Objective: To evaluate the sensitivity, specificity, predictive value, interrater reliability and intrarater reliability of using chest radiographs as a screening tool for the diagnosis of diffuse idiopathic skeletal hyperostosis (DISH).
Methods: After review of thoracic spine radiographs by 2 "gold standard" physicians, 45 patients with DISH meeting the criteria of Resnick and Niwayama were contrasted with 106 control patients consisting of 45 with thoracic spondylosis, 45 who lacked spondylosis and whose thoracic spine radiographs were otherwise normal for the age of the patient, and 16 with ankylosing spondylitis. The chest radiographs on the 151 subjects were placed in random order and read independently using an ordinal diagnostic certainty scale by 2 "test" radiologists, experienced in reading bone radiographs. Two months later the order of chest radiographs was rerandomized and the films reassessed by the same test radiologists.
Results: The averages for the diagnostic and population test characteristics were sensitivity = 77%; specificity = 97%; positive predictive value = 91%; and, negative predictive value = 91%. The area under the receiver operating characteristic curves was 0.975 and 0.976 for the radiologists, and kappa was 0.93, demonstrating that interrater reliability was high. On rereading the chest radiographs, intrarater reliability was exceptional (weighted kappa of 0.90 and 0.96 for the two test radiologists). DISH patients whose chest radiographs were read as not demonstrating DISH had significantly less extensive disease.
Conclusions: We conclude that chest radiographs are a reliable and valid screening tool for the diagnosis of DISH.