Sensitivity and specificity of different diagnostic criteria for Behçet's disease according to the latent class approach.

In this study, we estimated the sensitivity and specificity of different criteria (International, Japanese, Mason & Barnes, O'Duffy, James and clinical judgement) for Behçet's Disease using the latent class approach. Thirty-two consecutive Behçet's Disease patients who were initially diagnosed according to clinical judgement, and 56 control patients with other rheumatic diseases, were recruited from the rheumatology out-patient clinic. The patients had their charts reviewed, were interviewed and clinically examined. Various models including different combinations of the six diagnostic criteria were tested. In the model that included all the information gathered, the International (sensitivity = 0.95, specificity = 1.00), Japanese (sensitivity = 1.00, specificity - 0.95) and Mason & Barnes criteria (sensitivity = 1.00, specificity = 0.94) were the most accurate. Clinician judgement also performed very well in identifying a Behçet's Disease case (sensitivity = 1.00, specificity = 0.85). The International, Japanese and Mason & Barnes criteria were the most accurate. The latent class method led to these conclusions without making any initial assumptions about the accuracy of any of the competing criteria (including clinical judgement).