Francis Williams Ojara

Predictive value of tumour growth kinetics on overall survival during first-line paclitaxel-based combination chemotherapy in advanced non-small cell lung cancer patients

Francis W. Ojara (1,2), Andrea Henrich (1,2), Niklas Hartung (3), Wilhelm Huisinga (3), Markus Joerger (4), Charlotte Kloft (1)

(1) Dept. of Clinical Pharmacy & Biochemistry, Institute of Pharmacy, Freie Universitaet Berlin, Germany, (2) Graduate Research Training Program PharMetrX, Germany, (3) Institute of Mathematics, University of Potsdam, Germany, (4) Medical Oncology and Clinical Pharmacology, Dept. of Internal Medicine, Cantonal Hospital St. Gallen, Switzerland

Introduction: Platinum/paclitaxel doublet chemotherapy is a standard 1st-line option in patients with advanced (stage IIIB/IV) non-small cell lung cancer (NSCLC), with median overall survival (OS) of approximately 10 months [1]. Routine early tumour size assessment during treatment have proven prognostic value in NSCLC [2]. Adequate characterisation of the association between chemotherapy exposure and tumour growth kinetics would establish a quantitative exposure-OS relationship, enabling better prediction of treatment response. This would facilitate timely intervention in non-responders e.g. adopting alternative regimens to spare patients from needless toxicities while improving therapeutic benefit.

Objectives

  1. Characterise change in tumour size across time and quantify the impact of paclitaxel (PTX) exposure on tumour size in PTX-treated NSCLC
  2. Characterise OS in PTX-treated NSCLC
  3. Quantify the predictive value of early change in tumour size on OS in PTX-treated NSCLC

Methods: Patients (n=365) from the CEPAC-TDM study, receiving 3-weekly PTX in combination with carboplatin (target AUC 6 mg.min/mL) or cisplatin 80 mg/m2 for ≤ 6 cycles were analysed [3]. PTX was dosed either as 200 mg/m2 (n=183) or following a pharmacokinetic (PK)-guided dosing (n=182), with PTX PK only available in the PK-guided dosing arm. Longitudinal tumour size data (sum of longest diameters at baseline, week 6, 12, 18 and after treatment), and survival data (time of death) were documented. A tumour growth inhibition (TGI) model, describing tumour growth and drug-induced tumour decay [4], was developed based on the PK-guided dosing arm data. An exponentially declining drug effect, based on PTX area under the plasma concentration-time curve from start to the end of a cycle (AUCcycle) was estimated. To describe tumour growth kinetics in all patients, multiple imputation (MI, 50 replicates) [5] of PTX PK was performed using a PTX PK model developed from the PK-guided dosing arm data [6]. TGI model-predicted relative change in tumour size at week 8 (RS8) were averaged from the MI PTX AUCcycle. Different baseline time-to-event (TTE) models (constant, Weibull, Gompertz and log-normal) were evaluated to characterise OS. The predictive value of RS8 on OS was evaluated in a full covariate model (FCM) including sex, tumour histology, smoking history, platinum drug type, prior treatments (yes/no), alternative treatments (yes/no) and baseline covariates (ECOG performance status, lactate dehydrogenase, tumour stage and tumour size). Datasets were prepared in R 3.6.0 and analysis performed in NONMEM 7.4.3, with assistance of PsN 4.6.0 and Pirana 2.9.6.

Results: Tumour growth was best described by a linear process, with a growth rate of 0.276 (90% CI: 0.171, 0.452) mm/week. The typical drug effect driven by PTX AUCcycle was associated with a 31.5% RS8 (with baseline tumour size=8.3 cm (median), median PTX AUCcycle and typical growth rate) and a 48% decline in drug effect of PTX AUCcycle after 2 cycles of treatment was estimated. The PTX AUCcycle estimated from observed PTX concentrations largely fell within the interquartile range of MI PTX AUCcycle, showing reliability of the MI. The Weibull hazard baseline TTE model showed best prediction of the observed survival data, from Kaplan-Meier-based visual predictive checks, compared to the constant hazard, Gompertz and Weibull models. The model predicted median OS was 11.0 months versus 11.2 months observed. Inclusion of RS8 in the established Weibull FCM was associated with a statistically significant improvement in model fit (ΔOFV=21.4): a 20% reduction in RS8 was associated with 3.20 months increase in median OS.

Conclusions: We successfully developed a TGI model establishing a quantitative relationship between PTX AUCcycle and tumour growth kinetics, and subsequently quantified the predictive value of RS8 on OS. Linear tumour growth and exponentially decaying PTX AUCcycle effect adequately characterised the observed tumour size data. Using the MI approach enabled joint TGI model evaluation of the entire dataset, including patients with no PTX PK data. RS8 had a statistically significant impact on OS, emphasising its prognostic value [2]. The established modelling platform enables quantification and comparison of individual patient’s survival profiles from different clinically-relevant PTX dosing regimens in advanced NSCLC apriori for decision making.

References:
[1] A. Sandler et al. N. England. J. Med 355: 2542-2550 (2006)
[2] Y. Wang et al. Clin. Pharmacol. Ther. 86: 167-174 (2009)
[3] M. Joerger et al. Ann. Oncol. 27: 1895 (2016)
[4] L. Claret. J. Clin. Oncol. 27: 4103-4108
[5] A.M. Johansson et al. AAPS J. 15: 1035-1042 (2013)
[6] A. Henrich. J. Pharmacol. Exp. Ther. 362: 347-358 (2017)

Reference: PAGE () Abstr 9388 [www.page-meeting.org/?abstract=9388]

Poster: Drug/Disease Modelling - Oncology