AF and long-term mortality
All-cause mortality occurred in 568 out of 2,628 patients (21.6%) at a median follow-up of 4 [2-7] years. Preoperative AF was highly predictive of mortality (97/ 268 (36.1%) deaths in AF group vs 471/ 2,360 (19.9%) in SR group; HR: 2.24, 95% CI: 1.79-2.79, P<0.001). The Kaplan-Meier survival curves separated early and the difference remained constant for up to 10 years follow-up [Figure 1].
Of the baseline characteristics in Table 1, the following were associated with long-term all-cause mortality using univariate hazard regression model of Cox: increasing age (HR: 1.07, 95% CI: 1.06-1.08, P<0.001), bioprostheses (HR: 3.30, 95% CI: 2.28-4.77, P<0.001), hypertension (HR: 1.37, 95% CI: 1.14-1.63, P<0.001), diabetes (HR: 1.59, 95% CI: 1.31-1.95, P<0.001), raised preoperative serum creatinine (HR: 1.006, 95% CI: 1.005-1.007, P<0.001), chronic pulmonary disease (HR: 1.56, 95% CI: 1.29-1.89, P<0.001), peripheral vascular disease (HR: 2.03, 95% CI: 1.55-2.66, P<0.001), previous myocardial infarction (HR: 1.59, 95% CI: 1.16-2.17, P=0.006), poor left ventricular function (HR: 1.66, 95% CI: 1.24-2.23, P=0.002), and high EuroSCORE I (HR: 1.29, 95% CI: 1.26-1.34, P<0.001) [Supplemental table 1].
The following variables were then entered into the final baseline multivariate Cox proportional hazard model: age, hypertension, diabetes, preoperative serum creatinine, chronic pulmonary disease, peripheral vascular disease, pervious myocardial infarction, and left ventricular function. We did not include bioprostheses in the model, as we felt this reflects advanced age in the preoperative AF group rather than being a genuine risk factor. Similarly, we did not include EuroSCORE I in the model as we felt it would be a repetition, since EuroSCORE I incorporates risk factors already included in the model, such as age, chronic pulmonary disease, peripheral vascular disease, serum creatinine and left ventricular function. Multivariate analysis showed that AF remained significantly associated with long-term all-cause mortality after adjustment for other risk factors (HR: 1.54, 95% CI: 1.21-1.96, P<0.001) and after bootstrap resampling, the optimism-corrected c-index was -0.012 and the calibration slope was 0.906, which suggests no over fitting of the model. Other independent predictors of long-term mortality included advanced age (HR: 1.06, 95% CI: 1.05-1.08, P<0.001), presence of diabetes (HR: 1.46, 95% CI: 1.18-1.80, P=0.001), chronic pulmonary disease (HR: 1.44, 95% CI: 1.17-1.77, P=0.001), peripheral vascular disease (HR: 1.48, 95% CI: 1.11-1.97, P=0.001), poor left ventricular function (HR: 1.47, 95% CI: 1.04-2.08, P=0.031) and raised preoperative serum creatinine (HR: 1.005 95% CI: 1.003-1.007, P<0.001).
In a propensity score matching analysis, the risk of long-term all-cause mortality was higher in the preoperative AF cohort (OR: 1.47, 95% CI: 1.04-1.99, P=0.031) compared with the preoperative SR cohort after adjustment of baseline differences between the two groups [Supplemental Table 2, Supplement Table 3 and Supplemental Figure 1].