Sofiene Laarif(1), Sreenath M Krishnan(1), Brendan Bender(2), Angelica Quartino (2), Lena E Friberg(1)
(1) Dept. of Pharmaceutical Biosciences, Uppsala University, Sweden, (2) Genentech Inc., San Francisco, CA, USA
Objectives: In metastatic cancer, the growth and drug-induced shrinkage of individual tumor lesions may be highly dependent on the microenvironment of the hosting organ, and individual lesions may contribute differently to overall disease progression and survival. Lesions within the same organ may be more likely to evolve and respond in a similar manner than lesions from different organs. In the traditional analysis of tumor response to treatment, an overall measurement of patient tumor burden is used, i.e. the sum of the longest diameters (SLD) of up to 5 lesions (RECIST criteria v 1.1). Information of the impact of individual lesion dynamics on the outcome is hence ignored. The aim of this analysis was to develop population models to better understand and characterize the differences in tumor dynamics between lesions and between metastatic site.
Methods: The dataset consisted of lesion measurement data from 183 subjects with metastatic HER2-negative breast cancer receiving docetaxel at the dose of 100 mg/m2 on the first day of three-week treatment cycles. The treatment continued until disease progression as assessed by RECIST criteria (v 1.0) [1] or an intolerable toxicity was reached. The dataset included up to 10 lesions per individual which were followed up to 145 weeks (median 32 weeks). A kinetic/pharmacodynamic (K/PD) function was driving the effect in a tumor growth inhibition model [2] that characterizes both lesion growth and drug-induced lesion shrinkage. Inter-lesion (ILV), inter-organ (IORV), and inter-individual variability (IIV) were explored in lesion baseline, growth rate and drug-induced shrinkage parameters. Logistic regression models were developed to describe the observed dropout from lesion measurements and the appearance of new lesions. Evaluated predictors included SLD or lesion size, disease progression (defined as 20% increase in SLD from nadir or the appearance of a new lesion), number of total lesions, number of metastatic organs and treatment duration.
Results: In the study population, metastasis were located in 11 different morphological locations of the body. Liver (50% of patients), lymph nodes (46%) and lungs (26%) were the most frequent metastatic sites. The median number of lesions per subject was 3 (range: 1-10). Lesions showed diverse profiles of shrinkage and growth during the study, but were more similar within an organ than between organs.
The lesion model included a baseline value typical for the organ, along with IIV shared across all organs (28 %CV), and ILV that ranged from similar magnitude (kidney, soft tissue) and up to three times higher (pelvis) compared to the IIV. For growth rate, the model included IIV and IORV of similar magnitudes (~80 %CV). The drug-induced shrinkage rate was almost twice as high for liver compared to the other organs.
The probability to dropout from tumor size measurements increased with the appearance of a new lesion, 20% increase from SLD nadir, and treatment duration. The probability of a new lesion increased with a large tumor size at baseline, a large number of lesions at baseline and treatment duration.
The dropout model was applied for visual predictive checks. These demonstrated the models’ capability to adequately describe the typical trend and variability of lesion shrinkage and regrowth.
Conclusion: Inter-lesion, inter-organ, and inter-individual differences were well captured by the developed lesion model. This modeling approach, separating different levels of variability, has the potential to provide a better understanding of drug effect in different organs, and may be used to tailor treatments based on lesion location, lesion size and early lesion response. In a next step, the lesion model and the new lesion appearance model will be applied to explore relationships to survival.
References
[1] Therasseet al. J Natl Cancer Inst. (2000) 92:205-16
[2] Claret L et al. J ClinOncol. (2009) 27(25):4103-8.
Reference: PAGE 28 (2019) Abstr 9063 [www.page-meeting.org/?abstract=9063]
Poster: Drug/Disease Modelling - Oncology