1. DSMB Monitoring of the Therapeutic Hypothermia After Pediatric
Cardiac Arrest (THAPCA) Trials
  Presented by: R Holubkov (Salt Lake City, USA) Co-authors: AE Clark, A Atz, M Gausche-Hill, D Glidden, B Slomine, J Christensen, A Webster,, K Page, JM Dean, F Moler On behalf of: The Therapeutic Hypothermia After Pediatric Cardiac Arrest in Children (THAPCA) Study  

2. Funding
Supported by several grants from the United States National Heart, Lung, and Blood Institute (NHLBI), National Institutes of Health (NIH).
Contents are solely the responsibility of the authors and do not necessarily represent the official views of NHLBI or NIH.

3. Background
Cardiac arrest in children often results in death or in poor long-term functional outcome in survivors.
The THAPCA trials were designed to assess benefit of therapeutic hypothermia on functional outcome among children (in whom randomized evaluations were lacking) 1 year after cardiac arrest
Two separate, parallel trials among children arresting out of the hospital and those arresting in hospital (different arrest etiologies; interventions can be more rapidly initiated for children arresting in-hospital).

4. Subjects and Interventions
40 children’s hospital PICUs in U.S., Canada, and Great Britain
Children 3d - 17 yrs old, consented within 6 hours of return of circulation after cardiac arrest, randomized 1:1 to one of two five-day regimens of active temperature management:
Therapeutic hypothermia (target temp 33.0°C for first 48 hrs, followed by rewarming to 36.8°C), -or-
Therapeutic normothermia (target temp 36.8°C for all 5 days)

5. Primary Outcome and Sample Size
Primary outcome: One-year survival with Vineland Adaptive Behavior Scale-II (VABS) ≥ 70; this cutpoint reflects functional status no worse than 2 standard deviations below reference population mean of 100.
Out of hospital: Assume hypothermia improves good outcome rate by 20%, from roughly 20% to 40%. Need N=250 evaluable children enrolled
In hospital: Assume hypothermia improves good outcome rate by 15%, from roughly 45% to 60%. Need N=504 evaluable children enrolled

6. DSMB NIH-appointed 7-member DSMB
Met at 6-month intervals throughout enrollment
Interim superiority monitoring using symmetric O’Brien-Fleming boundaries with Lan-DeMets implementation
Interim futility monitoring per DSMB’s discretion, with 20% conditional power criterion
Two trials treated as “separate and parallel”
Unblinding to study arms per DSMB’s discretion, encouraged by NHLBI recommendations.

7. Recruitment
THAPCA was initiated at pilot sites in September 2009, with rollouts at other sites in following year.
Enrollment was much faster for the out of hospital than the in-hospital setting, as you will see.
Let’s look at recruitment concurrently with what occurred at each of the DSMB meetings:

8. Recruitment over Time

9. Five Initial Safety-Only Reviews
FEB 2010:
N=20
(DATA COMBINED
ACROSS TRIALS)
FEB 2011:
N=71 OH
N=43 IH
JULY 2010:
N=35 OH
N=23 IH
MAY 2012:
N=217 OH, N=135 IH
OCT 2011: N=148 OH, N=84 IH

10. Nov 2012: First Efficacy Look
NOV 2012: IN HOSPITAL
N=74 with 1-Year Data
OUT OF HOSPITAL
N=141 with 1-year Data

11. Nov 2012 Efficacy Look: In-Hospital
Assigned to
Treatment A
Treatment B
p-value
(age- stratified)
Alive at One Year With
VABS-II ≥ 70
15/39 (38%)
12/35 (34%)
0.57
14.7% of final information available. Proceed with trial.
As enrollment very slow, DSMB elected not to examine efficacy (or safety) data for In-hospital trial next time. “Skip a meeting for the In-hospital trial”

12. Nov 2012 Efficacy Look: Out of Hospital
Assigned to
Treatment C
Treatment D
p-value
(age- stratified)
Alive at One Year With
VABS-II ≥ 70
17/74 (23%)
6/67 (9%)
0.03
With 56.4% of final information available, O’Brien-Fleming boundaries recommend stopping if p< Proceed with trial.
DSMB did not choose to unmask identity of treatments.
DSMB to examine efficacy again at next meeting.

13. Aug 2013: Out of Hospital Efficacy Only
N=213 with 1-year Data

14. Aug 2013 Efficacy Look: Out of Hospital
Assigned to
Treatment C
Treatment D
p-value
(age- stratified)
Alive at One Year With
VABS-II ≥ 70
22/111 (20%)
12/102 (12%)
0.12
85.2% of final information available. “Proceed with trial.”
Enrollment was already complete! Stopping for efficacy at this meeting would not have affected conduct of this trial.
Final published results with N=260: Hypothermia 27/138 (20%), Normothermia 15/122 (12%), p=0.14

15. Mar 2014 In-Hospital Efficacy Look
N=214 Eligible Enrolled
N=149 with 1-Year Data

16. Mar 2014 Efficacy Look: In-Hospital
Assigned to
Treatment A
Treatment B
p-value
(age- stratified)
Alive at One Year With
VABS-II ≥ 70
26/74 (35%)
29/75 (34%)
0.83
29.6% of final information available. Proceed with trial.
Stopping for futility was not considered during this meeting.

17. Jan 2015 In-Hospital Efficacy Look
N=254 Eligible Enrolled
N=193 with 1-Year Data

18. Jan 2015 Efficacy Look: In-Hospital
Assigned to
Treatment A
Treatment B
p-value
(age- stratified)
Alive at One Year With
VABS-II ≥ 70
36/95 (38%)
38/98 (39%)
0.97
38% of final information available for target n.
Clearly, target n would not be reached. Optimistic n ~ 364.
95% CI for Arm B Tx Effect: (-12.8%, 14.6%) Recall original Effect Estimate was 15%.
Futility examined under realistic/optimistic scenarios:

19. Conditional Power Assuming final n=364
If true effect 0.9% (observed), conditional power <1%
If true effect 10%, conditional power 5%
If true effect 15% (hypothesized), conditional power 17%

20. Range of Conditional Power Estimates
Assumed Final
Analyzed N
ESTIMATED
CONDITIONAL
POWER if true
15% effect
10% effect
ESTIMATED PREDICTIVE POWER
GIVEN CURRENT DATA
320 7% 2% 1%
330 9% 3%
340
11%
350
13% 4%
360
16% 5%
364
17%
370
19% 6%
380
21%
390
24% 8%

21. Jan 2015 DSMB Meeting
DSMB extensively debated low conditional power and stopping rationales/implications.
Initial DSMB recommendation to NIH: Do not stop trial (power near 20% in very optimistic scenarios; trial results potentially useful for other outcomes)
NIH did not accept DSMB recommendations, reconvening the DSMB one month later.

22. February 2015 DSMB re-convened by NIH
Per NIH request, DCC calculated conditional power for In-Hospital trial for ad hoc outcomes, such as ranking survival times and one-year function, and doing a global rank-based test.
At this meeting, the DSMB did recommend terminating recruitment in the In Hospital trial, primarily due to perceived futility.

23. Summary: DSMB Actions
DSMB did not unblind to treatment identity throughout their tenure (also seen in other trials our DCC coordinates)
DSMB skipped a scheduled In Hospital interim analysis due to slow enrollment
DSMB was hesitant to terminate an underenrolling, thus-far-negative trial based on low conditional power alone.

24. Summary: THAPCA Design
Two “negative trials” – one enrolling at full speed, other systematically underenrolling
Fundamental basis for conservative efficacy monitoring boundaries demonstrated in Out of Hospital trial, ultimately negative though p=0.03 with >50% of target n.
Utility of conditional power criteria for futility monitoring, which motivates debate among DSMB members, was demonstrated in the In-Hospital trial.

25. Supplemental Slides

26. Additional Simulations: NIH request
Out of Hospital trial had trend of longer survival among hypothermia-treated children. What if such a trend also held in the In Hospital setting? (ad hoc analysis)
Sample, with replacement, from Out of Hospital data:
survival times for children who died
1-year VABS for survivors.
Under optimistic In Hospital assumptions (e.g., 50% total hypothermia survival versus 40% with normothermia), decent power could be eked out with optimistic final n.

27. Additional Power Simulations: NIH request
Scenario
Power
N=320
N=330
N=340
N=350
N=360
N=370
N=380
N=390
N=400
N=410
N=420
Primary Population, Mann-Whitney
59%
60%
62%
63%
66%
65%
67%
68%
70%
Primary Population,
Age-Stratified
50%
51%
52%
55%
56%
58%
57%
61%
Entire Population,
Mann-Whitney
71%
72%
73%
74%
76%
77%
78%
80%
81%
82%
75%

28. Original Power Calculations