One of the questions I am most often asked at workshops and by aspiring biosimilar developers is, which physicochemical and biological parameters should be included in a comparability study to demonstrate biosimilarity between a reference product and candidate biosimilar. The short an easy answer is to say that any attribute or characteristic that has an impact on safety and efficacy should be considered for inclusion in such a study. This, I know, is not very helpful as virtually any structural or functional attribute of a biologic can potentially affect its clinical performance. Since the SAHPRA has adopted the totality-of-evidence paradigm in the assessment and approval of biosimilars (as opposed to evaluating quality and clinical data packages separately by independent committees) the best way to begin the development of a biosimilar is to design the comparability study, which is the cornerstone for biosimilar approval, by using a stepwise approach. This approach starts with analytical studies that compare structural and functional properties of the reference and test products, which are relevant to clinical safety and efficacy followed by nonclinical studies, and a single clinical study, to confirm biosimilarity in terms of safety and efficacy.
The parameters to be included in the analytical comparability study are those chemical, physical, biological, and microbiological attributes that can be defined, measured and continually monitored during the manufacturing process to ensure that the final product remains within acceptable quality limits. From these are selected, based on risk ranking, the parameters that are most likely to affect clinical outcomes. These parameters or characteristics are referred to as the critical quality attributes (CQAs) of the product. Not all CQAs will have a high impact on clinical outcomes – some will have a greater effect than others. Hence, CQAs that affect clinical outcomes the most must be shown to be highly similar between the reference and candidate biosimilar, whereas those with a lesser impact should be shown to be comparable but not necessarily equivalent. In this regard a guidance document published by the FDA in 2012 (referred to in Chow et al, 2016), may be helpful in that it proposes that CQAs be divided into three tiers. Tier 1 includes CQAs that have the greatest influence on clinical performance while those in Tiers 2 and 3 have moderate and only slight effects, respectively. The FDA further proposes that for CQAs in Tier 1 an equivalence test be used to compare the reference and test products, a quality range approach for those in Tier 2 and a graphical presentation, or the actual raw data, for CQAs in Tier 3. The equivalence test (which is like that used for demonstrating bioequivalence between a generic and its innovator) is statistically more rigorous than the quality range approach, which in turn is more stringent than the raw data or graphical presentation of the data. As an example, if you should consider developing a biosimilar of erythropoietin then the amino acid sequence, glycan structure, biological activity, insoluble and high molecular mass aggregates, protein concentration and host cell impurities would be considered Tier 1 CQAs as they have a direct impact on safety and/or efficacy. Others such as isoform distribution and receptor binding will have a lesser impact and consequently will be classified as Tier 2, and finally, deaminated and oxidized erythropoietin variants will be regarded as Tier 3 CQAs as they have the least effect on clinical outcomes. In fact, the last two are often considered as product-related substances, rather than impurities.