Using predicted disease outcome to provide differentiated treatment of early rheumatoid arthritis

To determine the usefulness of a prediction model for making treatment decisions in early rheumatoid arthritis (RA).

In 152 patients with early RA, progression of radiological damage during the first year [Sharp-van der Heijde (SH) score > 0] was assessed and used to define actual disease outcome. Available variables at baseline were entered in a multivariate regression analysis with progression score as dependent variable. This model was used to predict disease outcome in every patient. Using the standard deviations of the predicted disease outcome, patients were divided into 3 groups: (1) severe disease: high probability (> or = 0.8) for progression > 0, (2) mild disease: high probability (> or =0.8) for progression < or = 0, and (3) not classified: no high probability for either option. It was determined how many patients could be classified by using this model.

One hundred nine patients (71.7%) showed joint damage progression during the first year. Baseline variables available were: age, sex, duration of symptoms, duration of morning stiffness, patient's global assessment of disease activity, Health Assessment Questionnaire score, swollen and painful joint count, bilateral compression pain in metatarsophalangeals, rheumatoid factor positivity, erythrocyte sedimentation rate, shared epitope positivity, SH-score, and the presence of erosions. The R2 value (approximately variation explained) of the prediction model was 0.36. By using this model 46.3% of patients could be classified as having severe disease, 0% as having mild disease, and 53.7% could not be classified.

To be able to make treatment decisions in early RA based on predicted disease outcome, a better prediction of disease outcome is needed, making the search for better prognostic variables urgent.