Answers are grounded in this paper and cite their sources.
Retrospective validation compares GCA prediction models' effectiveness in biopsy cohorts, suggesting their cautious use in triage.
Clinical prediction models are increasingly used to triage suspected giant cell arteritis (GCA), but their comparative performance in temporal artery biopsy (TAB) cohorts remains uncertain. We conducted a retrospective local external validation of three TAB prediction tools (González-López, Ing, Walvick & Walvick) in patients undergoing TAB for suspected GCA at a Swiss tertiary referral center (January 2018 to December 2021). Discrimination was quantified using the area under the receiver operating characteristic curve (AUC). Of 87 eligible patients, 38 (43.7%) had TAB-confirmed GCA; 81 had complete inputs for the primary head-to-head comparison. Discrimination was highest for the Ing model (AUC 0.877, 95% confidence interval [CI] 0.806 to 0.949), followed by the González-López model (AUC 0.832, 95% CI 0.745 to 0.919) and the Walvick & Walvick model (AUC 0.786, 95% CI 0.686 to 0.886). At an illustrative 0.40 probability threshold, sensitivity was 77.1% for both Ing and González-López, with specificity of 75.0% and 69.2%, respectively. At the Ing calculator’s specified 0.085 threshold, sensitivity was 100% and specificity 36.2%, with no false-negative and 30 false-positive classifications. In this biopsy-selected cohort, calculators showed moderate-to-good discrimination, but threshold-based use involved clinically important trade-offs between missed TAB-positive cases and unnecessary confirmatory evaluation. These tools may assist triage but should not be used as stand-alone arbiters to omit clinical review or confirmatory testing. Future work should prioritize externally validated models that integrate standardized vascular ultrasound and, where appropriate, other imaging modalities within structured diagnostic pathways.
No takes yet. Share an insight, caveat, or question.
Spoerri et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: