Optimal Dosing throughout and beyond Clinical Development

抄録

Background: Despite significant advances in Modeling and Simulation, dosing decisions during development and dosing instructions approved in the product label are not necessarily optimal. An adaptable approach (platform) focusing on formal optimization of dosing regimens on predefined efficacy and safety criteria, which accounts for all relevant information available at each stage of development through to approval, may help. The use case is based on KAE609, an antimalarial with relevant information in the public domain.

Methods: 1. Based on physicochemical and in vitro information (Charman, 2020) and concentration time courses in Asian patients (White, 2014), a calibrated PBPK model of KAE609 was built in a virtual population resembling the PoC population. 2. Using this model and the built-in human database of PK-Sim, a virtual population of 3000 Japanese (0-81 y/o) was generated, their concentration time courses exported, and a compartmental model fitted to them (NLME). 3. The individual PK parameters were supplemented with simulated PD parameters derived from individual parasitemia over time courses by White 2014 (supplement), yielding PKPD parameter vectors in each individual of the virtual population. 4. The efficacy criterion to be met was selected as maximal parasite reduction ratio (PRRmax) equal to the initial individual total parasite load, taking into account weight-based blood volume. 5. For each virtual patient, the individual dose achieving the efficacy criterion (=optimal dose) given different fractionation strategies was estimated, by fixing the dose (AMT) to 1 and estimating F (0<F<inf). 6. Target achievement was checked (DV vs. IPRED). 7. For the favored dosing regimen, discretization and a ML-based covariate search (MARS, R-package earth) was performed, the dosing vector upscaled for 95% target achievement of the covariate adjusted dosing regimen. 8. The performance of the final "real world" regimen was tested for efficacy (percent target achievement) and safety (distribution of AUCs and peak exposures).

Results: Given the chosen plausible model structure and parameterization, for identical efficacy, dose fractionation from single dose to q24h for 3 days results in a median reduction of the cumulative Dose (Cmax) by 25% (50%). Age and weight were covariates, the shape of the relationship automatically determined by MARS. As to be expected, achieving the efficacy criterion in 95% of the population with covariate-based dosing profoundly increases exposure compared to the ideal (unrealistic) regimen based on individual parameter vectors.

Discussion/Conclusions: Theoretically, a defensible dosing regimen across the entire population can be obtained as early as PoC information is available, potentially prior to first in man (input from in vitro/animal studies). Early input from domain specialists is mandatory. The highly modular and adaptable approach is implemented into sequential R-scripts (auditability).

Literature: Charman SA et al., Malar J. 2020 Jan 2;19(1):1; White NJ et al., N Engl J Med. 2014 Jul 31;371(5):403-10