advantage could include daily verification of RM transmission and/or the defined response mechanism to RM-generated alerts, and represents an area to investigate in future trials. Recent registry data have pointed to the possibility that intensity of RM use may be a predictor of outcomes (9). In the current study, transmission failure and delays were identified in early trials using alternative technologies (Table 3), and may represent a potential mechanism for the differences in survival benefit identified in our study.
Impact of RM
An important potential clinical advantage of remote device monitoring identified in observational studies is the potential for a reduction in hospitalization. In PREDICT RM, RM was associated with a significant reduction in the rate of hospitalization (HR: 0.81; 95% CI: 0.79 to 0.83) (8). In the current study, a statistically significant reduction in the rate of hospitalization was not observed in the included RCTs. Between-study heterogeneity was driven by a single trial, TRUST; excluding this trial reduced the trend toward a reduction in hospitalization (OR: 0.95; 95% CI: 0.84 to 1.08; p = 0.442) and eliminated between-study heterogeneity (I2 = 0.00%). Again, the reasons for this difference between the RCTs, and with previous observational registry data, are not clear. Possible explanations include differences in patient selection in RCTs and potentially more rigorous in-office follow-up occurring in the RCT context than in community-based follow-up.
A potential advantage for RM is its ability for continuous surveillance to identify problems early, thereby minimizing adverse outcomes including shocks. In the included RCTs, RM was associated with a similar level of shocks overall. Interestingly, however, a significant reduction in inappropriate shocks was recorded in the 3 RCTs providing information on this outcome. Possible reasons for this could include earlier detection of device/lead abnormalities or timely detection and treatment of atrial arrhythmias.
An important aspect of RM technology is its ability to detect abnormalities and clinical alerts early, presenting the possibility of expedited clinical intervention. In our study, RM had a significant advantage in time to detection of clinical events or time to clinical decision of –27.1 days (95% CI: –40.1 to –13.3 days; p < 0.001). Early detection of events may offer significant benefit in individual patients by allowing tailored clinical intervention to prevent clinical deterioration, such as early anticoagulation in patients with atrial fibrillation, or optimization of device programming to maximize biventricular pacing in patients with CRT-D devices.
In our study, we sought to evaluate the clinical impact of RM in ICD patients in RCTs, which represent the highest standard of clinical evidence. At the present time, RM of ICDs is an established technology, with broad, although not universal, acceptance and availability among many patients and clinicians worldwide. The results of our study provide further evidence toward the safety record of RM. In our study, patients undergoing RM and IO follow-up experienced comparable rates of all-cause mortality, CV mortality, and hospitalizations. At a minimum, our data provide convincing support for the noninferiority of RM compared with IO follow-up. An intriguing possibility of a substantial mortality benefit with RM has been seen in the IN-TIME trial, suggesting that daily verification of RM transmission and a defined response mechanism to RM-generated alerts may be important in producing maximal benefit from the technology. These benefits need verification in future studies. Additional advantages seen with RM included statistically significant reductions in inappropriate shock and earlier detection of clinical events. The relative equivalence in overall clinical outcomes with guidelines-consistent office-based follow-up should provide reassurance to patients and clinicians in health systems and geographic regions where RM is not available for logistical reasons or cost.
The study was conducted as a systematic review and meta-analysis of RCTs with available data reporting clinical outcomes with RM in ICD patients in the RCT setting. A limitation of this kind of study is that analyses are performed on reported data in the published reports rather than on primary study data. A further limitation of this kind of study is the possibility of publication or “file-drawer” bias, with the possibility that studies favorable to the intervention are made available for publication. Nevertheless, the results of our study do summarize the outcomes of a number of well-conducted randomized studies in >6,000 patients.
RM of ICD patients is a technology that is now established and part of routine clinical practice in many centers worldwide. The results of RM RCTs provide strong evidence of the safety of RM, with the suggestion of a potential mortality benefit with technologies using daily transmission verification. Future investigations are needed to further explore the key aspects of RM required to maximize the clinical benefit of this technology.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: Meta-analysis of randomized trials suggests that although RM may be associated with overall comparable levels of survival among patients with implanted cardioverter-defibrillators, verification of RM transmission data and appropriately coupled are potentially critical determinants of whether clinical benefit is maximized benefit.
TRANSLATIONAL OUTLOOK: Further studies are needed to clarify whether improved outcomes associated with RM technology derive from earlier detection of subclinical arrhythmias, avoidance of inappropriate shocks, or specific therapeutic interventions.
The authors acknowledge the support of Mr. Michael Draper, of the University of Adelaide Barr-Smith Library, for assistance with the development of the study search.
Drs. Mahajan and Pathak are supported by the Australian Postgraduate Award from the University of Adelaide. Drs. Mahajan, Twomey, and Pathak are supported by the Leo J. Mahar Electrophysiology Scholarship from the University of Adelaide. Dr. Lau is supported by a Postdoctoral Fellowship from the National Health and Medical Research Council of Australia. Dr. Sanders is supported by a Practitioner Fellowship from the National Health and Medical Research Council of Australia, and the National Heart Foundation of Australia; has served on the advisory board of Biosense-Webster, Medtronic, St. Jude Medical, Sanofi, and Merck, Sharpe and Dohme; has received lecture and/or consulting fees from Biosense-Webster, Medtronic, St. Jude Medical, Boston Scientific, Merck, Sharpe and Dohme, Biotronik, and Sanofi; and has received research funding from Medtronic, St. Jude Medical, Boston Scientific, Biotronik, and Sorin. Dr. Ganesan is supported by an Australian Early Career Health Practitioner Fellowship from the National Health and Medical Research Council of Australia. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- cardiac resynchronization therapy
- hazard ratio
- implantable cardioverter-defibrillator
- odds ratio
- randomized controlled trial
- remote monitoring
- Received September 10, 2014.
- Revision received April 14, 2015.
- Accepted April 15, 2015.
- American College of Cardiology Foundation
- Epstein A.E.,
- DiMarco J.P.,
- Ellenbogen K.A.,
- et al.
- Dubner S.,
- Auricchio A.,
- Steinberg J.S.,
- et al.
- Varma N.,
- Brugada P.
- ↵Akar JG, Bao H, Jones PW, et al. Use of remote monitoring is associated with improved outcomes among patients with implantable cardioverter defibrillators. Paper presented at: the Heart Rhythm Society Annual Scientific Sessions; May 10, 2014; San Francisco, CA.
- Varma N.,
- Piccini J.P.,
- Snell J.,
- Fischer A.,
- Dalal N.,
- Mittal S.
- Saxon L.A.,
- Hayes D.L.,
- Gilliam F.R.,
- et al.
- Deeks J.J.,
- Higgins J.P.T.,
- Altman D.G.
- Mabo P.,
- Defaye P.,
- Sadoul N.,
- et al.
- Al-Khatib S.M.,
- Piccini J.P.,
- Knight D.,
- Stewart M.,
- Clapp-Channing N.,
- Sanders G.D.
- Varma N.,
- Epstein A.E.,
- Irimpen A.,
- Schweikert R.,
- Love C.
- Crossley G.H.,
- Boyle A.,
- Vitense H.,
- Chang Y.,
- Mead R.H.
- Landolina M.,
- Perego G.B.,
- Lunati M.,
- et al.
- Guedon-Moreau L.,
- Lacroix D.,
- Sadoul N.,
- et al.
- Slotwiner D.,
- Wilkoff B.
- Akar J.G.,
- Bao H.,
- Jones P.,
- et al.