Anders Thorsted (1, 2), Anubha Gupta (2), Alienor Berges (2), Lena E. Friberg (1), Mats O. Karlsson (1)
(1) Uppsala University, Sweden, (2) GSK, United Kingdom
Introduction: Rheumatoid arthritis (RA) is a chronic immune-mediated disorder of the joints that progress from loss of self-tolerance to chronic synovitis, which ultimately drive joint damage [1]. Otilimab, which has undergone full clinical development [2], is a high-affinity recombinant IgG1 antibody that binds and inhibits soluble granulocyte-macrophage colony stimulating factor.
While several composite endpoints are available to measure RA disease activity and response to intervention, most are based upon a set of core disease activity measures established by the American College of Rheumatology (ACR) [3]. However, it is not straightforward to use the continuous endpoints from dose ranging trials (such as the disease activity score for 28 joints [DAS28]) to inform later registrational trials (which use the binary ACR responder criteria).
Objectives:
- To develop a longitudinal model based on item-response theory (IRT), capable of predicting both DAS28 response and the proportion of ACR20 responders.
- To increase understanding of response to otilimab across underlying ACR components, and the importance of the C-reactive protein (CRP) component on ACR20 response
Methods: Clinical data were available from phase 2a (n=39) [4] and phase 2b (n=222) [5] trials of subcutaneous otilimab in RA patients on a stable background treatment of methotrexate. Observations were available across five dose levels (22.5, 45, 90, 135, or 180 mg) and placebo.
The following dependent variables were included: Free otilimab concentration (drug exposure measure) and CRP in serum (inflammatory measure), and all standard ACR measures (28 and 66/68 swollen and tender joint counts [SJC/TJC], physician’s and patient’s global assessment of arthritis [PhGA, PtGA], patient’s assessment of pain [PtPAIN] and disability [HAQ-DI]).
A two-compartment PK model was used to predict individual otilimab exposure. The longitudinal CRP observations were reduced into integers (to be included in the IRT model), whereas SJC and TJC were split into the 28 joint counts (SJC/TJC) and the remaining 40 TJC and 38 SJC.
The independent occasion approach [6] was used to map all observed scores to underlying latent variables. Binomial models were used for all dependent variables aside from HAQ-DI, for which a two-parameter logit model with increasing difficulty parameters was used.
The individual latent variables were then exponentiated and used as the baseline of an indirect response model, to describe longitudinal changes in RA disease. A disease progression effect (either inhibitory or stimulatory at the individual level) and an inhibitory Emax drug effect (linked to free otilimab concentration) was implemented on Rin (describing ‘production’ of RA disease).
Results: The final IRT model described both composite endpoints and their underlying components well (assessed through visual predictive checks). While a separate latent variable was used for each component (seven in total), high correlations were observed among those that were reported by the patient (HAQ-DI, PtGA, PtPAIN), and those that were assessed by a physician (PhGA, TJC, SJC), whereas no correlations were observed for the CRP response.
In general, ACR20 response to otilimab treatment at Week12 was predicted to be similar for the phase3 dosing regimens of 90 and 150 mg once weekly (55 and 57%, respectively) indicating that both doses were at the maximum of the exposure-response curve. This corresponded well with the contRAst 1 phase3 responses of 55 and 51%, respectively.
Response in individual components was predicted to be highest for the PhGA, TJC and SJC components (with ~80% of individuals having a 20% decrease at Week12), while lower response rates were observed for HAQ-DI (~55%), PtGA (~65%), PtPAIN (~66%), and CRP (~60%). Disease progression, which drove improvement across all components except CRP, was also highest for PhGA, TJC and SJC (~45%). Nullifying the drug effect on the CRP component in the derivation of ACR20 response did not have a major impact on the predicted ACR20 response rate (54 and 56%, respectively).
Conclusion: An IRT model was developed to describe clinical efficacy measures in RA clinical trials, and was capable of describing response to otilimab treatment across both underlying components and composite endpoints. The model has the potential to be applied in clinical trial design and in the translation of phase2 to phase3 clinical endpoints.
References:
[1] Weyand CM and Goronzy JJ. The immunology of Rheumatoid Arthirtis. Nat Immunol 2021;22(1):10-18.
[2] Fleischmann RM, van der Heijde D, Strand V, Atsumi T, McInnes IB, Takeuchi T, et al. Anti-GM-CSF otilimab versus tofacitinib or placebo in patients with active rheumatoid arthritis and an inadequate response to conventional or biologic DMARDs: two phase3 randomized trials (contrast 1 and contRAst 2). Ann Rheum Dis. 2023;82(12):1516-1526.
[3] Felson DT, Anderson JJ, Boes M, Bombardier C, Chernoff M, Fried B, et al. The American College of Rheumatology preliminary core set of disease activity measures for rheumatoid arthritis clinical trials. The Committee on Outcome Measures in Rheumatoid Arthritis Clinical Trials. Arthritis Rheum. 1993;36(6):729-40.
[4] Genovese MC, Berkowitz M, Conaghan PG, Petefy C, Davy K, Fisheleva E, et al. MRI of the joint and evaluation of the granulocyte-macrophage colony-stimulating factor-CCL17 axis in patients with rheumatoid arthritis receiving otilimab: a phase 2a randomised mechanistic study. Lancet Rheumatol. 2020;2(11):e666-e676.
[5] Buckley CD, Simón-Campos JA, Zhdan V, Becker B, Davy K, Fisheleva E, et al. Efficacy, patient-reported outcomes, and safety of the anti-granulocyte macrophage colony-stimulating factor antibody otilimab (GSK3196165) in patients with rheumatoid arthritis: a randomised, phase 2b, dose-ranging study. Lancet Rheumatol. 2020;2(11):e677–e688
[6] Schindler E, Friberg LE, Lum BL, Wang B, Quartino A, Li C, et al. A pharmacometric analysis of patient-reported outcomes in breast cancer patients through item response theory. Pharm Res. 2018;35:122.
COI:
A.T.’s postdoctoral employment was financed by GSK through a research grant to Uppsala University. A.T., A.G., and A.B. are employees of GSK and hold GSK shares. M.O.K. and L.E.F. declare no competing financial interests.
Reference: PAGE 32 (2024) Abstr 11144 [www.page-meeting.org/?abstract=11144]
Poster: Drug/Disease Modelling - Other Topics