Purchase this article with an account.
H.-C. Hsu, M. Patane, G. Ousler, III, K. Kennedy, P. Johnston, R. Abelson; Leveraging Phase II Study Data to Effectively Select Primary Endpoints and Sample Size for a Phase III Study. Invest. Ophthalmol. Vis. Sci. 2010;51(13):6270.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We conducted a single-center, double-masked, randomized phase II study to evaluate the efficacy of ocular iontophoresis of EGP-437 (dexamethasone phosphate 40 mg/mL) at two current levels (7.5 mA-min at 2.5 mA and 10.5 mA-min at 3.5 mA), each compared to placebo for treating the signs and symptoms of dry eye. In order to select the endpoints with the strongest potential treatment effects as the primary endpoints for the next study, we employed a bootstrapping method utilizing the previous study’s data to simulate potential sample populations to identify the endpoints with stable treatment effect signals. Additional simulations were then employed to help select a final sample size.
In the phase II study, several efficacy endpoints showed significant or nearly significant treatment effects; these were selected as candidate primary endpoints for the next study. A bootstrapping method was used to sample subjects from each treatment group from the previous study’s ITT population. For each of the bootstrap samples, each endpoint was analyzed for statistical significance by an appropriate statistical method, and the endpoints with higher success rates were deemed more likely to be successful efficacy endpoints for the next study. Simulations for those endpoints were then rerun, varying the sample size to optimize power.
Signs that showed promising success rates included fluorescein and lissamine green staining (Ora scale), tear film break-up time (TFBUT), and ocular protection index (OPI). Promising symptoms included ocular discomfort during Controlled Adverse Environment (CAE) and stinging in the diary. The sign with the best intersection of success with the symptoms was fluorescein staining. The number of subjects per group for the promising endpoints was then varied in follow-up simulations to achieve success rates of approximately 80%.
The bootstrapping approach to ranking efficacy endpoints was helpful in prioritizing endpoint options and then selecting sample sizes for phase III development.
Clinical Trial: :
This PDF is available to Subscribers Only