Background Olokizumab (OKZ) is an anti-interleukin-6 (IL-6) monoclonal antibody in development for the treatment of rheumatoid arthritis (RA). In phase 2 studies, treatment with OKZ+MTX resulted in significant improvements in disease activity compared to placebo (PBO)+MTX.1,2 Here, we used dose-response modeling based on efficacy and safety data from phase 2 to determine the most appropriate OKZ doses for phase 3 development. This modeling approach utilizes all available data, and ensures that the decision is scientifically-based.
Objectives To test the dose-response relationship of OKZ with clinical efficacy and safety, and identify the optimal doses for phase 3.
Methods The nonparametric Jonkheere-Terpstra (JT) test was fitted to DAS28(CRP) data obtained from two double-blind, randomized, phase 2 dose-ranging (60 to 480mg 4-week cumulative dose [cd]) studies (RA0056 [NCT01242488] and RA0083 [NCT01463059]) to test for a monotonic dose-response curve without any assumptions on curve shape. The JT tests indicated that a dose-response model was statistically significant, allowing the Hill parametric model to be used, with study as a categorical covariate that affected both the ED50 and Emax parameters. Mean and 95% confidence intervals for the response were calculated. The Hill model was used since it fit the data adequately, and allowed estimation of the response curves using all available information. To select an appropriate dose, safety as well as efficacy must be included, and a balance between the two achieved. Thus, a dose-response model for the incidence of adverse events (AEs) was also required. A logistic regression model was fitted to the 17 most commonly observed preferred terms across pooled phase 2 data.
Results The JT test indicated a statistically significant monotonic dose-response relationship: p<0.005 for all dosing regimens (every 2 weeks [wks; Q2W] and every 4 wks [Q4W]) in each study, and across pooled data from the two studies. Categorical DAS28(CRP) endpoints also indicated a monotonic dose-response relationship (Table). A parametric dose-response curve fitted using a Hill model3 to combined data from the studies (Figure) adequately described the data, with an inflection point of 120mg 4-wk cd. This is the smallest dose that attains a response near the maximum. Dosing frequency had a small, non-statistically significant effect on categorical DAS28(CRP) response.
Logistic regression analysis of AEs did not indicate any evidence of a dose-related increase in AEs examined vs PBO+MTX, with the exception of injection site reaction.
Conclusions Efficacy increased with dose, without an increase in safety events, and plateaued at approximately 120mg cd. Modeling data were robust, with comparable results for individual studies vs pooled data, and continuous vs categorical responses. Thus a dose at the inflection point (128mg 4-wk cd given as 64mg Q2W) and a dose below the inflection point (64mg 4-wk cd) were selected for inclusion in the phase 3 program.
Genovese M. Ann Rheum Dis 2014;73(9):1607–15;
Takeuchi T. Mod Rheum 2015;10.3109/14397595.2015.1074648;
Reeve R. 2013. J Biopharm Stat 2013;23(3):648–61
Acknowledgement The authors acknowledge Costello Medical Consulting and ELM Group for medical writing and editorial assistance, funded by R-Pharm CJSC.
Disclosure of Interest R. Reeve Employee of: Quintiles Inc, A. Krotkova Employee of: R-Pharm CJSC, D. Weilert Employee of: Quintiles Inc, S. Fatenejad Consultant for: R-Pharm