By Michael H. Crawford, MD, Editor

SYNOPSIS: A large epidemiologic study with validated endpoints of patients with atrial fibrillation after non-cardiac surgery demonstrated such patients experience a higher incidence of subsequent atrial fibrillation, stroke, transient ischemic attacks, and all-cause mortality over five years of follow up.

SOURCE: Siontis K, Gersh BJ, Weston SA, et al. Association of new-onset atrial fibrillation after non-cardiac surgery with subsequent stroke and transient ischemic attack. JAMA 2020;324:871-887.

Atrial fibrillation (AF) after non-cardiac surgery may be an isolated phenomenon or a harbinger of future adverse events. Investigators from the Mayo Clinic employed the Rochester Epidemiology Project to ascertain the risk of stroke or transient ischemic attack (TIA) in patients with postoperative AF compared to similar patients without postoperative AF from 2000 to 2013.

From this population, patients with first-ever AF during or within 30 days after non-cardiac surgery were selected and matched with a comparator group by age, sex, year of surgery, and type of surgery. This resulted in 452 matched pairs. In those who developed AF, it occurred a median of two days after surgery; 90% occurred within 14 days. The median age of participants was 75 years, and 52% were men. Patients with AF had higher comorbidity indices and a mean CHA2DS2-VASc score of 4 vs. 3 for those without AF (P < 0.001). Over a median follow-up of five years, all adverse outcomes were more prevalent in those developing AF.

After adjustment for age and comorbidities, the risk of stroke or TIA was 10.7% (95% confidence interval [CI], 7.1-14.2) in the AF group vs. 6.0% (95% CI, 3.5-8.4) in the no AF group. The hazard ratio (HR) was 2.69 (95% CI, 1.35-5.37). Subsequent AF also was more common in the AF group: 51.4% (95% CI, 45.8-56.4) vs. 12.1% (95% CI, 9.0-15.1) for the no AF group. The HR was 7.94 (95% CI, 4.85-12.98). There was a significant difference in all-cause mortality in the AF group (HR, 1.66; 95% CI, 1.32-2.09), but not in cardiovascular mortality.

Anticoagulation therapy was prescribed in 49% of those with AF, and the median time on anticoagulants was 60 days. At one year, 35% were on anticoagulants, which did not appear to be related to the development of stroke or TIA. The authors concluded patients who developed AF after non-cardiac surgery had a higher incidence of subsequent stroke, TIA, AF, and all-cause mortality. However, the implications for the management of these patients cannot be ascertained from this observational study.


Previous studies that relied heavily on administrative databases and were of shorter duration than this analysis have shown that subsequent AF and stroke are more common in those who develop AF postoperatively. This study confirms and strengthens these findings in three important ways. First, all outcomes were manually verified in the patients’ charts rather than relying on coding alone. Second, there was a matched control group, so the effect of the type of surgery and other factors were balanced between the groups. Third, the follow-up was for five years. Thus, the results are perhaps not surprising, but serve to continue to cast doubt on the theory that postoperative AF is an isolated phenomenon caused by the stress of surgery and probably can be ignored for most patients. It seems more likely, as the authors noted, that surgery is a “stress test for AF.” Also, these could be patients with pre-existing paroxysmal AF that was undetected until the prolonged monitoring after surgery.

The most impressive result of this study is more than half of the postoperative AF patients experienced subsequent AF during follow-up. This raises the issue of whether these patients should be anticoagulated, starting when and for how long. Of course, postoperative bleeding concerns would delay the start of anticoagulants if they are considered necessary. In this analysis, few patients were started on anticoagulants, so these data cannot inform such decisions. A randomized trial would be necessary. However, it is interesting that 15% of patients who suffered a subsequent stroke were on anticoagulants. Perhaps these patients with more comorbidities than the patients who did not develop AF suffered stokes that were not thromboembolic. Unfortunately, the investigators could not fully categorize the mechanism of stroke in all patients. The fact that all-cause mortality was higher in the AF group supports the theory that they were more prone to adverse outcomes for a variety of reasons.

There were several limitations. The study population was largely white, so the results may not apply to other groups. The type of AF (paroxysmal, persistent, permanent) could not be determined. Postoperative AF patients may have been monitored more, so there may be an ascertainment bias toward finding AF subsequently. In addition, few patients in this study underwent vascular surgery, which would be expected to be a higher risk group for AF. Despite the matched control group, there always is the possibility of residual confounding in an observational study.

Since there are no randomized, controlled trials, watchful waiting may be one option. Or cardiologists could consider anticoagulant therapy in appropriate candidates for one to three months, and then conduct a monitoring study for a few weeks and base further therapy on the results. Perhaps in some patients, a loop recorder would make sense, with monthly interrogation. Perhaps anticoagulating appropriate candidates indefinitely if their CHA2DS2-VASc score was > 2 is the right approach. Unfortunately, a solid solution to this problem does not exist yet. Until then, cardiologists should act in a manner that seems most reasonable to us and patients.