The Tukey trend test: Multiplicity adjustment using multiple marginal models.

Area of Science:

• Biostatistics
• Pharmacometrics
• Toxicology

Background:

• Dose-response analysis with multiple, differently scaled endpoints presents challenges in selecting appropriate model shapes.
• Existing methods for handling multiple endpoints often rely on heuristic adjustments for multiple testing, which can be inadequate due to correlated parameter estimates.

Purpose of the Study:

• To develop a statistically rigorous method for dose-response analysis with multiple endpoints.
• To provide an accurate adjustment for multiple testing that accounts for correlations between dose effect estimates.

Main Methods:

• The proposed methodology derives an asymptotic correction for multiple testing from the score functions of marginal regression models.
• It utilizes a multivariate t-distribution to provide a one-step adjustment of p-values, accounting for inter-model correlations.
• The approach is demonstrated using generalized linear models, mixed-effect models, and simulation studies.

Main Results:

• The developed method effectively adjusts for multiple testing in dose-response analyses with multiple, correlated endpoints.
• Simulations and real-world examples confirm the accuracy and utility of the proposed p-value adjustment technique.
• The methodology is robust across different model types and endpoint scales.

Conclusions:

• The new methodology offers a significant advancement in dose-response analysis, particularly for complex datasets with multiple endpoints.
• It provides a reliable and efficient way to manage multiple testing issues, leading to more accurate conclusions.
• An R package is available for implementing this advanced statistical approach.