A curve similarity approach to parallelism testing in bioassay

Area of Science:

• Pharmacology
• Biostatistics
• Analytical Chemistry

Background:

• Bioassay potency determination relies on relative measures, necessitating parallelism assessment between reference standards and sample dose-response curves.
• Current methods like t-tests and equivalence tests statistically assess model parameters, not direct curve similarity.
• This indirect approach can be less effective in evaluating true parallelism.

Purpose of the Study:

• To propose a novel, direct curve similarity approach for evaluating parallelism in bioassays.
• To provide both frequentist and Bayesian implementations of this new method.
• To demonstrate the effectiveness of the curve similarity approach in detecting parallelism and non-parallelism.

Main Methods:

• Quantifying and normalizing the area between dose-response curves to create a composite parallelism measure.
• Testing the hypothesis that a sample is a simple dilution or concentration of a reference standard.
• Utilizing a simulation study to compare the new method with traditional approaches.

Main Results:

• The proposed curve similarity approach provides a direct measure of parallelism.
• Both frequentist and Bayesian versions of the test were developed.
• Simulation studies confirmed the approach's effectiveness in identifying parallelism and non-parallelism.

Conclusions:

• The curve similarity approach offers a more intuitive and effective method for assessing parallelism in bioassays compared to traditional statistical tests.
• This method directly evaluates the similarity of dose-response curves, overcoming limitations of parameter-based statistical assessments.
• The approach is robust and applicable for both frequentist and Bayesian analyses in bioassay development and quality control.