Expected vs observed effect sizes for survival endpoints in phase 3 oncology trials

Area of Science:

• Oncology
• Clinical Trial Design
• Biostatistics

Background:

• Randomized phase 3 oncology trials frequently fail primary endpoints.
• Phase 1 and 2 trial results do not always predict phase 3 success.
• Effect size estimation for trial power is a critical factor.

Purpose of the Study:

• To investigate effect size selection practices in phase 3 oncology trials.
• To determine if effect sizes for survival endpoints are overestimated.
• To assess the impact of these practices on trial power and failure rates.

Main Methods:

• Systematic review of phase 3 oncology trials published in 2023.
• Analysis of effect sizes for survival endpoints, specifically hazard ratios (HR).
• Comparison of expected HRs used for power calculations versus observed HRs from primary analyses.

Main Results:

• A pattern of effect size overestimation for survival endpoints was identified.
• Expected hazard ratios (average HR = 0.66) were stronger than observed (average HR = 0.72).
• 57.3% of observed hazard ratios were weaker than expected across 111 trials.

Conclusions:

• Phase 3 oncology trials may be systematically underpowered due to effect size overestimation.
• Underpowered trials contribute significantly to the high failure rate in oncology research.
• Revising effect size selection practices is crucial for improving the success of oncology clinical trials.