PI-069 - EXPOSURE-RESPONSE MODELING STRATEGIES IN ONCOLOGY: IMPACT OF DOSE REDUCTIONS ON INTERPRETATION OF RESULTS.

M. Yue^1,2, A. Lombard³, C. Pitou⁴, E. Chigutsa²; ¹Purdue University, West Lafayette, IN, USA, ²Eli Lilly and Company, Indianapolis, IN, USA, ³Eli Lilly and Company, Neuilly-sur-Seine, France, ⁴Eli Lilly and Company, London, England, United Kingdom.

Background: Dose adjustments due to adverse events are common in oncology clinical trials. Exposure-response (E-R) analysis using Kaplan-Meier (K-M) curve stratified by summary exposure metrics is prone to immortal time bias (ITB) when there is dose adjustment (1). Longitudinal pharmacokinetic/pharmacodynamic (PK/PD) modeling can address ITB but could be complex and time-consuming. The objective of this study is to determine the extent of dose reduction which could necessitate longitudinal PK/PD modeling instead of straightforward K-M quartile analysis.

Methods: Virtual trials were simulated using a longitudinal PK - tumor size – progression-free survival model. K-M curves were plotted using the simulated output and stratified by quartiles of different exposure metrics. To assess the impact of dose reductions, performance of the full longitudinal model was compared to reduced models which used summary exposure metrics (e.g. area-under-the-curve) for accurately capturing the E-R relationship.

Results: K-M quartile plots showed biased E-R relationships when trials had 80% of patients underwent ≥25% dose reduction, 40% of patients underwent ≥50% dose reduction, 20% of patients underwent ≥75% dose reduction, or 80% of patients had drug holidays for 28 days or longer. Dynamic concentrations from the longitudinal model provided better estimates of drug effect compared to those with static exposure metrics, particularly when there was high extent of dose reduction.

Conclusion: In cases of high percentage of patients with dose reduction, high magnitude of dose reduction, long drug holidays, or drug holidays resumed at a lower dose, longitudinal PK/PD modeling should be used to address ITB and improve estimation of E-R relationships and aid in decision-making.

1. Khandelwal A, Grisic AM, French J, Venkatakrishnan K. Pharmacometrics Golems: Exposure-Response Models in Oncology [published online ahead of print, 2022 Mar 14]. Clin. Pharmacol. Ther. 2022;10.1002/cpt.2564. doi:10.1002/cpt.2564