Color Doppler is modestly accurate compared with biopsy in patients with suspected giant cell arteritis

Reference

Monti S, Floris A, Ponte CB, et al. The proposed role of ultrasound in the management of giant cell arteritis in routine clinical practice. Rheumatology (Oxford) 2018;57(1):112-119.

Synopsis

These authors evaluated 293 patients with suspected GCA who were referred to their rheumatology clinic at Oxford. Each patient underwent a standard clinical assessment and color Doppler sonography. Although this study was not funded, 3 of the authors have significant relationships with industry. The sonographer was unaware of the patients' clinical history. The authors, however, included clinical parameters, laboratory parameters, and the results of the Doppler to establish a clinical diagnosis of GCA. The use of Doppler as part of the clinical diagnosis, also called incorporation bias, will overestimate the diagnostic accuracy of the included test. Only 55 of the patients were "fresh" (ie, had new onset of symptoms and less than 7 days of glucocorticoids before the consultation). The remainder were patients following up for an established diagnosis of GCA, patients who were suspected of having a flare up of their GCA, or those who had no definitive diagnosis but had been taking steroids for more than a week. Temporal artery biopsy was performed in only 31 of the new patients. The authors point out that over time, the number of patients undergoing biopsy dropped from approximately 40% to 25%. Although it is understandable to try to decrease the potential harms from invasive tests, this creates a problem of verification bias: the estimate for sensitivity is too high and the estimate for specificity is too low. Among the 31 patients with biopsies, only 5 had a confirmed diagnosis of GCA. The authors estimate that Doppler was modestly sensitive (63.3%; 95% CI 44 - 80) and highly specific (100%; 83 - 100). So, here we have conclusions that may be true, but the study's incorporation bias and verification bias make the estimates of test accuracy unreliable.