Using R to foster the communication with non-statisticians on Bayesian dose escalation models


Bayesian model-based dose-escalation designs, including one and two parameter logistic regression models, have meanwhile proven themselves in Phase I dose-escalation trials (Iasonos and O’Quigley, 2014 [1]). Compared to rule/algorithm-based designs such as the 3+3 design, model-based designs have the advantage of being more flexible in choosing the target toxicity rate and cohort size. Bayesian modeling allows to combine prior knowledge of the drug (e.g. from animal tox studies, data from comparator drugs or data from other studies) with the observed data from the current trial. The model-based approach accounts for uncertainty, optimize dose recommendations (balance of risk versus benefit for patients) and allows for dose de-escalation as well as dose re-escalation. Recent research has shown that model-based approaches are more reliable in estimating the maximum tolerated dose (MTD) and allocating less patients to ineffective or excessively toxic doses (Jaki et al., 2019 [2]). Because of their higher complexity, model-based designs are still seen critical among clinicians (Le Tourneau et al., 2012 [3]), thus the classical 3+3 design is still widely implemented due to its simplicity and transparency. Conaway et al., 2019 [4] have highlighted that the choice of design in early phase affects the outcome of the drug development process, therefore more attention should be paid to early-stage designs. In this talk we will showcase how we bring Bayesian dose-escalation models closer to non-statisticians by means of R. We will discuss how we plan, implement and communicate the dose-escalation design to the clinicians. Moreover, we show how we present our results to the safety monitoring committee (SMC) and how we support the SMC in dose escalation decisions. We use a simple data example to illustrate the proposed methodology. All will be discussed within the context of using R as primary tool. [1] Iasonos, A. and O’Quigley, J (2014). Adaptive Dose-Finding Studies A Review of Model-Guided Phase I Clinical Trials. Journal of clinical oncology, 2014; 32(23). [2] Jaki, T., Clive, S. and Weir, C.J. (2013). Principles of dose finding studies in cancer a comparison of trial designs. Cancer Chemother Pharmacol 711107-1114. [3] Le Tourneau C, Gan HK, Razak ARA, Paoletti X (2012). Efficiency of New Dose Escalation Designs in Dose-Finding Phase I Trials of Molecularly Targeted Agents. PLoS ONE 7(12) e51039. [4] Conaway, M.R. and Petroni, G.R. (2019). The impact of early phase trial design in the drug development process. Clinical Cancer Research.

Presented at 2019 Conference