The conclusions made by Shmerling and Delbanco1 regarding the utility of rheumatoid factor (RF) in the diagnosis of rheumatoid arthritis (RA) came as a bit of a surprise, seemingly reversing conventional thinking. Although we can agree, based on their data, that RF is, at times, ordered inappropriately (ie, in the absence of joint complaints), we find it difficult to accept the study's estimates for sensitivity (28%), specificity (87%), and positive predictive value (24%). Kelley et al2 quote 70% sensitivity of the RF for RA, and Wolfe et al3 quote 81.6%. We examined the study by Shmerling and Delbanco to locate possible explanations for this discrepancy.
The gold standard for the diagnosis of RA in the study was the use of chart review, applying the American Rheumatism Association's remaining six guidelines (excluding the RF).1 By this method, they claim to have discovered 10 seronegative RA patients out