Simulation and Design Considerations for Noninferiority Trials in Phase II
Tracy Higgins, Meg Bennetts, Patrick J Johnson.
Pfizer, Inc. UK
Objectives: Increasingly noninferiority designs are being used for Phase II trials to compare new drugs to standard treatment (i.e. a well proven standard therapy which has been shown to be superior to placebo, and has an established, predictable and quantifiable effect). It is often of interest to know early in the drug development process whether the new drug is equi-efficacious to standard treatment. In particular where the benefits of the new drug may be a better safety profile or an improved administration schedule. The large sample sizes required for noninferiority trials means they can be more costly and take longer to run, so it is important that the design is optimised to keep the impact on costs and timelines to a minimum. This poster outlines simulation methodology used to optimise trial design for a Phase II PoC study where the primary objective is to assess noninferiority against the standard treatment, and the secondary objective is to define the dose-response of the new drug in patients.
Methods: The noninferiority margin defines the largest difference that is clinically acceptable between treatments for those treatments to be regarded as clinically equivalent. Setting the margin is critical to the design and is based on the range of possible true differences between test drug and active control (from uncertainty in the mean response and the trial design). Noninferiority is usually assessed through confidence interval estimation with regard to this margin. In the example shown PK/PD models are available for both the new drug and the standard treatment and are used for simulation. The primary endpoint is FEV1 and the noninferiority margin is defined as -100 mL (a combination of the true mean difference and acceptable variability in the mean response based on the trial design). All data shown in the example presented are simulated data.
Technical success: Simulations were performed across a wide range of doses (incorporating parameter uncertainty from the PK/PD model but otherwise assuming an infinite number of subjects for each dose) and the distribution of the dose response presented graphically with the simulated response of the control treatment. For each replicate the equi-efficacious dose for the new drug was calculated and the distribution of equi-efficacious doses summarised across replicates.
Dose selection and optimisation: After establishing the possible dosing strength options, doses were selected using a modified version of PFIM 1.2  and a D-optimal iterative process that incorporates the influence of model uncertainty . The optimal design solutions were verified using classical trial simulations.
Trial success: The lower confidence interval of the difference between new drug and standard for each dose level was compared to the noninferiority margin (-100 mL) for each simulated trial. To calculate the probability of success and being correct the noninferiority margin was defined differently for technical success (probability correct) to eliminate the variability due to trial design. Operating characteristics were calculated for technical success noninferiority margins of -30 mL, -40 mL and -50 mL and the results compared.
Impact of dose-response approach: Each of the simulated trials was analysed using a dose-response approach to calculate the SE of the mean response of the new drug and using a pairwise comparison approach for each dose level separately, to assess the relative efficiency of the two methods. These comparisons were performed for the parallel group design and for the crossover design.
Interim Analysis:Simulations were performed incorporating an interim analysis to select a dose to carry forward to the second part of the trial for assessment of noninferiority with the control treatment.
Results: Not available yet
Conclusions: Not available yet
 Cornel P: Equivalence and noninferiority trials - are they viable alternatives for registration of new drugs? (III). Curr Control Trials Cardiovasc 2004, 5:8 (17 August 2004).
 Snapinn SM: Noninferiority Trials. Curr Control Trials Cardiovasc 2000, 1:19-21.
 Retout, S. & Mentré, F. Optimisation of individual and population designs using Splus. Journal of Pharmacokinetics and Pharmacodynamics, 2003, 30(6): 417-443.
 Johnson, P., Dai, H., Neelakantan, S. & Tensfeldt, T. Optimal dose & sample-size selection for dose-response studies. PAGE 16 (2007) Abstr 1127 [www.page-meeting.org/?abstract=1127]