Objectives: The Leiden clinical prediction rule (CPR) was developed in 2007 to predict disease progression in patients with recent-onset undifferentiated arthritis (UA). This systematic review and meta-analysis investigates the predictive ability of the rule at identifying patients who are at a high risk of developing rheumatoid arthritis (RA).
Methods: A systematic review of the literature search was conducted from 2007 to May 2013 to identify studies that validated the rule. This study adhered to the PRISMA guidelines. The methodological quality of studies was assessed using the QUADAS-2 tool. Pooled sensitivity and specificity values for each of the cut points were generated using a bivariate random-effects model. Heterogeneity was assessed using the variance of logit-transformed sensitivity and specificity. Bayes' theorem was used to calculate post-test probability of progression from UA to RA.
Results: The search identified four relevant studies, resulting in six data sets (n = 1084). A cut point of ≥ 9 was identified as the optimal cut point for determining progression to RA. It is associated with a greater pooled specificity (0.99, 95% CI 0.95-1.00) than sensitivity (0.31, 95% CI 0.24-0.37). Using Bayes' theorem, a score of ≥ 9 points increased the pre-test probability from 40.04% to 93.63%. A less stringent cut-off of ≥ 8 also identified a significant proportion of patients at risk of RA who have a high likelihood of progressing to RA (LR + 9.5, 95% CI 6.21-14.54).
Conclusion: A cut point of ≥ 9 offers an optimal estimate for identifying patients with UA who are at a high risk of developing RA and warrant intervention. However, a number of methodological limitations identified across studies suggest that the results should be interpreted cautiously and that further validation of the Leiden CPR is necessary.
Keywords: Clinical prediction rule; Rheumatoid arthritis; Undifferentiated arthritis.
© 2013 Elsevier Inc. All rights reserved.