CUSUMIN: A cumulative sum interval design for cancer phase I dose finding studies

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Model-assisted designs, such as the Bayesian optimal interval (BOIN) design, are increasingly used in cancer phase I studies for their balance of performance and simplicity.
• These designs often employ sequential test procedures, similar to those used in control charts for process monitoring.

Purpose of the Study:

• To propose a novel model-assisted design for cancer phase I trials utilizing cumulative sum (CUSUM) statistics.
• To evaluate the performance of the proposed CUSUM-based design against existing methods like BOIN, mTPI, and Keyboard.

Main Methods:

• The proposed design calculates the next cohort's dose based on CUSUM statistics derived from dose-limiting toxicity counts and pre-defined boundaries.
• Performance was assessed through intensive simulations, comparing the CUSUM design with BOIN, mTPI, and Keyboard.

Main Results:

• The CUSUM-based design demonstrated superior performance in controlling over-dosing rates compared to BOIN, mTPI, and Keyboard.
• The proposed method maintained comparable performance in determining the maximum tolerated dose.

Conclusions:

• The CUSUM statistic offers a more efficient approach for model-assisted designs in cancer phase I clinical trials.
• This novel design enhances safety by reducing over-dosing while effectively identifying the maximum tolerated dose.