New Quality-Range-Setting Method Based on Between- and Within-Batch Variability for Biosimilarity Assessment

Analytical biosimilarity assessment relies on two implicit conditions. First, the analytical method must meet a set of requirements known as fit for intended use related to trueness and precision. Second, the manufacture of the reference drug product must be under statistical quality control; i.e., the between-batch variability is not larger than the expected within-batch variability. In addition, the quality range (QR) method is based on one sample per batch to avoid biased standard deviations in unbalanced studies. This, together with the small number of reference drug product batches, leads to highly variable QR bounds. In this paper, we propose to set the QR bounds from variance components estimated using a two-level nested linear model, accounting for between- and within-batch variances of the reference drug product. In this way, the standard deviation used to set QR is equal to the square root of the sum of between-batch variance plus the within-batch variance estimated by the maximum likelihood method. The process of this method, which we call QR_ML, is as follows. First, the condition of statistical quality control of the manufacture process is tested. Second, confidence intervals for QR bounds lead to an analysis of the reliability of the biosimilarity assessment. Third, after analyzing the molecular weight and dimer content of seven batches of a commercial bevacizumab drug product, we concluded that the QR_ML method was more reliable than QR.