Comparison between continuous and discrete doses for model based designs in cancer dose finding.

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmacology

Background:

• Categorizing continuous variables is common in clinical research, particularly in phase I cancer trials for dose finding.
• This practice, despite known information loss, is frequently used to estimate the maximum tolerated dose (MTD).

Purpose of the Study:

• To quantify the information loss from using discrete doses versus continuous doses for MTD estimation in phase I trials.
• To compare the operating characteristics of continuous and discrete dose designs.

Main Methods:

• Utilized model-based designs: Escalation With Overdose Control and Continuous Reassessment Method.
• Employed Monte Carlo simulations across sixteen scenarios, varying dose set spacing and sample sizes.
• Evaluated safety and efficiency measures to quantify information loss.

Main Results:

• Continuous dose schemes significantly minimize information loss when the true MTD is unknown, a common scenario in phase I trials.
• Discrete dose schemes with 9 to 11 doses demonstrated comparable results to continuous schemes.
• The choice of discrete dose intervals and sample size impacts the degree of information loss.

Conclusions:

• Continuous dose escalation designs are preferable in phase I trials to minimize information loss for MTD estimation.
• If discrete doses are necessary, a selection of 9 to 11 doses can achieve similar efficiency and safety outcomes.
• Careful consideration of dose spacing is crucial when implementing discrete dose designs.