By Dr. Jonathan Shurlock
A recent publication in Circulation has generated online discussion regarding the use of so-called real-world data, the use of Medicare claims as a data source, the importance of true randomisation, and potential impact of bias on survival curves.
The study compared outcomes in patients with atrial fibrillation (AF) who either had standard treatment with anticoagulation or underwent left atrial appendage occlusion (LAAO). The authors highlighted their focus on so called real-world data in an attempt to account for differences in anticoagulation strategies and patient demographics than those seen in clinical trials.
Patients were identified from Medicare claims databases between 2015 and 2019. Individuals receiving LAAO were propensity matched on a 1:1 ratio with those receiving anticoagulation alone. Primary outcome measures were risks of mortality, stroke, systemic embolism, and bleeding. The authors describe attempting to adjust for confounding characteristics using Cox proportional hazard adjustment. 4,085 women and 5,378 men receiving LAAO were matched on a 1:1 basis to those receiving anticoagulation alone.
The authors reported a significant reduction in mortality risk associated with LAAO (women: HR, 0.509 [95% CI, 0.447–0.580]; men: HR, 0.541 [95% CI, 0.487–0.601], P<0.0001). With similar risk reduction reported for stroke (HR, 0.655 [95% CI, 0.555–0.772]) and systemic embolism (HR, 0.649 [95% CI, 0.552–0.762])
The authors conclude that these findings demonstrate an association between LAAO and a reduction in risk of death, stroke and systemic embolism and that these findings should be used to guide ‘shared decision-making with patients’.
Since publication there have been various responses querying the suitability of this study for reaching the author’s conclusions.The potential pitfalls of using medicare claims data has been discussed historically and more recently. Additional critique has focussed on the lack of randomisation in order to control for selection bias based on the clinician decision making at initiation of treatment. The same analysis written by Dr John Mandrola highlights the separation of survival curves within months of treatment, with a 50% reduction in risk of death at 12 months. Such a large risk reduction is not in keeping with previous randomised trials of LAAO, particularly when focussing on death from stroke. See the full critique from Dr Mandrola here.
Clinician’s involved in managing patients with AF who are likely to have frequent discussion about stroke risk reduction should take the time to read the full paper in order to properly analyse and draw conclusions.