Global tests for analysis of multiple tumour data from animal carcinogenicity experiments

Area of Science:

• Toxicology
• Biostatistics
• Animal Models

Background:

• Chronic animal carcinogenicity bioassays are crucial for assessing chemical carcinogens.
• Simultaneous evaluation of multiple tumor types and sites presents statistical challenges.
• Existing methods may not fully capture the overall dose-response effect across diverse tumor outcomes.

Purpose of the Study:

• To develop and present novel statistical test procedures for evaluating the overall dose effect on all tumor types and sites concurrently within a carcinogenicity bioassay.
• To provide a unified approach for analyzing tumor prevalence and onset rates.
• To illustrate the application of these tests using a real-world carcinogenicity experiment.

Main Methods:

• Utilized generalized estimating equations (GEE) for deriving global tests of tumor prevalence and onset rates.
• Employed logistic regression models for simultaneous testing of non-lethal tumors.
• Applied the Poly-3 test for analyzing mixtures of lethal or non-lethal tumors across different types and sites.
• The core statistical model assumes a common dose-effect alternative.

Main Results:

• The proposed generalized estimating equations approach provides a robust framework for simultaneous dose-effect testing.
• Logistic regression and Poly-3 tests offer specific solutions for different tumor characteristics (lethal/non-lethal) and multiplicity.
• Demonstrated the practical utility of these tests in analyzing a carcinogenicity experiment involving hepatocellular and bladder tumors.

Conclusions:

• The developed statistical tests enable a more comprehensive assessment of carcinogenicity by considering all tumor types and sites simultaneously.
• These methods enhance the ability to detect overall dose-related effects in chronic bioassays.
• The application confirms the utility of the proposed tests in toxicological risk assessment.