Simulation as a design tool for phase I/II clinical trials: an example from bone marrow transplantation.

We discuss the design and analysis of a proposed phase I/II clinical trial in bone marrow transplantation whereby dose modifications that decrease the risk of one complication increase the risk of another. Trials of this type are carried out to determine whether a dose can be found that balances the risks of each complication. Three different scenarios describing potential relationships between each risk and the treatment dose are postulated. The scenarios encompass both favorable situations in which several acceptable doses exist and unfavorable situations in which no acceptable dose exists. The operating characteristics of three sequentially developed trial designs were examined by simulation under each dose-response scenario. The first design was derived from seemingly reasonable rules, but simulations showed that performance fell far short of what was desired, thus motivating modifications. Subsequent designs showed improved performance characteristics compared to the original design. Without close examination of the operating characteristics, the original design would have been implemented, leading to high risk of an erroneous conclusion.