II-046 Zhe Huang

Model-integrated bioequivalence method for highly variable drugs with long half-life: a simulation study comparing complete washout and incomplete washout designs

Zhe Huang (1), Xiaomei Chen (1), Mark Donnelly (2), Lanyan Fang (2), Liang Zhao (2), Mats O. Karlsson (1) and Andrew C. Hooker (1)

(1) Department of Pharmacy, Uppsala University, Uppsala, Sweden (2) Division of Quantitative Methods and Modelling (DQMM), Office of Research and Standards (ORS), Office of Generic Drugs (OGD), Center for Drug Evaluation and Research (CDER), U.S. Food and Drug Administration (FDA), Silver Spring, MD, USA.

Objectives: 

An in vivo bioequivalence (BE) study with pharmacokinetic (PK) endpoints may be recommended to support the conclusion of therapeutic equivalence between generic and reference drug products1. The reference-scaled average BE approach (RSABE)2 may be used to evaluate highly variable drugs (HVD) following FDA guidance3, which recommends a replicate crossover design (partial or fully replicate) for comparison in absorption rate and extent by measuring Cmax and AUC, respectively, with non-compartmental analysis (NCA). However, BE studies using a replicate crossover design can be challenging for HVD with long half-lives, since a sufficient washout period (≥ 5 half-lives) between treatment periods is recommended for NCA analysis3.  One possible solution is to use an incomplete washout design to achieve shorter study duration, and to analyze the results with a model-integrated approach. Our group previously developed a model-integrated BE method4,5 that shows controlled type I error and higher power compared to NCA-based BE analysis. In this work, we expanded our method to analyze BE data for HVD and performed a simulation study to evaluate its performance for HVD with a long half-life in the absence of a complete washout period.    

Methods: 

The modified model-integrated BE method was developed to perform RSABE analysis for HVD. Specifically, the method includes the following steps:

  1. Model parameter estimation: A pre-defined PK model, including parameters accounting for the product differences in absorption, is employed to fit a BE dataset. An estimate of within-subject standard deviation of the reference product (S_WR) is obtained based on the estimated PK parameters.
  2. Parameter uncertainty estimation: The parameter uncertainty is assessed based on sampling importance resampling (SIR)6.
  3. Population simulation: A population simulation based on parameter estimation and uncertainty is performed to obtain distributions of the geometric mean ratios of Cmax and AUC.
  4. BE conclusion: The RSABE method is employed to adjust the BE acceptance criteria based on the estimated S_WR.

A simulation study (N=500) was performed to evaluate the modified methods adapted for RSABE. The model for generating the simulated datasets was a one-compartmental PK model with a typical half-life of 55.5 hr. Two sets of S_WR magnitudes were investigated on the relative bioavailability of the two tested compounds: 30% (CV30) and 48% (CV50). Fully replicate crossover studies with complete washout (≥5 half-life) and incomplete washout (a dosing interval of 72hr) were tested. A rich sampling point design (18 points) was used for both scenarios. Type I error and power were calculated as the percentage of cases concluding BE for the scenarios of bioinequivalence and bioequivalence, respectively. Standard NCA-based RSABE analysis methods were also evaluated for the BE data with complete washout.

Results: 

For the CV30 setting and the complete washout design, both the proposed model-integrated method for RSABE and the NCA-based RSABE controlled type I error at 5.6% (95%CI: [3.6%, 7.6%]) and 5.2% (95%CI: [3.4%, 7.2%]), respectively when the treatment effect on bioavailability was at the HVD expanded limit of 1.3. The model-integrated method exhibited slightly higher statistical power of 81% (95%CI: [77.6%, 84.4%]) than the NCA-based method (76.2% with 95%CI: [72.4%, 79.8%]). The model-integrated RSABE approach showed comparable performance for the incomplete washout design, with a controlled type I error of 5.2% (95%CI: [3.4%, 7.2%]) and a power of 84.4% (95%CI: [81.2%, 87.6%]). The NCA-based RSABE method could not be used for the evaluation of the incomplete washout design. Similar conclusions were found for the CV50 setting. The incomplete washout design had an overall study duration of 12 days in the simulation scenario, achieving a 68% reduction in study duration compared to the complete washout design (duration >= 37.5 days).

Conclusions: 

The use of a model-integrated BE method incorporating RSABE allows for both complete and incomplete washout designs with comparable power in the investigated scenarios. In situations with highly variable drugs with long half-lives, this can result in a substantial reduction in study duration. The model-integrated BE method for HVD was flexible enough to analyze data from both novel BE designs and the standard design for HVD with comparable controlled type I errors and power.  

In conclusion, the model-integrated method with incomplete washout designs serves as a promising strategy for the BE evaluation of drugs with high variability and long half-life by reducing the overall study duration substantially.

References:

  1. FDA Guidance for Industry: Statistical Approaches to Establishing Bioequivalence. (2022).
  2. Davit, B. M. et al. Implementation of a reference-scaled average bioequivalence approach for highly variable generic drug products by the US Food and Drug Administration. AAPS J 14, 915–924 (2012).
  3. FDA Draft guidance: Bioequivalence Studies With Pharmacokinetic Endpoints for Drugs Submitted Under an ANDA Guidance for Industry. (2021).
  4. Chen, X. et al. Development and comparison of model-integrated evidence approaches for bioequivalence studies with pharmacokinetic endpoints. Preprint at https://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-517824 (2023).
  5. Bjugård Nyberg, H. et al. Evaluation of model-integrated evidence approaches for pharmacokinetic bioequivalence studies using model averaging methods. Preprint at https://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-517823 (2023).
  6. Dosne, A.-G., Bergstrand, M. & Karlsson, M. O. An automated sampling importance resampling procedure for estimating parameter uncertainty. J Pharmacokinet Pharmacodyn 44, 509–520 (2017).

Reference: PAGE 32 (2024) Abstr 10872 [www.page-meeting.org/?abstract=10872]

Poster: Methodology - New Modelling Approaches

PDF poster / presentation (click to open)