Impact of Dose Selection Strategies Used in Phase II on the Probability of Success in Phase III

The purpose of this study was to assess the impact of phase II dose-selection strategies on the likelihood of success of phase III clinical programs, comparing both traditional and adaptive approaches. We evaluated the impact of the phase II approach to dose selection (including traditional, design-adaptive, and analysis-adaptive approaches), the sample size used in phase II, the number of doses studied in phase II, and the number of doses selected to advance into phase III on the probability of demonstrating efficacy, of demonstrating a lack of toxicity, of phase III trial success, and on the probability of overall success of the combined phase II/phase III programs. The expected net present value was used to quantify the financial implications of different strategies. We found that adaptive dose allocation approaches (in particular, the Bayesian general adaptive dose allocation method) usually outperformed other fixed dose allocation approaches with respect to both probability of success and dose selection. Design-adaptive approaches were more efficient than analysis-adaptive approaches. The allocation of additional resources into phase II improved the probability of success in phase III and the expected net present value. Bringing two doses forward into phase III testing also increased the probability of success and improved the expected net present value. The overall probability of success in phase III ranged from 35% to 65%, consistent with recent industry experience. This success rate could likely be improved with additional investment in phase II, the use of design-adaptive dose-finding designs when possible, increasing the power of phase III trials, more explicit consideration of toxicity concerns, and better dose selection.