Latent variable indirect response modeling of continuous and categorical clinical endpoints
C. Hu(1), P. Szapary(1), A. Mendelsohn (1), H. Zhou(1)
(1)Janssen Research & Development LLC, Spring House, PA, USA
Objectives: Parsimonious predictive exposure-response modelling is important in clinical drug development. This study aims to introduce a general latent variable representation of Indirect Response (IDR) models, and apply it to simultaneously model the efficacy endpoints of Psoriasis Area and Severity Index (PASI) scores and the 20%, 50%, and 70% improvement in the American College of Rheumatology disease severity criteria (ACR20/50/70), in psoriatic arthritis (PsA) patients treated with ustekinumab.
Methods: A general approach of using a latent-variable representation of IDR models in a format of change from baseline for clinical endpoints is developed from a continuous underlying process. The approach extends to general link functions that cover logistic/probit regression. Placebo effect parameters in the new representation are more readily interpretable and can be separately estimated from placebo data, thus allowing convenient and robust model estimation. When applying to ordered categorical endpoints, the approach allows the testing of baseline constraint that the probability of achieving endpoint equals zero. Inherent connections with baseline-normalized standard IDR models are derived. This approach was applied to data through the primary endpoint (week 24) from two phase III clinical trials of subcutaneously administered ustekinumab for the treatment of PsA, where PASI scores and ACR20/50/70 were jointly modelled with accounting of their correlations.
Results: An earlier approach using latent variable IDR [1,2] to model clinical endpoints is shown to be equivalent to a baseline-normalized representation [3,4]. This is used to further prove an equivalence between Type I/III IDR models for clinical endpoints. The equivalence properties hold under general link functions that cover logistic and probit regression or continuous clinical endpoint modelling. In the current application, 925 PsA patients contributed to nearly 3,000 ustekinumab serum concentration measurements, 5,000 ACR and 3,400 PASI scores. External validation showed reasonable parameter estimation precision and model performance.
Conclusion: The latent-variable IDR model representation provides a parsimonious approach for predictive modelling of clinical endpoints. In this framework, Type I and Type III IDR models are shown to be equivalent, therefore there are only three identifiable IDR models. The joint model could be used to predict both psoriasis and arthritis components.
References:
[1] Hu C, Szapary PO, Yeilding N, Zhou H (2011). Informative dropout modeling of longitudinal ordered categorical data and model validation: application to exposure-response modeling of physician's global assessment score for ustekinumab in patients with psoriasis. J Pharmacokinet Pharmacodyn 38(2):237-260.
[2] Hu C, Yeilding N, Davis HM, Zhou H (2011). Bounded outcome score modeling: application to treating psoriasis with ustekinumab. J Pharmacokinet Pharmacodyn 38(4):497-517.
[3] Woo S, Pawaskar D, Jusko WJ (2009). Methods of utilizing baseline values for indirect response models. J Pharmacokinet Pharmacodyn 40(1):81-91.
[4] Hu C, Xu Z, Mendelsohn A, Zhou H (2013). Latent variable indirect response modeling of categorical endpoints representing change from baseline. J Pharmacokinet Pharmacodyn 38(2):237-260.
