A dose-response test via closed-form solutions for constrained MLEs in survival/sacrifice experiments

Area of Science:

• Toxicology
• Biostatistics
• Cancer Research

Background:

• Estimating tumor onset in animal carcinogenicity studies is challenging as tumors are often not clinically observable.
• Current methods rely on complex, computer-intensive numerical solutions for survival and tumor onset functions.
• Existing approaches often require cause-of-death information, limiting their applicability.

Purpose of the Study:

• To derive closed-form solutions for nonparametric maximum likelihood estimators of tumor onset and survival functions.
• To develop a novel dose-response trend test that does not require cause-of-death data.
• To evaluate the performance of the new test through simulation and real-world examples.

Main Methods:

• Derived closed-form solutions for nonparametric maximum likelihood estimators under inequality constraints.
• Developed a modified Poly-k test using estimated tumor onset survival functions.
• Applied weighted least squares regression to construct the dose-response trend test.

Main Results:

• The proposed methods provide closed-form solutions, avoiding complex numerical computations.
• The new dose-response trend test effectively estimates trends without needing cause-of-death information.
• Simulation studies demonstrated the proposed test's performance compared to existing methods.

Conclusions:

• The derived closed-form estimators simplify the analysis of tumor onset and survival functions in carcinogenicity studies.
• The novel dose-response trend test offers a robust alternative, particularly when cause-of-death data is unavailable.
• These advancements facilitate more accurate and accessible risk assessment in preclinical studies.