Author + information
In a recent report, Villareal et al. (1) presented compelling data in support of an independent association between new-onset atrial fibrillation (AF) after coronary artery bypass graft (CABG) surgery and excessive long-term mortality up to 1,400 days. In the accompanying editorial, Levy and Kannel (2) correctly caution readers that this contention is not yet fully substantiated. This is because of study design and analysis limitations, including 1) potential residual confounding, and 2) lack of objective data supporting a plausible mechanism for how AF causes this late mortality. We have the following additional observations.
First, as partly noted in the editorial (2), entering age as a categorical variable >65 years is probably inadequate given its importance for both AF and mortality, which clouds the reported findings. Indeed, the fact that older age was not found to predict mortality in either the AF or no-AF cohort (refer to Table 6 in Villareal et al. ) may have been a side product of this methodology, and it is particularly surprising given that age >65 years was a highly significant predictor (p < 0.0001, OR = 2.04; refer to Table 3 in Villareal et al. ) of mortality for all patients combined (1).
Second, Villareal et al. (1) elaborated on a potential exciting observation, namely that antiarrhythmic drugs—which were associated with a significant 32% reduction in mortality—may have a yet undefined therapeutic role in improving outcomes in AF patients. This too was discussed in the editorial (2), but we contend that this potential role is significantly tempered by the fact that the derived protective effect of these drugs was identical in AF and no-AF patients (refer to Table 6 in Villareal et al. ). This presents the following question: If antiarrhythmic therapy improves outcome, shouldn't their protective effect be more pronounced in the AF population?
Our third observation relates to the fact that the case-matched groups were distinctly different from either prematching group. The employed methodology was very briefly described in the Methods section (1), and it is difficult to ascertain how the matching was done. Yet, almost always, propensity matching of patients will lead to matched cohorts with their variable values intermediate to the two baseline populations. This was not true of the case-matching used in this study (1), and its implications on the reported results remain an open question.
The above limitations not withstanding, to the extent that these long-term findings are true combined with the known substantial impact of AF on postoperative outcomes and resource utilization, this report (1) re-emphasizes a) the need to improve the clinical management of AF during and after hospitalization and, perhaps more importantly, b) the need to develop prophylactic measures (including surgical) to minimize its incidence. Accordingly, future efforts should focus on answering the critical questions: 1) in whom should prophylactic interventions (e.g., ablation) be done and how effective are they? and 2) are the associated costs justified?
Finally, the difficulty encountered by the investigators in matching AF patients to their no-AF counterparts—only 195 of 994 (20%) AF patients could be matched despite an abundance of no-AF (n = 5,481) patients—is an interesting yet underdiscussed aspect. We propose that this difficulty may prove to be a useful characteristic that may facilitate identification of multivariate models of high sensitivity and specificity for predicting high-risk AF patients. Such models may then be used to target prophylactic interventions toward patients who are most likely to benefit, and to do so in an objective and cost-effective manner.
- American College of Cardiology Foundation
- Villareal R.P.,
- Hariharan R.,
- Liu B.C.,
- et al.
- Levy D.,
- Kannel W.B.