Optimal sample sizes for phase II clinical trials and pilot studies.

Area of Science:

• Clinical Trial Design
• Biostatistics
• Bayesian Decision Theory

Background:

• Sample size calculation for phase III clinical trials is well-established.
• Methods for earlier phase clinical trials (phase II, pilot studies) lack consensus.
• Existing phase III methods are considered inappropriate for early phase trials.

Purpose of the Study:

• To explore appropriate sample size calculation methods for early phase clinical trials from a Bayesian decision-theoretic perspective.
• To determine optimal sample sizes for studies funded by large organizations (public sector, pharmaceutical companies).
• To minimize the average number of patients per identified effective therapy or maximize identified therapies over time.

Main Methods:

• Utilized a Bayesian decision-theoretic framework.
• Focused on optimizing sample sizes to balance patient numbers and successful therapy identification.
• Derived a relationship between phase II and phase III sample sizes.

Main Results:

• Proposed a specific ratio for phase II sample size relative to phase III.
• Indicated that phase II clinical trial sample sizes should be approximately 0.03 times that planned for the phase III study.
• This ratio is consistent with some existing proposals but smaller than commonly used in practice.

Conclusions:

• The Bayesian decision-theoretic approach offers a rational basis for sample size determination in early phase clinical trials.
• The recommended sample size ratio (0.03) provides an optimal balance for resource allocation and successful outcome identification.
• Current practices in phase II trial sample sizing may be larger than necessary according to this optimization framework.