Tumor burden modeling versus progression-free survival for phase II decision making

Area of Science:

• Oncology
• Clinical Trial Design
• Biostatistics

Background:

• Randomized Phase II oncology trials commonly use progression-free survival (PFS) and changes in tumor burden (sum of longest diameters, SLD) as endpoints.
• Tumor shrinkage and growth dynamics can be modeled from patient-specific SLD profiles over time.

Purpose of the Study:

• To compare the ability of PFS and SLD analyses to identify active oncology drugs in simulated Phase II trials.
• To determine the optimal endpoint for making decisions about initiating Phase III trials based on Phase II results.

Main Methods:

• Simulated 6 Phase II trials (5 positive, 1 negative) by resampling data from larger trials.
• Analyzed simulated trials using log-rank test for PFS, Wilcoxon rank-sum test for minimum SLD change, and two nonlinear mixed-effects models for SLD profiles.
• Assessed P-value distributions from each analysis in both positive and negative simulated trials.

Main Results:

• All four analysis methods yielded uniformly distributed P-values in the negative trial.
• PFS analysis consistently performed best or near-best in the 5 positive trials.
• SLD modeling analysis only outperformed minimum SLD analysis in one positive study.

Conclusions:

• Progression-free survival (PFS) is a more robust endpoint than sum of longest diameters (SLD) for guiding Phase III trial initiation in oncology.
• PFS analysis demonstrates superior ability to detect drug activity compared to various SLD analysis methods in Phase II oncology trials.