A methodological appraisal of the impact of different classification procedures used in three different phases of the australian rheumatoid arthritis twin survey.

Abstract

To compare methodological aspects of the impact of different classification procedures used in three phases of a twin study examining genetic factors in the aetiopathogenesis of rheumatoid arthritis (RA). We have previously reported the results of a study of the aetiopathogenesis of RA based on the Australian Twin Registry (ATR). In the original 258 pairs self-reporting a diagnosis of RA in twin, co-twin or both, a very high false positive self-reporting rate for RA was noted (Phase 1). Subsequent diagnostic information obtained by a disease-specific questionnaire, followed by telephone interviews with subjects and review of information obtained by mail and telephone interview from the patient's general practitioner or musculoskeletal specialist, identified 23 'true' RA pairs (Phase 2). Pairwise concordance percentages for RA based on those 20 discordant and 3 concordant pairs were as follows: monozygotic (MZ) 21% (95% confidence interval (CI)=6-44%), dizygotic (DZ) 0% (95% CI=0-25%) (probandwise concordance MZ 35% (8.9-67.3), DZ 0% (0-50.3)). Given the potential effects of misclassification on data interpretation, we have further pursued the accuracy of diagnosis by a systematic clinical, serological and radiographical evaluation of the 23 RA pairs (Phase 3). In only one instance did more intense diagnostic investigation of the 23 pairs result in recategorization. The probandwise concordance percentages were recalculated: MZ=37.5%, DZ=0%. Our original contention that genetic factors play some part in the aetiopathogenesis of RA, but do not account entirely for its determination, has been substantiated at a higher level of confidence and at almost identical levels of concordance.