Likelihood-based experimental design for estimation of ED50

Area of Science:

• Biostatistics
• Pharmacology
• Radiation Oncology

Background:

• Accurate estimation of the median effective dose (ED50) is crucial in pharmacology and toxicology.
• Confidence interval (CI) length is a key metric for assessing the precision of ED50 estimates.
• Existing dose allocation strategies have limitations in minimizing CI length.

Purpose of the Study:

• To present a dose allocation strategy that minimizes the length of likelihood-based confidence intervals for ED50 estimation.
• To compare this optimal strategy with alternative methods based on asymptotic variance and Fieller's Theorem.
• To explore methods for handling parameter-dependent dose allocations.

Main Methods:

• Development and presentation of a novel dose allocation method focused on minimizing likelihood-based CI length.
• Comparative analysis of the proposed method against existing ED50 estimation strategies.
• Investigation of techniques to manage parameter dependencies in dose allocation.

Main Results:

• The proposed dose allocation strategy effectively minimizes the length of likelihood-based confidence intervals for ED50.
• Demonstrated superiority of the new method over traditional approaches in terms of interval precision.
• Identified effective strategies for parameter-dependent dose allocations.

Conclusions:

• The presented dose allocation method offers improved precision for ED50 estimation.
• This approach is particularly relevant for experimental designs where minimizing uncertainty is paramount.
• Findings have implications for radiation tolerance studies and other dose-response assessments.