Development and Internal Validation of a Risk Score to Detect Asymptomatic Carotid Stenosis. (March 2021)

• Record Type: Journal Article
• Authors: Poorthuis, Michiel H.F.; Sherliker, Paul; Morris, Dylan R.; Massa, M. Sofia; Clarke, Robert; Staplin, Natalie; Lewington, Sarah; de Borst, Gert J.; Bulbulia, Richard; Halliday, Alison
• Is Part Of: European journal of vascular and endovascular surgery. Volume 61:Number 3(2021)
• Page Start: 365
• Page End: 373
• Publication Date: 2021-03
• Subjects: Atherosclerosis -- Carotid artery stenosis -- Ischaemic stroke -- Prevention -- Risk prediction model -- Targeted screening; Blood-vessels -- Endoscopic surgery -- Periodicals; Blood-vessels -- Surgery -- Periodicals; Vascular Surgical Procedures -- Periodicals; Vascular Surgical Procedures -- methods -- Periodicals; Vaisseaux sanguins -- Chirurgie -- Périodiques; Vaisseaux sanguins -- Chirurgie endoscopique -- Périodiques; Blood-vessels -- Endoscopic surgery; Blood-vessels -- Surgery; Endoscopy; Electronic journals; Periodicals; Electronic journals; 617.413005
• DOI: 10.1016/j.ejvs.2020.11.029
• Languages: English
• ISSNs: 1078-5884

Abstract

Abstract : Objective: Asymptomatic carotid stenosis (ACS) is associated with an increased risk of ischaemic stroke and myocardial infarction. Risk scores have been developed to detect individuals at high risk of ACS, thereby enabling targeted screening, but previous external validation showed scope for refinement of prediction by adding additional predictors. The aim of this study was to develop a novel risk score in a large contemporary screened population. Methods: A prediction model was developed for moderate (≥50%) and severe (≥70%) ACS using data from 596 469 individuals who attended screening clinics. Variables that predicted the presence of ≥50% and ≥70% ACS independently were determined using multivariable logistic regression. Internal validation was performed using bootstrapping techniques. Discrimination was assessed using area under the receiver operating characteristic curves (AUROCs) and agreement between predicted and observed cases using calibration plots. Results: Predictors of ≥50% and ≥70% ACS were age, sex, current smoking, diabetes mellitus, prior stroke/transient ischaemic attack, coronary artery disease, peripheral arterial disease, blood pressure, and blood lipids. Models discriminated between participants with and without ACS reliably, with an AUROC of 0.78 (95% confidence interval [CI] 0.77–0.78) for ≥ 50% ACS and 0.82 (95% CI 0.81–0.82) for ≥ 70% ACS. The number needed to screen in the highest decile of predicted risk to detect one case with ≥50% ACS was 13 and that of ≥70% ACS was 58. Targeted screening of the highest decile identified 41% of cases with ≥50% ACS and 51% with ≥70% ACS. Conclusion: The novel risk model predicted the prevalence of ACS reliably and performed better than previous models. Targeted screening among the highest decile of predicted risk identified around 40% of all cases with ≥50% ACS. Initiation or intensification of cardiovascular risk management in detected cases might help to reduce both carotid related ischaemic strokes and myocardial infarctions.