Niselman A V, Garcia Ben M, Rubio M C
Department of Mathematics, Faculty of Pharmacy and Biochemistry, Buenos Aires, Argentina.
Eur J Drug Metab Pharmacokinet. 1998 Apr-Jun;23(2):148-52. doi: 10.1007/BF03189331.
The aim of this study is to compare four statistical methods for outlier identification in Bioequivalence tests. The methods are based in four confidence intervals, 'parametric','non- parametric','robust with the asymptotic distribution of the M-estimator' and 'robust with the bootstrap distribution'. The drug we used was Diltiazen, in a two sequence randomized crossover study design. The pharmacokinetic parameters measured were the area under the plasma concentration curve (AUC), and the peak drug concentration (CMAX). Time to peak drug concentration (TMAX), was not used here in order to separate the efficiency of the methods from the efficiency of the measurements. The methods were applied to simulated and experimental data. We made two simulations, one with normal data and another one with outliers data. When simulating normal data all methods showed similar profiles and high power. On the contrary, when simulating experiments with outliers data, the parametric method showed low power, whereas robust methods showed just a slight decrement in power. When we analyzed experimental data of AUC, if we used the parametric method (recommended by U.S.P), we were not able to conclude Bioequivalence, but with the other methods, this was possible. This disagreement between parametric and robust procedures was a sign of outliers data. We conclude that the robust methods in bioequivalence assays help us in the identification of outliers as observations with weight equal zero.
本研究的目的是比较生物等效性试验中用于识别异常值的四种统计方法。这些方法基于四个置信区间，即“参数法”、“非参数法”、“基于M估计量渐近分布的稳健法”和“基于自助分布的稳健法”。我们使用的药物是地尔硫䓬，采用两序列随机交叉研究设计。所测量的药代动力学参数为血浆浓度曲线下面积（AUC）和药物峰浓度（CMAX）。为了将方法的效率与测量的效率区分开来，此处未使用达峰时间（TMAX）。这些方法应用于模拟数据和实验数据。我们进行了两次模拟，一次是正常数据模拟，另一次是异常值数据模拟。在模拟正常数据时，所有方法都显示出相似的特征和高功效。相反，在模拟含有异常值数据的实验时，参数法显示出低功效，而稳健法仅显示出轻微的功效下降。当我们分析AUC的实验数据时，如果使用参数法（美国药典推荐），我们无法得出生物等效性的结论，但使用其他方法则有可能得出结论。参数法和稳健法之间的这种不一致是异常值数据的一个迹象。我们得出结论，生物等效性试验中的稳健方法有助于我们将异常值识别为权重等于零的观测值。
