Van der Helm-van Mil AH et al. (2008) Validation of a prediction rule for disease outcome in patients with recent-onset undifferentiated arthritis: moving toward individualized treatment decision-making. Arthritis Rheum 58: 2241–2247

Van der Helm-van Mil AH et al. previously developed an algorithm that scored baseline laboratory, demographic and clinical findings to predict whether patients with recent-onset undifferentiated arthritis (UA) would go on to develop rheumatoid arthritis (RA). These same authors have now validated the algorithm and report accurate prediction of progression in 75% of patients with recent-onset UA.

Data from three cohorts of patients with recent-onset UA from the UK (n = 99), Germany (n = 155) and The Netherlands (n = 34) were used to validate the prediction rule. The original prediction rule included the parameter of baseline severity of morning stiffness; however, these data were only available in one cohort, so for the validation study the rule was rederived to include the duration (in minutes) of morning stiffness instead. A score of ≤6 correctly predicted that UA would not progress to RA in 83%, 83% and 86% of patients in the three cohorts, while a score ≥8 correctly predicted progression in 100%, 93% and 100% of patients.

Duration of morning stiffness was a less powerful predictor than its severity, and the rederived rule consequently had a slightly reduced diagnostic performance. Nonetheless, the accurate prediction of the development of RA in independent cohorts from different countries strongly supports the validity of the rule. A total of 25% of patients were in the intermediate group (scores between 6 and 8) for which no satisfactory estimation of risk could be made—emphasizing the need to identify additional predictive biomarkers in these patients.