The clinical utility of a rule for predicting rheumatoid arthritis in patients with early undifferentiated arthritis: Comment on the article by van der Helm-van Mil et al
Version of Record online: 26 FEB 2009
Copyright © 2009 by the American College of Rheumatology
Arthritis & Rheumatism
Volume 60, Issue 3, page 905, March 2009
How to Cite
Pratt, A. G., Isaacs, J. D. and Wilson, G. (2009), The clinical utility of a rule for predicting rheumatoid arthritis in patients with early undifferentiated arthritis: Comment on the article by van der Helm-van Mil et al. Arthritis & Rheumatism, 60: 905. doi: 10.1002/art.24348
- Issue online: 26 FEB 2009
- Version of Record online: 26 FEB 2009
- Arthritis Research Campaign
To the Editor:
We read with great interest the excellent article by van der Helm-van Mil et al, which provided convincing validation of a much-needed prediction rule for disease outcome in patients with recent-onset undifferentiated arthritis (UA) (1), and we have 2 observations.
First, the definition of early UA deserves comment. In our own practice, a patient is said to have UA when, in a consulting rheumatologist's opinion, “an inflammatory arthritis is suspected, with RA remaining a possibility, but established classification criteria for any rheumatologic condition are not fulfilled.” This is akin to the description of “undifferentiated polyarthritis” given by Wolfe et al, which states that it is not explicitly required that objectively observed joint inflammation be present, providing there is a history of inflammatory arthritis, especially in the presence of increased levels of inflammation markers (2). Among our own cohort of 53 patients with UA defined in this way (median symptom duration 14 weeks; followup ≥1 year), 18 (34%) would have been excluded if the 1958 “probable RA” criteria (3) favored by researchers in Leiden, The Netherlands, had been employed. In fact, the UA of 2 of these patients subsequently evolved into rheumatoid arthritis (RA), as defined by the American College of Rheumatology (formerly, the American Rheumatism Association) in its revised criteria (4). Consensus is certainly required in this area, but encouragingly, the prediction rule proposed by van der Helm-van Mil et al performed well in our UA cohort (positive predictive value [PPV] of a score ≥8 100%; negative predictive value [NPV] of a score ≤6 100%), as confirmed by the construction of a receiver operating characteristic (ROC) curve (area under the curve 0.88; SEM 0.048).
The second observation relates to the specific subset of UA patients in which a prediction rule of this kind is likely to be of most value. A previous study by other authors from the Leiden group demonstrated that the presence of circulating autoantibodies to citrullinated peptides (ACPAs) accurately predicted progression to RA among UA patients (PPV 93% over 3 years) (5). Given the heavy weighting (2 of a possible 13 points) attributed to ACPA status in the scoring system proposed by van der Helm-van Mil et al (1), one might anticipate that a “high-risk” score of ≥8 would simply identify ACPA-positive individuals whose progression to RA is already very likely. This is borne out in our own cohort. Of the 15 UA patients whose disease progressed to RA at ≥1 year, all 8 of those assigned “high-risk” prediction scores of ≥8 were ACPA positive. Furthermore, among the remaining 7 patients (in whom the assigned prediction score of 6–<8 was unhelpful), progression to RA was accurately predicted by the presence of ACPAs in an additional 3 patients. Therefore, in our small cohort, the prediction rule is less sensitive than ACPA status alone in identifying UA patients who will have developed RA at ≥1 year (8 of 15 versus 11 of 15).
Interestingly, van der Helm-van Mil et al showed through the construction of ROC curves that the diagnostic performance of their prediction rule was superior to that of autoantibody status considered in isolation. We suggest that this is attributable to its excellent NPV among ACPA-negative UA patients, the group in which prognostic markers are most needed. Of the 41 ACPA-negative UA patients in our cohort, 22 (54%) could be assigned prediction scores of ≤6 at inception, none of whom had progressed to RA at ≥1 year. This of course leaves a significant subset of UA patients for whom neither prediction score nor ACPA status can accurately predict outcome, and 21% (4 of 19) of this subset developed RA in our cohort.
Among our patients, the well-conceived prediction rule proposed by van der Helm-van Mil et al is most valuable as a negative predictor of RA among ACPA-negative UA patients. However, additional biomarkers are needed for appropriate diagnosis and prognosis among these patients, many of whom present at an early arthritis clinic with a challenging set of symptoms.
Dr. Pratt's work was supported by a clinical research fellowship from the Arthritis Research Campaign.
A. G. Pratt BSc, MRCP*, J. D. Isaacs BSc, PhD, FRCP*, G. Wilson RN, DipHE, * Newcastle University, and The Freeman Hospital, Newcastle-upon-Tyne, UK, The Freeman Hospital Newcastle-upon-Tyne, UK.