I read with interest the paper by Del Tacca et al. [1]. I agree that verification of the bioavailability of marketed pharmaceutical products is an interesting topic, since not all marketed products have been shown to be bioequivalent to the reference product. It is only possible to believe naively that all products on the market are bioequivalent with the reference product from the complete ignorance of the pharmaceutical legislation of the European Union [2]. Therefore, a list similar to the FDA Orange Book [3], which identifies the (bio)equivalent products and its reference product, is essential for prescribers. Interestingly, the legislation allowing the authorization of products with unknown bioavailability (e.g. bibliographical applications), but with the dosage instructions of the reference product, or the absence of a ‘European Orange Book’ are not criticized. On the contrary, the criticism is motivated by the substitution policies, which affect economically the reference products and prescribers, and are focused on the second-entry products that assure best the safety and efficacy profile and interchangeability with the reference product. Average bioequivalence is not only a surrogate of therapeutic equivalence but also of equivalent biopharmaceutical quality, which assures that generic and reference products will behave in the same way in all individual patients, irrespective of their demographics, concomitant medication or illnesses. From my point of view, this failed bioequivalence study has been used as an excuse to criticize the authorized generic pharmaceutical products and the substitution policies, but this is not scientifically correct for the following reasons. First, it would be convenient to distinguish between bioequivalence (i.e. 90% CI within the acceptance limits), bioinequivalence (90% CI completely outside of the acceptance limits) and non-equivalence (90% CI with some part inside and some part outside of the acceptance limits). In practical terms, it is necessary to conclude inequivalence in order to conclude that a generic is not similar to the reference product, since non-equivalence is inconclusive and another study with more statistical power (i.e. a lower variability or a higher sample size) might be able to conclude equivalence. In statistical terms, it is incorrect to conclude that one of the generic products is not equivalent simply because the authors were unable to conclude equivalence (90% CI for Cmax 0.7921, 1.0134). The inability to reject the null hypothesis does not support the validity of the null hypothesis. Only the alternative hypothesis can be proved in a test of hypothesis. In bioequivalence studies the null hypothesis is that the products are inequivalent and the alternative hypothesis is that the products are equivalent. In other words, absence of evidence (of bioequivalence) is not evidence of absence (of bioequivalence). On the contrary, the study results suggest that a study with lower variability or higher sample size would be able to show bioequivalence since the lower boundary of the 90% CI is very close to the acceptance range. Therefore, it is reasonable to believe that the marketing authorization holder will have sponsored a bioequivalence study that was able to conclude equivalence. In this respect, it is surprising that this paper was accepted without requiring the identification of the products under investigation. Presently, the study cannot be replicated, the Marketing Authorization Holders cannot defend themselves by submitting evidence of the bioequivalence of the products in another bioequivalence study and it is not possible to verify if they are products approved based on bioequivalence or not. Second, the authors claim that the reference product showed 8.5 and 5.4% greater AUC, but it is not statistically appropriate to give validity to a difference in point estimates that was not able to reach statistical significance. As all 90% confidence intervals included 100%, this study was not able to find any statistically significant difference between products, which, at the same time, is not enough to conclude equivalence because a large variability makes the study inconclusive. In addition, the study results show some inconsistencies that cast doubt on its correctness. According to the authors the AUC CV was 27.4% and the Cmax CV was 27.6%, which illustrates that the statistical analysis has been performed on the combined data from the three formulations. If this were the case, the width of the 90% CI would be the same for all three comparisons. However, the width of all three comparisons is different, which shows that the three comparisons have been calculated with a different CV, some of them higher than 30% in contrast to the authors' claim. Many other statements from the authors deserve a comment, but there is no room for more. In my opinion, this is enough to illustrate that the paper by Del Tacca et al. [1] describes an inconclusive study and only adds noise to the debate on the validity of bioequivalence as demonstration of therapeutic equivalence and interchangeability in case of generic substitution.
Read full abstract