Ida Netterberg (1), René Bruno (2), Yachi Chen (3), Helen Winter (3), Jin Y Jin (3), Lena E. Friberg (1)
(1) Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden, (2) Department of Clinical Pharmacology, Genentech-Roche, Marseille, France, (3) Department of Clinical Pharmacology, Genentech, South San Francisco, California, USA
Objectives: Atezolizumab (an immunooncology checkpoint inhibitor) was administered to 88 non-small cell lung cancer (NSCLC) patients in the phase I, dose-escalation study PCD4989g [1]. The patients received doses of 1-20 mg/kg or a fixed dose of 1200 mg. A large number of biomarkers were studied in this study, which opened up opportunities to explore biomarker relationships with tumor size (TS) and overall survival (OS). TS together with atezolizumab concentrations and biomarker concentrations (e.g. Interleukin 18, IL-18, and interferon-inducible T-cell alpha, I-TAC) were measured longitudinally. Based on these data we previously developed a pharmacokinetic (PK)-TS-IL-18 model, where the drug area under the curve (AUC) and model-predicted relative change from IL-18 baseline at day 21 (RCFBIL-18,d21) were identified as predictors of TS changes [2]. A large RCFBIL-18,d21 predicted a prolonged suppression of tumor growth, which is in line with a slower apparent growth rate and longer survival [3]. The aim of this study was to evaluate the potential of early biomarker effects, i.e. IL-18 and I-TAC, together with other model-derived metrics (i.e. PK and TS related) and baseline covariates, as predictors of OS in the same population. Since patients may be exposed to other anticancer drug treatment after atezolizumab treatment stopped, it was also investigated what impact the duration of the follow-up time has on the inclusion of predictors.
Methods: Five different distributions were evaluated to describe the baseline hazard for three different censoring times, i.e. (i) all available data (AAD, 69 deaths), (ii) censored 2 years after start of treatment (C2YASOT, 54 deaths) and (iii) censored 5 months after last dose (C5MALD, 28 deaths), using time-to-event modelling. The median time to death in the full data set (AAD) was 1.4 years and the follow-up times ranged from 16 days to 5.2 years (AAD), 2 years (C2YASOT) and 4.7 years (C5MALD). Baseline covariates (n=32) were evaluated in a stepwise covariate modelling (SCM) procedure for each of the three censoring times. Thereafter, PK, biomarker and TS model-derived variables were evaluated on top of the baseline covariates. Lastly, all added covariates were excluded one by one in a backward deletion step to arrive at a final model. A p-value of 0.05 was used for statistical significance in both forward and backward steps. The analysis was performed by joining the OS model with the PK-TS-IL18 model in a population PK parameter and data approach [4] in NONMEM 7.4.
Results: The baseline hazard functions were best described by the exponential (AAD and C2YASOT) and Gompertz (C5MALD) distributions. A summary of the included covariates at the different steps in the analysis is presented below. None of the evaluated IL-18 or I-TAC variables, or atezolizumab AUC, added predictive value on top of the baseline covariates, while all TS related variables resulted in p-values
|
AAD1 |
C2YASOT1 |
C5MALD1 |
AAD2 |
C2YASOT2 |
C5MALD2 |
AAD3 |
C2YASOT3 |
C5MALD3 |
|
LYM ALP PD-L1 |
NLR ALP SMK PD-L1 |
NLR ALP race SMK AST |
RCFB-TS(t) |
TSR6 |
TSR12 |
LYM ALP RCFB-TS(t) |
NLR TSR6 |
NLR ALP race AST TSR12 |
Included covariates: 1Baseline (SCM), 2Model-derived and 3Final model
LYM: lymphocyte count; ALP: alkaline phosphatase; PD-L1: programmed death-ligand 1 expression; NLR: neutrophil/lymphocyte ratio; SMK: smoking status; RCFB-TS(t): relative change from baseline tumor size time course; tumor size ratio week 6; TSR12: tumor size ratio week 12
Conclusions: Relationships between the tumor time-course and OS were established based on early Phase I study data in a large cohort of patients with NSCLC, regardless of censoring time. However, although IL-18 was used as a predictor of tumor size changes, the biomarker was not directly a predictor of OS. The included covariates (both baseline and model-derived) varied slightly dependent on the censoring time, but those selected were in general correlated.
References:
[1] Herbst RS, Soria J-C, Kowanetz M, Fine GD, Hamid O, Gordon MS, et al. Predictive correlates of response to the anti-PD-L1 antibody MPDL3280A in cancer patients. Nature 2014;515:563–7.
[2] Netterberg I, Li C-C, Molinero L, Budha N, Sukumaran S, Stroh M, et al. A PK/PD Analysis of Circulating Biomarkers and Their Relationship to Tumor Response in Atezolizumab-Treated non-small Cell Lung Cancer Patients. Clin Pharmacol Ther 2018.
[3] Claret L, Jin JY, Ferté C, Winter H, Girish S, Stroh M, et al. A Model of Overall Survival Predicts Treatment Outcomes with Atezolizumab versus Chemotherapy in Non-Small Cell Lung Cancer Based on Early Tumor Kinetics. Clin Cancer Res 2018;24:3292–8
[4] Zhang L, Beal SL, Sheiner LB. Simultaneous vs. sequential analysis for population PK/PD data I: best-case performance. J Pharmacokinet Pharmacodyn 2003;30:387–404.
[5] Hooker AC, Karlsson MO, The Kaplan-Meier Mean Covariate plot (KMMC): a new diagnostic for covariates in time-to-event models, PAGE 21 (2012) Abstr 2564
Reference: PAGE 28 (2019) Abstr 8886 [www.page-meeting.org/?abstract=8886]
Poster: Drug/Disease Modelling - Oncology