Author + information
- Received March 13, 2000
- Revision received November 3, 2000
- Accepted December 6, 2000
- Published online March 15, 2001.
- Alfred P Hallstrom, PhD∗,* (, )
- John H McAnulty, MD, FACC†,
- Bruce L Wilkoff, MD, FACC§,
- Dean Follmann, PhD#,
- Merritt H Raitt, MD, FACC†,
- Mark D Carlson, MD, FACC‡,
- Anne M Gillis, MD¶,
- Hue-Teh Shih, MD, FACC∥,
- Judy L Powell, RN∗,
- Hank Duff, MD¶,
- Blair D Halperin, MD, FACC†,
- the Antiarrhythmics Versus Implantable Defibrillator (AVID) Trial Investigators
- ↵*Reprint requests and correspondence:
Dr. Alfred P. Hallstrom, University of Washington, 1107 Northeast 45th, Suite 505, Seattle, Washington 98105
The goal of this study was to identify subgroups of arrhythmia patients who do not benefit from use of the implantable cardiac defibrillator (ICD).
Treatment of serious ventricular arrhythmias has evolved toward more common use of the ICD. Since estimates of the cost per year of life saved by ICD therapy vary from $25,000 to perhaps $125,000, it is important to identify patient subgroups that do not benefit from the ICD.
Data for 491 ICD patients enrolled in the Antiarrhythmics Versus Implantable Defibrillators Study were used to create a hazards model relating baseline factors to time to first recurrent arrhythmia. The model was used to predict the hazard for recurrent arrhythmia among all trial patients. A priori cut points provided lower and higher recurrent arrhythmia risk strata. For each stratum the incremental years of life due to ICD versus antiarrhythmic drug therapy were calculated.
Factors that predicted recurrent arrhythmia were: ventricular tachycardia as the index arrhythmia, history of cerebrovascular disease, lower left ventricular ejection fraction, a history of any tachyarrhythmia before the index event and the absence of revascularization after the index event. Survival times (over a follow-up of three years) were identical in each arm of the lowest risk sextile (survival advantage 0.03 ± 0.12 [se] years), while the survival advantage for patients above the first sextile was 0.27 ± 0.07 (se) years (two-sided p = 0.05).
Patients presenting with an isolated episode of ventricular fibrillation in the absence of cerebrovascular disease or history of prior arrhythmia who have undergone revascularization or who have moderately preserved left ventricular function (left ventricular ejection fraction > 0.27) are not likely to benefit from ICD therapy compared with amiodarone therapy.
Treatment of serious ventricular arrhythmias has evolved toward more common use of the implantable cardiac defibrillator (ICD). A number of randomized trials have now reported a reduction in mortality ranging between 20% and 40% among patients treated with an ICD compared with patients treated with amiodarone or other antiarrhythmic therapies (1–4). Since estimates of the cost per year of life saved by ICD therapy vary from $25,000 to perhaps $125,000 (5,6), it is important to identify patient subgroups—if they exist—that do not benefit from the ICD (7).
During planning for the Antiarrhythmics Versus Implantable Defibrillators study (AVID), we recognized that stored electrograms in ICDs provide natural history data that can be used to investigate whether there are important predictors of arrhythmia recurrence among baseline variables. If it is possible to find baseline criteria identifying patients at low risk for recurrent arrhythmia, then we can evaluate how those patients would have survived had they not received an ICD by examining mortality of the corresponding group in the drug arm. If little or no survival advantage of the ICD was found, then this would constitute a subgroup that might do well without an ICD, thus avoiding potential complications of the ICD and yielding substantial cost savings.
Of the 1,016 patients enrolled in AVID, 507 patients were randomized to ICD and 509 patients were randomized to antiarrhythmic drugs (AAD) (95% were discharged on amiodarone). Nineteen patients who died during the baseline hospital stay (9 in the ICD and 10 in the AAD arms) were excluded from this analysis. There were seven patients in the ICD arm excluded because they were discharged without an ICD. Data for the remaining 491 ICD patients were entered into a proportional hazards failure time model to identify baseline factors (that is, patient characteristics, histories and therapies and procedures up to the time of baseline hospital discharge) that predicted time to first recurrent arrhythmia. The time to first recurrent arrhythmia has the advantage that ICD patients were generally antiarrhythmic drug-free until that time (only 19 patients in the ICD arm were placed on antiarrhythmic drugs before the first recurrent arrhythmia). An end point of recurrent arrhythmia was defined as: 1) any ICD therapy (either antitachycardia pacing [ATP] or shock) for ventricular fibrillation (VF) or ventricular tachycardia (VT) documented by the ICD stored electrogram; 2) an episode of VF or VT identified by hospitalization for recurrent arrhythmia or out-of-hospital emergency care, documented by emergency medical personnel with external monitoring; or 3) death classified by the death events committee as arrhythmic (8).
We used the resulting model to predict the hazard for recurrent arrhythmia among all patients regardless of treatment assignment. Six patients (two in the ICD and four in the AAD arms) had to be excluded because data were missing on factors found to be predictive of recurrence of arrhythmia. We chose to use the sextiles of the hazard distribution as a priori cut points based on our belief that subgroups smaller than 10% to 15% of the population for whom the benefit of ICD was questionable would be of little practical importance. A priori cut points reduce the problem of using the “best” cut point based on the data, and sextiles gave us potential cut points of 16.7%, 33.4% and 50.1%—reasonable portions of the population. For each stratum defined by the sextile cut point, estimates of the mortality distributions and, hence, estimates of the mean survival time and associated standard error were computed in the ICD and AAD arms (ostensibly equivalent because of randomization) by the Kaplan-Meier method, and, from these, the incremental years of life due to ICD versus AAD therapy were calculated. Although randomization was terminated in April of 1997, for analysis of factors predictive of time to arrhythmia recurrence in the ICD arm, we included the observational follow-up period from April 7, 1997 through August 31, 1998, which allowed a minimum follow-up of 15 months and a maximum of five years. However, survival was compared only during the follow-up obtained until the trial was terminated on April 7, 1997, to avoid the impact that implantation of ICDs in the drug arm might create when the trial was terminated. Survival was truncated at three years because very few patients were followed in the trial beyond three years.
By three years a recurrent arrhythmia occurred in 313 (63.7%) of the 491 ICD patients. These were identified by ICD shock for VF in 53 (16.9%), ICD shock for VT in 166 (53.0%), ATP for VT in 67 (21.4%), VF/VT documented by hospital or EMS recordings in 17 (5.4%) and death classified as arrhythmic in 10 (3.2%). In the proportional hazards model, factors that contributed to prediction of recurrent arrhythmia in the ICD group were: VT as the index arrhythmia, history of cerebrovascular disease, lower left ventricular ejection fraction (LVEF), a history of any tachyarrhythmia (including atrial fibrillation) before the index event and the absence of revascularization, either coronary artery bypass grafting or percutaneous transluminal coronary angiography, after the index event during baseline hospitalization (Table 1). The lowest risk (of recurrent arrhythmia) sextile in this model consists mostly (78.9%) of patients whose index arrhythmia was VF, with no prior arrhythmia, no cerebrovascular disease and either: 1) revascularization, or 2) no revascularization and LVEF >27%. Estimates of the time to recurrent arrhythmia are displayed in Figure 1. Over a mean 2.6 years (range 0 to 5.2) of follow-up, 35.6% of ICD patients in the first sextile had a recurrent arrhythmia compared with 70.2% of ICD patients in the second through sixth sextiles. Estimates of survival by the treatment arm for the first and second through sixth sextiles are displayed in Figure 2. Patients in the lowest sextile for risk of recurrent arrhythmia had identical mean survival times (over a follow-up of three years) in each arm (survival difference 0.03 ± 0.12 [se] years), while the survival advantage for patients above the first sextile was 0.27 ± 0.07 [se] years over a follow-up period of three years (two-sided p = 0.05). Use of any other sextile as a cut point did not provide significant differential incremental survival estimates; for example, use of the second sextile as a cut point yielded survival differences of 0.16 ± 0.10 years in the first and second sextiles and 0.26 ± 0.08 years in the third through sixth sextiles. The recurrent arrhythmia rates and the mean survival time to three years for each sextile are shown in Table 2.
Baseline characteristics of the subgroup defined by the first sextile identified as “potentially not benefiting” from an ICD are shown in Table 3. As expected from the prediction model, there are substantial differences in a number of characteristics between the first and higher sextiles: LVEF: 46.8% versus 28.6%; index arrhythmia VF: 92.2% versus 34.1%; no prior tachyarrhythmias: 89.8% versus 29.5%; and history of cerebrovascular disease: 0.6% versus 17.0%. Many other factors, such as a history of congestive heart failure, are also different, but not independently. Procedures during the index hospitalization and therapies at discharge from the index hospitalization are shown in Table 4. Revascularization during the index hospitalization was three times as common in the first sextile (25% vs. 8%). Although some procedures and discharge therapies differed between the first and higher sextile groups, differences in use by treatment arm were relatively the same in the sextile groups, with the possible exception of calcium channel blockers.
A substantial subgroup (16.9%) of AVID patients has been identified for whom the ICD rendered no survival advantage over amiodarone during three years of follow-up. This finding needs to be verified in an independent sample. Unfortunately, the AVID registry cannot be used for this purpose because LVEF was obtained on only a small subset of the registry patients, and no data on cerebrovascular history were gathered. The predictors of arrhythmia recurrence in AVID were: 1) measures of arrhythmia based on historical information or the index event; 2) ejection fraction (EF) (related to the general level of left ventricular dysfunction); 3) a history of cerebrovascular disease and 4) revascularization after the index event but during hospitalization for the index event and declared to be necessary before randomization. A patient’s risk for arrhythmia recurrence can be determined from the formula in Table 1, but nearly 80% of the low-risk patients were survivors of VF with no prior arrhythmia, no cerebrovascular disease and either had an EF >27% or were revascularized after the VF. These criteria are not as simple as the 35% EF cut off Domanski et al. (9)derived from the AVID data. However, the report by Domanski et al. (9)does not recognize the strong correlation between LVEF and presenting arrhythmia as a result of the AVID entry criteria. To be eligible for AVID, VT patients had to have EFs <40%, while there was no EF restriction for VF patients. Thus, VF patients were over-represented in the EF >35% patient population. Our result suggests that presenting arrhythmia may be at least as important as LVEF in determining who may not benefit from ICD implantation. A comparison of classification by LVEF and sextile is shown in Table 5. Of the patients, identified at low risk for recurrent arrhythmia 28% had LVEF ≤0.35, while 22.7% of the patients identified at high risk for recurrent arrhythmia had LVEF >0.35. Thus, there is substantial nonoverlap.
Mortality was similar and moderately low in the sextile at low risk of recurrent arrhythmia whether the patient was treated with ICD or AAD (e.g., at two years, survival in the ICD and AAD arms was 87% and 91% in the first sextile and 82% and 74% in those not in the first sextile). The equivalence of survival cannot be attributed to crossover by addition of an ICD since only two of the 76 low arrhythmia risk (first sextile) patients in the drug arm had received an ICD by three years. It may be that amiodarone is relatively more effective in this subgroup. Looking at clinically relevant arrhythmias (criteria 2 and 3 in the first paragraph of the Methods section), the arrhythmia recurrence rate at three years in the drug arm was 12% in the first sextile and 35% in sextiles two through six, a ratio of 0.33 compared with a ratio of 0.48 for the total arrhythmia recurrence rates in the ICD arm. These ratios suggest amiodarone is eliminating or rendering nonclinical proportionately more of the “potential” arrhythmia or that proportionately more of the arrhythmia are self-terminating (without an ICD to intervene) with insufficient symptoms to cause hospitalization in the first sextile.
We can only speculate as to why the identified criteria might select patients who are at lower risk for recurrent arrhythmia and seem to do as well on amiodarone as after ICD implant. Survivors of VF might be expected to be the most likely to benefit from an ICD given that survival after a recurrence of VF is much less likely than from a recurrence of VT. On the other hand, recurrences of VF, though perhaps more likely to be fatal, are much less common than recurrences of VT. On electrophysiologic testing, VT is a much more reproducible finding than VF. Perhaps in many patients VF is a one-time chance occurrence or is the result of severe ischemia that is made less likely to occur by revascularization. A history of cerebrovascular disease suggested a higher rate of recurrence and higher mortality and may similarly identify patients with more aggressive vascular disease who are more likely to have a recurrence of an ischemia triggered VF. A history of a prior arrhythmia was also a risk factor and, though this group includes patients with a history of atrial fibrillation, it may be most important as a marker of a prior VT. In contrast with VF, VT is usually a reproducible arrhythmia that is due to reentry around myocardial scarring that is usually not eliminated by anti-ischemic therapy. In this substrate, which would be present in patients that present with VT or had had an episode of VT in the past, arrhythmia is more likely to recur and may be more resistant to the antiarrhythmic effects of amiodarone and, thus, make the ICD relatively more effective.
It is intriguing to consider whether these patients need specific antiarrhythmic treatment at all. Indeed, of the 17 deaths in the 166 patients, 11 were classified by the events committee to be nonarrhythmic. Based on a Cox proportional hazards survival model of mortality in the 166 patients at low risk of recurrent arrhythmia, higher Canadian Cardiovascular Class for angina at baseline (relative hazards [rh] approximately 1.5), index event in-hospital (rh = 7.6) and a history of renal disease (rh = 7.0) were predictive of mortality. Adjusting for these baseline predictors, among discharge medications and procedures performed during the baseline hospitalization, nitrates (rh = 2.9) and calcium channel blockers (rh = 3.0) were predictive of mortality. Most of these factors suggest that the deaths may have been secondary to ischemia. Again, perhaps amiodarone is relatively more effective in these patients because of its antiadrenergic or anti-ischemia properties. The similar efficacy of amiodarone and beta-adrenergic blocking agents in the Hamburg ICD trial may be further evidence that the anti-ischemic effects of amiodarone may be more important in this subgroup of patients.
A weakness of this study is that, although the general approach to determining our subgroup criteria were planned before collecting the data, the specific criteria for defining the subgroups were data-derived. Thus, the results must be considered speculative and need to be verified prospectively in other datasets. However, while the method used in this paper has the potential to produce a biased estimate of the benefit of ICD related to the inability to develop the prediction model in both groups, the bias tends to inflate the benefit of the ICD at the lowest sextile of risk, relative to the other sextiles. Thus, the conclusions drawn in this paper may be conservative.
A second weakness is that these data do not speak to the issue of a relative benefit of the ICD beyond three years. It does, though, seem unlikely that there will be a large number of patients who will have their first recurrence of VF or VT more than three years after their index event.
In summary, patients presenting with an isolated episode of VF in the absence of cerebrovascular disease or history of prior arrhythmia who have undergone revascularization or who have moderately preserved left ventricular function (LVEF > 0.27) are not likely to benefit from ICD therapy compared with amiodarone therapy. Treatment considerations for these patients should include: 1) amiodarone as a reasonable alternative therapy to ICDs, especially if there are clinical, financial, logistical, social or psychological reasons to avoid ICD therapy; and 2) focus on therapy for nonarrhythmic mortality risks, especially ischemia.
☆ Supported by a contract (N01-HC-25117) with the National Heart, Lung and Blood Institute, Bethesda, Maryland.
- antiarrhythmic drug
- antitachycardia pacing
- Antiarrhythmics Versus Implantable Defibrillators study
- ejection fraction
- implantable cardioverter defibrillator
- left ventricular ejection fraction
- relative hazards
- ventricular fibrillation
- ventricular tachycardia
- Received March 13, 2000.
- Revision received November 3, 2000.
- Accepted December 6, 2000.
- American College of Cardiology
- Antiarrhythmics Versus Implantable Defibrillators (AVID) Investigators
- Connolly S.J.,
- Gent M.,
- Roberts R.S.,
- et al.
- CASH Investigators,
- Kuck K.-H.,
- Cappato R.,
- Siebels J.,
- Rüppel R.
- Mushlin A.I.,
- Hall W.J.,
- Zwanziger J.,
- et al.
- Laupacis A.,
- Feeny D.,
- Detsky A.S.,
- Tugwell P.X.
- AVID Investigators,
- Greene H.L.
- Domanski M.J.,
- Sakseena S.,
- Epstein A.E.,
- et al.