Comment on: “Levothyrox® New and Old Formulations: Are they Switchable for Millions of Patients?”

Abstract

I read with interest the article published in Clinical Pharmacokinetics by Didier Concordet and colleagues [1], who reported an independent revised statistical analysis of the bioequivalence study between old and new formulations of Levothyrox® [2]. Using the approach of individual bioequivalence (IBE) instead of average bioequivalence (ABE), more than 50% of healthy volunteers enrolled in this bioequivalence study were outside the a priori bioequivalence range and thus the authors questioned whether the switchability of these two drugs was a safe option in patients. Meanwhile, a ‘Levothyrox scandal’ has legitimately agitated the French public and media due to an unexpectedly large number of adverse events occurring soon after the launch of the new formulation. In this sensitive context, the conclusion of this reanalysis could be falsely interpreted and therefore counterproductive. So, I feel the need to raise some major concerns. First, their article perfectly illustrates a decade of scientific research that eventually led to IBE being abandoned, in particular when the same bioequivalence study could end up with a discordant conclusion: ABE accepted but IBE rejected or, conversely, ABE rejected but IBE accepted. No satisfactory regulatory decision/rule could be established. In the past, several generic versions of levothyroxine received marketing approval on the basis of ABE in the standard acceptance range (80–125%), with public health concern focused not on safety but on quality due to loss of the active ingredient over the duration of the shelf life. On the basis of a US Food and Drug Administration (FDA) initiative, the assay range specification for levothyroxine sodium tablets was reduced from 90–110% to 95–105% of the label claim in 2007 and classification of levothyroxine as a narrow therapeutic index (NTI) drug was adopted in 2014. To achieve this, the FDA required a bioequivalence study designed as a fully replicated crossover approach in order to scale the bioequivalence limits to the variability of the reference product and compare the test and reference product for within-subject variability [3], meaning that each subject should receive each formulation on two occasions. Statistical details can be found in Jiang et al. [4]. In Europe, the European Medicines Agency (EMA) did not endorse this scaled ABE approach, preferring a direct tightening of ABE limits to 90–111.11%. This historic recap should help provide an understanding of the bias of the revised statistical analysis by Concordet et al. [1]. Replicated crossover designs are critical when an IBE approach is used to allow separate estimation of within-subject variances for test and reference products and the subject-by-formulation interaction variance component. As noted by the authors, the study design was not replicated and details that would allow period and sequence fixed effects to be taken into account were not available. Thus, the implemented model looked like IBE but was not IBE. Then, applying the ABE limits (90–111.11%) as IBE limits is too conservative and could lead to the old formulation of Levothyrox® being declared not individually bioequivalent to itself. With this suggested IBE model, and considering Table 1 in the article by Concordet et al. [1], 202 patients (99%) would meet IBE (80–125%) criterion on unadjusted data and 125 patients (61%) on baseline-adjusted data. In my opinion, and in keeping with the EMA approach, the correct way to show bioequivalence with levothyroxine should be the use of ABE limits (90–111.11%) with IBE limits (80–125%) as supportive data, in a real replicated design. I acknowledge that the authors did not claim a lack of bioequivalence between the old and new formulations of Levothyrox® but suggested that ABE may not guarantee their switchability as the study was performed in healthy volunteers. This is a redundant and unfounded criticism that has persisted during the long story of bioequivalence with This comment refers to the article available at https ://doi. org/10.1007/s4026 2-019-00747 -3.