Uncertainty and key factors in assessing drug effect in cell-based and xenograft tumor models in rodents

Simon Zhou (1), Yan Li (1), Yue Gao (1), Ho-pi Lin (1), Angela James (1), Maria Palmisano (1) and Chee Ng (2)

(1) Celgene Cooperation, Summit, NJ, USA, (2)Children’s Hospital of Philadelphia, Philadelphia, PA, USA

Objectives: There have been poor concordance of IC50s by same drugs on identical cancer cell lines by different labs[1]. The goal of this investigation is to develop a mathematical model to capture the key driving forces of tumor growth/shrinkage and essential drug properties in order to better understand their impacts on the estimation of drug effect, eg. IC50.

Methods: A general mathematical model was developed and it well-characterized the drug exposure and tumor growth inhibition in both cell-based models and xenografted tumor models in mice. Closed form mathematical solutions were obtained for drug effect characterization in cell-based models and subjected to parameter sensitive analysis using derivatives. For the xenograft tumor models, numerical sensitivity analysis was conducted by simulations and subsequent PK/PD analyses to better understand the key drivers for tumor growth and shrinkage.

Results: The modeling and analyses indicated that innate tumor growth and death rate are the key driving force for tumor growth and shrinkage. In typical cell-based and xenografted models, the experimental observations are limited to cell number or tumor volume increase without any explicit measure of tumor death. IC50 cannot be accurately identified if high drug concentrations or high doses do not bring the cell number or tumor volume below the initial cell numbers or tumor volumes before drug treatment. And experiments producing qualitatively similar and clustered effects of tumor growth inhibition are not conducive for delineation of drug effect. Given usual experimental variability, different combinations of IC₅₀, K_g, and K_d could produce similar tumor inhibition profiles, further confounding the interpretation of the “true” drug effect.

conclusion: Large variation in drug effect estimation is inherent in either in vitro and in vivo models of tumor growth. Estimation of drug IC50 on tumor growth inhibition is highly sensitive to tumor growth rate and death rate. Reporting the baseline tumor growth rate with IC50 estimated will provide a less-biased measure for comparing different anti-cancer agents in vitro and in vivo.

References:
[1] Haibe-Kains, B. et al. Inconsistency in larger pharmacogenomic studies. Nature: 504,389-394.

Reference: PAGE 23 () Abstr 3257 [www.page-meeting.org/?abstract=3257]

Poster: Methodology - Model Evaluation