We used data from the statewide NYCAS cohort study of 9,308 CEAs performed by 482 surgeons in 167 hospitals to identify independent patient risk factors for death and stroke within 30 days of surgery. NYCAS is the largest, clinically detailed, population-based study of CEA outcomes and risk factors in community practice. Among the 25 potential patient factors that were examined, we identified four domains of variables that were independently associated with higher risk of perioperative death and stroke, several of which represent prognostic factors that have not previously been assessed or reported.
We were able to use the large NYCAS dataset to stratify patients into several distinct neurological acuity subgroups which represent new findings. Most prior work focused on differences in complications between patients operated for symptomatic v. asymptomatic carotid disease. Our results confirm the well-documented finding that symptomatic patients have twice the risk of perioperative death or stroke compared to asymptomatic ones.6, 7, 18, 35
While prior studies and the national guidelines largely consider asymptomatic patients as a homogenous low risk group, this study shows that asymptomatic patients with a history of distant cerebrovascular disease (stroke or TIA or stroke >1 year prior to surgery) have one-third higher risk adjusted complication rates compared to patients with no history of stroke or TIA. This is important because three-quarters of CEAs in the US are done in asymptomatic patients, and these patients have less to gain from surgery.32,43, 44
Among symptomatic patients, stroke as the indication for surgery (compared to TIA) has also been identified in some,33, 35, 36, 39
but not all22, 24, 18
prior investigations. Some of the heterogeneity in the literature appears influenced by whether ocular TIAs (low risk) are lumped together with cerebral TIAs or not.18, 27
Unfortunately, we were not able to distinguish ocular from hemispheric TIAs in our dataset. The current study expands this work by identifying three distinct prognostic subgroups among symptomatic patients who have a stepwise increase in the risk of complications—those operated on for TIA, stroke, and the acute syndromes (crescendo TIA or stroke-in-evolution). This also confirms the finding of a systematic review which combined data from 10 studies and concluded that patients with crescendo TIA and stroke-in-evolution constitute a very high risk group.18
The NYCAS study had nearly as many of these unusual cases as were present in all of these 10 studies combined.
Our multivariable model also highlighted two other poor prognostic factors (admission from the ED and severe disability) that are additional measures of neurological acuity. Admission from the ED was a poor prognostic factor even after stratifying for recent carotid symptoms so this factor may capture ways in which patients admitted from the ED may differ in other ways regarding subtle differences in neurologic severity, trajectory of symptoms, or comorbid illness burden, among other possible factors. Severe disability probably represents substantial loss of brain function due to a large territory major stroke. Patients who had severe disability had triple the complication risk (13.08% rate of death or stroke) confirming the recommendations of our national expert panel who felt that such patients were inappropriate candidates for CEA because the harms of surgery outweighed the benefits.31
We also identified two anatomic risk factors. Patients with 50% to 99% stenosis of the contralateral internal carotid artery (significant, but non-occluded disease on the non-operated side) had 44% greater risk-adjusted complication rates probably due to diminished collateral blood flow capacity. Most prior work focuses on the impact of total contralateral occlusion,11, 13, 24, 45
though we have previously reported worse outcomes with 50-99% contralateral stenosis in other patient populations.18, 20, 39
We are uncertain about what to conclude from the trend towards double the risk of adverse events among patients with deep carotid artery ulcers. The borderline finding (p=.07) is likely due to its rarity as a risk factor. Ulcerated plaques of any severity increased the risk of complications in NASCET,24
was of borderline significance in ECST,23
and was not a risk factor Academic Medical Center Consortium observational study.38
NYCAS provided a unique opportunity to evaluate the impact of advanced age on outcomes because the mean age was 75 years. In NYCAS, patients ≥80 years old had one-third higher risk-adjusted odds of death or stroke confirming the results a prior registry20
and VA study.29
While many studies examined age ≥80 years as a univariate risk factor, the literature is mixed on this topic46
and interpretation limited by lack of formal multivariable analyses in most cases.14, 46
Additionally, most of the RCTs excluded patients older than 80 years (as well as those with major comorbidities) because of concerns about higher risk and more limited life expectancy. The RCTs of CEA v. carotid stenting reported much higher risk of perioperative complications in patients ≥80 years old.5, 47
Taken together, this suggests that octogenarians comprise a high risk group for whom the benefits of any carotid revascularization (CEA or stenting) may be greatly diminished compared to their younger counterparts.
Our finding that coronary artery disease and diabetes increases the risk of complications was expected and consistent with the prior literature on CEA, as well as the larger cardiac risk assessment literature.48
That diabetes requiring insulin was a more robust prognostic variable (compared to any type of diabetes) is a novel finding, though one that makes sense clinically as a marker of more severe diabetes and vascular disease burden.
The fact that Black and Hispanic patients had worse outcomes even after adjusting for age, neurologic and comorbidity factors was unexpected and the reasons for such potential disparities in surgical outcomes should be the subject of further investigations. The few previous studies that examined racial and ethnic disparities in CEA outcomes found conflicting results.12,19, 21, 28
It is worth noting that we did not find differences in results by gender, degree of ipsilateral stenosis, history of heart failure or atrial fibrillation, among other characteristics that have sometimes found to be risk factors in other studies.45, 46
These differences may be due to variations in the type of study samples or use of multivariable techniques.
Several strengths and limitations are worth noting. NYCAS is the largest, most clinically detailed, population-based study of CEA outcomes in unselected, community practice. The very large number of cases enabled us to examine the independent impact of over 25 potential sociodemographic, neurological, and comorbidity risk factors among both symptomatic and asymptomatic patients. All data were based on detailed independent chart review, and we ascertained deaths and strokes within 30 days of surgery (not just those that occurred during the index hospitalization).
However, like all observational cohort studies, we relied on information on risk factors and complications documented in the medical records during usual practice. There was no standard approach to pre- or post-surgical assessment as could have done in a prospective trial. That said, we had access to the full complement of inpatient notes, diagnostic imaging results, and operative reports, and all deaths and strokes were confirmed by physician over-reading. While the data reflects practice in 1998-1999, operative techniques and perioperative management for CEA have been consistent over the intervening period, and there is no reason to believe that association between risk factors and outcomes would change considerably over time. Finally, just because certain subgroups had higher risks of adverse events after CEA does not mean that such patients should not have surgery. Whether patients with risk factors we identified as increasing the short term risk of death or stroke due to surgery would also be at higher long term risk of death or stroke if they were managed with medical therapy alone is unknown. The decision to have surgery must balance benefits and harms.
These results have several practical implications. From a clinical standpoint, information about risk factors should help referring physicians, neurologists, surgeons, and anesthesiologists better weigh the risks and benefits of CEA for an individual patient. This prognostic information may also help identify those who might be considered potential candidates for carotid stenting because they are too high risk from CEA. From a research and quality improvement perspective, there is a need for CEA-specific risk-adjustment models so that outcomes among different patients and providers can be fairly compared. Similarly, since RCTs of various surgical and anesthesia techniques are rarely undertaken, observational data are often used to highlight processes of care associated with better outcomes—something that requires appropriate risk-adjustment. CEA-specific risk models appear to be superior to the standard generic cardiac risk assessment tools.48
Finally, most of the prognostic factors we identified (indication for surgery, contralateral stenosis, neurologic disability, and diabetes on insulin) are only knowable from the medical record. This has implications for risk adjustment models and surgical audit studies based solely on hospital discharge databases.