2018 - Montreux - Switzerland

PAGE 2018: Drug/Disease modelling - Oncology
Johan Wallin

Validation of Xenograft Dose Predictions for Clinical Efficacy in NSCLC

Eva Hanze (1), Lars Lindbom (1), Johan Wallin (2)

(1) qPharmetra, (2) Lilly PKPD&PMx

Objectives:

The xenograft mouse model is widely used to study the response to cancer therapy. Pharmacokinetic-pharmacodynamic (PKPD) modelling of xenograft data can be performed to predict the exposure level to target in the clinical setting. However, how well these dose predictions translate to clinical efficacy is not well studied although previous attempts have been done. One example is the work done by Rochetti et al[1] where they demonstrated a good correlation between preclinical potency parameters and exposure obtained at therapeutic doses in the clinic for a range of cytotoxic agents.

The objective with this work was to further evaluate the correlation between preclinical and clinical efficacy estimates, including targeted agents representing both small molecules and biologics. Preclinical efficacy estimates derived from xenograft experiments across a range of cell lines, as well as clinical efficacy estimates from failed or successful late phase drug development programs in NSCLC, were collected from both in-house data as well as published data.

Methods:

One central task in this work was to collect data on preclinical and clinical efficacy estimates. The literature was surveyed for oncology compounds with Phase 3 results in NSCLC published in the last 10 years (2006-2016). In addition, a selected number of standard of care compounds were included in the analysis. Substances of interest were targeted kinase inhibitors, monoclonal antibodies (mAbs) and cytotoxics.

The preclinical anti-tumor potency parameter k2, derived using the model developed by Simeoni et al [2], was of primary interest as an efficacy estimates. Ideally, preclinical k2 values should be directly compared to translated clinical k2 estimates. However, due to limitations in the clinical information, clinical EC50-values were used. For approved compounds with no identified exposure-response (ER) it was assumed that EC50 was less than lowest exposure quartile in the effective dose. For failed compounds with no identified ER it was assumed that the EC50 was higher than highest exposure quartile for the tested dose.

The analysis of correlation of preclinical-clinical efficacy estimates was performed in NONMEM (version 7.3.0). The regression was performed on log-log scale and the M3 method was used to account for the cases where the EC50 was assumed to be less than lowest exposure quartile or higher than the highest exposure quartile observed.

Results:

In total, 33 approved, 17 failed and 15 standard of care compounds in various indications were identified. The type of clinical ER information reported varied between the compounds, including Emax type of modelling, time-to-event analysis, logistic regression using a cut-off concentration to cases with no published ER information.  In total, 9 NSCLC compounds had sufficient information to be included in the correlation analysis. The correlation between preclinical k2 estimates and clinical EC50 estimates was found to be relatively high (r = 0.90).

Conclusions:

A relatively strong correlation was found between the preclinical and clinical efficacy parameters, supporting the use of xenograft models to predict clinical therapeutic doses. However, this analysis was limited to data from only 9 compounds and based on only one k2 estimate from one single cell-line for each compound. Further work is ongoing to include data from additional cell-lines for each compound. 



References:
[1] M. Rocchetti, M. Simeoni, E. Pesenti, G. De Nicolao, I. Poggesi. Predicting the active doses in humans from animal studies: A novel approach in oncology. European journal of cancer August 2007, Volume 43, Issue 12, Pages 1862–1868. DOI: https://doi.org/10.1016/j.ejca.2007.05.011
[2] Simeoni M, Magni P, Cammia C, De Nicolao G, Croci V, Pesenti E, Germani M, Poggesi I, Rocchetti M. Predictive pharmacokinetic-pharmacodynamic modeling of tumor growth kinetics in xenograft models after administration of anticancer agents. Cancer Research. 2004 Feb 1;64(3):1094-101. DOI: 10.1158/0008-5472.CAN-03-2524
[3] Bergstrand M, Karlsson MO, Handling data below the limit of quantification in mixed effect models. AAPS J. 2009 Jun;11(2):371-80. DOI: 10.1208/s12248-009-9112-5.


Reference: PAGE 27 (2018) Abstr 8433 [www.page-meeting.org/?abstract=8433]
Poster: Drug/Disease modelling - Oncology
Click to open PDF poster/presentation (click to open)
Top