Enhanced Survival With Surgical Ablation of Atrial Fibrillation During Mitral Valve Surgery

Surgical ablation (SA) at the time of isolated mitral valve surgery (MVS) is recommended in patients with preexisting atrial fibrillation (AF). However, SA remains infrequently utilized during MVS with a poorly quantified impact on stroke and survival.

Medicare claims (2008-2019) were queried to identify beneficiaries with preexisting AF undergoing MVS. All-cause mortality and the postoperative incidence of stroke/transient ischemic attack were evaluated as separate endpoints. Overlap propensity score weighting was used to risk-adjust for measured confounding variables. Analyses were repeated using surgeon frequency of SA as an instrumental variable to adjust for unmeasured confounding variables.

From 2008-2019, 41,795 Medicare beneficiaries with a preexisting diagnosis of AF underwent MVS. Surgeons were categorized, with 1326 infrequently (bottom quartile) performing SA (<30%; 10,364 beneficiaries) and 740 frequently (top quartile) performing SA (≥62%; 10,476 beneficiaries) during MVS. Beneficiaries undergoing MVS with SA ("as-treated" analysis) had a risk-adjusted median survival advantage of 0.56 (95% CI, 0.33-0.81) years (8.85 [95% CI, 8.64-9.04] vs 8.29 [95% CI, 8.11-8.47] years, P < .001 for risk-adjusted survival comparison) compared to those without. Beneficiaries undergoing MVS by frequent SA surgeons ("surgeon-preference" analysis) had a risk-adjusted median survival advantage of 0.35 (95% CI, 0.05-0.71) years (8.59 [95% CI, 8.40-8.85] vs 8.24 [95% CI, 7.97-8.40] years, P = .0015 for risk-adjusted survival comparison) compared to surgeons who infrequently performed SA.

In Medicare beneficiaries with preexisting AF, concomitant SA during MVS is associated with improved survival, as is undergoing surgery by a frequent SA surgeon. When analyzed based on surgeon preference for SA, the magnitude and time-dependent nature of the treatment effect of SA were substantially different compared to the "as-treated" analysis, suggesting that "as-treated" analyses may be subject to bias from unmeasured confounding variables.