LAUSR

I congratulate Lyons et al. [1] on their work. However, the summary section implies that temporal artery biopsy (TAB) is no longer the “gold standard” for the diagnosis of giant cell arteritis (GCA). In 2018 the European League Against Rheumatism first suggested that ultrasound and MRI can be used for the first-line diagnosis of GCA at centres with sufficient expertise in performing and interpreting these studies [2]. However, ultrasound machines cost at least £15,000–60,000 [3], and require extensive training. The July 2019 draft British Society of Rheumatology guidelines recommended that the confirmatory diagnostic test for GCA could be either a TAB at least 1 cm in length, or an ultrasound of the temporal and axillary arteries, or both [4]. Most neuroophthalmologists still prefer TAB over ultrasound for the work-up of GCA [5]. On meta-analysis the estimated sensitivity of TAB is 77% [6]. In contrast, systematic review of the literature found that the hypoechoic halo sign on temporal artery ultrasound had 68% sensitivity and 81% specificity compared with a positive TAB [7]. Atherosclerosis and other conditions can cause a false-positive halo on ultrasound [8]. TAB may reveal alternative diagnoses including syphilis, sarcoidosis, metastases, amyloidosis [9], zoster, and granulomatosis with polyangiitis [10], which may not be discovered in an expedient fashion without tissue pathology. The TABUL study [11] is referenced a dozen times by Lyons et al. to support their statement that, “investigations, along with clinician insight, produce the highest sensitivity and specificity” for GCA. However, the TAB in the TABUL study were substandard. Seven percent of the attempted TAB did not retrieve a temporal artery, and 43% of the TAB specimens were <1 cm in length. “Clinician insight” is valuable in the diagnosis of GCA, but is subject to bias. TRIPOD-compliant [12], externally validated prediction models [13] more objectively estimate the pretest probability in patients with suspected GCA. Clinicians may have difficulty estimating the risk of non-linear continuous variables such as age and bloodwork, and may not appreciate that the platelet level is a stronger predictor for GCA than ESR or CRP [13]. Algorithms cannot replace physicians but can make them better [14]. Finally, inefficiency of the PD-1/PD-L1 (programmed cell death protein 1/programmed death-ligand 1) immune checkpoint has been implicated in the pathophysiology of GCA [15].