1. CLEAR III Monthly Broadcast January 2012

3. Agenda Enrollments and Announcements…………………..Bing Cao CLEAR III Goals for 2012…………………………Natalie Ullman Impact of Adaptive Randomization…….Gayane Yenokyan

4. CLEAR III RECENT ENROLLMENTS December 2011- January 2012

5. Congratulations to December 2011-January 2012 Enrollments Dr. Sagi Harnof (PI), Yulia Wasserman, Tali Kimchi Shiri Fischler (coordinators), Chaim Sheba Medical Center 10 th enrollment on 12-15-2011 (the trial's 188 th enrollment)

6. Congratulations to December 2011-January 2012 Enrollments Dr. George Lopez (PI), Kari Langford (coordinators), University of Texas, Houston 6 th enrollment on 12-21-2011 (the trial's 189 th enrollment)

7. Congratulations to December 2011-January 2012 Enrollments Dr. Opeolu Adeoye (PI), Becky Reinert (coordinators), University of Cincinnati 5 th enrollment on 12-22-2011 (the trial's 190 th enrollment)

8. Congratulations to December 2011-January 2012 Enrollments Dr. Venkatesh Aiyagari(PI), Maureen Hillmann(coordinator), University of Illinois at Chicago 4 th enrollment on 12-30-2011 (the trial's 191 st enrollment)

9. Congratulations to December 2011-January 2012 Enrollments Dr. Fuat Arikan (PI), Alba Peña Paris (coordinators), Vall d'Hebron University Hospital, Barcelona 4 th enrollment on 1-04-2012 (the trial's 192 nd enrollment)

10. Congratulations to December 2011-January 2012 Enrollments Dr. R. Scott Graham (PI), Dr. Randall Merchant (coordinators), Virginia Commonwealth University 3 rd enrollment on 1-05-2012 (the trial's 193 rd enrollment)

11. Congratulations to December 2011-January 2012 Enrollments Dr. David B. Seder (PI), Barbara McCrum (coordinators), Maine Medical Center 2 nd enrollment on 1-05-2012 (the trial's 194 th enrollment)

12. Congratulations to December 2011-January 2012 Enrollments Dr. Panayiotis Varelas (PI), Kathleen Mays-Wilson (coordinators), Henry Ford Health System 9 th enrollment on 1-10-2012 (the trial's 195 th enrollment)

13. Congratulations to December 2011-January 2012 Enrollments Dr. Robert Hoesch (PI), Julie Martinez(coordinator), University of Utah 8 th enrollment on 1-13-2012 (the trial's 196 th enrollment)

14. Congratulations to December 2011-January 2012 Enrollments Dr. Harold Adams (PI), Heena Olalde(coordinator), University of Iowa 4 th enrollment on 1-14-2012 (the trial's 197 th enrollment)

15. Covariate-Adaptive Design in Clear III trial: Status Report Presented by Gayane Yenokyan, Assistant Scientist, Biostatistics Center Simulations by Jonathan Gellar, PhD candidate, Department of Biostatistics

16. Presentation Plan Part I: – Goals of the (covariate-)adaptive design – Results of the simulations – Review of the current randomization design Part II: – Adaptive randomization status report

17. PART I: COVARIATE-ADAPTIVE RANDOMIZATION

18. Goals of the Adaptive Design The primary goal of adaptive randomization in the CLEAR III trial is to ensure that factors highly predictive of outcome are well-balanced between treatment groups A secondary goal is to reduce the amount of imbalance between treatment and control within each site Both of these goals aim to improve our estimate of the treatment effect, as differences between groups are more likely due to the treatment itself

19. Terminology Adaptive Designs: Response – adaptive – The likelihood of being assigned to each treatment arm changes given the information accumulated in the trial about the response Covariate – adaptive – The likelihood of being assigned to each treatment arm depends on the value of the important predictors

20. Simulation Work For the Adaptive Design in CLEAR III Data from CLEAR II studies on 100 patients Assigned patients to 2 treatment arms in many different ways Evaluated these methods to see how similar the two treatment arms are 2 main factors: size of the IVH and location of the clot

21. Predictors Preliminary analyses (from CLEAR A, B, and Safety data) indicate that two factors are highly associated with outcome (6-month Rankin): 1.IVH Volume 3 categories: – 0-20 cc (small) – 20-50 cc (medium) – >50 cc (large) 2.ICH Location Thalamic vs. non-thalamic bleed

22. Adaptive Algorithm in Clear III For each new patient, look at their IVH and clot location Look at all patients who have been previously randomized – Count the number of A’s and B’s in the study that have the same IVH volume category as the current patient – Count the number of A’s and B’s in the study that have the same ICH location as the current patient Combine this information (giving more weight towards IVH volume imbalance) to calculate the total imbalance The patient is more likely assigned to the arm that has less patients of the “same kind”

23. A note on clinical sites… In any multi-site randomized trial, site is an important variable “Institution effect” Need to stratify by site: – Use a randomization method that ensures equal distribution of patients at site level

24. Full Study Design 1.First 104 patients were randomized by study- wide block randomization 2.Address within-site imbalance first – Do not adaptively randomize unless there are at least 2 treatment and two placebo at that site – If less then two of either group, treatment is determined by site-specific block randomization (ABBA, AABB, BAAB, etc.) – Whenever we reach the 2/2 criteria, we move on to adaptive randomization, even if the block is not yet complete 3.Adaptive Randomization

25. Information regarding the Predictors Best possible data: Reading center-determined IVH volume Reading center-determined ICH location Both from the Stability CT DICOM If IVH volume is not available, follow “best information available” principle

26. Priority for Volume Information Volumetrics – Stability (most recent pre-randomization scan available) – Diagnostic Graeb Score – Reading Center or Surgical Center Stability (most recent pre-randomization scan available) Diagnostic Use RC over SC if both are available for the same scan – Site Stability (most recent pre-randomization scan available) Diagnostic

27. Priority for Clot Location Information All clot locations are taken from the diagnostic scan – Adjudicated location – Reading Center – Surgical Center – Site

28. From Graeb to IVH: the Data 16 GRAEB Scores Age Gender

29. From Graeb to IVH: the Model IVHvolume = ( 2.63586 + -0.55172*sex + -0.00849*age + 0.7062*GRAEB[1] + 0.26839*GRAEB[2] + - 0.4269*GRAEB[3] + 0.39977*GRAEB[4] + 0.17565*GRAEB[7] + 0.0029*GRAEB[8] + - 0.20547*GRAEB[9] + -0.03099*GRAEB[10] + 0.13624*GRAEB[12] + 0.11667*GRAEB[14] + 0.23831*GRAEB[1]*GRAEB[9] + -0.56245*GRAEB[1]*sex + 1.00483*GRAEB[9]*sex + 0.00511*GRAEB[1]*age + -0.01804*GRAEB[9]*age + 0.45083*GRAEB[2]*GRAEB[10] + - 0.19025*GRAEB[10]*sex + 0.002*GRAEB[2]*age + -0.00433*GRAEB[3]*GRAEB[11] + - 0.02657*GRAEB[3]*sex + 0.4268*GRAEB[11]*sex + 0.01366*GRAEB[3]*age + 0.46692*GRAEB[4]*GRAEB[12] + -0.2008*GRAEB[12]*sex + 0.42216*GRAEB[5]*GRAEB[13] + - 0.61784*GRAEB[13]*sex + 0.00716*GRAEB[5]*age + 0.00714*GRAEB[13]*age + 0.44686*GRAEB[6]*GRAEB[14] + -0.00483*GRAEB[6]*sex + -0.075*GRAEB[14]*sex + 0.00107*GRAEB[6]*age + 0.10993*GRAEB[7]*GRAEB[15] + -0.0578*GRAEB[7]*sex + 0.47792*GRAEB[15]*sex + -0.01424*GRAEB[15]*age + 0.00717*GRAEB[8]*sex + 0.31504*GRAEB[16]*sex + 0.00196*GRAEB[8]*age + -0.00902*GRAEB[16]*age )^2

30. PART II: COVARIATE-ADAPTIVE RANDOMIZATION STATUS REPORT

31. Date: October 2011 171 patients

32. Patient Randomization Over Time

33. Number of Subjects Randomized at Each Site Each site needs to have at least 2 subjects in each treatment arm to be “ready for adaptive algorithm”

34. IVH VOLUME

35. Balance on IVH volume Presented are the differences in IVH volume between the 2 treatment arms over time

36. Sources of Information on IVH

38. CLOT LOCATION

39. Balance on Clot Location Presented are the differences in proportions with thalamic location between the 2 treatment arms over time

40. Source of Information on Clot Location

42. SITE LEVEL

43. Within-site Imbalance

44. Within-Site Imbalance after 171 patients

45. Summary Adaptive randomization seems to be proceeding as expected – Only 13 AR patients in this report. The majority of the effects will be seen later in the study. Current level of imbalance is within normal rang. Both volumetrics and clot locations are relatively current, although there is of course always room for improvement – Especially in reducing the frequency of Graeb being used

46. Pre-Show Questions Answered!

47. Question: Why do we need adaptive randomization in CLEAR III? Response: The primary goal of covariate-adaptive randomization in the CLEAR III trial is to ensure that factors highly predictive of outcome are well-balanced between treatment groups

48. Question: What is the preferred source of information for both IVH volume and clot location? Response: Stability CT IVH volume and adjudicated location

49. Question: What is the target value for IVH volume difference between the two treatment arms? Response: No more than 5 cc

50. Question: Why within-site balance is important? How current randomization design ensures within-site balance? Response: The within-site balance is important, because a site is source of variation and can be a predictor of functional outcomes. Therefore, it is important to account for sites. We use within-site blocks to ensure that there are at least 2 patients in each treatments prior to starting adaptive algorithm.

52. Goals of 2012 Enroll 12 patients per month Continue to Improve Protocol Performance – Decrease protocol-associated SAEs – Increase protocol-associated removal of clot Conduct all follow-up visits inside the windows Maintain training of new team members

53. Quarterly Refresher Webinar February 9, 2012 9am-4pm

54. Thanks, see you next month!