1 Timing and Duration of Relapse Prevention Trials in Psychiatric New Drug Development David Michelson, M.D. Executive Director, Neuroscience Medical Research Eli Lilly and Company

2 Background The FDA has asked several questions, including particularly whether –evidence of the long-term (maintenance) efficacy of new psychiatric drug treatments should be provided earlier than current practice –3-month lead-ins adequately assess efficacy during maintenance treatment Changes to current practice could provide more data at approval, but –It is not established that current practice provides insufficient data –How interpretable/valuable data from a longer/earlier study would be is unclear The changes being discussed would make the conduct of studies more difficult –Large increases in sample size and issues around signal detection –Time to study completion –When studies are conducted during the development process –Analysis of previous studies illustrates some of problems raised by such changes

3 Attrition During Initial Treatment Predicted by Previous Relapse Prevention Studies at Lilly Indication ( Study Identifier ) OL Tx / DurationEstimated Attrition at 12 weeks Estimated attrition at 26 Weeks* Depression (HCEX) Fluox 20 mg 12 weeks 53%73% Panic Disorder** (HCHG) Fluox 20 mg 10 weeks 56%69% ADHD (LYAF) Atmx mg/kg 12 weeks 31%54% Bipolar Mania (HGHL) Olanz 5-20 mg 12 weeks 51%72% * Attrition for 26 weeks are estimates derived by extrapolating from patient disposition or Kaplan-Meier curves during blinded treatment phase ** Acute phase was double-blind rather than open label

4 Attrition During Initial Treatment in Olanzapine Schizophrenia Studies Relapse prevention for olanzapine was assessed by –switching patients stable on any antipsychotic to olanzapine, –observing for 14 weeks –randomizing to discontinuation or continued olanzapine Data are not directly comparable to data from long-term studies of the mood and anxiety disorders –Patients not acutely ill at study outset Historical attrition rates in acute studies of olanzapine over ~12 weeks are approximately 50% –Suggests that the discontinuation rates of approximately 50% at 3 months and 70% at 6 months seen in other disorders studied with standard designs also apply for schizophrenia

5 Duration of Initial Treatment and Sample Size: Historical Data Duration Of Treatment Depression Relapse Rates in Historical Data (HCEX) ADHD Relapse Rates in Historical Data (LYAF) DrugPlaceboN needed at Study Entry DrugPlaceboN needed at Study Entry 12 Weeks37%58%26419%35% Weeks22%33%790n/a 52 Weeksn/a 5%15%456

6 Duration of Treatment Prior to Withdrawal: Other Considerations Treatment effect sizes in historical data do not appear to decrease markedly with longer ‘run-ins’ Longer treatment periods require signficantly larger samples –Attrition prior to randomization rises with increasing time –Event rates decrease with time (longer periods of stability appear to predict future stability) It is unknown whether longer stabilization time makes patients more stable or simply selects for patients with a more stable course of illness ‘Negative’ studies could only be interpreted in the presence of a positive control –Would increase sample size and time to completion even more Increasing stringency around response criteria or restricting excursions would necessitate further increases in sample size

7 Study Timing Relapse prevention studies take significantly longer to conduct than acute efficacy studies Requiring completion of such studies at NDA filing will delay approvals and patient access to novel treatments

8 Time Required to Conduct Acute and Relapse Prevention Studies: Recent Data Atomoxetine Study Identifier Study Description 1 Time from First Patient Visit to Last Patient Visit LYAT 2 Pivotal acute efficacy Two arm, 6-week acute placebo controlled study N = days (3.9 months) LYAC 3 Pivotal acute efficacy Four arm, 8–week, placebo-controlled dose response study N = days (8.1 months) LYAF 4 Registration quality study to support labeling for long-term efficacy Placebo-controlled withdrawal study with two randomizations (after 3 months and after 1 year) N = days (31.3 months) 1 N = all patients with at least 1 post-baseline visit 2 Michelson et al. Am J Psychiatry Michelson et al Pediatrics Michelson et al J Am Acad Child Adol Psychiatry 2004

9 Requiring Completion of Relapse Prevention at NDA Filing: Concerns Optimal dosing may be poorly understood until late in phase III –Potential for suboptimal/excessive dose in relapse prevention Relapse prevention studies involve extended acute open label treatment, and early in phase III –can confound interpretation/attribution of adverse events Large numbers of patients will be exposed to drug for extended periods prior to a definitive demonstration of acute efficacy Relapse prevention studies started after the completion of pivotal trials in phase III will significantly delay in patient access to novel treatments

10 What is the Risk of Chronically Inefficacious Drugs Being Approved and Used Under Current Practice? Predictive value of 3 month lead-ins for longer efficacy –Available data are limited, but the evidence we do have suggests that efficacy at 3 months predicts efficacy after longer periods –Conversely, we are unaware of convincing data demonstrating a later loss of drug effect for any antidepressant or antipsychotic effective acutely and in continuation treatment after 3 months Predictive value of sub-acute data for longer efficacy –The likelihood that a novel compound capable of inducing but not maintaining a symptom response chronically would fare well in 8-12 week acute, parallel design trials seems low Marked and early ‘therapeutic tachyphylaxis’ would make it very difficult to sustain superiority to placebo through the trial’s endpoint

11 Conclusions Requiring longer relapse prevention studies and making them a condition of initial filing will delay patient access to new drugs without providing a clear offsetting benefit over current practice –Access to novel drugs will be delayed, in some cases considerably –The proposed changes could significantly increase barriers to approval, discouraging sponsors from undertaking some psychiatric drug development programs. –The programs most affected may be drugs with highly novel and unprecedented mechanisms, which are most likely to provide breakthroughs for patients, but also have the highest likelihood of failure