Development and Evaluation of Disease Activity Measures in Rheumatoid Arthritis Using Multi-Level Mixed Modeling and Other Statistical Methodologies: A Dissertation

Remarkable progress has been made in the development of effective treatments for patients with rheumatoid arthritis (RA). To ensure that a patient is optimally responding to treatment, consistent monitoring of disease activity is recommended. Established composite and individual disease activity measures often cannot be computed due to missing laboratory values. Simplified measures that can be calculated without a lab value have been developed and previous studies have validated these new measures, yet differences in their performance compared with established measures remain. Therefore, the goal of my doctoral research was to examine and evaluate disease activity and composite measures to facilitate monitoring of response in clinical care settings and inclusion of patients with missing laboratory values in epidemiological research. In the first study, the validity of two composite measures, the Clinical Disease Activity Index (CDAI) and the Disease Activity Score with 28 joint count (DAS28) was examined and both were significantly associated with a rheumatologist’s decision to change therapy (CDAI OR=1.58; 95% CI: 1.42, 1.76) (DAS28 OR=1.34; 95% CI 1.27,1.56). However, further evaluation using receiver operating characteristic (ROC) analysis found that they were not strong predictors of physician decisions to change therapy (AUC=0.75, 0.76, respectively). Thus, they should not be used to guide treatment decisions in the clinic. Two measures of disease activity, erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) are often not measured and impede the computation of composite measures of disease activity. In the second study, significant factors which may predict the measurement of the ESR and CRP were identified and included physician and clinical variables but no quantitative disease activity measures. Thus the suitability of the ESR and CRP as measures of disease activity is suspect. In the final study, I created a new composite measure, the modified disease activity score with 28 joint count (mDAS28), by replacing the laboratory value in the DAS28. The mDAS28 was then validated by comparing its performance with the DAS28. The measures were strongly correlated (r=0.87), and strong agreement was found between the two measures when categorizing patients to levels of disease activity (ĸ=0.77) and treatment response (ĸ=0.73). Therefore, the mDAS28 could be used in place of the DAS28 when laboratory values needed to compute the DAS28 are missing. In summary, I found that the CDAI and DAS28 were not strong predictors of the rheumatologist’s decision to change therapy. I also found that the variability in the measurement of ESR and CRP was not associated with disease activity. I was able to modify the DAS28 by replacing the laboratory measure and create a new simplified measure, the mDAS28. I also validated the mDAS28 for use in the clinic and in epidemiological research when the DAS28 is unavailable