Yubo Xiao1, Xiao Zhang1, Junyi Wu1, Xuan Zhou1, Scott Marshall2
1 GSK, Shanghai, China; 2 GSK, Stevenage, Hertfordshire, UK
Introduction: Global simultaneous drug development has become a common practice in multinational pharmaceutical companies to reduce the time lag to launch and improve patient access to novel treatments. Learning how intrinsic and extrinsic factors impact drug intervention throughout the exploratory phases of drug development, including Phase 2 trials, increases the confidence in future confirmatory multi-regional clinical trials.[1] Specifically, in some East Asian countries, an assessment of ethnic sensitivity between Asians and non-Asians is required. In this regard, early characterization of the D-E-R relationship across a more ethnically diverse population would be advantageous. Balancing this goal versus the inherent risks associated with early drug development requires careful consideration of how to efficiently allocate sample size across regions to detect clinically relevant differences prior to confirmatory trials.
To illustrate some of the considerations in ongoing programs we have utilised an established model based on a Phase 1 study data from an anti-psoriatic drug (drug A) to design a hypothetical multi-regional dose-ranging study.
Objectives: To evaluate regional sample size allocation in a planned Phase 2 dose-ranging trial using recognized statistical and model-informed approaches, assuming borderline clinically relevant efficacy differences in the region of interest (Region X).
Methods: In this planned trial, 175 patients (in alignment with similar phase 2 trial sizes for this disease) were to be randomly assigned to four dose groups of drug A or placebo based on standard dose-ranging considerations. The primary endpoint was the percentage improvement from baseline in the Psoriasis Area and Severity Index (PASI) score at week 12. Longitudinal pharmacokinetics (PK) and clinical response data were simulated using an established D-E-R model for Drug A in line with the objective. [2] A defined covariate effect of region X on IC50 was included in the simulations to allow the patients from region X following the top dose to have a minimal clinically meaningful improvement. [3] Simulations were performed with different proportions of sample size from Region X to allow for assessing the impact of sample size allocation on the power to detect clinically relevant differences.
A pairwise statistical approach and a model-based approach using the Monte-Carlo Mapped Power method [4] were employed for sample size assessments using the simulated data. When using the pairwise statistical approach, 20000 trials were simulated for each specific proportion of sample size from Region X. Every of the simulated trials was compared in Region X and the rest of the regions for the week 12 efficacy with t-test in the top-dose group, and power was calculated as the percentage of trials with p-value < 0.05. When using the model-based approach, one large dataset (N=21000) was simulated for each specific proportion of sample size from Region X. Difference in individual objective function values (ΔiOFV) was calculated after the large dataset was re-estimated with a full model and a reduced model. The ΔiOFV was sampled according to the trial sample size (N=175) and summed (∑ΔiOFVs). The percentage with ∑ΔiOFVs greater than the significance criterion of 3.84 was taken as the power to identify the difference between regions. With both methods, the power was estimated for different proportions of sample size from Region X.
Results: Based on the planned trial design, using the pairwise approach, the maximum power (when Region X takes 50% of the total sample size) to detect statistically significant treatment differences between Region X and the rest of the regions was below 40%. In contrast, with the model-based approach, 48 subjects from Region X, taking 27% of the total sample size, provided >80% power to detect the regional difference in the treatment effect.
Conclusions: This case study demonstrates that the recognized statistical approach based on the end-of-trial observations from one dose group is inefficient in detecting regional heterogeneity. In contrast, the use of a D-E-R model-based approach that utilizes data from across the dose groups and duration of the trial increases trial efficiency to detect and importantly characterize potential regional differences.
References:
[1] ICH E17. General Principles for Planning and Design of Multi-Regional Clinical Trials.
[2] Salinger DH, Endres CJ, Martin DA, Gibbs MA. A semi-mechanistic model to characterize the pharmacokinetics and pharmacodynamics of brodalumab in healthy volunteers and subjects with psoriasis in a first-in-human single ascending dose study. Clin Pharmacol Drug Dev. 2014 Jul;3(4):276-83. doi: 10.1002/cpdd.103.
[3] Carlin CS, Feldman SR, Krueger JG, Menter A, Krueger GG. A 50% reduction in the Psoriasis Area and Severity Index (PASI 50) is a clinically significant endpoint in the assessment of psoriasis. J Am Acad Dermatol. 2004 Jun;50(6):859-66. doi: 10.1016/j.jaad.2003.09.014.
[4] Vong C, Bergstrand M, Nyberg J, Karlsson MO. Rapid sample size calculations for a defined likelihood ratio test-based power in mixed-effects models. AAPS J. 2012 Jun;14(2):176-86. doi: 10.1208/s12248-012-9327-8.
Reference: PAGE 32 (2024) Abstr 11182 [www.page-meeting.org/?abstract=11182]
Poster: Methodology - Study Design