The Degrees of Correlation in Bioavailability Parameters and the Confidence Regions in Bioequivalence Study

生體可用率有關的參數間之相關度及生體相等性試驗的可信區域

許興智(Hsing-Chu Hsu) ； 陸海林(Hai-Lin Lu)

《藥物食品分析》 3卷1期 (1995/03) Pp. 17-26

摘要

傳統上，判定二種藥劑是否具有生體相等性皆沿用簡易之單變數統計解析法。對C(下標 max)，AUC(下標 ∞)等參數分別推定標準品和試驗品間之平均差異度的95%或90%可信間距作為是否具有類同性之判定基準。
二個95%可信間距構成一可信區域。倘若C(下標 max)和AUC(下標 ∞)互不相關，為各自獨立之變數則此一可信區域涵蓋C(下標 max)和AUC(下標 ∞)之推定母平均向量的下限機率為(0.95)^2=90.25%。但是在藥物動力學已知C(下標 max)及AUC(下標 ∞)為高度相關之參數，於是此一可信區域之機率隨相關度之大小而異。
本文以實例闡明在每一個生體相等性試驗，C(下標 max)和AUC(下標 ∞)之相關度皆不同且差異頗大。因之，以簡易之單變數統計解析法來判定二種藥劑是否具有生體相等性所據的機率在各個試驗並不相同。由C(下標 max)和AUC(下標 ∞)各別之95%可信間距構成之矩形可信區域顯然地有異於90%可信橢圓。90%可信橢圓正確地規範C(下標 max)和AUC(下標 ∞)之推定母平均向量存在之90%可信區域而二個95%可信問距構成之矩形可信區域涵蓋推定母平均向量之可信機率到底有多大不但不清楚而且隨著不同之試驗而不同。
因此，若欲以明確而一致的可信機率為基準來判定二種藥劑是否具有生體相等性，雙變數可信橢圓法是較合理可行的方法。然而，每一種統計分析方法皆有其缺陷，現階段生體相等性的判定最好參照多種分析方法的結果比較適宜。

並列摘要

The parameters to define the shape of the 1-α confidence ellipse constructed by the C(subscript max) and AUC(subscript ∞), the physical meanings of the shape of the ellipse, correlation of the C(subscript max) and AUC(subscript ∞), and confidence regions of two 95% confidence intervals and the 90% confidence ellipse was described and discussed. The ratio of the lengths of the major and minor axes for the confidence ellipse was a proportional function of the coefficient of correlation between the two relevant bioavailability variables. The degree of correlation could vary significantly from one bioequivalence study to another. The degree of correlation has great impact on the probability of confidence region and, hence, in univariate analysis each bioequivalence decision-making would be unavoidably based on the different probability. Examples of the bioequivalence study of nine drugs were demonstrated. In order to assess bioequivalence with a consistent probability, the bivariate method using two 90% confidence ellipses of reference and test products is proposed. Since each available statistical method has its limitation and shortocoming, it is desirable to evaluate bioequivalence using multiple assessment methods.

並列關鍵字

Correlation ； C(subscript max) ； AUC(subscript ∞) ； Confidence ellipse ； Confidence region