While a positive TAB is the gold standard for a diagnosis of GCA, its sensitivity ranges from ~ 70 to > 90%, which underscores that a negative biopsy does not exclude the diagnosis of GCA . This sensitivity rates may be even lower in the large-vessel phenotype of GCA with reported rates being as low as 52%.  Skip lesions may contribute to a negative TAB in the presence of GCA; as well in patients with predominant large vessel disease . Therefore, there has been a longstanding interest in the search for serological markers to better aid the diagnosis of GCA with a focus on inflammatory markers [11,12,13,14,15,16,17,18,19,20,21]. In this study we have confirmed that ESR, CRP and platelet counts each have moderate diagnostic utility for a subsequent clinical diagnosis of GCA in the most relevant context, which is all patients referred for a TAB. Further, we have estimated cut-off values for the interpretation of test results. These cut-off values were estimated at 50 mm/hr. for ESR, 20 mg/L for CRP and 300 × 109/L for platelet counts. Importantly, we found no difference in these optimum cut-off values between TAB positive and negative GCA patients, as these tests are likely to be the most useful in TAB negative patients.
The findings of our study are broadly consistent with findings of multiple previous studies, yet direct comparisons are complicated by differences in patients and control definition, and particularly, cut-off values used to define a positive test. Our study identified a cut-off of 50 mm/hr. for the ESR, which is the same as that used in the ACR Classification Criteria for GCA , and which has been utilised by a number of similar studies [14, 16, 17]. In comparison, other studies have utilized the upper limit of the normal laboratory range [12, 19], which is substantially lower than either CRP or ESR levels generally seen in GCA. Overall, there has been limited research on appropriate cut-off criteria for interpretation of a positive test for GCA. Importantly, the cut-off values derived from our study for ESR and CRP are comparable to those derived by Heyreh et al  who identified a cut-off of 47 mm/hr. for ESR and 24.5 mg/L for CRP, and also similar to those derived by Kermani et al  who identified a cut-off of 56 mm/hr. for ESR and 26.9 mg/L for CRP. Studies evaluating platelet count for the diagnosis of GCA have generally utilized a value of 400 × 109/L, derived from laboratory estimates of the normal range [15, 17, 19], whereas, in contrast to ESR and CRP, our estimated cut-off for platelets was within the normal laboratory range.
Studies which report AUC estimates for ESR, CRP and platelets can be directly compared to our study because these are independent of the cut-off values used. We reported an AUC for the ESR of 0.65 (95% CI 0.57, 0.72), and previous point estimates of 0.62 , 0.67 , 0.59  and 0.71  from four previous studies are within the confidence intervals of our estimate. Our AUC estimate for CRP, 0.72 (95% CI 0.65, 0.79), although identical to one previous study , was higher than in two other previous studies, 0.63  and 0.61  respectively. Similarly, our AUC estimate for platelets, 0.72 (95% CI 0.65, 0.79) was virtually identical to the point estimate from three previous studies [15, 17, 19], but higher than a fourth (0.63) . An important caveat for the comparison of these studies to ours is that these previous studies all compared the TAB positive patients to TAB negative patients, which most likely included some TAB negative GCA patients. Regardless, all studies suggest that ESR, CRP and platelets have, at best, moderate ability to distinguish between GCA and non-GCA patients, and as demonstrated by Toren et al , the utility of these three diagnostic tests in predicting positive biopsy is decreased for patients who have been initiated on glucocorticoids at the time of referral for biopsy.
Discordance between positive ESR and CRP results is a recognised phenomenon and an evaluation of discordant ESR/CRP laboratory tests in adults indicated clinical differences, with infections, myocardial infarction and venous thrombosis more prevalent in the high CRP/low ESR group, and connective tissue disease, ischemic strokes and transient ischemic attacks more prevalent in the high ESR/low CRP group . This discordance is also observed in GCA, with one study reporting that the CRP has a significantly better sensitivity for GCA compared to the ESR . In our study, this discordance also extended to positive platelet count results, with a kappa agreement between the three tests of only 67%. Although there was a trend for a lower AUC and lower specificity for the ESR test compared to the CRP and platelet tests in our study, this did not reach statistical significance, and we conclude that the tests are in fact comparable at the cut-off values used. It is also quite possible that discordant results may reflect underlying meaningful clinical differences between GCA patients, although this remains to be properly evaluated.
The discordance between ESR, CRP and platelet results in GCA suggest the possibility that a combination of tests may provide the best utility for the diagnosis of GCA. In our study, a multivariable analysis indicated that, given CRP and platelet results, the ESR was essentially redundant, and that specific combinations of CRP and platelet results resulted in high sensitivity and specificity for GCA. Of the three previous studies which evaluated ESR, CRP and platelets by multivariable regression, two concluded, as in our study, that CRP and platelets were the best predictors of GCA [15, 20], whereas the other concluded that ESR and platelets were the best predictors . This latter study also included other blood count markers such as neutrophil: lymphocyte ratio, and monocyte: lymphocyte ratio which, in addition to CRP, which did not reach statistical significance in multivariable regression.
A strength of our study was that it consisted of a state-wide cohort of patients from 5 tertiary referral and peripheral centres hence capturing the full spectrum of patients. Our study included not only patients with a positive TAB but included patients with a clinical diagnosis of GCA, despite having a negative TAB. This is crucial as studies have shown TAB results do not affect the management of patients with suspected GCA  and as the ACR Classification criteria were not designed as diagnostic criteria, patients may still be diagnosed with GCA on clinical grounds, especially if there is a good response to glucocorticoid therapy. Hence, we believe our findings more accurately reflect real-world clinical practice. Limitations of this study were that a third of the study population was excluded due to missing data, however the excluded patients had similar age and gender distribution to the included cohort. Further data on concomitant steroid treatment was not available.