Maria Garcia-Cremades (1), Victor Mangas SanJuan (1), Iñaki F Troconiz (1), Gary Mo (2), Celine Pitou (2), Philip W Iversen (3), Johan E Wallin (2)
(1) Department of Pharmacy and Pharmaceutical Technology, School of Pharmacy, University of Navarra, Pamplona, Spain, (2) Global PK/PD & Pharmacometrics, Eli Lilly and Company (3) Lilly Research laboratories, Eli Lilly and Company
Objectives: Pharmacokinetic/Pharmacodynamic characterization of anti-tumor drug effects using xenograft studies [1] is an important process during early stages of oncology drug development. A common feature in those studies is the presence of experimental variability (i. e., same tumor cell lines), and it is not uncommon to note discrepancies in model-related parameters and response outcomes (i.e., % tumor growth inhibition). Proper characterization of different levels of variability (individual and experimental) may facilitate interpretation of individual experiments, as well as optimal design. The objectives of this work are to quantify the extent and the impact of inter-study variability on the model parameters reflecting initial tumor conditions (TS0) and proliferation rate (lambda1).
Methods: Longitudinal tumor volume data from animals (n=239) that received saline administration were used for this analysis. The selected analysis dataset represent different lung cancer tumor cell lines (n=12). For each type of cell line the number of experiments varied from 1 to 13. All the analyses were performed with NONMEM 7.3. Different structural models were fit to the data to characterize the unperturbed tumor growth kinetics. Once the base population model was selected for each cell line, the study variable was incorporated into the model either as a non-ordered categorical covariate, or as second level of variability through the $LEVEL functionality available in NONMEM 7.3.
Results: The unperturbed tumor growth model proposed by Simeoni et al., provided a fair description of the data in all the twelve different tumor cell lines. Including the study variable as second level of variability was significant (p<0.001) in those studies with more than three repeated experiments, and for the two main parameters of the model. The magnitude of the inter-study variability in TS0 and l1 was estimated between 20-30% and was in the same range compared to the inter-animal variability (10-42%).
Conclusions: This analysis shows that inter-study variability is present in typical xenograft experiments with a moderate magnitude and can be detected with precision in cases with at least four repeated experiments. Current available tools in Pharmacometrics allow proper handling of study effects beyond the covariate effects. Ongoing research focuses on the role of study effects in the perturbed growth model, and its similarities across different types of cancers.
References:
[1] Simeoni M, Magni P, Cammia C, De Nicolao G, Croci V, Pesenti E, Germani M, Poggesi I, Rocchetti M. Cancer Research 64: 1094-101.
Reference: PAGE 25 (2016) Abstr 4699 [www.page-meeting.org/?abstract=4699]
Poster: Methodology - New Modelling Approaches