Background Giant cell arteritis (GCA) is a systemic, inflammatory vasculitis primarily affecting people over the age of 50 years. GCA is treated as a medical emergency due to the potential for sudden, irreversible visual loss. Temporal artery biopsy (TAB) is one of the five criteria of the American College of Rheumatology (ACR) 1990 classification, which is used to aid the diagnosis of GCA. TAB is an invasive test, and it can be slow to obtain a result due to delays in performing the procedure and the time taken for histopathologic assessment. Temporal artery ultrasonography (US) has been demonstrated to show findings in people with GCA such as the halo sign (a hypoechoic circumferential wall thickening due to oedema), stenosis or occlusion that can help to confirm a diagnosis more swiftly and less invasively, but requiring more subjective interpretation. This review will help to determine the role of these investigations in clinical practice. Objectives To evaluate the sensitivity and specificity of the halo sign on temporal artery US, using the ACR 1990 classification as a reference standard, to investigate whether US could be used as triage for TAB. To compare the accuracy of US with TAB in the subset of paired studies that have obtained both tests on the same patients, to investigate whether it could replace TAB as one of the criteria in the ACR 1990 classification. Search methods We used standard Cochrane search methods for diagnostic accuracy. The date of the search was 13 September 2022. Selection criteria We included all participants with clinically suspected GCA who were investigated for the presence of the halo sign on temporal artery US, using the ACR 1990 criteria as a reference standard. We included studies with participants with a prior diagnosis of polymyalgia rheumatica. We excluded studies if participants had had two or more weeks of steroid treatment prior to the investigations. We also included any comparative test accuracy studies of the halo sign on temporal artery US versus TAB, with use of the 1990 ACR diagnostic criteria as a reference standard. Although we have chosen to use this classification for the purpose of the meta-analysis, we accept that it incorporates unavoidable incorporation bias, as TAB is itself one of the five criteria. This increases the specificity of TAB, making it difficult to compare with US. We excluded case-control studies, as they overestimate accuracy, as well as case series in which all participants had a prior diagnosis of GCA, as they can only address sensitivity and not specificity. Data collection and analysis Two review authors independently assessed the studies for inclusion in the review. They extracted data using a standardised data collection form and employed the QUADAS-2 tool to assess methodological quality. As not enough studies reported data at our prespecified halo threshold of 0.3 mm, we fitted hierarchical summary receiver operating characteristic (ROC) models to estimate US sensitivity and also to compare US with TAB. We graded the certainty of the evidence using the GRADE approach. Main results Temporal artery ultrasound was investigated in 15 studies (617 participants with GCA out of 1479, 41.7%), with sample sizes ranging from 20 to 381 participants (median 69). There was wide variation in sensitivity with a median value of 0.78 (interquartile range (IQR) 0.45 to 0.83; range 0.03 to 1.00), while specificity was fair to good in most studies with a median value of 0.91 (IQR 0.78 to 1.00; range 0.40 to 1.00) and four studies with a specificity of 1.00. The hierarchical summary receiver operating characteristic (HSROC) estimate of sensitivity (95% confidence interval (CI)) at the high specificity of 0.95 was 0.51 (0.21 to 0.81), and 0.84 (0.58 to 0.95) at 0.80 specificity. We considered the evidence on sensitivity and specificity as of very low certainty due to risk of bias (-1), imprecision (-1), and inconsistency (-1). Only four studies reported data at a halo cut-o% > 0.3 mm, finding the following sensitivities and specificities (95% CI): 0.80 (0.56 to 0.94) and 0.94 (0.81 to 0.99) in 55 participants; 0.10 (0.00 to 0.45) and 1.00 (0.84 to 1.00) in 31 participants; 0.73 (0.54 to 0.88) and 1.00 (0.93 to 1.00) in 82 participants; 0.83 (0.63 to 0.95) and 0.72 (0.64 to 0.79) in 182 participants. Data on a direct comparison of temporal artery US with biopsy were obtained from 11 studies (808 participants; 460 with GCA, 56.9%). The sensitivity of US ranged between 0.03 and 1.00 with a median of 0.75, while that of TAB ranged between 0.33 and 0.92 with a median of 0.73. The specificity was 1.00 in four studies for US and in seven for TAB. At high specificity (0.95), the sensitivity of US and TAB were 0.50 (95% CI 0.24 to 0.76) versus 0.80 (95% CI 0.57 to 0.93), respectively, and at low specificity (0.80) they were 0.73 (95% CI 0.49 to 0.88) versus 0.92 (95% CI 0.69 to 0.98). We considered the comparative evidence on the sensitivity of US versus TAB to be of very low certainty because specificity was overestimated for TAB since it is one of the criteria used in the reference standard (-1), together with downgrade due to risk of bias (-1), imprecision (-1), and inconsistency (-1) for both sensitivity and specificity. Authors' conclusions There is limited published evidence on the accuracy of temporal artery US for detecting GCA. Ultrasound seems to be moderately sensitive when the specificity is good, but data were heterogeneous across studies and either did not use the same halo thickness threshold or did not report it. We can draw no conclusions from accuracy studies on whether US can replace TAB for diagnosing GCA given the very low certainty of the evidence. Future research could consider using the 2016 revision of the ACR criteria as a reference standard, which will limit incorporation bias of TAB into the reference standard.
