Scientific Director, Quantitative Pharmacology EMD Serono, Billerica, MA, USA, Massachusetts, United States
Objectives: The idiopathic inflammatory myopathies are a heterogenous group of rare connective tissue diseases with the hallmark clinical manifestation of progressive muscle weakness with subgroups including polymyositis (PM), dermatomyositis (DM) or immune mediated necrotizing myopathy (IMNM). The objectives of this analysis were to develop models describing the disease progression or trajectory reflected by clinical endpoints and to identify predictors of a change in disease activity.
Methods: The data source of the Myositis Registry from the University of Pittsburgh was used to develop the model. After data exploration, two clinical endpoints, physician global activity (PGA) and cutaneous disease activity (CUTNA) visual analog scales were chosen for modeling based on the availability of the longitudinal data and the data trajectories. Different approaches were tested using the original scale values, transformation of integer values for bounded integer model (BIM) and Markov model. Various model functions have been evaluated including linear progression, exponential progression, an inverse Bateman function and a remission/progression model. Covariates of demographics and clinical characteristics were evaluated based on the likelihood ratio test, general linear model and covariate plots. Statistically significant covariates were included in a full model and retained in the final model in a stepwise regression during the back elimination. Visual predictive checks to assess frequency of scores and proportion of data at each score over time and multiple diagnostic approaches including agreement plots were performed to evaluate the model performance.
Results: The PGA model was developed using pooled dataset from PM, DM and IMNM (N=50, 122 and 13, respectively) after determining the trajectories were similar while the CUTNA model was based on DM population only (N=120). The time course of PGA and CUTNA scores over time were best described using a BIM approach, and a function describing remission and progression was best for both endpoints. For the PGA endpoint, positive AJo1 antibody status, higher baseline PGA score, and the presence of heliotrope rash were identified as covariates which were predictive of poorer outcomes than the reference patients. For patients with muscle weakness in the neck there was a greater likelihood of a better outcome than for the reference patient. For the CUTNA endpoint, only covariates on Baseline could be estimated. Heliotrope rash, symmetrical proximal weakness of upper extremities and other dermatomyositis rash were predictive of higher CUTNA score.
Conclusions: Disease progression models have been successfully developed for PGA and CUTNA using the real-world data. Covariates have been identified for each model which could provide valuable information to inform future clinical trial design in myositis patients.
