Sreenath M. Krishnan (1), Brendan C. Bender (2), Jin Jin (2), and Lena E. Friberg (1)
(1) Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden. (2) Genentech Inc, San Francisco, CA
Introduction: Population modelling is increasingly used for describing tumor responses to anticancer treatments and different model-derived tumor metrics have been suggested as predictors of overall survival (OS) [1,2]. Bender et al. [3] developed a novel semi-mechanistic population kinetic/pharmacodynamic (KPD) model to characterize tumor growth in HER2–negative metastatic breast cancer patients receiving either docetaxel or paclitaxel treatment. Since the study populations were comparable in terms of cancer type, tumor baseline and patient characteristics, the analysis was here extended by developing a combined model for tumor response for the two taxanes, and tumor size metrics and time-course of tumor size, were explored as predictors of OS.
Objectives: The aims of this study were (a) to develop a combined tumor size model of docetaxel and paclitaxel in HER2- metastatic breast cancer patients, and (b) to investigate predictors including time-dependent covariates, such as tumor size ratio (TSR) and tumor time-course, for overall survival.
Methods:
Tumor data: The combined dataset consisted of tumor response from 185 patients receiving docetaxel treatment (100 mg/m2 infused over 1 hour on day 1 of each 3-week cycle) and 219 patients treated with paclitaxel (90 mg/m2 of paclitaxel infused over 1 hour on days 1, 8, and 15 every 4 weeks). Patients were HER2- metastatic breast cancer women with a median age of 55 years (range 27–85 years). The tumor size was followed for a maximum of 2.8 years (median=0.6 years) and the survival data was collected for a maximum of 4.8 years (median=2.2 years).
Tumor model: The tumor model developed by Bender et al. [3], was applied to fit the combined dataset and evaluated for shared tumor-related parameters between docetaxel and paclitaxel treated patient populations.
OS model: Parametric time-to-event (TTE) models with different probability density functions such as exponential, Weibull, Gompertz, log-normal and log-logistic functions were evaluated for describing the observed survival data. A joint model of tumor-OS was applied for investigating the predictors of OS [4]. The tested predictors were
- Patient baseline characteristics:
- Age
- Tumor baseline
- Tumor model parameters, and tumor model metrics:
- Model-predicted tumor time-course (TS(t))
- Time-varying relative change in tumor size (rTS(t))
- Tumor size ratio week 6 and week 8 (TSRw6 & TSRw8)
- Derivative of TS(t)
- Log-transformed growth rate (log(kg))
- Time-to-tumor-growth (TTG)
Covariates were investigated as time-varying until the time of tumor nadir (TTG), until w6 and w8 occurred (TSR) or until the last tumor size observation (derivative of TS(t)). NONMEM 7.4.3 software was used for model development.
Results: The tumor model consisted of six compartments mimicking tumor quiescence, drug–resistance, and tumor drug–sensitivity. The total observed tumor size is represented as the sum of all 6 compartments. All tumor-related parameters could be shared between the docetaxel and paclitaxel datasets, without causing an apparent deterioration in the model fit, while tumor shrinkage was drug-specific. The model described the data from both drugs well and the uncertainty of the estimated parameters were acceptable.
A parametric TTE model with a Weibull distribution described the observed OS data the best. In univariate analysis, baseline tumor size, derivative of TS(t), tumor time-course (log transformed), TSRw6, and TTG were significant predictors of survival for both docetaxel and paclitaxel treatment. After the best predictor, baseline tumor size, was included, the derivative of TS(t) was the most significant predictor and included in the final OS model. No other predictor improved the model fit further.
Conclusion: The combined model for HER2- breast cancer patients, incorporating tumor shrinkage, tumor quiescence, and tumor regrowth upon resistance to taxane drug exposures, described the individual time-courses well. The results from OS analysis indicated that a large tumor baseline (βbaseline = 0.004 mm-1) and higher tumor burden at dropout from study (βderivative = 0.01 mm-1) were associated with poorer survival. The developed models, with shared tumor-related and OS parameters, may support development of other drugs in this indication.
References:
[1] Bruno et al., Clin Pharmacol Ther. 2014.
[2] Bender et al., Br J Clin Pharmacol 2016.
[3] Bender et al., PAGE 26 (2017) Abstr 7344 [www.page-meeting.org/?abstract=7344]
[4] Zhang et. al., J Pharmacokinet Pharmacodyn (2003).
Reference: PAGE 28 (2019) Abstr 8930 [www.page-meeting.org/?abstract=8930]
Poster: Drug/Disease Modelling - Oncology