Determination of sample size and similarity criteria using regional error rate in a multi-regional clinical trial

Abstract

There is growing interest in multi-regional clinical trials (MRCTs), which are those conducted in many regions using the common protocol. MRCTs simplify the approval and registration processes for treatment in all regions, and provide an opportunity to reduce costs and time consumption by not repeating similar clinical trials. Accomplishing the goal of MRCTs depends on consistency in the effect size of individual regions following verification of overall treatment effects. However, there are currently no criteria to assess the similarity of treatment effects across regions, or standards for calculating the required number of clinical trial subjects in the region of interest to demonstrate such similarity. In 2007, Japanese MHLW provided guideline on similarity criteria and the required number of clinical trial subjects in Japan for MRCTs. However, this guideline does not offer a statistical perspective. Based on the MHLW guideline, Ko et al. (2010) proposed a method based on the concept of the assurance probability to calculate the sample size in the region of interest. But this method does not focus on the second purpose of MRCTs, which concerns the similarity of treatment effects across regions. This thesis introduces a method standardized by effect size, which was originally suggested by Kang et al. (2016), as a statistical hypothesis testing procedure to address the second purpose of MRCTs. This thesis also discusses approaches using the regional type II error rate to calculate critical values for a hypothesis testing on the similarity, as well as the required number of clinical trial subjects in the region of interest through the suggested method. Using calculated regional type II error rates according to similarity criteria, it was shown that it is easier to control the regional type II error rate if the difference in effect sizes between the region of interest and other regions excluding the region of interest is great or the critical value for the similarity increases. If a pre-determined regional type II error rate is satisfied and parameters such as effect size and the required number of patients in the region of interest are the same respectively, the critical value for the similarity criterion D1≥ρD is considered more conservative compared to the critical value for the similarity criterion D1≥ρD1c. The proportion of the patients in the region of interest did not monotonically decrease with the regional type II error rate. Such an undesirable property needs to be improved by developing new similarity criteria. Furthermore, the method used to control the regional type I error rate is an expanded form of the method proposed by Ko et al. (2010), in the sense that it is applicable in cases where the effect sizes across regions are heterogeneous. Recently, there has been an increasing trend of globalization in developing new drugs using MRCTs in Korea. As the frequency of conducting MRCTs increases, it has become important to determine critical value to evaluate the similarity in treatment effects between ethnic groups and to calculate the sample size in the region of interest, such as Korea. In such clinical development environments, the regional type II error method proposed in this thesis will be useful for MRCTs, and should be further studied for continued improvement.

다지역 임상시험은 다양한 지역에서 동일한 계획서를 가지고 진행함으로써 모든 지역에서 의약품 허가 및 등록을 용이하기 할 뿐만 아니라 각 지역에서 유사한 임상시험을 중복하여 수행하지 않음으로써 시간과 비용을 줄일 수 있는 기회를 제공하기 때문에 그 관심도가 지속적으로 증가하고 있다. 다지역 임상시험의 목적을 달성할 수 있는 능력은 전체 치료 효과를 입증한 이후 참역한 개별 지역에서 효과가 얼마나 유사한지에 달려 있다. 하지만 치료 효과의 지역 간 유사성을 보이기 위한 유사성 기준에 대한 정의와 이러한 유사성을 보이기 위한 특정 관심 지역에 필요한 임상시험 대상자 산출에 대한 기준에 대한 정의가 이뤄지지 않고 있다. 2007년 일본 후생성은 다지역 임상시험에서 유사성 기준과 이러한 기준을 근거로 일본에 필요한 임상시험 대상자수 산출에 대해 가이드라인을 제공하고 있으나, 통계적인 관점에서는 제안하고 있지는 않다. Ko et al.(2010)은 후생성의 가이드라인을 기반으로 보장 확률개념을 도입하여 특정 관심 지역의 임상시험 대상자수 결정에 대해 제안했다. 하지만 다지역 임상시험의 두 번째 목적인 지역 간 치료 효과의 유사성 대한 가설에 중점을 두고 평가하지 않았다는 점에서 적절하지 않다. 본 논문에서는 다지역 임상시험의 두 번째 목적을 통계적 가...