Author + information
- Received April 13, 2015
- Revision received July 14, 2015
- Accepted July 16, 2015
- Published online October 1, 2015.
- David S. Frankel, MD∗ (, )
- Jackson J. Liang, DO,
- Gregory Supple, MD,
- Sanjay Dixit, MD,
- Mathew D. Hutchinson, MD,
- Melissa A. Elafros, PhD,
- David J. Callans, MD and
- Francis E. Marchlinski, MD
- Cardiovascular Division, Electrophysiology Section, Hospital of the University of Pennsylvania, Philadelphia, Pennsylvania
- ↵∗Reprint requests and correspondence:
Dr. David S. Frankel, Cardiovascular Division, Electrophysiology Section, Hospital of the University of Pennsylvania, 9 Founders Pavilion, 3400 Spruce Street, Philadelphia, Pennsylvania 19104.
Objectives This study sought to identify predictors of transplantation/left ventricular assist device (LVAD)-free survival among patients with left ventricular nonischemic cardiomyopathy (NICM) and ventricular tachycardia (VT).
Background Outcomes vary widely among these patients.
Methods The derivation cohort consisted of patients with NICM undergoing VT ablation from 2007 to 2011. Scar percentage was defined as the area of low voltage divided by total surface area. Cox proportional hazard modeling was performed to identify predictors of shorter time to the primary endpoint of death, transplantation, or LVAD. A risk score was created using β regression coefficients. The risk score was then validated in a separate cohort of patients undergoing ablation from 2004 to 2007.
Results Of 100 patients with NICM undergoing VT ablation, 41 experienced an endpoint during 1.2 years mean follow-up. In multivariate modeling, VT storm, wider native QRS duration, greater endocardial/epicardial bipolar scar percentage, and lower left ventricular ejection fraction (LVEF) identified worse transplantation/LVAD-free survival. The risk score = (VT storm × 83) + (greater of endocardial/epicardial bipolar scar percentage × 4) + (QRS duration) − (LVEF × 3). A score >180 identified patients at high risk for the endpoint, whereas a score <100 identified low risk. Among the 50-patient validation cohort, the high-risk group again had worse transplantation/LVAD-free survival during a mean 3.0 years of follow-up (<10% vs. 50% for intermediate and >80% for low, p < 0.001).
Conclusions Wider native QRS duration, greater bipolar scar percentage, VT storm, and lower LVEF are potent predictors of time to death, transplantation, or LVAD. By combining these variables into an “electrophysiological risk score,” patient risk can be refined.
The incidence of death, heart transplantation, or left ventricular assist device (LVAD) implantation among patients with nonischemic cardiomyopathy (NICM) presenting with ventricular tachycardia (VT) is substantial (1,2). The ability to better risk stratify these patients beyond left ventricular ejection fraction (LVEF) could provide valuable prognostic information and potentially guide more intensive heart failure therapies.
Greater extent of myocardial delayed enhancement on cardiac magnetic resonance imaging (MRI) was recently shown to predict worse outcomes in patients with NICM (3); however, this finding has yet to be replicated with other scar assessment modalities. Similarly, wider QRS duration has been inconsistently related to mortality in patients with NICM (4,5). We hypothesized that larger areas of scar defined by low voltage during electroanatomic mapping and wider QRS durations would predict worse transplantation- and LVAD-free survival among patients with NICM and VT. We further sought to develop and then validate an electrophysiological risk score to aid in risk stratifying patients with NICM and VT.
The risk score was derived from consecutive patients with left ventricular NICM and sustained VT who had been referred to the Hospital of the University of Pennsylvania for ablation between July 2007 and April 2011. Patients with idiopathic VT, right ventricular cardiomyopathy, or ischemic cardiomyopathy, as defined by history of myocardial infarction or obstructive coronary artery disease on angiography, were excluded. Patients undergoing multiple ablation procedures during this period were included once, with analysis of the voltage maps from the first procedure only. However, if epicardial mapping was not performed during the first procedure but was performed during a subsequent procedure, the subsequent epicardial map was analyzed. The risk score was validated in a separate cohort of consecutive patients with identical inclusion and exclusion criteria who had been referred to our institution for ablation between March 2004 and June 2007. VT storm was defined as 3 or more separate episodes of VT within 24 h. All patients provided written informed consent for catheter ablation and for their anonymized medical information to be included in research studies.
Measurement of QRS duration and classification of QRS morphology
The presenting QRS duration, whether paced or conducted, was measured using digital calipers on the 12-lead electrocardiogram (ECG) immediately preceding VT ablation. For patients presenting in VT, QRS duration was measured on a prior ECG or the recordings from the VT ablation (CardioLab, GE Medical Systems, Waukesha, Wisconsin) not during VT. For patients presenting in paced rhythm, prior ECGs/ablation recordings were similarly searched to identify and measure native QRS complexes. QRS durations were compared between those experiencing death, transplantation, or LVAD and those not experiencing the primary endpoint. Because QRS durations would be expected to vary widely on the basis of whether bundle branch block or paced rhythm were present, QRS duration comparisons were further stratified by QRS morphology category (left bundle branch block, right bundle branch block, biventricular paced, right ventricular paced). Left bundle branch block, right bundle branch block, and intraventricular conduction delay were defined according to standard ECG criteria established by the American Heart Association, the American College of Cardiology, and the Heart Rhythm Society (6). ECGs were analyzed blinded to patient outcome.
Electroanatomic mapping and VT ablation
Electroanatomic mapping (CARTO, Biosense Webster, Diamond Bar, California) was performed during sinus or paced rhythm to define areas of low voltage and abnormal electrograms, consistent with scar (7), using a 3.5-mm open irrigation tip catheter (Navistar Thermocool, Biosense Webster) maintaining a fill threshold of 15 mm to ensure adequate sampling of the entire represented surface area. Bipolar electrograms were recorded between the distal and ring electrodes of the mapping catheter and filtered at 30 to 400 Hz. Endocardial unipolar electrograms were recorded between the distal electrode of the mapping catheter and Wilson’s central terminal and filtered at 1 to 240 Hz (8). All patients underwent endocardial mapping; epicardial mapping was performed when an ECG of spontaneous or induced VT suggested an epicardial exit (9) and/or endocardial ablation failed to eliminate targeted VT. Voltage maps were analyzed off-line using the CARTO area measurement software, with endocardial bipolar low voltage defined as <1.5 mV, endocardial unipolar as <8.3 mV, and epicardial bipolar as <1.0 mV (10,11). To avoid misclassification of epicardial fat or coronary arteries as scar, areas of epicardial bipolar voltage <1.0 mV were only considered scar if wide, split, or late electrograms were also present (12). Scar percentage was defined as the area of low voltage divided by the entire surface area of the mapped chamber, multiplied by 100. Because some patients did not undergo epicardial mapping and to account for the fact that some patients have greater endocardial bipolar scar percentage whereas others have greater epicardial bipolar scar percentage, a single combined variable, “greater of endocardial or epicardial bipolar scar percentage,” was created to most accurately reflect total scar burden. Voltage maps were analyzed blinded to patient outcome.
After electroanatomic mapping, VT induction was performed. Programmed stimulation was delivered from the right and left ventricles, with up to 3 extrastimuli delivered to refractoriness. When a 12-lead ECG of spontaneous VT was available, clinical VT was defined by match in all 12 leads. When a 12-lead ECG of spontaneous VT was not available, clinical VT was defined by match in near-field and far-field implantable cardioverter-defibrillator electrogram morphology, as well as cycle length within 30 ms of stored electrograms from spontaneous VT episodes. Every spontaneously occurring VT was considered clinical; thus, a single patient could have multiple clinical VTs. Special attention was paid to elimination of clinical VT. Additionally, all mappable VT and VT with cycle length >250 ms were also considered relevant and routinely targeted for ablation. When hemodynamically tolerated, entrainment mapping was used to define critical components of the VT circuit. If VT was not mappable, substrate modification was performed with linear and/or cluster lesions targeting exit sites identified by pace-mapping as well as late potentials. Ablation settings included power up to 50 W, with temperature limit 42° and goal 12 to 15 ohm impedance drop. After ablation, programmed stimulation was repeated in patients who were medically stable, with up to 3 ventricular extrastimuli delivered from 2 sites at 2 pacing cycle lengths. Acute procedural success was defined as elimination of all VT. Partial success was defined as elimination of all clinical VT. Failure was defined as persistent inducibility of clinical VT.
The primary endpoint was defined as death, heart transplantation, or LVAD implantation. Patients were routinely evaluated at 6 weeks after ablation and then at 3- to 6-month intervals. Defibrillators were interrogated at each visit to assess for VT recurrence. For patients not followed at our institution, referring cardiologists were contacted and telephone interviews performed with patients or family members every 6 to 12 months. The social security death index was also queried.
Statistical analysis, derivation of risk score, and validation of risk score
Continuous variables are expressed as means with standard deviations, and categorical variables are expressed as percentages. The independent samples Student t test and Pearson chi-square test were used to compare normally distributed continuous and dichotomous variables, respectively. The Fisher exact test was used to compare dichotomous variables with expected cell values <5. The Mann-Whitney U test was used to compare non-normally distributed continuous variables. Patients were followed until death, transplantation, LVAD, or the most recent clinical evaluation. We identified predictors of time to the primary endpoint using Cox proportional hazard modeling. Variables subjected to univariate screening included age, LVEF, New York Heart Association (NYHA) heart failure class, creatinine, VT storm, number of spontaneous and induced VTs, acute procedural success, diuretic use, amiodarone use, endocardial unipolar low voltage percentage, greater of endocardial or epicardial bipolar scar percentage, and native QRS duration. The assumption of proportional hazards was tested and confirmed. Patients without natively conducted QRS complexes were excluded from QRS duration survival analysis. Variables showing marginal associations with time to death, transplantation, or LVAD on univariate testing (p < 0.10) were assessed in a multivariate model.
Variables retaining significant, independent associations with time to death, transplantation, or LVAD on multivariate testing were included in the risk score. These variables were assigned weighted points in proportion to their β regression coefficient values, rounded to the nearest integer. By adding these points, an overall risk score was calculated. Cutoffs were established to divide the population into equally sized high-, intermediate-, and low-risk groups. We then constructed Kaplan-Meier curves to illustrate transplantation and LVAD-free survival and compared the high-, intermediate-, and low-risk groups using a log-rank test.
Next, we validated the risk score in a separate population with NICM and VT. Risk scores were calculated and subjects assigned to high-, intermediate-, and low-risk groups using the cutoffs established in the derivation cohort. Again, Kaplan-Meier curves were constructed and transplantation/LVAD-free survival compared between the risk groups using a log-rank test. A receiver operating characteristic curve was plotted for the risk score to assess sensitivity and specificity for 3-year outcomes. Analyses were performed using SPSS (version 20.0, SPSS Inc., Chicago, Illinois) and R (Foundation for Statistical Computing, Vienna, Austria). We considered p values ≤0.05 to indicate statistical significance.
Baseline characteristics and outcomes of derivation cohort
Of 100 patients with NICM undergoing VT ablation, 49 recurred over a mean follow-up of 1.2 years (range 3 days to 4.8 years). Among these 49, recurrent VT was treated exclusively with antitachycardia pacing in 44% and with at least 1 implantable cardioverter-defibrillator shock in 56%. Forty-one patients experienced the primary study endpoint (3 LVADs, 12 transplantations, and 29 deaths). Two patients underwent LVAD implantation followed by transplantation. One patient underwent LVAD implantation and later died. The majority of LVAD implantations and transplantations were performed for end-stage heart failure; 3 were performed for refractory ventricular arrhythmias. Compared with those who survived the study period without transplantation or LVAD, those who experienced an endpoint were older at the time of presentation, with lower LVEF, higher NYHA heart failure class, higher creatinine, more likely to present in VT storm, with a greater number of VTs occurring spontaneously or induced, more likely to have a cardiac resynchronization therapy device, and more likely to be taking a diuretic or amiodarone (Table 1).
QRS duration and morphology
The presenting rhythm on ECG immediately preceding ablation was biventricular paced in 42 patients, right ventricular paced in 7 patients, left bundle branch block in 7 patients, right bundle branch block in 7 patients, and normal or intraventricular conduction delay in 37 patients (Figure 1). In 25 patients, ECGs of native conduction were not available on the day of ablation but were available on previous ECGs, performed a median of 29 days earlier (interquartile range: 3 to 509 days). Of the 49 patients presenting in a paced rhythm, 18 patients had underlying left bundle branch block, 10 patients had right bundle branch block, and 13 patients had normal or intraventricular conduction delay. Eight patients had no native rhythm underlying pacing and were not included in the QRS duration survival analysis. Left bundle branch block was more common among patients who experienced the primary endpoint than among those who did not (41.7% vs. 17.9%, p = 0.01) (Table 1, Figure 1).
Native QRS duration was greater among those who experienced death, transplantation, or LVAD. The difference was significant when combining all native QRS morphologies (151.7 ± 38.2 ms vs. 128.9 ± 31.5 ms, p = 0.002) (Figure 2A) and when limiting the comparison to patients with normal or intraventricular conduction delay morphologies (123.8 ± 27.1 ms vs. 109.1 ± 16.1 ms, p = 0.02) (Figure 2B). There was a directionally consistent, nonsignificant trend toward greater QRS duration among patients experiencing the endpoint in each additional QRS morphology category (left bundle branch block, right bundle branch block, biventricular paced, and right ventricular paced, p = 0.2, 0.7, 0.2, and 0.2, respectively) (Figures 2C to 2F).
Voltage mapping and ablation
A mean of 357 points were sampled per endocardial voltage map and 506 points per epicardial voltage map. Patients who went on to experience the endpoint had a greater percentage of endocardial bipolar scar (16.7 ± 12.8 vs. 8.7 ± 8.6, p = 0.001) (Table 1), endocardial unipolar low voltage area (54.4 ± 30.3 vs. 28.8 ± 21.4, p < 0.001), and greater of endocardial or epicardial bipolar scar (18.4 ± 12.3 vs. 10.9 ± 8.7, p = 0.001) (Figure 3). Programmed stimulation was performed at the end of ablation in 80% of patients. Among these patients, complete success was achieved in 50%, partial success in 38%, and failure in 12%.
Cox proportional hazard modeling and derivation of risk score
Using univariate testing, older age, lower LVEF, higher NYHA heart failure class, higher creatinine, VT storm, greater number of spontaneous or induced VTs, diuretic use, amiodarone use, larger endocardial unipolar low voltage area, larger endocardial or epicardial bipolar scar percentage (whichever was greater), acute procedural success and longer native QRS duration were all associated with shorter time to death, heart transplantation, or LVAD (Table 2). In multivariate analysis, lower LVEF, VT storm, larger endocardial or epicardial bipolar scar percentage (whichever was greater), and longer native QRS duration remained independently associated with shorter time to death, heart transplantation, or LVAD.
Accordingly, these 4 variables were weighted in proportion to their β regression coefficients and combined to create a risk score. To maximize ease of use and thereby clinical utility, β regression coefficients were converted to points by dividing all coefficients by the value of the lowest coefficient (0.012) and then rounding to the nearest integer, as follows.
Risk scores were then calculated for each patient and cutoffs selected to divide the population into equally sized thirds. High risk was defined as risk score >180 points, intermediate risk 100 to 180, and low risk <100. For example, a patient who presented with VT storm, bipolar scar percentage 25, QRS duration 170 ms, and LVEF 20 would have a risk score = 83 + 100 + 170 − 60 = 293 (high risk). A patient who presented without VT storm, with bipolar scar percentage 5, QRS duration 120 ms, and left ventricular ejection fraction 20 would have a risk score = 0 + 20 + 120 – 60 = 80 (low risk).
Kaplan-Meier curves were then constructed comparing 2-year transplantation/LVAD-free survival among the high-, intermediate-, and low-risk groups. Survival was worse in the high-risk group compared with the intermediate- and low-risk groups (<40% vs. 70% and >90% respectively, p < 0.001 for trend across groups) (Figure 4). Rates of 1-year transplantation/LVAD-free survival were 50%, 75%, and >90%, respectively. The area under the risk score receiver operating characteristic curve was 0.79 for the derivation cohort.
Validation of risk score
Of the 50 patients in the validation cohort, 20 patients were classified as high risk, 15 patients as intermediate risk, and 15 patients as low risk. Baseline characteristics of the validation cohort, stratified by risk group, are provided in Table 3. During 3.0 years mean follow-up (range 9 days to 10.3 years), 17 (85%) patients in the high-risk group died or underwent transplantation/LVAD, compared with 7 (47%) patients in the intermediate-risk group and 2 (13%) patients in the low-risk group (Table 3). One patient in the high-risk group underwent LVAD implantation followed by transplantation and subsequently died 6 months later. Of the 49 deaths in the derivation and validation cohorts, the cause of death could be identified in 14 patients. Of these 14, causes of death were infection in 3 patients, cardiogenic shock in 2 patients, stroke in 2 patients, complication of an unrelated procedure in 2 patients, pulseless electrical activity in 2 patients, complication of VT ablation in 1 patient, refractory VT in 1 patient, and trauma in 1 patient.
In Kaplan-Meier analysis, transplantation and LVAD-free survival was worse in the high-risk group compared with the intermediate- and low-risk groups (<10% vs. 50% and >80% respectively, p < 0.001 for trend across groups) (Figure 5). The area under the risk score receiver operating characteristic curve was 0.82 for the validation cohort. A cutoff score of 100 was highly sensitive for development of the primary endpoint by 3 years (sensitivity 93.5%, specificity 54.7%), whereas a cutoff of 180 was more specific (sensitivity 64.7%, specificity 85.8%).
In a moderate-sized group of contemporary patients with NICM presenting for VT ablation, we identified wider native QRS duration, greater endocardial or epicardial bipolar scar percentage, VT storm, and LVEF as independent predictors of shorter time to death, transplantation, or LVAD. By combining these variables into an “electrophysiological risk score,” we were able to discriminate among patients at high-, intermediate-, and low-risk for major adverse events over medium-term follow-up. We subsequently validated the risk score in a separate group of patients with longer-term follow-up, and it again predicted risk with a high degree of accuracy.
Whereas left bundle branch block has long been established as a risk factor for adverse heart failure outcomes (13), we further showed that QRS width, within each category of QRS morphology, further predicts risk. For example, among those with normal QRS morphology or intraventricular conduction delay, greater QRS width predicted shorter time to death, transplantation, or LVAD. This observation was directionally consistent, although not statistically significant, across all QRS morphology categories. Within a given QRS morphology category, increasing QRS width is likely a manifestation of slower cell-to-cell conduction, potentially a result of increased fibrosis (14).
Scar size as assessed by MRI has been shown to predict death and need for heart transplantation among patients with ischemic cardiomyopathy, independently of LVEF (15). More recently, this finding was replicated among patients with NICM (3). Thus, it is not surprising that larger scars, as identified by bipolar voltage mapping, would predict worse prognosis among patients with NICM presenting with VT. In fact, careful voltage mapping with good contact and sampling density may be a more sensitive technique than MRI for detection of myocardial scarring. In particular, unipolar voltage mapping may have even greater sensitivity for identifying subtle areas of abnormality. For example, larger areas of unipolar voltage abnormality were recently shown to predict irreversibility of NICM in patients without scar detected by MRI or bipolar voltage mapping (16).
In contrast to a recent multicenter report (17), acute procedural success was not significantly associated with the primary endpoint in our multivariate model (p = 0.06). It is likely that with a larger sample size, the association would have been statistically significant.
Through the incorporation of electrophysiological risk factors, we aimed to improve risk stratification of patients with NICM beyond LVEF. Should our risk score be validated in other populations, measurement of QRS duration and voltage mapping could be used as part of a baseline electrophysiological evaluation, which, when combined with LVEF, could provide valuable prognostic information. This information could then be used to counsel patients and guide the intensity of their heart failure follow-up and interventions. Physicians caring for high-risk patients could intensify neurohormonal modulators as aggressively as possible and have a lower threshold for referral to advanced heart failure programs specializing in LVAD, transplantation, and other cutting-edge/experimental treatments.
Although our electrophysiological risk score performed well in this single-center study, it should be validated in other centers. Further, to determine the range of its applicability, the risk score would also need to be tested in different patient populations, such as patients with NICM without VT and patients with ischemic cardiomyopathy. Our risk score cannot be calculated in those without intrinsic atrioventricular conduction, as native QRS duration is required. Not all patients underwent epicardial voltage mapping. However, all patients did undergo endocardial unipolar voltage mapping, which has been shown to predict the extent and distribution of midmyocardial and epicardial scar (8). Lastly, the threshold to implant an LVAD or perform a heart transplantation may vary from institution to institution.
We identified wider native QRS duration, greater endocardial or epicardial bipolar scar percentage, and VT storm, in addition to LVEF, as independent predictors of shorter time to death, transplantation, or LVAD in patients with NICM and VT. By combining these variables into an electrophysiological risk score, patients can be classified as high-, intermediate-, or low-risk for major adverse outcomes.
COMPETENCY IN MEDICAL KNOWLEDGE: In addition to LVEF, we identified wider native QRS duration, increased scar percentage detected by electroanatomic mapping, and VT storm as independent predictors of shorter time to death, transplantation, or LVAD in patients with nonischemic cardiomyopathy. We combined these factors to create an electrophysiological risk score that accurately divided patients into high-, intermediate-, and low-risk categories.
TRANSLATIONAL OUTLOOK: Should our risk score be validated in different populations, measurement of QRS duration and voltage mapping could be performed as a baseline electrophysiological evaluation, which, when combined with LVEF, could provide more accurate prognostic information. This could then be used to guide the intensity of heart failure follow-up and interventions.
Supported in part by the F. Harlan Batrus Research Fund and the Susan and Murray Bloom Fund. Dr. Marchlinski has served on scientific advisory boards and received research support from St. Jude Medical and Biosense Webster. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- left ventricular assist device
- left ventricular ejection fraction
- magnetic resonance imaging
- nonischemic cardiomyopathy
- ventricular tachycardia
- Received April 13, 2015.
- Revision received July 14, 2015.
- Accepted July 16, 2015.
- American College of Cardiology Foundation
- Tokuda M.,
- Tedrow U.B.,
- Kojodjojo P.,
- et al.
- Shamim W.,
- Yousufuddin M.,
- Cicoria M.,
- Gibson D.G.,
- Coats A.J.,
- Henein M.Y.
- Surawicz B.,
- Childers R.,
- Deal B.J.,
- et al.
- Cassidy D.M.,
- Vassallo J.A.,
- Miller J.M.,
- et al.
- Hutchinson M.D.,
- Gerstenfeld E.P.,
- Desjardins B.,
- et al.
- Marchlinski F.E.,
- Callans D.J.,
- Gottlieb C.D.,
- Zado E.
- Cano O.,
- Hutchinson M.,
- Lin D.,
- et al.
- Zimetbaum P.J.,
- Buxton A.E.,
- Batsford W.,
- et al.
- Kwon D.H.,
- Halley C.M.,
- Carrigan T.P.,
- et al.
- Campos B.,
- Jauregui M.E.,
- Park K.M.,
- et al.
- Yokokawa M.,
- Kim H.M.,
- Baser K.,
- et al.