By Charles DiLiberti and Michael T. Boyne
The comparison of a new product or formulation to an existing one is a critical aspect of the drug development process affecting developers of innovator drugs, biologics, generics, and biosimilars. While bioequivalence is the standard method of comparison for most conventional drugs, comparisons of complex heterogeneous drugs and biologics are generally approached differently via a process that can be thought of broadly as “establishing similarity.” For most complex pharmaceutical products, such similarity comparisons involve extensive data sets with numerous variables. Historically, most similarity comparisons of complex conventional drugs and biosimilars have been accomplished by comparing test and reference product results for each variable or attribute individually using conventional univariate statistical methods, such as tests of equivalence, differences, intervals, visualization, and summary statistics.
The selection of appropriate statistical tools for comparing complex pharmaceutical products is a challenge often guided by past practice and intended use. However, recent regulatory language about establishing similar structural sameness and analytical similarity has led to increasing discussion about the “right” statistical approach to use. While univariate approaches are well established for showing significant differences or statistical equivalence for a single attribute, they can minimize the impact that correlated small changes in multiple attributes may have on the purity, potency, safety, or efficacy of complex products. This has led to increasing interest in multiple attribute (multivariate) statistical approaches and a discussion of the value they may add to this specific regulatory challenge.
Although a wide variety of multivariate approaches—such as principal component analysis, partial least squares, chemometric modeling, and various distance metrics such as Euclidean, Mahalanobis, and Kolmogorov–Smirnov—have been proposed for the evaluation of complex pharmaceutical similarity in scientific literature, the practical application of such multivariate methods to the analysis of complex pharmaceuticals has been mostly limited to other contexts such as surveillance/pharmacovigilance, quality by design, and design of experiments. By evaluating the interrelationships among analytical variables, multivariate methods can “see” important patterns that more conventional univariate methods might miss and, thus, could be important tools to enhance the evaluation of similarity for complex pharmaceutical products.
While much of the discussion of similarity metrics has focused on the evaluation of biosimilars, the challenges of establishing similarity for complex conventional pharmaceuticals with heterogeneous active ingredients (such as low molecular weight heparins, conjugated estrogens, pentosan, and glatiramer), etc. are much alike. The complexity of some drug substances has been a barrier to development of follow-on products, touching upon both the fundamental definitions of what constitutes an active pharmaceutical ingredient and what the demonstration of pharmaceutical equivalence/analytical similarity should entail. The use of appropriate statistical tools can help guide the regulatory process and scientific justification for a variety of possible conclusions about how much alike two complex pharmaceutical products or formulations are, e.g., (1) indistinguishable from one another, (2) significantly different analytically, but without clinically meaningful differences, or (3) significantly different both analytically and clinically, and thus not comparable or similar. The statistical methods used for such similarity comparisons are critically important to ensure the safety, efficacy, purity, and potency of new products, but also to do so in a way that does not impose undue clinical or analytical burdens on sponsors.
We welcome you to join us to learn more about the latest statistical methods to evaluate the similarity of complex pharmaceuticals on October 27 at our symposium Improved Statistical Approaches for Comparability/Equivalence Assessments of Biosimilars and Complex Drugs. This symposium builds on last year’s successful symposium on complex drugs by delving into the quantitative metrics that many have inquired about. Our lineup of three outstanding speakers, Matej Horvat, Ph.D.; Ajaz Hussain, Ph.D.; and Shein-Chung Chow, Ph.D.; will provide insights from the industry, regulatory agency, and academic perspectives. This session is the result of a cross-disciplinary collaboration between the Generic Pharmaceuticals, CMC Statistics, and Biosimilars focus groups, and is a must-see for anyone working with complex pharmaceutical products. Michael and CMC Statistics Chair Helen Strickland, Ph.D., will co-moderate this relevant and timely session.