P-111 - PREDICTION OF LONG-TERM MORTALITY FOR PATIENTS WITH SEVERE ASYMPTOMATIC DE-NOVO CAROTID STENOSIS UNDERGOING CAROTID ENDARTERECTOMY (PREMY2SE-CEA) RISK SCORE: DEVELOPMENT, INTERNAL VALIDATION, AND COMPARISON WITH EXISTING PREDICTION MODELS.

TOPIC:
Other
AUTHORS:
D'Oria M. (University Hospital of Trieste ~ Trieste ~ Italy) , Mastrorilli D. (University Hospital of Verona ~ Verona ~ Italy) , Mezzetto L. (University Hospital of Verona ~ Verona ~ Italy) , Veraldi G. (University Hospital of Verona ~ Verona ~ Italy) , Calvagna C. (University Hospital of Trieste ~ Trieste ~ Italy) , Lepidi S. (University Hospital of Trieste ~ Trieste ~ Italy)
Introduction:
Currently endorsed clinical practice guidelines from the European Society for Vascular Surgery and European Stroke Organization suggest that carotid endarterectomy (CEA) in asymptomatic patients should only be proposed if the risk of perioperative stroke is less than 3% and life expectancy after the procedure is estimated to be at least five years. The aim of this study was to develop a risk prediction model for "PREdiction of long-term MortalitY for patients with severe asYmptomatic de-novo carotid StEnosis undergoing Carotid EndArterectomy (PREMY2SE-CEA)" and to externally validate all available risk scoring systems comparing them against the PREMY2SE tool.
Methods:
Data were collected retrospectively from a dedicated database on consecutive patients who underwent elective CEA for severe (>70% according to NASCET criteria) asymptomatic carotid stenosis at two Italian University Hospital from 2011 through 2016. The primary endpoint of the PREMY2SE risk score was five-year mortality. Discrimination and calibration of the risk prediction models were evaluated.
Results:
A total of 901 patients were included in the final study cohort (mean age 74.2 ± 7.9 years, 66.6% males). By performing multivariable logistic regression with backward elimination, a parsimonious model was derived. The model showed excellent discrimination with an Area Under the Curve (AUC) of 0.791 (SE 0.02, 95% CI 0.74 - 0.83) and good calibration with integrated calibration index (ICI) of 0.010. Internal validation by bootstrapping 2000 samples estimated a statistically significant discrimination with P= <.001 and Bootstrapped 95% CI of 0.78-1.25. Six other risk prediction models were identified; as outlined in Table I, the PREMY2SE risk score significantly outperformed these other risk models in this cohort (P=<.001). Based on the analysis of five-year survival curves with Kaplan-Meier estimates, the study cohort was divided into four categories as described in Figure 1 (log rank <.001): very low risk (92.1%, CI 88.6-95.7), low risk (90.5%, CI 89.3-91.7), moderate risk (72.2%, CI 65.9-78.4), high risk (64.4%, CI 50.2-78.6).
Conclusion:
This study describes development, evaluation, and internal validation of a risk prediction model (PREMY2SE) for long-term mortality after CEA in asymptomatic patients. Our model demonstrated acceptable predictive performance and good calibration metrics. Physicians could use the PREMY2SE-CEA risk scoring tool to complement their estimates of life-expectancy, and prompt selective consideration prophylactic CEA to improve benefit of the interventions. Further external validation of our prediction model in an independent cohort of patients might be required.
ATTACHMENTS: