A comparison of the performance of different methods of disease classification for rheumatoid arthritis. Results of an analysis from a nationwide twin study.

Objective: We have assessed the importance of the distinction between classification criteria for rheumatoid arthritis (RA) that recognize the presence of currently active disease from those that, in addition, incorporate evidence of past disease activity in ascertaining disease occurrence. We applied 7 classification schemes to a population of twins with inflammatory arthritis to determine (a) the number of individuals classified as RA positive by each scheme and hence the effect on the estimate of disease concordance in the twins and (b) their performance in correctly assigning a diagnosis compared with a physician's opinion.

Methods: The schemes assessed were the 1958 ARA (Rome) criteria which detect active disease, the 1966 New York (using both the 2/4 and 3/4 published cutoffs) which detect "ever" disease and 4 variants of the 1987 ARA criteria. These were the 4/7 and decision tree approaches applied on the basis of the relevant features (1) being present at the time of the study and (2) being present ever and allowing current joint deformity to substitute for absent joint swelling.

Results: In all, 283 individuals with a history of joint swelling were assessed, 255 of whom were considered to have RA by their physician. Criteria used to recognize "current" RA identified only about 70% of those which recognized RA ever. These differences in ascertainment level produced a marked effect on the monozygotic twin RA concordance estimates with percentages ranging from 10 to 18%. The results from receiver operating curves confirmed that criteria used to assess only current RA were too insensitive to be of value. Of the criteria that recognized RA status ever the 1987 ARA performed best overall.

Conclusion: The use of classification methods that incorporate past as well as current evidence of disease activity is essential to avoid important misclassification in epidemiological and family studies. The 1987 criteria, applied retrospectively and allowing joint deformity to substitute for swelling, are of enhanced value over other existing schemes.