Maxwell Tawanda Chirehwa (1), Tarjinder Sahota (1), Stefano Zamuner (1), Núria Buil-Bruna (1)
CPMS, Precision Medicine, GSK, Stevenage, UK
Introduction: Getting the dose right in Ph3 is an important requisite for the successful registration of new medicines highlighting the need for high quality fit-for-purpose Ph2 programs. Traditional methods for dose selection in Ph2 studies, such as pairwise comparison and empirical regression methods (e.g., direct D-R models) face limitations in accounting for longitudinal changes in PK and PD, comparing multiple regimens, and incorporating prior data [1].
Full end-to-end pharmacometric Dose-Exposure-Target Engagement-Response (D-E-TE-R) based analyses, as a form of longitudinal D-E-R analysis, has the potential to overcome these limitations by modelling the impact of dose and dose frequency together on longitudinal clinical response. The feasibility of such an approach however is highly dependent on design choices made in Ph2b. A robust design methodology is needed to enable assessment of the feasibility DETER methods on prospective Ph2b designs to ensure a fit-for-purpose protocol and DETER-based analysis plans for accurate selection of Ph3 dose regimens. We present a viable methodological approach applied to a hypothetical monoclonal antibody targeting a soluble target elevated in an immune-inflammation indication.
Methods: The proposed methodology builds on the simulation re-estimation approach of Kowalski [2]. It requires defining a prior DETER model, and study design variables (such as number of arms and patients per arm, dose regimens, randomization ratio). We assume informative prior distributions for each parameter in the D-E-TE model based on previous studies and weakly informative priors for the TE-R part of the model reflecting the frequent absence of earlier studies evaluating the clinical response.
The simulation re-estimation exercise involves defining the number of designs (j), and trials (k) to be generated from parameter prior distribution. For each design (j), the algorithm is as follows:
- Generate k random draws of model parameters from the parameter prior distribution.
- Simulate subject-level PK, TE, and R data for each k in step 1 based on the Ph2b design variables.
- Estimate pre-specified DETER model parameters.
- Evaluate model performance using model convergence, condition number, and parameter, accuracy, and precision.
- Evaluate endpoint success by:
- For each k, simulate m expected response for each dosing regimen using the estimated parameters and uncertainty from step 3, without variability.
- For each k, simulate the true response using parameters generated in step 1 without variability.
- Evaluate the bias (mean error), accuracy (mean absolute error), precision (half confidence interval (CI) width).
- Assurance: Assess the joint probability of attaining the required target and minimum value (TV and MV) as prob(quantile(R, q1) ≤ TV & quantile(R, q2) ≤ MV).
Results: To illustrate how the approach can be used, we have constructed possible designs of a Ph2b study with placebo-corrected percent change from baseline (PCPCFB) in response as the endpoint. The prior DETER model was defined based on PK from Davda [3], PD from Gibiansky [4], and the response model included a continuous endpoint characterized by a placebo sub-model and a drug effect model with parameters. A power model described the TE-R part of the model and the reduction in the response was driven by level of the unbound soluble target (unbT^gamma). The prior distributions of the model were used to generate k=300 sets of parameters. Simulation and re-estimation were performed in NONMEM v7.5, and post-processing was performed in R v4.2.
The primary study design consists of placebo and 4 active doses (5, 20, 60 and 120 mg), with 60 subjects/arm. Alternative designs had reduced number of active arms or patients/arm. For the primary design, convergence was achieved, and the covariance step was successful for all 300 models. All estimated parameters had high accuracy except the gamma parameter (%MAE = 40). Relative MAE for PCPCFB was below 23% for 20 mg dose and above. The maximum half CI for PCPCFB was 45% or lower. The probability of success for the 120 mg was 83%. The analysis shows that the primary design enables robust characterization of the DER and provides high confidence to select a Ph3 regimen.
Conclusion: By designing adequate Ph2b studies for DETER modelling enable robust characterization of granular D-R relationships, the proposed methodological approach ensures the accurate and precise selection of Ph3 dose regimens.
References:
- Musuamba FT, Manolis E, Holford N, Cheung SYA, Friberg LE, Ogungbenro K, et al. Advanced Methods for Dose and Regimen Finding During Drug Development: Summary of the EMA/EFPIA Workshop on Dose Finding (London 4–5 December 2014). CPT Pharmacometrics Syst Pharmacol. 2024;6(7):418–29
- Kowalski KG. Integration of Pharmacometric and Statistical Analyses Using Clinical Trial Simulations to Enhance Quantitative Decision Making in Clinical Drug Development. 2019;11(1):85–103.
- Davda JP, Dodds MG, Gibbs MA, Wisdom W, Gibbs JP. A model-based meta-analysis of monoclonal antibody pharmacokinetics to guide optimal first-in-human study design. MAbs. 2014;6(4):1094
- Gibiansky L, Gibiansky E. Modeling approaches to characterize target-mediated pharmacokinetics and pharmacodynamics for therapeutic proteins. In: Quantitative Pharmacology and Individualized Therapy Strategies in Development of Therapeutic Proteins for Immune-Mediated Inflammatory Diseases. John Wiley & Sons, Ltd; 2019. p. 149–72.
Reference: PAGE 32 (2024) Abstr 11028 [www.page-meeting.org/?abstract=11028]
Poster: Drug/Disease Modelling - Other Topics