Beyond the two-trials rule.

Area of Science:

• Biostatistics
• Pharmacoeconomics
• Regulatory Science

Background:

• The standard drug approval process relies on the "two-trials rule," requiring at least two independent, adequate, and well-controlled studies to demonstrate efficacy.
• This rule, while robust, may not be optimal when more than two trials are available or when considering nuanced error rate control.

Purpose of the Study:

• To explore alternatives to the traditional two-trials rule for drug approval.
• To investigate methods for controlling partial Type-I error rates while maintaining overall Type-I error control, especially with up to three studies.
• To evaluate the performance of different p-value combination methods in terms of statistical power and efficiency.

Main Methods:

• Examined p-value combination techniques including Pearson's method, Edgington's method, and the harmonic mean p-value test.
• Analyzed the properties of these methods concerning partial and overall Type-I error rates.
• Assessed methods for sequential trial assessment and compared project power and the number of studies required.

Main Results:

• Edgington's method demonstrated a practical balance between ease of implementation and statistical performance.
• While showing moderate inflation in partial Type-I error rates, Edgington's method significantly enhanced project power compared to other approaches.
• The study highlighted the utility of less-common p-value combination methods for optimizing drug approval processes.

Conclusions:

• Alternative p-value combination methods, particularly Edgington's, offer viable enhancements to the standard two-trials rule for drug approval.
• These methods provide flexibility in utilizing data from multiple trials while maintaining rigorous statistical control.
• Edgington's method is recommended for its implementability, communication ease, and improved power in demonstrating drug efficacy.