Postoperative atrial fibrillation predicts cause-specific late mortality after coronary surgery

Abstract

Objectives. To investigate the association between postoperative atrial fibrillation (POAF) and cause-specific death after coronary artery bypass grafting (CABG) over time. Design. The cohort included 6821 patients undergoing primary isolated CABG between 1996 and 2009. Survival analyses using Cox proportional hazards determined the association between POAF and late mortality and cause of death. Four categories of mortality were examined: cardiac mortality; and death related to arrhythmia, cerebrovascular disease, and heart failure. Results. Median follow-up was 9.8 years and 2152 of 6821 patients (32%) developed POAF. During follow-up, 2302 of 6821 patients (34%) died. For all mortality categories, lower survival rates were found among POAF patients. After adjustment for baseline characteristics, medical history, and preoperative status, POAF was related to increased mortality in all four categories: cardiac mortality (HR 1.4; 95% CI 1.3–1.5); death related to arrhythmia (HR 1.8; 95% CI 1.6–2.0); cerebrovascular disease (HR 1.4; 95% CI 1.2–1.6); and heart failure (HR 1.4; 95% CI 1.3–1.6). The effect remained more than 8 years after surgery. Conclusions. POAF predicts cause-specific late mortality after CABG, with a sustained effect many years postoperatively. This suggests that POAF-episodes are not merely an indication of more advanced disease at surgery, but predicts a persistent negative effect on cause-specific survival.

Key words::

• atrial fibrillation
• coronary artery bypass grafts
• CABG
• outcomes

Introduction

Postoperative atrial fibrillation (POAF), a common complication after coronary artery bypass grafting (CABG), is associated with a higher incidence of late mortality (1–8). The mechanisms behind the effect of POAF on mortality are still not thoroughly explored. The few existing studies that examine cause of death include solely the direct (primary) cause of death and show a higher incidence of overall cardiovascular mortality among patients with POAF (9–11). No study has analyzed to what extent morbidity that is plausibly related to POAF contribute to death (i.e. contributing cause of death). Analyses including both primary and contributing causes of death should better capture the impact of the diseases leading to the death of the patient (12), and approach the causal relationship between POAF and mortality.

An implied connection has been found between POAF and the occurrence of late arrhythmias (9), and POAF has been associated with an increased risk of stroke after CABG (13,14). In non-surgical populations, atrial fibrillation (AF) has also been associated with conditions such as stroke (15,16) and heart failure (15). The cause of death from these conditions can be assumed as connected to the effect of POAF in a surgical cohort.

The present study aimed at a thorough examination of cause-specific death associated with POAF after CABG, with analyses of both primary and contributing causes of death. The effects of POAF were followed over time, and adjustments were made for preoperative conditions. Our hypothesis was that POAF specifically predicts deaths related to arrhythmia, cerebrovascular disease, and heart failure. If POAF was simply an indication of more advanced age and disease, its effect on long-term survival would diminish over time. Therefore separate analyses that allowed the effect to change over time were performed.

Material and methods

Patients

From January 1996 through December 2009, 7428 consecutive patients underwent primary isolated CABG at the Department of Cardiothoracic Surgery, Uppsala University Hospital, Uppsala, Sweden.

The exclusion criteria included preoperative AF (n = 281) and other atrial arrhythmias (n = 5), in order to focus on postoperative, and not preexisting, arrhythmias; patients with preoperative atrial arrhythmias had an increased mortality in the examined categories of death compared with the patients with preoperative sinus rhythm (SR). Other exclusion criteria were pacemaker (n = 30), non-Swedish citizenship (n = 144), and operative morality (n = 147), which included deaths occurring during hospitalization (n = 130), or after hospitalization but within 30 days of surgery (n = 16), or after 30 days but clearly related to the surgical procedure (n = 1).

The final study cohort included 6821 patients. The majority of the operations (96%) were performed on-pump through a routine technique for cardiopulmonary bypass and with moderate hypothermia (32° to 34°C).

The study complies with the declaration of Helsinki and was approved by the local Ethics Review Board of Uppsala, Sweden (2010/453); individual informed consent was waived.

Data collection

Prospectively collected data of all patients were obtained through the department's research database. The quality of the database is regularly validated and retains a high quality. It contained very few missing data, which were then retrieved from patient charts or otherwise replaced by mean or median values. The variables included in the analyses are presented in Table I, and covered baseline characteristics, medical history, and preoperative status. New York Heart Association (NYHA) Functional Classification describes the extent of heart failure based on severity of symptoms. Canadian Cardiovascular Society (CCS) Angina Grading Scale classifies the severity of angina. European System for Cardiac Operative Risk Evaluation (EuroSCORE) is a method of calculating risk of operative mortality in cardiac surgery. EuroSCORE was added to the database in 2001 and was only available for half of the patients, with an even distribution between the two groups.

Table I. Patient characteristics.

Variables	POAF n = 2152 (32%) %	No POAF n = 4669 (68%) %	p-value
Age (years), mean ± SD	69.0 ± 7.7	65.2 ± 9.1	< 0.0001
Male gender	79	77	0.08
Hypertension	52	50	0.2
Diabetes	21	22	0.4
Prior MI	56	52	0.006
Diseased coronary vessels			0.5
1	4	5
2	21	21
3	75	74
Left main stenosis	34	32	0.2
LVEF			0.2
Good (> 50%)	69	70
Reduced (30–50%)	26	26
Poor (< 30%)	5	4
NYHA class			0.008
I/IIIA	47	51
IIIB/IV	53	49
CCS angina grading scale			0.1
0–2	32	35
3–4	68	65
EuroSCORE, median (25th to 75th pctl)^a	3 (2–5)	2 (1–4)	< 0.0001

CCS, Canadian Cardiovascular Society; LVEF, left ventricular ejection fraction; MI, myocardial infarction; NYHA class, New York Heart Association functional classification; POAF, postoperative atrial fibrillation.

^aVariable only available for half of the patients, even distribution between groups.

All patients were monitored by continuous five-lead telemetry for a minimum of 3 days postoperatively, longer in case of arrhythmia. Twelve-lead electrocardiograms (ECGs) were registered on admission, on postoperative Day 4, and additionally in case of arrhythmia. POAF was defined as an episode of AF after CABG lasting more than 30 s that was recorded postoperatively on a 12-lead ECG or on continuous telemetry.

No specific prophylactic treatment for POAF was used during the study period. Baseline medications, including beta-blockers, were reinstated the day after surgery, if permitted by the clinical status of the patient. Patients received low-dose dalteparin until discharge. Warfarin was not routinely given to POAF patients. Episodes of POAF were treated medically with sotalol or amiodarone, and/or electrical cardioversion. Patients who developed POAF were often started on beta-blockers postoperatively. According to local routines, all patients admitted in SR should be discharged in SR. In order to ensure adherence to this routine in at least 95% of all cases, 100 randomly selected medical records in the POAF group were reviewed. In two cases, the patients were discharged without documented SR, and with warfarin. Thereby, the proportion discharged without verified SR should not be more than 5% (p = 0.046).

Outcome

All patients were identified as being dead or alive on 31 December 2012. The cause of death for all Swedish citizens is registered in the Cause of Death Registry by the Swedish National Board of Health and Welfare, and are classified according to the International Statistical Classification of Diseases and Related Health Problems, Ninth or Tenth Revision (ICD-9 or 10). The registry contains multiple causes of death, in descending order of influence. One primary (underlying) cause of death is registered, defined as the disease or injury that initiated the train of morbid events leading directly to death (www.who.int). Diseases considered to have had a substantial impact on the process leading to the death of the patient are also registered, and termed the contributing causes of death.

Late mortality was defined as death from any cause more than 30 days after surgery. Cardiac mortality was defined as a cardiac diagnosis being the primary cause of death, and included ICD-9 codes 394–404, 411–414, and 420–425, or ICD-10 codes I05–I13, I20–I25, and I30–I52.

To better understand the impact of disease on mortality among POAF patients, additional analyses included both primary and contributing causes of death. The focus was on causes of death with a plausible connection to POAF. The hypothesis was that POAF might lead to late arrhythmia, cerebrovascular disease, and heart failure, which in turn was associated with increased mortality. Therefore, three categories of mortality were defined based on both primary and contributing causes of death: death related to arrhythmia (ICD-9 codes 426–427; ICD-10 codes I44–I49), cerebrovascular disease (ICD-9 codes 430–438; ICD-10 codes I60–I69), and heart failure (ICD-9 codes 428; ICD-10 codes I50).

Statistical analysis

The SAS statistical software (version 9.3, SAS Institute Inc., Cary, NC, USA) was used for data processing and statistical analyses.

The effect of POAF on long-term survival was determined by Kaplan–Meier survival curves. A log-rank test was performed to compare differences between the groups with and without POAF. Univariable, age-adjusted and multivariable Cox proportional hazard analyses were used to investigate predictors of late mortality. Known major risk factors for mortality after CABG surgery were considered, including baseline characteristics, medical history, and preoperative status. The variables were evaluated for linearity and categorized in case of nonlinearity, and were then used in the format that provided the best discriminatory power.

The variables entered into the univariable Cox analyses are presented in Table I. Variables with a p-value of ≤ 0.05 in the univariable analyses were included in the multivariable analyses. Models were developed for each category of mortality examined (cardiac mortality, and death related to arrhythmia, cerebrovascular disease, and heart failure). The results of the Cox analyses are presented as hazard ratio (HR) and 95% confidence intervals (CI). Adjusted survival curves were standardized to the distributions of adjustment covariates in the POAF cohort, and were estimated from the cumulative hazard functions analyzed in multivariable Cox’ regression models stratified by POAF status.

Any possible interactions between POAF and the independent risk factors for each outcome were tested by the introduction of an interactions term.

As the basic models assume the HR is constant over time, three separate models were performed on follow-up: after less than 3 years, 3–8 years, and more than 8 years after surgery. Few patients in the higher age groups were at risk during the later time periods and analyses included three age groups (≤ 60, 61–70, and > 70 years).

The robustness of the models was checked by risk factor analyses of late mortality, where all patients with one or more complications (n = 1266) after surgery were excluded. These complications included sternum insufficiency (n = 141), re-operation because of bleeding (n = 183), perioperative myocardial injury (n = 259), postoperative dialysis (n = 55), aortic balloon pump (n = 49), prolonged need of inotropic drugs (n = 656), respiratory complications (n = 64), infection (n = 22), and stroke (n = 175).

Results

The study cohort included 6821 patients. The median follow-up time was 9.8 years (range 0.1–17 years). During follow-up, 2302 of 6821 patients (34%) died. POAF was associated with an increased late mortality, overall. The absolute difference was 4.2% for 5-year survival, 9.5% for 10-year survival, and 14% for 15-year survival.

The 2152/6821 patients (32%) who developed POAF were older, had more often suffered from a prior myocardial infarction (MI), and had a more advanced NYHA class than patients without POAF (Table I). A higher EuroSCORE was also observed (n = 3197; variable added in 2001, Table I). Both groups had the same degree of coronary vessel disease: a third of the patients with left main stem stenosis, and a median of three diseased vessels (Table I). Perioperative variables did not differ between the two groups, while postoperative complications were more frequent among POAF patients (Table II).

Table II. Peri- and postoperative variables.

Variables	POAF n = 2152 (32%) %	No POAF n = 4669 (68%) %	p-value
Perioperative
Aortic cross-clamp time, min^a	35 (43–53)	34 (42–52)	0.3
CPB time, min^a	66 (82–97)	65 (80–96)	0.5
Use of IMA	93	92	0.6
Number of distal anastomoses^a	3 (3–4)	3 (3–4)	0.2
Perioperative myocardial injury^b	8.8	9.3	0.5
Postoperative
Re-operation, bleeding	3.5	2.3	0.005
IABP	0.9	0.6	0.2
Respiratory complication^c	4.1	2.2	< 0.0001
Infection^d	7.3	5.0	< 0.0001
Stroke^e	3.7	2.1	< 0.0001
Permanent pacemaker	0.8	0.7	0.6

CPB, cardiopulmonary bypass; IABP, intra-aortic balloon pump; IMA, internal mammary artery; POAF, postoperative atrial fibrillation.

^aMedian (25th to 75th pctl).

^bPlasma creatine kinase-band value > 50 μg postoperative Day 1.

^cProlonged mechanical ventilation > 24 h, or re-intubation.

^dTreatment with antibiotics, and/or positive blood culture.

^eSevere neurological symptoms lasting ≥ 6 h, verified by computer tomography.

Cause of death

Cardiac mortality was the most common primary cause of death, affecting 929/2302 patients (40%). The most common non-cardiovascular cause of death was cancer, followed by endocrine disease (including diabetes), respiratory disease, gastrointestinal disease, and infectious disease. Based on both primary and contributing causes of death, 420 deaths (18%) were related to arrhythmia, 396 deaths (17%) to cerebrovascular disease, and 706 deaths (31%) to heart failure. The primary cause of death was classified as cardiac in 30% of the arrhythmia-related deaths, 13% of the cerebrovascular disease-related deaths, and 52% of the heart failure-related deaths. Overall, one of these three diseases contributed to 1144 deaths (50%), and 710 deaths (31%) were related to either arrhythmia or cerebrovascular disease.

Predictors of late mortality (> 30 days)

POAF was related to an increased cardiac mortality in univariable analyses (HR 1.7; 95% CI 1.6–1.9) (Figure 1A). The effect was reduced after adjusting for age (HR 1.4; 1.3–1.5), but remained after adjusting for other baseline characteristics, medical history, and preoperative status (Table III, Figure 1B).

Figure 1. Kaplan–Meier survival curve based on cardiac mortality (A), and survival curve adjusted for baseline characteristics, medical history, and preoperative status (B), by occurrence of POAF. The figure shows 95% CI at 5, 10, and 15 years. Log-rank p < 0.0001. POAF = postoperative atrial fibrillation.

Table III. Multivariable predictors of cardiac mortality after CABG. Follow-up 10 (6–13) years.^a

Variables	n	Dead	HR	95% CI
POAF
No	4669	548	1.0	ref.
Yes	2152	381	1.4	(1.3–1.5)
Age (years)
≤ 60	1629	113	1.0	ref.
61–70	2536	276	1.7	(1.5–1.9)
71–80	2458	489	3.9	(3.7–4.1)
> 80	198	51	8.2	(7.9–8.6)
Diabetes
No	5365	676	1.0	ref.
Yes	1456	253	1.8	(1.6–1.9)
Number of prior MIs
≤ 1	5925	710	1.0	ref.
2	682	159	1.4	(1.3–1.6)
3	163	42	1.5	(1.2–1.8)
≥ 4	51	18	3.0	(2.5–3.4)
LVEF
Good (> 50%)	4732	471	1.0	ref.
Reduced (30–50%)	1798	374	1.9	(1.8–2.1)
Poor (< 30%)	291	84	3.4	(3.2–3.7)

CABG, coronary artery bypass grafting; HR, hazard ratio; LVEF, left ventricular ejection fraction; MI, myocardial infarction; POAF, postoperative atrial fibrillation.

^aMedian (25th to 75th pctl). All variables in Table I were considered, and the final analysis adjusted for age, diabetes, number of prior myocardial infarctions, and left ventricular function.

POAF was also related to an increased cause-specific mortality, based on both primary and contributing causes; arrhythmia-related death HR 2.1 (1.9–2.3), cerebrovascular disease-related death HR 1.9 (1.7–2.1), and heart failure-related death HR 1.8 (1.7–2.0). After adjusting for age, the influence of POAF was somewhat reduced; arrhythmia-related death HR 1.7 (1.5–1.9), cerebrovascular disease- related death HR 1.4 (1.2–1.6), and heart failure-related death HR 1.4 (1.3–1.6). After adjusting for other baseline characteristics, medical history, and preoperative status, the effects of POAF remained virtually the same as after adjusting only for age (Table IV). Survival curves (Figure 2) adjusted for the distributions of the covariates of the POAF cohort indicate diverging curves over time in all mortality categories.

Figure 2. Adjusted survival curves based on death related to: (A) arrhythmia, (B) cerebrovascular disease, and (C) heart failure, by occurrence of POAF. The figures show 95% CI at 5, 10, and 15 years. Log-rank p < 0.0001. POAF = postoperative atrial fibrillation.

Table IV. Multivariable predictors of death related to arrhythmia, cerebrovascular disease, and heart failure after CABG. Follow-up 10 (6–13) years.^a

Variables	n	Primary or contributing cause of death
		Arrhythmia^b			Cerebrovascular disease^c			Heart failure^d
		Dead	HR	95% CI	Dead	HR	95% CI	Dead	HR	95% CI
POAF
No	4669	228	1.0	ref.	224	1.0	ref.	408	1.0	ref.
Yes	2152	192	1.8	(1.6–2.0)	172	1.4	(1.2–1.6)	298	1.4	(1.3–1.6)
Age (years)
≤ 60	1629	50	1.0	ref.	21	1.0	ref.	54	1.0	ref.
61–70	2536	119	1.6	(1.3–2.0)	116	3.9	(3.4–4.3)	209	2.8	(2.5–3.1)
71–80	2458	230	4.2	(3.9–4.5)	231	9.9	(9.5–10)	402	6.9	(6.6–7.2)
> 80	198	21	8.9	(8.4–9.4)	28	24	(24–25)	41	15	(15–16)
Hypertension
No	3345	192	1.0	ref.
Yes	3476	228	1.2	(1.1–1.4)
Diabetes
No	5365	292	1.0	ref.	275	1.0	ref.	512	1.0	ref.
Yes	1456	128	2.1	(1.9–2.4)	121	2.2	(2.0–2.4)	194	1.9	(1.7–2.0)
Number of prior MIs
≤ 1	5925							533	1.0	ref.
2	682							127	1.4	(1.2–1.6)
3	163							33	1.5	(1.1–1.8)
≥ 4	51							13	2.9	(2.4–3.5)
Diseased coronary vessels
1–2	1726	70	1.0	ref.
3	5095	350	1.4	(1.1–1.7)
Left main stenosis
No	4602							470	1.0	ref.
Yes	2219							236	1.2	(1.0–1.3)
LVEF
Good (> 50%)	4732	211	1.0	ref.	226	1.0	ref.	329	1.0	ref.
Reduced (30–50%)	1798	171	2.1	(1.9–2.3)	150	1.7	(1.5–1.9)	310	2.3	(2.2–2.5)
Poor (< 30%)	291	38	4.0	(3.6–4.3)	20	1.8	(1.4–2.3)	67	4.1	(3.9–4.4)

CABG, coronary artery bypass grafting; HR, hazard ratio; LVEF, left ventricular ejection fraction; MI, myocardial infarction; POAF, postoperative atrial fibrillation.

^aMedian (25th to 75th pctl). All variables in Table I were considered, and the final analyses adjusted for ^bage, hypertension, diabetes, diseased coronary vessels, and left ventricular function, ^cage, diabetes, and left ventricular function, and ^dage, diabetes, number of prior myocardial infarctions, left main stenosis, and left ventricular function.

Interaction analyses revealed no significant interactions between the independent risk factors and PO AF for any of the four categories of mortality.

An episode of POAF in relation to the CABG procedure was associated with increased mortality in all categories during the three examined time-periods (Table V). The effect of POAF remained over time, with a 30–80% increase more than 8 years after surgery (Table V).

Table V. Effect of POAF on late mortality over time. Follow-up 10 (6–13) years.^a

	< 3 years after CABG				3–8 years after CABG				> 8 years after CABG
	POAF	No POAF	HR	95% CI	POAF	No POAF	HR	95% CI	POAF	No POAF	HR	95% CI
Patients at risk	2152	4669			2005	4480			1312	3018
Primary cause of death
Cardiac mortality	72	69	2.0	(1.7–2.3)	126	180	1.3	(1.1–1.5)	182	300	1.3	(1.1–1.5)
Primary or contributing cause of death:
Arrhythmia	17	24	1.5	(0.9–2.2)	63	66	1.8	(1.4–2.1)	111	139	1.8	(1.6–2.1)
Cerebrovascular disease	30	29	1.9	(1.3–2.4)	66	83	1.4	(1.1–1.8)	75	113	1.4	(1.1–1.7)
Heart failure	34	33	1.9	(1.4–2.4)	95	129	1.3	(1.0–1.5)	169	246	1.5	(1.3–1.7)

CABG, coronary artery bypass grafting; HR, hazard ratio; POAF, postoperative atrial fibrillation.

^aMedian (25th to 75th pctl).

In a modified cohort of patients without other major complications (n = 5555), the effect of POAF on death in all four mortality categories was similar to the total cohort. After multivariable adjustment, POAF was associated with increased cardiac mortality (HR 1.4; 1.3–1.6), arrhythmia-related death (HR 1.8; 1.6–2.0), cerebrovascular disease-related death (HR 1.4; 1.1–1.6), and heart failure-related death (HR 1.5; 1.3–1.7) in the modified cohort.

Discussion

An episode of POAF in relation to CABG was associated with increased late cardiac mortality, and specifically death related to arrhythmia, cerebrovascular disease, and heart failure. The effect of POAF on survival was independent of established risk factors, including medical history, and preoperative status. More than 8 years after surgery POAF was still associated with an increased mortality between 30% and 80%.

This study was one of the largest studies investigating the effect of POAF on cause of death, and added data on contributing cause of death to explain the reason behind the negative impact of POAF after CABG. The incidence of POAF was in the upper range of other reports after CABG (19% to 31%) (2–11), indicating that the method used for arrhythmia recording was sufficient. Differences in observation time can have led to an information bias, but were not correlated to the presence of other complications.

The analysis of multiple causes of death, that is both primary and contributing causes, is a methodological approach for improving knowledge on the conditions examined and providing an accurate understanding of the causal complexity involved (12). The cause of death was unbiased regarding POAF, as the occurrence was unknown when the cause of death was registered. The registration of causes of death, perhaps especially contributing causes, involves an interpretation that can be partly subjective. This leads to a risk of random information bias. Although the cause of death obtained from the Swedish registry did not contain any missing data, the cause of death is not always based on autopsy. However, the quality of the registry is continually assessed (17).

The negative effect of POAF on cardiovascular survival was confirmed (9–11). In the prior studies concerning cause of death, analyses were unadjusted for age and co-morbidities (9,10) or were not large enough to permit analyses on subcategories of cardiac death (11). In this large patient cohort, a subdivision was possible and provides a more thorough picture of how POAF modified the effect on mortality and cause of death, with focus on causes of death plausibly connected to POAF.

As hypothesized, this study showed that arrhythmia-related death was higher in the POAF group, with a 50% higher mortality during the first 3 years following surgery. There was no tendency for the negative effect of POAF on survival to diminish over time. In fact, an episode of POAF at surgery almost doubled the risk of arrhythmia-related death more than 8 years after CABG. After CABG, POAF has previously been associated with later development of AF 5 years postoperatively (9). It may be later occurring episodes of AF that effect survival.

The increased incidence of death related to cerebrovascular disease in the POAF group might also be a consequence of recurring arrhythmias: 20% of patients with cerebrovascular disease-related death also suffered from arrhythmia as a contributing cause of death. POAF has previously been independently associated with a higher risk of delayed stroke after CABG (13,14).

The effect of POAF on heart failure-related death was most prominent during the first years after surgery, but was still significant during later time periods. AF and heart failure are known to co-occur and influence mortality in non-surgical populations (18). The persistent effect of POAF on heart failure-related death may, at least partly, be connected to reoccurring arrhythmias.

Our findings do not prove the notion of a direct causal attribution. In order to propose causality, a strong biologic assumption of potential mechanisms would be mandatory, and whether a decreased incidence of POAF would increase survival after CABG requires randomized, intervention trials.

Analyses concerning morbidity during the period between hospital discharge and death could contribute to clarifying whether the relationship between POAF and mortality is due to an increased incidence of late arrhythmia. Information regarding medication, especially related to POAF, such as anti-arrhythmic drugs and anti-coagulants, might also add important knowledge. We adjusted for a number of available risk factors, but pathology not detected preoperatively may still have influenced the results.

Analyses restricted to uncomplicated cases, that is patients without other major postoperative complications, were performed to verify that any influence on long-term survival appeared driven by POAF and not its association to other postoperative complications. As the effect of POAF was the same in the uncomplicated cohort, the associations could be assumed specific to POAF. Thus, the hypothesis that the negative influence of POAF is driven by the increased occurrence of other complications could be rejected.

In this study at least 95% of the POAF patients were discharged in SR. This implies that POAF is to be considered as an episodic event and not a beginning of a permanent arrhythmia. The finding that the effect of POAF remained the same on otherwise uncomplicated cases, further strengthened the hypothesis that the negative effect of POAF is valid also in patients with a transient episode POAF as the only complication after CABG.

In conclusion, an episode of POAF is associated with increased late mortality after CABG, specifically death related to arrhythmia, cerebrovascular disease, and heart failure, which have a plausible relationship to POAF. The effect of POAF was independent of other important risk factors (e.g. age, diabetes, and left ventricular function) and was still present more than 8 years after the surgical procedure. Based on these findings, possible treatments and follow-up should be based on a more long-term approach and not solely on early interventions, with the risk of overlooking later-occurring POAF-related mortality. Our findings suggest that the occurrence of POAF is not merely a sign of more advanced disease at the time of surgery, but predicts a persistent negative effect on cause-specific survival after CABG.

Declaration of interest: The authors report no declarations of interest. The authors alone are responsible for the content and writing of the paper.