best of jacc: clinical electrophysiology · need for pericardiocentesi s. post-procedurally, dapt...

Best of JACC: Clinical Electrophysiology

Enjoy the Top Cited Articles* of 2017

*Top cited articles in JACC: Clinical Electrophysiology 2017

jacc-electrophysiology.org

Thrombus Formation AfterLeft Atrial Appendage OcclusionWith the Amplatzer Amulet DeviceAlexander Sedaghat, MD, Jan-Wilko Schrickel, MD, René Andrié, MD, Robert Schueler, MD, Georg Nickenig, MD,Christoph Hammerstingl, MD

ABSTRACT

OBJECTIVES This study sought to define the ideal post-procedural anticoagulant regime and to systematically study

the incidence of device-related thrombus.

BACKGROUND Left atrial appendage occlusion (LAAo) is an alternative to life-long oral anticoagulation in selected

patients with atrial fibrillation.

METHODS This study included 24 atrial fibrillation patients (ages 79� 8 years; 75%male, CHA2DS2VASc [Congestive Heart

Failure, Hypertension, Age$75 Years, Diabetes Mellitus, Previous Stroke or Transient Ischemic Attack or Thromboembolism,

Vascular Disease, Age 65 to 74 Years. Sex] score: 4.3 � 1.5, HAS-BLED [Hypertension, Abnormal Renal and Liver Function,

Stroke, Bleeding, Labile International NormalizedRatio, Elderly, Drugs or Alcohol] score: 3.6�0.8) after LAAowith theuse of

the Amplatzer Amulet system. Dual antiplatelet therapy for 3 months was prescribed in 95.6% of the cases.

RESULTS Transesophageal echocardiography identified a high rate of device adherent thrombi (16.7%, n ¼ 4 of 23)

after a mean of 11.0 � 8.2 weeks. Thrombus formation occurred under dual antiplatelet therapy (3 of 4) or clopidogrel

monotherapy (1 of 4). When compared with patients without thrombi, echocardiography showed higher degrees of

spontaneous echo contrast grades within the LAA (3.0 � 1.0 vs. 1.3 � 1.1), lower LAA peak emptying velocities

(17.5 � 5.0 cm/s vs. 48.3 � 21.1 cm/s), and decreased left ventricular function (39 � 10% vs. 50 � 13%) in patients

with device-related thrombus. All thrombi were observed within the untrabeculated region of the LAA ostium between

the left upper pulmonary vein ridge and the occluder disc, indicating suboptimal LAA occlusion.

CONCLUSIONS Device-related thrombus is a frequent finding after LAAo with the Amplatzer Amulet device (St. Jude

Medical, St. Paul, Minnesota). Our results emphasize the need for an optimized post-LAAo anticoagulation regimen, a

revised implantation strategy, and possibly modified patient selection criteria. (J Am Coll Cardiol EP 2017;3:71–5) © 2017

by the American College of Cardiology Foundation.

L eft atrial appendage occlusion (LAAo) hasbeen introduced in clinical practice as analternative to life-long oral anticoagulation

(OAC) focused on patients that are unamenable forOAC (1). The concept of LAAo is intriguing as it prom-ises the combination of reduced thromboembolicevents and a potential benefit concerning bleedingevents when OAC is stopped after implantation.

However, the optimal regimen of post-proceduralanticoagulation or platelet inhibition after LAAo isan unresolved issue and manufacturers’ recommen-dations vary from initiation of vitamin K antagonisttherapy for 45 days followed by dual antiplatelettherapy (DAPT) after implantation of the Watchmandevice (Boston Scientific, Natick, Massachusetts), toDAPT alone up to 6 months after LAAo with the

From the Medizinische Klinik und Poliklinik II, Herzzentrum, Universitätsklinikum Bonn, Rheinische-Friedrich-Wilhelms Uni-

versität Bonn, Bonn, Germany. The authors have reported that they have no relationships relevant to the contents of this paper to

disclose. Drs. Sedaghat and Schrickel contributed equally to this work.

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.

Manuscript received March 21, 2016; revised manuscript received April 19, 2016, accepted May 20, 2016.

J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y VO L . 3 , N O . 1 , 2 0 1 7

ª 2 0 1 7 B Y T H E AM E R I C A N C O L L E G E O F C A R D I O L O G Y F O U N D A T I O N

P U B L I S H E D B Y E L S E V I E R

I S S N 2 4 0 5 - 5 0 0 X / $ 3 6 . 0 0

h t t p : / / d x . d o i . o r g / 1 0 . 1 0 1 6 / j . j a c e p . 2 0 1 6 . 0 5 . 0 0 6

Amplatzer Amulet system (St. Jude Medical,St. Paul, Minnesota) (2,3). Also, LAAo hasyet to prove its efficacy concerning thereduction of major bleeding events in high-risk atrial fibrillation patients (4) and is incompetition with the improved safety ofdirect OAC (5).

Thrombus formation on the AmplatzerCardiac Plug device (St. Jude Medical) hasbeen described in up to 17% of cases (6) andthe new Amplatzer Amulet device has been

recently introduced with novel design features alsoaiming to reduce thrombogenicity (7).

The incidence of device-related thrombus (DRT)formation with the new Amulet occluder has not beenexamined in clinical routine with serial trans-esophageal echocardiographic examinations. There-fore, we investigated safety and efficacy of LAAo withthe Amulet system in a prospective cohort study withregard to thrombus formation.

METHODS

We performed clinical and echocardiographic follow-up in patients who underwent LAAo with the use ofthe Amulet device at our center between November2014 and November 2015. All patients had to provideinformed consent for participation to our LAAo regis-try, whichwas approved by the local ethics committee.

STATISTICAL ANALYSIS. We present a series of 24patients, which does not allow for detailed, inferen-tial statistical analysis. Values are presented as mean� SD if normally distributed; categorical variables aregiven as frequencies and percentages.

RESULTS

PATIENTS UNDERGOING LAAO. Twenty-fourpatients underwent LAAo with the Amplatzer Amuletdevice between January 2014 and May 2015. Thesepatients were 78 � 9 years of age, predominantlymale (75%), and at a high thromboembolic as well asbleeding risk (CHA2DS2VASc [Congestive Heart Fail-ure, Hypertension, Age $75 Years, Diabetes Mellitus,Previous Stroke or Transient Ischemic Attack orThromboembolism, Vascular Disease, Age 65 to 74Years. Sex] score: 4.3 � 1.5, HAS-BLED [Hypertension,Abnormal Renal and Liver Function, Stroke, Bleeding,Labile International Normalized Ratio, Elderly, Drugsor Alcohol] score: 3.6 � 0.8). The indication for LAAowas driven by a history of bleeding events in 75% ofpatients, or an anticipated high risk of bleeding in

25% of patients. Levels of serum creatinine were 1.5 �0.6 mg/dl.

LAAo was performed under conscious sedation inall patients with a mean procedure time of 37 � 12min; the mean device size was 24 � 3 mm. DuringLAAo, 37 � 36 ml of contrast dye were used.

LAAo was successful in all patients. Peri-interventional complications occurred in 2 of 24 ofpatients (8.3%), consisting of minor pericardial effu-sion in 1 patient and 1 pericardial tamponade with theneed for pericardiocentesis. Post-procedurally, DAPTwas prescribed for a minimum of 3 months in themajority of patients (23 of 24, 95.8%). One patient wasmaintained on OAC based on patient’s preferences.

COMPARISON OF PATIENTS WITH AND WITHOUT

DRT. Transesophageal echocardiography identified 4patients (16.7%) with large DRT after a mean of 11.0 �8.2 weeks (4, 4, 16, and 20 weeks). DRT were detectedprior to electrocardioversion in 1 case (Patient #1) andduring routine follow-up in the other 3 patients (Pa-tients #2 to #4). Of note, Patient #4 had suffered froma transient ischemic attack (TIA) 2 weeks beforefollow-up. Three of the 4 patients were still underdual antiplatelet therapy when DRT was diagnosed; 1patient was on monotherapy with clopidogrel afterplanned cessation of DAPT (Table 1).

PREDICTORS FOR THROMBUS FORMATION. Whencomparing groups according to the occurrence ofDRT, baseline echocardiography showed that patientswith thrombi had lower left ventricular ejectionfractions (39 � 10% vs. 50 � 13%), higher degrees ofspontaneous echo contrast within the LAA (3.0 � 0.8vs. 1.3 � 1.1), and lower LAA peak emptying velocities(17.5 � 5.0 cm/s vs. 48.3 � 21.1 cm/s). Additionally, pa-tients with thrombi on the Amulet device more oftenhadahistoryofLAA thrombi (75.0%vs. 19.2%) (Table 2).

Of interest, all thrombi were located within theangle between the left upper pulmonary vein–ridgeand the occluder disc in the untrabeculated transitionbetween LAA ostium and neck, resulting in incom-plete sealing of the LAA ostium in these patients(Figure 1). In contrast, the rate of incomplete LAAsealing in patients without thrombi was 41.2%(Table 2).

TREATMENT REGIMEN OF DRT. Due to the safetyprofile and their potential to resolve LAA thrombi (8),we initiated OAC by use of direct acting OAC in allpatients. Patient #1 was treated with apixaban (5 mgtwice daily), whereas Patient #2 was prescribed 30 mgof edoxaban, due to chronic renal disease. In 2patients (Patients #3 and #4), reduced-dose dabiga-tran therapy with 110 mg twice daily was initiated.

SEE PAGE 76

ABBR EV I A T I ON S

AND ACRONYMS

DAPT = dual antiplatelet

therapy

DRT = device-related thrombus

LAA = left atrial appendage

LAAo = left atrial appendage

occlusion

OAC = oral anticoagulation

TIA = transient ischemic attack

Sedaghat et al. J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 1 , 2 0 1 7

Device-Related Thrombus With the Amulet Device J A N U A R Y 2 0 1 7 : 7 1 – 5

72

Follow-up echocardiography was performed after amean of 6 � 2 weeks and documented resolution ofthrombi in all patients. During this period, no stroke/TIA or bleeding events were observed.

DISCUSSION

In our study on patients undergoing LAAo with thenovel Amplatzer Amulet occluder, we found a highrate of DRT of 16.7% despite continued antiplatelettherapy. One patient developed DRT under clopi-dogrel monotherapy after planned cessation of DAPT3 months after LAAo. Our findings suggest a similarthrombogenicity of the novel Amulet device ascompared to its precursor, despite device modifica-tions. Following our results, occurrence of DRTappears to be related to patient characteristics as wellas implantation technique.

RISK FACTORS FOR THROMBUS FORMATION. Overall,the clinical data concerning thrombus formation ondevices used for occlusion of the LAA are scarce.Among several case reports, only Plicht et al. (6)performed a systematic follow-up of patients treatedwith the Amplatzer Plug device and reported anincidence of 17.6% of patients. In their analysis,several risk factors for thrombi were identified,including higher CHA2DS2-VASc and HAS-BLEDscores as well as reduced left ventricular function. Inour study, we found lower left ventricular ejectionfraction in patients with thrombi, whereas thrombo-embolic and bleeding risk scores were not different.In contrast to the data by Plicht et al. (6), we foundechocardiographic markers of left atrial hemostasis tobe risk factors for thrombus formation on theoccluder device, including higher degrees of sponta-neous echo contrast within the LAA and lower LAApeak emptying velocities.

Interestingly, with the Watchman device, DRT arereported in a significantly lower percentage ofpatients, approximately 5.7% (9,10). In addition todifferences in device design, the reduced incidencemay be due to the mandatory initial phase of OACafter LAAo with the Watchman device, potentiallyallowing for complete endothelialization of theoccluder (11). The debate on post-procedural therapyis further stimulated by a recent analysis on clopi-dogrel nonresponse after LAAo (10). In this study of 4patients with DRT, Ketterer et al. (10) identifiedthree-quarters of patients as clopidogrel non-responders, deeming the current concepts of DAPTinappropriate and insufficient therapy after LAAo.

DEVICE-RELATED THROMBI AFTER LAAO: AN AVOIDABLE

ISSUE? The presented data underline the unmet need

to define an optimized post-procedural anticoagulantregimen after LAAo. Given the common use of LAAo inpatients at prohibitive bleeding risk under OAC, DRTleaves both patients and physicians with a dilemma: itprovides an iatrogenic indication for therapeuticanticoagulant therapy. Although there is hardly anyscientific evidence on the relevance of DRT (9),effective resolution of these thrombi is usually aimed

TABLE 1 Characteristics of Patients With DRT

Patient #1 Patient #2 Patient #3 Patient #4

Age, yrs 80 78 69 82

Sex F M M M

LVEF, % 25 39 48 45

Indication for LAAo GI bleeding Intracranial bleeding GI bleeding GI bleeding

Creatinine, mg/dl 1.0 2.4 0.9 1.4

Atrial fibrillation Persistent Persistent Persistent Permanent

Diabetes Y N N N

Arterial hypertension Y Y Y Y

History of stroke/TIA N N Y Y

CHA2DS2VASc score 7 3 4 3

HAS-BLED score 4 3 3 5

History of LAA thrombus N Y Y Y

Device size, mm 31 25 26 22

Time after LAAo, weeks 20 16 4 4

Therapy at time ofthrombus detection

Clopidogrel DAPT DAPT DAPT

CHA2DS2VASc ¼ Congestive Heart Failure, Hypertension, Age $75 Years, Diabetes Mellitus, Previous Stroke orTransient Ischemic Attack or Thromboembolism, Vascular Disease, Age 65 to 74 Years. Sex; DAPT ¼ dual anti-platelet therapy; DRT ¼ device-related thrombus; GI ¼ gastrointestinal; HAS-BLED ¼ Hypertension, AbnormalRenal and Liver Function, Stroke, Bleeding, Labile International Normalized Ratio, Elderly, Drugs or Alcohol;LAA ¼ left atrial appendage; LAAo ¼ left atrial appendage occlusion; LVEF ¼ left ventricular ejection fraction;TIA ¼ transient ischemic attack.

TABLE 2 Comparison of Patients With and Without DRT After

Implantation of the Amplatzer Amulet Occluder

DRT(n ¼ 4)

No Thrombus(n ¼ 20)

Age, yrs 77 � 6 78 � 9

Male 66.7 76.2

CAD 75.0 41.0

Diabetes 25.0 17.6

CHA2DS2-VASc 4.3 � 1.9 4.3 � 1.5

HAS-BLED 3.8 � 1 3.6 � 0.8

Creatinine, mg/dl 1.6 � 0.7 1.5 � 0.8

Device size, mm 26.0 � 3.7 23.7 � 2.6

History of bleeding 66.7 76.2

History of LAA-thrombus 75.0 19.2

Incomplete LAA sealing 100.0 41.2

LVEF, % 39 � 10 50 � 13

SEC grade 3.0 � 0.8 1.3 � 1.1

LAApev, cm/s 17.5 � 5.0 48.3 � 21.1

LAV, ml 97 � 44 75 � 33

Values are mean � SD or %.

CAD ¼ coronary artery disease; LAApev ¼ left atrial appendage peak emptyingvelocity; LAV ¼ left atrial volume; SEC ¼ spontaneous echo contrast; otherabbreviations as in Table 1.

J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 1 , 2 0 1 7 Sedaghat et al.J A N U A R Y 2 0 1 7 : 7 1 – 5 Device-Related Thrombus With the Amulet Device

73

for in daily clinical routine (12). Due to the potentialthromboembolic risk associated with DRT, we feltobliged to initiate OAC in all patients with thrombi.The need for medical treatment was further empha-sized by the fact that 1 patient suffered a TIA. Althoughcausality between the DRT and the TIA in this patientcannot be proven retrospectively, the early presenta-tion of the patient as well as the scientific evidence onthe increased thromboembolic risk associated withLAA thrombi (13) indicate a direct relationship.

Our study indicates that more individualized anti-coagulation regimens might be warranted after LAAowith the Amulet device. Thrombi occurred in mostpatients with echocardiographic evidence/indicatorsof left atrial hemostasis and/or a history of LAAthrombus. In fact, three-quarters of patients with DRTin our study had a history of LAA thrombi, suggesting apredisposition for (local) thrombus formation andunderlining the role of left atrial hemodynamics. As aclinical consequence of our findings, we introduced acourse of direct OAC treatment of 6 weeks after LAAo inour practice, followed by aspirin monotherapy aftertransesophageal echocardiogram control examination.Additionally, our study emphasizes, that completesealing of the LAA ostium should be aimed for by

means of optimized device sizing and implantationtechnique, because all thrombi were found betweenthe device disc and the large left upper pulmonary veinridge. The observed predilection site strongly suggeststhat incomplete sealing of the LAA ostium (14) and thecreation of a cul-de-sac or “neo-appendage” may act asa nidus for thrombi and increase thrombogenicity(Figure 1H). In this context, intraprocedural 3-dimen-sional echocardiography as well as future technologicalimprovements such as steerable transseptal sheathsmay lead to optimal procedural results.

Overall, our analysis indicates a multifactorialcause of DRT with the Amulet occluder device. Giventhe limited number of patients included in this anal-ysis, large-scale randomized studies are needed tofurther address this topic and general recommenda-tions cannot be given at this point. However, whenDRT are encountered on the Amulet device, a tem-porary regimen of direct OAC appears to be safe andeffective, at least in selected patients. Until theseissues are addressed, close echocardiographic andclinical follow-up should be performed.

STUDY LIMITATIONS. The sample size in our studyis clearly an important limitation. However, the

FIGURE 1 TEE Images of DRT

Two- and 3-dimensional transesophageal echocardiography (TEE) images of device-related thrombi (DRT) in Patients #1 (A, B), #2 (C, D), #3 (E, F), and #4 (G). Typical

localization of DRT between the occluder disc and the left upper pulmonary vein (LUPV) ridge is depicted in (H). Red arrows indicate device-related thrombus formation.

AV ¼ aortic valve; LAA ¼ left atrial appendage.

Sedaghat et al. J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 1 , 2 0 1 7

Device-Related Thrombus With the Amulet Device J A N U A R Y 2 0 1 7 : 7 1 – 5

74

presented finding could have an impact on dailyclinical practice and treating physicians should beaware of this fact.

CONCLUSIONS

DRT formation after LAAo with the Amulet systemseems a frequent finding, which appears to be asso-ciated with patient-derived risk factors, implantationtechnique, and post-procedural anticoagulantregimen. Our observations and the cited data under-score the need for future prospective studies onoptimized implantation goals followed by individu-alized anticoagulant treatment strategies after LAAo.

REPRINT REQUESTS AND CORRESPONDENCE: Dr.Christoph Hammerstingl, Medizinische Klinik und Poli-klinik II–Herzzentrum, Universitätsklinikum Bonn,Sigmund-Freud-Strasse 25, 53105 Bonn, Germany. E-mail:[email protected].

RE F E RENCE S

1. Reddy VY, Möbius-Winkler S, MillerMA, et al. Leftatrial appendage closure with the Watchman devicein patients with a contraindication for oral anti-coagulation: the ASAP study (ASA Plavix FeasibilityStudyWithWatchmanLeftAtrialAppendageClosureTechnology). J Am Coll Cardiol 2013;61:2551–6.

2. Tzikas A, Shakir S, Gafoor S, et al. Left atrialappendage occlusion for stroke prevention inatrial fibrillation: multicentre experience with theAMPLATZER Cardiac Plug. EuroIntervention 2015;11:1170–9.

3. Holmes DR, Reddy VY, Turi ZG, et al., for thePROTECT AF Investigators. Percutaneous closureof the left atrial appendage versus warfarin ther-apy for prevention of stroke in patients with atrialfibrillation: a randomised non-inferiority trial.Lancet 2009;374:534–42.

4. Price MJ, Reddy VY, Valderrábano M, et al.Bleeding outcomes after left atrial appendageclosure compared with long-term warfarin: apooled, patient-level analysis of the WATCHMANrandomized trial experience. J Am Coll Cardiol Intv2015;8:1925–32.

5. Ezekowitz MD, Kent AP. Novel anticoagulantseliminate the need for left atrial appendage

exclusion devices. Circulation 2014;130:1505–14.

6. Plicht B, Konorza TF, Kahlert P, et al. Risk fac-tors for thrombus formation on the AmplatzerCardiac Plug after left atrial appendage occlusion.J Am Coll Cardiol Intv 2013;6:606–13.

7. Freixa X, Chan JLK, Tzikas A, Garceau P,Basmadjian A, Ibrahim R. The Amplatzer� CardiacPlug 2 for left atrial appendage occlusion: novelfeatures and first-in-man experience. Euro-Intervention 2013;8:1094–8.

8. Hammerstingl C, Pötzsch B, Nickenig G. Reso-lution of giant left atrial appendage thrombuswith rivaroxaban. Thromb Haemost 2013;109:583–4.

9. Main ML, Fan D, Reddy VY, et al. Assessment ofdevice-related thrombus and associated clinicaloutcomes with the WATCHMAN left atrialappendage closure device for embolic protectionin patients with atrial fibrillation (from thePROTECT-AF trial). Am J Cardiol 2016;117:1127–34.

10. Ketterer U, D’Ancona G, Siegel I, Ortak J,Ince H, Kische S. Percutaneous left atrialappendage occlusion: device thrombosis in

clopidogrel non-responders. Int J Cardiol 2016;204:196–7.

11. Schwartz RS, Holmes DR, Van Tassel RA, et al.Left atrial appendage obliteration: mechanisms ofhealing and intracardiac integration. J Am CollCardiol Intv 2010;3:870–7.

12. Qazi AH, Wimmer AP, Huber KC, Latus GG,Main ML. Resolution (and late recurrence) ofWATCHMAN device-related thrombus followingtreatment with dabigatran. Echocardiography2016;33:792–5.

13. Lowe BS, Kusunose K, Motoki H, et al. Prog-nostic significance of left atrial appendage“sludge” in patients with atrial fibrillation: a newtransesophageal echocardiographic thromboem-bolic risk factor. J Am Soc Echocardiogr 2014;27:1176–83.

14. Wunderlich NC, Beigel R, Swaans MJ, Ho SY,Siegel RJ. Percutaneous interventions for leftatrial appendage exclusion: options, assessment,and imaging using 2D and 3D echocardiography.J Am Coll Cardiol Img 2015;8:472–88.

KEY WORDS atrial fibrillation, left atrialappendage occlusion, thrombus

PERSPECTIVES

COMPETENCY IN MEDICAL KNOWLEDGE: DRT on the

Amulet device appear to be a frequent finding. In clinical prac-

tice, the occurrence of these thrombi should prompt echocar-

diographic follow-up of patients after LAAo, especially in

patients with a high risk for DRT, including those with echo

contrast within the atrium, reduced LAA emptying velocities, and

a history of LAA thrombi. When DRT are encountered, temporary

treatment with direct OAC is a safe and effective treatment

option.

TRANSLATIONAL OUTLOOK: The results of this analysis

emphasize the need for future studies on the post-procedural

anticoagulation after LAAo. It is for randomized studies to

determine the optimal duration and type of anticoagulation after

LAAo, including the use of direct OAC, with regard to safety and

efficacy.

J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 1 , 2 0 1 7 Sedaghat et al.J A N U A R Y 2 0 1 7 : 7 1 – 5 Device-Related Thrombus With the Amulet Device

75

Recurrent Post-Ablation ParoxysmalAtrial Fibrillation Shares SubstratesWith Persistent Atrial FibrillationAn 11-Center Study

Junaid A.B. Zaman, MA, BMBCH,a,b Tina Baykaner, MD, MPH,a,c Paul Clopton, MS,c Vijay Swarup, MD,d

Robert C. Kowal, MD, PHD,e James P. Daubert, MD,f John D. Day, MD,g John Hummel, MD,h Amir A. Schricker, MD,i

David E. Krummen, MD,c Moussa Mansour, MD,j Gery F. Tomassoni, MD,k Kevin R. Wheelan, MD,d

Mohan Vishwanathan, MD,a Shirley Park, MD,a Paul J. Wang, MD,a Sanjiv M. Narayan, MD,a John M. Miller, MDl

ABSTRACT

OBJECTIVES The purpose of this study was to determine the mechanistic overlap between paroxysmal and persistent

forms of atrial fibrillation (AF), focusing on AF sources as a classification approach.

BACKGROUND The role of AF substrates is unclear in patients with paroxysmal AF (PAF) that recurs after pulmonary

vein isolation (PVI). We hypothesized that patients with recurrent post-ablation (redo) PAF despite PVI have electrical

substrates marked by rotors and focal sources and structural substrates that resemble persistent AF more than patients

with (de novo) PAF at first ablation.

METHODS In 175 patients at 11 centers, we compared AF substrates in both atria using 64-pole basket catheters and phase

mapping, and indices of anatomical remodeling between patients with de novo or redo PAF and first ablation for persistent AF.

RESULTS Sources were seen in all patients. More patients with de novo PAF (78.0%) had sources near pulmonary veins

(PVs) than patients with redo PAF (47.4%; p ¼ 0.005) or persistent AF (46.9%; p ¼ 0.001). The total number of

sources per patient (p ¼ 0.444), and number of non-PV sources (p ¼ 0.701) were similar between groups, indicating that

redo PAF patients had residual non-PV sources after elimination of PV sources by prior PVI. Structurally, left atrial size did

not separate de novo from redo PAF (49.5 � 9.5 mm vs. 49.0 � 7.1 mm; p ¼ 0.956) but was larger in patients with

persistent AF (55.2 � 8.4 mm; p ¼ 0.001).

CONCLUSIONS Patients with PAF despite prior PVI show electrical substrates that resemble persistent AF more closely

than patients with PAF at first ablation. Notably, these subgroups of PAF are indistinguishable by structural indices.

These data motivate studies of trigger versus substrate mechanisms for patients with recurrent PAF after PVI.

(J Am Coll Cardiol EP 2017;3:393–402) © 2017 by the American College of Cardiology Foundation.

From the aStanford University, Stanford, California; bImperial College, London, United Kingdom; cSan Diego VA Medical Center,

San Diego, California; dArizona Heart Rhythm Center, Phoenix, Arizona; eBaylor University Medical Center, Dallas, Texas; fDuke

University Medical Center, Durham, North Carolina; gIntermountain Medical Center, Salt Lake City, Utah; hOhio State University,

Columbus, Ohio; iUniversity of California, San Diego, San Diego, California; jMassachusetts General Hospital, Boston, Massa-

chusetts; kCentral Baptist Hospital, Lexington, Kentucky; and the lIndiana University School of Medicine, Indianapolis, Indiana.

Dr. Zaman is the recipient of a Fulbright British Heart Foundation Scholarship 2015–2016 (68150918) and British Heart Foundation

Travel Grant 2014–2015 (FS/14/46/30907). Dr. Baykaner is the recipient of the Josephson and Wellens Heart Rhythm Society

Fellowship 2015–2016. Dr. Swarup received consulting fees and honoraria from Biosense Webster, Inc; and research grants from

Medtronic, Inc., Boston Scientific Corp., St. Jude Medical, and Biotronik. Dr. Kowal has received consulting fees and honoraria

from Medtronic, Inc. Dr. Daubert has received consulting fees and honoraria from ARCA Biopharma, Biosense Webster, Inc.,

Medtronic, Inc., Biotronik, St. Jude Medical, Boston Scientific Corp., Sorin Group, CardioFocus, Inc., Gilead Sciences Inc.,

Northwestern University, Orexigen Pharmaceuticals, VytronUS, Heart Metabolics, and Zoll; has received support from Biosense

Webster, Inc., Boston Scientific, Medtronic Inc., Gilead Sciences Inc., St. Jude Medical, Biotronik, and Bard Electrophysiology; and

has equity interest and stock options from Biosense Webster. Dr. Hummel has received consulting fees/honoraria from Biosense

Webster, Inc. Dr. Schricker was the recipient of an ACC/Merck fellowship in 2012–2013; and has received consulting fees/honoraria

J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y VO L . 3 , N O . 4 , 2 0 1 7



I S S N 2 4 0 5 - 5 0 0 X / $ 3 6 . 0 0


P ulmonary vein isolation (PVI) is cen-tral to the ablation of paroxysmalatrial fibrillation (PAF) and persistent

atrial fibrillation (AF), yet its results remainsuboptimal even in recent clinical trials(1–4). An increasingly recognized fact is thatPAF patients may do well after PVI evenwhen the pulmonary veins (PVs) have recon-nected (5,6), suggesting that PVI lesion setsinterrupt other mechanisms. Indeed, studiessuggest that PAF is a heterogeneous popula-tion that may overlap with persistent AF (7),which may have substrate mechanismsremote from the PVs. However, these mech-anisms are yet unidentified in PAF.

We hypothesized that patients withrecurrent PAF after prior PVI are more likelyto have substrates remote from the PVs, andmore closely resemble patients with persis-tent AF than patients with PAF at their first

PVI procedure. This may follow for several reasons.First, patients with recurrent PAF may have peri-PVmechanisms not eliminated at initial ablation. Sec-ond, it may be artificial to “dichotomize” populationswith PAF and persistent AF given their overlap in trueAF burden (7), left atrial (LA) size, and atrial struc-tural abnormalities (8). Third, recent mapping of PAFshows substrates in the form of rotors and focalsources remote from the PVs where targeted ablationcan eliminate AF acutely and long-term in manycenters (9–12) with different techniques (13). More-over, AF sources have now been demonstrated in

human optical mapping studies (14) with many simi-larities to these clinical studies.

We tested our hypothesis by examining electricalsubstrates of rotors or focal sources, and structuralsubstrates by echocardiography, in patients with PAFat first ablation, recurrent PAF despite prior PVI, andfirst-time persistent AF ablation in an 11-center pro-spective observational study.

METHODS

ENROLLMENT AT CONTRIBUTING CENTERS. Between2012 and 2014, we enrolled 175 patients undergoingAF ablation for routine indications at 11 centers in theUnited States (Table 1). All studies and data analyseswere performed with local institutional review boardapproval and patients provided written consent fordata collection.

PATIENT CLASSIFICATION. We classified the 175patients prospectively into de novo PAF, redo PAF,

TABLE 1 Contributing Centers

Stanford Medical Center, Palo Alto, California

Arizona Heart Rhythm Center, Phoenix, Arizona

Baylor University Medical Center, Dallas, Texas

Duke University Medical Center, Durham, North Carolina

Intermountain Medical Center, Salt Lake City, Utah

Ohio State University Wexner Medical Center, Columbus, Ohio

University of California San Diego Medical Center, San Diego,California

San Diego Veterans Affairs Medical Center, California

Massachusetts General Hospital, Boston, Massachusetts

Central Baptist Hospital, Lexington, Kentucky

Indiana University Health University Hospital, Indianapolis, Indiana

SEE PAGE 403


AND ACRONYMS

AF = atrial fibrillation

ANOVA = analysis of variance

AT = atrial tachycardia

CHADSVASc = congestive

heart failure, hypertension,

age, diabetes, stroke, vascular

risk factors, age, sex

CMR = cardiac magnetic

resonance

FIRM = focal impulse and

rotor mapping

LA = left atrium

PAF = paroxysmal atrial

fibrillation

PV = pulmonary vein

PVI = pulmonary vein isolation

RA = right atrium

Redo PAF = recurrent post-

ablation paroxysmal atrial

fibrillation

from Abbott Electrophysiology. Dr. Krummen has received fellowship support from Boston Scientific, Biotronik, Biosense

Webster, Inc., Medtronic, and St. Jude Medical; and has received consulting fees/honoraria from Topera Medical and Pacific

Blue Innovations. Dr. Monsour has received consulting fees/honoraria from Biosense Webster, Inc., St. Jude Medical, Sen-

treheart, and Medtronic; and has received grants from Biosense Webster, Inc., St. Jude Medical, and Boston Scientific Corp. Dr.

Tomassoni has received consulting fees/honoraria from Stereotaxis Inc., Topera Medical, and St. Jude Medical; has been a

speaker for Abbott/Topera, St. Jude Medical, Biosense Webster, Inc., and Biotronix; and has been an advisor for Abbott/Topera,

St. Jude Medical, Biosense Webster Inc., BXS, Medtronic, and Biotronix. Dr. Wheelan has received consulting fees/honoraria

from Medtronic, Inc.; has equity interest/stock options in Medtronic, Inc.; and has received support from St. Jude Medical,

Boston Scientific Corp., and Medtronic, Inc. Dr. Viswanathan has received consulting fees/honoraria from Biosense Webster,

Inc. Dr. Park has received consulting fees/honoraria from Medtronic, Inc. Dr. Wang has received consulting fees/honoraria from

Medtronic, Inc.; and has received grant and fellowship support from Medtronic, Inc. Dr. Narayan has received funding from the

National Institutes of Health (R01 HL83359; R01 HL 122384; K24HL103800); consulting fees/honoraria fromMedtronic Inc., St. Jude

Medical, Biotronik, Boston Scientific Corp.; has ownership, equity interest, and stock optionswith ToperaMedical; and has received

modest consulting fees from Abbott and University of California Regents. Dr. Miller has received consulting fees/honoraria from

Topera Medical, Medtronic Inc., Boston Scientific Corp., St. Jude Medical, and Biosense Webster, Inc.; has received support from

Medtronic Inc., Boston Scientific Corp., St. JudeMedical, Biotronik, and BiosenseWebster, Inc.; and is an advisor for ToperaMedical

and BiosenseWebster, Inc. All other authors have reported that they have no relationships relevant to the contents of this paper to

disclose. Francis Marchlinski, MD, served as Guest Editor for this paper.

All authors attest they are in compliancewith human studies committees and animalwelfare 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.

Manuscript received May 16, 2016; revised manuscript received September 28, 2016, accepted October 3, 2016.

Zaman et al. J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 4 , 2 0 1 7

Post-Ablation PAF A P R I L 2 0 1 7 : 3 9 3 – 4 0 2

394

and persistent AF groups. PAF was defined as AF thatterminates spontaneously or with intervention within7 days of onset, whereas persistent AF was defined ascontinuous AF that is sustained >7 days (15). The denovo PAF group consisted of patients presenting fortheir first catheter ablation of AF (n ¼ 48). RedoPAF patients had recurrent PAF despite 1 prior PVI(n ¼ 31) and persistent AF patients were presentingfor their first ablation for persistent AF (n ¼ 96).We excluded those presenting for repeat persistentAF ablation.

ELECTROPHYSIOLOGICAL STUDY. Patients discontinuedantiarrhythmic medications for >5 half-lives exceptamiodarone, which was stopped as early as possibleprior to the procedure. During the procedure, heparinwas used to maintain activated clotting time >350 s. A64-pole basket catheter was advanced to the rightatrium (RA), then transseptally to the LA (Figure 1A),with attention paid to basket location to prevent errorssuch as inadvertent ventricular placement as notedin some recent studies (16). Our practice is to move thebasket during successive focal impulse and rotormapping (FIRM)map epochs to ensure basket coverageof the majority of the atrial surface (Figure 1B). Elec-trodes were referenced to electroanatomic shells(NavX, St. Jude Medical, St Paul, Minnesota, Minne-sota, or Carto, Biosense-Webster, Inc., Diamond Bar,California). In both NavX and Carto shells, distanceswere estimated based on the fixed distance betweenelectrodes along basket splines, and cartographicrulers on the atrial shell when available.

FIRM MAPPING OF AF SUBSTRATE. FIRM mapsidentify electrical rotors as phase singularities withsurrounding disorganization. Rotors were consideredAF sources if stable within <1 electrode for multiplerecording epochs over 2 to 5 min. Focal AF sourceswere defined as origins with centrifugal activationand breakdown into meandering wavelets.

FIRM mapping uses algorithms to map propagationsequences from observed AF activations. For unipolardeflections that are noncomplex, mapping can bestraightforward. Figure 2A depicts activation mappingfrom a basket catheter where points of maximumnegative dV/dt (green line) indicate each AF activationcycle. These activations identify a counterclockwiserotational circuit in the inferior LA (shown in iso-chrones in Figure 2B and on electroanatomic map inFigure 2D) where ablation terminated AF to atrialtachycardia (AT) (Figure 2C). Conversely, in caseswhere multiple AF deflections are seen for any cycle,classical rules often misidentify signals within repo-larization which are by definition far-field (17). Insuch cases, FIRM determines local activation by

analyzing variations in signal morphology betweeneach basket electrode and neighboring channels.A sawtooth-shaped wave of normalized amplitude isobtained from action potential duration (18), conduc-tion restitution data (19), and rate adapted to estimatephase to identify rotors. If no signal is detected, thatregion of the atria remains black on FIRMmaps. FIRM-identified rotors show many similarities to micro–re-entrant AF drivers found in optical mapping of AF inhuman atria (14), with early data showing concordancebetween FIRM-mapped and optically mapped AFsources in human hearts (20).

CHARACTERIZING SUBSTRATE BY FUNCTIONAL

MAPPING. We described functional substrate in eachpatient prospectively using the following criteria:source number, the total number of sources reportedby FIRM mapping, whether or not ablated; and sourcelocation, which was further subcategorized into PVversus non-PV location (those within 1 cm of a PVwhere considered PV, all others being remote fromthe PVs) and RA versus LA. These were identifiedprospectively at each case. Reproducibility of thisassignment by individual operators has recently beenreported to be good with kappa ¼ 0.89 (21). Thesedata were collected prospectively by each investi-gator during each case, then collated and retrospec-tively blinded to patient group.

CHARACTERIZING THE ANATOMICAL SUBSTRATE. Wemeasured LA size on 2-dimensional echo, a well-validated index of LA structural remodeling, in allpatients pre-ablation. Additional imaging tests suchas late-gadolinium enhanced cardiac magnetic reso-nance (CMR) were not in widespread use at the timeof the study.

FIRM-GUIDED ABLATION. Ablation commenced withFIRM-guided radiofrequency ablation for AF, using3.5-mm (Biosense-Webster, Inc.) or 4-mm (St. JudeMedical) tipped irrigated catheters or a non-irrigatedcatheter (Boston Scientific, Inc., Marlborough, Massa-chusetts) in some patients with heart failure. Ablationwas delivered to the organized domain of rotors (areasof 2 cm2 to 3 cm2, similar to AF driver areas in humanoptical mapping) (14). Ablation typically commencedin the RA then proceeded to the LA sites and wasrepeated to eliminate rotors on remapping (Figure 1A).Rotors and focal sources were not ablated if near sen-sitive structures such as the phrenic nerve.

PULMONARY VEIN ISOLATION. Ablation was per-formed to isolate left and right PVs in pairs, with veri-fication of PV entrance block using a circular mappingcatheter. Ablation was avoided near sensitive struc-tures such as the esophagus or phrenic nerve. Theablation protocol was FIRM first, until no more

J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 4 , 2 0 1 7 Zaman et al.A P R I L 2 0 1 7 : 3 9 3 – 4 0 2 Post-Ablation PAF

395

sources, then PVI. If AF sources (identified prospec-tively) fell within the operator’s planned PVI lesion set,this sequence could be changed. At redo-ablation, ifveins were reconnected they were re-isolated.

STATISTICAL ANALYSIS. Continuous variables with anormal distribution are presented as mean� SD unlessotherwise noted and evaluated with 1-way analysis ofvariance (ANOVA), Bonferroni and Tukey post hoc

FIGURE 1 Technique for FIRM-Guided Ablation

(A) Typical workflow for FIRMmapping, with basket placement by fluoroscopic images. Sequential atrial basket recordings were taken with source

elimination prior to PVI. (B)Multiple basket positions used per FIRM-mapping epoch to minimize unmapped atrial regions. Positions 1-2 are in the RA;

positions 3-4 are in the LA. AF¼ atrial fibrillation; AT¼ atrial tachycardia; FIRM¼ focal impulse and rotor mapping; LA¼ left arterial; LIPV¼ left inferior

pulmonary vein; LSPV¼ left superior pulmonary vein; PV¼ pulmonary vein; PVI¼ pulmonary vein isolation; RA ¼ right atrium; RSPV¼ right superior

pulmonary vein; SVC¼ superior vena cava.



396

tests where indicated. Counts of sources and conges-tive heart failure, hypertension, age, diabetes, stroke,vascular risk factors, age, sex (CHADS2VASc) scores arereported with medians and quartiles and comparedamong groups with Kruskal-Wallis tests. Nominalvariables are reported as counts and percentages andevaluated with chi-square tests. Logistic regressionwas used to evaluate the group difference in the pres-ence of PV sources controlling for site differences.Probabilities below 0.05 were considered significant.

RESULTS

A total of 175 patients were included in the presentstudy; Table 2 summarizes patient characteristics.

NUMBER OF SOURCES. There were 144 total FIRMidentified rotors/sources in the de novo PAF groupcompared to 129 in the redo PAF and 304 in thepersistent AF group. This corresponded to roughlyequal numbers of sources per patient of 3 (range 2 to5) in the de novo PAF group, 3 (range 2 to 4) in theredo PAF group, and 3 (range 2 to 4) in the persistent

group (p ¼ 0.444). Overall, 91% were rotors and 9%focal sources.

SOURCE LOCATIONS. Only 54.3% of patients hadany sources within 1 cm of PVs. The presence of anyPV-localized sources differed by group (p ¼ 0.002).More patients with PAF at first ablation (78.0%)showed 1 or more PV sources than patients with redoPAF (47.4%; p ¼ 0.005) and persistent AF (46.9%;p ¼ 0.001). Group difference in the presence of any PVsources remained significant (p ¼ 0.008) in a logisticregression model controlling for study-site differ-ences. When analyzed as a count, the number of PV-localized sources per patient showed a similarpattern of differences (p ¼ 0.002). By contrast, therewas no difference among the groups in the number ofnon-PV sources (p ¼ 0.701).

The distributions of PV and non-PV sources bygroup are illustrated in Figure 3. In patients with redoPAF, sources near PVs lay near gaps in prior PVisolation lesion sets or just outside prior lesion sets,such that electrograms were detected to yield sources(e.g., see case in Figure 2). Figure 4 illustrates rotor

FIGURE 2 Rotational Activity at FIRM-Mapped Rotor, Where Ablation Terminated AF to Atrial Tachycardia

In this 61-year-old man with persistent AF, unipolar AF electrograms confirm rotation around rotor core. (A) Electrograms (unipoles) and dV/dt (first derivative, green)

shown, with 1 cycle annotated (red lines). (B) LA shell with activation map from the annotated cycle, demonstrating earliest-latest interaction in a rotational pattern.

(C) Termination panel shows abrupt termination of AF to organized AT with ablation at this site, prior to any PVI. Black bar represents 1,000 ms. (D) Location on LA

posterior wall of termination site (red) during ablation of rotor area (white). ECG ¼ electrocardiogram; other abbreviations as in Figure 1.


397

sites near and remote from PVs with direct termina-tion to sinus rhythm during ablation.

There was a nonsignificant trend towards a differ-ence among the groups in the number of LA sources(p ¼ 0.070) with medians of 3 (range 2 to 3) for denovo PAF and 2 (range 1 to 3) for both redo PAF andpersistent AF (Figure 5).

STRUCTURAL REMODELING. There was a significantdifference among groups in LA size (p < 0.001). LAsize did not separate de novo from redo PAF(49.5 � 9.5 mm vs. 49.0 � 7.1 mm, respectively; p ¼0.956) but was significantly larger in patients with

persistent AF (55.2 � 8.4 mm; p ¼ 0.001 vs. each using3-group ANOVA with post hoc pairwise comparisons).Thus, despite the electrical similarities between redoPAF and persistent AF, they were structurally distinctusing traditional indices (Figure 6). The CHADS2VAScscore did not differentiate among the 3 groups (p ¼0.681) with a median of 1 (range 0 to 2) in each group.Likewise, left ventricular ejection fraction did notdifferentiate among the 3 groups (p ¼ 0.777, using3-group ANOVA with post hoc pairwise comparisons)(Table 2).

ACUTE IMPACT OF ABLATION. The acute impact ofablation is summarized in Table 3. Acute terminationwas seen in 83 patients, of which 41% were to AT and59% to sinus rhythm directly. Figure 2 shows anexample of a phase-identified AF rotor, in which goodunipolar signal quality enabled confirmation of FIRM-mapped rotors by showing rotational activation usingtraditional analysis of minimum dV/dt (green line), ina patient with persistent AF. Targeted ablation of thisrotor in the inferior LA (Figure 2D) terminated AF toan AT prior to PVI. Figure 4 shows another patientwith abrupt termination to sinus rhythm duringablation of a rotor adjacent to the PVs.

FIGURE 3 Higher Number of PV Sources in De Novo PAF

(Left) The bar chart shows a higher percentage of patients with PV rotors/sources in de novo PAF compared to redo PAF and persistent AF

(p ¼ 0.002). Differences in the number of non-PV sources (right) were not significant (p ¼ 0.701). PAF ¼ paroxysmal atrial fibrillation;

PV ¼ pulmonary vein.

TABLE 2 Population Demographics

De Novo Paroxysmal AF(n ¼ 41)

Redo Paroxysmal AF(n ¼ 38)

Persistent AF(n ¼ 96)

Age (yrs) 58.8 � 12.5* 61.3 � 9.5 63.5 � 10.1

LA diameter (mm) 49.5 � 9.5 49.0 � 7.1 56.1 � 8.0†

LVEF (%) 57.6 � 9.7 56.9 � 7.9 56.4 � 10.3

CHADS2VASc 1 (0–2) 1 (0–2) 1 (0–2)

Values are mean � SD or n (range). *p < 0.05 vs. persistent AF. †p ¼ 0.001 using 3-groupanalysis of variance with post hoc pairwise comparisons.

AF ¼ arterial fibrillation; CHADS2VASc ¼ congestive heart failure, hypertension, age, diabetes,stroke, vascular risk factors, age, sex score; LA ¼ left arterial; LVEF ¼ left ventricular ejectionfraction.



398

FIGURE 4 Ablation of AF Sources Adjacent and Remote to PVs Causing Acute Termination to Sinus Rhythm

(A) Antero-posterior bi-atrial NavX map with RA, non-PV, and PV LA sources labelled in a patient at redo ablation for PAF. (B) AF signals on

ablation catheter at FIRM-mapped rotor adjacent to prior left superior PVI lesion set, where AF was terminated by FIRM-guided ablation.

Black bar represents 1,000 ms. ABL ¼ ablation catheter; other abbreviations as in Figures 1 and 2.

FIGURE 5 Nonsignificant Differences in RA and LA Sources

There was a trend toward a difference among the groups in the number of LA sources (p ¼ 0.070) which was somewhat higher in de novo PAF

than in the other groups. There was no group difference in the number of RA sources (p ¼ 0.541). Abbreviations as in Figures 1 and 3.


399

DISCUSSION

This study uses mapping of functional AF substratesin patients at 11 U.S. centers to show that patientswith PAF despite prior PVI (redo PAF) are moresimilar electrophysiologically to patients withpersistent AF than those with de novo PAF at firstablation. Compared to patients with de novo PAF,those with redo PAF had rotor distributions awayfrom the PVs in the LA and in RA, more like persistentAF. Notably, this electrophysiological difference wasnot reflected in structural remodeling because pa-tients with de novo and redo PAF had similar LA di-mensions that were each lower than in patients withpersistent AF. These data suggest the possibility ofidentifying a priori the 35% to 50% of patients withPAF who may not be arrhythmia-free after a singlePVI. Mechanistically, these data motivate studies toexamine the role of trigger versus substrate ablationin patients at repeat ablation procedures.

ROLE OF ROTORS AND FOCAL SOURCES IN HUMAN

AF. Evidence continues to mount that human AF issustained by localized rotors and focal sources. Thecharacteristics of rotors on FIRM mapping are similarto those from optical mapping in human atria in thatthey show stable endocardial rotors in w2-cm2 areaswhere ablation can acutely terminate AF, yet areunstable and transient on the epicardium (14). Thesedata may help reconcile differences between endo-cardial FIRM mapping and less stable rotors onepicardial electrocardiographic imaging (13). Earlydata show concordance between concurrent FIRMand optically mapped AF in human hearts (20).

FIRM-guided ablation (FIRM plus PVI) has beenreported in many patients with many results in largeseries consistent with the original CONFIRM (Con-ventional Ablation for Atrial Fibrillation With orWithout Focal Impulse and Rotor Modulation) trial(10–12,22). However, not all studies support the rotormechanism. Some studies suggest that rotors do notexist, but typically used empirical rules with tech-nical errors such as reported cycle lengths of 250 msto 500 ms in AF (rates 2 to 4 Hz) that are lessconsistent with AF, and the use of Shannon entropyto unipolar signals although it is likely valid only forbipolar signals (16). The recent OASIS (Outcome ofDifferent Ablation Strategies in Persistent andLong Standing Persistent Atrial Fibrillation) trialinitially reported FIRM þ PVI success of 52% in40 patients (23) but has subsequently been retracteddue to nonrandomization issues which may impactany comparisons between groups. Even despite this,those authors’ recently reported 20% 1-year successfrom PV antrum isolation alone in similar persistentAF patients (24) suggests that FIRM substantiallyimproves the results of PV antrum isolation atthat center. Direct comparisons of FIRM þ PVI toPVI alone (akin to the STAR-AF2 [Substrate andTrigger Ablation for Reduction of Atrial Fibrillation]trial) are warranted and ongoing. Interestingly, otherless impressive studies show a substantial acutetermination rate of persistent AF to AT (>30%) withFIRM-guided ablation alone (25), supporting themechanistic presence of sources. Large multicenterrandomized trials are ongoing to test thesequestions.

SHARED AND DISCORDANT MECHANISMS IN PAF

AND PERSISTENT AF. This study defines potentialmechanisms that may separate PAF patients with andwithout AF recurrence after index PVI that mayoutline a spectrum for AF phenotypes, from de novoPAF to redo PAF then to persistent AF. Of particularnote is the electrical but not anatomical separation of

TABLE 3 Acute Impact of Ablation

Total Cases WithTerminations During Ablation

CasesTerminating

to SR

CasesTerminating

to AT

De novo PAF (n ¼ 41) 27 (65%) 18 9

Redo PAF (n ¼ 38) 24 (64%) 15 9

Persistent AF (n ¼ 96) 32 (33%) 16 16

Values are n (%).

AT ¼ atrial tachycardia; PAF ¼ paroxysmal atrial fibrillation; SR ¼ sinus rhythm.

FIGURE 6 LA Size Did Not Separate De Novo From Redo PAF

LA size was higher only in the persistent AF group, despite the electrophysiological

differences in this study. Mean � SEM. Abbreviations as in Figures 1 and 3.



400

de novo from redo PAF, which shared AF source dis-tributions and characteristics with persistent AF.

Despite effective antral ablation at index ablationof PAF, up to 70% of patients may have reconnectionat repeat electrophysiological study 3 monthslater (6), which could be due to gaps in radio-frequency ablation lines, or incomplete circumferen-tial contact with balloon-based technologies. WhilePV reconnection may occur in patients withor without clinical AF recurrence, few studieshave examined if substrates may differ between suchgroups. Other candidate differences in substrate be-tween groups include fibrosis and CMR-identifiedscar that are the subject of intense investigation.

STRUCTURE-FUNCTION DISSOCIATION IN AF

PROGRESSION. Structurally, there was no differencein LA size between de novo and redo PAF despitedifferences in atrial source numbers and atrial dis-tributions. In contrast, persistent AF showed asignificantly larger LA size. This structure-functiondissociation is supported by recent data with asimilar mix of AF phenotypes (26) and lone PAF pa-tients compared to controls (27). Atrial CMR imagingusing delayed enhancement CMR also reveals a widespectrum of fibrotic substrates that do not conform tocurrent classification schemes based on arbitrary timecut offs. It is hoped such delineation of functionalsubstrates will allow for patient-tailored risk stratifi-cation to improve outcomes (28).

FUTURE DIRECTIONS. These data contribute to thediscussion on how to improve the results from PVantral isolation. This is timely. Recent trials showthat empiric linear ablation or ablation of complexfractionated electrograms may not improve the re-sults of PV isolation in persistent AF patients(1,29,30). Success in PAF is limited to w65% at 1 yearand w50% at 2 years with PVI by cryoballoon orradiofrequency energy using force-sensing catheters(1 or more procedures permitted in blanking period)(31). Although more durable PVI may furtherenhance outcomes, the present data demonstratethe presence of AF substrates that are often remotefrom the PVs or empirical line sites—such as in theRA—that could be targeted for ablation. In a retro-spective analysis of the CONFIRM trial, the presenceof rotors that were not directly or inadvertentlytargeted by ablation portended a worse arrhythmia-free prognosis (32).

STUDY LIMITATIONS. Ideally, it would have been usefulto FIRMmap all patients with PAF atfirst ablation, onlyperform PVI, and then assess what substrates werepresent on a repeat ablation. Such a trial is planned.

More generally, the current study did not prospec-tively randomize substrates to ablation or nonablation;however, such a study is also underway. Echocardio-graphic indices of structure are increasingly beingreplaced by CMR, which is more widely available now,and neither LA volume data nor computerized tomo-graphic data were universally available from all sites.Echocardiographic data were reported from clinicalrecords at each site, not from a core lab. Further elec-trophysiological characterization with voltage or frac-tionation indices would be attenuated by priorablation in redo PAF, and was not performed.

CONCLUSIONS

Across 11 U.S. centers, we found that PAF patients atrepeat ablation are electrically more similar to pa-tients with persistent AF with numerous extra-PVsources than to PAF patients at their first ablation,despite similar LA size. Our findings motivate studiesto further define “functional substrates” and moti-vate trials to test the benefit of substrate ablationover repeat PVI in patients at repeat procedures.These findings may also help improve knowledge ofAF progression and improve clinical outcomes.

ACKNOWLEDGMENTS The authors thank Dr. VivekReddy and Mount Sinai School of Medicine, NewYork, New York, for contributing to this study.

ADDRESS FOR CORRESPONDENCE: Dr. Sanjiv M.Narayan, Department of Cardiology, Stanford Uni-versity, Falk CV275, 300 Pasteur Drive, Palo Alto,California 94305. E-mail: [email protected].

PERSPECTIVES

COMPETENCY IN MEDICAL KNOWLEDGE: This study ad-

vances the increasing literature highlighting heterogeneity

within patients with PAF, suggesting that those with recurrent

AF may overlap with persistent AF patients.

TRANSLATIONAL OUTLOOK: The scientific field continues to

find growing evidence of organization within fibrillating atria,

with a hierarchy of spatial sites identified by the methods in this

study and corroborated by others that may guide ablation to

improve outcomes in persistent AF. By showing similarities be-

tween patients with PAF who recurred despite prior PVI and

those with persistent AF, these data start to define a functional

spectrum of AF that may improve the current binary classification

of AF based on detected duration. We believe this will allow for

more tailored, patient specific therapies to help understand and

treat this disease.


401

RE F E RENCE S

1. Verma A, Jiang C, Betts TR, et al. Approaches tocatheter ablation for persistent atrial fibrillation.N Engl J Med 2015;372:1812–22.

2. Natale A, Reddy VY, Monir G, et al. ParoxysmalAF catheter ablation with a contact force sensingcatheter: results of the prospective, multicenterSMART-AF trial. J Am Coll Cardiol 2014;64:647–56.

3. Calkins H. Demonstrating the value of contactforce sensing more difficult than meets the eye.Circulation 2015;132:901–3.

4. Dukkipati SR, Cuoco F, Kutinsky I, et al. Pul-monary vein isolation using the visually guidedlaser balloon. J Am Coll Cardiol 2015;66:1350–60.

5. Jiang R-H, Po SS, Tung R, et al. Incidence ofpulmonary vein conduction recovery in patientswithout clinical recurrence after ablation ofparoxysmal atrial fibrillation: mechanistic impli-cations. Heart Rhythm 2014;11:969–76.

6. Kuck K-H, Hoffmann BA, Ernst S, et al. Impactof complete versus incomplete circumferentiallines around the pulmonary veins during catheterablation of paroxysmal atrial fibrillation. CircArrhythmia Electrophysiol 2016;9:e003337.

7. Charitos EI, Pürerfellner H, Glotzer TV,Ziegler PD. Clinical classifications of atrial fibrilla-tion poorly reflect its temporal persistence: in-sights from 1,195 patients continuously monitoredwith implantable devices. J Am Coll Cardiol 2014;63:2840–8.

8. Oakes RS, Badger TJ, Kholmovski EG, et al.Detection and quantification of left atrial struc-tural remodeling with delayed-enhancementmagnetic resonance imaging in patients withatrial fibrillation. Circulation 2009;119:1758–67.

9. Narayan SM, Krummen DE, Shivkumar K,Clopton P, Rappel W-J, Miller JM. Treatment ofatrial fibrillation by the ablation of localizedsources. J Am Coll Cardiol 2012;60:628–36.

10. Tomassoni G, Duggal S, Muir M, et al. Long-term follow-up of FIRM-guided ablation of atrialfibrillation: a single-center experience. J InnovCard Rhythm Manag 2015;6:2145–51.

11. Rashid H, Sweeney A. Approaches for focalimpulse and rotor mapping in complex patients : aUS private practice perspective. J Innov CardRhythm Manag 2015;6:2193–8.

12. Miller JM, Kowal RC, Swarup V, et al. Initialindependent outcomes from focal impulse androtor modulation ablation for atrial fibrillation:Multicenter FIRM Registry. J Cardiovasc Electro-physiol 2014;25:921–9.

13. Haissaguerre M, Hocini M, Denis A, et al. Driverdomains in persistent atrial fibrillation. Circulation2014;130:530–8.

14. Hansen BJ, Zhao J, Csepe TA, et al. Atrialfibrillation driven by micro-anatomic intramuralre-entry revealed by simultaneous sub-epicardialand sub-endocardial optical mapping in explan-ted human hearts. Eur Heart J 2015;36:2390–401.

15. January CT, Wann LS, Alpert JS, et al. 2014AHA/ACC/HRS Guideline for the Management ofPatients With Atrial Fibrillation: Executive Sum-mary: A Report of the American College of Cardi-ology/American Heart Association Task Force onPractice Guidelines and the Heart Rhythm Society.J Am Coll Cardiol 2014;64:2246–80.

16. Jalife J, Filgueiras-Rama D, Berenfeld O. Letterby Jalife et al regarding article, “QuantitativeAnalysis of Localized Sources Identified by FocalImpulse and Rotor Modulation Mapping in AtrialFibrillation” nothing. Circ Arrhythmia Electro-physiol 2015;8:1296–8.

17. Narayan SM, Wright M, Derval N, et al. Clas-sifying fractionated electrograms in human atrialfibrillation using monophasic action potentials andactivation mapping: evidence for localized drivers,rate acceleration, and nonlocal signal etiologies.Heart Rhythm 2011;8:244–53.

18. Narayan SM, Krummen DE, Enyeart MW,Rappel W-J. Computational mapping identifieslocalized mechanisms for ablation of atrial fibril-lation. PLoS One 2012;7:e46034.

19. Narayan SM, Franz MR, Clopton P, Pruvot EJ,Krummen DE. Repolarization alternans revealsvulnerability to human atrial fibrillation. Circula-tion 2011;123:2922–30.

20. Hansen BJ, Briggs C, Moore BT, et al. Abstract18402: human atrial fibrillation drivers seensimultaneously by focal impulse and rotor map-ping and high-resolution optical mapping. Circu-lation 2015;132 Suppl 3:A18402.

21. Lalani GG, Coysh T, Baykaner T, et al. Orga-nized sources are spatially conserved in recurrentcompared to pre-ablation atrial fibrillation: furtherevidence for non-random electrical substrates.J Cardiovasc Electrophysiol 2016;27:661–9.

22. Sommer P, Kircher S, Rolf S, John S, Arya A.Successful repeat catheter ablation of recurrentlongstanding persistent atrial fibrillation with rotorelimination as the procedural endpoint : a case seriesshort title : rotor elimination as endpoint for AF ab-lations. J Cardiovasc Electrophysiol 2015;27:274–80.

23. Natale A, Mohanty S, Gianni C, et al. LBCT02-01 / LBCT02-01 impact of rotor ablation in

non-paroxysmal AF patients: results from a ran-domized trial (OASIS). Heart Rhythm 2016;13:1373–6.

24. Bai R, DiBiase L, Mohanty P, et al. Provenisolation of the pulmonary vein antrum with orwithout left atrial posterior wall isolation in pa-tients with persistent atrial fibrillation. HearRhythm 2016;13:132–40.

25. Gianni C, Mohanty S, Di Biase L, et al. Acuteand early outcomes of FIRM-guided rotors-onlyablation in patients with non-paroxysmal atrialfibrillation. Hear Rhythm 2016;13:830–5.

26. Maury P, Thomson E, Rollin A, et al. Lack ofcorrelations between electrophysiological andanatomical-mechanical atrial remodeling in pa-tients with atrial fibrillation. Pacing Clin Electro-physiol 2015;38:617–24.

27. Stiles MK, John B, Wong CX, et al. Paroxysmallone atrial fibrillation is associated with anabnormal atrial substrate: characterizing the“second factor”. J Am Coll Cardiol 2009;53:1182–91.

28. Marrouche NF, Wilber D, Hindricks G, et al.Association of atrial tissue fibrosis identified bydelayed enhancement MRI and atrial fibrillationcatheter ablation: the DECAAF study. JAMA 2014;311:498–506.

29. Wong KCK, Paisey JR, Sopher M, et al. Nobenefit of complex fractionated atrial electrogram(CFAE) ablation in addition to circumferentialpulmonary vein ablation and linear ablation :BOCA study. Circ Arrhythmia Electrophysiol 2015;8:1316–24.

30. Vogler J, Willems S, Sultan A, et al. Pulmonaryvein isolation versus defragmentation: the CHASE-AF clinical trial. J Am Coll Cardiol 2015;66:2743–52.

31. Kuck K-H, Brugada J, Fürnkranz A, et al. Cry-oballoon or radiofrequency ablation for parox-ysmal atrial fibrillation. N Engl J Med 2016;374:2235–45.

32. Narayan SM, Krummen DE, Clopton P,Shivkumar K, Miller JM. Direct or coincidentalelimination of stable rotors or focal sources mayexplain successful atrial fibrillation ablation:on-treatment analysis of the CONFIRM (CONven-tional ablation for AF with or without FocalImpulse and Rotor Modulation) Trial. J Am CollCardiol 2013;62:138–47.

KEY WORDS ablation, atrial fibrillation,sources



402

High Remission Rates inVasovagal SyncopeSystematic Review and Meta-Analysis ofObservational and Randomized Studies

Payam Pournazari, MD, MSC, Inderjeet Sahota, MD, MSC, Robert Sheldon, MD, PHD

ABSTRACT

OBJECTIVES The aims of this study were to quantify the degree of improvement in vasovagal syncope after

assessment and to identify predictive factors.

BACKGROUND No treatments for vasovagal syncope have been proved effective, but patients in all prospective

studies appear to show a reduction in the likelihood of fainting.

METHODS A systematic review and meta-analysis was performed of studies published from 1993 through 2013.

Inclusion criteria were: 1) vasovagal syncope frequency in the preceding 1 to 2 years; and 2) the proportion of subjects

with syncope in at least the first follow-up year. Random-effects methods were used.

RESULTS Of 338 screened studies, 17 were analyzed, with a mean of 112 subjects (range 9 to 511 subjects). In the preceding

epoch, 97% of subjects fainted, with 2.6 � 1.0 syncopal spells per year. In the follow-up year, the proportion of patients

with$1 syncope recurrence was 677 of 1,912 (35.4%), and in the meta-analysis, the proportion of subjects fainting was only

0.44 (95% confidence interval: 0.41 to 0.46; p < 0.001). Subjects in larger studies were less likely to faint than those in

randomized trials (relative risk: 0.35 vs. 0.55; p ¼ 0.004). The probabilities of$1 syncope recurrence in the observational

versus randomized studies were 0.30 (95% confidence interval: 0.24 to 0.37) and 0.54 (95% confidence interval: 0.46 to

0.62), respectively (p < 0.001). None of the degree of blinding, type of intervention, age, sex, and number of recent faints

predicted the probability of syncope recurrence. Heterogeneity was very high in all analyses (I2 ¼ 60% to 96%).

CONCLUSIONS The spontaneous remission rate in highly symptomatic syncope patients is high, and remission oc-

curs in all types of studies. Improvement was more likely in larger and observational studies. (J Am Coll Cardiol EP

2017;3:384–92) © 2017 by the American College of Cardiology Foundation.

V asovagal syncope is a common problem thatreduces quality of life (1,2), but no treat-ments have been proved to be effective in

prospective, randomized, placebo-controlled clinicaltrials (3–6). However, patients in observational cohortstudies and both arms of randomized studies appear toshow a reduction in the likelihood of syncope (5–18).Both the magnitude and the cause of this effect areunknown. It appears to be large, and given the in-effectiveness of many treatments (19), it should beexplored further.

The magnitude of improvement is sufficientlylarge that randomized clinical trials of patientsrequire several hundred subjects for adequate po-wer, rather than the dozens that might be expected(20–22). Although regression to the mean is possible,the finding also might be due to a placebo effect (23).The latter is a complex effect with roots in patientexpectancy and interactions with caregivers. If so, itmight be possible to improve outcome by deliber-ately modifying how caregivers interact withpatients.

From the Libin Cardiovascular Institute of Alberta, Calgary, Alberta, Canada. The authors have reported that they have no

relationships relevant to the contents of this paper to disclose.




Manuscript received June 8, 2016; revised manuscript received September 27, 2016, accepted October 20, 2016.

J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 4 , 2 0 1 7

ª 2 0 1 7 B Y T H E A M E R I C A N CO L L E G E O F C A R D I O L O G Y F O U N DA T I O N


I S S N 2 4 0 5 - 5 0 0 X / $ 3 6 . 0 0


We performed a systematic review and meta-analysis of patient outcomes in observational studiesand the control arms of randomized control trials.Both types of studies were included because theyfeatured populations of patients prospectively studiedand not currently receiving proven effective therapies.The primary comparison was the proportion ofpatients fainting in the pre-study observational periodwith the study observation period.

METHODS

The study report is based on the guidelines ofPreferred Reporting Items for Systematic Reviewsand Meta-Analyses (24) and Meta-Analysis of Obser-vational Studies in Epidemiology (25).

ELIGIBILITY CRITERIA. Observational cohort studiesand randomized clinical trials of vasovagal syncopewere included. Patients were diagnosed withvasovagal syncope according to local institutionalstandards. The criteria stipulated adult studies with

full text availability, published in any lan-guage between 1993 and 2013, and if theyincluded the proportions of subjects whohad fainted in the proximate years precedingand following entry in the study.

INFORMATION SOURCES. We searched the Cochrane,PubMed, MEDLINE, and Embase online databases forobservational studies and randomized controlledtrials meeting the inclusion criteria. This wasfollowed with a hand search of the reference lists andthe investigators’ files.

SEARCH STRATEGY AND STUDY SELECTION. Wesearched for “vasovagal syncope” and related termsand subheadings, then separately searched for“neurally mediated syncope” and “neurocardiogenicsyncope.” We initially screened titles and abstracts toexclude irrelevant studies and those without primarydata. The remaining reports were subjected to a full-text review. The search results were reviewed by 2independent investigators (P.P., R.S.) and discrep-ancies resolved by consensus.

FIGURE 1 Flow Diagram of Study Selection for Systematic Review

SR ¼ systematic review.

AB BR E V I A T I O N S

AND ACRONYM S

RR = risk ratio

J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 4 , 2 0 1 7 Pournazari et al.A P R I L 2 0 1 7 : 3 8 4 – 9 2 Syncope Remission Systematic Review

385

DATA EXTRACTION AND ITEMS. A pilot data extrac-tion sheet was tested on the first 5 studies. P.P.extracted the data and R.S. validated them. Discrep-ancies were settled by consensus. All studies werescrutinized for potential overlap with others. Infor-mation was extracted on: 1) patient demographics(age, sex, number of years of syncope, number oflifetime faints, number and frequency of faints in thepreceding 0.5 to 2 years); 2) study structure (publi-cation year, enrollment period, randomized orobservational, degree of blinding, and type of inter-vention in the experimental and control arms); and 3)

study outcome (duration of follow-up and proportionof subjects fainting).

The studies did not report data completely or ho-mogeneously. Two studies (17,26) had variable recentobservation periods, and for these we report theminimum periods. Estimates for the duration of thefollow-up period were recorded as a pre-specifiedfixed observation period if possible; if not possible,estimates were taken from the maximum observationperiod from actuarial survival curves, from recordedranges, and in 2 studies (13,26) from means that hadnarrow SDs. Where not reported, recent syncopefrequency was estimated from syncope counts andobservation duration.

ASSESSMENT OF STUDY QUALITY. Methodologicalquality was assessed using the Newcastle-OttawaScale. This was selected because our interest was inthe outcomes of cohorts, not the effects of particularinterventions. Not all measures were applicable; the 6that were used were: 1) cohort representativeness;2) exposure ascertainment; 3) outcome absence atbaseline; 4) outcome assessment; 5) adequacy ofobservation duration; and 6) completeness of cohortfollow-up. The originally described measures do notcohere completely with the methodology of thestudies, and 2 modifications were made. The qualitiesof exposure ascertainment were: 1) electrocardiog-raphy during syncope and a case report form; 2)structured interview; 3) written self-report; and 4)verbal self-report. The first 2 levels were deemed to begood quality. The qualities of outcome assessmentwere ranked as: 1) blinded outcome assessment; 2) any

TABLE 1 Design and Structure of the Studies From Which the Cohorts Were Derived

First Author (Ref. #) (Year) Type Blinding Intervention Prior Duration, yrs Follow-Up Duration, yrs Patients, n

Raviele et al. (28) (2004) RCT DB Pacemaker 1 2.5 13

Brignole et al. (5) (2012) RCT DB Pacemaker 2 2 39

Raviele et al. (14) (1999) RCT DB Drug 2 1 55

Ventura et al. (18) (2002) RCT SB Drug 0.5 1 28

van Dijk et al. (17) (2006) RCT OL Physical 2 1.5 110

Madrid et al. (10) (2001) RCT DB Drug 1 1 24

Duygu et al. (9) (2008) RCT OL Physical 0.5 1 41

Sheldon et al. (15) (2006) RCT DB Drug 1 1 100

Sheldon et al. (6) (2016) RCT DB Drug 1 1 105

Occhetta et al. (13) (2004) RCT SB Pacemaker 1 3 9

Sutton et al. (16) (2000) RCT OL Pacemaker 2 6 23

Moya et al. (11) (2001) OBS – NA 2 1.5 82

Brignole et al. (8) (2006) (phase 1) OBS – NA 2 2 392



Aydin et al. (7) (2009) OBS Conservative 1 2 276

Natale et al. (26) (1996) OBS – Drug, 46; conservative, 8 0.5 3 54

DB ¼ double-blind; NA ¼ not available; OL ¼ open-label; RCT ¼ randomized controlled trial; SB ¼ single-blind.

TABLE 2 Study Participants With Proportion of Subjects in Cohorts With Syncope in the

Observation Periods

First Author(Ref. #) (Year)

Mean Age,yrs

Female,%

PriorFrequency

BaselineSyncope, %

Patients WithSyncope

Outcome, %

van Dijk et al. (17) (2006) 39 72 1.5 100 51 (56/110)

Madrid et al. (10) (2001) 31 63 2 100 46 (11/24)

Duygu et al. (9) (2008) 40 51 NA 100 56 (23/41)

Sheldon et al. (15) (2006) 41 68 3 94 35 (36/100)

Sheldon et al. (6) (2016) 28 71 4 96 60 (63/105)

Occhetta et al. (13) (2004) 58 44 4 100 78 (7/9)

Sutton et al. (16) (2000) 56 43 3 100 61 (14/23)

Moya et al. (11) (2001) 63 45 2 100 37 (30/82)

Brignole et al. (8) (2006) (phase 1) 66 55 2 100 36 (143/392)

Brignole et al. (8) (2006) (phase 2) 64 50 0.5 100 34 (17/50)

Brignole et al. (5) (2012) (phase 1) NA NA 2.5 100 36 (185/511)

Aydin et al. (7) (2009) 51.6 54 NA 100 24 (65/276)

Natale et al. (26) (1996) 48 54 NA 100 19 (10/54)

NA ¼ not available.

Pournazari et al. J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 4 , 2 0 1 7

Syncope Remission Systematic Review A P R I L 2 0 1 7 : 3 8 4 – 9 2

386

outcome assessment by study personnel; and 3) self-reporting or telephone call with no stated subsequentreview. The first 2 levels were deemed to be goodquality. The reports were assessed independently byall authors and discrepancies resolved by consensus.

SUMMARY MEASURES. The primary measure was therelative reduction of the proportion of subjects withsyncope in the follow-up period compared with thosein the pre-entry period. The primary analysis pooledall studies, and secondary subgroup analyses wereperformed on study type and descriptors and onbaseline patient demographics and clinical data.

STATISTICAL ANALYSIS. Comprehensive Meta-Analysis version 3 (Biosoft, Englewood, NewJersey) was used. For the primary analysis, wecalculated the risk ratio (RR) with 95% confidenceintervals using a random-effects analysis. We esti-mated heterogeneity with I2 in a secondary fixed-effect analysis, but in keeping with our a priorihypothesis, we do not report fixed-effect analysisestimates of risk reduction.

RESULTS

STUDY SELECTION AND CHARACTERISTICS. Therewere 338 citations, of which 17 cohorts in 15 reports

were included (Figure 1). Two studies had 2 cohortseach. The cohorts included 11 randomized controlledtrials (6 double-blind, 2 single-blind, and 3 open-label trials) and 6 observational studies (Table 1).Four randomized trials studied devices, 5 studieddrugs, and 2 studied nonbiomedical interventions(5–18).

STUDY QUALITY. Eleven studies had the highestquality-level indicators in all 6 measures. Threestudies (7,16,26) did not document details ofoutcome assessment, and 1 study (27) did not reportthe proportion of patients with follow-up data. Onestudy (10) did not document how outcomes wereassessed and had a high rate of loss to follow-up(20%). Three studies (9,17,18) had written self-reporting with no documentation of details ofoutcome assessment.

STUDY PARTICIPANTS. The cohorts had 9 to 511subjects, totaling 1,912 subjects (Table 2). The meanage was 49 � 12 years and 58 � 10% was women. Ofthe subjects, 547 (29%) were in randomized trials,and 1,315 were in prospective observational studies.Almost all subjects (1,859 of 1,912 [97%]) fainted inthe year preceding enrollment. There were 2.6 � 1.0faints in the year before enrollment, and symptomslasted 7.1 � 2.6 years. Only 43 of 1,912 subjects

FIGURE 2 Relative Risk for Syncope in Studies Grouped According to Sample Size

The median sample size was 54 subjects. 0.00 ¼ <54 subjects; 1.00 ¼ $54 subjects. CI ¼ confidence interval.


387

(2.2%) were lost to follow-up before a studyoutcome. Madrid et al. (10) reported 10 patientsprematurely lost but did not specify study arm.Raviele et al. (14), van Dijk et al. (17), POST (Pre-vention of Syncope Trial) I (15), and POST II reported8, 7, 4, and 5 patients, respectively, prematurely lostto follow-up in their control arms.

OUTCOMES AND OBSERVATION PERIODS. The me-dian recent observation period was 1 year, and themedian follow-up period was 2 years. Data for thepreceding period were available for 6 months (n ¼ 3),1 year (n ¼ 6), and 2 years (n ¼ 8). Data for theobservation period were available for 1 year (n ¼ 6),1.5 years (n ¼ 2), 2 years (n ¼ 5), 3 years (n ¼ 2), and 2.5and 6 years (n ¼ 1 each).

PRIMARY OUTCOME. Only 677 of 1,912 patients (36%)fainted in the observation period (p < 0.0001)(Table 2, Figure 2). In the random-effects analysis(Figure 2), the RR for having syncope in the follow-upobservation period compared with the period beforestudy entry was 0.44 (95% confidence interval: 0.41to 0.46; p < 0.001). There was a high degree of het-erogeneity (I2 ¼ 92%).

EFFECT OF STUDY DESIGN. Subjects in largerstudies were less likely to faint (RR: 0.35 vs. 0.55;p ¼ 0.004) (Figure 2). Study design had a significantassociation with the proportion of subjects withrecurrences (Table 3, Figure 3). Subjects in observa-tional studies were less likely to faint than those inrandomized trials (RR: 0.30 vs. 0.54; p < 0.001).However, the nature of blinding in the randomizedtrial was not statistically associated with syncoperecurrence. In the control arms of double-blind,single-blind, and open-label studies, the RRs of thefollow-up periods compared with the baselineperiods were 0.44, 0.65, and 0.59, respectively(p ¼ 0.09).

The nature of the intervention in the randomizedtrials had no significant effect on the likelihood ofoutcomes in their cohort control arms (Table 3,Figure 4). Four studies assessed device therapy, 5tested drug therapy, and 2 examined the effect ofnonmedical measures in the prevention of syncope.In the control arms of device, drug, and lifestylestudies, the RRs of the follow-up periods comparedwith the baseline periods were 0.56, 0.50, and 0.55,respectively.

EFFECT OF CLINICAL FACTORS. Neither patient agenor patient sex predicted a higher remission rate inthe following year (Table 3). Studies with mean fre-quencies of >2 faints in the preceding year reportedinsignificantly smaller reductions in syncope fre-quency compared with those reporting mean fre-quency of #2 episodes in the preceding year. In allsubgroups the degree of heterogeneity was high, withI2 values of 60% to 96%. This included the 2 signifi-cant subgroup analyses (sample size and studydesign), which both had I2 ¼ 89%.

DISCUSSION

We documented a high remission rate of syncope inobservational studies and the control arms of ran-domized control trials. We sought to provide an es-timate of the remission rate to guide future powercalculations and to uncover potential variables thataffect this rate.

HIGH REMISSION RATES. The remission rate, whichwe calculated as the probability of the absence ofsyncope during follow-up in patients who fainted in arecently preceding epoch, was high, highly signifi-cant, and ubiquitous. There was high heterogeneityseen in all groups and subgroups, suggesting thepresence of several variables affecting the likelihoodof remission. In the random-effects analysis, no clinical

TABLE 3 Risk Ratios for Fainting in the Period Before Study Enrollment

Compared With the Observation Period in the Study

Studies, nRiskRatio 95% CI

Significance,p Value

Total sample 17 0.44 0.41–0.46 <0.001

Study type

Cohort 6 0.30 0.24–0.37 <0.001

RCT 11 0.54 0.46–0.62

Blinding 0.09

Open label 2 0.59 0.48–0.72

Single blind 3 0.65 0.52–0.82

Double blind 6 0.44 0.33–0.58

Intervention 0.80

Lifestyle 2 0.55 0.48–0.64

Drug 5 0.50 0.37–0.66

Pacemaker 4 0.56 0.41–0.76

Age, yrs 0.27

<50 8 0.51 0.40–0.65

>50 8 0.41 0.30–0.55

Sex 0.31

Female #55% 6 0.39 0.27–0.57

Female >55% 8 0.49 0.41–0.59

Recent faints 0.068

#2/yr 7 0.37 0.27–0.52

>2/yr 7 0.53 0.44–0.65

Sample size 0.004

<54 8 0.55 0.46–0.66

$54 9 0.35 0.27–0.45

Publication year 0.69

Before June 2006 9 0.44 0.33–0.59

After June 2006 8 0.41 0.32–0.53



388

FIGURE 3 Relative Risk for Syncope in the Periods Before and After Enrollment in Cohorts

The cohorts are grouped according to whether they were derived from observational cohort studies (OBS) or randomized controlled trials

(RCT). CI ¼ confidence interval.

FIGURE 4 Relative Risk for Syncope in the Control Arms of Randomized Trials Grouped According to the Nature of the Intervention in

the Treatment Arms of the Trials

Abbreviations as in Figure 3.


389

variables predicted outcome. Patients who fainted lessoften at baseline might be more likely to not faint duringfollow-up, in keeping with our previous reports (29,30).Patients in open-label, nonrandomized cohort studieswere considerably more likely to improve than patientsin randomized trials, while the degree of blinding in therandomized studies had no significant effect on out-comes. The type of intervention in randomized armsalso did not predict improvement in control arms. Het-erogeneity might also be due to differences in inclusioncriteria among studies, several small and possibly highlyselective studies, and different degrees of autonomicnervous system improvement and patient learningamong studies.

POSSIBLE MECHANISMS AFFECTING IMPROVEMENT. Onepossibility is regression to the mean. Randomlyrecurring events occur in a Poisson distribution. Someevents occur earlier and some later, leading to theappearance of long and short intervals between eventsand periods of more and less frequent fainting. Anapparent improvement in syncope frequencywould bedescribed as regression to the mean. The hypothesis ofregression to the mean cannot be tested directly withstudy summary data but seems unlikely given that allthe studies reported such marked improvement.

A second possibility is the placebo effect, a complexcollection of factors (31–34). These include behavioralfactors such as conditioning (31), expectancy (34), andphysician behavior (32,33,35). Here the invasivenessof the intervention did not affect apparent improve-ment. Interestingly, 2 blinded studies of arthroscopicknee and back surgery concluded that much ofsymptom suppression was due to expectancy on thepart of patients and surgeons, not to the biomedicalconsequences of the operations (36,37). Indeed, 1possibility for the high degree of statistical heteroge-neity might be heterogeneity among local styles ofinteraction among patients and caregivers. The pa-tients in nonrandomized studies have better outcomesthan those in randomized trials. This is consistent withthe notion that patients in randomized trials knowthat they are possibly exposed to inactive agents,whereas those in cohort studies believe that theirtreatment is active. This might lead to increasedexpectation of improvement (31,34,38).

EXPECTANCY. Kirsch originally hypothesized thatplacebo effects are produced by a self-fulfilling effecttermed expectancy (34). In this concept, if a personbelieves that he or she will improve, eventuallyimprovement occurs. Subjects who expect to improvedemonstrate subject response expectancy, whilethose whose physicians expect them to improvedemonstrate observer response expectancy. Given

the large magnitude of the effect sizes in this report,it may be that understanding and harnessing theseeffects will lead to patient improvement withoutbiomedical interventions.

IMPLICATIONS FOR FUTURE RESEARCH. Given thestrikingheterogeneity in howstudieswere conducted andreported, future comparisons and integrations wouldbenefit from uniform standards for study design, patientinclusion, and data collection. A pressing issue is todetermine whether the substantial patient improvementin the absence of proven effective biomedical in-terventions is due to regressions to themeans or to factorsrelated to the placebo effect. Most patients present after arecent worsening (39), and prognosis can be predicted bytheir recent syncope frequency (30). These results andthose of this study suggest that there are intrapatientfactors that canworsen, that persist, and that can improve.Future work might address these issues with the intentof developing therapeutic approaches targeting thesenovel factors. These studies may require patient-specificdata.

STUDY LIMITATIONS. This approach combinedpurely observational cohort studies with the controlarms of randomized trials. Unidentified differences inpatient composition, style of care, or cultural expec-tations might have contributed to the heterogeneity,and we are unable to identify all the factors. Theremight be publication bias, in that negative random-ized studies might not be published, and these mightbe negative because of large improvements in controlsubjects. We captured only the likelihood that sub-jects fainted, not the frequency of fainting. Therefore,if anything, our results underestimate the magnitudeof improvement. In contrast, the very high proportionof patients who faint at baseline means that effectscan only have a neutral or negative effect on the pro-portion of patients who faint. However, these are thepatients most likely to merit treatment attempts, andquantifying and understanding this effect is impor-tant. Whether these 1,912 study subjects are repre-sentative of the larger population of patients withfrequent vasovagal syncope is unknown. The follow-up observation period was slightly longer than thebaseline period, which if anything would increase thelikelihood of syncope in follow-up. There was variablereporting, and in some cases it was necessary to esti-mate observation duration from survival curves. Twoof the reports each contributed 2 cohorts, which weretherefore not completely statistically independent,although the values fall well within the range of theothers. Finally, we modified the Newcastle-OttawaScale to suit this study, and the modifications(as with the scale itself) have not been validated (40).



390

CONCLUSIONS

Syncope likelihood decreases in both observationalstudies and the control arms of randomized trials. Itslarge magnitude has important implications for trialdesign and suggests a heretofore unexplored thera-peutic effect that, if improved, might enhance patientoutcomes without specific medical therapy.

ADDRESS FOR CORRESPONDENCE: Dr. Robert Shel-don, Libin Cardiovascular Institute of Alberta, Uni-versity of Calgary, 3280 Hospital Drive NW, Calgary,Alberta T2N 4Z6, Canada. E-mail: [email protected].

RE F E RENCE S

1. Rose MS, Koshman ML, Spreng S, Sheldon R.The relationship between health-related quality oflife and frequency of spells in patients with syn-cope. J Clin Epidemiol 2000;53:1209–16.

2. Linzer M, Gold DT, Pontinen M, Divine GW,Felder A, Brooks WB. Recurrent syncope as achronic disease: preliminary validation of adisease-specific measure of functional impairment.J Gen Intern Med 1994;9:181–6.

3. Saklani P, Krahn A, Klein G. Syncope. Circulation2013;127:1330–9.

4. Kuriachan V, Sheldon RS, Platonov M.Evidence-based treatment for vasovagal syncope.Heart Rhythm 2008;5:1609–14.

5. Brignole M, Menozzi C, Moya A, et al. Pacemakertherapy in patients with neurally mediated syncopeand documented asystole: Third International Studyon Syncope of Uncertain Etiology (ISSUE-3): a ran-domized trial. Circulation 2012;125:2566–71.

6. Sheldon RS, Raj SR, Rose MS, et al. The SecondPrevention of Syncope Trial (POST 2): a random-ized, placebo-controlled study of fludrocortisonein the prevention of vasovagal syncope. J Am CollCardiol 2016;68:1–9.

7. Aydin MA, Maas R, Mortensen K, et al. Pre-dicting recurrence of vasovagal syncope: a simplerisk score for the clinical routine. J CardiovascElectrophysiol 2009;20:416–21.

8. Brignole M, Sutton R, Menozzi C, et al. Earlyapplication of an implantable loop recorder allowseffective specific therapy in patients with recur-rent suspected neurally mediated syncope. EurHeart J 2006;27:1085–92.

9. Duygu H, Zoghi M, Turk U, et al. The role of tilttraining in preventing recurrent syncope in pa-tients with vasovagal syncope: a prospective andrandomized study. Pacing Clin Electrophysiol2008;31:592–6.

10. Madrid AH, Ortega J, Rebollo JG, et al. Lack ofefficacy of atenolol for the prevention of neurallymediated syncope in a highly symptomatic popu-lation: a prospective, double-blind, randomizedand placebo-controlled study. J Am Coll Cardiol2001;37:554–9.

11. Moya A, Brignole M, Menozzi C, et al. Mecha-nism of syncope in patients with isolated syncopeand in patients with tilt-positive syncope. Circu-lation 2001;104:1261–7.

12. Natale A, Newby KH, Dhala A, Akhtar M, Sra J.Response to beta blockers in patients with neuro-cardiogenic syncope: how to predict beneficial ef-fects. J Cardiovasc Electrophysiol 1996;7:1154–8.

13. Occhetta E, Bortnik M, Audoglio R,Vassanelli C. Closed loop stimulation in preventionof vasovagal syncope. Inotropy Controlled Pacingin Vasovagal Syncope (INVASY): a multicentrerandomized, single blind, controlled study. Euro-pace 2004;6:538–47.

14. Raviele A, Brignole M, Sutton R, et al. Effect ofetilefrine in preventing syncopal recurrence in pa-tients with vasovagal syncope: a double-blind, ran-domized, placebo-controlled trial. The VasovagalSyncope International Study. Circulation 1999;99:1452–7.

15. Sheldon R, Connolly S, Rose S, et al. Preven-tion of Syncope Trial (POST): a randomized,placebo-controlled study of metoprolol in theprevention of vasovagal syncope. Circulation2006;113:1164–70.

16. Sutton R, Brignole M, Menozzi C, et al., for theVasovagal Syncope International Study (VASIS)Investigators. Dual-chamber pacing in the treat-ment of neurally mediated tilt-positive car-dioinhibitory syncope: pacemaker versus notherapy: a multicenter randomized study. Circula-tion 2000;102:294–9.

17. van Dijk N, Quartieri F, Blanc JJ, et al. Effec-tiveness of physical counterpressure maneuvers inpreventing vasovagal syncope: the PhysicalCounterpressure Manoeuvres Trial (PC-Trial). J AmColl Cardiol 2006;48:1652–7.

18. Ventura R, Maas R, Zeidler D, et al.A randomized and controlled pilot trial of beta-blockers for the treatment of recurrent syncopein patients with a positive or negative response tohead-up tilt test. Pacing Clin Electrophysiol 2002;25:816–21.

19. Sheldon RS, Grubb BP, Olshansky B, et al. 2015Heart Rhythm Society expert consensus statement

on the diagnosis and treatment of posturaltachycardia syndrome, inappropriate sinus tachy-cardia, and vasovagal syncope. Heart Rhythm2015;12:e41–63.

20. Raj SR, Faris PD, McRae M, Sheldon RS.Rationale for the Prevention of Syncope Trial IV:assessment of midodrine. Clin Auton Res 2012;22:275–80.

21. Raj SR, Rose S, Ritchie D, Sheldon RS. TheSecond Prevention of Syncope Trial (POST II)—arandomized clinical trial of fludrocortisone for theprevention of neurally mediated syncope: ratio-nale and study design. Am Heart J 2006;151:1186.e11–7.

22. Sheldon R, Rose S, Connolly S. Prevention ofSyncope Trial (POST): a randomized clinical trial ofbeta blockers in the prevention of vasovagal syn-cope; rationale and study design. Europace 2003;5:71–5.

23. Sahota I, Sheldon R, Pournazari P. Clinicalimprovement of vasovagal syncope in the absenceof specific therapies: the Seinfeld effect. Cardiol J2014;21:637–42.

24. Liberati A, Altman DG, Tetzlaff J, et al. ThePRISMA statement for reporting systematic re-views and meta-analyses of studies that evaluatehealth care interventions: explanation and elabo-ration. J Clin Epidemiol 2009;62:e1–34.

25. Stroup DF, Berlin JA, Morton SC, et al., for theMeta-Analysis of Observational Studies in Epide-miology (MOOSE) Group. Meta-analysis of obser-vational studies in epidemiology: a proposal forreporting. JAMA 2000;283:2008–12.

26. Natale A, Geiger MJ,Maglio C, et al. Recurrenceof neurocardiogenic syncope without pharmaco-logic interventions. Am J Cardiol 1996;77:1001–3.

27. Brignole M, Ungar A, Bartoletti A, et al. Stan-dardized-care pathway vs. usual management ofsyncope patients presenting as emergencies atgeneral hospitals. Europace 2006;8:644–50.

28. Raviele A, Giada F, Menozzi C, et al.A randomized, double-blind, placebo-controlledstudy of permanent cardiac pacing for the treat-ment of recurrent tilt-induced vasovagal syncope.

PERSPECTIVES

COMPETENCY IN MEDICAL KNOWLEDGE: The sponta-

neous remission rate in patients with frequent vasovagal syncope

is very high. For many patients with frequent vasovagal syncope,

it might be reasonable to follow them expectantly for a brief

period before starting specific medical therapy.

TRANSLATIONAL OUTLOOK: The high rate of remission will

help guide power calculations in future studies and suggests that

a better understanding of the placebo effect might improve pa-

tient care.


391

The Vasovagal Syncope and Pacing Trial (SYN-PACE). Eur Heart J 2004;25:1741–8.

29. Sheldon R, Rose S, Flanagan P, Koshman ML,Killam S. Risk factors for syncope recurrence aftera positive tilt-table test in patients with syncope.Circulation 1996;93:973–81.

30. Sumner GL, Rose MS, Koshman ML, Ritchie D,Sheldon RS. Recent history of vasovagal syncope in ayoung, referral-based population is a stronger predic-tor of recurrent syncope than lifetime syncopeburden.J Cardiovasc Electrophysiol 2010;21:1375–80.

31. Voudouris NJ, Peck CL, Coleman G. The role ofconditioning and verbal expectancy in the placeboresponse. Pain 1990;43:121–8.

32. Miller FG, Rosenstein DL. The nature and po-wer of the placebo effect. J Clin Epidemiol 2006;59:331–5.

33. Margo CE. The placebo effect. Surv Oph-thalmol 1999;44:31–44.

34. Kirsch I. Response expectancy as a determi-nant of experience and behavior. Am Psychol1985;40:1189–202.

35. Hrobjartsson A, Gotzsche PC. Is the placebopowerless? An analysis of clinical trials comparingplacebo with no treatment. N Engl J Med 2001;344:1594–602.

36. Sihvonen R, Paavola M, Malmivaara A, et al.Arthroscopic partial meniscectomy versus shamsurgery for a degenerative meniscal tear. N Engl JMed 2013;369:2515–24.

37. Kallmes DF, Comstock BA, Heagerty PJ, et al.A randomized trial of vertebroplasty for osteopo-rotic spinal fractures. N Engl J Med 2009;361:569–79.

38. Colagiuri B, Smith CA. A systematic review ofthe effect of expectancy on treatment responsesto acupuncture. Evid Based Complement AlternatMed 2012;2012:857804.

39. Sheldon RS, Sheldon AG, Serletis A, et al.Worsening of symptoms before presentation withvasovagal syncope. J Cardiovasc Electrophysiol2007;18:954–9.

40. Stang A. Critical evaluation of the Newcastle-Ottawa Scale for the assessment of the quality ofnonrandomized studies in meta-analyses. Eur JEpidemiol 2010;25:603–5.

KEY WORDS observational, randomizedclinical trial, systematic review, vasovagalsyncope



392

Does the ImplantableCardioverter-Defibrillator BenefitVary With the Estimated Proportional Riskof Sudden Death in Heart Failure Patients?Wayne C. Levy, MD,a Yanhong Li, MD, MS,b Shelby D. Reed, PHD,b Michael R. Zile, MD,c Ramin Shadman, MD,d

Todd Dardas, MD,a David J. Whellan, MD,e Kevin A. Schulman, MD,b Stephen J. Ellis, PHD,b Matthew Neilson, PHD,f

Christopher M. O’Connor, MD,b for the HF-ACTION Investigators

ABSTRACT

OBJECTIVES In this study, the authors developed the Seattle Proportional Risk Model (SPRM) to estimate the pro-

portion of total mortality due to sudden death. We prospectively validated the model in HF-ACTION (Participants in Heart

Failure: A Controlled Trial Investigating Outcomes of Exercise Training) and tested whether the implantable cardioverter-

defibrillator (ICD) benefit varied with the SPRM.

BACKGROUND Prediction of which heart failure patients are most likely to die of sudden death versus nonsudden

death is an important factor in determining who will benefit the most from an ICD.

METHODS Among 2,331 patients enrolled, 1,947 patients were retained for analysis over a median follow-up of 2.5

years. The SPRM was calculated using age, gender, diabetes, body mass index, systolic blood pressure, ejection fraction,

New York Heart Association functional class, sodium, creatinine, and digoxin use.

RESULTS An ICD (ICD or CRT-D) was in use before death in 1,204 patients (62%). SPRM was predictive of sudden

death versus nonsudden death in those without an ICD (p ¼ 0.002). The hazard ratio representing ICD versus no ICD was

0.63 for all-cause mortality (p ¼ 0.0002). The ICD benefit varied with the SPRM for all-cause mortality (p ¼ 0.001),

with a greater benefit in those with a higher conditional probability of sudden death.

CONCLUSIONS In population of ambulatory patients with a New York Heart Association functional class II-IV HF and

ejection fraction of #35%, the SPRM was predictive of the proportional risk of sudden versus nonsudden death. ICDs

were associated with a decreased risk of all-cause mortality by 37% and the ICD benefit varied with the SPRM. The

SPRM may be useful in risk stratifying patients for a primary prevention ICD. (Exercise Training Program to Improve

Clinical Outcomes in Individuals With Congestive Heart Failure; NCT00047437) (J Am Coll Cardiol EP 2017;3:291–8)

© 2017 by the American College of Cardiology Foundation.

S udden death comprises one-half of all deaths inpatients with chronic heart failure (1). Meta-analysis of primary prevention implantable

cardioverter defibrillator (ICD) trials suggests thatICDs decrease sudden death by approximately 60%

(2) (relative risk reduction). In many patients, suddendeath is amarker of the progression of their underlyingheart failure. As a result, the prevention of suddendeathmaymerely alter themode of death from suddento pump failure, as seen in post myocardial infarction

From the aUniversity of Washington, Seattle, Washington; bDuke Clinical Research Institute, Duke University, Durham, North

Carolina; cMedical University of South Carolina and RHJ Department of Veterans Affairs Medical Center, Charleston, South Car-

olina; dSouthern California Permanente Medical Group, Los Angeles, California; eJefferson Medical College, Philadelphia, Penn-

sylvania; and the fUniversity of Glasgow, Glasgow, United Kingdom. HF-ACTION was funded by grants 5U01HL063747,

5U01HL066461, 5U01HL068973, 5U01HL066501, 5U01HL066482, 5U01HL064250, 5U01HL066494, 5U01HL064257, 5U01HL066497,

5U01HL068980, 5U01HL064265, 5U01HL066491, and 5U01HL064264 from theNationalHeart, Lung, andBlood Institute; and grants

R37AG018915 and P60AG010484 from the National Institute on Aging. Dr. Levy has received research funding from HeartWare,

Medtronic, Resmed, Amgen, andGEHealthcare; and is a consultant to Novartis, GEHealthcare, Pharmin, Abbot, Relypsa, andMicro

Systems Engineering. Drs. Reed, Li, Schulman and Levy received funding from grant 5R01NR011873 from the National Institute of

Nursing Research. All other authors have reported that they have no relationships relevant to the contents of this paper.




I S S N 2 4 0 5 - 5 0 0 X / $ 3 6 . 0 0


trials with ICDs (3). ICDs are a Class I indica-tion by American College of Cardiology/American Heart Association/Heart RhythmSociety guidelines to prevent sudden deathin New York Heart Association (NYHA) func-tional class II and III patients with an ejectionfraction (EF) of#35% and other selected heartfailure patients with a life expectancy of >1year (4). However, the usefulness of a primaryprevention ICDmay be diminished in patientswho are older, women, have chronic kidneydisease, or multiple comorbidities (5–8). The2013 guidelines were updated to include thefollowing statement, “The usefulness of im-plantation of an ICD is of uncertain benefitto prolong meaningful survival in patientswith a high risk of non-sudden death aspredicted by frequent hospitalizations,

advanced frailty, or comorbidities such as systemicmalignancy or severe renal dysfunction” (Class IIb).In the National Cardiovascular Data Registry forprimary prevention ICDs, one-half of patients are inNYHA functional class III/IV (9) and one-third are $75years of age, patients in whom a primary preventionICD may have a diminished benefit (7).

Prediction of which heart failure patients are mostlikely to die of sudden death versus nonsudden deathmay provide better risk stratification than NYHAfunctional class and EF. For example, at the sameannual mortality a patient who has a 70% likelihood ofdying from sudden death, conditional on dying, wouldbe expected to derive more benefit from an ICD than asimilar patient who has a 30% likelihood of dying fromsudden death (10). To facilitate incorporating suchinformation into treatment decisions, we derived theSeattle Proportional Risk Model (SPRM) in a separatecohort of patients without an ICD (9,985 patients with2,552 deaths and 48% sudden death) not to predict therisk of death, but rather if a patient dies, the mode ofdeath (sudden vs. nonsudden) (11). The model foundthe proportion of sudden death was greater withyounger age, male gender, lack of diabetes mellitus,lower EF, better NYHA functional class (i.e., II vs. III orIV), higher body mass index, digoxin use, and valuesof systolic blood pressure (SBP), sodium, and creati-nine closer to the normal range (Figure 1).

We prospectively applied the SPRM to data fromthe HF-ACTION (Participants in Heart Failure: AControlled Trial Investigating Outcomes of ExerciseTraining) clinical trial (12). Our aim was to: 1) validatewhether the model predicts the proportion of suddenversus nonsudden death; and 2) determine if thebenefit of an ICD varied with the estimated condi-tional probability of sudden death. We hypothesizedthat there would be a greater relative ICD benefit onsudden death and total mortality in those patientswith a higher predicted proportion of mortality fromsudden death.

METHODS

HEART FAILURE. The HF-ACTION was a clinical trial(NCT00047437) of exercise training in 2,331 ambula-tory patients with NYHA functional class II, III, and IVheart failure and an EF of #35% (12). We excludedpatients who received a left ventricular assist deviceor underwent cardiac transplantation (n ¼ 78),patients who were missing baseline variables neces-sary to calculate the SPRM score (n ¼ 281) and missingvalues necessary to calculate the SHFM score (n ¼ 25),resulting in a sample of 1,947 patients. The mode ofdeath was adjudicated by a clinical events committee(13). Sudden death was defined as unexpected andotherwise unexplained death in a previously stablepatient, including patients who were comatose andthen died after attempted resuscitation. Patients inthis category should have had recent human contactbefore the event. Patients who died and had been outof contact for prolonged periods of time were classi-fied as ‘unknown’mode of death. For this analysis, theendpoint of sudden death included those classified bythe clinical events committee as sudden death or un-known mode of death as described. We combinedthese endpoints because it is more similar to themethods used in the trials within which the SPRM wasderived. The SPRM score was calculated as previouslydescribed (11). We defined “ICD use” if an ICD or car-diac resynchronization therapy with ICD (CRT-D) waspresent at baseline or implanted before the end offollow-up (death or end of the trial). It is our antici-pation that the CRT benefit of a device that is alreadypresent on mortality is already reflected in the SHFMby improvements in the SBP, EF, and NYHA functionalclass (14). Thus, in a CRT-D device present at baseline,

SEE PAGE 299


AND ACRONYMS

CRT-D = cardiac

resynchronization therapy with

ICD

EF = ejection fraction

HR = hazard ratio

ICD = implantable

cardioverter-defibrillator

NYHA = New York Heart

Association

SBP = systolic blood pressure

SHFM = Seattle Heart Failure

Model

SPRM = Seattle Proportional

Risk Model

VF = ventricular fibrillation

VT = ventricular tachycardia




Manuscript received April 18, 2016; revised manuscript received August 25, 2016, accepted September 1, 2016.

Levy et al. J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 3 , 2 0 1 7

The SPRM-HF-ACTION Trial M A R C H 2 0 1 7 : 2 9 1 – 8

292

the additional benefit of the device is due to the ICDpart of the CRT-D. Consequently, ICDs and CRT-Dswere treated as “ICDs” in this analysis as the major-ity of CRT-Ds were present at baseline. ICD or CRT-Dhad to be placed $6 weeks before enrollment per theHF-ACTION protocol. Patients who may have had anICD explant remained in the ICD group. For the firstaim—to validate the SPRM—we used logistic regressionto compare the SPRM-predicted versus the observedproportional risk of sudden death by quartiles of theSPRM among those without an ICD, in the total cohortwith an expectation that the risk of sudden deathwould be less in those with an ICD. Calibration wasassessed by the Hosmer-Lemeshow goodness of fit testand discrimination by receiver operating character-istic area under the curve in patients without an ICD.

For the second aim, to ascertain if the benefit ofan ICD varied with the estimated conditional proba-bility of sudden death, we used a Cox proportionalhazards model to examine the effect of ICDs on all-cause mortality. We used the 3-year survival in theICD group in each SPRM quartile and the Cox modeldevice hazard ratio (HR) to estimate life-years addedwith an ICD over a patients lifespan (Gompertzmethod), number needed to treat for 3 years, and yearsneeded to treat to add 1 year of life (15). To examine theimpact of ICDs on sudden death and nonsudden death,we applied a regression technique that uses pseudo-values fromcumulative incidence functions to accountfor competing risks from the othermodes of death (16).With both the Cox model and the pseudovaluescompeting risk models, we applied an interaction termto evaluate whether the effects of ICDs varied acrossthe conditional risk of sudden death as estimated usingthe SPRM and included a baseline covariate repre-senting a history of ventricular tachycardia/ventricu-lar fibrillation (VT/VF) as a proxy for indication for ICDimplantation, because the specific ICD indication wasnot available. We also adjusted for Seattle Heart Fail-ure Model (SHFM) scores (14) to evaluate whether ICDbenefits varied across individuals with varying SPRMscores when adjusting for risk of all-causemortality (aspredicted using SHFM) as previously described (11).For a sensitivity analysis, we adjusted for 18 individualvariables in the Cox model for all-cause mortality. Theanalyses were performed using SAS version 9.3 (SASInstitute Inc., Cary, North Carolina) with p # 0.05considered significant. The analyses were approved bythe Duke University Institutional Review Board.

RESULTS

Among the 1,947 patients who met the study’s inclu-sion criteria, the mean age was 59 years, 73% male, EF

of 25%, diabetes in 33%, and NYHA functional class IIor III in 99% of subjects, similar to the SCD-HeFT(Sudden Cardiac Death in Heart Failure Trial). Sixty-two percent had an ICD before the end of follow-up.Of the patients with an ICD at the end of follow-up,893 patients (74%) had an ICD at baseline (ICD 28%,CRT-D 18% of all patients) and 311 patients (26%)received an ICD during follow-up (ICD 13%, CRT-D 5%of all patients). The baseline demographics of thosewith and without an ICD before the end of follow-upare shown in Table 1. Patients with an ICD werehigher risk as judged by many markers, includingolder age, history of VT/VF, higher NYHA functionalclass, creatinine, diuretic dose, and lower EF and SBP.This is reflected by an approximately 33% higherSHFM estimated 1-year all-cause mortality (notincluding the benefit of the ICD). Patients with a CRT-D versus ICD had a higher SHFM estimated 1-yearmortality with medical therapy (8.6 � 6.5% vs. 7.5 �6.7%; p ¼ 0.0033) and a lower SPRM predicted pro-portion of sudden death (53.8 � 13.6% vs. 56.3� 14.1%;p ¼ 0.0022).

During a mean of 2.5 years of observation, 328patients (16.8%) died, with 138 deaths among 743patients (18.6%) without an ICD and 190 deaths

FIGURE 1 Impact of Clinical Risk Factors on the Mode of Death

The Seattle Proportional Risk Model (SPRM) is a logistic regression model that provides a

patient level estimate of the proportion (not the absolute risk) of sudden versus non-

sudden death based on age, gender, ejection fraction (EF), New York Heart Association

(NYHA) functional class, diabetes mellitus, systolic blood pressure (SBP), sodium, creati-

nine, digoxin use, and body mass index (BMI).

J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 3 , 2 0 1 7 Levy et al.M A R C H 2 0 1 7 : 2 9 1 – 8 The SPRM-HF-ACTION Trial

293

among 1,204 patients (15.8%) with an ICD. Amongthose who died, the distributions of the variousmodes of death differed between those with andwithout an ICD: the proportion of patients with sud-den death was lower in those with an ICD comparedwith those without an ICD (31% vs. 59%; p # 0.0001),and the proportion of patients with pump failuredeaths was higher in those with an ICD comparedwith those without an ICD (39% vs. 18%; p # 0.0001).Other modes of death were similar (Table 2).

AIM 1: DETERMINE WHETHER SPRM PREDICTS THE

PROPORTION OF SUDDEN VERSUS NONSUDDEN

DEATHS. Overall, the SPRM-predicted conditionalprobability of sudden death was 56%. The SPRM-predicted conditional risk of sudden death was 57%in patients alive at the end of follow-up in the trial,55% in those who died of sudden death, 46% in thosewith pump failure death, and 48% in those who diedof other cardiovascular or noncardiovascular death(p < 0.0001). Among patients without an ICD, thepredicted and observed proportions of deaths thatwere sudden were similar (57% vs. 59%). The SPRMwas predictive sudden versus nonsudden death inpatients without an ICD (logistic regression, oddsratio: 2.05; p ¼ 0.002) providing validation of theSPRM for prediction of sudden versus nonsuddendeath. The Hosmer-Lemeshow goodness of fit test inthe patients without an ICD showed adequate SPRMcalibration (p ¼ 0.51). The receiver operating charac-teristic area under the curve for sudden versus non-sudden death was 0.64 in the cohort of patients whonever had an ICD, the same as the SPRM derivationcohort (0.64). For comparison, an EF of #35% andNYHA functional classes II and III had a receiveroperating characteristic area under the curve of 0.52in the SPRM derivation cohort. Among patients withICDs, the observed proportions of deaths attributableto sudden death were lower than predicted (oddsratio: 0.32; p < 0.0001), representative of the ICDbenefit on sudden death (Figure 2).AIM 2: EVALUATE WHETHER ICD BENEFIT VARIES

WITH THE SPRM SCORE. In the Cox proportionalhazards model, adjusted for a history of VT/VF, all-cause mortality (per SHFM), and proportion of mor-tality due to sudden death (per SPRM), an ICD wasassociated with a 37% reduction in all-cause mortality(HR: 0.63; p ¼ 0.0002). When an interaction term wasadded to test whether the benefit of ICDs variedacross SPRM scores, the results indicated that ICDswere associated with a lower relative risk of deathamong patients with higher SPRM scores (higherproportion of sudden death; HR: 0.58; interactionp ¼ 0.001 with SPRM modeled as a continuous vari-able) (Figure 3A). The SPRM interaction was similarfor both ICD (HR: 0.56; p ¼ 0.002) and CRT-D (HR:0.61; p ¼ 0.029). When modeling the risk of all-causedeath using SPRM quartile groups instead, treatmenteffects of ICDs varied across SPRM quartiles with thegreatest benefit in the highest SPRM quartile(Figure 3A), validating the hypothesis. The point of noICD benefit in the fitted Cox model occurred at aSPRM of approximately 32% (Figure 3A). Within thefirst quartile of SPRM, patients with a SPRM of #32%predicted sudden death had a trend for ICD harm (HR:

TABLE 2 The HF-ACTION Mode of Death

All Patients(N ¼ 328)

No ICD(n ¼ 138)

ICD(n ¼ 190) p Value

Sudden cardiac death 140 (43) 81 (59) 59 (31)

<0.0001Pump failure death 99 (30) 25 (18) 74 (39)

Other cardiovascular death 32 (10) 10 (7) 22 (12)

Noncardiovascular death 57 (17) 22 (16) 35 (18)

Values are n (percent of total deaths). A chi-square test was used to compare distributionsbetween patients with and without implantable cardiac defibrillators (ICDs).

HF-ACTION ¼ Participants in Heart Failure: A Controlled Trial Investigating Outcomes ofExercise Training.

TABLE 1 Baseline Demographics

No ICD(n ¼ 743)

ICD(n ¼ 1,204) p Value*

Age (yrs) 58 (13) 60 (12) 0.03

Male 492 (66) 926 (77) <0.0001

Ejection fraction (%) 27 � 8 24 � 7 <0.0001

Systolic blood pressure(mm Hg)

118 � 19 112 � 18 <0.0001

NYHA functional class

II 517 (70) 720 (60)

III 221 (30) 469 (40) <0.0001

IV 5 (1) 15 (1)

History of VT/VF 16 (2) 283 (24) <0.0001

Diabetes mellitus 227 (31) 413 (34) 0.09

Body mass index (kg/m2) 31 � 8 31 � 7 0.54

Sodium (mEq/l) 139.4 � 3.1 139.1 � 3.7 0.07

Creatinine (mg/dl) 1.24 � 0.74 1.39 � 0.87 <0.0001

ACEI/ARB 707 (95) 1130 (94) 0.23

Beta-blocker 702 (94) 1143 (95) 0.66

Aldosterone blocker 285 (38) 587 (49) <0.0001

Digoxin 279 (38) 603 (50) <0.0001

Furosemide equivalent dose(mg/kg/day)

0.5 � 0.6 0.7 � 0.7 <0.0001

SPRM-predicted risk of SCDconditional on death

57 (15) 55 (14) 0.01

SHFM-predicted risk of 1-yrmortality†

5.9 (4.7) 7.9 (6.7) <0.0001

Values are n (%) or mean � SD. *Student t test for continuous variables and chi-square test for categorical variable. †Estimated without the effect of ICD im-plantation included.

ACEI/ARB ¼ angiotensin-converting enzyme/angiotensin receptor blocker;ICD ¼ implantable cardioverter-defibrillator; NYHA ¼ New York Heart Association;SCD ¼ sudden cardiac death; SD ¼ standard deviation; SHFM ¼ Seattle HeartFailure Model; SPRM ¼ Seattle Proportional Risk Model; VT/VF ¼ ventriculartachycardia/ventricular fibrillation.



294

1.44; p ¼ 0.31), whereas those with an SPRM of >32%within the first quartile had a trend for ICD benefit(HR: 0.73; p ¼ 0.23; interaction p ¼ 0.10). If we adjustfor age, sex, EF, NYHA functional class, SBP, ischemicetiology, diuretic dose, angiotensin-convertingenzyme inhibitor, angiotensin receptor blockers,beta blockers, aldosterone antagonists, statins, serumsodium, total cholesterol, hemoglobin, lymphocytes,uric acid, and VT/VF in the all-cause mortality Coxmodel, the ICD overall benefit was 0.62 (p ¼ 0.0002)and the SPRM had an interaction p value of 0.0002,very similar to the SHFM-adjusted model. The life-years added with ICD treatment for a lifetime forthe lowest to highest SPRM quartile is estimated at0.9, 1.5, 2.7, and 5.7 years. The number needed totreat for 3 years was 16.9, 18.2, 21.2, and 9.1. Analternative metric, the years needed to treat to add1 year of life is 6.7, 5.5, 4.4, and 2.3 (Online Table 1).

When applying the competing risks analysis toexamine the risk of sudden death, adjusted for ahistory of VT/VF, all-cause mortality (per SHFM), andproportion of mortality due to sudden death (perSPRM), an ICD was associated with a 67% reduction inthe hazard of sudden death (HR: 0.33; p < 0.0001).When an interaction term between SPRM scores andICDs was included in the model, the results suggestedthat ICDs may impart a greater relative reduction onthe risk of sudden death in individuals with higherSPRM scores (i.e., higher proportion of mortality fromsudden death) to a greater extent than individualswith lower SPRM scores (interaction p ¼ 0.07 withSPRM modeled as a continuous variable) (Figure 3B).

When applying the competing risks analysis toexamine the risk of nonsudden death, an ICD was notassociated with an increase in the hazard of non-sudden death (HR: 1.05; p ¼ 0.81, and adjusted forVT/VF, SHFM score, and SPRM score).

A history of VT/VF was associated with a 1.7-foldincreased risk of all-cause mortality (p ¼ 0.0006), a1.8-fold increased risk of sudden death (p ¼ 0.05), anda 1.6-fold increased risk of nonsudden death(p ¼ 0.01). Thus, within this cohort, a history of VT/VFwas associated with a similar increase in the risk ofsudden and nonsudden death and did not predictpatients with a greater proportion of sudden versusnonsudden death.

To address confounding by ICD placement duringthe trial, a sensitivity analysis was performed withexclusion of patients who received an ICD afterrandomization (i.e., comparison of no ICD before theend of follow-up and ICD at baseline). The relativeICD benefit remained greater in those with a higherSPRM scores (interaction p ¼ 0.011) (Online Figures 1Aand 1B).

If we start a typical HF-ACTION patient, a 58-year-old male, NYHA functional class II, EF 27%, SBP118 mm Hg, sodium 138 mEq/l, creatinine 1.2 mg/dl,no diabetes mellitus, no digoxin, and a body mass in-dex of 31 kg/m2, the SPRM is approximately 64% pre-dicted proportion of sudden death. Changing any 1 ofthe following variables has a similar effect on theSPRM predicted proportion of sudden death and willdecrease it by approximately 7% (NYHA functionalclass III vs. II, EF 42% vs. 27%, female vs. male, aging12 years, body mass index 25 vs. 31 kg/m2, SBP 88vs. 118 mm Hg, adding diabetes mellitus, sodium132 mEq/l vs. 138 mEq/l, or increasing creatinine by0.6 mg/dl). Adding 1 to 4 of these risk changes for thispatient would change the SPRM from approximately64% to 57%, 51%, 44%, and 37%, respectively. Fromthe fitted curve in Figure 2, the associated ICD HR withthe previous SPRM values would be 0.51, 0.58, 0.65,0.74, and approximately 0.84, respectively. It is thecombination of several risk markers, rather than asingle risk marker, that is associated with an attenu-ation of the ICD benefit.

DISCUSSION

Our analysis has: 1) validated that the previouslyderived SPRM predicts the proportion of patients who

FIGURE 2 Observed Versus SPRM-Predicted Conditional Risk of Sudden Death

The Seattle Proportional Risk Model (SPRM) predicted versus observed proportion of

sudden death within the patients who died is shown for SRPM quartiles for patients with

and without ICDs. The diagonal black line is the line of identity.


295

will die of sudden death within HF-ACTION (Figure 2);and, more important, 2) demonstrated a highly sig-nificant interaction of benefit of an ICD with theSPRM on all-cause mortality (Figure 3A). There was a2.7-fold greater relative ICD benefit (63% vs. 23%) inthose with a higher conditional probability of suddendeath (SPRM quartile 4 vs. 1) within a cohort of largelyNYHA functional class II and III patients with an EFof #35% who were considered eligible to participatein an exercise training program. In the fitted Coxmodel, patients with a SPRM estimated #32% pro-portional risk of sudden death had no ICD benefit ontotal mortality. This is similar to the 31% residual riskof clinically adjudicated sudden death in those whohad an ICD present before death in this cohort.

Many prior analyses have suggested ICD inter-actions with many of the SPRM variables (5–8,17). TheSPRM is a method to allow integration of these mul-tiple variables into a single variable (11). Moreimportant, the ICD benefit varied with the SPRM withgreater ICD benefit in those with a higher predictedproportion of sudden death consistent with a priorMarkov model (10).

Prior research has shown the proportion of suddendeath varies based on the clinical scenario. The pro-portion of sudden death is approximately 50% in

chronic systolic heart failure (1), approximately 33%in the first year after a myocardial infarction (18), andapproximately 25% in patients with heart failure withpreserved EF (19). Many clinicians assume that, withsuch a high risk of sudden death, an ICD will abrogatethis increased risk of sudden death and reduce totalmortality. The DINAMIT (Defibrillator in AcuteMyocardial Infarction Trial) trial tested the effects ofICDs in patients after a myocardial infarction; ICDsdecreased sudden death by 67%. However, there wasa corresponding 70% increase in nonarrhythmicdeath and no benefit on total mortality (3).

In patients with chronic heart failure, the propor-tion of deaths attributable to sudden death varies byNYHA class: from approximately one-third in NYHAfunctional class IV to approximately two-thirds inNYHA functional class II patients (1). The proportionof clinically adjudicated sudden death due to atachyarrhythmic event is one-half of clinically adju-dicated sudden death in hospitalized heart failurepatients and as high two-thirds of clinically adjudi-cated sudden death in ambulatory heart failurepatients (1). Thus, the proportion of total mortalitydue to a tachyarrhythmic event that may be pre-vented by an ICD is approximately 17% (50% of the33% sudden death) in NYHA functional class IV

FIGURE 3 Effect of ICDs on All-Cause Mortality and Sudden Death Across SPRM Score Quartiles

A B

(A) The benefit associated with an implantable cardioverter defibrillator (ICD) on all-cause mortality. The points are the hazard ratio for the quartiles and the associated

95% confidence interval (CI). The ICD was associated with an overall mortality benefit of 37% (p ¼ 0.0002). There was greater benefit in the patients with a higher

predicted proportion of sudden death (interaction p value in the fitted Cox model ¼ 0.001). The overall benefit of the ICD on all-cause mortality is illustrated by

the dotted line (no interaction) and the interaction with the Seattle Proportional Risk Model (SPRM) in an adjusted Cox model is shown with the solid line. (B) The benefit

associated with an ICD on sudden death. The points are the hazard ratio for the quartiles and the associated 95% CI. The ICD was associated with an overall sudden death

benefit of 67% (hazard ratio: 0.33; p < 0.0001). There was a suggestion of potentially greater benefit in the patients with a higher predicted proportion of sudden

death (interaction p value in the fitted model with SPRM as a continuous variable ¼ 0.07). The overall benefit of the ICD on sudden death is illustrated by the dotted line

(no interaction) and the interaction with the SPRM is shown with the solid line.



296

patients and approximately 50% (approximately 67%of the approximately 67% of sudden deaths) inambulatory heart failure patients. These observationslikely explain why we saw an interaction of the SPRMwith both sudden death (p ¼ 0.07) and total mortality(p ¼ 0.001) with greater benefit in patients with ahigher proportional risk of sudden death.

In the average Medicare patient receiving an ICD,the mortality in the first year after an ICD is 13.5%(20). It is anticipated the average Medicare patientwill have a much lower proportion of deaths due tosudden death than seen in clinical trials, due to olderage and higher creatinine and the relative benefit ofan ICD is likely lower than was observed in MADIT II(Multicenter Automatic Defibrillator Implantation),SCD-HeFT, or observationally in HF-ACTION.

STUDY LIMITATIONS. There are numerous limita-tions to the current analysis. This post hoc analysiswas not a randomized trial of primary preventionICDs. The reason for ICD and CRT-D implantation(primary vs. secondary prevention) and ICD pro-gramming was not available. However, the magnitudeof ICD benefit within this ambulatory largely NYHAfunctional class II to III HF population with an EFof #35% (HR: 0.33 for sudden death; HR: 0.63 for all-cause mortality) is similar to the meta-analyses ofprimary prevention ICDs (HR: 0.40 for sudden death;HR: 0.73 all-cause mortality). The benefit of an ICDover the lifespan of the device and the patient lifespanmay be greater than observed during the 2.4 yoursof observation in HF-ACTION (Online Table 1). It ispossible that patients who were otherwise potentiallyeligible for a primary prevention ICD (essentially allpatients in HF-ACTION) had unmeasured differencesthat we could not account for in our risk adjustmentmodels. The benefit from a primary prevention ICDcan be better identified by applying the SPRM to ran-domized ICD trials. The SPRM should not be used fordecision making for secondary prevention ICDs.

The goal of a model like SPRM within a validationcohort is not to predict the absolute or relative risk ofsudden death, because that can vary with the defini-tion of sudden death used by an adjudication

committee. Rather, the goal is to demonstrate that amodel like SPRM can identify patients who will derivethe most all-cause mortality benefit from a primaryprevention ICD, as well as those who will derive nomeaningful mortality benefit (10).

CONCLUSIONS

Patients who have a higher SPRM proportional risk ofsudden death derive greater relative all-cause mor-tality benefit from an ICD than their counterpartswith a lower predicted risk. Patients with a <32%predicted proportion of total mortality due to suddendeath had no benefit of the ICD on total mortality.The SPRM may help to identify those patients whowill derive the greatest benefit from a primary pre-vention ICD and allow more appropriate use of thiseffective but expensive therapy.

ADDRESS FOR CORRESPONDENCE: Dr. Wayne C.Levy, Division of Cardiology, University of Washing-ton, Box 356422, 1959 NE Pacific Street, Seattle,Washington 98195. E-mail: [email protected].

RE F E RENCE S

1. Effect of metoprolol CR/XL in chronic heartfailure: metoprolol CR/XL randomised interventiontrial in congestive heart failure (MERIT-HF). Lan-cet 1999;353:2001–7.

2. Theuns DA, Smith T, Hunink MG, Bardy GH,Jordaens L. Effectiveness of prophylactic implanta-tion of cardioverter-defibrillators without cardiacresynchronization therapy in patients with ischaemic

or non-ischaemic heart disease: a systematic reviewand meta-analysis. Europace 2010;12:1564–70.

3. Hohnloser SH, Kuck KH, Dorian, et al. Prophylacticuse of an implantable cardioverter-defibrillator afteracute myocardial infarction. N Engl J Med 2004;351:2481–8.

4. Yancy CW, Jessup M, Bozkurt B, et al. 2013ACCF/AHA guideline for the management of heart

failure: a report of the American College of Cardi-ology Foundation/American Heart Association TaskForce on practice guidelines. J Am Coll Cardiol2013;62:e147–239.

5. Lee DS, Tu JV, Austin PC, et al. Effect of cardiacand noncardiac conditions on survival after defi-brillator implantation. J Am Coll Cardiol 2007;49:2408–15.

PERSPECTIVES

COMPETENCY IN MEDICAL KNOWLEDGE: PC5 (Request

and provides consultative care), MK2 (Knowledge of diagnostic

testing and procedures), and SBP3 (identifies factors that impact

the cost of health care, and advocates for, and practices

cost-effective care.

TRANSLATIONAL OUTLOOK: The SPRM is a method to

integrate common clinical variables into a risk model to predict

the proportion of sudden death and potentially to assist in

selection of patients who are most appropriate for primary

prevention ICDs. Observational ICD use in HF-ACTION suggests

patients with a greater predicted proportion of death due to

sudden death derive more benefit from an ICD. Other comorbid-

ities that may not be prevalent in clinical trial databases, such as

cancer, dementia, cirrhosis, are likely to decrease the proportion

of sudden death similar to the effect of diabetes mellitus.


297

6. Ghanbari H, Dalloul G, Hasan R, et al. Effec-tiveness of implantable cardioverter-defibrillatorsfor the primary prevention of sudden cardiacdeath in women with advanced heart failure: ameta-analysis of randomized controlled trials.Arch Intern Med 2009;169:1500–6.

7. Mezu U, Adelstein E, Jain S, Saba S. Effective-ness of implantable defibrillators in octogenariansand nonagenarians for primary prevention ofsudden cardiac death. Am J Cardiol 2011;108:718–22.

8. Williams ES, Shah SH, Piccini JP, et al. Pre-dictors of mortality in patients with chronic kidneydisease and an implantable defibrillator: an EPGENsubstudy. Europace 2011;13:1717–22.

9. Hammill SC, Stevenson LW, Kadish AH, et al.Review of the registry’s first year, data collected,and future plans. Heart Rhythm 2007;4:1260–3.

10. Owens DK, Sanders GD, Heidenreich PA,McDonaldKM,HlatkyMA. Effect of risk stratificationon cost-effectiveness of the implantable car-dioverter defibrillator. AmHeart J 2002;144:440–8.

11. Shadman R, Poole JE, Dardas TF, et al. A novelmethod to predict the proportional risk of suddencardiac death in heart failure: derivation of theSeattle Proportional Risk Model. Heart Rhythm2015;12:2069–77.

12. O’Connor CM, Whellan DJ, Lee KL, et al.,HF-ACTION Investigators. Efficacy and safety ofexercise training in patients with chronic heartfailure: HF-ACTION randomized controlled trial.JAMA 2009;301:1439–50.

13. Reed SD, Li Y, Dunlap ME, et al. In-hospitalresource use and medical costs in the last year oflife by mode of death (from the HF-ACTION ran-domized controlled trial). Am J Cardiol 2012;110:1150–5.

14. Levy WC, Mozaffarian D, Linker DT, et al. TheSeattle Heart Failure Model: prediction of survivalin heart failure. Circulation 2006;113:1424–33.

15. Levy WC, Mozaffarian D, Linker DT, et al.,COMET Investigators. Years-needed-to-treat toadd 1 year of life: A new metric to estimatetreatment effects in randomized trials. Eur J HeartFail 2009;11:256–63.

16. Klein JP, Andersen PK. Regression modeling ofcompeting risks data based on pseudovalues ofthe cumulative incidence function. Biometrics2005;61:223–9.

17. Bardy GH, Lee KL, Mark DB, et al. SuddenCardiac Death in Heart Failure Trial I. Amiodaroneor an implantable cardioverter-defibrillator forcongestive heart failure. N Engl J Med 2005;352:225–37.

18. Pouleur AC, Barkoudah E, Uno H, et al. Path-ogenesis of sudden unexpected death in a clinicaltrial of patients with myocardial infarction and leftventricular dysfunction, heart failure, or both.Circulation 2010;122:597–602.

19. Zile MR, Gaasch WH, Anand IS, et al.,I-Preserve Investigators. Mode of death in patientswith heart failure and a preserved ejection frac-tion: results from the irbesartan in heart failurewith preserved ejection fraction study (I-Preserve)trial. Circulation 2010;121:1393–405.

20. Al-Khatib SM, Greiner MA, Peterson ED,Hernandez AF, Schulman KA, Curtis LH. Patient andimplanting physician factors associated with mor-tality and complications after implantablecardioverter-defibrillator implantation, 2002-2005. Circ ArrhythmElectrophysiol 2008;1:240–9.

KEY WORDS heart failure, ICD, nonsuddendeath, prognosis, proportional risk,regression analysis, risk prediction model,sudden death

APPENDIX For supplemental figures andtables, please see the online version of thisarticle.



298

The Impact of First ProcedureSuccess Rate on the Economics ofAtrial Fibrillation AblationMoussa Mansour, MD,a Edward Karst, MS,b E. Kevin Heist, MD, PHD,a Nirav Dalal, MS, MBA,c

Jason H. Wasfy, MD, MPHIL,a Douglas L. Packer, MD,c Hugh Calkins, MD,d Jeremy N. Ruskin, MD,a

Srijoy Mahapatra, MDe

ABSTRACT

OBJECTIVES The purpose of this study was to compare health care costs associated with repeat ablation of atrial

fibrillation (AF) with health care costs associated with a successful first procedure.

BACKGROUND Catheter ablation has become established as a rhythm control strategy for symptomatic paroxysmal

and persistent AF. The economic impact of ablation is not completely understood, and it may be affected by repeat

procedures performed for recurrent AF.

METHODS The source of data was the MarketScan (Truven Health, Ann Arbor, Michigan) administrative claims dataset

from April 2008 to March 2013, including U.S. patients with private and Medicare supplemental insurance. Patients who

underwent an outpatient atrial ablation procedure and a diagnosis of AF were identified. Total health care cost was

calculated for 1 year before and after the ablation. Patients were categorized as having undergone a repeat ablation if an

additional ablation was performed in the following year.

RESULTS Of 12,027 patients included in the study, repeat ablation was performed in 2,066 (17.2%) within 1 year.

Patients with repeat ablation had higher rates of emergency department visits (43.4% vs. 32.2%; < 0.001) and sub-

sequent hospitalization (35.6% vs. 21.5%; p < 0.001), after excluding hospitalizations for the repeat procedure. Total

medical cost was higher for patients with repeat ablation ($52,821 vs. $13,412; p < 0.001), and it remained 46% higher

even after excluding the cost associated with additional ablations ($19,621 vs. $13,412; p < 0.001).

CONCLUSIONS Health care costs are significantly higher for patients with a repeat ablation for AF than for patients with

only a single ablation procedure, even though both groups have similar baseline characteristics. The increased costs persist

even after excluding the cost of the repeat ablation itself. These results emphasize the economic benefit of procedural

success in AF ablation. (J Am Coll Cardiol EP 2017;3:129–38) © 2017 by the American College of Cardiology Foundation.

From the aHeart Center, Massachusetts General Hospital, Boston, Massachusetts; bHealth Economics Outcomes Research, St. Jude

Medical, Sylmar, California; cDivision of Cardiovascular Diseases, Mayo Clinic, Rochester, Minnesota; dDivision of Cardiology,

Department of Medicine, Johns Hopkins Hospital, Baltimore, Maryland; and eMedical Affairs, St. Jude Medical, St. Paul, Min-

nesota. This work was partially supported by the Deane Institute for Integrative Research in Atrial Fibrillation and Stroke at

Massachusetts General Hospital; and by St. Jude Medical, Inc. Dr. Mansour has received consulting fees from Biosense Webster,

St. Jude Medical, Medtronic, Biotronik, and Sentreheart; and receives research support from Biosense Webster, Boston Scientific,

and St. Jude Medical. Mr. Karst is an employee of St. Jude Medical. Dr. Heist has received consulting fees from Boston Scientific,

Sanofi, Sorin, and St. Jude Medical; receives research support from Biotronik, Boston Scientific, and St. Jude Medical; and receives

honoraria from Biotronik, Boston Scientific, Medtronic, Sorin, and St. Jude Medical. Mr. Dalal is an employee of St. Jude Medical.

Dr. Wasfy receives honoraria from the New England Comparative Effectiveness Public Advisory Council. Dr. Packer has provided

consulting services, for which he received no personal compensation, for Abbott Laboratories/Topera, Aperture Diagnostics,

BiosenseWebster, Boston Scientific, CardioFocus, CardioInsight Technologies, Johnson & Johnson Healthcare Systems, Johnson &

Johnson, MediaSphere Medical, Medtronic, Siemens, and St. Jude Medical; receives research funding from the American Heart

Association Foundation Award, Biosense Webster, Boston Scientific/EPT, CardioInsight, CardioFocus, Endosense, Hansen

Medical, Medtronic CryoCath, National Institutes of Health, St. Jude Medical, Siemens, and Thermedical; and receives royalties

from St. Jude Medical. Dr. Calkins has received consulting fees from AtriCure, Boehringer Ingelheim, Daiichi Sankyo, Medtronic,

St. Jude Medical, and Tropera. Dr. Ruskin has received consulting fees from Advanced Medical Education, Astellas/Cardiome,

BiosenseWebster, Bristol-Myers Squibb, Medtronic, Pfizer, InfoBionic, Daiichi Sankyo, Gilead Sciences, and Portola Pharmaceuticals;




I S S N 2 4 0 5 - 5 0 0 X / $ 3 6 . 0 0


A trial fibrillation (AF), the most com-mon arrhythmia, represents a largeclinical and economic burden with

total health care costs of $26 billion annually(1–8). Catheter ablation has become estab-lished as a rhythm control strategy for the

arrhythmia (9) that relieves symptoms, improvesquality of life, and possibly reduces hospitalizationsand emergency department (ED) visits (10,11). Earlierstudies suggested that AF ablation is cost-effective(11–14), but the economic impact on the health caresystem is not fully understood. One reason is thatablation is not always successful in eliminating AF.The success rate of first-time ablation ranges from50% to 90%, thus often leading to a repeat procedure(15–18). The broad economic consequences of repeatablations have not been studied. It is possible thatthe impact of a repeat ablation extends even beyondthe cost of the additional procedure. Furthermore,limited hospital experience with AF ablation is associ-ated with higher readmission rates and proceduralcomplications, a finding suggesting a link betweenprocedural quality and downstream economic impact(13). Repeat ablation may be associated with addi-tional hospital stays and cardioversions even beforethe repeat ablation procedure. The aims of this studywere to compare health care costs associated withrepeat ablation with health care costs associatedwith a single ablation procedure and to investigatethe components responsible for the disparity.

METHODS

DATA SOURCE. This retrospective cohort study usedclaims data over a 5-year period (April 2008 to March2013) from MarketScan (Truven Health, Ann Arbor,Michigan). The dataset includes U.S. patients withprivate insurance or Medicare supplemental insur-ance, and it contains diagnosis and procedure codes,along with the amount paid, for inpatient hospitali-zations, outpatient services, and prescription drugs.

COHORT DERIVATION. Patients with a Current Pro-cedural Terminology (CPT) code of 93651 for atrialablation from 2009 to 2012 at age 18 years or older were

included. (Online Tables 1 to 3 provide the codesused.) Patients were required to have 12 months ofprevious insurance coverage andnoprevious inpatientor outpatient ablation in the available history. Addi-tional exclusion criteria were atrioventricular nodeablation, concomitant diagnosis of supraventriculartachycardia or Wolff-Parkinson-White syndrome, orno AF diagnosis on day of ablation. Patients were alsoexcluded if there was <1 year of continuous insurancecoverage after the index event.

Patients were categorized as having undergonerepeat ablation or single ablation. A patient wasplaced in the repeat-ablation group if an additionalcardiac ablation procedure, excluding primary atrio-ventricular node and ventricular ablation, had beenperformed within days 4 to 368 after the first abla-tion. The first 3 days were omitted to avoid countingmultipart procedures as repeat ablations. All otherpatients were included in the single-ablation group.Comorbid conditions, CHA2DS2-VASc score, andCharlson Comorbidity Index (19,20) were determinedat the time of first ablation by using diagnosis codes.Acute complications in the 30 days after ablationwere compared for the two groups.

OUTCOMES. Clinical outcomes were expressed bycalculating the number and percentage of patients inthe single-ablation and repeat-ablation groups withevents within 1 year before and after the index event.Event types included inpatient hospitalization, hos-pitalization not for ablation, ED visit, Holter moni-toring, electrical cardioversion, and prescription forantiarrhythmic drugs (AADs). Total medical costfrom the payer’s perspective was evaluated by addingpayments from all sources for all-cause inpatient hos-pitalizations and outpatient medical services. Cumu-lative daily cost accumulation was evaluated starting1 year before the index event, excluding the day of theindex event and the following 3 days, and continuingfrom days 4 to 368 after the index event. Costs asso-ciated with the index event were removed by omittingthe 3 days after the procedure. For patients with 1 ormore repeat ablations, total nonablation cost wascalculated by excluding all inpatient hospitalizationswith an atrial ablation and all costs incurred on the day


AND ACRONYMS

AAD = antiarrhythmic drug

AF = atrial fibrillation

ED = emergency department

receives fellowship support from Biosense Webster, Boston Scientific, Medtronic, and St. Jude Medical; serves on the scientific

advisory board of CardioInsight and InfoBionic; serves on a scientific steering committee for Pfizer; and holds equity in InfoBionic

and Portola Pharmaceuticals. Dr. Mahapatra is an employee of and has stock options in St. Jude Medical. Francis Marchlinski, MD,

served as Guest Editor for this paper.




Manuscript received March 28, 2016; revised manuscript received June 6, 2016, accepted June 13, 2016.

Mansour et al. J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 2 , 2 0 1 7

Procedure Success and Economics of Atrial Fibrillation Ablation F E B R U A R Y 2 0 1 7 : 1 2 9 – 3 8

130

of any outpatient ablations. In a subgroup of patientswith 2 years of enrollment before and after the indexevent, cumulative costs were evaluated over the2 years before and 2 years after the ablation. As asensitivity analysis, medical cost per patient-year inthe subsequent yearwas compared for the entire groupof patients who had an AF ablation, including thosewho died or who changed insurance within 1 year afterthe index event.

MODELING THE IMPACT OF REPEAT ABLATION. Alinear model was developed to show mean health carecost in the year after ablation as a function of repeatablation rate. The model was projected to a nationalestimate of 75,000 ablations for atrial fibrillationperformed in the United States per year.

STATISTICAL ANALYSIS. Baseline patient character-istics of the single-ablation and repeat-ablation groups

were summarized using frequency and percentage forcategorical variables and mean � SD for continuousvariables. To test for differences in variables betweenthe two groups, chi-square tests and unpaired Studentt tests, respectively, were used. Cumulative incidenceof repeat atrial ablation was calculated with theKaplan-Meier method. The proportion of patientsexperiencing clinical outcomes in the year before andafter ablation was plotted with Clopper-Pearson con-fidence intervals and compared between the single-ablation and repeat-ablation groups by using theFisher exact test. After confirming with the Shapiro-Wilk statistic that health care costs did not havenormal distribution, we employed nonparametricstatistics to summarize costs. Comparisons in the yearbefore and after ablation used a Wilcoxon matched-pair signed rank test. Health care costs are summa-rized using box-and-whisker plots with median,

FIGURE 1 Consort Diagram

Of 83,792 patients with an outpatient atrial fibrillation (AF) ablation in the 4-year period, a first outpatient ablation was identified in 12,027

patients with 1 year of pre- and post-procedure enrollment. There were 2,066 patients (17.2%) with a repeat ablation in the following year.

CPT ¼ Current Procedural Terminology; SVT ¼ supraventricular tachycardia; WPW ¼ Wolff-Parkinson-While syndrome.

J A C C : C L I N I C A L E L E C T R O P H Y S I O L O G Y V O L . 3 , N O . 2 , 2 0 1 7 Mansour et al.F E B R U A R Y 2 0 1 7 : 1 2 9 – 3 8 Procedure Success and Economics of Atrial Fibrillation Ablation

131

quartiles, and extreme values, along with an annota-tion for trimmed means excluding the top and bottom1%. To compare cost accumulation up to 2 years beforeand after the index event in patients with a single orrepeat ablation, a Wilcoxon signed rank test was usedin each separate year. A multivariable logistic regres-sion model compared odds for having nonablationcosts higher than the median value. Covariatesincluded patients’ characteristics and clinical events inthe year before the index event. Level of significancewas 0.05, and Bonferroni correction was used in mul-tiple comparisons for yearly cost accumulation. SASversion 9.3 (SAS Institute, Cary, North Carolina) wasused for statistical analysis.

RESULTS

Of the 83,792 patients in MarketScan who underwentan outpatient ablation procedure from 2009 to 2012,12,027 met all inclusion and exclusion criteria(Figure 1). A repeat ablation was performed within 1year in 2,066 patients (17.2%); the remaining 9,961patients had a single ablation. There were 2 or 3repeat ablations in 221 patients (1.8% of all patients).The cumulative risk of repeat ablation reached 28.4%of patients at 3 years after the index event.

Patients who underwent a repeat ablation in thefirst year were slightly younger (59.4 � 9.4 years vs.60.8 � 11.1 years; p < 0.001), more likely to haveprivate insurance (73% vs. 66%; p < 0.001), and lesslikely to have ischemic heart disease (40% vs. 43%;p < 0.001) than patients with a single ablation(Table 1). There was no significant difference by sexor prevalence of hypertension, diabetes, or heartfailure. The CHA2DS2-VASc score and CharlsonComorbidity Index were both lower in the repeat-ablation group, although the magnitude of differ-ences was small.

Complications occurred in 6.6% (661 of 9,961) ofpatients with a single ablation and in 6.6% (136 of2,066) of patients with a repeat ablation (p ¼ 0.79).There were no significant differences in rate ofvascular complication, perforation or tamponade, orstroke or transient ischemic attack. Pneumothorax orhemothorax occurred in 0.3% of patients (25 of 9,961)with a single ablation and in 0.1% (3 of 2,066) of pa-tients with a repeat ablation (p ¼ 0.006).

HEALTH CARE USE. Figure 2 shows the percentageof patients who had hospitalizations, ED visits,Holter monitoring, cardioversion, and AAD pre-scription in the groups with a single ablation or arepeat ablation 1 year before and 1 year after abla-tion. Although the study design ensured no

hospitalizations for ablation in the year before theindex event, 40.7% of the patients in the single-ablation group and 40.9% of the patients in therepeat-ablation group were hospitalized for anotherreason (p ¼ 0.89). In the year after the index event,patients with repeat ablation had higher hospitali-zation rates for ablation (26.1% vs. 0.0%; p < 0.001)and for all other reasons (35.6% vs. 21.5%; p < 0.001).ED visits, Holter monitoring, electrical cardioversion,and prescription for AADs were all higher for therepeat-ablation group in the year after ablation. Therewere also statistically significant differences, butwith smaller magnitude, in clinical events in the yearbefore the index event.

Median medical cost was down 47% from $11,528 inthe year before ablation to $6,102 the following yearfor patients with a single ablation, but it increasedfrom $11,657 to $42,898 for patients with a repeatablation (p < 0.001 for matched-pair test of before-and-after difference) (Table 2). Trimmed mean costwas $18,036 before and $13,412 after ablation whenthere was a single procedure, compared with $18,320before and $52,821 after for patients with repeatablation. More interestingly, even after the cost ofablation procedures was removed, median medicalcost remained 88% higher in the repeat-ablationgroup ($11,456 vs. $6,102, p < 0.001), and mean costwas higher by 46% ($19,621 vs. $13,412). There weresignificant increases in both inpatient and outpatientcosts, even after excluding repeat ablations.

Mean total cost accumulation for patients with asingle or repeat ablation is shown in Figure 3. Thevalue of each point on the curve shows mean accu-mulated cost, whereas the slope indicates the rate ofincurring costs. Although costs overlapped in the yearbefore the index event ($1,683/month for single

TABLE 1 Patients’ Characteristics

Single Ablation(n ¼ 9,961)

Repeat Ablation(n ¼ 2,066) p Value

Age, yrs 60.8 � 11.1 59.4 � 9.4 <0.001

Private insurance 6,620 (66%) 1,516 (73%) <0.001

Medicare 3,341 (34%) 550 (27%)

Female 2,833 (28%) 563 (27%) 0.28

Male 7,128 (72%) 1,503 (73%)

Hypertension 7,001 (70%) 1,415 (68%) 0.11

Ischemic heart disease 4,279 (43%) 818 (40%) 0.005

Diabetes 2,269 (23%) 444 (21%) 0.21

Heart failure 2,073 (21%) 403 (20%) 0.18

CHA2DS2-VASc 1.66 � 1.40 1.56 � 1.34 0.005

Charlson ComorbidityIndex

1.68 � 1.98 1.52 � 1.76 <0.001

Values are mean � SD or n (%). The p values are from matched-pair test, whichcompares the change from year before to year after for single vs. repeat ablation.



132

ablation vs. $1,679/month for repeat ablation;p ¼ 0.39), a significant divergence occurred after theindex event ($1,151/month for single ablation and$4,428 for repeat ablation; p < 0.001). The slope ofthe curve for patients with a repeat ablation waselevated throughout the year after ablation,

compared with patients with a single ablation. For thesubgroup of 2,494 patients with 2 years before andafter the index event, elevated costs persisted in therepeat-ablation group beyond the year of the repeatablation. There were no significant differences ineither year before the index event.

FIGURE 2 Clinical Outcomes Before and After AF Ablation

The percentage of patients undergoing single and repeat atrial fibrillation (AF) ablations with events in the year before (lighter shading) and

after (darker shading) the initial ablation. ED ¼ emergency department.


133

Undergoing a repeat ablation was associated withhigher odds of having nonablation medical cost abovethe median (odds ratio: 2.64; p < 0.001), as shown inTable 3. Significant but weaker associations were alsoseen with covariates of patient-related characteristicsand previous clinical events, except AAD prescrip-tion. The association of repeat ablation with higherthan average nonablation medical cost persisted inthe multivariable regression.

There were 9,201 patients with AF ablation wholacked 1 year of continuous insurance coverage afterthe index event and were excluded from the studycohort. Considering all 21,228 patients with an AFablation, median expenditures after ablation wentfrom $11,855 before ablation to $47,689 per patient-year in the 2,854 (13.4%) patients who underwent arepeat ablation in the available follow-up period, butthey went from $11,881 to $6,056 per patient-year forthe remainder with a single ablation.

ECONOMIC MODEL OF IMPACT OF REPEAT ABLATION.

The basis of the economic model for cost impact ofrepeat ablation was the finding that the mean medicalcost in the year following ablation was $52,821 for the17.2% of patients having repeat ablation and $13,412for the remainder of patients having a single proce-dure. Therefore, incremental cost associated withhaving a repeat procedure was $39,409. Overall, themedical cost per patient is $20,182 in the year after firstablation procedure. Using a linear model of averagecost of care, an improvement of 1% in the rate of repeat

ablations corresponds to an expected savings of $394for each patient undergoing ablation. Projected to anational average of 75,000 AF ablations per year in theUnited States, the total savings to the health caresystem for each 1% decrease in rate of repeat ablationswould amount to $29.6 million per year.

DISCUSSION

The major finding of this study is that repeat ablationis associated with greatly increased medical cost. Theexcess burden on the health care system extends overand above the cost of the repeat procedure itself.Additionally, this study confirmed that repeat abla-tion is common, occurring in 17.2% of the patients at1 year and 28.4% at 3 years. A reduction of just 1% inthe rate of repeat procedures can result in cost sav-ings of nearly $30 million per year to the U.S. healthcare system.

AF is associated with a substantial economicburden, contributing nearly half a million hospitali-zations each year and adding billions of dollars indirect and indirect costs to the U.S. health care system(4–8). Data from randomized studies show that cath-eter ablation is more effective at eliminating AFrecurrence than are AADs (9,21,22). Catheter ablationis not always effective, and the greatest benefit of thisintervention occurs when it is successful at elimi-nating AF. Imperfect treatment success may be onereason for the mixed findings on cost effectiveness

TABLE 2 Table Showing Median (Q1-Q3) and Trimmed Mean for Health Care Cost and Use Before and After Ablation*

Single Ablation (n ¼ 9,961) Repeat Ablation (n ¼ 2,066)

p ValueYear Before Year After Year Before Year After

Total medical cost $11,528 (5,662-22,039)$18,036

$6,102 (2,633-14,872)$13,412

$11,657 (5,944-23,295)$18,320

$42,898 (26,495-69,704)$52,821

<0.001

Total medical cost with AF diagnosis $3,870 (905-10,547)$8,216

$964 (329-3,267)$4,057

$4,986 (1,612-11,821)$9,277

$26,154 (8,438-50,475)$34,450

<0.001

Cost of inpatient hospitalizations $0 (0-7,666)$6,400

$0 (0-0)$3,946

$0 (0-7,765)$6,527

$3,677 (0–24,328)$16,823

<0.001

Cost of outpatient services $7,812 (4,287-14,017)$11,016

$5,325 (2,505-10,878)$8,770

$8,209 (4,580-14,140)$11,222

$26,352 (1,759-37,132)$35,070

<0.001

Number of visits 25 (16-37)28.3

24 (13-37)27.4

27 (18-39)29.7

33 (24-47)36.5

<0.001

Visits with AF diagnosis 7 (3-13)9.0

6 (3-11)7.8

9 (5-15)10.7

13 (8-21)15.1

<0.001

Total pharmaceutical cost† $2,357 (909-43,539)$3,155

$2,252 (801-4,477)$3,143

$2,624 (1,129-4,552)$3,254

$3,181 (1,129-4,552)$3,838

<0.001

AAD cost† $65 (0-547)$465

$0 (0-317)$356

$168 (0-772)$572

$195 (0-997)$650

<0.001

Number of prescriptions† 31 (18-49)36.0

32 (17-50)36.4

32 (20-49)36.9

38 (24-56)42.7

<0.001

Values are median (interquartile range [Q1 to Q3]), and trimmed mean. *The p values are from a matched-pair test, which compares the change from year before to year afterfor single versus repeat ablation. †On the basis of 8,242 patients with single ablation and 1,676 with repeat ablation with accompanying drug coverage information.

AAD ¼ antiarrhythmic drug; AF ¼ atrial fibrillation.



134

of catheter ablation compared with AADs or ratecontrol (23–27).

RATE OF REPEAT ABLATION. Previous studiesdemonstrated that repeat ablation is needed in manypatients (12,15,28,29). The rate of repeat ablation de-pends on many factors, including the type of AF,baseline comorbidities, the operator’s skill, and thetype of the technology and technique used for theintervention. Even beyond the procedure itself,modifiable risk factors have become an enticing targetfor treatment, if behavioral changes and comorbiditymanagement can be achieved (30). The current studydemonstrated that in the general population in theUnited States, the risk of repeat ablation is 17.2% at1 year and 28.4% at 3 years. These percentages aresubstantial, highlight the magnitude of the problem,and present an opportunity for maximizing the suc-cess rate of ablation for AF. Although rates of repeatprocedures have been reported in Medicare patientsand for inpatient hospitalizations (13,14), this reportdescribes the consequences of repeat proceduresand includes broad private-insurance and MedicareU.S. populations.

INCREASED COST ASSOCIATED WITH REPEAT

ABLATION. Outcomes and cost implications of AFablation have been previously reported using claimsdata (11–14). However, previous reports did notconsider the incremental cost of repeat ablationacross the population, including patients with Medi-care and private insurance. The current study dem-onstrates that repeat ablation results in a dramaticincrement in health care use. Although the compli-cation rate is identical, mean medical cost in the yearfollowing the first ablation is $13,412 if the patient hasonly a single ablation procedure, compared with$52,821 for patients undergoing a second ablation.Higher medical cost in patients needing a secondablation is not unexpected, but it is notable that theincreased cost is not only the result of expensesrelated to the repeat procedure. After excluding ex-penditures for ablation, the average medical cost inthe year following the first ablation is $13,412 in pa-tients with single ablation compared with $19,621 forpatients needing a second ablation. The cost increaseof 46% results from elevated use of different com-ponents of the health care system, including hospi-talization, ED visits, electrical cardioversion,ambulatory Holter monitoring, AAD prescriptions,and any consequences of the additional procedure.A sensitivity analysis including patients without acomplete year of follow-up demonstrates costs thatare similar to those in the main findings on thestudy cohort.

The accumulated medical cost plots in Figure 3reveal important information. The cost profile in theyear before the ablation was similar for the single-ablation and repeat-ablation groups. This finding ar-gues against inherent patient-related characteristicsthat could have led to the difference in cost after theprocedure. Covariate adjustment in the multivariablelogistic regression in Table 3 did not reduce the as-sociation of repeat ablation with above-median non-ablation medical cost. In both groups, there was anincrease in cost in the months leading to ablation, asillustrated in the slope of the curves. This finding may

FIGURE 3 Accumulated Medical Costs Before and After AF Ablation

The upper panel shows cost accumulation in U.S. dollars (USD) in all 12,027 patients in the

year before and year after atrial fibrillation (AF) ablation, with red indicating patients

with repeat ablation and green indicating those with a single procedure. The lower panel

shows cost accumulation in 2,494 patients (441 with repeat, 2,053 with single ablation)

with 2 years before and after ablation. Level of significance with Bonferroni correction is

0.025 for the upper panel and 0.0125 for the lower panel. Differences in each cost

comparison after ablation, but none before ablation, are significant.


135

reflect increased symptom severity leading toincreased health care use and ultimately to the deci-sion to undergo ablation for AF. However, afterablation the cost increment moderates in patientswith a single procedure, in contrast to the explosiveincrease in patients with repeat ablation. The accel-erated increment in cost in the repeat-ablation grouptapers off after 1 year after the first ablation. Thisreduction in cost acceleration may possibly beexplained by improved rhythm control obtained withsubsequent procedures.

The number of ablations for AF performed in theUnited States is approaching 75,000 per year. Repeatablation is performed in many of those patients. Pa-tients with an additional ablation were found in thisstudy to have $39,409more in costs the following year,a finding suggesting that a mere 1% reduction in therate of repeat ablation could save the U.S. health caresystem nearly $30 million. This presents an importantopportunity at a time of massive health care changes,many driven by the need for cost containment.

Most recurrences of AF after ablation result fromgap formation along ablation lines that leads toreconnection of the pulmonary veins (31). Multipletools and techniques have been used to improvecatheter stability with the intention of reducing thechance of AF recurrence after the first ablation. Thesetechniques include the use of high-frequency jetventilation (32), contact force sensing on the ablationcatheter (33), the use of steerable sheaths (34), andinfusion of adenosine. The use of some of these tools

has been limited in some situations because of theadditional cost they may incur. This study suggeststhat the use of tools to improve lesion formation byskilled operators could potentially improve theoverall cost effectiveness of AF ablation, in additionto the obvious gain of alleviating symptomatic AF.

CLINICAL IMPLICATIONS. In a time of rising healthcare costs, there is a focus on cost-effective care.Although AF ablation is cost-effective in selectedpatients, this study shows that medical costs for pa-tients undergoing repeat ablation are substantiallyhigher than for patients undergoing only a singleprocedure within a year. Because measurable clinicalcharacteristics and pre-procedural costs appear to besimilar in the two groups of this study, it may be thattechnical success depends on procedural quality inaddition to underlying patient-related characteristics.Efforts aimed at reducing the rate of repeat pro-cedures are important to both patient and provider.These efforts provide a clinical benefit to the patientand are associated with a reduction in direct medicalcosts. There may be even further consequences ofarrhythmia recurrence on indirect medical costs, suchas lost productivity and consequences of early death.

STUDY LIMITATIONS. The number of covered lives inthe MarketScan research databases amounts to nearly20% of the U.S. population. However, there are limi-tations to using administrative claims. In absence ofclinical data, it is not possible to compare left ven-tricular ejection fraction, New York Heart Associationfunctional class, or type and severity of AF in thesingle-ablation and repeat-ablation groups. Market-Scan contains death information only if deathoccurred in a hospital. Censoring from the dataset isnoninformative because it can occur whether a pa-tient died or changed insurance. Many patients withan ablation lacked 1 year of follow-up data and wereexcluded. Although MarketScan is a nationwidesample, it is weighted toward persons working forlarge employers and their dependents. Medicare pa-tients without supplemental insurance, Medicaidpatients, and patients lacking any insurance are ab-sent from the dataset.

Additionally, it cannot be determined whetherpatients were treated according to guidelines or withoptimal ablation techniques. At least 1 year ofcontinuous enrollment was required before the indexablation, but some patients may have had previousablations before the study period. The single-ablationgroup included patients with recurrent arrhythmiatreated with medications, as well as those whochose not to undergo a repeat procedure in thefollowing year.

TABLE 3 Predictors of Having Medical Cost Above the Median in the Year After Ablation,

Excluding Cost of Any Repeat Ablations

Univariate LogisticRegression

Multivariable LogisticRegression

Odds Ratio(95% CI) p Value

Odds Ratio(95% CI) p Value

Repeat ablation 2.64 (2.38-2.92) <0.001 3.01 (2.70-3.34) <0.001

Age (per yr) 1.03 (1.02-1.03) <0.001 1.02 (1.01-1.02) <0.001

Medicare insurance 1.51 (1.39-1.63) <0.001 0.89 (0.78-1.03) 0.11

Male 0.72 (0.66-0.78) <0.001 0.77 (0.70-0.86) <0.001

Hypertension 1.65 (1.52-1.78) <0.001 1.17 (1.05-1.29) 0.004

Ischemic heart disease 1.64 (1.52-1.76) <0.001 1.14 (1.05-1.24) 0.002

Diabetes mellitus 1.77 (1.62-1.93) <0.001 0.97 (0.87-1.09) 0.66

Heart failure 2.03 (1.85-2.23) <0.001 1.17 (1.04-1.32) 0.008

CHA2DS2-VASc (per point) 1.26 (1.23-1.29) <0.001 0.99 (0.94-1.04) 0.62

Charlson Comorbidity Index 1.33 (1.30-1.36) <0.001 1.25 (1.21-1.29) <0.001

ED visit in previous year 1.53 (1.42-1.64) <0.001 1.30 (1.20-1.41) <0.001

Hospitalization in previous year 1.71 (1.59-1.84) <0.001 1.24 (1.14-1.34) <0.001

Cardioversion in previous year 1.29 (1.20-1.39) <0.001 1.15 (1.06-1.25) <0.001

Holter monitoring in previous year 1.12 (1.04-1.20) 0.004 1.13 (1.04-1.22) 0.003

AAD prescription in previous year 1.02 (0.95-1.10) 0.54 0.98 (0.91-1.06) 0.66

CI ¼ confidence interval; ED ¼ emergency department.



136

CONCLUSIONS

Repeat ablation is associated with higher costs,over and above any subsequent procedural costs,compared with a single ablation procedure for AF.Because patient-related characteristics and healthcare costs are similar before initial ablation, theresults suggest that factors beyond baselinecharacteristics may have an impact on ablationsuccess. These results emphasize the importance ofprocedural quality in AF ablation not only forpatients’ outcomes but also for subsequent healthcare costs.

ADDRESS FOR CORRESPONDENCE: Dr. MoussaMansour, Cardiac Arrhythmia Unit, Heart Center,Massachusetts General Hospital, 55 Fruit Street, Bos-ton, Massachusetts 02114. E-mail: [email protected].

RE F E RENCE S

1. Miyasaka Y, Barnes ME, Gersh BJ, et al. Seculartrends in incidence of atrial fibrillation in OlmstedCounty, Minnesota, 1980 to 2000, and implica-tions on the projections for future prevalence.Circulation 2006;114:119–25.

2. Naccarelli GV, Varker H, Lin J, Schulman KL.Increasing prevalence of atrial fibrillation andflutter in the Unites States. Am J Cardiol 2009;104:1534–9.

3. Colilla S, Crow A, Petkun W, Singer DE,Simon T, Liu X. Estimates of current and futureincidence and prevalence of atrial fibrillation inthe U.S. adult population. Am J Cardiol 2013;112:1142–7.

4. Healey JS, Connolly SJ, Gold MR, et al. Sub-clinical atrial fibrillation and the risk of stroke.N Engl J Med 2012;366:120–9.

5. Miyasaka Y, Barnes ME, Bailey KR, et al. Mor-tality trends in patients diagnosed with first atrialfibrillation: a 21-year community-based study.J Am Coll Cardiol 2007;49:986–92.

6. Mamas MA, Caldwell JC, Chacko S, Garratt CJ,Fath-Ordoubadi F, Neyses L. A meta-analysis ofthe prognostic significance of atrial fibrillation inchronic heart failure. Eur J Heart Fail 2009;11:676–83.

7. Go AS, Mozaffarian D, Roger VL, et al. Heartdisease and stroke statistics—2014 update: areport from the American Heart Association. Cir-culation 2014;129:e28–292.

8. Kim MH, Johnston SS, Chu BC, Dalal MR,Schulman KL. Estimation of total incrementalhealthcare costs in patients with atrial fibrillationin the United States. Circ Cardiovasc Qual Out-comes 2011;4:313–20.

9. January CT, Wann LS, Alpert JS, et al. 2014AHA/ACC/HRS guideline for the management ofpatients with atrial fibrillation: a report of theAmerican College of Cardiology/American Heart

Association Task Force on Practice Guidelines andthe Heart Rhythm Society. J Am Coll Cardiol 2014;64:e1–76.

10. Wokhlu A, Monahan KH, Hodge DO, et al.Long-term quality of life after ablation of atrialfibrillation: the impact of recurrence, symptomrelief, and placebo effect. J Am Coll Cardiol 2010;55:2308–16.

11. Ladapo JA, David G, Gunnarsson CL, et al.Healthcare utilization and expenditures in pa-tients with atrial fibrillation treated with cath-eter ablation. J Cardiovasc Electrophysiol 2012;23:1–8.

12. Kim MH, Lin J, Kreilick C, Foltz Boklage SH.Total costs and atrial fibrillation ablation successor failure in Medicare-aged patients in the UnitedStates. Adv Ther 2010;27:600–12.

13. Shah RU, Freeman JV, Shilane D, Wang PJ,Go AS, Hlatky MA. Procedural complications,rehospitalizations, and repeat procedures aftercatheter ablation for atrial fibrillation. J Am CollCardiol 2012;59:143–9.

14. Piccini JP, Sinner MF, Greiner MA, et al. Out-comes of Medicare beneficiaries undergoingcatheter ablation for atrial fibrillation. Circulation2012;126:2200–7.

15. Weerasooriya R, Khairy P, Litalien J, et al.Catheter ablation for atrial fibrillation: are resultsmaintained at 5 years of follow-up? J Am CollCardiol 2011;57:160–6.

16. Hussein AA, Saliba WI, Martin DO, et al. Nat-ural history and long-term outcomes of ablatedatrial fibrillation. Circ Arrhythm Electrophysiol2011;4:271–8.

17. Ganesan AN, Shipp NJ, Brooks AG, et al. Long-term outcomes of catheter ablation of atrialfibrillation: a systematic review and meta-analysis.J Am Heart Assoc 2013;2:e004549.

18. Wazni O, Wilkoff B, Saliba W. Catheter ablationfor atrial fibrillation. N Engl J Med 2011;365:2296–304.

19. Charlson ME, Pompei P, Ales KL,MacKenzie CR. A new method of classifyingprognostic comorbidity in longitudinal studies:development and validation. J Chronic Dis 1987;40:373–83.

20. Quan H, Sundararajan V, Halfon P, et al. Cod-ing algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care2005;43:1130–9.

21. Nielsen JC, Johannessen A, Raatikainen P, et al.Radiofrequency ablation as initial therapy inparoxysmal atrial fibrillation. N Engl J Med 2012;367:1587–95.

22. Scherr D, Khairy P, Miyazaki S, et al. Five-yearoutcome of catheter ablation of persistent atrialfibrillation using termination of atrial fibrillation asa procedural endpoint. Circ Arrhythm Electro-physiol 2015;8:18–24.

23. Chan PS, Vijan S, Morady F, Oral H. Cost-effectiveness of radiofrequency catheter ablationfor atrial fibrillation. J Am Coll Cardiol 2006;47:2513–20.

24. Eckard N, Davidson T, Walfridsson H, Levin L.Cost-effectiveness of catheter ablation treatmentfor patients with symptomatic atrial fibrillation.J Atr Fibrillation 2009;1:461–70.

25. McKenna C, Palmer S, Rodgers M, et al. Cost-effectiveness of radiofrequency catheter ablationfor the treatment of atrial fibrillation in the UnitedKingdom. Heart 2009;95:542–9.

26. Reynolds MR, Zimetbaum P, Josephson ME,Ellis E, Danilov T, Cohen DJ. Cost-effectiveness ofradiofrequency catheter ablation compared with anti-arrhythmic drug therapy for paroxysmal atrial fibrilla-tion. Circ Arrhythm Electrophysiol 2009;2:362–9.

PERSPECTIVES

COMPETENCY IN MEDICAL KNOWLEDGE: Repeat ablation

for AF is associated with increased hospitalizations, ED visits, Hol-

ter monitoring, electrical cardioversion, and AAD use, compared

with single ablation. These increases result in higher costs, over and

above any subsequent procedural costs of repeat ablation.

TRANSLATIONAL OUTLOOK: To the extent that the proce-

dural success of AF ablation can be affected by tools and oper-

ators’ skill, improvements in the success rate enhance the cost

effectiveness of ablation and make it increasingly important as

an option for patients with AF. Future investigations should

determine whether there are identifiable clinical factors in

patients who go on to require repeat ablation, both to improve

selection of patients and to help determine which tools are

needed to improve outcomes of the procedure.


137

27. Blackhouse G, Assasi N, Xie F, et al. Cost-effectiveness of catheter ablation for rhythmcontrol of atrial fibrillation. Int J Vasc Med 2013;2013:262809.

28. Haines DE. Atrial fibrillation ablation in thereal world. J Am Coll Cardiol 2012;59:150–2.

29. Morillo CA, Verma A, Connolly SJ, et al.Radiofrequency ablation vs. antiarrhythmic drugsas first-line treatment of paroxysmal atrial fibril-lation (RAAFT-2): a randomized trial. JAMA 2014;311:692–700.

30. Parthak RK, Middeldorp ME, Lau DH, et al.Aggressive risk factor reduction study for atrialfibrillation and implications for the outcome of

ablation: the ARREST-AF cohort study. J Am CollCardiol 2014;64:2222–31.

31. Kowalski M, Grimes MM, Perez FJ, et al. His-topathologic characterization of chronic radio-frequency ablation lesions for pulmonary veinisolation. J Am Coll Cardiol 2012;59:930–8.

32. Williams JL, Valencia V, Lugg D, et al. Highfrequency jet ventilation during ablation of sup-raventricular and ventricular arrhythmias. J InnovCardiac Rhythm Manag 2011;2:528–35.

33. Reddy VY, Dukkipati SR, Neuzil P, et al. Ran-domized, controlled trial of the safety and effec-tiveness of a contact force-sensing irrigatedcatheter for ablation of paroxysmal atrial

fibrillation: results of the TactiCath contact forceablation catheter study for atrial fibrillation(TOCCASTAR) study. Circulation 2015;132:907–15.

34. Hutchinson MD, Garcia FC, Mandel JE, et al.Efforts to enhance catheter stability improve atrialfibrillation ablation outcome. Heart Rhythm 2013;10:347–53.

KEY WORDS atrial fibrillation, catheterablation, outcomes research

APPENDIX For supplemental tables, pleasesee the online version of this article.



138

JACC.org/Submit©2018, American College of Cardiology I18101

Editor-in-Chief

Podcasts

Expedited publication

for high-impact findings

Central illustrations

summarizing key findings

International editorial boards,

readers and authors

Newly redesigned,

mobile-friendly website

Top-ranked in Impact

Factor

Five specialty journals

Get the Broadest Reach for Your Practice-Changing Research

Publish in

Follow us

@JACCJournals

JACCJournals

best of jacc: clinical electrophysiology · need for pericardiocentesi s. post-procedurally, dapt...

Documents