临床试验随机化方法的优选
|
摘要
目的近十几年来涌现出一些新的随机化方法,能有效提升随机化的性能,但其在临床试验中真正应用的情形并不多见,仍然以传统的区组随机化设计占据主导,存在着严重的潜在选择性偏倚风险。本研究将通过探讨多种随机化方法的统计性能,为临床试验随机化方法的选择提供实用性指引。方法选择简单随机化设计(SR)方法,以及目前常见的五种限制性随机化方法,包括固定区组设计(PBD)、变化区组设计(VBD)、大棒设计(BSD)、带不平衡限制的偏币设计(BCDWIT)、区组瓮设计(BUD),借助随机模拟方法,在不同的最大容许不平衡性(MTI)设定下,分别从随机性和均衡性两个方面对各方法的统计性能进行整体评价。本研究涉及的随机化方法优选仅在限制性随机化方法中进行,并将各方法的随机化性能指标作为评价优选的依据。结果 PBD和VBD方法的固定分配概率、猜对分配概率最高,相比之下,BSD、BUD、BCDWIT方法具有更好的随机化性能。具体看, BUD有着最小的固定分配概率,BCDWIT其次;BSD有着最小的猜对分配概率,BUD其次。结论为了减少试验的选择性偏倚,区组随机化设计不应再作为开放性临床试验随机化的推荐方法,而应考虑在其他几种限制性随机化方法中进行选择。在随机化方法选定后,还可通过大量模拟筛选出最佳的随机分配序列。
Objective Some new randomization methods have emerged in the past decade,and these methods can effectively improve the randomization performance,but the actual application in clinical trials is rare.The traditional block design still dominates,but block design is easy to cause serious and potential risk of selective bias.This study will provide practical guidance for the selection of randomization methods for clinical trials by exploring the statistical performance of multiple randomization methods.Methods This paper chooses the simple randomization(SR),and several common restrictive randomization methods,including permuted block design(PBD),variable block design(VBD),big stick design(BSD),biased coin design with imbalance tolerance(BCDWIT),and block urn design(BUD),under maximal tolerated imbalance(MTI) settings,we comprehensively evaluated the statistical performance of each method from two aspects of randomness and balance with random simulation method.This study is to optimize randomization method in the restrictive randomization method with the randomization performance index of each method as the basis for evaluation.Results The deterministic assignment probability and the correct guessing probability of PBD and VBD methods are the highest.In contrast,BSD,BUD and BCDWIT methods have better randomization performance.Specifically,BUD has the smallest deterministic assignment probability,BCDWIT second; BSD has the smallest correct guessing probability,BUD second.Conclusion In order to reduce the selective bias,block design should not be used as a recommended method for randomization of open clinical trials,consideration should be given to selecting among several other restrictive randomization methods.After the randomization method is selected,the best random allocation sequence can be screened by a large number of simulations.
Objective Some new randomization methods have emerged in the past decade,and these methods can effectively improve the randomization performance,but the actual application in clinical trials is rare.The traditional block design still dominates,but block design is easy to cause serious and potential risk of selective bias.This study will provide practical guidance for the selection of randomization methods for clinical trials by exploring the statistical performance of multiple randomization methods.Methods This paper chooses the simple randomization(SR),and several common restrictive randomization methods,including permuted block design(PBD),variable block design(VBD),big stick design(BSD),biased coin design with imbalance tolerance(BCDWIT),and block urn design(BUD),under maximal tolerated imbalance(MTI) settings,we comprehensively evaluated the statistical performance of each method from two aspects of randomness and balance with random simulation method.This study is to optimize randomization method in the restrictive randomization method with the randomization performance index of each method as the basis for evaluation.Results The deterministic assignment probability and the correct guessing probability of PBD and VBD methods are the highest.In contrast,BSD,BUD and BCDWIT methods have better randomization performance.Specifically,BUD has the smallest deterministic assignment probability,BCDWIT second; BSD has the smallest correct guessing probability,BUD second.Conclusion In order to reduce the selective bias,block design should not be used as a recommended method for randomization of open clinical trials,consideration should be given to selecting among several other restrictive randomization methods.After the randomization method is selected,the best random allocation sequence can be screened by a large number of simulations.
引文
[1] Berger VW.Do not use blocked randomization.Headache,2006,46(2):343.
[2] BergerVW.Misguided precedent is not a reason to use permuted blocks.Headache,2006,46(2):1210-1211.
[3] Berger VW,Bejleri K,Agnor R.Comparing MTI randomization procedures to blocked randomization.Statistics in Medicine,2016,35(5):685-694.
[4] Berger VW,Agnor R.Delayed unmasking and selection bias.Statistics in Medicine,2016,35:2111-2112.
[5] Soares JF,Wu CF.Some restricted randomization rules in sequential designs.Communications in Statistics-Theory and Methods,1983,12:2017-2034.
[6] Chen YP.Biased coin design with imbalance tolerance.Communications in Statistics-Stochastic Models,1999,15:953-975.
[7] Zhao W,WengY.Block urn design—a new randomization algorithm for sequential trials with two or more treatments and balanced or unbalanced allocation.Contemporary Clinical Trials,2011,32 (1):953-961.
[8] Rosenberger WF,Lachin JM.Randomization in Clinical Trials:Theory and Practice,Second Edition.New Jersey:John Wiley & Sons,Inc,2016:38-40,152-155.
[9] Blackwell D,Hodges JL.Design for the control of selection bias.Annals of Mathematical Statistics,1957,28:449-460.
[10] Berger VW.Randomization,Masking andjAllocation Concealment.Boca Ration:Taylor & Francis Group,LLC.,2018:129-130.
[11] 刘玉秀.没有分配隐蔽则没有随机化.医学研究生学报,2016,29(6):561-565.
[2] BergerVW.Misguided precedent is not a reason to use permuted blocks.Headache,2006,46(2):1210-1211.
[3] Berger VW,Bejleri K,Agnor R.Comparing MTI randomization procedures to blocked randomization.Statistics in Medicine,2016,35(5):685-694.
[4] Berger VW,Agnor R.Delayed unmasking and selection bias.Statistics in Medicine,2016,35:2111-2112.
[5] Soares JF,Wu CF.Some restricted randomization rules in sequential designs.Communications in Statistics-Theory and Methods,1983,12:2017-2034.
[6] Chen YP.Biased coin design with imbalance tolerance.Communications in Statistics-Stochastic Models,1999,15:953-975.
[7] Zhao W,WengY.Block urn design—a new randomization algorithm for sequential trials with two or more treatments and balanced or unbalanced allocation.Contemporary Clinical Trials,2011,32 (1):953-961.
[8] Rosenberger WF,Lachin JM.Randomization in Clinical Trials:Theory and Practice,Second Edition.New Jersey:John Wiley & Sons,Inc,2016:38-40,152-155.
[9] Blackwell D,Hodges JL.Design for the control of selection bias.Annals of Mathematical Statistics,1957,28:449-460.
[10] Berger VW.Randomization,Masking andjAllocation Concealment.Boca Ration:Taylor & Francis Group,LLC.,2018:129-130.
[11] 刘玉秀.没有分配隐蔽则没有随机化.医学研究生学报,2016,29(6):561-565.