Independent Predictors of Stroke After CABG Identified

Unique predictors of stroke following coronary artery bypass graft (CABG) surgery have been found in a study. According to the researchers, identification and modification of these risk factors for stroke could help prevent this outcome and aid in surgical planning.

“Age, preoperative hematocrit levels, smoking status and a history of congestive heart failure are all significant predictors of stroke following CABG surgery,” said Ziyad Knio, BS, a biostatistician at Beth Israel Deaconess Medical Center and Harvard Medical School, both in Boston. “Looking out for these in patients’ medical history is definitely worthwhile so that they can be better monitored.

“In the future,” said Mr. Knio, who reported the findings at the Society of Cardiovascular Anesthesiologists’ 2016 annual meeting (abstract 13), “we can use these variables, as well as others possibly, to score patients based on their risk of suffering a stroke following CABG.”

Co-author Rabya Saraf, BA, said successfully identifying patients with coronary artery disease who have an increased risk for stroke may modify their treatment. According to Ms. Saraf, one study suggests that an elevated risk for stroke may be mitigated by treatment using off-pump CABG or percutaneous intervention instead of traditional on-pump CABG (Am J Cardiol 2015;115:1382-1388).

Refining the Model

To investigate independent predictors of stroke in CABG surgeries, Mr. Knio and his colleagues used the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) national database.

The researchers extracted 8,666 CABG cases seen between 2007 and 2013 by querying for Current Procedural Terminology code. These cases were then cross-checked for stroke occurring within a 30-day postoperative period. Stroke was defined as an embolic, thrombotic or hemorrhagic vascular accident or stroke with motor, sensory or cognitive dysfunction lasting more than 24 hours. Of the patients who underwent CABG, 148 (1.4%) experienced a stroke within 30 days of surgery.

“This is not a considerable number,” said Mr. Knio, “but it was enough for us to begin looking for associations.” Variables potentially associated with stroke were examined using univariate and multivariate analyses, comparing patients with and without stroke.

“We looked at the potential predictors by identifying those that had a significantP value on univariate analysis,” Mr. Knio reported. “However, we felt that a 0.05 cutoff would be too strict and that a degree of leniency would be necessary since we were only beginning to look for associations.”

Variables significantly associated with stroke on univariate analysis included age, sex, American Society of Anesthesiologists class, diabetes, smoking, ventilator dependence, history of congestive heart failure, preoperative renal failure, emergency surgery, blood urea nitrogen, white blood cell count and hematocrit. These factors were then entered into a stepwise logistic regression to determine which were independently associated with stroke after CABG.

“We eliminated predictors one by one using backwards stepwise elimination with a 0.05 cutoff,” Mr. Knio explained.

The final model included age (odds ratio [OR], 1.03; 95% CI, 1.02-1.05; P=0.0002), preoperative hematocrit (OR, 0.96; 95% CI, 0.93-0.99; P=0.0034), smoking (OR, 1.86; 95% CI, 1.25-2.71; P=0.0016) and history of congestive heart failure (OR, 2.21; 95% CI, 1.47-3.25; P<0.0001).

According to Mr. Knio, the final c-statistic of 0.67 shows “a pretty good predictive accuracy. It’s not stellar,” he added, “but this model is definitely better than guessing whether a patient will suffer a stroke simply by chance.”

Data Limitations

The researchers acknowledged several limitations to the study, including the small number of patients who had a stroke within 30 days of CABG surgery. In addition, despite the “fairly complete” data, some of the variables that might be good predictors were missing information and were excluded from further analysis.

“We will definitely investigate these variables further in the future to help refine our predictive model,” Mr. Knio concluded.

The moderator of the ses sion, Thomas Floyd, MD, CM, FRCPC, professor of anesthesiology at Stony Brook School of Medicine, in New York, expressed concern about the ACS NSQIP database’s underestimation of overall stroke frequency and underrepresentation of blacks.

“African-Americans make up 13% of the population, but the ACS NSQIP database contains only 6% representation, and this is an at-risk population,” said Dr. Floyd. “We’re not doing a good job with the acquisition of data in that population, for some reason.”