Screening for possible human carcinogens and mutagens. False positives, false negatives: statistical implications.

Area of Science:

• Toxicology
• Biostatistics
• Genetics

Background:

• Carcinogen screening assays yield four outcomes: true positive, true negative, false positive, and false negative.
• These outcomes resemble statistical hypothesis testing results, involving false positive and false negative rates.

Purpose of the Study:

• To analyze the statistical issues in interpreting carcinogen screening assay results.
• To explore appropriate statistical methods for analyzing assay data and selecting effective test batteries.
• To differentiate statistical considerations in assay interpretation from those in hypothesis testing.

Main Methods:

• Comparison of assay outcome categorization with statistical hypothesis testing.
• Application of statistical methods like Bayes' theorem and multivariate clustering for assay data analysis.
• Evaluation of statistical significance versus biological importance in genotoxicity studies.

Main Results:

• Assay interpretation involves distinct statistical issues compared to hypothesis testing.
• Bayes' theorem and clustering techniques can enhance the prediction of potential carcinogens.
• Conclusions are sensitive to estimates of sensitivity, specificity, statistical methods, and data characteristics.

Conclusions:

• Statistical analysis of carcinogen screening should focus on effect estimation and exploration over formal hypothesis testing.
• Appropriate statistical designs and analyses are crucial for genotoxicity experiments.
• Accurate interpretation requires careful consideration of statistical methods and data quality.