Abstract
Pre-clinical cancer research often involves animal studies to evaluate the safety and efficacy of new treatments before they reach human clinical trials. However, these studies raise significant ethical concerns about animal welfare and have led to stringent regulations, in particular those that adhere to the ""3Rs"" principle in research: Replacement, Reduction, and Refinement of animal use. In the study of cancer growth-curves, a common application of the 3Rs principle is to ensure that animals are euthanized when their pain level is considered too high, as evaluated by a veterinary, or when the tumor size is larger than a set threshold. Another application of the 3Rs principle concerns the statistical planning, design, and sample size calculation of the experiment, since errors made at this level can have severe consequences on both results and the design of future experiments.
From a statistical modelling and inferential point of view, with cancer research experimental data, the potential animal culling/dropout (or data censoring) needs to be considered in order to assess, for example, the treatment effect measured by the difference in growth curves between treatment and control groups. In particular, false discovery rates can increase drastically (or test sizes decrease drastically) when the censoring mechanism is not correctly modeled. Therefore, within a general framework for inference with complex data settings, we propose a simple statistical procedure to produce correct inferential tools when data are censored, below, above or by intervals, for censoring data mechanisms that can be modeled according to the experimental settings. The proposed approach does not require separate analytical developments for different models and/or censoring data mechanisms, making it easy to implement in a variety of experimental situations. Additionally, the censoring thresholds are also allowed to be random and to depend on the growth size of the curves, which is a more realistic practical situation, as attested by a case study on tumour growth inhibition that is used as a motivating example. The theoretical results are based on Zhang et al. (2023) and Orso et al. (2024).
Collegamento Microsoft Teams
Organizzazione
Silvia Cagnone