Tumor Dynamic Model-Based Decision Support for Phase Ib/II Combination Studies: A Retrospective Assessment Based on...

Area of Science:

• Oncology
• Clinical Trials
• Biostatistics

Background:

• Model-based tumor growth inhibition (TGI) metrics are crucial for go/no-go decisions in early clinical studies.
• Evaluating TGI metrics in Phase III trials of immunotherapy combinations is essential for validating their predictive power.

Purpose of the Study:

• To independently evaluate model-based TGI metrics using data from a Phase III trial of atezolizumab plus bevacizumab plus chemotherapy (ABCP) versus bevacizumab plus chemotherapy (BCP) in non-small cell lung cancer.
• To assess the performance of TGI metrics in mimicking Phase Ib/II study decision-making.

Main Methods:

• Resampled data from the IMpower150 Phase III trial (1,202 patients) to simulate smaller Phase Ib/II studies (15-40 patients/arm, 6-24 weeks follow-up).
• Estimated TGI metrics using a bi-exponential TGI model and calculated effect sizes (geometric mean ratio [GMR], objective response rate [ORR] difference, progression-free survival [PFS] hazard ratio [HR]).
• Evaluated correct and incorrect go decisions based on simulated study designs and TGI metric thresholds.

Main Results:

• For 40 patients and 24 weeks follow-up, correct go decision rates were 83% (KG GMR <0.90), 69% (dORR >0.10), and 58% (PFS HR <0.70), with low incorrect go decision rates (4-12%).
• TGI metrics consistently outperformed RECIST endpoints across various simulated designs.
• Predicted overall survival (OS) hazard ratio was consistently around 0.80.

Conclusions:

• Model-based estimation of tumor growth rate (KG) GMR serves as a valuable exploratory endpoint.
• These findings support the use of TGI metrics for informing early clinical decisions in investigational combination studies for non-small cell lung cancer.