The bioequivalence studies on thirteen drugs reported after 1980 were analyzed with the conventional analysis of variance (ANOVA), the power approach, the 95% confidence intervals and FDA's the two one-sided tests. The powers of test for the bio-availability relevant parameters were generally poor that among 33 parameters studied, only two parameters of which the power exceeded 0.8 and there were fourteen parameters of which the powers were less than 0.3 The ANOVA and the power approach are not the satisfactory framework for the bioequivalence assessment. The confidence intervals and the two one-sided tests are more informative and can facilitate the professional judgement on the physical meaning of the estimated mean difference. It is therefore recommended that the conventional ANOVA and the power approach should not be regarded as the crucial criteria for the bioequivalence assessment of drug products. 傳統上慣用變異數分析法( ANOVA )來推斷不同製造源而藥劑學上同等的二製劑間是否具有生體相等性。然而事實上,倘若P>0.05而無法否定歸雩假說時,並不能据此而主張二者間有可能類同或相等。所以需要使用能檢出有意義的差異度的機率,稱為檢出力( 1-β)者來支持二者有可能不具有意義的差異之主張。
以往在主體相等性之研究報告中明示檢出力者為數極少,大部份僅以簡易的 t 檢定或ANOVA之無顯著差異之結果就冒然主張類同,致使其主張無可信性。近年來因統計科學的進步闡明了歸零假說之t檢定及ANOVA並不適用於主張類同性,而發展出Bayes氏法, 95%可信問距法及二個單側檢定法。 本報告以傳統的ANOVA,檢出力法, 95%可信閭距法及二個單側檢定法,分別檢討1980年以後被發表的十三種藥物製劑之主體可用率及生體相等性報告,獲得如下知見。 1.報告中有關生體可用率之各種指標值( parameters ),雖在ANOVA獲得無顯著差異的結果,但其檢出力皆很小。三十三個指標值中竟有十四個之檢出力小於0.3,只有二個指標值之檢出力大於0.8。照國內現行生體相等性基準,全部皆屬於非生體相等性。 2.有部分指標值雖然檢出力很小,但卻在FDA推荐的二個單側檢定法顯示合乎小於+/- 20%差異之生體相等性條件,故不能認為是非主體相等性。相反的,也有一個指標值( MRT, sulpiride),其20%差異度之檢出力高達0.88,但以95%可信間距法推定其差異度時,卻大於十/- 20%的範圍。更有一指標值( AUC. warfarin ),ANOVA剌定為有極顯著差異( P<O.O1),但95%可信間距法及二個單側檢定法推定其差異度則小於十/- 11%。若照ANOVA之結果則是不具生體相等性,但若依後二法判定則不能認定不是生體相等性。這種矛盾及困擾必須加以辨明。3.本質上,ANOVA及檢出力的統計解析並無醫藥上的相等性或非相等性的判別功能。況且歸零假說原本是為期待被否定而設立,因之不適合作為無差異性之主張依据,只能供作選篩性的參考資訊。生體相等佐的制定應使用較合理且可讓醫藥專家衡量其差異度的意義的95%可信間距或二個單側檢定法。因為醫藥問題並非單純的統計解析所能解答,統計解析的結果須賴專業知識的考量才能賦與其意義。 4.二個單側檢定法所推定的可信問距常比95%可信間距更為狹窄,此乃二個單側檢定法是相當於兩側檢定之90%可信問距的緣故。