Bojan Lalovic (1), Oneeb Majid (2), Ishani Savant Landry (1), Larisa Reyderman (1), Margaret Moline (1), Jim Ferry (1), Ziad Hussein (2)
(1) Eisai Inc., Woodcliff Lake, NJ, USA, (2) Eisai Ltd., Hatfield, UK
Objectives: To describe lemborexant PK and quantify the effects of intrinsic and extrinsic factors in healthy adults, elderly, and in subjects with insomnia disorder, based on six extensively sampled Phase 1, one Phase 2 and two Phase 3 sparsely sampled studies.
Methods: The 12230-observation dataset included 1892 subjects who were 18 to 88 years old weighing 37 to 168 kg and predominantly female (66%). The dataset included lemborexant doses up to 100 mg, administered QD in the morning or at bedtime. Inclusion of known absorption-related predictors based on Phase 1 data, preceded forward addition (p<0.01) and backward elimination (p<0.001) of covariates to constitute the final lemborexant population model. The pre-defined factors included were body weight, age, sex, BMI, race, food, formulation, creatinine clearance, laboratory parameters, concomitant medications (e.g. gastric pH modifiers) and disease characteristics (primary/secondary insomnia). NONMEM and R were used to illustrate clinical relevance covariates under several simulation scenarios and derive PK measures. Visual predictive checks (VPC) and nonparametric bootstrap were used for model qualification and validation. NONMEM SAEM/IMP estimation under MPI parallelization enabled the analysis of this large pooled PK dataset, not readily feasible using the conventional estimation approach (FOCEI).
Results: Informed by extensively sampled Phase 1-2 data, initial lemborexant population pharmacokinetics were described using a linear 3-compartment model with mixed zero (D1) and first order (Ka), lag-time absorption process. The model included a combined (additive/proportional) residual error with distinct parameters for periods before and after 3 hrs post-dosing. This base model also incorporated absorption-related covariate effects: bedtime dosing on D1, the effect of food on Ka and relative bioavailability (F1), and the effect of formulation on D1 and Ka. Due to the absence of PK sampling in the first 9 hours following bedtime dosing for the Phase 2 and two Phase 3 sparsely sampled studies, the inclusion of covariates in the final model was constrained to the clearance parameter (CL). High eta (η) shrinkage (>60%) was noted for all between-subject variability parameters except CL. Final model lemborexant parameters were oral clearance (CL/F) of 22.7 L/hr, steady-state volume of distribution (Vss/F) of 1070 L and inter-compartmental clearances (Q3 and Q4) of 32.1 and 31.0 L/hr, respectively. For the tablet, D1 was 0.118 hrs with bedtime dosing increasing D1 by 2.33-fold; Ka was estimated to be 0.595 (1/hr). The effect of food decreased Ka by 30% while increasing F1 by 21%. Lemborexant CL/F decreased with increasing BMI (exponent = – 0.428), increasing ALP levels (exponent = – 0.118) and was 26% lower in the elderly (≥ 65 years).
Conclusions: Simulations assuming an elderly, high BMI/ALP cohort were predicted not to exceed the observed range of exposures across studies, supporting no specific lemborexant dose adjustment.
Reference: PAGE 28 (2019) Abstr 8829 [www.page-meeting.org/?abstract=8829]
Poster: Drug/Disease Modelling - CNS