Comparison of the costs of drug development under different methods of calculation

This paper compares two methods for estimating drug development costs: the conventional cost per expected approval and the probabilistic cost to achieve at least one approval with a given confidence level. The paper shows that, for typical success rates and high confidence levels, the probabilistic approach requires 2–4 times more trials—implying that standard average-cost figures can underestimate reliable investment by a similar factor. Extending the model to multiple phases, the analysis demonstrates that while the gap between the two objectives narrows over time, stochastic transitions often leave portfolios underfunded for subsequent phases. Thus, expected-value estimates carry a high risk of failing to meet confidence-based portfolio goals, with direct implications for research funding and policy design.