A VALIDATION OF CURRENT CLASSIFICATION CRITERIA FOR THE DIAGNOSIS OF PSORIATIC ARTHRITIS – PRELIMINARY RESULTS OF THE CASPAR STUDY

Background: Of several proposed classification criteria for PsA, none have been adequately validated.Objectives: To validate and compare existing classification criteria for PsA in a large international cohort.Methods: Data was collected prospectively from 589 consecutive clinic attendees with PsA and 535 control patients (next clinic attendee with inflammatory arthritis), from 29 rheumatology clinics in 12 countries. The diagnosis was based upon the physician''s opinion and verified by an examination of randomly selected case-record forms by the data quality committee. Data included clinical, laboratory and plain radiographic findings of the spine, pelvis, hands and feet. Clinical features were deemed to be positive if they were currently or ever present. Subjects were classified by each of 7 existing criteria (Vasey & Espinoza, Bennett, Gladman, McGonagle, Moll & Wright, ESSG, and Fournie). Sensitivity and specificity were compared using conditional logistic regression analysis (referent to Vasey). Latent class analysis was used to check the test performance, without knowledge of the physician diagnoses. The performance of individual variables was described in terms of diagnostic odds ratio (the ratio of the odds of a positive test in cases compared to the odds of a positive test in controls). A DOR of 1 indicates that the test does not discriminate between disease and non-disease.Results: Controls had rheumatoid arthritis (n=383), ankylosing spondylitis (n=72), undifferentiated arthritis (n=38), connective tissue disorders (n=14) or other disease (n=28). The specificity of each method was high; statistically, the method of Bennett (1.00) was more specific than Vasey (0.99), which was similar to Gladman (0.98) and Moll (0.99), and more specific than McGonagle (0.93) and ESSG (0.94). The sensitivity was more variable; statistically the method of Vasey (0.96) was similar to McGonagle (0.98) and more sensitive than Bennett (0.48), Moll (0.91), ESSG (0.56) and Gladman (0.90). LCA confirmed similar test performance characteristics. The method of Fournie (sensitivity 0.99, specificity 0.74) was not included in this comparison since 52% of subjects could not be classified by this method. The items with the highest DOR were features of psoriasis itself (DOR of 581 for history of psoriasis and 328 for psoriasis currently present), including nail dystrophy (59) and family history of psoriasis (9.2) and features characteristic of PsA such as dactylitis (19.4), DIP involvement (6.4), radiographic juxta-articular new bone formation (4.8), radiographic interphalangeal ankylosis (4.4), unilateral radiographic sacroiliitis (3.4) and clinical enthesitis (3.8 for any tender enthesis, 2.7 for heel pain). Absence of rheumatoid factor (29.4) or subcutaneous nodules (109) continue to be associated with PsA. Radiographic ray involvement was moderately discriminatory (4.6).Conclusion: In this preliminary analysis, the criteria-set of Vasey and Espinoza appears to be the most accurate. It is possible that it could be improved by incorporating other features with high diagnostic odds ratios.Citation: Ann Rheum Dis, volume 64, supplement III, year 2005, page 107Session: Advances in psoriatic arthritis