Predictive models for thromboembolic events in giant cell arteritis: A US veterans health administration population-based study

Abstract

Giant cell arteritis (GCA) that affects older patients is an independent risk factor for thromboembolic events. The objective of this study was to identify predictive factors for thromboembolic events in patients with GCA and develop quantitative predictive tools (prognostic nomograms) for pulmonary embolism (PE) and deep venous thrombosis (DVT). A total of 13,029 patients with a GCA diagnosis were included in this retrospective study. We investigated potential predictors of PE and DVT using univariable and multivariable Cox regression models. Nomograms were then constructed based on the results of our Cox models. We also assessed the accuracy and predictive ability of our models by using calibration curves and cross-validation concordance index. Age, inpatient status at the time of initial diagnosis of GCA, number of admissions before diagnosis of GCA, and Charlson comorbidity index were each found to be independent predictive factors of thromboembolic events. Prognostic nomograms were then prepared based on these predictors with promising prognostic ability. The probability of developing thromboembolic events over an observation period of 5 years was estimated by with time-to-event analysis using the method of Kaplan and Meier, after stratifying patients based on predicted risk. The concordance index of the time-to-event analysis for both PE and DVT was > 0.61, indicating a good predictive performance. The proposed nomograms, based on specific predictive factors, can accurately estimate the probability of developing PE or DVT among patients with GCA.