Development of a diagnostic prediction model for giant cell arteritis by sequential application of Southend Giant Cell Arteritis Probability Score and ultrasonography: a prospective multicentre study

Background Giant cell arteritis is a critically ischaemic disease with protean manifestations that require urgent diagnosis and treatment. European Alliance of Associations for Rheumatology (EULAR) recommendations advocate ultrasonography as the first investigation for suspected giant cell arteritis. We developed a prediction tool that sequentially combines clinical assessment, as determined by the Southend Giant Cell Arteritis Probability Score (SGCAPS), with results of quantitative ultrasonography. Methods This prospective, multicentre, inception cohort study included consecutive patients with suspected new onset giant cell arteritis referred to fast-track clinics (seven centres in Italy, the Netherlands, Spain, and UK). Final clinical diagnosis was established at 6 months. SGCAPS and quantitative ultrasonography of temporal and axillary arteries with three scores (ie, halo count, halo score, and OMERACT GCA Score [OGUS]) were performed at diagnosis. We developed prediction models for diagnosis of giant cell arteritis by multivariable logistic regression analysis with SGCAPS and each of the three ultrasonographic scores as predicting variables. We obtained intraclass correlation coefficient for inter-rater and intra-rater reliability in a separate patient-based reliability exercise with five patients and five observers. Findings Between Oct 1, 2019, and June 30, 2022, we recruited and followed up 229 patients (150 [66%] women and 79 [34%] men; mean age 71 years [SD 10]), of whom 84 were diagnosed with giant cell arteritis and 145 with giant cell arteritis mimics (controls) at 6 months. SGCAPS and all three ultrasonographic scores discriminated well between patients with and without giant cell arteritis. A reliability exercise showed that the inter-rater and intra-rater reliability was high for all three ultrasonographic scores. The prediction model combining SGCAPS with the halo count, which was termed HAS-GCA score, was the most accurate model, with an optimism-adjusted C statistic of 0·969 (95% CI 0·952 to 0·990). The HAS-GCA score could classify 169 (74%) of 229 patients into either the low or high probability groups, with misclassification observed in two (2%) of 105 patients in the low probability group and two (3%) of 64 of patients in the high probability group. A nomogram for easy application of the score in daily practice was created. Interpretation A prediction tool for giant cell arteritis (the HAS-GCA score), combining SGCAPS and the halo count, reliably confirms and excludes giant cell arteritis from giant cell arteritis mimics in fast-track clinics. These findings require confirmation in an independent, multicentre study. Funding Royal College of Physicians of Ireland, FOREUM.