Metzler C M
Upjohn Company, Kalamazoo, Michigan 49001.
Stat Med. 1991 Jun;10(6):961-9; discussion 969-70. doi: 10.1002/sim.4780100617.
In recent years a number of decision rules, based on sound statistical principles, have been proposed for deciding if a test formulation is bioequivalent to a reference formulation. The decision rule based on confidence intervals has been accepted by regulatory agencies, at least by the Food and Drug Administration of the United States. A useful property of this decision rule is that the regulatory agency need not require a certain sample size, since the level of protection against wrongly deciding bioequivalence is set by the choice of the alpha level used to compute the confidence intervals. The manufacturer claiming bioequivalence is concerned about sample size, for sample size determines the probability of falsely deciding non-bioequivalence when the test formulation does indeed have an acceptable relative bioavailability. Curves of probability of rejecting bioequivalence have been computed for error coefficient of variation of 10, 20 and 30 per cent, for relative bioavailability from 70 to 130 per cent, and for protection levels of 90 and 95 per cent. These curves can be used for choosing the sample size for a bioequivalence study.
近年来，基于合理的统计原则，人们提出了一些判定规则，用于确定一种测试制剂是否与参比制剂生物等效。基于置信区间的判定规则已被监管机构所接受，至少在美国食品药品监督管理局是如此。该判定规则的一个有用特性是，监管机构无需规定特定的样本量，因为用于计算置信区间的α水平的选择决定了防止错误判定生物等效性的保护水平。声称生物等效的制造商关注样本量，因为样本量决定了在测试制剂确实具有可接受的相对生物利用度时错误判定非生物等效性的概率。对于变异系数为10%、20%和30%的误差、相对生物利用度为70%至130%以及保护水平为90%和95%的情况，已经计算出了拒绝生物等效性的概率曲线。这些曲线可用于为生物等效性研究选择样本量。
