Incorporating Historical Control Data When Comparing Tumor Incidence Rates

Area of Science:

• Toxicology
• Statistical analysis
• Carcinogenesis research

Background:

• The U.S. National Toxicology Program (NTP) collects extensive tumor data from control animals in carcinogenicity studies.
• Current NTP methods do not statistically integrate historical control data, limiting sensitivity in trend detection.
• Informal use of historical control data lacks statistical rigor for evaluating carcinogenicity.

Purpose of the Study:

• To propose a novel survival-adjusted statistical test for detecting dose-related tumor trends.
• To incorporate historical control tumor rates and their inter-study variability into carcinogenicity assessments.
• To improve the interpretation of rare tumor occurrences in treated animals compared to historical controls.

Main Methods:

• Development of a survival-adjusted statistical test for tumor incidence.
• Incorporation of historical control data and adjustment for study-to-study variation.
• Extensive simulations using realistic scenarios to compare the proposed test with the current NTP method.

Main Results:

• The proposed test demonstrates strong performance in simulations compared to the standard NTP approach.
• The new method significantly increases statistical power for detecting dose-related tumor trends.
• A case study on benzophenone demonstrated higher significance (P = 0.004) using the proposed test versus the current method (P = 0.074) for histiocytic sarcoma.

Conclusions:

• The proposed survival-adjusted test offers a statistically robust method for analyzing carcinogenicity data.
• Integrating historical control data enhances the detection of subtle dose-related tumor trends.
• This approach provides a more sensitive and accurate evaluation of potential carcinogens in animal bioassays.