Effective Early Phase Oncology Study Design

The main objectives of the study were to:

• Investigate the safety profile
• Identify the Recommended Phase II Dose (RP2D)
• Gain preliminary evidence of anti-tumor activity

Client Challenge

The main challenge was optimizing the study design to consider the client needs—reasonable total number of enrolled patients and controlled study duration—while ensuring patient safety and close data follow-up.

Additionally, the study needed to provide sufficient evidence to support further development and be used for regulatory purposes.

Our Approach

Given the multiple objectives, the study was divided into three parts: dose-escalation, dose-expansion, and Phase II, each with different design elements using a Bayesian approach leveraging the accumulation of information during the trial.

For the dose-escalation part to assess the RP2D, we implemented a Bayesian Optimal Interval (BOIN) design. BOIN design is as easy to implement and understand as traditional 3+3 designs and shows better performance equivalent to more complex model-based designs but without complexity. It optimizes the allocation of patients to effective dose levels while controlling for overdosing and considers the ongoing evaluation of toxicity for all treated patients when assigning dose levels to new patients.

In addition, an accelerated titration phase, where dose escalation can be conducted after a single patient has been dosed at a dose level without any safety concern, was applied to minimize patient allocation to the lowest dose levels and reduce underdosing.

For the dose-expansion part, a toxicity monitoring design following the same principle as dose escalation was implemented with stopping boundaries for overdose control based on dose-limiting toxicities.

Finally, we implemented a Bayesian Optimal design for Phase II with efficacy monitoring based on Objective Response Rate (ORR) allowing more than one interim analysis while controlling the type I error and maximizing the study power under the null hypothesis for a given fixed sample size. Two interim looks were planned with stopping boundaries for futility based on ORR to ensure the study would only continue if the treatment was deemed promising.

All design elements were based on target toxicity rate and ORR estimated from literature for similar products and available client data and simulations performed in R to assess different scenarios.

Outcomes

By working with our biostatisticians and implementing a Bayesian Optimal design, we were able not only to ensure a successful FDA submission for the client, with the design being accepted as part of an IND submission but also to design a Phase I/II study that delivers quality data to guarantee patient safety.

References

• Liu S. and Yuan, Y. (2015). Bayesian optimal interval designs for phase I clinical trials, Journal of the Royal Statistical Society: Series C , 64, 507-523.Cancer Research, 24(18), 4357-4364.
• Zhou, H., Lee, J. J., & Yuan, Y. (2017). BOP2: Bayesian optimal design for phase II clinical trials with simple and complex endpoints. Statistics in Medicine, 36: 3302-3314