Giant-cell arteritis is a diagnostic challenge. To determine the diagnostic performance of ultrasonography for giant-cell arteritis. Studies published up to April 2004 in the MEDLINE, EMBASE, and Cochrane databases; reference lists; and direct contact with investigators. Studies in any language that examined temporal artery ultrasonography for diagnosis of giant-cell arteritis, enrolled at least 5 patients, and used biopsy or the American College of Rheumatology (ACR) criteria as the reference standard. Two reviewers independently graded methodologic quality and abstracted data on sensitivity and specificity of ultrasonography for giant-cell arteritis. Diagnostic performance was determined for the halo sign, stenosis, or occlusion and for any of these ultrasonographic abnormalities. Weighted sensitivity and specificity estimates and summary receiver-operating characteristic (ROC) curve analysis were used. Twenty-three studies, involving a total of 2036 patients, met the inclusion criteria. The weighted sensitivity and specificity of the halo sign were 69% (95% CI, 57% to 79%) and 82% (CI, 75% to 87%), respectively, compared with biopsy and 55% (CI, 36% to 73%) and 94% (CI, 82% to 98%), respectively, compared with ACR criteria. Stenosis or occlusion was an almost equally sensitive marker compared with either biopsy (sensitivity, 68% [CI, 49% to 82%]) or ACR criteria (sensitivity, 66% [CI, 32% to 89%]). Consideration of any vessel abnormality nonsignificantly improved diagnostic performance compared with ACR criteria. Between-study heterogeneity was significant, but summary ROC curves were consistent with weighted estimates. When the pretest probability of giant-cell arteritis is 10%, negative results on ultrasonography practically exclude the disease (post-test probability, 2% to 5% for various analyses). The primary studies were small and of modest quality and had considerable heterogeneity. Ultrasonography may be helpful in diagnosing giant-cell arteritis, but cautious interpretation of the test results based on clinical presentation and pretest probability of the disease is imperative.