Article under review:
Calvillo-King L, Xuan L, Zhang S, Tuhrim S, Halm EA. Predicting risk of perioperative death and stroke after carotid endarterectomy in asymptomatic patients: derivation and validation of a clinical risk score. Stroke 2010;41(12):2786-94.
Purpose: The purpose of this report was to develop and validate a clinical prediction rule to estimate risk of stroke or death within 30 days of carotid endarterectomy (CEA) for asymptomatic carotid stenosis.
Methods: The authors analyzed asymptomatic cases in the New York Carotid Artery Surgery (NYCAS) study, a population-based cohort of carotid endarterectomies performed in Medicare beneficiaries in New York state between January 1, 1998, and June 30, 1999. Cases were identified by using Medicare and New York state hospital discharge data for the 157 hospitals in the state. "Asymptomatic cases" were considered to be those with no history of carotid-territory stroke or TIA in the 12 months preceding surgery.
Hospital charts were abstracted by trained nurse abstractors for relevant sociodemographic data, results of neurologic examinations, functional status, laboratory results, diagnostic imaging results, and neurologic, medical, and surgical histories. Potential perioperative deaths, strokes, and TIAs were identified by abstracting the medical records of the index admission, as well as charts for all subsequent admissions to any New York state hospital within 30 days of surgery (identified through record linkage in Medicare and New York state hospital discharge databases). Cases identified as having a death, stroke, or TIA based on review of admission and progress notes, discharge summaries, or brain imaging were independently reviewed and confirmed by 2 physicians including a neurologist. Initial agreement was very high (95%) and any disagreements were resolved by consensus.
Independent predictors of perioperative stroke or death were identified using multivariate logistic regression analysis.
Results: NYCAS included 6553 carotid endarterectomies performed on asymptomatic patients. The mean age was 75 years (range 42 to 98), and 55% were male. The 30-day risks of death, any stroke, or stoke or death were 0.8%, 2.5%, and 3.1%, respectively.
Eight variables were identified as independent predictors of perioperative death or stroke among patients following carotid endarterectomy for asymptomatic carotid stenosis, including sociodemographic factors (female; nonwhite), neurologic factors (history of remote stroke or TIA, ie, more than 12 months prior to surgery; nonoperated stenosis of at least 50%; severe disability), and cardiac comorbidities (coronary artery disease; congestive heart failure; valvular heart disease).
The independent predictors of any perioperative stroke were similar to those for combined stroke and death, except that congestive heart failure did not meet statistical significance in any of the stroke models.
The authors produced 2 clinical prediction rules based on their results, an 8-point "CEA-8 Clinical Risk Score" and a 7-point "Patient-friendly Risk Score" (see Tables 1, 2, and 3). The patient-friendly version includes only items that could be ascertained by clinical history (ie, it does not include the degree of nonoperated stenosis).
Table 1: Clinical prediction rules for perioperative stroke or death following carotid endarterectomy for asymptomatic carotid stenosis
|Risk Factors||Points, CEA-8 Score||Points, Patient-friendly Score|
|Remote stroke or TIA (more than 12 months prior to CEA)||1||1|
|Nonoperated stenosis of 50% or more||1||NA|
|Congestive heart failure||1||1|
|Coronary artery disease||1||1|
|Valvular heart disease||1||1|
Table 2: Qualitative surgical risk categories as a function of total scores on the clinical prediction rules
|Qualitative Surgical Risk||Total CEA-8 Score||Total Patient-friendly Score|
|Low||0 to 2||0 to 2|
|High||4 or higher||4 or higher|
Table 3: Observed rates of stroke or death within 30 days of carotid endarterectomy for asymptomatic carotid stenosis as a function of surgical risk category
|Surgical Risk Category||CEA-8 Model||Patient-friendly Model|
|Low||0.6% to 2.2%||1.4% to 2.9%|
|High||6.6% to 9.6%||8.3%|
Using the defined qualitative surgical risk categories, 71% of patients in NYCAS who underwent carotid endarterectomy for asymptomatic carotid stenosis were at low risk, 20% were at moderate risk, and 10% were at high risk.
Discussion: The management of asymptomatic carotid stenosis has been and remains controversial, in part because the demonstrated statistically significant results of the Asymptomatic Carotid Artery Stenosis (ACAS) study are of small magnitude, and no consensus has developed concerning their clinical importance and generalizability. The results of ACAS, published in 1995, depended on stringent surgeon selection (less than 3% post-operative stroke or death rate) and an excellent perioperative stroke and death rate (1.5%, excluding the risk of angiography). Subsequent national guidelines have underscored the importance of restricting carotid endarterectomies to surgeons and hospitals with adequate surgical volume and at most a 3% risk of stroke or death within 30 days of carotid endarterectomy for asymptomatic carotid stenosis.
Less attention has been given to development of patient-level clinical prediction rules to provide an overall evidence-based assessment of the risks of carotid endarterectomy for asymptomatic patients. However, surgical outcomes depend not only on the operative skill of the surgical team and appropriate postoperative management, but also fundamentally on careful selection of appropriate cases for surgery, ie, patients who have a reasonable chance of benefiting from the procedure.
The present study provides an easy way of categorizing patients by preoperative risk. The present results suggest that carotid endarterectomy for asymptomatic carotid stenosis is unlikely to benefit patients in the high-risk category (as defined here) and even in the moderate category carotid endarterectomy would have to be considered questionable.
Further validation of this method is needed, especially with more contemporary samples in both Medicare and non-Medicare (eg, Veterans Administration) populations.