1. 1 Chapter 9 Bridging Studies & Multi-Regional Trials

2. 2 Outline  Introduction  Taiwan ’ s Situations  An Bayesian Approach  Discussion

3. 3 Global Drug Development Bridging Studies –For clinical trials without including Taiwan patients, how to conduct a bridging study in Taiwan for assessment of similarity of treatment effect between Taiwan and other counties? Multi-regional Trials –For clinical trials including Taiwan patients, how to assess the similarity of treatment effect between Taiwan and all regions?

4. 4 Introduction ICH (International Conference on Harmonisation) E5 Ethnic Factors in the Acceptability of Foreign Clinical Data The purpose of this guidance is to facilitate the registration of medicines among ICH regions by recommending a framework for evaluating the impact of ethnic factors upon a medicine’s effect, i.e., its efficacy and safety at a particular dosage and dose regimen.

5. 5 Asian (n = 342) HR = 0.66 (0.48, 0.91), P =.011 RR = 12.0% Non-Asian (n = 1350) HR = 0.93 (0.81, 1.08), P =.364 RR = 6.5% 0123456789101112131415 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0123456789101112131415 ——IRESSA ® ------Placebo Time, mo Patients surviving (%) Ethnic Difference

6. 6 Objectives of ICH E5  To describe the characteristics of foreign clinical data that will facilitate their extrapolation to different populations and support their acceptance as a basis for registration of a medicine in a new region  To describe regulatory strategies that minimize duplication of clinical data and facilitate acceptance of foreign clinical data in the new region  To describe the use of bridging studies, when necessary, to allow extrapolation of foreign clinical data to a new region  To describe development strategies capable of characterizing ethnic factor influences on safety, efficacy, dosage, and dose regimen

7. 7 Bridging Data Package A bridging data package consists of 1)Selected information from the complete clinical data package (CCDP) that is relevant to the population of the new region, including pharmacokinetic data, and any preliminary pharmacodynamic and dose-response data, and 2)If needed, a bridging study to extrapolate the foreign efficacy and/or safety data to the new region.

8. 8 Complete Clinical Data Package A clinical data package intended for registration containing clinical data that fulfill the regulatory requirements of the new region and containing pharmacokinetic data relevant to the population in the new region

9. 9 Bridging Study A bridging study is defined as a supplemental study performed in the new region to provide pharmacodynamic or clinical data on efficacy, safety, dosage, and dose regimen in the new region that will allow extrapolation of the foreign clinical data to the new region.

10. 10 Ethnic Factors  Intrinsic Ethnic Factors are more genetic and physiologic in nature e.g., genetic polymorphism, age, gender, height, weight, lean body mass, body composition, and disease conditions, etc.  Extrinsic Ethnic Factors are more social and cultural in nature e.g., environment, culture, medical practice, health insurance, practices in clinical trials or conduct

11. 11 Bridging Studies ICH E5 Only after the medicine is approved in the original region Performed in the new region

12. 12 Taiwan’s Situations

13. 13 Taiwan Before Bridging Study  An approved local clinical trial study report is required for the new drug application in Taiwan—July 7 Announcement in 1993 Disadvantage: A sample size of 40 as required would be difficult to demonstrate significant importance clinically or statistically The study design of the local trial usually only repeated a study that has been done in the foreign countries but in a smaller sample size;The study has not been designed based on the medical situation in Taiwan

14. 14 Taiwan’s Strategy to Implement Bridging Study  Smoothly convert compulsory Local Clinical Trial (LCT) to meaningful bridging study  Gradually, stepwise announce waived local clinical trial  Create an environment: (1) meet international regulation, ICH (2) require optimized dosage for Taiwanese patient  Communicate with local and international pharmaceutical industry  Announce new regulation according to the international norm and the consensus from communications  Create an international platform “APEC – Taipei”  Implement Double Twelve Announcement – Bridging Study

15. 15 Stepwise Implementation  1998 Announce: two years later, switch from LCT to bridging study  Many communications and negotiations with local and international pharmaceutical industry  2000, Dec.12, (Double Twelve Announcement) – public announce bridging study regulation  1998 Five announcements of LCT wavier  Two years transition periods: both LCT and bridging studies acceptable from 2000 ~ 2002  Many international conferences held in Taipei and other Asian countries, regarding BS, through the APEC platform  Ask CDE to complete the practical issues related to implementation of BS  2004, Jan. 1, Bridging evaluation

16. 16 Available Statistical Methods 1. Hierarchical Model (Liu, Hsueh, and Chen, 2002, Biometrical Journal, 44: 969-981) 2. Takeuchi, M. Controlled Clinical Trials 23: 55-57, 2002 3. Shao, J. and Chow, S. C. Statistics in Medicine, 21: 1727-1742, 2002 4. Population Similarity (Chow, Shao, Hu, 2002, JBS, 12: 385-400) 5. Consistency Approach (Shih, 2001, Controlled Clinical Trials, 22: 357-366) 6. Bayesian Positive Treatment Approach (Liu, Hsiao, and Hsueh, 2002, JBS 12: 297-294) 7. Bayesian Noninferiority Approach (Liu, Hsueh and Hsiao, 2004, JBS accepted) 8. Group Sequential Approach (Hsiao, Xu and Liu, 2003, JBS, 13: 793-801) 9. Two-Stage Approach (Hsiao, Xu and Liu, 2004, submitted)

17. 17 INFO Y 3 N Data Package Vol., page 1 Checking List for the evaluation of Bridging Study by the Sponsor I. The current status of clinical study of the drug in the world □ □□ □ II. NDA expert report or Investigator’s Brochure 2 □ □□ □ III. Pharmacokinetics, safety and efficacy data related to Asian population □ □□ □ IV.Comparative analysis of Pharmacokinetics, safety and efficacy data between Asian population and others. □ □□ □ V. Self evaluation (please provide reference materials or literature) □ □□ □ Y N U 1.Does the drug show a Non-linear pharmacokinetics at the therapeutic dose? □ □□ □ □ □ □□ □ □ 2.Is the drug with a steep pharmacodynamic curve for both efficacy and safety (a small change in dose results in a large change in effect) in the range of the recommended dosage and dose regimen? □ □□ □□ □ □□ □ □ □ □□ □□ □ □□ □ □ 3.Is the drug with narrow therapeutic dose range? Note ： 1.To speed up reviewing process, please clearly indicate the volume and page number as requested. In addition to the page number, the related paragraph may be highlighted when necessary. 2.Please provide the comparative analysis of different ethnic groups, if it’s available. Please also explain if there is no comparative analysis of different ethnic groups in NDA expert report. 3.Y=yes; N=no; U=unknown Checking List for Sponsors (1 of 3)

18. 18 Note ： 1.To speed up reviewing process, please clearly indicate the volume and page number as requested. In addition to the page number, the related paragraph may be highlighted when necessary. 2.Y=yes; N=no; U=unknown INFO Y 2 N Data Package Vol., page 1 Checking List for the evaluation of Bridging Study by the Sponsor V. Self evaluation (please provide reference materials or literature) □ □□ □ Y N U 4. Is the drug highly metabolized, especially through a single pathway, thereby increasing the potential for drug-drug interaction ? □ □□ □ □ □ □□ □ □ 5.Is the drug metabolized by enzyme known to show genetic polymorphism? □ □□ □□ □ □□ □ □ □ □□ □ □ □ □□ □ □ 6. Is the drug administered as a prodrug, with the potential for ethnically variable enzymatic conversion ? □ □□ □ □ □ □□ □ □ 7. Is the drug with high inter-subject variation in bioavailability ? □ □□ □ □ □ □□ □ □ 8. Is the drug with low bioavailability, thus more susceptible to dietary absorption effects? □ □□ □ □ □ □□ □ □ 9. Is the drug with high likelihood of use in setting of multiple co- medications ? □ □□ □ □ □ □□ □ □ 10. Is the drug with high likelihood for inappropriate use, e.g. analgesics and tranquilizers ? Checking List For Sponsors (2 of 3)

19. 19 INFO Y 3 N Data Package Vol., page 1 Checking List for the evaluation of Bridging Study by the Sponsor V. Self evaluation (please provide reference materials or literature) □ □□ □ Y N U 11. Is there any difference in epidemics of applied indication between the major study population and our population (including medical history, mechanism of disease development and the rate of occurrence, the efficacy and safety of other drugs in the same class)? □ □□ □ □ □ □□ □ □ 12. Other important ethnic sensitive factors, such as “Is there any difference in the medical practice?” □ □□ □□ □ □□ □ □ Note ： 1.To speed up reviewing process, please clearly indicate the volume and page number as requested. In addition to the page number, the related paragraph may be highlighted when necessary. 2.Y=yes; N=no; U=unknown 3.Please according the checking list provide an integrate summary or a brief description of all the information submitted. VI. Post-marketing surveillance information □ □□ □ Overall conclusion of self evaluation (Is it clinically insignificant? What is the risk and benefit of the drug applied (such as, “Does the indication applied belong to severe disease”, “Is there a alternative therapy?”, “Are the differences of the data in ethnic factors acceptable ?) □ □□ □ Summary 3 □ □□ □ Checking List for Sponsors (3 of 3)

20. 20 Bridging Data Package Summary for the Consideration of Bridging Study Accept submission Checking List Technical Review (Designate reviewer) Review meeting Sponsor meeting Supplement Clinical Review Committee Review report and Recommendation: 1. No Bridging study required 2. Bridging study is required – Type of Bridging study Result of Evaluation: 1. No Bridging study required 2. Bridging study is required – Type of Bridging study Notification SponsorBoPACDE CDE acceptance verification Expert Consultants (Statistical, Clinical, Pharmacokinetics reviewers) Schedule Sponsor meeting

21. 21

22. 22

23. 23 Does bridging strategy of ICH E5 warrant further implementation? Is Taiwan on the right way?

24. 24 Case I  Drug A is a fixed combination of two anti-platelet agents with indication for secondary prevention of thromboembolic stroke ( 200mg dipyridamole/25mg aspirin 1bid)  After the standard process of BSE, we decided to request a bridging study due to an ethnic difference in medical practice (much lower dose for one of the components in Taiwan) and higher headache-associated dropout rate in previous Philippine study

25. 25 Case I  Headache drop out rate: Phillipino > Caucasian  Local Bridging Study Result : first 4 weeks Group Placebo Reduced Dose 2wk Full Dose Full Dose 2wk 4wk Headache 8.7% 6.7% 16.3% drop out rate  Risk Management: Change labeling’s instruction for use

26. 26 Case II  Drug B is a new potent lipid-lowering agent  The PK study in Japanese shows that Cmax of Japanese is 1.9~2.5 times of that for Caucasian while AUC is 2~2.5 times  Although the mean interracial difference is not substantial, Taiwan approved the drug with reduced maximal dosage due to the dose-dependent, drug-related rare SAE of rhabdomyolysis

27. 27 Case II The decision is further echoed by US FDA After reviewing the results of a Phase IV PK study in Asian-Americans, FDA urged the physician to reduce the starting dose and prescribe high dose with caution for Asians in Labeling in March, 2005

28. 28 Bayesian Approach

29. 29 Bayesian Approach For bridging studies  Small sample size  No power  Information on dose response, efficacy and safety of the original region can not be concurrently obtained from the local bridging studies but are available in the trials conducted in the original region  Need to borrow “strength” from CCDP of the original region  Information on dose response, efficacy and safety of the original region can and should be incorporated in a statistically sound manner to evaluate bridging evidence by local bridging studies

30. 30 Before Experiment After Experiment Bayesian Approach Past experience about similar situations involving similar  Prior information P (  Posterior information P(  prior&data  Observed results ( Data ) Make Statistical inference Treatment effect 

31. 31 Assumption, Notation and Hypotheses  We focus on the trials for comparing a test product and a placebo control  X i and Y j are some efficacy responses for patients i and j receiving the test product and the placebo control respectively in the new region  X i ’s and Y j ’s are normally distributed with known variance σ 2  μ NT and μ NP are the population means of the test and placebo, respectively, and let Δ N = μ NT - μ NP  H 0 : Δ N  0 vs. H A : Δ N > 0

32. 32 Parameters Test Product Effect Placebo Effect Original Region  OT  OP New Region  NT  NP

33. 33 Before Bridging Study Original region data to estimate   OT Original region data to estimate   OP Conclude   OT >   OP

34. 34 Bayesian Positive Treatment Approach Original region information to form prior Posterior information P{  NT -  NP  prior & bridging  Bridging data New region treatment effect   NT —   NP Conclude  NT >  NP if P{  NT -  NP >0  prior & bridging  is large

35. 35 Previous Statistical Approach  Use the estimate of treatment effect from the original region formulated as a normal prior  Compute the posterior treatment effect with the data from the new region

36. 36 Results from Original Region Change from baseline in sitting DBP at week 12 Region StatisticsTest Placebo I n138132 Mean-18 -3 SD 11 12 II n185179 Mean-17 -2 SD 10 11 III n141143 Mean-15 -5 SD 13 14 Previous Statistical Approach

37. 37 Results from New Region: Change from baseline in Sitting DBP at week 12 Region StatisticsTest Placebo New n 64 65 Mean-4.5 -3.8 SD 11 11 Posterior probability of similarity: Psp  1 Previous Statistical Approach

38. 38 Previous Statistical Approach  Original region: Efficacy of the test drug is superior to the placebo  New Region: Reduction of sitting BP of the test drug is same as that of the placebo  Conclusion: The results of the original region can be extrapolated to the new region despite of inconsistent results between original and new regions  Evaluation of bridging studies is overwhelmingly by the results of original region due to imbalance of information provided by the two regions

39. 39 Use of Prior Distribution The proposed mixture model of the prior distribution for Δ N is a weighted average of the noninformative and normal priors as given below π(Δ N ) =γπ 1 (Δ N ) + (1-γ)π 2 (Δ N )  π 1 (.) ≡c is a non-informative prior  π 2 (.) is a normal prior with mean θ 0 and variance σ 0 2 which summarizes the foreign clinical data about the treatment difference provided in the CCDP  0 ≦ γ ≦ 1

40. 40 Marginal Density Based on the clinical responses from the bridging study in new region, Δ N can be estimated by The marginal density is where

41. 41 Posterior Distribution Given the bridging data and prior distribution, the posterior distribution of Δ N is

42. 42 Bridging Evaluation Similarity on efficacy in terms of a positive treatment effect for the new region can be concluded if the posterior probability of Similarity for some pre-specified 0 <  < 0.5.

43. 43 Example  The CCDP provides the results of three randomized, placebo controlled trials for a new antidepressant (test drug) conducted in the original region  The primary endpoint is the change from baseline of sitting diastolic blood pressure (mmHg) at week 12  A bridging study was conducted in the new region to compare the difference in efficacy between the new and original region

44. 44 Three Scenarios  The first scenario presents the situation where no statistically significant difference in the primary endpoint exists between the test drug and placebo (2- sided p-value = 0.6430  The second situation is that the mean reduction of sitting diastolic blood pressure at week 12 of the test drug is statistically significantly greater than the placebo group (2-sided p-value < 0.0001)  The third scenario is the situation where due to the insufficient sample size of the bridging study, no statistical significance is found between the test drug and placebo although the magnitude of the difference between the test drug and placebo observed in the original region is preserved in the new region (2-sided p-value = 0.0716)

45. 45

46. 46

47. 47 Scenario I  If the regulatory agency allows all information of the original region to be used for evaluation of similarity between the new and original region, γ is set to be 0 and hence P SP  1.00  If γ ≧ 0.1, then P SP always drops to around 0.6789

48. 48 Scenario II  The values of P SP in Example 2 appear to be close to 1.00 regardless of the choice of γ

49. 49 Scenario III The values of P SP are all greater than 0.9675 for all values of γ between 0 and 1 With the strength of the substantial evidence of efficacy is borrowed from the CCDP of the original region, our procedure can prove the similarity of efficacy between the new and original region when a non-significant efficacy result but with a similar magnitude is observed in the bridging study

50. 50 Final Remarks  The proposed prior is a weighted average of a non-informative prior and a normal prior  The proposed procedure can avoid the situation of concluding similarity between the new and original region when the efficacy result of the test drug observed the bridging study of the new region is same as or even worse than that of the placebo group  Our proposed procedure can reach a conclusion that is more consistent with the results obtained from the bridging study

51. 51 Final Remarks  Selection of weight γ by the regulatory agency in the new region should consider all differences in both intrinsic and extrinsic ethnical factors between the new and original regions and at the same time should also reflect their belief on the evidence of efficacy provided in the CCDP of the original region

52. 52 Final Remarks  We use a normal prior for summarization of the results in CCDP of the original region  We also use other prior distributions I) double exponential distribution II) lognormal distribution  Other different distributions used for π 2 reach the same conclusion

53. 53 The 4th Multitrack Workshop in Japan Pursuing Global Joint Development, Drug Discovery and Fostering Promotion

54. 54 Drug Lag Number of products waiting for marketing among world’s top 88 selling-products in 2004 Japan: 28 Russia: 27 Singapore: 22 Taiwan: 12 France: 9 Korea: 5 Germany: 2 UK: 1 USA:0

55. 55

56. 56

57. 57 ICH-E5 Q11 It may be desirable in certain situations to achieve the goal of bridging by conducting a multi-regional trial under a common protocol that includes sufficient numbers of patients from each of multiple regions to reach a conclusion about the effect of the drug in all regions…

58. 58 Key Words in Q11 Answer Multiregional Trials serving as a bridging study Hierarchy of persuasiveness –The most persuasive results Statistical significant in overall result AND Statistical significant in region of interest –Still persuasive results Statistical significant in overall result AND Region of interest could not reach statistical significance BUT Consistent trends across regions

59. 59 Questions How do we draw inferences concerning the treatment effect in a multi-regional trial? How do we assess “consistent trends” across regions? What is a rational way to determine the minimum sample size for a region?

60. 60 Fundamental Premises 1.No region should proclaim itself to be the region of interest and demand the treatment to show a statistically significant result at the usual significance level within that region. 2.The study’s primary objective is to demonstrate an overall treatment effect. 3.Satisfying regional requirement is a key secondary objective.

61. 61 Assumption and Notation  We focus on the trials for comparing a test product and a placebo control  X i and Y j are some efficacy responses for patients i and j receiving the test product and the placebo control respectively in the new region  X i ’s and Y j ’s are normally distributed with known variance σ 2  μ T and μ P are the population means of the test product and placebo, respectively, and let Δ = μ T - μ P

62. 62 Assumption and Notation Effect size (Δ/σ) is uniform across regions Hypothesis: H 0 :Δ ≦ 0 vs. H A :Δ ＞ 0 N is the total sample size for each group planned for detecting an expected treatment difference Δ=δ at the desired significance level αand with power 1 － β. N=2{(z 1-α +z 1-β )/δ} 2, where z 1-α is the (1 － α)th percentile of the standard normal distribution.

63. 63 Questions How do we draw inferences concerning the treatment effect in a multi-regional trial? How do we assess “consistent trends” across regions? What is a rational way to determine the minimum sample size for a region?

64. 64 Consider a Specific Region Suppose that we are interested in judging whether the treatment is effective in a specific region, say Region 1

65. 65 Multi-Regional Results Overall Results Sample Size 2N Trt. Diff. D p 1, D 1 p2, D2p2, D2 p K-1, D K-1 pK, DKpK, DK..... Region 1 DataRegion 2 DataRegion K-1 DataRegion K Data..... Results other than Region 1 Trt. Diff. D 1C

66. 66 Four Criteria Given that the overall result is significant at α level, we will judge whether the treatment is effective in the Region 1 by the following four criteria. (i) D 1 ≧ ρD 1c, for some ρ>0 (ii) D 1 ≧ ρD, for some ρ>0 (iii) ρ ≦ D 1 /D 1c ≦ 1/ρ, for some 0<ρ<1 (iv) ρ ≦ D 1 /D ≦ 1/ρ, for some 0<ρ<1

67. 67 Assurance Probabilities

68. 68 Assurance Probabilities

69. 69 Aim To determine p 1 to ensure that the assurance probabilities of criteria (1)-(4) given Δ=δ are maintained at a desired level, say 80%.

70. 70 α= 0.025, β=0.1, and ρ=0.5

71. 71 Concluding Remarks Selection of the magnitude, ρ, of consistency trend may vary from disease to disease The Japanese MHLW suggests that ρ be 0.5 or greater for the second criteria

72. 72 Assumption and Notation p i is the proportion of patients in the i th region Z i is the test statistic in the i th region For simplicity, we will work with 3 regions Z is the test statistic for the overall results

73. 73 Strategies Positive consistent trends across all regions Positive consistent trends at some significant level for all regions Positive consistent trend for a specific region

74. 74 Notations Region 3 [Largest] Region 2 [Second smallest] Region 1 （ Japan? ） [Smallest] Sample Size/Group: N p3p3 p2p2 p1p1 The number N is determined to provide an 80% or 90% power for the primary analysis at the one-sided 2.5% level. Due to the constraints of p 1 ≤ p 2 ≤ p 3 and p 1 +p 2 +p 3 =1 D3D3 D2D2 D1D1 Estimated treatment effect (New treatment - Placebo)

75. 75 Positive Consistent Trends Across All Regions Probability of consistent trends across 3 regions ψ(x) is the cumulative distribution of the standard normal distribution at x

76. 76 Observations P cs is the largest when p 1 = p 2 = p 3, i.e., the 3 regions are equally represented. For a given p 1,  P cs is the smallest when p 2 = p 1 ;  P cs is the largest when p 2 = p 3 = (1-p 1 )/ 2.

77. 77 P cs When p 2 = p 1 (p 1, p 2, p 3 )80% power90% power (0.05, 0.05, 0.9)53.758.6 (0.1, 0.1, 0.8)65.671.7 (0.15, 0.15, 0.7)73.479.9 (0.2, 0.2, 0.6)78.985.3 (0.25, 0.25, 0.5)82.588.8 (0.3, 0.3, 0.4)84.590.7

78. 78 P cs When p 2 = p 3 (p 1, p 2, p 3 ) 80% power90% power (0.05, 0.475, 0.475)69.674.6 (0.1, 0.45, 0.45)76.482.2 (0.15, 0.425, 0.425)80.486.5 (0.2, 0.4, 0.4)82.888.9 (0.25, 0.375, 0.375)84.290.3 (0.3, 0.35, 0.35)84.891.0

79. 79 A Reasonable Question Assuming that treatment effect is positive and uniform across regions, what is the minimum number of subjects the smallest region should contribute so that the probability of observing consistent trends across the 3 regions is 80% (90%)?

80. 80 Plots of P cs against p 1 with p 2 =p 1 0.8 0.9 0.1510.2130.277 Power:90% Power:80% P cs never reaches 90%

81. 81 But … In practice, inference concerning regional results in a confirmatory trial is relevant only if the overall treatment effect is statistically significant. The above calls for looking at P cs conditional on first concluding a significant overall treatment effect at the one-sided 2.5% level.

82. 82 Conditional P cs vs Unconditional P cs 80% power90% power (0.05, 0.05, 0.9)57.5 (53.7)64.6 (58.6) (0.1, 0.1, 0.8)71.1 (65.6)73.8 (71.7) (0.15, 0.15, 0.7)82.5 (73.4)82.5 (79.9) (0.2, 0.2, 0.6)86.1 (78.9)90.0 (85.3) (0.25, 0.25, 0.5)90.9 (82.5)92.2 (88.8) (0.3, 0.3, 0.4)93.2 (84.5)94.4 (90.7) Treatment effect = 0.250,  =1