Introduction: There is a need for tools that are easy to use in clinical practice supporting decision making upon treatment in early rheumatoid arthritis (RA). Aim: The aim was to identify components of prognosticators in early RA and to identify individual patients with a poor prognosis as early as possible.
Methods: Two cohorts from the Swedish TIRA project including 320+408 patients with recent onset RA were included in the study. Disease activity was measured by C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), and the 28-joint count disease activity score (DAS-28), and by the physicians’ global assessment of disease activity (PGA). Disability was assessed as activity limitation by the Swedish version of the Health Assessment Questionnaire (HAQ) and impairment was reported by pain on a visual analogue scale of 0–100 mm. Serological markers were rheumatoid factor (RF) and anti-CCP. RF was measured at the time for diagnosis, and anti-CCP at the time of diagnosis or at one or some of the follow-ups. If at least one anti-CCP test was positive, the patient was judged to be anti-CCP-positive. Assuming different clinical practice in the different cohorts, two different treatment strategies were assumed based on clinical practice in real-world settings. Principal Component Analysis and Multiple Linear Regression Analysis were used to identify prognosticators. Prediction rules were identified by data-driven approach, controlling for different treatment strategies.
Results: Progression of disease and disability measures and inflammation measures the first three months after inclusion predicted a considerable part of DAS-28 at the 1-year follow-up. Serological markers had a larger explanatory power for men than for women. Anti-CCP was a significant predictor for men, but not for women. Two versions of rules, one for women and one for men, predicting good or poor prognosis at one year after inclusion were produced by using measures of disability (Health Assessment Questionnaire), DAS-28, relative change in DAS-28 during first three months, sex, and test of anti-CCP. The rules demanded high prognostic specificity but the prognostic sensitivity was moderate.
Conclusion: A considerable part of DAS-28 at one year after inclusion could be explained by the first 3 months’ progression of disease, disability and inflammation. Anti-CCP was predictive for men but not for women, and needs further investigation. A decision tree predicting poor prognosis among individual early RA-patients showed high specificity and moderate sensitivity on a validationcohort. The medical informatics approach used, controlling for different treatment strategies, yields promising results and further studies will control for more specific differences in treatment strategies, e.g. different DMARDs initiated.