The analysis of longitudinal data from life-span carcinogenicity bioassays on Sprague-Dawley rats

Main Article Content

Daria Sgargi
Simona Panzacchi
Daniele Mandrioli
Fiorella Belpoggi
Rossella Miglio

Keywords

longitudinal analysis, body weights, Sprague-Dawley rats, mixed-effects models, carcinogenicity studies

Abstract

Background and aim of the work: Long Term Carcinogenicity Bioassays (LTCB) are among the best instruments to strengthen the evidence on which regulatory agencies base their decision to classify harmful agents as human carcinogens, so they are fundamental to protect public health. The statistical analysis is essential to validate the results from cancer and non-cancer outcomes in carcinogenicity bioassay. This work proposes and applies some methodologies for the analysis of non-cancer outcomes, such as body weights. Methods: We use data from studies already concluded, evaluated and published: 4 bioassays aimed at testing the carcinogenic potential of Coca-Cola on Sprague-Dawley rats of different ages. The analysis of body weights of the second generation of rats was performed using mixed-effects models: linear models were fitted for nonlinear models we considered human non-linear growth functions. Results: Linear models were fitted using the log-transformation of time and polynomial term of third order for time. Sex and treatment influence body weight, age of dams during gestation doesn’t. Growth models: Jenns-Bayley, Count and 1st order Berkey-Reed growth functions were evaluated; the latter best describes the data. Sex and treatment significantly influence all parameters. The direction, magnitude and significance of the effect variable is substantially similar in all models. The analysis of residuals highlights the same issues for all models: the extreme trends in the last part of life heavily affect the models’ performance. Conclusions: Mixed-effects models allowed to account for the structural effect of covariates that act the same way on all individuals, and to add random effects that introduce a correlation among subjects if clustering happens; nonlinear human growth models added information about the whole growth process, therefore these may be useful methods in studies focused on development and sexual maturation.

Abstract 111 | PDF Downloads 44