A Systematic Evaluation of Single Metrics in Discriminating Changes in Rheumatoid Arthritis Disease Activity

Background: The treat-to-target approach in the management of rheumatoid arthritis involves frequent assessments of a patient’s disease activity to guide drug dose and regimen adjustments until a pre-defined low disease activity state and/or remission has been achieved. However, which metric (single or a composite) that best reflects response to therapy and a change in disease activity, or a lack thereof, is yet to be systematically examined.
Aims: To evaluate the ability of clinically used single rheumatoid arthritis metrics in discriminating disease activity changes in response to therapy using population-modelling methods.
Methods: A total of 11 metrics of rheumatoid arthritis disease activity (tender and swollen joint counts, acute phase reactants, and global health, pain and physical function assessments) were obtained from 203 early rheumatoid arthritis patients who attended the Royal Adelaide Hospital. Participants were initiated with combination DMARDs according to a treat-to-target approach with pre-defined triggers for treatment intensification, and were followed up for approximately 2 years. Population models describing the population typical value and the variability for both baseline (pre-treatment) scores and the magnitude of score change from baseline to a single “treated” state (patient’s latest clinic visit at least 26 weeks after the initiation of therapy) were developed for each metric using NONMEM®. Various base and covariate models were tested to describe the distribution of scores at baseline and the treated state in the population and subpopulations. Metrics were ranked by their discriminatory capacity based on plotting their estimates for change between baseline and treated states against its variability, and calculating their proximity to a point representing the largest change and lowest variability.
Results: All models incorporated a single estimable parameter (normally distributed in the population) additive to baseline scores to describe the change between states. The effect of the physician’s decision to administer intra-articular or intra-muscular corticosteroids significantly improved the fit (P < 0.0001 as determined by the likelihood ratio test) of all models and was associated with higher disease activity (either at baseline or the treated state). Swollen joint counts and the physician’s assessment of the patient’s global health demonstrated a greater ability to discriminate changes in rheumatoid arthritis disease activity than others. Non-specific markers of inflammation such as ESR and CRP were shown to be less discriminatory.
Conclusion: Understanding the discriminatory capacity of single disease activity metrics is the first step in evaluating and developing a new composite index for the treat-to-target approach in rheumatoid arthritis.