Mortality in Patients Treated with Flecainide and Encainide for Supraventricular Arrhythmks Edward L.C. Pritchett, MD, and William E. Wilkinson, PhD
In a recent clinical trial, the class lc antiarrhythmic drugs encainide and flecainide were found to be associated with an increased mortality risk in patients with new myocardial infarction and ventricular arrhythmias. The purpose of this study was to assess whether an increased mortality risk also accompanied the use of these drugs to treat patients with supraventricular arrhythmias. Data were obtained from the respective pharmaceutical sponsors on the mortality observed with each drug in United States and foreign protocols enrolling patients with supraventricular arrhythmias. Mortality in the encainide population (343 patients) and the flecainide population (236 patients) was compared with that in a research arrhythmia clinic, the Duke population (164 patients). Nine deaths occurred in the combined encainide-flecainide population and 10 deaths occurred in the Duke population; the follow-up periods averaged 466 days and 1,285 days, respectively. The 6-year survival functions of these 2 populations, estimated by the KaplanMeier technique, did not dtffer significantly (p = 0.62). The hazard ratio for the combined encainide-flecainide population relative to the Duke population was estimated to be 0.6 with a 95% confidence interval of 0.2, 1.7. These descriptive comparisons did not demonstrate any excess mortality when flecainide and encainide were used in patients with supraventricular arrhythmias. (Am J Cardiol 1991;67:976-980) From the Department of Medicine (Divisions of Clinical Pharmacology and Cardiology) and the Department of Community and Family Medicine (Division of Biometry and Medical Informatics), Duke University Medical Center, Durham, North Carolina. Data management was partially supported by grant MO1 RR30 from the National Center for Research Resources, National Institutes of Health, Bethesda, Maryland. Additional support was provided by grants from Bristol-Myers Company, Wallingford, Connecticut, and 3M Pharmaceuticals, St. Paul, Minnesota. Manuscript received November 15, 1990; revised manuscript received and accepted January 8, 199 1. Address for correspondence: Edward L.C. Pritchett, MD, Box 3477, Room 2000, Duke South Hospital, Duke University Medical Center, Durham, North Carolina 27710.
976
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 67
lecainide and encainide are class Ic antiarrhythmic drugs that were developedinitially for treatment of ventricular arrhythmias. Before marketing approval by the Food and Drug Administration, both drugs were found to be effective for the management of patients with ventricular arrhythmias including premature ventricular beats and nonsustainedand sustained ventricular tachycardias; both compounds were labeled at the time of marketing (flecainide in 1985 and encainide in 1986) by the Food and Drug Administration for all these indications. In 1989 the Cardiac Arrhythmia SuppressionTrial showed that there were some patient populations in which these drugs might paradoxically increasethe risk of arrhythmic death; in patients with a recent myocardial infarction and premature ventricular beats, treatment with flecainide or encainide was associated with a 2.5-fold increase in mortality compared with placebo treatment. t At the time that the results of the Cardiac Arrhythmia Suppression Trial were released, both flecainide and encainide were under development for supraventricular arrhythmias; and both drugs probably were being used widely in clinical practice to treat supraventricular arrhythmias. Because of concern about the safety of these compoundsin these other patient populations, this report examines the mortality observed with flecainide and encainide during development by their respective sponsorsfor supraventricular arrhythmias. A description of mortality in other relevant populations is provided for comparison.
F
METHODS Study populations: We obtained from Bristol-Myers
Company and from the 3M Pharmaceuticals sections of their data basesfor patients with any supraventricular arrhythmias who were treated with encainide and flecainide, respectively. We specifically requestedfrom each sponsorthe following information about each patient: the length of follow-up on drug treatment, age, race, sex, and whether the patient had any heart diseasein addition to his arrhythmia. The information was provided to us by the sponsorin a computer file; we did not have accessto original casereport forms. The data file that we obtained did not specify the preciseelectro-
cardiographic diagnosisof the patient’s supraventricular arrhythmia, the nature of any associatedheart disease, whether other medications were used, nor why patients discontinued follow-up. To make the study populations as comparable as possible, we selectedfrom the available data basespatients who were aged L 18 years, who were ambulatory, and who receivedorally administered antiarrhythmic therapy. We obtained from Bristol-Myers Company a data base of 415 patients with supraventricular arrhythmias who were treated with encainide in the United States and foreign protocols between February 1977 and June 1989. We excluded 23 patients who receivedonly intravenous encainide therapy and 49 other patients who were l of these drugs. Data analysis: For the primary analyses, we combined the flecainide and encainide populations into a single population (called the combined encainide-flecainide population) as was done in the Cardiac Arrhythmia Suppression Trial. l The survival functions of this combined population and the Duke population were estimated using the well-known product-limit method developed by Kaplan and Meier3 and compared using a log-rank test and a proportional hazards model.4 Becauseof the small number of deaths observedin these study populations, the precision associatedwith estimated survival functions and the power associated with tests comparing survival functions are major considerations. Both of these concernscan be partially ad-
dressedby the use of confidence intervals. However, if survival functions are estimated using the KaplanMeier method, the standard techniques used in con-’ strutting confidence intervals for survival probabilities are based on large-sample (asymptotic) theory. An alternative to the nonparametric estimation of survival functions is to assume a specific parametric model for the risk of death during the course of the study. In particular, if the hazard function (instantaneous failure rate) is assumedto be constant for the duration of the study, the failure-time distribution is exponential. In this case,survival functions can be estimated, 2 survival functions can be compared and confidenceintervals can be constructed using “exact” methods which do not depend on large-sample theory.5)6 Since death rates increasewith age, the assumptionof a constant hazard function is not appropriate over an extended period of time. However, if the observationperiod of interest is sufficiently short that aging effects can be ignored, then the exponential distribution may provide a reasonableapproximation to the failure-time distribution for a mortality study. The adequacy of the exponential model was partially checked by fitting the Duke and combined encainide-flecainide data to a Weibull model4 in which the 2 failure-time distributions were assumedto have a common shape parameter; this parameter was not significantly different from unity, the value of the shape parameter for which the Weibull model reducesto an exponential model. For the study populations reported here, nonparametric analysesusing the Kaplan-Meier technique, the log-rank test and the proportional hazards model were supplementedby parametric analysesusing procedures derived from the exponential model. The latter procedures were usedprimarily to construct confidenceintervals for specific survival probabilities and hazard ratios. To provide a frame of reference for the survival function of each of these patient populations, the expected survival of each group was calculated based on the age-race-sex-specificmortality rates in the United States in 1980.’ Finally, the mortality experienceof the study populations was compared with the mortality observed in the Cardiac Arrhythmia SuppressionTrial.’ In that report, the survival experienceof patients treated with encainide or flecainide was compared with patients given placebo. In particular, the relative risk of death from any cause for patients receiving encainide or flecainide was estimated by forming the ratio of the 2 gross mortality rates (number of deaths divided by the number of patients), as if the observation time for each surviving patient had been the same. In a proportional hazards model for the comparison of 2 groups, relative risk can be defined as the hazard THE AMERICAN JOURNAL OF CARDIOLOGY MAY 1, 1991
977
the observed 6-year survival of this combined encainide-flecainide population calculated by the KaplanMeier life table method. Flecainide Encainide Duke Among the 154 patients in the Duke population, 10 (n=236) (n = 343) (n = 154) deaths were observedin the follow-up period that averNo. No. No. (%I (%) (%) aged 1,285 days (Table I). None of the deaths in the Race* Duke population occurred in a patient who was taking White 144 (94) 218 (96) 215 (91) flecainide, encainide, or any class Ic antiarrhythmic Other 10 (6) 9 (4) 21 (9) drug. Figure 1 shows the observed 6-year survival of Sex* Male 87 (56) 176 (59) 123 (52) this population. The comparison of the 2 survival funcFemale 67 (44 124 (41) 113 (48) tions in Figure 1 using the log-rank test did not reveal a Heart disease* Present 54 (35) 141 (45) 136 (58) statistically significant difference between them (p = Absent 100 (65) 174 (55) 100 (42) 0.62). The proportional hazards model was used to Mean age f SD 45.6 zk 16.3 46.7 f 15.8 52.5 zt 14.6 Follow-up (days) 1,285 877 609 634 311 325 make the samecomparisonwhile controlling for the effects of age and concomitant heart disease;not surpris* In the encainide data base, race was not recorded for 116 patients, sex for 43 and heart disease for 28. ingly, becauseof the distribution of these covariates in SD = standard deviation. the study populations, the difference between the 2 groups remained insignificant. For additional compariratio, the constant ratio of the 2 instantaneous failure son, the observed6-year survival in the Duke popularates. If the 2 failure-time distributions are exponential, tion is shown along with the survival expected on the the hazard ratio is simply the ratio of the parametersof basisof age-race-sex-specificmortality rates in the U.S. these 2 distributions. Our estimates of relative risk population in Figure 2. based on the exponential model are estimates of this Using the exponential failure-time model, the estihazard ratio. mated hazard ratio for the combined encainide-flecainide population relative to the Duke population is 0.6; a RESULTS 95% confidence interval for this hazard ratio is (0.2, Among the 236 patients in the flecainide population, 1.7). In contrast, the model-basedestimate of the hazonly 1 death was observedin the follow-up period that ard ratio associatedwith encainide or flecainide relative averaged311 days (Table I); among the 343 patients in to placebo in the Cardiac Arrhythmia SuppressionTrithe encainide population, 8 deaths were observedand al is 2.6 with a 95% confidence interval of (1.6, 4.3). the averagelength of follow-up was 609 days (Table I). (The reported relative risk for death from any cause Thus, for the combined encainide-flecainidepopulation, was 2.5 with a 95% confidence interval of [1.6, 4.51.) 9 deaths were observed among 579 patients with a mean follow-up interval of 488 days. Figure 1 shows TABLE I Demographic Populations
Characteristics
of the Study
100
G c fj ‘i s
90
Sd
z.- I
‘5 85 3 v)
95
p=O.62
-2 8
f
In a recent clinical trial, the class lc antiarrhythmic drugs encainide and flecainide were found to be associated with an increased mortality risk in patients with new myocardial infarction and ventricular arrhythmias. The purpose of this study was to assess whether an increased mortality risk also accompanied the use of these drugs to treat patients with supraventricular arrhythmias. Data were obtained from the respective pharmaceutical sponsors on the mortality observed with each drug in United States and foreign protocols enrolling patients with supraventricular arrhythmias. Mortality in the encainide population (343 patients) and the flecainide population (236 patients) was compared with that in a research arrhythmia clinic, the Duke population (164 patients). Nine deaths occurred in the combined encainide-flecainide population and 10 deaths occurred in the Duke population; the follow-up periods averaged 466 days and 1,285 days, respectively. The 6-year survival functions of these 2 populations, estimated by the KaplanMeier technique, did not dtffer significantly (p = 0.62). The hazard ratio for the combined encainide-flecainide population relative to the Duke population was estimated to be 0.6 with a 95% confidence interval of 0.2, 1.7. These descriptive comparisons did not demonstrate any excess mortality when flecainide and encainide were used in patients with supraventricular arrhythmias. (Am J Cardiol 1991;67:976-980) From the Department of Medicine (Divisions of Clinical Pharmacology and Cardiology) and the Department of Community and Family Medicine (Division of Biometry and Medical Informatics), Duke University Medical Center, Durham, North Carolina. Data management was partially supported by grant MO1 RR30 from the National Center for Research Resources, National Institutes of Health, Bethesda, Maryland. Additional support was provided by grants from Bristol-Myers Company, Wallingford, Connecticut, and 3M Pharmaceuticals, St. Paul, Minnesota. Manuscript received November 15, 1990; revised manuscript received and accepted January 8, 199 1. Address for correspondence: Edward L.C. Pritchett, MD, Box 3477, Room 2000, Duke South Hospital, Duke University Medical Center, Durham, North Carolina 27710.
976
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 67
lecainide and encainide are class Ic antiarrhythmic drugs that were developedinitially for treatment of ventricular arrhythmias. Before marketing approval by the Food and Drug Administration, both drugs were found to be effective for the management of patients with ventricular arrhythmias including premature ventricular beats and nonsustainedand sustained ventricular tachycardias; both compounds were labeled at the time of marketing (flecainide in 1985 and encainide in 1986) by the Food and Drug Administration for all these indications. In 1989 the Cardiac Arrhythmia SuppressionTrial showed that there were some patient populations in which these drugs might paradoxically increasethe risk of arrhythmic death; in patients with a recent myocardial infarction and premature ventricular beats, treatment with flecainide or encainide was associated with a 2.5-fold increase in mortality compared with placebo treatment. t At the time that the results of the Cardiac Arrhythmia Suppression Trial were released, both flecainide and encainide were under development for supraventricular arrhythmias; and both drugs probably were being used widely in clinical practice to treat supraventricular arrhythmias. Because of concern about the safety of these compoundsin these other patient populations, this report examines the mortality observed with flecainide and encainide during development by their respective sponsorsfor supraventricular arrhythmias. A description of mortality in other relevant populations is provided for comparison.
F
METHODS Study populations: We obtained from Bristol-Myers
Company and from the 3M Pharmaceuticals sections of their data basesfor patients with any supraventricular arrhythmias who were treated with encainide and flecainide, respectively. We specifically requestedfrom each sponsorthe following information about each patient: the length of follow-up on drug treatment, age, race, sex, and whether the patient had any heart diseasein addition to his arrhythmia. The information was provided to us by the sponsorin a computer file; we did not have accessto original casereport forms. The data file that we obtained did not specify the preciseelectro-
cardiographic diagnosisof the patient’s supraventricular arrhythmia, the nature of any associatedheart disease, whether other medications were used, nor why patients discontinued follow-up. To make the study populations as comparable as possible, we selectedfrom the available data basespatients who were aged L 18 years, who were ambulatory, and who receivedorally administered antiarrhythmic therapy. We obtained from Bristol-Myers Company a data base of 415 patients with supraventricular arrhythmias who were treated with encainide in the United States and foreign protocols between February 1977 and June 1989. We excluded 23 patients who receivedonly intravenous encainide therapy and 49 other patients who were l of these drugs. Data analysis: For the primary analyses, we combined the flecainide and encainide populations into a single population (called the combined encainide-flecainide population) as was done in the Cardiac Arrhythmia Suppression Trial. l The survival functions of this combined population and the Duke population were estimated using the well-known product-limit method developed by Kaplan and Meier3 and compared using a log-rank test and a proportional hazards model.4 Becauseof the small number of deaths observedin these study populations, the precision associatedwith estimated survival functions and the power associated with tests comparing survival functions are major considerations. Both of these concernscan be partially ad-
dressedby the use of confidence intervals. However, if survival functions are estimated using the KaplanMeier method, the standard techniques used in con-’ strutting confidence intervals for survival probabilities are based on large-sample (asymptotic) theory. An alternative to the nonparametric estimation of survival functions is to assume a specific parametric model for the risk of death during the course of the study. In particular, if the hazard function (instantaneous failure rate) is assumedto be constant for the duration of the study, the failure-time distribution is exponential. In this case,survival functions can be estimated, 2 survival functions can be compared and confidenceintervals can be constructed using “exact” methods which do not depend on large-sample theory.5)6 Since death rates increasewith age, the assumptionof a constant hazard function is not appropriate over an extended period of time. However, if the observationperiod of interest is sufficiently short that aging effects can be ignored, then the exponential distribution may provide a reasonableapproximation to the failure-time distribution for a mortality study. The adequacy of the exponential model was partially checked by fitting the Duke and combined encainide-flecainide data to a Weibull model4 in which the 2 failure-time distributions were assumedto have a common shape parameter; this parameter was not significantly different from unity, the value of the shape parameter for which the Weibull model reducesto an exponential model. For the study populations reported here, nonparametric analysesusing the Kaplan-Meier technique, the log-rank test and the proportional hazards model were supplementedby parametric analysesusing procedures derived from the exponential model. The latter procedures were usedprimarily to construct confidenceintervals for specific survival probabilities and hazard ratios. To provide a frame of reference for the survival function of each of these patient populations, the expected survival of each group was calculated based on the age-race-sex-specificmortality rates in the United States in 1980.’ Finally, the mortality experienceof the study populations was compared with the mortality observed in the Cardiac Arrhythmia SuppressionTrial.’ In that report, the survival experienceof patients treated with encainide or flecainide was compared with patients given placebo. In particular, the relative risk of death from any cause for patients receiving encainide or flecainide was estimated by forming the ratio of the 2 gross mortality rates (number of deaths divided by the number of patients), as if the observation time for each surviving patient had been the same. In a proportional hazards model for the comparison of 2 groups, relative risk can be defined as the hazard THE AMERICAN JOURNAL OF CARDIOLOGY MAY 1, 1991
977
the observed 6-year survival of this combined encainide-flecainide population calculated by the KaplanMeier life table method. Flecainide Encainide Duke Among the 154 patients in the Duke population, 10 (n=236) (n = 343) (n = 154) deaths were observedin the follow-up period that averNo. No. No. (%I (%) (%) aged 1,285 days (Table I). None of the deaths in the Race* Duke population occurred in a patient who was taking White 144 (94) 218 (96) 215 (91) flecainide, encainide, or any class Ic antiarrhythmic Other 10 (6) 9 (4) 21 (9) drug. Figure 1 shows the observed 6-year survival of Sex* Male 87 (56) 176 (59) 123 (52) this population. The comparison of the 2 survival funcFemale 67 (44 124 (41) 113 (48) tions in Figure 1 using the log-rank test did not reveal a Heart disease* Present 54 (35) 141 (45) 136 (58) statistically significant difference between them (p = Absent 100 (65) 174 (55) 100 (42) 0.62). The proportional hazards model was used to Mean age f SD 45.6 zk 16.3 46.7 f 15.8 52.5 zt 14.6 Follow-up (days) 1,285 877 609 634 311 325 make the samecomparisonwhile controlling for the effects of age and concomitant heart disease;not surpris* In the encainide data base, race was not recorded for 116 patients, sex for 43 and heart disease for 28. ingly, becauseof the distribution of these covariates in SD = standard deviation. the study populations, the difference between the 2 groups remained insignificant. For additional compariratio, the constant ratio of the 2 instantaneous failure son, the observed6-year survival in the Duke popularates. If the 2 failure-time distributions are exponential, tion is shown along with the survival expected on the the hazard ratio is simply the ratio of the parametersof basisof age-race-sex-specificmortality rates in the U.S. these 2 distributions. Our estimates of relative risk population in Figure 2. based on the exponential model are estimates of this Using the exponential failure-time model, the estihazard ratio. mated hazard ratio for the combined encainide-flecainide population relative to the Duke population is 0.6; a RESULTS 95% confidence interval for this hazard ratio is (0.2, Among the 236 patients in the flecainide population, 1.7). In contrast, the model-basedestimate of the hazonly 1 death was observedin the follow-up period that ard ratio associatedwith encainide or flecainide relative averaged311 days (Table I); among the 343 patients in to placebo in the Cardiac Arrhythmia SuppressionTrithe encainide population, 8 deaths were observedand al is 2.6 with a 95% confidence interval of (1.6, 4.3). the averagelength of follow-up was 609 days (Table I). (The reported relative risk for death from any cause Thus, for the combined encainide-flecainidepopulation, was 2.5 with a 95% confidence interval of [1.6, 4.51.) 9 deaths were observed among 579 patients with a mean follow-up interval of 488 days. Figure 1 shows TABLE I Demographic Populations
Characteristics
of the Study
100
G c fj ‘i s
90
Sd
z.- I
‘5 85 3 v)
95
p=O.62
-2 8
f
