Abstract A critical component of the early development of anticancer agents is the assessment of the Maximum Tolerated Dose during Phase I. In practice, clinical toxicities are graded on a scale of severity, ranging from Grade 1 (least severe) to Grade 4. This categorical scale is also used for continuous readouts, such as neutropenia. As the underlying pathophysiology can be expected to remain the same, and vary by degree or intensity alone, the dose-response (or concentration- response) of lower-grade toxicities should foreshadow the incidence of higher-grade toxicities, which represent different thresholds on the same concentration-response curve. If the concentration- response of Grade 4 toxicity can be predicted from that of Grade 1 toxicity, then the accurate estimation of the MTD should be feasible from a careful estimation of the Grade 1 toxicity. In this work, we demonstrate the development of anticipatory toxicity models that are capable of predicting higher-grade toxicities during dose-escalation. First, we used a simple empirical approach that relies on a minimum of assumptions. We have previously shown that the nadir of neutropenia is strongly correlated with the maximum moving average concentration over the dosing window. Using this pharmacokinetic parameter as an independent variable, we developed a set of analytical expressions to connect the concentration- response of each lower grade of toxicity to the corresponding higher grades of toxicity. Next, we tested this family of anticipatory toxicity models on a simulated dataset of neutropenia generated from a previously published model of clinical neutropenia for a range of chemotherapeutics (Friberg et al., 2002). Finally, we validated our anticipatory toxicity models with in-house preclinical data. Next, we extended the development of anticipatory toxicity models to a model-based approach. We fitted the lower grades of toxicity from the simulated dataset described above to fit a semi-mechanistic model of neutropenia, showing that the model can predict higher grades of toxicity. Next, we again validated the anticipatory toxicity models built on a semi-mechanistic basis using in-house preclinical data. Each of the two approaches presented here may be better suited for different situations. The empirical approach is better suited for toxicities with a poorly understood mechanistic basis, while the semi-mechanistic approach is well suited for neutropenia and other hematological toxicities. In either case, the anticipatory toxicity models were able to accurately predict the incidence of higher-grade toxicities in the given datasets. Thus, a careful analysis of these lower-grade toxicities may provide the opportunity to accelerate clinical development through the design of alternative clinical dose-escalation schemes. Such alternative escalation schemes may also be safer, as they focus on the forward prediction of the MTD before it is reached. Citation Format: Ekta Kadakia, Snehal Samant, Christopher J. Zopf, Dean Bottino, Greg Hather, Santhosh Palani, Wen Chyi Shyu, Arijit Chakravarty. Predicting Maximum Tolerated Dose during Phase I using anticipatory toxicity models. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 3747. doi:10.1158/1538-7445.AM2014-3747
Read full abstract