etween subgroups in quantitative variables with normal distribution and equality of variance were disclosed by One-Way ANOVA analysis. Significant test statistics were followed with Scheffe post hoc test. Quantitative variables with normal distribution but inequality of variance were tested with Brown Forsythe and Welch test and post hoc Games-Howell test.
Quantitative variables with a distribution other than normal were log transformed to normalise distribution to satisfy the One-Way ANOVA condition. If the log transformed distribution did not approximate normality Kruskal Wallis test was used. In case of significant test statistics, pairwise Mann-Whitney test with Bonferroni correction, was performed.
Kaplan Meier survival curves with time to event late cardiac death were drawn for each of the five subgroups. Differences in estimates between subgroups were tested with Log-Rank.
Variables found to be significantly different between subgroups were tested in regard to potential confounding. Each variable was tested using univariable Cox proportional-hazard regression with time to event late cardiac death and event AF as the dependent variables, respectively. Confounders were identified at a significance level of 5% (p-value < 0.05). The proportionality assumption for the Cox models was tested with Schoenfeld residuals.
Univariable Cox proportional-hazard regression was subsequently used to evaluate the relationship between CK-MB and both ECC and CC.
A multivariable Cox proportional-hazard regression model was produced for late cardiac death and AF, respectively. Potential confounders identified in the previous analysis were adjusted for in the two models.
Cumulative hazard functions assessed the correlation between POAF and AF. Differences in baseline characteristics between the POAF patients and patients in postoperative sinus rhythm were considered with Pearson- x2 coefficient or Fishers exact test for categorical variables and independent sample T-test for normal distributed quantitative variables. Variables with a distribution other than normal were assessed with non-parametric Mann-Whitney test. A multivariable Cox proportional-hazard regression model with time to event AF adjusted for possible confounders was produced.
Cox proportional-hazard regression was subsequently used for evaluating the risk of all-cause death and cardiac death following POAF and AF.
The proportionality assumptions for the overall multivariable Cox models were tested with global Schoenfeld residuals test. The goodness-of-fit of the models were estimated by Martingale residuals and associated Cox-Snell residuals.
Values with a two-sided alpha level of p < 0.05 were considered statistically significant.
IBM SPSS 25.0 and STATA 15 statistical software were used for statistical analysis.
2.6. Interobserver Analysis
An interobserver analysis was performed to evaluate the validity of the information from The Western Denmark Heart Register. From each CK-MB subgroup, 10% of patients were chosen at random. Two researchers (KKJ and LPR) independently reviewed patient records. Discrepancies were resolved by consensus. The information from The Western Denmark Heart register was compared to the information obtained from patient records by calculating Cohen’s -coefficient, variable sensitivity and specificity.
The Danish Data Protection Agency and the Danish Patient Safety Authority approved the study. The Regional Committees on Health Research Ethics for Southern Denmark waived consent requirements.
The study evaluated 414 patients. Baseline characteristics for the five CK-MB subgroups were balanced with respect to preoperative baseline characteristics (Table 1).
3.1. CK-MB and Late Cardiac Death
In the 10-year follow-up period 105 of 414 patients died, whereof 33 deaths were of cardiac origin. Three patients died from unknown causes.
The Kaplan Meier survival estimates of time to late cardiac death generally decreased with an increasing level of CK-MB especially pronounced after 5 years in subgroups with CK-MB > 75 μg/l (Figure 1). However, Log-Rank test did not identify any statistically significant difference between subgroups.
Neither univariable nor multivariable Cox proportional-hazard regression analysis identified any statistical significant association between postoperative CK-MB level and late cardiac death (Table 2).
Subsequent analysis showed a statistically significant decrease in both ECC and CC in subgroup CK-MB < 25 compared to subgroup CK-MB ≥ 100 (Table 3).
3.2. CK-MB and AF
AF occurred in 54 out of 414 patients, within 10 years. No association between postoperative CK-MB level and time to AF was identified in univariable or multivariable Cox proportional-hazard analysis (Table 4).
No differences between CK-MB subgroups regarding postoperative CNS injury, sternal infection or reoperation for bleeding were found. An increased postoperative CK-MB level was associated with a significantly higher risk of postoperative acute myocardial infarction, new Q-wave, higher EuroSCORE, reoperation due to ischemia and prolonged intensive care unit admission time.
3.3. POAF and AF
POAF occurred in 84 out of 414 patients, whereof 21 later developed AF. Among the 330 patients in postoperative sinus rhythm only 33 patients later developed AF (Figure 2).
Beta-blocker treatment was significantly more frequent in patients with postoperative sinus rhythm (76%) compared to patients with POAF (64%; p-value =
Table 1. Baseline characteristics.
Normally distributed variables are noted as mean with standard deviation, others as median with interquartile range.
0.035). There was no correlation between beta-blocker treatment and later diagnosis of AF.
Age and EuroSCORE were significantly different between the POAF patients and patients in postoperative sinus rhythm, while also being independently associated with late AF. However, multivariable Cox proportional-hazard analysis with time to event AF remained significantly associated with POAF when adjusting for age and EuroSCORE (Table 5).
Figure 1. Kaplan Meier survival analysis.
Table 2. Risk of late cardiac death.
Risk of late cardiac death evaluated with univariable and multivariable Cox analysis. Multivariable Cox analysis is adjusted for potential confounders: left ventricular ejection fraction, previous cardiac surgery, extracorporeal circulation time and EuroSCORE. ECC and EuroSCORE remained significantly associated with late cardiac death. Abbreviations: CK-MB: creatine phosphokinase MB, HR: hazard ratio.
3.4. AF and late Cardiac Death
Cardiac death was more frequent in POAF and AF patients compared to patients in sinus rhythm (Table 6).
Median admission time of patients in the POAF group was 175 hours (95% CI 178 - 200), which was significantly longer compared to 159 hours (95% CI 160 -
Table 3. Risk of prolonged extracorporeal circulation time and aortic cross-clamp time.
Effect of extracorporeal circulation time and aortic cross-clamp time on CK-MB. Abbreviations: CK-MB: creatine phosphokinase MB, HR: hazard ratio.
Table 4. Risk of later diagnosis of paroxysmal or persistent atrial fibrillation.
Risk of later diagnosis of AF evaluated with univariable and multivariable Cox analysis. Multivariable Cox analysis is adjusted for potential confounders: thrombocytevolume and EuroSCORE. Abbreviations: AF: paroxysmal or persistent atrial fibrillation, CK-MB: creatine phosphokinase MB, HR: hazard ratio.
168; p < 0.001) in patients with sinus rhythm. Median intensive care unit admission time in the POAF group was 35 hours (95% CI 38 - 47) compared to 34.5 hours (95% CI 34 - 40; p = 0.008) for patients in postoperative sinus rhythm. Mean EuroSCORE in POAF patients were 5 (95% CI 4.2 - 5.6), while 4 (95% CI 3.7 - 4.3; p = 0.013) in patients with sinus rhythm.
3.5. Interobserver Analysis
The interobserver analysis showed a general underreporting in The Western Denmark Heart Register. The underreporting of comorbidities was moderate with Cohens -coefficient ranging from 0.53 in reporting of chronic obstructive pulmonary disease to 1.0 for diabetes mellitus. Notification regarding medicinal usage was varying from unreliability of hypertensive treatment ( = 0.05) to the well performing beta-blocker treatment ( = 0.79) and ACE inhibitor treatment ( = 0.80). Underreporting also applied for postoperative variables such as CNS injury, sternal infection and new Q-wave on ECG.
Figure 2. Hazard functions for time to paroxysmal or persistent atrial fibrillation (AF).
Table 5. Risk of later diagnosis of paroxysmal or persistent atrial fibrillation confer postoperative rhythm.
Risk of later diagnosis of AF evaluated with univariable and multivariable Cox analysis. Multivariable Cox analysis is adjusted for potential confounders: age and EuroSCORE. Abbreviations: AF: paroxysmal or persistent atrial fibrillation, HR: hazard ratio, POAF: postoperative atrial fibrillation.
Table 6. Risk of late cardiac death in patients with POAF or AF compared to patients in sinus rhythm.
Risk of late all-cause and cardiac death in patients with POAF or AF compared to patients in sinus rhythm, evaluated with univariable and multivariable Cox analysis. Multivariable Cox analysis is adjusted for age as a continuous variable. Abbreviations: AF: paroxysmal or persistent atrial fibrillation, HR: hazard ratio, POAF: postoperative atrial fibrillation.
The variables ventricular dysfunction, renal function and POAF performed well with the latter having a Cohens -coefficient of 0.80, sensitivity of 72.7% and specificity of 100%.
Two patient records reported CABG to be acute and one reported a preoperative AF diagnosis, why they were excluded from analysis confer exclusion criteria.
4.1. CK-MB and Late Cardiac Death
There was a numeric increased cardiac mortality among higher CK-MB subgroups, progressively pronounced beyond 5 years postoperative. However, the result must be interpreted with caution. As time progress after the CABG procedure, there is a growing risk of residual confounding influencing the result. Furthermore, a higher postoperative CK-MB level was significantly associated with an increased EuroSCORE, which indicates a more vulnerable cardiac risk profile in these patients. Hence, a shorter lifespan due to co-morbidities is to some extent expected in these patients.
Left ventricular ejection fraction, previous cardiac surgery, ECC and EuroSCORE were found to be significantly different between CK-MB subgroups and independently associated with late cardiac death. Presumably mirroring different steps in the cascade of mechanisms. Previous cardiac surgery might result in vulnerability of the myocardium. A high EuroSCORE indicates reduced resistance due to comorbidities, while decreased ejection fraction reflects declining cardiac function, which complicates sufficient cardioplegia and reperfusion of the myocardium. Hence, resulting in prolonged ECC and CC. Further, inadequate cardioplegia and thus suboptimal preservation of the myocardium during surgery results in increased CK-MB release.
4.2. CK-MB and AF
The study did surprisingly not identify any association between CK-MB and late AF. As elevated CK-MB correlated well with impaired cardiac function and arguably insufficient cardioplegia of the myocardium, then CK-MB would theoretically have been associated with increased risk of arrhythmia . This finding suggests that development of AF is less dependent on the invasiveness of surgery and more so on pre-existing pathological substrate.
4.3. POAF and AF
The 2.4-fold increase in later diagnosis of AF in patients with POAF compared with patients in postoperative sinus rhythm could likewise indicate existence of pathological substrate in atrial myocardium that enables atrial arrhythmia. Thus leaving the myocardium prone for later development of AF regardless of immediate conversion to sinus rhythm.
The higher frequency of beta-blocker treatment among patients in postoperative sinus rhythm compared to patients with POAF is presumably a manifestation of the well known antiarrhythmic property of beta-blocker therapy .
The association between POAF and AF remained significant after adjusting for age. Furthermore, the hazard function for patients experiencing POAF increases at a higher rate from around 5 years postoperative compared to that of patients in postoperative sinus rhythm. If age were the main cause of AF, one would expect a more constant increase in AF with age.
These findings are consistent with the findings of Ahlsson et al. , who showed POAF to be associated with late AF and speculated if heart surgery could be viewed as a ’stress test’ that unmasks a predisposition for AF. Melduni et al.  shared the hypothesis of a predisposing pathological substrate and showed that likelihood of recurrence of AF was coupled to the extent of underlying disease quantified by left ventricular diastolic dysfunction.
Although non-significant, patients with POAF were more likely to die from cardiac cause within 10 years after CABG compared to patients in sinus rhythm. This correlates well with previously found increased all-cause mortality due to POAF in large cohort studies  . Ahlsson et al.  and Schwann et al.  found POAF to be correlated with increased cardiovascular mortality. Melduni et al.  did not find POAF significantly associated with all-cause mortality but did however find AF to be.
Therefore, the deficient significance regarding this matter in the present study is most likely due to underpower. However, accumulation of events in POAF patients could merely be a surrogate measure of the significantly higher EuroSCORE in POAF patients.
A number of other postoperative characteristics were considered. POAF patients were admitted significantly longer to both the regular hospital ward and intensive care unit. Furthermore, there was a higher incidence of postoperative CNS injury in POAF patients. POAF has previously been proved to increase the length of hospital stay in CABG patients   and accumulate costs . POAF prophylactic regimes have shown to significantly reduce risk of POAF and length of stay .
4.4. Limitations and Strengths
Data was collected prospectively. It is the first study to consider the long-term predictive capability of CK-MB on the risk of death confined to cardiac cause.
Knowledge of an association between POAF and AF is sparse, why this study reveals new data.
The cohort originates from a single-centre, which impair the external validity. Though it is noted, that the CABG procedure is relatively standardized. Therefore, extrapolation of the results might be reasonable.
Missing register data were to a large extent acquired from patient files, reaching a low percentage of missing values. Register data are subjected to interobservational variation.
The requirement of including more patients to increase power and prevent type 2 errors was weighed against the resulting higher risk of residual confounding if patients from earlier surgical years were included in the study. If so, prompting a higher risk of confounding from differences in patient demographics and procedural factors.
4.7. Descriptive Variables
Postoperative CK-MB measurement is a standardized procedure and blood sample analyses were performed at the same laboratory for all patients. In addition, it has to be taken into consideration that even though CK-MB is considered specific to cardiac tissue, some amount of the CK-MB is produced in the brain tissue and striated skeletal muscle. As mentioned, elevated CK-MB can also be measured in other clinical conditions .
4.8. Outcome Variables
The Western Denmark Heart Register does not distinguish between atrial flutter and atrial fibrillation and the occurrence is only registered during hospital admission. However, a study by Zebis et al.  showed no occurrences of atrial flutter after CABG in a similar population and only a risk of 1% in developing atrial fibrillation after hospital discharge.
Munkholm et al.  evaluated the POAF variable in The Western Denmark Heart Register and calculated variable sensitivity 75.2% (95% CI 72.2 - 78.7), specificity 90.9% (95% CI 88.8 - 92.6) and concluded that the POAF variable was reliable to be used in epidemiological studies. These findings correspond well with the results from the interobserver analysis.
The Danish Cause of Death Register was fully updated, why all patients had a complete 10-year followed-up. All patients were accounted for in The National Patient Register.
No association between elevated creatine phosphokinase MB level after solitary CABG and late cardiac death within 10 years was evident. However, creatine phosphokinase MB level ≥100 was significantly associated with both prolonged extracorporeal circulation time and aortic cross-clamp time compared to a creatine phosphokinase MB level < 25.
Creatine phosphokinase MB was not associated with a later diagnosis of paroxysmal or persistent atrial fibrillation within 10 years after solitary CABG. However, patients with postoperative atrial fibrillation presented with a statistically significant 2.4-fold increase in later diagnosis of paroxysmal or persistent atrial fibrillation within 10 years postoperatively.
Long-term follow-up in patients, who experience postoperative atrial fibrillation, may be beneficial to avoid later complications.