Model-based evaluation of DAS28 as a potential surrogate for ACR20 to establish the dose-response relationship for disease modifying anti-rheumatic drugs. A case study using tasocitinib (CP-690,550), an oral JAK inhibitor.
Yoon Jung Lee (1), Sriram Krishnaswami (2), Lutz Harnisch (3), Jonathan French (2), Kyungsoo Park (1)
(1) Department of Pharmacology, College of Medicine, Yonsei University, Seoul, Korea (2) Pfizer Global R&D, New London, CT, USA (3) Pfizer Global R&D, Sandwich, UK
Background: Tasocitinib (CP-690,550), a selective inhibitor of the Janus kinase (JAK) family, is being developed for the treatment of several autoimmune diseases such as rheumatoid arthritis (RA) and prevention of allograft rejection. The primary clinical endpoint for the response of RA to treatment is American College of Rheumatology response criterion of 20% improvement (ACR20) by which a patient is considered a responder or non-responder. Disease Active Score measured by 28 tender and swollen joint assessments (DAS28-3(CRP)), developed as a clinical index of RA to objectively evaluate a patient's response, has a continuous scale ranging from 0 to 9.4, reflecting the extent of underlying inflammation [1], [2]. Using DAS28 as a simplified notation for DAS28-3(CRP), the modeling hypothesis is that the continuous DAS28 may provide more precise dose-response curves than the binary ACR20, thus potentially improving the efficiency of dose response studies with new therapies.
Objectives: This study assesses the feasibility of using DAS28 as a more efficient measure than ACR20 for establishing dose-response of tasocitinib.
Methods: Data were obtained from a parallel, randomized, double-blinded, and placebo-controlled phase 2a study of CP-690,550 [3]. This 6-week study consisted of 4 dose arms, 0 (placebo), 5, 15 and 30mg bid, given orally in 252 patients with RA. Drug effects were assessed by ACR20 and DAS28 at baseline and at weeks 1, 2, 4, and 6, with ACR20 measured as a binary and DAS28 as a continuous variable. Both types of data were analyzed using NONMEM 7 within a mixed-effect model framework. For ACR20, using a logistic regression, the logit of the probability of ACR20 response was modeled as a sum of the latent variable threshold, placebo effect, and "pure" drug effect [4], [5]. Similarly for DAS28, the observed score was modeled as a sum of baseline, placebo effect, and pure drug effect. Model parameters were estimated sequentially in two steps as this approach estimated model parameters better than the simultaneous estimation of placebo and drug parameters. At the first step, placebo parameters were estimated from placebo group data only. Then, drug parameters were estimated from treatment group data by fixing placebo parameters at their estimates obtained at the first step. Dose-response information derived from DAS28 and ACR20 were compared using a time-averaged standard error of the model prediction normalized by the estimate (TASE, %). In addition, the percent change in DAS28 from baseline (DDAS) translatable to 30% difference from placebo for ACR20 (DACR) was estimated as follows. In Step 1, P30, the probability of DACR > 30%, was computed using the normal approximation. In Step 2, DDAS satisfying Prob(t ≥ Tx) = P30 was computed using the t-distribution, with Tx being the t-statistic corresponding to a certain value of DDAS.
Results: The placebo effect was best described by a monotone increasing exponential function for ACR20 and an inverse Bateman function for DAS28. The drug effect was described as an inhibitory indirect response model to account for an inhibitory drug effect on the inflammatory RA process. Overall, the model represented the data well. For doses of 0, 5, 15 and 30 mg, TASE (%) estimates were 26.5, 12.1, 8.3 and 7.2% for the probability of being a responder according to ACR20, and 7.5, 4.8, 4.4 and 4.4% for DDAS, indicating ACR20 yields more variation and wider confidence intervals at all doses as compared to DAS28. The estimated DACR and DDAS increased with dose, ranging from 21 to 54% and from 16 to 41%, respectively, with DDAS being smaller than DACR at the same dose level. DDAS translatable to 30% of DACR was estimated to be 22%, which was obtained by regressing the estimates of DACR from all weeks and doses on those of DDAS.
Conclusion: These findings support that a continuous DAS28 endpoint may provide more precise dose-response curve estimates than the binary ACR20 response measure, thus offering the potential of achieving a similar level of precision in the effect size of clinically useful doses at lower sample sizes. To validate the above results, further analyses will be necessary, which will include the assessment of DAS28 for data from other studies with the same agent and other DMARDs in the same or different class. Further work would be to assess not just the precision but also the accuracy (reliability) of predicting doses that will ultimately be evaluated in Phase 3 studies on the ACR scale. In addition, covariate effects that might influence DAS28 variation will also be assessed.
References:
[1] http://www.das-score.nl/www.das-score.nl/DAS_CRP.html
[2] Fransen J, van Riel PLCM. The disease activity score and the EULAR response criteria. Clin Exp Rheumatol (2005) 23(suppl.39): S93-S99.
[3] Kremer JM, Bloom BJ, Breedveld FC, Coombs JH, Fletcher MP, Gruben D, Krishnaswami S, Burgos-Vargas R, Wilkinson B, Zerbini CA, Zwillich SH. The safety and efficacy of a JAK inhibitor in patients with active rheumatoid arthritis: Results of a double-blind, placebo-controlled phase IIa trial of three dosage levels of CP-690,550 versus placebo. Arthritis Rheum. 2009 Jul;60(7):1895-905.
[4] Hutmacher MM, Krishnaswami S, Kowalski KG. Exposure-response modeling using latent variables for the efficacy of a JAK3 inhibitor administered to rheumatoid arthritis patients. J Pharmacokinet Pharmacodyn (2008) 35:139-157.
[5] Park K, Verotta D, Gupta SK, Sheiner LB. Use of a pharmacokinetic/ pharmacodynamic model to design an optimal dose input profile. J. Pharmacokinet Pharmacodyn (1998) 26: 471-492.