Probit和Logit模型计算半数反应量的差异研究
|
摘要
基于Bliss方法的Probit和Logit模型是计算半数反应量最常用的模型。本文比较了《美国药典》和《欧洲药典》对Probit和Logit模型选用的原则,并通过计算机模拟实验比较了2种模型计算结果的区别。从实用的角度,Probit和Logit模型计算ED_(50)得出的结果差异很小,因此本文建议参考《美国药典》,按统计软件的功能选取Bliss法的具体模型。选取Probit和Logit模型外的其他模型,必须有实验或理论证据的支持并得以验证,不能仅根据日常实验数据的拟合优度确定。
The Bliss method-based Probit model and Logit model are the most commonly used ones in calculating the median effective dose. In this article,the choices between the Probit model and the Logit model in the US Pharmacopeia and the European Pharmacopoeia are compared. The calculation results from the two models are analyzed. The Probit model and the Logit model deliver similar median effective dose estimations without meaningful difference. Which model to choose should be based on the availability of software that meets the laboratory's analysis and reporting needs,as suggested by the US Pharmacopeia. Choosing a model other than Probit or Logit is appropriate,only if it is supported and validated by experimental or theoretical evidence,and can not be justified simply by the fitness of data from routine tests.
The Bliss method-based Probit model and Logit model are the most commonly used ones in calculating the median effective dose. In this article,the choices between the Probit model and the Logit model in the US Pharmacopeia and the European Pharmacopoeia are compared. The calculation results from the two models are analyzed. The Probit model and the Logit model deliver similar median effective dose estimations without meaningful difference. Which model to choose should be based on the availability of software that meets the laboratory's analysis and reporting needs,as suggested by the US Pharmacopeia. Choosing a model other than Probit or Logit is appropriate,only if it is supported and validated by experimental or theoretical evidence,and can not be justified simply by the fitness of data from routine tests.
引文
[1]TREVAN JW.The error of determination of toxicity[J].Proc RSoc Lond B,1927,101(712):483-514.
[2]王洋,项荣武,田百会,等.Bliss法四种模型计算半数反应量和治疗指数的差异探究[J].沈阳药科大学学报,2015,32(6):463-469.
[3]BLISS,CHESTER I.The method of Probits[J].Sci,1934,79(2037):38-39.
[4]BLISS,CHESTER I.The method of Probits-a correction[J].Sci,1934,79(2053):409-410.
[5]孙瑞元.简捷实用的半数致死量综合计算法[J].药学学报,1963,10(2):65-74.
[6]黎七雄,汪晖.半数致死量(LD50)Bliss法的评价及计算[J].数理医药学杂志,1995,8(4):318-320.
[7]COX DR.The regression analysis of binary sequences[J].J RStat Soc Series B Stat Methodol,1958:215-242.
[8]Pharmacopeia US.USP 40 NF 35[S].Rockville:USP,2017.
[9]European directorate for the quality of medicines.European pharmacopoeia 9th edition[S].EDQM,2016.
[10]ALDRICH,JOHN H,FORREST DN.Linear probability,Logit,and Probit models.Vol.45[J].Sage,1984:30-47.
[11]KNUDSEN LF,CURTIS JM.The use of the angular transformation in biological assays[J].JASA,1947,42(238):282-296.
[2]王洋,项荣武,田百会,等.Bliss法四种模型计算半数反应量和治疗指数的差异探究[J].沈阳药科大学学报,2015,32(6):463-469.
[3]BLISS,CHESTER I.The method of Probits[J].Sci,1934,79(2037):38-39.
[4]BLISS,CHESTER I.The method of Probits-a correction[J].Sci,1934,79(2053):409-410.
[5]孙瑞元.简捷实用的半数致死量综合计算法[J].药学学报,1963,10(2):65-74.
[6]黎七雄,汪晖.半数致死量(LD50)Bliss法的评价及计算[J].数理医药学杂志,1995,8(4):318-320.
[7]COX DR.The regression analysis of binary sequences[J].J RStat Soc Series B Stat Methodol,1958:215-242.
[8]Pharmacopeia US.USP 40 NF 35[S].Rockville:USP,2017.
[9]European directorate for the quality of medicines.European pharmacopoeia 9th edition[S].EDQM,2016.
[10]ALDRICH,JOHN H,FORREST DN.Linear probability,Logit,and Probit models.Vol.45[J].Sage,1984:30-47.
[11]KNUDSEN LF,CURTIS JM.The use of the angular transformation in biological assays[J].JASA,1947,42(238):282-296.