Surgical ablation of atrial fibrillation: early survival benefit or residual confounding?

Dear Editor,

We read with interest the study by Dr Pasierski and colleagues regarding outcomes of surgical ablation (SA) in patients with atrial fibrillation undergoing cardiac surgery [1]. The authors report immediate (within 30 days) and persistent (up to 6 years) adjusted survival benefits associated with SA.

The authors reasonably hypothesize the long-term survival benefit relates to the maintenance of sinus rhythm, delaying heart failure progression. However, the authors do not offer an explanatory mechanism for the early survival benefit. We struggle to identify a clinically-plausible explanation for such an early divergence in survival curves; thus, we question whether this finding is due to unmeasured confounding variables not addressed by the well-executed inverse probability of treatment weighting design. Indeed, the findings are in contrast to a study from Malaisrie and colleagues noting similar early survival and a late (after 2 years) signal of improved survival in SA recipients [2].

Many factors influence surgeons’ decision to perform SA, including duration of atrial fibrillation, degree of left atrial dilation, ‘eyeball test’ of individual patient’s ability to tolerate additional time on cardiopulmonary bypass, and each surgeon’s beliefs about the efficacy of SA [3]. Although the authors controlled for many variables that influence early outcomes in their model, these other factors, some of which are impossible to quantify, may have resulted in residual confounding. In other words, the patients receiving SA may have remained ‘healthier’, even after statistical adjustments, and thus less likely to suffer early mortality (which, we contend, is independent from receiving SA).

An alternative study design, termed provider-preference instrumental variable analysis (PP-IVA), affords an opportunity to adjudicate whether residual confounding underlies this finding [4]. Under a set of reasonable assumptions specific to PP-IVA, comparing outcomes of patients operated by surgeons who frequently versus infrequently perform SA would balance these difficult-to-measure variables and thus yield a more accurate measurement of any treatment effect associated with SA [5]. We have previously demonstrated the utility of PP-IVA to answer other relevant questions in cardiac surgery [6]. We encourage the authors to consider PP-IVA to adjudicate whether residual confounding may instead explain the early treatment effect they have attributed to SA.

Conflict of interest: none declared.

REFERENCES