Background The Disease Activity Score (DAS) and American College of Rheumatology (ACR) response criteria are the 2 most commonly used measures to assess clinical response in rheumatoid arthritis (RA) trials. DAS28 is a modified DAS to indicate 28 specified joints assessed for tenderness/swelling. ACR20, a responder variable of ≥20% improvement, measures change over time in RA activity. Both are key efficacy endpoints in RA trials. Biosimilars are structurally highly similar versions of marketed biologics that are supported by comparative analytical testing, pre-clinical/clinical trials to demonstrate similarity to their innovator biologics. The study population and design of Phase 3 clinical trial is chosen based on its sensitivity to detect a difference between potential biosimilar and reference product. A notable recent development is CT-P13 (Remsima™; Inflectra™), a biosimilar of Remicade® (infliximab) and the first biosimilar monoclonal antibody approved by the European Medicines Agency (EMA). As biosimilar drugs and trials become more prevalent, the choice of a margin to demonstrate equivalent efficacy between a biosimilar and its reference remains the most critical element in the design and analysis in biosimilar clinical trials.
Objectives Establish an equivalence margin (EM) for a biosimilar trial in RA that excludes clinically meaningful differences in ACR20 or DAS28.
Methods Following the principle provided in ICH E9, we derived EMs for ACR20 using a meta-analysis (MA) of historical therapeutic effects demonstrated in 5 randomised, placebo-controlled clinical trials of infliximab. This approach is different to the “weighted average of historical data” method utilised in the CT-P13 biosimilar pivotal trial that yielded an EM of ±15%. We present herein a rigorous and conservative method, ie, MA of historical data, to determine EMs following guidelines for non-inferiority studies for new molecular entities. We present a side-by-side comparison between ACR20 and DAS28 to demonstrate different characteristics of the 2 endpoints, including sample size, metric, margin size, and available historic data.
Results MA results suggest that, for 2-treatment comparisons, rate difference (RD) is superior to ratio or odds ratio in terms of heterogeneity and reporting bias. EMs of ±14% to ±12% were obtained. These EMs represent 50% of the lower bound of the 95% confidence interval (CI) for random effects of historical effect size.
Conclusions The difference in ACR20 response rates is more accepted in terms of clinical relevance. MA of historical trial data is appropriate to determine EMs. Relevance of historical trials, heterogeneity, and reporting bias must be carefully assessed to develop EMs for biosimilar trials. Therefore, the derivation of EM, must account for the unique characteristics of the proposed biosimilar/reference product combination. There is support for the use of either the 2-sided 95% vs 90% CI for the observed RD in ACR20 and DAS28 falling within the margin to conclude equivalence. Ultimately, a globally harmonized approach to CI's will simplify equivalence analyses in biosimilar trials.
Disclosure of Interest S. Hua Shareholder of: Pfizer Inc, Employee of: Pfizer Inc, K. Barker Shareholder of: Pfizer Inc, Employee of: Pfizer Inc, P. Qu Employee of: Pfizer Inc, M. Rehman Shareholder of: Pfizer Inc, Employee of: Pfizer Inc, R. Schaum Shareholder of: Pfizer Inc, Employee of: Pfizer Inc, S. Visvalingam Shareholder of: Pfizer Inc, Employee of: Pfizer Inc, J. McClellan Shareholder of: Pfizer Inc, Employee of: Pfizer Inc, S. Salts Shareholder of: Pfizer Inc, Employee of: Pfizer Inc, A. AL-Sabbagh Shareholder of: Pfizer Inc, Employee of: Pfizer Inc