Perioperative risk prediction using the POSSUM and V-POSSUM models in symptomatic carotid stenosis

Abstract

Summary: Background: This study aimed to evaluate risk factors for adverse outcomes and perioperative stroke and death in patients with symptomatic carotid stenosis undergoing open endarterectomy (CEA). The second objective was to assess the predictive value of the POSSUM and V-POSSUM models for predicting morbidity and mortality from CEA in symptomatic carotid stenosis. Patients and methods: A retrospective observational study of all patients admitted to a single center who underwent CEA for symptomatic carotid stenosis was performed. 320 patients from 1999 to 2013 were included. Postoperative complications, 30-day survival, and stroke rates were recorded. The observed outcomes were compared to the POSSUM and V-POSSUM expected mortality (observed to expected ratio (O:E)). Results: The mean age was 68.1±10.0 years. 215 patients were male (67%). Risk factors for surgical complications were: age, with a higher risk in both groups of less than 60 years and more than 75 years of age (p=0.04), a higher ASA score (p=0.04), and hyperlipidemia (p=0.017). Risk factors for the combined endpoint stroke or death were a higher ASA category (p<0.001), stroke as indication for CEA (p 0.022), and a high degree of stenosis (p=0.019). For POSSUM predicted mortality, there was a good O:E ratio in the two lowest risk groups, but a 2-fold overprediction of death or stroke in the two high-risk strata. The area under the curve (AUC) was 0.58 (95% CI: 0.43–0.73). The V-POSSUM showed a better fit in the high-risk groups, but an underprediction of mortality in the low-risk strata. Conclusions: Age and comorbid conditions are risk factors for adverse outcomes after CEA. The V-POSSUM model is better than POSSUM to predict postoperative death and stroke after CEA in patients with symptomatic carotid stenosis and a high preoperative physiological score. In patients with low physiological scores, both POSSUM and V-POSSUM show a limited predictive value.