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

摘要: OBJECTIVE: We have assessed the importance of distinction between classification criteria for rheumatoid arthritis (RA) that recognize presence currently active disease from those that, in addition, incorporate evidence past activity ascertaining occurrence. applied 7 schemes to a population twins with inflammatory determine (a) number individuals classified as RA positive by each scheme and hence effect on estimate concordance (b) their performance correctly assigning diagnosis compared physician's opinion. METHODS: The were 1958 ARA (Rome) which detect disease, 1966 New York (using both 2/4 3/4 published cutoffs) "ever" 4 variants 1987 criteria. These 4/7 decision tree approaches basis relevant features (1) being present at time study (2) ever allowing current joint deformity substitute absent swelling. RESULTS: In all, 283 history swelling assessed, 255 whom considered physician. Criteria used "current" identified only about 70% recognized ever. differences ascertainment level produced marked monozygotic twin estimates percentages ranging 10 18%. results receiver operating curves confirmed assess too insensitive be value. Of status performed best overall. CONCLUSION: use methods well is essential avoid important misclassification epidemiological family studies. criteria, retrospectively swelling, are enhanced value over other existing schemes.