Current models are not reliable when it comes to predicting events after carotid revascularization in acute patients. Peri-procedural events seem to be particularly hard to predict. The development of models that can be externally validated is essential for the decision-making process in patients with high event rates, both during the procedure or while the case is “cooling off.”
That is why the aim of this work was to evaluate the short- and long-term external validity of previously published prediction models, which may have worked very well in the studies that featured them and (above all else) with the operators involved, but which may not be reproducible in a real-world setting with a patient with symptomatic severe carotid lesions.
For that purpose, researchers analyzed data on 2184 patients who underwent angioplasty and 2261 patients who underwent endarterectomy included in 4 randomized studies (EVA-3S, SPACE, ICSS, and CREST). Investigators assessed 23 models to estimate stroke or death within 30 days from the procedure and 7 models to predict long-term risk.
Death or stroke occurred in 158 patients who underwent angioplasty (7.2%) and 84 patients who underwent endarterectomy (3.7%). Most models to predict short-term events after angioplasty (n = 4) or endarterectomy (n = 19) had very poor discriminative performance and poor calibration with small absolute risk differences between the lowest and highest risk patients and, in general, overestimation of risk in the highest risk patients.
As regards long-term outcome models, they had a slightly better performance and reasonable calibration. The problem is that we are dealing with acute patients: decisions must be made fast and what really tips the scales are peri-procedural events or events within 15 days, if a decision is made against revascularization.
None of the current models is reliable when it comes to making the decision to carry out a carotid revascularization in an acute patient. In particular, prediction of short-term outcomes, where the highest risk lies regardless of the decision made, seems to be a difficult task.
Original title: Prediction Models for Clinical Outcome After a Carotid Revascularization Procedure. An External Validation Study.
Reference: Eline J. Volkers et al. Stroke. 2018 Jul 16. Epub ahead of print.
We are interested in your opinion. Please, leave your comments, thoughts, questions, etc., below. They will be most welcome.