________________________________________________________________________

 

DEPARTMENT OF HEALTH AND HUMAN SERVICES

FOOD AND DRUG ADMINISTRATION

CENTER FOR DEVICES AND RADIOLOGICAL HEALTH

 

Division of Cardiovascular Devices

Pacing, Defibrillator & Leads Branch

________________________________________________________________________

 

CLINICAL REVIEW

 

PMA:                          P030005                               

Prepared By:              Scott Proestel, M.D.

Date:                           06-14-2004                                                               

Product:                      CONTAK CD

Sponsor:                     Guidant Corporation

Team Leader:            Owen Faris, Ph.D.

________________________________________________________________________

 

TITLE

 

COMPANION (Comparison of Medical Therapy, Pacing, and Defibrillation in Heart Failure)

 

SYNOPSIS

 

This was a prospective, multi-center, controlled trial intended to enroll up to 2200 subjects with moderate to severe heart failure and randomize them 1:2:2 to optimal pharmacologic therapy (OPT), OPT plus cardiac resynchronization therapy without defibrillation (CRT-P), or OPT plus CRT-P with defibrillation (CRT-D), respectively.  There were four efficacy endpoints: (1) all-cause mortality plus all-cause hospitalization (primary endpoint), (2) total survival (secondary endpoint), (3) cardiac morbidity (secondary endpoint), and (4) change in maximal oxygen consumption from baseline to six months in an exercise substudy (secondary endpoint).

 

The COMPANION protocol was approved by FDA on October 1, 1999, and the first subject was enrolled on January 20, 2000.  Following the enrollment of 1638 subjects, the Data and Safety Monitoring Board (DSMB) recommended termination of further enrollment, which the Steering Committee implemented on December 1, 2002.  The basis provided by the DSMB for the recommendation was that the study had:

 

“…met the criteria specified in the protocol for success both for the primary endpoint of mortality and hospitalization combined and for mortality alone for biventricular pacing with defibrillator as compared to optimal medical therapy alone.  In addition, the comparison of biventricular pacing alone to optimal medical therapy nearly crossed the boundary for statistical significance for the primary endpoint of mortality and hospitalization combined and was consistent but not significant for mortality alone.”

 

Based on the results of COMPANION, the sponsor is currently seeking the following expanded indications for the approved CONTAK CD device (P010012, approved May 2, 2002):

 

“Guidant Cardiac Resynchronization Therapy Defibrillators (CRT-Ds) are indicated for reduction of all-cause mortality and symptoms of moderate to severe heart failure (NYHA III/IV) in patients who remain symptomatic despite stable, optimal heart failure drug therapy, and have left ventricular dysfunction (EF £ 35%) and QRS duration ³ 120 ms.”

 

“Guidant CRT-Ds provide ventricular antitachycardia pacing and ventricular defibrillation for the treatment of life-threatening ventricular arrhythmias.”

 

INCLUSION CRITERIA

 

  1. Moderate or severe heart failure, defined as symptomatic heart failure for at least six months with New York Heart Association (NYHA) class III or IV symptoms at the time of enrollment, and at least one of the following events in the previous 12 months:
    1. Hospitalization for heart failure management
    2. Outpatient visit in which intravenous (IV) inotropes or vasoactive infusion were administered continuously for at least 4 hours
    3. Emergency room visit of at least 12 hours duration in which IV heart failure medications were administered (including diuretics)
  2. QRS duration ³ 120 ms and PR interval > 150 ms from any two leads of a 12-lead ECG
  3. Left ventricular ejection fraction (EF) £ 35%
  4. Left ventricular end diastolic dimension ³ 60 mm (required only if LVEF measured by echo) or > 3.0 cm/m2
  5. Age ³ 18 years
  6. Optimal pharmacological therapy for heart failure as defined below.  All subjects were to be on optimal therapy for at least one month prior to randomization.  Exceptions to this requirement were made for diuretics and beta blockers.  Investigators were allowed to adjust diuretic doses at any time.  Subjects on beta blockers must have completed up-titration to a stable dose for at least three months prior to randomization.  All subjects were required to be receiving the following at the time of enrollment:
    1. A loop diuretic unless contraindicated or not tolerated
    2. An ACEI unless previously failed or contraindicated
    3. Spironolactone if a candidate, unless contraindicated or not tolerated
    4. A beta blocker if a candidate, unless contraindicated or not tolerated

 

Note: The ECG, LVEF, and LVEDD (if applicable) must have been measured within the three month period prior to randomization.  Furthermore, if a revascularization procedure occurred after the LVEF was measured, the LVEF measurement must be repeated at least 30 days after the revascularization procedure.

 

EXCLUSION CRITERIA

 

  1. Meet the general indications for an implantable cardioverter defibrillator
  2. Meet the general indications for antibradycardia pacing
  3. Expected to receive a heart transplant in the next six months
  4. Chronic, medically refractory atrial tachyarrhythmias.  “Chronic” is defined as not terminating spontaneously and cannot be terminated with medical intervention, or terminated with medical intervention but does not terminate spontaneously
  5. Unexplained syncope
  6. Myocardial infarction within 60 days of randomization
  7. History of non-compliance with oral heart failure therapy
  8. Progressive or unstable angina
  9. Uncontrolled blood pressure: systolic BP > 160 mm Hg or < 85 mm Hg, or diastolic BP > 90 mm Hg
  10. Hypersensitivity to a 0.7 mg dose of dexamethasone acetate
  11. Surgically uncorrected primary valvular heart disease
  12. Coronary artery disease with surgical or percutaneous correction within 60 days of randomization
  13. Women who are pregnant or not using medically acceptable birth control
  14. Hypertrophic obstructive cardiomyopathy
  15. Amyloid disease
  16. Hospitalization for heart failure or IV inotropic or vasoactive therapy in excess of four hours in the 30 days prior to enrollment
  17. Tricuspid prosthesis
  18. Life expectancy < 6 months due to any other medical conditions

 

SUBJECT MONITORING

 

Subjects were evaluated during a clinic visit at one week, one month, and at three-month intervals thereafter until study termination or subject withdrawal.  The data that was collected at each of these visits included a physical assessment, 12-lead EKG, device evaluation, and a global assessment (combines physician and subject assessments).  In addition, at the 3- and 6-month visits a six-minute walk and Minnesota Living With Heart Failure Questionnaire (MLWHFQ) was administered.

 

RESULTS

 

Baseline Characteristics

 

A total of 1638 subjects were enrolled in COMPANION, with 1520 subjects (92.8%) randomized to a treatment group.  There were 118 subjects enrolled but not randomized due to changes in clinical status between enrollment and randomization such that the eligibility criteria were no longer satisfied.

 

The baseline characteristics of enrolled subjects were analyzed by treatment group (TABLE 1).  The cohorts appear to be well matched overall.  Of note, there was a modestly higher percentage of NYHA functional class IV subjects in the OPT cohort compared with CRT-D.  As class IV subjects had a 47.2% mortality rate during the trial, and class III a 16.1% mortality rate, this baseline imbalance favored the device arm.  In addition, there was a higher percentage of ischemic subjects in the OPT arm.  As ischemic subjects had a 25.2% mortality rate during the trial, and non-ischemic subjects a 15.0% mortality rate, this imbalance also favored the device arm.  However, other baseline characteristics appeared to be well matched.

 

 

TABLE 1: Baseline Characteristics

 

 

CRT-D

CRT-P

OPT

 

Number Randomized

595

617

308

Age (years)

65.6 ± 11.2

65.3 ± 11.7

66.7 ± 10.7

Gender

 

 

 

      Male

401(67.4%)

415(67.3%)

211(68.5%)

      Female

194(32.6%)

202(32.7%)

97(31.5%)

NYHA Functional Class

 

 

 

      Class III

512(86.1%)

537(87.0%)

253(82.1%)

      Class IV

83(14.0%)

80(13.0%)

55(17.9%)

Etiology of Cardiomyopathy

 

 

 

      Ischemic

325(54.6%)

331(53.7%)

181(58.8%)

      Non-Ischemic

270(45.4%)

285(46.3%)

127(41.2%)

Ejection Fraction

22.5 ± 6.9

22.0 ± 6.8

22.8 ± 7.2

Conduction Abnormality

 

 

 

      LBBB

434(72.9%)

426(69.2%)

215(69.8%)

      RBBB

61(10.3%)

76(12.3%)

27(8.8%)

      IVCD

100(16.8%)

114(18.5%)

66(21.4%)

QRS Duration (ms)

159 ± 24

159 ± 25

156 ± 24

PR Interval (ms)

206 ± 40

203 ± 38

202 ± 35

Heart Rate (beats/min)

73 ± 13

73 ± 13

72 ± 12

ACEI/ARB Use

533(89.6%)

548(88.8%)

273(88.6%)

Beta-Blocker Use

402(67.6%)

421(68.2%)

204(66.2%)

Spironolactone Use

328(55.1%)

329(53.3%)

169(54.9%)

Digoxin Use

422(70.9%)

471(76.3%)

207(67.2%)

Diuretic Use

575(96.6%)

581(94.2%)

291(94.5%)

 


Primary Efficacy Endpoint

 

The protocol defined the primary efficacy endpoint as “…a combination of all-cause mortality and all-cause hospitalization, where all-cause mortality is defined as death from all causes and all-cause hospitalization is defined as admission to a hospital for any reason.”  The protocol excluded the hospitalization associated with the investigational device implant, and included emergency department visits or unscheduled office visits that resulted in treatment with intravenous inotropes or vasoactive drugs for worsening heart failure.  However, during the trial the definition of hospitalization was altered three times.  These changes were not documented within a protocol revision, and the FDA was not notified.  The definition was initially changed in March, 2001, to include only hospitalizations lasting greater than 24 hours.  Subsequently, the definition was changed in February, 2002, to include only hospitalizations for which the discharge date differed from the date of admission.  The third change was a requirement that the inotrope or vasoactive infusion last greater than 4 hours.  The timing of the third change is unknown to the FDA at the time of this review.

 

It should also be noted that while the protocol only states that “…the hospitalization associated with the investigational device implant…” would not be considered, this was expanded to include all subsequent additional hospitalizations for reimplant attempts.  Finally, episodes where subjects had their implant/reimplant hospitalizations prolonged due to adverse events were not included as a hospitalization.  It could be argued that such subjects should have been included as having reached the primary endpoint, as they were no longer in the hospital to receive a device.

 

The changes in hospitalization definition may have been in response to multiple early hospitalizations related to device implant issues.  While the Morbidity and Mortality Committee was blinded to treatment assignment and did not have access to the data in aggregate form, the Committee was aware of the nature of the adverse event.  It is reasonable to believe that the Committee would realize that admissions for lead revision and other device-related adverse events would be specific to subjects in the device arm of the trial.  Based on the revised definition of the primary endpoint, these hospitalizations were adjudicated as “adverse events” if they were not associated with a calendar date change, and therefore no longer counted towards the primary endpoint.  In addition, these events were no longer captured as hospitalizations, so the data provided by the sponsor to FDA could not be used to calculate the original primary endpoint.

 

The definition of hospitalization ultimately used for the trial was the following:

 

  1. All hospitalizations associated with a calendar date change, without regard to whether the hospitalization was considered elective or related to heart failure, with the following exceptions:
    1. index hospitalizations;
    2. reimplant attempt hospitalizations;
    3. hospitalizations that were considered elective and associated with the device (the sponsor further clarified on June 9, 2004, during a teleconference with this reviewer, that episodes such as lead revisions were not considered elective.  The only hospitalizations excluded based on this criterion were ones in which an OPT subject was implanted with a CRT-P or CRT-D device without a documented medical need, or a CRT-P subject received a CRT-D device without a documented medical need)
  2. All outpatient (i.e., clinic and emergency department) infusions of inotropic or vasoactive therapy for greater than four hours for worsening heart failure

 

The FDA statistician will provide formal statistical analyses of the primary and secondary efficacy endpoints.  This review will utilize only descriptive statistics of the efficacy data to provide additional insight into the results of the trial.  It should be noted that the protocol assigned 0.05 of alpha to four different endpoints.  The sponsor justified this excessive alpha allocation by stating that they would “…be conservative in the interpretation of the multiple analyses, looking for consistency across variables.”  No definition was provided for what would constitute consistency, and therefore there was no specific definition for what would constitute statistical success for the trial.

 

Withdrawals

 

There was a higher than anticipated dropout rate during the trial, particularly within the OPT arm.  Prior to November 30, 2002, there were 80 withdrawals in the OPT arm and 39 within CRT-D.  The sponsor has attributed this imbalance to the desire of OPT subjects to receive marketed CRT-P or CRT-D devices.  Therefore, the sponsor attempted to obtain information on subjects beyond the time at which they were withdrawn from the trial.  It should be noted that crossovers to a different therapy were considered Class I deviations against the investigator, defined as, “Investigator compromised patient’s right and/or welfare and/or study analysis groups.”  Apparently, in order to avoid a Class I deviation, investigators sometimes withdrew subjects from the trial for the purpose of crossing over to a different therapy.  However, withdrawal of subjects from a trial in order to obtain alternative therapy is not consistent with the principles underlying an intent to treat analysis.  Subjects should only have been withdrawn if further participation was felt to compromise their welfare, they refused further participation, or participation was no longer possible.

 

The magnitude of the effect of including data obtained following subject withdrawal was analyzed.  Of the 595 CRT-D subjects, 390 had a reported primary endpoint event (372 hospitalizations, 18 deaths).  Of these, 4 had the primary endpoint event following withdrawal (2 hospitalizations, 2 deaths).  The mean time to the primary endpoint following withdrawal in these 4 subjects was 78.1 days.  Of the 308 OPT subjects, 216 had a reported primary endpoint event (196 hospitalizations, 20 deaths).  Of these, 14 had their primary endpoint event following withdrawal (10 hospitalizations, 4 deaths).  The mean time to the primary endpoint following withdrawal in these 14 subjects was 237.7 days.

 

Reasons for not using the data that was collected on subjects following their withdrawal include:

 

  1. Withdrawal is intended to mark the end of a subject’s trial participation.
  2. Allowing the inclusion of data collected after withdrawal increases the type I error rate due to the additional chances for an intervention to “win.”
  3. Additional information was obtained on subjects following their withdrawal without necessarily obtaining consent, if the sponsor did not believe that consent had been rescinded.  However, please see #1.
  4. The quality of subject follow-up after withdrawal is unclear.  This follow-up included asking investigators to document retrospectively the primary endpoint events that they remembered.

 

For the above reasons, only the events from randomization to subject withdrawal, death, or November 30, 2002, (whichever came first) were included in the analyses for this review.  The only exception is for the mortality analysis, which was performed prior to this reviewer’s awareness that the data included events subsequent to a subject’s withdrawal.  Upon further review, it was found that this additional follow-up increased the overall mortality rate in the CRT-D arm from 15.8% (94/595) to 17.6% (105/595), and in the OPT arm from 21.1% (65/308) to 25.0% (77/308).  While this change modestly favored the device arm, the difference was not expected to change the overall conclusions of this review with respect to mortality.  In addition, vital status is more readily verifiable than whether a hospitalization occurred, so issue #4 may not be a concern.  However, issues regarding informed consent and type I error remain, and it may be reasonable to only include data up to subject withdrawal, death, or November 30, 2002, in the device label.

 

Mortality

 

All-cause mortality was decreased in the CRT-D cohort as compared to OPT (TABLE 2).  While this TABLE does not take into account the modest difference in mean observed time between the two cohorts, it does allow observation of the effects of the intervention in absolute terms (please see TABLE 4 for an adjustment for duration of observation).  There was a 7.4 % absolute reduction in all-cause mortality associated with CRT-D therapy.  CRT-D therapy was also associated with a decrease in cardiac deaths, due to important decreases in pump failure death and sudden cardiac death (SCD).  The pump failure mortality rates in the CRT-D and CRT-P cohorts were quite similar (8.7% and 8.6%, respectively), and were both less than the 11.0% pump failure death rate associated with OPT.  Given the known physiologic effects of CRT-P, it appears reasonable to believe that these results represent an improvement in pump failure mortality due to CRT-P.

 

There was a modest increase in SCD associated with CRT-P in comparison with OPT.  This difference is not likely due to chance, as the median time to SCD in the CRT-P cohort was substantially shorter than in either the OPT or CRT-D cohorts (TABLE 3 [Please note that this TABLE only includes subjects who died during the trial, and is not intended to characterize the median mortality for the cohorts as a whole]).  However, the decrease in pump failure deaths associated with CRT-P in both of the device arms overwhelms the increase in SCD associated with CRT-P.  It is possible that the increase in SCD associated with CRT-P is due to an improvement in pump function from CRT-P leading to fewer pump failure deaths, therefore rendering subjects more available to die from another cause such as SCD.  However, if this were the case, one would not expect the median time to SCD to be substantially shorter with CRT-P compared to OPT.  As might be expected, the SCD rate was lowest in the CRT-D arm, and the impact of the CRT-P component of the CRT-D device is that it may lead to a decrease in pump failure death but an increase in defibrillation therapy delivery.  It should also be noted that there were 10 deaths secondary to sepsis in the CRT-D cohort, 5 in the CRT-P cohort, and only one death attributed to sepsis in the OPT cohort.  As the device therapy arms involve implantation and maintenance of an intravascular device, it is reasonable to believe that this increase in septic deaths was not due to chance.

 

There was a modest difference in mean observation time per subject between the three cohorts (CRT-D 1.31 years/subject, CRT-P 1.32 years/subject, and OPT 1.22 years/subject.  As stated previously, the time of observation that was calculated for mortality included time beyond subject withdrawal if additional follow-up was possible.)  Therefore, mortality was adjusted for duration of observation (TABLE 4).  This adjustment improved the mortality results for the CRT-D and CRT-P cohorts relative to OPT, due to the slightly longer periods of observation in these groups.  However, the effect was modest and did not change any of the overall conclusions regarding the effect of CRT-D and CRT-P on mortality.

 

 

TABLE 2: Mortality Rates

 

 

CRT-D

CRT-P

OPT

 

Total Subjects

595

617

308

Total Deaths

105(17.6%)

131(21.2%)

77(25.0%)

      Cardiac Deaths

76(12.8%)

109(17.7%)

58(18.8%)

            Pump Failure Deaths

52(8.7%)

53(8.6%)

34(11.0%)

            Sudden Cardiac Deaths

17(2.9%)

48(7.8%)

18(5.8%)

            Ischemic Deaths

4(0.7%)

2(0.3%)

4(1.3%)

            Implant Procedure Deaths

2(0.3%)

6(1.0%)

0(0.0%)

            Other Cardiac Procedure Deaths

0(0.0%)

0(0.0%)

2(0.6%)

            Hyperkalemic Deaths

1(0.2%)

0(0.0%)

0(0.0%)

      Deaths Within 30 Days of Randomization

7(1.2%)

11(1.8%)

3(1.0%)

 


 

TABLE 3: Median Days to Mortality*

 

 

CRT-D

CRT-P

OPT

 

All Deaths

258.0

236.0

253.0

Cardiac Deaths

265.5

205.0

215.5

Pump Failure Deaths

282.0

289.0

182.5

Sudden Cardiac Deaths

341.0

186.5

253.0

*Includes only subjects who died

 

 

TABLE 4: Mortality Rate Adjusted for Time Observed

 

 

CRT-D

CRT-P

OPT

 

Total Subjects

595

617

308

Cumulative Follow-Up (Years)

777.2

812.9

374.8

Deaths/100 Subject-Years

13.5

16.1

20.5

Cardiac Deaths/100 Subject-Years

9.8

13.4

15.5

Pump Failure Deaths/100 Subject-Years

6.7

6.5

9.1

Sudden Cardiac Deaths/100 Subject-Years

2.2

5.9

4.8

 

 

An evaluation of all-cause mortality with respect to subject baseline characteristics was performed (TABLE 5).  For continuous variables, data was separated into quartiles for analysis by treatment assignment.  Overall, CRT-D was associated with substantial improvement in all-cause mortality across the majority of baseline characteristic subgroups.  However, it should be noted that improvement in survival associated with CRT-D therapy diminished markedly with improving EF at baseline.  This is biologically plausible as subjects with a higher EF would on average be healthier, and thus less able to benefit from the intervention.  With respect to baseline conduction abnormality, the majority of benefit was associated with subjects who had LBBB, and there was a slight increase in mortality associated with both CRT-D and CRT-P as compared to OPT in subjects with IVCD.  Based on the relatively small size of the IVCD subgroup, small differences in mortality, and the issue of multiple comparisons, it is not reasonable to conclude that CRT-D caused an increase in mortality in this subgroup.  The only other subgroup with an increased mortality associated with CRT-D was in the first quartile of PR interval duration.  A consistent increase in mortality can be seen in the OPT cohort with increasing PR interval, which is biologically plausible as PR interval prolongation may be an indicator of worsening cardiac disease.  Therefore, it may be reasonable to believe that this subgroup is less likely to benefit from CRT-D than the trial population as a whole, due to being more healthy and thus less able to benefit from the therapy.

TABLE 5: All-Cause Mortality and Baseline Characteristics

 

 

CRT-D

CRT-P

OPT

Age (years)

 

 

 

      Q1*(20-58) 

20/150(13.3%)

29/163(17.8%)

11/61(18.0%)

      Q2 (59-66)

28/156(17.9%)

29/141(20.6%)

17/74(23.0%)

      Q3 (67-74)

24/150(16.0%)

30/160(18.8%)

23/97(23.7%)

      Q4 (75-92)

33/139(23.7%)

43/153(28.1%)

26/76(34.2%)

Gender

 

 

 

      Male

83/401(20.1%)

108/415(26.0%)

59/211(28.0%)

      Female

22/194(11.3%)

23/202(11.4%)

18/97(18.6%)

NYHA Functional Class

 

 

 

      Class III

68/512(13.3%)

95/537(17.7%)

47/253(18.6%)

      Class IV

37/83(44.6%)

36/80(45.0%)

30/55(54.5%)

Etiology of Cardiomyopathy

 

 

 

      Ischemic

77/325(23.7%)

80/331(24.2%)

54/181(29.8%)

      Non-Ischemic

28/270(10.4%)

51/285(17.9%)

23/127(18.1%)

Ejection Fraction

 

 

 

      Q1 (5-17)

27/153(17.6%)

54/169(32.0%)

25/75(33.3%)

      Q2 (18-21)

30/138(21.7%)

29/160(18.1%)

20/74(27.0%)

      Q3 (22-28)

29/160(18.1%)

30/155(19.4%)

19/76(25.0%)

      Q4 (29-35)

19/144(13.2%)

18/133(13.5%)

13/83(15.7%)

Conduction Abnormality

 

 

 

      LBBB

67/434(15.4%)

81/426(19.0%)

55/215(25.6%)

      RBBB

15/61(24.6%)

22/76(28.9%)

8/27(29.6%)

      IVCD

23/100(23.0%)

28/114(24.6%)

14/66(21.2%)

QRS Duration (ms)

 

 

 

      Q1 (120-140)

30/144(20.8%)

47/177(26.6%)

20/93(21.5%)

      Q2 (141-160)

37/203(18.2%)

35/184(19.0%)

25/92(27.2%)

      Q3 (161-175)

17/102(16.7%)

13/94(13.8%)

13/58(22.4%)

      Q4 (176-276)

21/146(14.4%)

36/162(22.2%)

19/65(29.2%)

PR Interval (ms)

 

 

 

      Q1 (150-178)

27/143(18.9%)

23/161(14.3%)

12/84(14.3%)

      Q2 (179-200)

30/205(14.6%)

43/201(21.4%)

22/97(22.7%)

      Q3 (201-221)

15/87(17.2%)

23/111(20.7%)

14/52(26.9%)

      Q4 (222-560)

33/160(20.6%)

42/144(29.2%)

29/75(38.7%)

Heart Rate (beats/min)

 

 

 

      Q1 (44-64)

24/172(14.0%)

32/189(16.9%)

20/92(21.7%)

      Q2 (65-72)

27/143(18.9%)

29/129(22.5%)

20/69(29.0%)

      Q3 (73-80)

23/131(17.6%)

29/154(18.8%)

17/83(20.5%)

      Q4 (81-122)

31/149(20.8%)

41/145(28.3%)

20/64(31.3%)

Q1 = first quartile, Q2 = second quartile, Q3 = third quartile, Q4 = fourth quartile
All-Cause Hospitalization

 

The total number of hospitalizations and the total amount of time spent in the hospital were not endpoints for this trial.  However, they are clinically meaningful, and relevant to the concerns of patients and physicians.  It has been claimed that the index hospitalization for the device implant procedure, as well as the additional hospitalizations following failed implant attempts, are not recurring events and should therefore be excluded for the purposes of calculating the rate of “all cause hospitalization.”  While they are not recurring events with the same frequency as heart failure hospitalizations, they none the less need to be considered if one is to understand what happened to subjects during the trial with respect to all causes of hospitalization.  One of the presumed strengths of a claim of reduction in all cause hospitalization (or time to hospitalization) is that one would spend less time in the hospital.  If in fact there is simply an exchange from heart failure hospitalizations to device hospitalizations for implants, reimplants, and adverse events related to the implant procedure or device, the clinical benefit is far from clear.  In addition, while it is true that heart failure hospitalizations recur at a much higher rate than implant procedures, the strength of the data is not compelling with respect to understanding an annualized hospitalization rate anyway.  Subjects in the CRT-D group were followed on average approximately 1.29 years, and in the OPT group approximately 1.01 years (only considering the follow-up that occurred up to the time of withdrawal).  In addition, only 122 of the original 903 CRT-D and OPT subjects had 2 years of follow-up, and no subjects had 3 years of follow-up.  This should give some pause to the interpretation of annualized hospitalization rates based on this trial, regardless of whether the implant procedures are included.

 

Finally, it has also been claimed that one should not calculate the overall number of hospitalizations for the CRT-D and OPT cohorts due to the issue of competing risks of death and hospitalization.  Therefore, it may be that the CRT-D device keeps people alive who would otherwise have died, therefore enriching the CRT-D cohort over time with sicker subjects who may be more likely to require hospitalization.  While this is a reasonable concern and may very well have occurred during COMPANION, it does not undermine the value of such an analysis.  This is because whatever hospitalization rate occurred during the trial, it represents what the subjects experienced, regardless of the underlying cause.  It is only if one were to attempt to assign a specific cause for an increase in hospitalizations associated with the device, such as a direct effect of the CRT-P component, that the issue of competing risks would apply.

 

An analysis was performed which included all hospitalizations documented by the sponsor (i.e., hospitalizations for which there was a date change, including elective device hospitalizations, all index hospitalizations, all re-implant attempt hospitalizations, and all outpatient inotropic infusions for worsening heart failure).  All events from randomization to subject withdrawal, death, or November 30, 2002, (whichever came first) were included.  Hospitalizations that were not associated with a date change or were not associated with an implant attempt were not included as these events were not captured as hospitalizations during the trial.

 

Within the CRT-D cohort, there were 1548 hospitalizations (mean duration 5.42 +/- 8.82 days, median 3.0 days, minimum 1 day, maximum 166 days).  A total of 314 of these hospitalizations were attributed to heart failure.  Within the OPT cohort, there were 507 hospitalizations (mean duration 6.57 +/- 7.32 days, median 4.0 days, minimum 1 day, maximum 88 days).  A total of 216 of these hospitalizations were attributed to heart failure.

 

However, the preceding description does not take into consideration the difference in overall duration of follow-up between the two cohorts.  For CRT-D, the mean duration of follow-up was 469.3 +/- 233.3 days, median 471 days, minimum 2 days, and maximum 1020 days.  For OPT, the mean duration of follow-up was 370.5 +/- 241.0 days, median 356 days, minimum 1 day, maximum 1042 days.

 

Therefore, the mean all-cause hospitalization rate in the CRT-D cohort was 2.02 hospitalizations/year, and in the OPT cohort 1.62 hospitalizations/year.  The mean days spent hospitalized for any cause in the CRT-D cohort was 11.0 days/year, and in the OPT cohort 10.7 days/year. 

 

Listed in TABLE 6 is the frequency of hospitalizations subjects experienced in the two cohorts.

 

 

TABLE 6: Frequency of Hospitalizations by Cohort

 

# OF HOSPITALIZATIONS

CRT-D

(n = 595)

OPT

(n = 308)

 

1

199 (33.4%)

71 (23.1%)

2

160 (26.9%)

43 (14.0%)

3

96 (16.1%)

26 (8.4%)

4

56 (9.4%)

20 (6.5%)

5

37 (6.2%)

15 (4.9%)

6

19 (3.2%)

4 (1.3%)

> 6

23 (3.9%)

10 (3.2%)

 

 

Cardiac Morbidity

 

Cardiac morbidity was a secondary endpoint for the trial.  The definition that was provided in the protocol was the occurrence of the following events:

 

  1. Worsening heart failure resulting in use of intravenous vasoactive or inotropic therapy exceeding four hours
  2. Mechanical respiratory or cardiac support
  3. Any cardiac surgery, including heart transplant
  4. Resuscitated cardiac arrest or sustained ventricular tachycardia requiring intervention (e.g., chest thump, external cardioversion, or external defibrillation)
  5. Hospitalization for acute decompensation of heart failure
  6. Hospitalization that results in death from cardiac causes
  7. Significant device-related events resulting in
    1. Permanent disability
    2. Hospitalization for pending death or permanent disability

 

However, the cardiac morbidity endpoint was analyzed by hospitalizations, not by cardiac morbid events as described in the protocol.  Therefore, each hospitalization was determined to be “cardiac morbid” if any of the above events occurred, and counted as a single event regardless of the number of events that took place during the hospitalization.  This change was not documented within a protocol revision, and the FDA was not notified.  The datasets provided to FDA did not allow the possibility of calculating the endpoint as described in the protocol.

 

In addition, the inclusion of external cardioversion/defibrillation in the definition of cardiac morbidity is problematic.  This is because subjects with an ICD would not be anticipated to receive external cardioversion/defibrillation, as the device already provides internal cardiac defibrillation.  If one is to assume that the delivery of such an intervention signals a worsened clinical status, then the episodes of cardioversion/defibrillation by the device should also be included.  However, this would also be problematic as the device is continuously monitoring cardiac rhythm, and would therefore be likely to cardiovert/defibrillate on many occasions that would otherwise not have been captured in the absence of a device.  Due to the multiple difficulties in interpretation, criterion #4 should not have included external cardioversion/defibrillation.

 

Based on the definition of cardiac morbidity used by the sponsor, there were 368 cardiac morbid events in the CRT-D cohort, and 312 in the OPT cohort.  The number of times cardiac event descriptions were reported is provided (TABLE 7).  However, single events could have multiple descriptions within TABLE 7.  Therefore, no percentages were calculated, and this data must be interpreted with caution.

 

 

TABLE 7: Reported Descriptions of Cardiac Morbid Events

 

Description

CRT-D

(n=595)

OPT

(n=308)

 

Any cardiac surgery

33

88

Hospitalization for acute decompensation of HF

312

215

Hospitalization resulting in death from cardiac causes

27

24

Mechanical respiratory/cardiac support

33

17

Resuscitation for cardiac arrest/sustained VT

13

5

Worsening HF resulting in vaso/inotropic tx >4 hrs (Inpat)

203

166

Worsening HF resulting in vaso/inotropic tx >4 hrs (Outpt)

9

14

 

 

Implant Hospitalizations

 

Characterization of the implant hospitalization was not an endpoint for this trial.  However, as subjects must undergo this hospitalization to receive the device, it was felt to be reasonable to provide descriptive statistics related to these events.  Of the 595 CRT-D subjects, 541 underwent a successful implant procedure, 47 had unsuccessful attempts, and 7 were intents (randomized to CRT-D, but no implant procedure ever occurred).  Additionally, within the CRT-D cohort there were a total of 588 subjects who underwent 641 implant hospitalizations.  The mean duration of these hospitalizations was 2.91 +/- 4.93 days, median 2.0 days, minimum 1.0 day, maximum 116 days).  Of the 588 CRT-D subjects who underwent an implant attempt, 537 underwent 1 attempt, 49 underwent 2 attempts, and 2 underwent 3 attempts.

 

Additional Efficacy Data

 

Additional efficacy data collected for all subjects during the trial included a six-minute walk, global assessment (composed of physician and subject assessments combined), NYHA functional class, and Minnesota Living With Heart Failure Questionnaire (MLWHFQ).  However, as of June 14, 2004, only the NYHA functional class and MLWHFQ data have been submitted for FDA review (TABLE 8).  These results appear to suggest an improvement in both NYHA functional class and MLWHFQ scores associated with the device.  However, it should be noted that (1) these are subjective assessments performed in the context of an unblinded trial, (2) there is a large amount of missing data, particularly at 6 months, and (3) the missing data is heavily weighted towards the OPT arm (TABLE 9).

 

 

TABLE 8: Mean NYHA and MLWHFQ Scores

 

 

CRT-D

OPT

 

NYHA             Baseline

3.14 +/- 0.35

3.18 +/- 0.38

                        3 Months

2.49 +/- 0.70

2.95 +/- 0.66

                        6 Months

2.44 +/- 0.71

2.78 +/- 0.69

 

 

 

MLWHFQ      Baseline

62.7 +/- 22.9

60.6 +/- 22.7

                        3 Months

38.8 +/- 26.2

51.5 +/- 26.4

                        6 Months

37.5 +/- 26.5

47.6 +/- 26.9

 


 

TABLE 9: NYHA and MLWHFQ Missing Data

 

 

CRT-D

(n=595)

OPT

(n=308)

 

NYHA             Baseline

0 (0%)

0 (0%)

                        3 Months

52 (8.7%)

66 (21.4%)

                        6 Months

97 (16.3%)

109 (35.4%)

 

 

 

MLWHFQ      Baseline

16 (2.7%)

9 (2.9%)

                        3 Months

77 (12.9%)

65 (21.1%)

                        6 Months

115 (19.3%)

101 (32.8%)

 

 

Medication Use

 

The percentage of subjects on various cardiac medications at the time of their final study visit was analyzed by treatment group (TABLE 10).  The proportion of subjects on these medications appears to be well matched.

 

 

TABLE 10: Medication Use at Final Study Visit

 

 

CRT-D

(n=556)

OPT

(n=272)

 

ACEI/ARB Use

 460 (82.7%)

 237 (87.1%)

Beta-Blocker Use

 416 (74.8%)

 191 (70.2%)

Aldosterone Antagonist Use

 257 (46.2%)

 133 (48.9%)

Digoxin Use

 380 (68.3%)

 183 (67.3%)

Diuretic Use

502 (90.3%)

 251 (92.3%)

 

Safety

 

There were no safety endpoints for the CONTAK CD system, only for the EASYTRAK left ventricular lead (P010012, approved May 2, 2002).  The sponsor therefore proposed a post hoc safety endpoint of system-related complication-free rate.  However, as there was no pre-specified safety endpoint, it is more reasonable to compare all adverse events (AEs) between the CRT-D and OPT cohorts.  In addition, the system-related complication-free rate did not include subjects with failed implant attempts.  As the events surrounding a failed implant would not have occurred if the subject had received optimal pharmacologic therapy only, it is not reasonable to exclude them from a safety analysis.

 

All AEs through November 30, 2002, were included.  It should be noted that the adverse event data in this review was not adjusted for the difference in observation time between the two cohorts (mean observation time for CRT-D subjects, 1.29 years, and for OPT subjects, 1.01 years).  There were a total of 3732 AEs in the CRT-D arm and 1230 in the OPT arm.

 

An observation was defined as a clinical adverse event that was correctable by noninvasive measures.  This included reprogramming of the pulse generator as well as antibiotic treatment of a pocket infection.  A complication was defined as a clinical adverse event which required invasive measures to correct, such as evacuation of a hematoma or lead dislodgment.  Of the CRT-D AEs, 540 (14.5%) were complications and 3192 (85.5%) observations.  Of the 1230 AEs in the OPT cohort, 241 (19.6%) were complications and 989 (80.4%) observations.

 

The number of AEs experienced by subjects was evaluated (TABLE 11).  For example, 202 subjects experienced between 1 and 3 AEs during the trial.  A total of 562 of the 595 subjects in CRT-D arm experienced at least one AE, while 247 of 308 subjects in the OPT arm experienced at least one AE.

 

 

TABLE 11: Number of AEs Experienced by Cohort

 

Number of AEs

CRT-D

(n=595)

OPT

(n=308)

 

1-3

202 (33.9%)

118 (38.3%)

4-6

150 (25.2%)

63 (20.5%)

7-9

100 (16.8%)

32 (10.4%)

>9

110 (18.5%)

34 (11.0%)

TOTAL

562 (94.5%)

247 (80.2%)

 

The sponsor provided an adjudicated classification of the AEs (TABLE 12).  However, this adjudication allowed for events such as heart failure episodes and ventricular arrhythmias to be classified in multiple different ways.  For instance, congestive heart failure could be classified as shortness of breath, volume overload, physiological reaction, pulmonary edema, loss of left ventricular lead capture, or congestive heart failure.  Therefore, this data should be interpreted with caution.

 

With respect to arrhythmias, there was a marked increase associated with CRT-D.  However, this was likely due to ascertainment bias due to the device’s ability to capture such events.  In addition, there has been concern that ICD therapy may be associated with an increased risk of depression or anxiety.  While there were episodes of depression and anxiety specifically attributed to ICD therapy during COMPANION, the overall reported rate of episodes in the CRT-D cohort (36 episodes/595 subjects, 6.1%), was fairly similar to OPT (14 episodes/308 subjects, 4.5%).

 

 

TABLE 12: Adverse Events by Cohort

 

Sponsor Adjudicated

AE Classifications

 

CRT-D

(n=595)

OPT

(n=308)

 

AV Block - heart block, complete

10 (1.7%)

1 (0.3%)

Aneurysm

0 (0%)

2 (0.6%)

Angina

0 (0%)

6 (1.9%)

Allergic reaction

12 (2.0%)

0 (0%)

Altered hemodynamic status

1 (0.2%)

0 (0%)

Arrhythmia

20 (3.4%)

9 (2.9%)

Arrhythmia - atrial fibrillation

74 (12.4%)

21 (6.8%)

Arrhythmia - atrial flutter

12 (2.0%)

3 (1.0%)

Arrhythmia - sinus tachycardia

8 (1.3%)

2 (0.6%)

Bleeding/fluid accumulation

6 (1.0%)

0 (0%)

Brady capture - LV

55 (9.2%)

2 (0.6%)

Brady capture - RV

10 (1.7%)

0 (0%)

Brady capture - atrium

16 (2.7%)

1 (0.3%)

Bruise

1 (0.2%)

0 (0%)

Cancer, other

14 (2.4%)

12 (3.9%)

Cardiac arrest

7 (1.2%)

4 (1.3%)

Cardiac tamponade

2 (0.3%)

0 (0%)

Change in arrhythmia

1 (0.2%)

0 (0%)

Change in arrhythmia - brady

46 (7.7%)

12 (3.9%)

Change in arrhythmia - junctional

7 (1.2%)

1 (0.3%)

Change in arrhythmia - SVT

49 (8.2%)

8 (2.6%)

Change in arrhythmia - VT

0 (0%)

1 (0.3%)

Change in drug therapy

1 (0.2%)

0 (0%)

Change in physical status

59 (9.9%)

17 (5.5%)

Chest pain

202 (33.9%)

90 (29.2%)

Chronotropic incompetence

1 (0.2%)

0 (0%)

Claudication

1 (0.2%)

2 (0.6%)

Coagulopathy

6 (1.0%)

2 (0.6%)

Congestive heart failure

391 (65.7%)

251 (81.5%)

Diabetes complications

2 (0.3%)

1 (0.3%)

Digoxin level

1 (0.2%)

0 (0%)

Dissection, coronary sinus

15 (2.5%)

3 (1.0%)

Distal thromboemboli

1 (0.2%)

0 (0%)

Dizziness, cause undetermined

92 (15.5%)

40 (13.0%)

Dysfunctional dialysis catheter

2 (0.3%)

0 (0%)

Dyspnea

165 (27.7%)

67 (21.8%)

ERI, premature (early)

1 (0.2%)

0 (0%)

Edema

23 (3.9%)

10 (3.2%)

Electrolyte/lab

85 (14.3%)

30 (9.7%)

Endocrine

91 (15.3%)

23 (7.5%)

Epistaxis

8 (1.3%)

5 (1.6%)

Fatigue

62 (10.4%)

24 (7.8%)

Febrile

10 (1.7%)

3 (1.0%)

Finishing wire stripped

1 (0.2%)

0 (0%)

Fistula

1 (0.2%)

0 (0%)

GI

243 (40.8%)

54 (17.5%)

GI bleed

23 (3.9%)

7 (2.3%)

Genitourinary

16 (2.7%)

13 (4.2%)

Helix related (screw tip), broken or stretched

1 (0.2%)

0 (0%)

Hematoma

44 (7.4%)

3 (1.0%)

Hematuria

6 (1.0%)

2 (0.6%)

Hemoglobin drop

29 (4.9%)

10 (3.2%)

Hemorrhage

4 (0.7%)

0 (0%)

High DFTs - tachy

4 (0.7%)

0 (0%)

Hypertension

19 (3.2%)

7 (2.3%)

Hypotension

90 (15.1%)

26 (8.4%)

Inappropriate shock above rate cutoff

59 (9.9%)

2 (0.6%)

Inappropriate shock due to oversensing

19 (3.2%)

0 (0%)

Infection

125 (21.0%)

31 (10.1%)

Insulation breach suspected

1 (0.2%)

0 (0%)

Insulation breech observed

1 (0.2%)

0 (0%)

Insulation problem, cut or damage

1 (0.2%)

0 (0%)

Intermittent sensing - atrium pace - brady

1 (0.2%)

0 (0%)

Ischemia

6 (1.0%)

1 (0.3%)

Malfunction, memory problem

1 (0.2%)

0 (0%)

Malfunction, tone/alarms, appropriate

1 (0.2%)

0 (0%)

Malfunction, tone/alarms, inappropriate

1 (0.2%)

0 (0%)

Materials unretrieved in body

1 (0.2%)

0 (0%)

Migration of device

2 (0.3%)

0 (0%)

Multiple counting - tachy

30 (5.0%)

0 (0%)

Muscle stimulation

2 (0.3%)

0 (0%)

Myocardial infarction

13 (2.2%)

7 (2.3%)

Neurologic

34 (5.7%)

14 (4.5%)

Occlusion

1 (0.2%)

0 (0%)

Oversensing - atrium pace - tachy

1 (0.2%)

0 (0%)

Oversensing - atrium pace sense

2 (0.3%)

0 (0%)

Oversensing - vent. pace - tachy

2 (0.3%)

0 (0%)

Oversensing - ventricle pace sense

3 (0.5%)

0 (0%)

Pacemaker syndrome

1 (0.2%)

0 (0%)

Pacemaker-mediated tachycardia (PMT)

13 (2.2%)

1 (0.3%)

Pain

185 (31.1%)

72 (23.4%)

Palpitations

21 (3.5%)

4 (1.3%)

Pectoral muscle stimulation

1 (0.2%)

0 (0%)

Perforation, coronary venous

5 (0.8%)

2 (0.6%)

Perforation, venous

2 (0.3%)

0 (0%)

Pericardial effusion

2 (0.3%)

0 (0%)

Pericarditis

3 (0.5%)

0 (0%)

Phantom shock

3 (0.5%)

0 (0%)

Phrenic nerve/diaphragm stimulation

98 (16.5%)

1 (0.3%)

Physical trauma

70 (11.8%)

15 (4.9%)

Physiological reaction

145 (24.4%)

51 (16.6%)

Placement difficulty, stylet related

1 (0.2%)

0 (0%)

Pleural effusion

15 (2.5%)

2 (0.6%)

Pleurisy

2 (0.3%)

0 (0%)

Pneumonia (respiratory infection)

0 (0%)

4 (1.3%)

Pneumothorax

11 (1.8%)

5 (1.6%)

Pocket erosion/extrusion

3 (0.5%)

0 (0%)

Pocket infection

22 (3.7%)

1 (0.3%)

Post-surgical wound discomfort

60 (10.1%)

1 (0.3%)

Pseudoaneurysm

1 (0.2%)

0 (0%)

Psychological effects

51 (8.6%)

23 (7.5%)

Pulmonary edema

10 (1.7%)

3 (1.0%)

Pulmonary embolism

3 (0.5%)

2 (0.6%)

Rate response insufficient - brady

2 (0.3%)

0 (0%)

Renal failure

104 (17.5%)

31 (10.1%)

Respiratory arrest

6 (1.0%)

0 (0%)

Respiratory distress

2 (0.3%)

0 (0%)

Respiratory failure

11 (1.8%)

2 (0.6%)

Respiratory related

256 (43.0%)

101 (32.8%)

Sensing T-wave - tachy

1 (0.2%)

0 (0%)

Sensing myopotential - tachy

1 (0.2%)

0 (0%)

Set screw, not tight

1 (0.2%)

0 (0%)

Shock

3 (0.5%)

2 (0.6%)

Statement from witness - HCP

1 (0.2%)

0 (0%)

Stroke syndrome or CVA

12 (2.0%)

3 (1.0%)

Syncope

27 (4.5%)

12 (3.9%)

Thrombosis

14 (2.4%)

4 (1.3%)

Transient heart block

3 (0.5%)

1 (0.3%)

Transient ischemic attack (TIA)

5 (0.8%)

3 (1.0%)

Tumors, growths

4 (0.7%)

2 (0.6%)

Ulceration

3 (0.5%)

2 (0.6%)

Undersensing - atrium pace sense - brady

4 (0.7%)

0 (0%)

Undersensing - vent. pace sense - tachy

1 (0.2%)

0 (0%)

Vascular related

21 (3.5%)

8 (2.6%)

Ventricular fibrillation

17 (2.9%)

2 (0.6%)

Ventricular tachycardia

115 (19.3%)

32 (10.4%)

Volume overload

30 (5.0%)

6 (1.9%)

 


 

Dehydration appeared to be more common in the CRT-D cohort.  In particular, there were 24 episodes of renal failure attributed to dehydration in the CRT-D cohort compared to 1 episode in the OPT cohort.  Given the presumed mechanism of action of CRT-P in improving cardiac function, it is plausible that dehydration might be more common with this intervention if the diuretic dose were not decreased.  TABLE 13 lists the sponsor’s adjudication with the number of episodes attributed to dehydration for CRT-D and OPT.

 

TABLE 13: Episodes of Dehydration

 

Sponsor Adjudication

CRT-D

(n=595)

OPT

(n=308)

 

Renal failure

24 (4.0%)

1 (0.3%)

Electrolyte/lab

10 (1.7%)

4 (1.3%)

Physiological reaction

10 (1.7%)

2 (0.6%)

Hypotension

20 (3.4%)

9 (2.9%)

Dizziness, cause unspecified

10 (1.7%)

3 (1.0%)

Physiological effect

1 (0.2%)

0 (0%)

Syncope

7 (1.2%)

1 (0.3%)

Fatigue

6 (1.0%)

2 (0.6%)

TOTAL

88 (14.8%)

22 (7.1%)

 

AGREEMENT MEETINGS

 

The preceding analyses are consistent with the agreements reached between the sponsor and FDA during the Agreement Meetings as described in the FDA letters dated September 8, 1999, and November 14, 2002.  The latter agreement letter included the COMPANION Trial Revised Analysis Plan dated November 1, 2002.  Neither agreement makes any reference to whether the CRT-D and CRT-P data could be compared, or of using any available CRT-P data in the analysis of the CRT-D device.  In fact, the COMPANION protocol on page 17 states, “Descriptive statistics will be used to describe any similarity or difference between CONTAK TR and CONTAK CD.”  It should be noted that descriptive statistics were used exclusively in this review.