Author byline as per print journal: Hillel P Cohen, PhD; Caridad Pontes, MD, PhD; Fernando de Mora, MBA, PhD
Abstract: |
Submitted: 20 December 2024; Revised: 6 January 2025; Accepted: 8 January 2025; Published online first: 14 January 2025
Comment on ‘Misinformation about interchangeable biosimilars undermines US health policy, physician confidence, and patient health’ (GaBI Journal, 2024;13(2):55-60)
Dear Editor,
We have concerns about an article from Reilly and McKibben entitled ‘Misinformation about interchangeable biosimilars undermines US health policy, physician confidence, and patient health’ that was published in GaBI Journal on 15 July 2024 [1]. In this article Reilly and McKibben criticize the meta-analysis published by Herndon et al. [2] and conclude that the conclusions reached by Herndon et al. are not supported by the data. We believe that there are numerous severe factual errors within Reilly and McKibben that question their conclusions.
Reilly and McKibben warn the readers about what they consider a reduction of the quality standards required by the US Food and Drug Administration (FDA) for a biosimilar being awarded the interchangeability designation. The authors fundamentally claim that interchangeability, which in the US refers to the possibility for the pharmacist to dispense either the biosimilar or the originator without the prescriber’s intervention, should rely on pre-authorization switching studies in patients. The underlying view of the publication is that allowing pharmacists to foster transition from the originator to the biosimilar in a given patient without such studies having been conducted, may put the patient at an unexpected safety risk.
First and foremost, the publication does not accurately portray the science of ‘biosimilarity’ and of the evidence needed to warrant safety of the ‘exchange’ between the original reference biological and a biosimilar.
The authors do not acknowledge that the concepts of analytical and functional similarity are the key facts underpinning biosimilar and interchangeability [3, 4]. Drugs with the same structure and function, by definition, will have the same clinical impact, including efficacy and safety events [5-7]. Instead, Reilly and McKibben focus on the meta-analysis published by Herndon et al. as if the clinical data evaluated in the meta-analysis were the sole basis supporting single and multiple switches between reference product and biosimilar.
Throughout their paper, Reilly and McKibben allude to the premise that ‘without conducting switching studies, it is unknown whether efficacy diminishes or if side effects arise more frequently following a switch’.
The view that without pre-authorization switching studies the preservation of the therapeutic benefit-to-risk balance in a patient being changed to a biosimilar brand is uncertain is against the very basic principle of biosimilars regulatory science.
Indeed, as the authors state ‘biosimilars are not identical to their reference biologicals’, in as much as any original biological medicine is not identical to itself from batch-to-batch. Variability of product characteristics is intrinsic to complex living cell production processes, and clinically acceptable changes in laboratory analysed structural and functional attributes occur in any original biological. In light of the inherent inter-patient variability of the therapeutic response, a minor variability of those analytical attributes is generally undetectable and has not been of concern to healthcare professionals for decades [8].
Based on the critical reliance upon the certainty that in pharmacology function follows form, the most sensitive, and scientifically justified approach to anticipating equivalent efficacy and safety between a biosimilar preparation and the reference product is therefore the demonstration that differences in critical structural and functional attributes do not exceed the intrinsic structural variability of the original medicine [8].
Another major omission is the indisputable fact that there is no data nor has there ever been any data to support the theoretical immunogenicity concern related to multiple switching between reference product and biosimilar.
The basis for requesting a multiple-switch study to obtain an interchangeability designation in the US was based on a theoretical concern of an increased incidence of immunogenicity after multiple switching between reference product and biosimilar. When considering immunogenicity, the primary determinant of immunogenicity elicited by proteins is the primary amino acid sequence. The amino acid sequence of the reference product and biosimilar must always be identical. Other attributes may also have an impact but given the essentially overlapping structural attributes based on the inherent variability to the originator, the risk of differences in anti-drug antibody formation is very low.
Given their essentially overlapping analytical features, no increase in immunogenicity is expected nor has any increase in immunogenicity ever been detected with biosimilars approved in the US or EU when given to naïve patients or when switched one or more times with the reference product [2, 9, 10]. We believe that this is a glaring omission that colors the conclusions that are reached in the publication.
Factually, there are numerous important scientific flaws in the Reily and McKibben arguments, that we comment on in detail below. These include the following:
- Omission of real-world observational studies in the meta-analyses
Reily and McKibben claim that the meta-analysis conducted by Herndon et al. is flawed because it includes only randomized clinical trials and omits real-world data. However, besides the obvious issue of heterogeneity in designs, randomized clinical trials are the best source of direct comparisons allowing conclusions to be drawn on causality of the exposure and the outcomes, while observational data is inherently subject to potential biases.
- Reliance on single-switch versus multiple-switch studies
Reilly and McKibben argue that the Herndon publication places too great reliance on single-switch studies,whereas the US concept of interchangeability involves multiple switches. This is true when considering the total number of studies included by Herndon et al. However, Reilly and McKibben do not mention that Herndon et al. identified and included 16 multiple-switch randomized clinical trials in their meta-analysis in addition to the 28 single-switch studies. The combination of single-switch and multi-switch studies by Herndon et al. is a pragmatic approach that summarizes differences to controls in a number of clinically plausible situations including single and multiple switching. Furthermore, and very importantly, the supplemental analyses by subgroups that are provided by Herndon et al. do not show statistically meaningful differences between single and multiple switching studies.
There is also a growing literature of biosimilar-to-biosimilar switching studies that are effectively multiple switches of a single molecule. To date, there are more than 30 published biosimilar-to-biosimilar switching studies of various sizes and with a variety of molecules that cumulatively enrolled more than 6,000 patients [11]. There were no unexpected safety observations and no increase in immunogenicity was detected. While it is acknowledged that direct comparison of observational data and interventional studies is not methodologically appropriate, intellectual honesty demands that the totality of existing multi-switch data should be acknowledged in the critical interpretation of the results of the meta-analysis.
- Heterogeneity in types of molecules studied and in patient populations.
Reilly and McKibben criticize Herndon et al. for including many different molecules in their analysis. While the meta-analysis does include diverse classes of products with different physicochemical properties ranging from small proteins (insulin, epoetin, filgrastim) to fusion proteins (etanercept) and monoclonal antibodies (chimeric, humanized and human), again their approach is pragmatic in the sense that it summarizes many different controlled scenarios that contribute to the estimation of differences to controls, thus reflecting wider external validity of the conclusions. Importantly, the supplemental analyses by subgroups that are provided by Herndon et al. does not show statistically meaningful differences between product types.
Reilly and McKibben claim that individual patient populations are not evaluated or even discussed within Herndon et al. However, the controlled nature of the studies and the variety of clinical situations that were included in the meta-analysis highlight that there are no clinically meaningful effects of switching when compared to patients that were not switched. Besides, since trials may detect only large clinical differences which are unlikely to happen with highly similar structural attributes, comparative clinical data should not be necessarily deemed superior to the analytical, biological function and clinical pharmacokinetic/pharmacodynamic (PK/PD) data used to establish biosimilarity. It is an established fact that clinical studies are poorly sensitive to detect small clinical effects and are, at best, a blunt tool when compared to the other data used to establish biosimilarity [6].
- Insufficient safety variables
Reilly and McKibben claim that the three safety profiles ‘do not equate to the other multifactorial considerations.’ This statement minimizes the value and robustness of the safety factors that were considered: (1) death; (2) severe adverse events; and (3) discontinuation. The three main variables used are widely standardized in clinical trials and can be deemed as comparable across studies. Risks are relative to control groups, so conceptually they can be properly summarized across studies.
Reilly and McKibben criticize the fact that the meta-analysis of Herndon et al. is limited to safety events and omits efficacy data. However, Reilly and McKibben overlook the fact that all randomized clinical trials included the meta-analysis of Herndon et al. were statistically powered for principal efficacy endpoints, were already assessed regarding their main variables for concluding biosimilarity, and the overwhelming majority of these studies met their efficacy endpoints. Of the very few studies that missed the endpoints, FDA and the European Medicines Agency (EMA) concluded that the endpoints were not missed by much, and that totality of evidence from all other data sources, including analytical biological function, clinical PK/PD data, supported a conclusion that those products were biosimilar [6, 12]. Post-approval surveillance data supports the decisions of EMA and FDA [13, 14]. To note, loss of efficacy, if occurred, could be indirectly reflected in early discontinuation, which was one of the primary variables evaluated in the meta-analysis, and in the evaluation of discontinuations there were no reasons for concern, either in the main analysis nor in the subgroups analyses.
In contrast, the safety data from the pivotal randomized clinical trials were presented in a descriptive manner because the frequency of the individual safety events was too low to be analysed statistically. Actually, clinical studies would need to enrol exceptionally large numbers of patients to quantify safety events with the low incidence observed in biosimilar studies. For example, massive clinical studies are often conducted to assess the safety of novel vaccines before they are licensed to be administered to healthy individuals in order to have sufficient statistical power to quantify relatively rare safety adverse events. It is only by combining the safety data from multiple studies that the overall safety profile of biosimilars may be compared to the overall safety of reference products.
- Confidence intervals that are too wide
Reilly and McKibben claim that the conclusions reached by Herndon et al. are invalid because of ‘the substantial variability and uncertainty in the data, as indicated by large variances in sample sizes and broad error bars’. To note, while individual trials might provide large confidence intervals, the final sample sizes were large, and the meta-analysis yielded weighed precise estimates for the three safety variables that were evaluated.
- Conclusions based on ‘lack of differences’
Reilly and McKibben repeat the aphorism that ‘absence of evidence is not evidence of absence is acutely applicable to the importance of switching studies in the demonstration of interchangeability’. But when evaluating safety in a large meta-analysis, a statistical absence of differences is very much a valid piece of data, especially in the context of consistent trends across products and outcomes. There is a lack of adverse signals in Herndon et al. so that none of the relevant potential adverse consequences of non-exchangeability (i.e. death, severe adverse events or early discontinuation) show differences of magnitude enough as to be detected. If there were additional adverse events missed in the meta-analysis, these are definitely not translated into relevant differences in the important clinical consequences of death, severe adverse events or treatment discontinuation.
Conclusions
Reilly and McKibben state that ‘third-party biosimilar substitution raises a variety of issues pertaining to patient and physician consent, maintenance of accurate patient records, clear product identification and traceability, adequate pharmacist education, and clear communication between healthcare providers’. At face value, these statements are a reasonable position. However the major issue with their article is: (1) that their position is substantiated on unfounded methodological criticisms; (2) the way to prevent such potential effects is not to deny the basic science of biosimilarity and question FDA’s decision drivers, but to explain why the science behind biosimilars allows for an exchange without disproportionate requests for large clinical studies, as is being widely discussed among regulatory bodies from multiple countries [15].
By questioning the methodology and conclusions of Herndon et al. by use of flawed reasoning, Reilly and McKibben are contributing to spread misinformation and distrust in biosimilars and may be contributing to the detrimental consequences that they themselves highlight.
Funding sources
No funding was received for the drafting of this letter.
Disclosures
HPC was an employee of Sandoz until his retirement in early 2024. While employed at Sandoz he was an author of several systematic reviews on the topic of biosimilar switching. Sandoz had no role whatsoever in any aspect of this Letter-to-the-Editor.
FdeM is full professor of pharmacology at the Universidad Autónoma de Barcelona, Spain. Besides being principal investigator in government-funded research projects on novel approaches to biosimilar characterization, he acts as a speaker and a consultant in biosimilar science, regulation and market for the biopharmaceutical industry. He also collaborates with healthcare administrations and with numerous public and private organizations.
CP is clinical pharmacologist at the Hospital de la Santa Creu i Sant Pau and associate professor of pharmacology at the Universitat Autònoma de Barcelona, Spain. She is an external expert to the Spanish Agency of Medicines and Medical Devices for clinical pharmacology and has long-lasting experience in drug evaluation. CP has no industrial conflicts of interest.
Competing interests: The authors declare no conflicts of interest.
Provenance and peer review: Not commissioned; internally peer reviewed.
Authors
Hillel P Cohen, PhD
Caridad Pontes, MD, PhD
Fernando de Mora, MBA, PhD
Author for correspondence: Hillel P Cohen, PhD, Highland Park, NJ 08904, USA |
Disclosure of Conflict of Interest Statement is available upon request.
Copyright © 2025 Pro Pharma Communications International
Permission granted to reproduce for personal and non-commercial use only. All other reproduction, copy or reprinting of all or part of any ‘Content’ found on this website is strictly prohibited without the prior consent of the publisher. Contact the publisher to obtain permission before redistributing.