Simulation-based evaluation of personalized dosing approaches for anti-FGFR/KLB bispecific antibody fazpilodemab

Abstract

Personalized dosing approaches play important roles in clinical practices to improve benefit: risk profiles. Whereas this is also important for drug development, especially in the context of drugs with narrow therapeutic windows, such approaches have not been fully evaluated during clinical development. Fazpilodemab (BFKB8488A) is an agonistic bispecific antibody which was being developed for the treatment of nonalcoholic steatohepatitis. The objective of this study was to characterize the exposure-response relationships of fazpilodemab with the purpose of guiding dose selection for a phase II study, as well as to evaluate various personalized dosing strategies to optimize the treatment benefit. Fazpilodemab exhibited clear exposure-response relationships for a pharmacodynamic (PD) biomarker and gastrointestinal adverse events (GIAEs), such as nausea and vomiting. Static exposure-response analysis, as well as longitudinal adverse event (AE) analysis using discrete-time Markov model, were performed to characterize the observations. Clinical trial simulations were performed based on the developed exposure-response models to evaluate probability of achieving target PD response and the frequency of GIAEs to inform phase II dose selection. Dynamic simulation of personalized dosing strategies demonstrated that the AE-based personalized dosing is the most effective approach for optimizing the benefit-risk profiles. The approach presented here can be a useful framework for quantifying the benefit of personalized dosing for drugs with narrow therapeutic windows.

FIGURE 1

Schematic representation of the simulation workflow for the evaluation of the treatment personalization approaches. See “Dynamic clinical trial simulation” in the section 2 and Text S1 for details. AE, adverse event; PK, pharmacokinetic.

FIGURE 2

(a) Simulated relationships between dose and liver fat reduction or occurrence of persistent GIAE. Lines and colored shaded areas are median and 95% prediction intervals. Vertical lines and gray shaded areas delineate the expected therapeutic window on a population level with fixed‐dosing approaches. (b) Simulated probability of mean liver fat reduction exceeding target threshold at different dose levels. At the dose levels of 100 mg q2w, for example, the probability of exceeding mean liver fat reduction of 30% after 12 weeks of treatment in the phase Ib NAFLD population was calculated as 80%. In both panels, prediction intervals include between‐subject variability in the pharmacokinetics of BFKB8488A. GIAE, gastrointestinal adverse event; NAFLD, non‐alcoholic fatty liver disease.

FIGURE 3

Comparison of various treatment personalization approaches on the mean liver fat reduction and cumulative percentages of treatment discontinuation at the end of the simulated period. Each symbol represents these metrics from the simulation of the corresponding treatment approaches. Gray symbols represent fixed dosing approaches at 75, 100, and 130 mg q2w. Colored symbols represent PK‐, AE‐based, or the combination approaches. See text for the detailed description of the personalization approaches. AE, adverse event; PK, pharmacokinetic.