Following the thought-provoking short course some other methodological issues were addressed next day. One of them was Balaam’s design which was not very popular in the past (perhaps due to its inefficiency in terms of sample size) but nowadays it is recommended for interchangeability studies. Júlia Singer presented a summary of apparently different models applied which often lead to the same result, pointing out that inefficiency continues to be a drawback of this design.

Switchability of biosimilars continues to be a controversial topic because the relevance and sensitivity of the proposed PK/PD approach was often questioned. László Tóthfalusi raised the possibility of characterizing switchability using long-term observational data. This type of analysis is closer to real life, and the need of switching occurs in a natural way (whereas in clinical studies switches occur after pre-defined periods). Time to need of switch (for which multiple reasons like decreasing efficiency, side effects etc. can be identified) is the primary endpoint in such observational studies. The results presented at the conference were convincing and informative. The hazard of treatment failure for a certain product is a clinically meaningful and sensitive parameter, enabling also the comparison of more biosimilars of a same originator, and also the analysis of risk factors (co-treatment, gender) from a relatively small clinical database. The variables necessary for this analysis can be obtained from health-insurance databases.

During the sample size evaluation of clinical studies in biosimilar development programs (like phase 1 equivalence of PK parameters, and phase 3 therapeutic equivalence) sample sizes are often underestimated due to the assumption of completely identical distributions for the originator and the biosimilar product. Alternatively, some sponsors routinely assume a small difference between means (like a typical ratio of 1.05 for the geometric means of PK parameters). Arne Ring presented a review of these approaches and proposed a more structured way of handling small differences between distributions by introducing the concept of assurance (expected power). The method is taking into account the uncertainty in the assumptions of sample size assessment, and usually leads to feasible sample sizes.

Maybe the most expected session was the one in which regulators talked about challenges in biosimilar development. The keynote presentation was held by Andrea Laslop, starting with an overall picture of biosimilars in Europe: 61 products registered (out of which 7 were already withdrawn), corresponding to 15 originators. One of the differences between the regulation in EU compared to other regions is the decreasing role of in vivo studies, shifting the emphasis more and more to the quality attributes, and towards a deeper understanding of the relationship between quality attributes and clinical outcome. All this can be summarized in the principle of 3Rs, meaning: Replace, Reduce, Refine.

Ina-Christine Rondak presented reflections on statistical aspects at quality level (CHMP. Draft Reflection paper on statistical methodology for the comparative assessment of quality attributes in drug development. 2017) After a short historical overview the comments received on the draft Reflection Paper were summarized. These comments were summarized and discussed extensively last year during a 5-day workshop (CHMP. Meeting Report: Workshop on the draft reflection paper on statistical methodology for the comparative assessment of quality attributes in drug development. 2018) where 31 representants of different pharma industry associtations, 31 EU regulators from 21 national competent authorities (+1 FDA representant) exchanged ideas. Some of the commenters pointed out that statistical decision should not be limited to a pass/fail criterion. Some of the other questions and concerns raised by commenters: the scope of reflection paper is too broad, more specific reflections dependent on the setting would be needed. The reflection paper uses the term similarity but it does not give an exact definition. There might be an inherent contradiction between „consistent manufacturing“ and shift/drifts in means.

Andreas Brandt continued these thoughts by analysing the questions and concerns from statistical point of view, pointing out that similarity can’t be identified with equivalence in mean, but with equivalence in distribution. And the main distinction between descriptive and inferential statistics does not consist in different methodologies but in the conclusions drawn from the results (valid only for the observed batches or extrapolating for other batches as well).

Stephan Lehr’s interactive presentation (during which – like in a quiz – the audience had to guess and recognize some quotations from different guidelines) formulated some challenging questions preparing in a way also the next session. Stimulating thoughts were grouped around 4 topics:

• Revision of „products containing recombinant granulocyte-colony stimulating factor”
• Guideline on estimand in clinical equivalence trials (Appendix to ICH E9)
• Bayesian approaches to demonstrate biosimilarity
• Development in rare diseases

This session was followed by a regulatory panel discussion and Q&A type of interaction in which replies were given to critical questions from the audience with a focus on the current regulatory requirements, approval process and burning issues of debate facing clinical development teams specialized on biosimilars.

It seemed that the main challenge for companies developing biosimilars was still the “moving target” phenomenon, i.e. the fact that the originator also has its own life cycle in which changes in the manufacturing process may occur. Some of the questions referred to the new terminology defined in the appendix of the old statistical guideline (ICH E9 (R1) addendum on estimands and sensitivity analysis in clinical trials to the guideline on statistical principles for clinical trials) which is expected to become effective in the near future, with some changes compared to the draft version (due to the large number of comments received to the draft version).

Discussions were open, colloquial, thought-provoking, and could be continued „till cows come home”, but had to be censored once. However, the participants were confident that they will meet next year at the 5^th forum.