Author: Michael Vlessides
Predicting ischemic stroke after surgery may have gotten easier, thanks to a team of researchers. In a recent study, their novel scoring system—the STRAS, or stroke after surgery score—accurately predicted the risk for this sometimes deadly adverse event in patients with and without a history of ischemic stroke.
Of note, their study also found that in addition to well-established predictors, either a history of migraines or patent foramen ovale (PFO) was associated with the incidence of postoperative ischemic stroke.
“Ischemic stroke after surgery affects a large patient population,” said Katharina Platzbecker, MD, the lead author of the study and a research fellow at Massachusetts General Hospital, in Boston. “Earlier this year, our group reported that ischemic stroke risk after surgery is not only high in the immediate postoperative period but remains elevated for up to two years after the procedure.
“Therefore, we aimed to provide a tool that helps identify patients that would benefit from prevention in terms of mitigating ischemic stroke risk,” she added. “Our primary aim was to focus on stroke risk at one year, because several stroke prevention measures cannot be implemented immediately after the procedure and may only show their full potential after some time. On the other hand, health care providers tend to evaluate 30-day outcomes, so we also tested the short-term performance of our model, as well.”
Development of a Dual Model
The retrospective cohort study comprised 357,565 adult patients undergoing noncardiac surgery between 2005 and 2017 at independent health care networks in Massachusetts. For score development, selected predictors were based on available literature and comprised baseline patient characteristics and demographics, comorbidities, and procedure-related factors. In the development cohort, stepwise backward regression was utilized for predictor elimination, followed by bootstrap resampling to avoid overfitting.
Each final predictor was weighted by the value of its beta coefficient, divided by the overall smallest one of the final model. The predictive ability of the model was then evaluated in the independent external validation cohort by applying C-statistics and calibration plots.
“Overall, we screened more than 400,000 patients in the two cohorts for eligibility,” Dr. Platzbecker explained at the 2018 annual meeting of the American Society of Anesthesiologists (abstract A3014). “We only looked at adult patients, and excluded those with an ASA physical status of VI and patients undergoing cardiac surgery.” The final development cohort comprised more than 150,000 patients, whereas the final validation cohort included over 207,000 patients.
A preliminary prediction instrument demonstrated excellent diagnostic ability, with an area under the curve (AUC) of 0.91. “Postoperative stroke incidence was highly conditional on previous history of ischemic stroke, which was a very strong predictor,” Dr. Platzbecker explained. “To enhance model fit, we decided to stratify our instrument by this very strong predictor and build two models for patients, [for those] with and without a history of ischemic stroke.”
The final development cohort included 5,650 patients (3.8%) with a history of ischemic stroke and 144,690 (96.2%) without such a history. Among these, 38.6% of patients with a history of ischemic stroke (n=2,182) experienced an ischemic stroke within one year after surgery, compared with 1.1% of those without such a history (n=1,644).
Independent predictors of ischemic stroke were identified to build the final score for patients without a history of ischemic stroke (Table 1). A separate list of factors was identified for those patients with a history of ischemic stroke (Table 2).
|Table 1. Independent Predictors of Ischemic Stroke in Those With No Such History|
|Age 41 to 60 years (2 points)|
|Age 61 to 70 years (3 points)|
|Age at least 70 years (4 points)|
|ASA physical status greater than II (4 points)|
|TIA within three months of surgery (7 points)|
|TIA within no more than one year and greater than three months before surgery (3 points)|
|Carotid artery stenosis (3 points)|
|PFO (3 points)|
|Migraine (2 points)|
|Hypertension (1 point)|
|Atrial fibrillation (1 point)|
|Valvular heart disease (1 point)|
|Peripheral vascular disease (1 point)|
|Chronic kidney disease (1 point)|
|Neurosurgery (5 points)|
|Vascular surgery (2 points)|
|Emergency surgery (2 points)|
|PFO, patent foramen ovale; TIA, transient ischemic attack|
|Table 2. Independent Predictors of Ischemic Stroke in Those With a History of Ischemic Stroke|
|Ischemic stroke within three months of surgery (8 points)|
|Ischemic stroke greater than three months and within one year before surgery (4 points)|
|Carotid artery stenosis (1 point)|
|PFO (1 point)|
|Migraine (1 point)|
|Emergency surgery (1 point)|
|Inpatient surgery (1 point)|
|PFO, patent foramen ovale|
“Among patients with a history of ischemic stroke, what is interesting is that even if the patient does not arrive at any other additional score points, the baseline risk is as high as 8.5%,” Dr. Platzbecker said. “Nevertheless, the time point of the previous stroke really matters. Among patients who had their previous stroke within the three months prior to surgery, the risk of stroke after surgery is almost 50%.” The model showed good discriminative ability with an AUC of 0.75 (95% CI, 0.74-0.77) in patients with a history of ischemic stroke and an AUC of 0.81 (95% CI, 0.80-0.82) in those without.
These results were confirmed by external validation in a cohort with a different case mix and different postoperative ischemic stroke incidence, which yielded an AUC of 0.73 (95% CI, 0.72-0.75) for patients with a history of ischemic stroke and an AUC of 0.82 (95% CI, 0.81-0.83) among those without a history of ischemic stroke.
“In conclusion, we are convinced that this instrument can be used by clinicians to identify and manage patients with a high risk of postoperative ischemic stroke,” Dr. Platzbecker said. “We also believe that it can be used by researchers to facilitate recruitment for stroke prevention trials.”
A Surprising Connection
The researchers were also intrigued by the finding that PFO and migraine had relatively higher score point values than conventional risk factors, such as atrial fibrillation and hypertension. “We speculate an underlying mechanism might be the fact that atrial fibrillation and hypertension are well-known risk factors and very likely to be treated, whereas migraine and PFO are rather nominal risk factors with no recommendations for prevention in the surgical population.”
The findings prompted considerable discussion. “I’m curious that you found PFO to be an important factor,” said session moderator Matthew K. Whalin, MD, PhD, an assistant professor of anesthesiology at Emory University School of Medicine, in Atlanta. “I don’t know how important it is in everyone else’s health system, but where I practice it’s not routinely tested. From what I understand, it’s pretty common; maybe 10% or more of our patients have one and might not know it.”
“In our cohort, the prevalence was only around 1%,” Dr. Platzbecker said. “Yet we know from studies that it is actually much higher than that—as high as 25%.”
“Once someone is diagnosed with a PFO at our institution, it’s fairly standard for them to get an echocardiogram to rule out PFO as a source of future stroke,” Dr. Whalin said. “In your data set, were you able to tell how often a PFO was discovered in patients?”
“It would be nice to do an echocardiogram on all patients prior to surgery, but this is obviously not going to happen,” said study senior author Matthias Eikermann, MD, PhD, the vice chair of faculty affairs in the Department of Anesthesia, Critical Care and Pain Medicine at Massachusetts General Hospital. Dr. Eikermann and his colleagues published a recent paper on PFO as a risk factor of postoperative ischemic stroke (Eur Heart J 2018. doi: 10.1093/eurheartj/ehy402). “We were able to substantiate our findings in subcohorts of patients who had undergone adequate diagnostics, meaning an echocardiogram with a bubble study. Of course, in this subgroup of patients, the PFO prevalence was much higher.”