Presented to The Fourth Margaret River Region Forum 28 April – 1 May 2010

1. Rare diseases – Small Populationsthe challenge of clinical trial design Albert Farrugia Plasma Protein Therapeutics Association & ALMAR Therapeutics Pty Ltd The Fourth Margaret River Region Forum 28 April – 1 May 2010

2. The drug development process

4. IF

6. Regulators are reasonableRare plasma protein deficiencies FDA approval of Baxter Protein C – pivotal study n=18 patients with deficiency FDA approval of CSL fibrinogen – pivotal study n=15 patients with afibrinogenemia EMEA approval for Novoseven in FVII deficiency - data on 32 patients with FVII deficiency treated in 28 different sites in 6 countries between 1988 and 1999.

9. Required Size, based on:

10. Significance level (usually 5%)

11. Minimal clinically worthwhile difference

12. Power (usually 80-90%)

13. Results: Test of significance

14. P<0.05 = Positive Study

16. The ideal Randomization Equipoise Controls Blinding 100 100 Clinical Trial

17. The use of RCTs should be questioned when… Results are readily observable or dramatic High morbidity/mortality occurs even with best available therapy Patients/physicians show reluctance to placement in non-treatment control arm Crossing-over from positive observable results ‘Orphan-type’ unmet indication where RCT costs are prohibitive and may stifle innovation Professor Archie Cochrane 1909 - 1988 “The [randomised controlled trial] is a very beautiful technique, of wide applicability, but as with everything else there are snags.” IN OTHER WORDS — MANY OF THE INDICATIONS FOR PLASMA THERAPIES

18. Thinking outside the boxAlternative approaches There must be some other way…. Interim analysis N of 1 trials Sequential designs Bayesian analysis RCT There is !

19. Bayesian methodology Less direct evidence is required for decision when prior evidence is taken into account – SO Uses Prior Evidence (a crucial component in the interpretation of any finding) Bayesian statistics allows us to combine prior evidence with trial results Prior information  probability distribution of the likely effect of the experimental treatment Posterior Probability distribution of the likely effect of the experimental treatment (range of plausible effects) + = Trial results (if necessary and possible)

20. IVIG vs Plasma Exchange for children with GBSBayesian approach Initial 86% probability that IVIg was not inferior to PE (Prior). Also include adult experience through previous trials Trial population size then needed to show non-inferiority with power of 77% is 160 kids With conventional design population needs to be 750 kids! Using Bayes we show that IVIG can be used instead of PE in children, and spare 500 kids from having to be trialed! Clinical Trials 2005; 2:305-310

21. 6.8 0.98 0.99 0.94 0.999 0.999 0.73 Inhibitors in previously transfused hemophiliacs (PTPs) Problem - Inhibitors in PTPs Need – to identify risky products FDA trial proposal (2003) – conventional method BUT most FVIII would be off the market Bayesian alternative (Lee & Roth 2005) - can identify risky products and keep the other products on the market Upper threshold of the true population incidence of inhibitors (%) Product ITT Observedinhibitorincidence (%) A B C D E F 2.3 2.3 2.9 0.9 1.0 5.7

22. The role of sequential analysis in RCTs In sequential trials: data are analysed after each participant’s results become available the trial continues until a clear benefit is seen in one of the comparison groups, or it is unlikely that any difference will emerge duration is shorter than fixed-length trials when there is a large difference in the effectiveness of the interventions being compared Use of this methodology is restricted to conditions where the outcome of interest is known relatively quickly

24. Using Z and V statistics

25. Boundaries specified through correction for discrete looks – crossing boundary completes trial

26. Reached in March

27. By using a series of interim looks –

28. A strong conclusion could be drawn from treating 26 subjects

29. A fixed sample size trial would have required 57 subjectsFeb 1, 1996 Jan 1, 1996 NEUROLOGY 1998;51:1629-1633

30. N of 1 trailsTrial of therapy Randomized clinical trial used in just one patient Patient undergoes pairs of treatment periods one period is the experimental treatment other period is the comparator (placebo or alternative treatment Considered to provide the strongest level of evidence Do not permit any generalization of the findings BUT - may be combined through meta‐analysis through a Bayesian random effects model Combination provides a population estimate for treatment effectiveness Retains the capacity to provide a distinct effectiveness estimate for each individual patient ?New and emerging indications for IgG

31. Some thoughts on………… ......Meta Analysis

32. Problems with meta-analyses The outcomes of 12 large randomized, controlled trials were not predicted accurately 35 percent of the time by the meta-analyses published previously on the same topics. (N Engl J Med 1997;337:536-42.) Adjusting for random error through sequential analysis reveals false positive results in many meta-analyses ieprediction of an effect when none exists (International Journal of Epidemiology 2008;1–12)

33. Use of Trial Sequential Analysis (TSA) to analyze Meta - Analyses Cochrane Neonatal Reviews !!!!!! !!!!!! !!!!!! International Journal of Epidemiology 2008;1–12

35. FDA approval of CSL fibrinogen – pivotal study n=15 patients with afibrinogenemia