新药国际同步研发中类三臂非劣效试验设计研究与应用

发布时间：2018-06-09 14:00

  本文选题：药物国际同步研发 + 桥接　； 参考：《第四军医大学》2017年博士论文

【摘要】：种族敏感性(Ethnic Sensitivity)是药物研发中的重要问题之一,药物的代谢动力学和药效学特征在不同种族间往往具有不同的表现。鉴于这种种族差异,在原区域(Original Region)批准上市的药品,在新区域(New Region)申报上市时,需要阐明药品在种族间的差异,及其对药物安全性和有效性的影响。为了充分获得药品对新区域种族人群最直接的药动/药效学、疗效、安全性数据,药品申办方往往被监管机构要求在新区域重新进行I~IV期临床试验,这种重复试验不仅浪费医疗资源,更延迟了有效新药惠及患者的时间。对此,ICH E5提出桥接试验(Bridging Study)的概念:如果药物在不同种族人群具有相似性,即可将该药物其他种族临床试验数据外推至新区域的目标种族(Targeted Ethnic,TE),根据情况在新区域进行小规模附加试验,说明药物对该区域人群的安全性、有效性。然而,由于缺乏明确的种族“相似性”定义,这种序贯地国际研发模式在实践中存在诸多亟待解决的问题。随后,多区域临床试验(Multi-regional Clinical Trial,MRCT,又称为国际多中心临床试验)提出了在国际范围内由多个区域共同参与、同步试验的策略,以促进药物在全球范围内的开发与上市。为使MRCT中每个区域符合当地药品监管机构的要求,MRCT的设计和区域内样本含量估计异常复杂。为确保每个局部区域以足够的把握度确证药品对本区域种族人群的疗效,MRCT的样本含量往往被设计得非常庞大。因此,相对于一个大规模MRCT,在适量样本的MRCT框架外,根据情况同步或序贯地补充针对TE人群小规模的局部区域临床试验(Local Clinical Trial,LCT),变成了新药国际研发的最新趋势。而总结以上现状,新药国际研发项目中有三个亟待解决的焦点问题:(1)如何估计局部区域试验样本含量;(2)如何分析药物针对TE人群的疗效;(3)如何解释局部区域的试验结果。针对这些问题,我们提出一种类三臂非劣效设计,采用MRCT附加小规模LCT的桥接设计,以评价国际研发中局部区域的药物疗效。具体研究内容和结果如下:1类三臂非劣效设计框架的提出假设已有一个安慰剂对照的MRCT,在药物有效性已确证的基础上,目标区域将独立开展一个只招募TE人群的临床试验,除区域不同外,LCT的设计要素与MRCT完全一致。基于不同种族人群的生物共性(Biological Commonality of Humankind),我们假设在一定程度上LCT与MRCT具有相似的安慰剂效应,将两个安慰剂组类比为三臂非劣效设计中的安慰剂臂,对药物是否有效进行定性判断;同时,将LCT中的试验组类比为三臂非劣效设计中的试验臂;MRCT中的试验组类比为阳性对照臂(Active Reference Arm),对药物的有效性进行定量度量。在这个类三臂(Quasi-three-arm)结构中,假设LCT的药物疗效是非劣效于MRCT的,以MRCT试验组与安慰剂组疗效差的百分比作为非劣效界值,通过非劣效检验可以确证药物在LCT的疗效:(1)相对于安慰剂的有效性;(2)保留了MRCT疗效的一定百分比;(3)非劣效于MRCT的整体疗效。2不同资料类型下类三臂非劣效设计的桥接方案基于LCT与MRCT具有相似安慰剂效应的假设,本研究针对不同类型的试验终点(连续型、二分类和生存资料),基于安慰剂效应在LCT与MRCT间相等和不等两种情况,以LCT与MRCT组间疗效差之比的形式构建检验假设,提出桥接设计方案和样本量估计方法,以一次检验分别对药物疗效进行定性判断和定量度量;并通过模拟研究与实例应用探讨了各检验方法的可行性、适用性、稳健性,以及样本含量估计随参数变化的趋势。(1)连续型资料:近似正态假设下,分别提出安慰剂效应相等和不等的非劣效检验方法和样本量估计公式。采用Monte Carlo模拟实验,(1)在单侧检验水准α=0.025,目标把握度1-β=0.80的设定下,不同参数组合109个模拟情景的平均检验效能为0.7993±0.0068,平均I类错误为0.0250±0.0007。(2)样本含量估计的模拟研究发现,MRCT样本含量M_(RCTN)、MRCT组间疗效差与方差之比δ_M/σ、LCT与MRCT真实疗效差之比ρ、非劣效界值θ是影响样本含量的关键因素。其中,随着θ的增加,N_(LCT)呈明显的上升趋势;随着ρ的增大,N_(LCT)会逐步下降;当固定其他参数时,δ_M/σ每增大一倍,则N_(LCT)缩小到原来的1/4,而当δ_M/σ时,N_(LCT)的下降趋于缓慢,直至δ_M/σ时,N_(LCT)趋于稳定;同时,N_(LCT)也随着MRCTN的增大而减小,但伴随着/Mds的增大,变化幅度逐步减小。(3)基于错误的安慰剂效应假设,低估的样本含量引起了条件效能的降低,当N_(LCT)=500 1000 1500,,时,条件检验效能平均降低至69.04%,67.60%,66.02%;而相同条件下,I类错误并未出现明显膨胀,维持在名义检验水准0.025附近随机波动。(2)二分类资料:(1)二分类资料样本含量随参数变化的情况与正态资料类似,且并不会随M_(RCT)安慰剂组预期终点事件发生率MCp的变化产生明显改变。(2)大样本情况下,二分类资料的非劣效检验统计量近似正态分布,然而由于对统计量中方差σ~2(?)的估计依赖于ψ(π)的估计值(?),有学者提出用约束极大似然法/限制性极大似然估计(Restricted Maximum Likelihood Estimation,RMLE)估计方差更为准确。因此,本部分采用模拟试验,重点比较了基于极大似然估计(Maximum Likelihood Estimation,MLE)的Wald检验与RMLE检验下的条件检验效能和I类错误。结果发现156个模拟情景中,Wald检验的模拟结果准确性较差,且变异更大、稳定性欠佳。类三臂非劣效设计的二分类资料适宜采用RMLE方法,估计方差及检验统计量更加准确、稳健。(3)生存资料:服从指数分布的生存资料中,试验组与对照组的风险比(Hazard Ratio,HR)经过对数转换后呈正态分布;服从Weibull分布的生存资料,通过Cox比例风险模型,不同处理组间系数即为对数风险比,且近似正态。通过这种正态转换,本节对服从指数分布、Weibull分布的非劣效检验和样本含量估计进行了理论研究。本研究的主要创新点有:(1)提出一种新的MRCT+LCT桥接思路,由安慰剂、MRCT试验组、LCT试验组构造类似于三臂非劣效设计的框架,通过一次检验达到对药物疗效的定性判断和定量度量的目的;(2)探讨了不同数据资料类型下的统计分析方法和样本量估计,给出了影响样本量估计的参数设置,并推荐了不同条件下的样本量估计方案及适用范围;(3)编写了基于不同数据资料类型下类三臂非劣效设计的模拟研究程序,给出可以直接应用于临床实践的桥接方案。本文以新药国际研发中药物在目标区域的疗效评价为研究目的,提出类三臂非劣效设计的桥接策略及其在不同数据资料类型下的桥接方案和样本含量估计公式,重点探讨了样本量估计的参数设置及各种分布检验统计量的可行性、准确性、稳健性,以期为桥接试验设计提供参考依据,节约医疗资源,促进有效药物尽快上市。
[Abstract]:Ethnic Sensitivity is one of the important issues in drug development. Pharmacokinetic and pharmacodynamic characteristics are often different among different races. In view of the racial differences, drugs approved in the original region (Original Region) and the declaration of the new region (New Region) need to be clarified. The differences between races and their effects on drug safety and effectiveness. In order to obtain the most direct pharmacokinetic / pharmacodynamics, efficacy, safety data, and safety data for the new regional population, the drug applicant is often required to reconduct the I~IV clinical trials in the new area, which is not only a waste of medical resources, More delayed the time for effective new drugs to benefit the patient. In this case, ICH E5 proposed the concept of Bridging Study: if the drug is similar in different ethnic groups, other racial clinical trials of the drug can be extrapolated to the target race of the new region (Targeted Ethnic, TE), and in the new area, a small scale is added to the new area. The test shows the safety and effectiveness of the drug for the population in the region. However, there are many problems to be solved in the practice of the sequential international R & D model due to the lack of a clear definition of racial "similarity". Then, multi regional clinical trials (Multi-regional Clinical Trial, MRCT, also known as international multicenter clinical trials) In order to make every region of MRCT conform to the requirements of local drug regulators, the design of MRCT and the estimation of the sample content in the region are very complex to ensure that each local area is sure enough. The sample content of MRCT is often very large in assessing the effectiveness of drugs for ethnic groups in the region. Therefore, relative to a large-scale MRCT, a small scale local regional clinical trial (Local Clinical Trial, LCT) for the TE population (Local Clinical Trial, LCT) has been transformed into a new drug international research. There are three key issues in the new drug international R & D project: (1) how to estimate the content of local regional test samples; (2) how to analyze the efficacy of drugs for the TE population; (3) how to explain the experimental results in the local area. In view of these problems, we propose a non inferiority complex of three arms. The MRCT plus small scale LCT bridge design is used to evaluate the drug efficacy of the local area in international R & D. The specific research content and results are as follows: the 1 category of three arm non inferiority design framework has a placebo-controlled MRCT. On the basis of the validation of the drug effectiveness, the target area will be independently recruited to recruit only TE The clinical trials of the population, except for regional differences, are exactly the same as the LCT design elements. Based on the Biological Commonality of Humankind of different racial groups, we assume that LCT and MRCT have a similar placebo effect to a certain extent, and the two placebo groups are analogized to the placebo arm in the three arm non inferiority design. Qualitative judgment of the effectiveness of the drug was made; at the same time, the experimental group in the LCT group was analogous to the three arm non inferiority design arm; the test group in MRCT was analogous to the positive control arm (Active Reference Arm) to measure the effectiveness of the drug. In this class of three arm (Quasi-three-arm) structure, the efficacy of the LCT was assumed to be not inferior. MRCT, the percentage of the difference between the MRCT test group and the placebo group was a non inferiority value, and the efficacy of the drug in LCT was confirmed by the non inferiority test: (1) the efficacy of the placebo relative to the placebo; (2) a certain percentage of the efficacy of the MRCT was retained; (3) the non inferiority of the three arm non inferiority design under the different data types of the MRCT was not inferior to the overall efficacy of MRCT. The bridging scheme is based on the hypothesis that LCT and MRCT have similar placebo effect. This study aims at different types of test endpoints (continuous, two classification and survival data), based on the placebo effect equal and unequal between LCT and MRCT, and constructs a bridge design scheme based on the ratio of the difference of the curative effect between the LCT and the MRCT group. And the method of sample size estimation, the qualitative judgment and quantitative measurement of the drug effect are carried out by one test, and the feasibility, applicability, robustness, and the trend of the sample content estimation with the parameter change are discussed by the simulation study and the case application. (1) the continuous data, under the approximate normal assumption, the placebo The effect equality and unequal non inferiority test method and the sample size estimation formula. Using the Monte Carlo simulation experiment, (1) in the unilateral test level alpha =0.025, the target grasp degree 1- beta =0.80 is set, the average test efficiency of the 109 simulated scenarios with different parameters combination is 0.7993 + 0.0068, the average I type error is 0.0250 + 0.0007. (2) sample content estimation The simulation study found that the MRCT sample content is M_ (RCTN), the ratio of the curative effect and the variance of the MRCT group is delta _M/ sigma, LCT and the true effect of MRCT, and the non inferiority boundary value theta is the key factor affecting the sample content. Among them, N_ (LCT) increases with the increase of theta; N_ (LCT) decreases gradually with the increase of P, when other parameters are fixed. When delta _M/ is doubled, then N_ (LCT) is reduced to the original 1/4, and when N_ (LCT) decreases slowly when delta _M/ sigma, N_ (LCT) tends to stabilize until Delta _M/ sigma; meanwhile, N_ (LCT) decreases with the increase, but with the increase, the variation amplitude decreases gradually. (3) based on the false placebo-effect hypothesis, the undervalued sample contains The quantity caused the reduction of the condition efficiency. When N_ (LCT) =500 10001500, the condition test efficiency was reduced to 69.04%, 67.60%, 66.02%; and under the same condition, the I class error did not expand obviously, and maintained at the nominal test level 0.025 random fluctuation. (2) two classification data: (1) two classification data sample content varies with parameter. The normal data is similar and does not change significantly with the change of the incidence of the expected end point of the M_ (RCT) placebo group. (2) the non inferiority test statistics of the two classification data are approximately normal distribution under the large sample case. However, some scholars put forward the constraints due to the estimation of the difference of the Chinese difference ~2 (?) on the statistics. The maximum likelihood method / restrictive maximum likelihood estimation (Restricted Maximum Likelihood Estimation, RMLE) is more accurate in estimating the variance. Therefore, the simulation experiment is used in this part to compare the condition test effectiveness and the I class error based on the Wald test and RMLE test based on the maximum likelihood estimation (Maximum Likelihood Estimation, MLE). In the present 156 simulated scenarios, the accuracy of the simulation results of the Wald test is poor, and the variation is greater and the stability is not good. The two classification data of the three arm non inferiority design are suitable for the use of the RMLE method, and the estimation variance and the test statistics are more accurate and robust. (3) the survival data: the risk ratio of the experimental group and the control group in the subsistence data of the digital distribution (Hazard Ratio, HR) is a normal distribution after logarithmic conversion; subsistence data obeys Weibull distribution, and through Cox proportional hazard model, the coefficient between different processing groups is the logarithmic risk ratio, and it is approximately normal. Through this normal transformation, this section has a theory that obeys the exponential distribution, the non inferiority test and the sample content estimation of the Weibull distribution. The main innovations of this study were as follows: (1) a new MRCT+LCT bridging idea was proposed, which consisted of a placebo, MRCT test group, and the LCT test group structure similar to the three arm non inferiority design framework to achieve qualitative judgment and quantitative measurement of the efficacy of the drug by one test; (2) the statistical analysis under different data types was discussed. The method and sample quantity estimate, give the parameter setting of the estimation of sample size, and recommend the sample size estimation scheme under different conditions and the scope of application. (3) the simulation research program based on the three arm non inferiority design based on different data types is written, and the bridging scheme which can be directly applied to clinical practice is given. The therapeutic evaluation of drugs in the target area in international R & D is the purpose of research. The bridging strategy of three arm non inferiority design and its bridge scheme and sample content estimation formula under different data types are proposed. The parameters setting of the sample size estimation and the feasibility, accuracy and robustness of various distribution test statistics are discussed. In order to provide reference for bridge test design, save medical resources and promote effective drug listing as soon as possible.
【学位授予单位】：第四军医大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：R95
，

本文编号：1999966