Comment on: British Society for Rheumatology guideline on diagnosis and treatment of giant cell arteritis: reply

Dear Editor, We thank Dr Ing [1] for his comments in response to our guideline [2]. He communicates his enthusiasm for novel ‘computerized algorithms’ for diagnosis of GCA and his ‘preference [for] temporal artery biopsy (TAB) over ultrasound for the work-up of GCA’.

Patient partners were deeply involved in the development of our clinical guideline, and indicated the high value placed by patients on a fast, accurate diagnosis of GCA. In preparing our guideline, we reviewed the evidence for the accuracy of temporal artery US (TAUS). We concluded that, where TAUS is available, it can provide fast, actionable clinical information for patients with suspected GCA. Of course, performing TAUS does not preclude going on to also perform TAB if necessary. Indeed, if the TAUS result does not fit with the clinical picture, we would recommend proceeding to TAB: in this situation TAB would add further, valuable clinical information. In our guideline, we also highlight the potential role of aortic imaging for GCA diagnosis, but this is less often used as a first-line test. As always in medicine, no test is perfect, and all medical test results should be interpreted in light of the whole clinical picture.

Dr Ing reminds us that clinicians can sometimes get a diagnosis wrong. This is a truism; all humans are subject to cognitive biases. This is why in our guideline we advise assessing pre-test probability in light of all available clinical information and then performing a confirmatory test for GCA (TAB, imaging or both), rather than advocating long-term steroids based purely on clinical judgement. This approach is consistent with that advocated by the latest EULAR guidelines for GCA, which states that imaging and biopsy have similar diagnostic value if the assessors are proficient in the techniques and that a suspected diagnosis of GCA should be confirmed by either imaging or histology [3].

Dr Ing goes on to make the far stronger claim that a computerized algorithm can outperform physician judgement. He does not provide any convincing evidence for this. Clinical skills and diagnostic acumen should not be reduced to mere ‘physician intuition’. Crucially, what Ing’s 10-factor prediction model cannot do is account for other possible explanations for the presenting features, including rarer ‘do not miss’ conditions such as infection or cancer. An alternative approach has been suggested that prompts the physician to consider other possible explanations [4]. Its utility in clinical practice is currently being evaluated prospectively.

We avoided citing any prediction rule in the guideline to avoid the appearance of prematurely endorsing them. Particularly where ‘black box’ algorithms are concerned, the potential for harm is well-documented. There is also a risk of discarding relevant clinical information. For GCA, nuances of the history or rarer clinical features such as scalp necrosis may not contribute much to accuracy or efficiency of a diagnostic algorithm at the group level, but could significantly upgrade judgement of clinical probability for an individual patient.

The Grading of Recommendations Applicability, Development and Evaluation (GRADE) guideline process we followed suggests predefining categories of pre-test probability of disease, in order to conduct simple decision analysis to estimate how many false-negatives and false-positives would result from different testing strategies [5]. We defined these categories of pre-test probability using an expert elicitation process within the guideline group, supported by clinical vignettes. Furthermore, we feel the chosen 20% cutoff for ‘low’ probability of GCA makes clinical sense. If on average 20–25% of TABs are positive, then this implies that physicians calibrate their judgement such that on average, the benefit–harm ratio for investigation (TAB) and treatment of GCA is favourable if the pre-test probability is around 20–25%. The 50% probability threshold between ‘medium’ and ‘high’ probability corresponds to our earlier definition of ‘strongly suspected’ GCA as warranting immediate steroid treatment (‘GCA diagnosis more likely than any other diagnosis’). Referring for TAB and starting immediate steroids are clinical decisions that specialists who regularly assess patients for GCA are already used to making. Whether an algorithm would outperform these specialists in predicting test results remains to be seen.

In comparison with TAB, TAUS is faster, safer, more acceptable to patients and examines a far wider portion of the relevant vasculature. If TAUS is available, then patients can be investigated more safely for GCA than if only TAB were available. Hence, TAUS may still be indicated for patients with a pre-test probability <20%. As well as reassuring patients and clinicians if the TAUS is negative, a positive TAUS in this situation could be of great value in prompting further investigation and treatment in those patients with atypical presentations who otherwise might face visual loss due to delayed diagnosis [6].

Conversely, in the setting of strongly suspected GCA (defined as >50% probability), a positive TAUS can be considered confirmatory of the diagnosis. This avoids the morbidity associated with TAB and allows the TAUS to be repeated later for monitoring of response to therapy: a key question for future research is whether this could be used to help assess tocilizumab-treated patients and predict further relapse where inflammatory markers are suppressed by the drug.

We agree that there can be a ‘grey area’ of equivocal test results for TAUS; this is also true for TAB [7]. Future advances in the clinical utility of ultrasound may include use of quantitative measurements of halo size and extent [8] rather than a single cut-off for TAUS positivity as advocated by Ing in his letter.

Academics and clinicians should be working in partnership with patient groups to improve our GCA diagnostic clinical pathways in order to confirm or refute a diagnosis of GCA faster and more accurately. Perhaps the most important point to make for our readership is that when US has been introduced to GCA diagnostic pathways, the incidence of GCA-related sight loss has fallen [9, 10]. Even now, there are far too many patients who are found to have irreversible visual loss due to GCA. We must act to address this.

Funding: No specific funding was received from any funding bodies in the public, commercial or not-for-profit sectors to carry out the work described in this manuscript.

Disclosure statement: The clinical authors of this letter have all successfully implemented TAUS into their routine clinical practice and we still request TAB for some, but not all, of our patients with suspected GCA. The two patient authors co-wrote the GCA guidelines and work in a voluntary capacity for groups supporting patients with polymyalgia rheumatica and GCA.

References