Physicians underestimate stroke, bleeding risk in AF patients

Physicians tend to underestimate both stroke and bleeding risk when deciding whether or not to treat atrial fibrillation (AF) patients with oral anticoagulants. An analysis published online March 29 in Circulation found that in up to 80 percent of cases, physician-assessed risk differed from objective risk scores.

Cardiologists, electrophysiologists and primary care physicians who care for patients with AF must weigh the benefit of stroke prevention against the risk of bleeding in patients with AF. Benjamin A. Steinberg, MD, of the Duke Clinical Research Institute in Durham, N.C., and colleagues designed a study using data from the U.S.-based ORBIT-AF (Outcomes Registry for Better Treatment of Atrial Fibrillation) registry to assess if physicians’ estimate of risk aligned with estimates in empirical risk models.

ORBIT-AF included a baseline report from each patient’s physician that provided the physician’s subjective assessed risk for stroke and bleeding. Risk was categorized as low (less than 3 percent), intermediate (3 to 6 percent) or high (more than 6 percent). Patients also were evaluated using CHADS2, a stroke risk assessment tool, and ATRIA, a bleeding risk assessment tool. 

They stratified 10,094 patients with AF who enrolled in the registry between June 2010 and August 2011 by risk using both the subjective and objective risk assessments and then calculated the agreement between scores. They also determined the rate of treatment with oral anticoagulants across groups.

Physicians rated 16 percent of patients as high risk for stroke while the CHADS2 score placed 72 percent as high risk. By physician assessment, 7 percent of patients were at high risk of bleeding compared with 17 percent by ATRIA score.

Factors such as prior stroke or transient ischemic attack and severe AF symptoms seemed to weigh more heavily in physician’s assessment of stroke risk than hypertension, heart failure and age. They seemed less likely to consider anemia and significant renal disease in bleeding risk assessment compared with ATRIA scoring.

“One explanation is that physicians simply value certain risk factors differently from empirical scores,” Steinberg et al wrote. “Alternatively, this could be another manifestation of poor calibration of risk by clinicians.”

Eighty percent of patients deemed high risk of stroke by CHADS2 score received oral anticoagulants and 81 percent seen as high risk by subjective assessment were treated with oral anticoagulants. Oral anticoagulant treatment didn’t differ greatly between patients with low or high ATRIA scores or with low or high physician risk estimates.  

Stroke risk may trump bleeding risk in physicians’ decision making. They also noted that the rate of oral anticoagulation overall was high but “[u]p to 70 percent of patients at low risk of stroke receive OAC and nearly 20 percent of patients at high risk did not receive OAC,” they wrote. “There remains room for improvement in the selection of AF patients for OAC.”

The use of a registry might introduce sampling bias, they wrote. Editorialists Alex Y. Tan, MD, and Kenneth A. Ellenbogen, MD, of Virginia Commonwealth University School of Medicine in Richmond, added that more than 80 percent of the patients were treated by a cardiologist or electrophysiologist, and results may not reflect real-world practice that includes more primary care physicians.