Abstract
Objectives: This study evaluated the predictive performance of age, creatinine, and ejection fraction (ACEF) I and II scores for the development of postoperative atrial fibrillation (PoAF) after isolated on-pump coronary artery bypass grafting (CABG) surgery and compared them with a novel nomogram model developed for PoAF prediction. Subjects and Methods: This retrospective multicenter study involved 511 patients who underwent isolated on-pump CABG. Their ACEF scores were calculated, and multivariate logistic regression analysis was performed to develop a nomogram model. The discriminative performance of the ACEF scores and the novel nomogram model was assessed using the area under the receiver operating characteristic curve (AUC). Results: Of the 511 patients, 169 (33.1%) developed PoAF. The ACEF I and II scores showed moderate discriminative ability (AUC = 0.642 and 0.647, respectively), with no significant difference between them (p = 0.787). Logistic regression analyses identified age, preoperative hemoglobin levels, emergency procedure, chronic kidney disease or need for dialysis, preoperative β-blocker use, preoperative angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use, inotrope requirement, postoperative stroke, and postoperative potassium levels as independent predictors of PoAF. The novel nomogram model demonstrated greater predictive ability than the ACEF scores (AUC = 0.742, p < 0.001). Conclusion: ACEF scores could be helpful risk stratification tools for PoAF after on-pump CABG procedures. Additional validation studies are required to confirm their clinical utility in diverse surgical procedures and patient populations.
Age, creatinine, and ejection fraction (ACEF) I and II scores can predict atrial fibrillation after on-pump coronary artery bypass grafting procedures with moderate accuracy.
A novel nomogram model with ten variables was developed, which showed better discriminative ability than both ACEF scores.
These findings suggest potential applications of ACEF scores in preoperative risk assessment and patient management.
Introduction
Atrial fibrillation (AF) is a common arrhythmia after cardiac surgery, with varying rates based on surgery type. It can occur in 15%–40% of patients after coronary artery bypass grafting (CABG) surgery and 33%–49% of patients after valve surgery [1, 2]. Various factors, such as oxidative stress, inflammation, electrical remodeling, pain, electrolyte imbalance, and ischemia, can contribute to postoperative AF (PoAF) [3, 4]. While many cases may be asymptomatic and self-limiting, PoAF may sometimes be associated with stroke, myocardial infarction, and death, with an increased risk of permanent AF within 5 years after surgery [5]. The factors mentioned above increase the susceptibility of the myocardium to AF development in the postoperative period. Therefore, clinicians require an accessible risk score based on daily parameters to predict PoAF development. The age, creatinine, and ejection fraction (ACEF) I score includes three factors that can be considered reasonably practical and were originally employed to evaluate the risk of mortality in elective cardiac surgeries [6]. Over time, this scoring system has been validated in various procedures and clinical scenarios in the field of cardiovascular interventions, including AF recurrence after ablation [7‒12]. Later, preoperative anemia and emergency surgery were incorporated into the ACEF model, and the updated ACEF II score demonstrated superiority over the original ACEF I score [13].
Previous studies have reported that advanced age and existing comorbidities such as renal and heart failure significantly increase the risk of PoAF after CABG [14]. Since these parameters are the same as those incorporated into the ACEF I and II models, both scores may predict PoAF development after CABG. Therefore, this study assessed whether ACEF I and II scores can predict the risk of PoAF development in patients after isolated on-pump CABG.
Subjects and Methods
This observational study was conducted according to the guidelines outlined in the STROBE Statement and Checklist for observational studies [15].
Study Participants and Study Design
This multicenter retrospective study enrolled 511 consecutive patients who underwent isolated on-pump CABG procedures between January 2023 and December 2023. The exclusion criteria were as follows: (i) off-pump CABG or combined procedures, (ii) preoperative AF rhythm or other atrial/ventricular arrhythmias, (iii) preoperative hemodynamic instability, (iv) a history of thyroid disorders, (v) a history of previous cardiac surgeries, (vi) reinterventions due to cardiac arrest or tamponade or other reasons during the in-hospital follow-up period, (vii) left atrial thrombus on the preoperative echocardiography, and (viii) missing data. After obtaining approval from the Local Ethics Committee, the data of patients in the Cardiovascular Surgery Departments of Bozok University Hospital, Ankara City Hospital, and Tokat GOP University Hospital were obtained from their files and medical records. Patients’ demographic characteristics, preoperative comorbidity factors, perioperative laboratory values, echocardiography findings, intraoperative parameters, postoperative outcomes, and ACEF I and II scores were recorded. Patients were divided into two groups: the PoAF group consisted of those who developed PoAF after surgery during the hospital stay, and the non-PoAF group included patients who did not develop PoAF. The primary endpoint was to evaluate the ability of ACEF I and II scores to predict PoAF development and their discrimination performance. The secondary endpoint was developing a novel nomogram model to predict PoAF and comparing its predictive accuracy to the ACEF I and II scores.
PoAF Diagnosis
AF was diagnosed based on the guidelines of the European Society of Cardiology [16]. AF was recognized when a standard 12-lead electrocardiography (ECG) recording or a single-lead ECG tracing lasting more than 30 s showed a cardiac rhythm with irregular RR intervals and undetectable recurrent P waves. During the in-hospital follow-up, patients received daily 12-lead ECGs with additional recordings if they reported symptoms such as palpitations or dyspnea or if rhythm irregularities were suspected during telemetry monitoring.
ACEF I and II Score Calculations
The ACEF I score was calculated using the method established by Ranucci et al. [6] (age/left ventricular ejection fraction [LVEF] + 1 point for serum creatinine >2 mg/dL). The ACEF II score was calculated using the following method: age/LVEF + 2 points for serum creatinine level >2 mg/dL + 3 points for emergency surgery + 0.2 points × (36% − hematocrit) [13]. The formulas used values from the preoperative data.
Statistical Analysis
The data are presented as frequencies and percentages for categorical variables and as means with standard deviations or medians with interquartile ranges for numerical variables. The distribution of the variables was analyzed using the Kolmogorov-Smirnov test. Normally and non-normally distributed numerical variables were compared between groups using an independent sample t test or Mann-Whitney test, respectively, and categorical variables were compared using a chi-square test or Fisher’s exact test. A multivariate logistic regression analysis was conducted to identify independent predictors of PoAF after CABG. The included variables were selected based on a comprehensive review of relevant literature and the availability of data in our study. All available preoperative, intraoperative, and postoperative variables were included in the model. ACEF I and II scores were not included in the model to avoid multicollinearity and were analyzed separately. The stepwise selection method was used to add and remove variables based on their significance (p < 0.05 and p > 0.10, respectively). The Hosmer-Lemeshow test was used to assess the model’s goodness-of-fit, which indicated an adequate fit (p = 0.333). The discrimination abilities of the model and the ACEF I and II scores in predicting PoAF were assessed using receiver operating characteristic (ROC) curves and the areas under the ROC curve (AUCs) and compared using a nonparametric approach. A nomogram was created to calculate PoAF probabilities using the regression model. A p value of < 0.05 was considered statistically significant. The statistical analyses were performed using SPSS Statistics for Windows (version 25.0; IBM Corp., Armonk, NY, USA), and the ROC curve analyses, and nomogram model creation were conducted using Stata (release 17; Stata Corp LLC, College Station, TX, USA).
Results
This study included 511 patients, of which 169 (33.1%) developed PoAF and while 342 (66.9%) did not develop PoAF. There were 133 (78.7%) males in the PoAF group and 251 (73.4%) males in the non-PoAF group. The mean age was significantly higher in the PoAF group than in the non-PoAF group (65.3 vs. 61.7 years, p < 0.001). The median European System for Cardiac Operative Risk Evaluation (EuroSCORE) II was also significantly higher in the PoAF group than in the non-PoAF group (0.8 vs. 0.7, p = 0.001). There was a significant difference between groups in procedure priority (p = 0.001), with higher PoAF rates in urgent and emergency cases. There was a higher incidence of patients with preoperative chronic kidney disease (CKD) in the PoAF group than in the non-PoAF group (5.9% vs. 1.5%, p = 0.009). The LVEF differed significantly between groups (p = 0.012), although the difference was not clinically significant. In addition, the median ACEF I and II scores were significantly higher in the PoAF group than in the non-PoAF group (1.4 vs. 1.1 [p < 0.001] and 1.3 vs. 1.2 [p < 0.001], respectively). The patients’ baseline characteristics and preoperative laboratory parameters are presented in Table 1.
Baseline characteristics, comorbidities, and preoperative laboratory parameters of both groups
. | Total, n = 511 . | PoAF, n = 169 (33.1%) . | Non-PoAF, n = 342 (66.9%) . | p value . |
---|---|---|---|---|
Age, mean (±SD), years | 62.9 (8.4) | 65.3 (7.8) | 61.7 (8.5) | <0.001 |
Gender, n (%) | ||||
Male | 344 (75.1) | 133 (78.7) | 251 (73.4) | 0.192 |
Female | 127 (24.9) | 36 (21.3) | 91 (26.6) | |
BMI, median (IQR), kg/m2 | 27.9 (4) | 27.9 (3.2) | 27.9 (5) | 0.593 |
Euroscore II, median (IQR), % | 0.7 (0.4) | 0.8 (0.4) | 0.7 (0.4) | 0.001 |
HT, n (%) | 309 (61) | 109 (64.5) | 200 (58.5) | 0.191 |
HL, n (%) | 214 (42) | 71 (42) | 143 (41.8) | 0.966 |
DM, n (%) | 222 (43.4) | 82 (48.5) | 140 (40.9) | 0.104 |
Smoking, n (%) | 218 (43) | 77 (45.6) | 141 (41.2) | 0.351 |
COPD, n (%) | 126 (25) | 49 (29) | 77 (22.5) | 0.110 |
Peripheral artery disease, n (%) | 41 (8) | 11 (6.5) | 30 (8.8) | 0.376 |
CKD or requirement for dialysis, n (%) | 15 (2.9) | 10 (5.9) | 5 (1.5) | 0.009 |
History of stroke/TIA, n (%) | 21 (4.1) | 8 (4.7) | 13 (3.8) | 0.617 |
Perioperative MI | 80 (16) | 33 (19.5) | 47 (13.7) | 0.090 |
Priority of procedure, n (%) | ||||
Elective | 388 (76) | 118 (69.8) | 270 (78.9) | 0.001 |
Urgent | 114 (22.3) | 43 (25.4) | 71 (20.8) | |
Emergency | 9 (2) | 8 (4.7) | 1 (0.3) | |
β-Blocker therapy | 273 (53.4) | 84 (49.7) | 189 (55.3) | 0.236 |
ARB/ACE-I therapy | 183 (36) | 77 (45.6) | 106 (31) | 0.001 |
Statins | 133 (26) | 48 (28.4) | 85 (24.9) | 0.390 |
Laboratory parameters | ||||
Creatinine, median (IQR), mg/dL | 0.9 (0.3) | 0.9 (0.3) | 0.9 (0.2) | 0.078 |
Platelet, mean (±SD), ×109/L | 234 (67.7) | 242.1 (81.2) | 229.7 (59.6) | 0.079 |
Neutrophil, median (IQR), ×109/L | 5.1 (2.7) | 5.1 (2.6) | 5.1 (2.7) | 0.549 |
Lymphocyte, median (IQR), ×109/L | 2.2 (2) | 2.2 (1.9) | 2.2 (1.9) | 0.996 |
Hemoglobin, mean (±SD), g/dL | 13.3 (1.6) | 13 (1.7) | 13.5 (1.6) | 0.005 |
Hematocrit %, mean (±SD) | 39.8 (4.8) | 39.2 (4.8) | 39.6 (4.8) | 0.052 |
TSH level, median (IQR), mU/L | 1.3 (1.1) | 1.2 (1) | 1.3 (1.1) | 0.095 |
Albumin, median (IQR), g/dL | 42 (5) | 42 (4) | 43 (5) | 0.023 |
C-reactive protein, median (IQR), mg/L | 5 (14.7) | 6 (16.5) | 5 (14) | 0.243 |
Echocardiography features | ||||
LVEF, median (IQR), % | 55 (10) | 55 (15) | 55 (10) | 0.012 |
SPAP, median (IQR), mm Hg | 26 (7) | 26 (7) | 25 (6) | 0.216 |
LVEDD, mean (±SD), mm | 45.8 (7.2) | 47 (7) | 45 (7.2) | 0.005 |
LAD, median (IQR), mm | 37 (5) | 38 (5) | 37 (5) | 0.198 |
Risk models | ||||
ACEF I, median (IQR) | 1.2 (0.4) | 1.4 (0.4) | 1.1 (0.3) | <0.001 |
ACEF II, median (IQR) | 1.2 (0.5) | 1.3 (0.7) | 1.2 (0.4) | <0.001 |
. | Total, n = 511 . | PoAF, n = 169 (33.1%) . | Non-PoAF, n = 342 (66.9%) . | p value . |
---|---|---|---|---|
Age, mean (±SD), years | 62.9 (8.4) | 65.3 (7.8) | 61.7 (8.5) | <0.001 |
Gender, n (%) | ||||
Male | 344 (75.1) | 133 (78.7) | 251 (73.4) | 0.192 |
Female | 127 (24.9) | 36 (21.3) | 91 (26.6) | |
BMI, median (IQR), kg/m2 | 27.9 (4) | 27.9 (3.2) | 27.9 (5) | 0.593 |
Euroscore II, median (IQR), % | 0.7 (0.4) | 0.8 (0.4) | 0.7 (0.4) | 0.001 |
HT, n (%) | 309 (61) | 109 (64.5) | 200 (58.5) | 0.191 |
HL, n (%) | 214 (42) | 71 (42) | 143 (41.8) | 0.966 |
DM, n (%) | 222 (43.4) | 82 (48.5) | 140 (40.9) | 0.104 |
Smoking, n (%) | 218 (43) | 77 (45.6) | 141 (41.2) | 0.351 |
COPD, n (%) | 126 (25) | 49 (29) | 77 (22.5) | 0.110 |
Peripheral artery disease, n (%) | 41 (8) | 11 (6.5) | 30 (8.8) | 0.376 |
CKD or requirement for dialysis, n (%) | 15 (2.9) | 10 (5.9) | 5 (1.5) | 0.009 |
History of stroke/TIA, n (%) | 21 (4.1) | 8 (4.7) | 13 (3.8) | 0.617 |
Perioperative MI | 80 (16) | 33 (19.5) | 47 (13.7) | 0.090 |
Priority of procedure, n (%) | ||||
Elective | 388 (76) | 118 (69.8) | 270 (78.9) | 0.001 |
Urgent | 114 (22.3) | 43 (25.4) | 71 (20.8) | |
Emergency | 9 (2) | 8 (4.7) | 1 (0.3) | |
β-Blocker therapy | 273 (53.4) | 84 (49.7) | 189 (55.3) | 0.236 |
ARB/ACE-I therapy | 183 (36) | 77 (45.6) | 106 (31) | 0.001 |
Statins | 133 (26) | 48 (28.4) | 85 (24.9) | 0.390 |
Laboratory parameters | ||||
Creatinine, median (IQR), mg/dL | 0.9 (0.3) | 0.9 (0.3) | 0.9 (0.2) | 0.078 |
Platelet, mean (±SD), ×109/L | 234 (67.7) | 242.1 (81.2) | 229.7 (59.6) | 0.079 |
Neutrophil, median (IQR), ×109/L | 5.1 (2.7) | 5.1 (2.6) | 5.1 (2.7) | 0.549 |
Lymphocyte, median (IQR), ×109/L | 2.2 (2) | 2.2 (1.9) | 2.2 (1.9) | 0.996 |
Hemoglobin, mean (±SD), g/dL | 13.3 (1.6) | 13 (1.7) | 13.5 (1.6) | 0.005 |
Hematocrit %, mean (±SD) | 39.8 (4.8) | 39.2 (4.8) | 39.6 (4.8) | 0.052 |
TSH level, median (IQR), mU/L | 1.3 (1.1) | 1.2 (1) | 1.3 (1.1) | 0.095 |
Albumin, median (IQR), g/dL | 42 (5) | 42 (4) | 43 (5) | 0.023 |
C-reactive protein, median (IQR), mg/L | 5 (14.7) | 6 (16.5) | 5 (14) | 0.243 |
Echocardiography features | ||||
LVEF, median (IQR), % | 55 (10) | 55 (15) | 55 (10) | 0.012 |
SPAP, median (IQR), mm Hg | 26 (7) | 26 (7) | 25 (6) | 0.216 |
LVEDD, mean (±SD), mm | 45.8 (7.2) | 47 (7) | 45 (7.2) | 0.005 |
LAD, median (IQR), mm | 37 (5) | 38 (5) | 37 (5) | 0.198 |
Risk models | ||||
ACEF I, median (IQR) | 1.2 (0.4) | 1.4 (0.4) | 1.1 (0.3) | <0.001 |
ACEF II, median (IQR) | 1.2 (0.5) | 1.3 (0.7) | 1.2 (0.4) | <0.001 |
PoAF, postoperative atrial fibrillation; IQR, interquartile range; SD, standard deviation; BMI, body mass index; EuroSCORE, European System for Cardiac Operative Risk Evaluation; TIA, transient ischemic attack; MI, myocardial infraction; ARB/ACE-I, angiotensin receptor blocker/angiotensin-converting enzyme inhibitor; CKD, chronic kidney disease; HT, hypertension; HL, hyperlipidemia; DM, diabetes mellitus; COPD, chronic obstructive pulmonary disease; TSH, thyroid stimulating hormone; LVEF, left ventricular ejection fraction; SPAP, systolic pulmonary artery pressure; LVEDD, left ventricle end-diastolic diameter; LAD, left atrial diameter.
Table 2 presents a detailed analysis of the intraoperative and postoperative findings and laboratory parameters in both groups. The PoAF group had a significantly longer median total perfusion time than the non-PoAF group (144.8 vs. 105 min, p = 0.029). Additionally, the need for intra-aortic balloon pump and inotropic support was significantly greater in the PoAF group than in the non-PoAF group (p = 0.001 and 0.003, respectively). Furthermore, the incidence of cerebrovascular accidents, pulmonary complications, intensive care unit (ICU) and hospital stay lengths, and mortality rates were significantly higher in the PoAF group than in the non-PoAF group (p < 0.05).
Intraoperative and postoperative outcomes and laboratory parameters of both groups
. | Total, n = 511 . | PoAF, n = 169 (33.1%) . | Non-PoAF, n = 342 (66.9%) . | p value . |
---|---|---|---|---|
Cross-clamp time, mean (±SD), min | 75.3 (21.7) | 77.5 (23.4) | 74.1 (20.8) | 0.096 |
Total perfusion time, median (IQR), min | 107 (39) | 114.8 (44) | 105 (35) | 0.029 |
IABP requirement, n (%) | 43 (8.4) | 24 (14.2) | 19 (5.6) | 0.001 |
Inotrope requirement, n (%) | 196 (38.4) | 80 (47.3) | 116 (33.9) | 0.003 |
Number of diseased vessels, median (IQR) | 3 (2) | 3 (2) | 3 (2) | 0.072 |
RCA lesion, n (%) | 297 (58.1) | 99 (58.6) | 198 (57.9) | 0.883 |
Cardioplegia solution, n (%) | ||||
Blood cardioplegia | 267 (52.3) | 92 (54.4) | 175 (51.2) | 0.487 |
Del Nido cardioplegia | 244 (47.7) | 77 (45.6) | 167 (48.8) | |
Laboratory parameters | ||||
Creatinine, median (IQR), mg/dL | 0.9 (0.4) | 1 (0.5) | 0.9 (0.3) | 0.021 |
Platelet, median (IQR), ×109/L | 176 (70) | 173 (77) | 178 (71) | 0.125 |
Neutrophil, median (IQR), ×109/L | 10 (3.7) | 10 (4.9) | 9.9 (3.6) | 0.039 |
Lymphocyte, median (IQR), ×109/L | 0.7 (0.6) | 0.8 (0.6) | 0.7 (0.6) | 0.232 |
Hemoglobin, median (IQR), gr/dL | 9.7 (1.5) | 9.6 (1.5) | 9.8 (1.7) | 0.275 |
Hematocrit, median (IQR), % | 28 (4) | 28.5 (5) | 29 (4) | 0.314 |
Potassium, median (IQR), mEq/L | 4.4 (0.6) | 4.3 (0.7) | 4.4 (0.5) | 0.022 |
Magnesium, median (IQR), mg/dL | 2 (0.4) | 2 (0.5) | 2 (0.3) | 0.682 |
Complications | ||||
Stroke/TIA, n (%) | 15 (2.9) | 12 (7.1) | 3 (0.9) | <0.001 |
Pulmonary complications, n (%) | 30 (5.9) | 16 (9.5) | 14 (4.1) | 0.015 |
Renal dysfunction, n (%)a | 26 (5.1) | 13 (7.7) | 13 (3.8) | 0.060 |
ICU length of stay, median (IQR), days | 2 (2) | 3 (2) | 2 (2) | <0.001 |
Hospital length of stay, median (IQR), days | 7 (4) | 8 (4) | 6 (4) | <0.001 |
Mortality, n (%) | 24 (4.7) | 15 (8.9) | 9 (2.6) | 0.002 |
. | Total, n = 511 . | PoAF, n = 169 (33.1%) . | Non-PoAF, n = 342 (66.9%) . | p value . |
---|---|---|---|---|
Cross-clamp time, mean (±SD), min | 75.3 (21.7) | 77.5 (23.4) | 74.1 (20.8) | 0.096 |
Total perfusion time, median (IQR), min | 107 (39) | 114.8 (44) | 105 (35) | 0.029 |
IABP requirement, n (%) | 43 (8.4) | 24 (14.2) | 19 (5.6) | 0.001 |
Inotrope requirement, n (%) | 196 (38.4) | 80 (47.3) | 116 (33.9) | 0.003 |
Number of diseased vessels, median (IQR) | 3 (2) | 3 (2) | 3 (2) | 0.072 |
RCA lesion, n (%) | 297 (58.1) | 99 (58.6) | 198 (57.9) | 0.883 |
Cardioplegia solution, n (%) | ||||
Blood cardioplegia | 267 (52.3) | 92 (54.4) | 175 (51.2) | 0.487 |
Del Nido cardioplegia | 244 (47.7) | 77 (45.6) | 167 (48.8) | |
Laboratory parameters | ||||
Creatinine, median (IQR), mg/dL | 0.9 (0.4) | 1 (0.5) | 0.9 (0.3) | 0.021 |
Platelet, median (IQR), ×109/L | 176 (70) | 173 (77) | 178 (71) | 0.125 |
Neutrophil, median (IQR), ×109/L | 10 (3.7) | 10 (4.9) | 9.9 (3.6) | 0.039 |
Lymphocyte, median (IQR), ×109/L | 0.7 (0.6) | 0.8 (0.6) | 0.7 (0.6) | 0.232 |
Hemoglobin, median (IQR), gr/dL | 9.7 (1.5) | 9.6 (1.5) | 9.8 (1.7) | 0.275 |
Hematocrit, median (IQR), % | 28 (4) | 28.5 (5) | 29 (4) | 0.314 |
Potassium, median (IQR), mEq/L | 4.4 (0.6) | 4.3 (0.7) | 4.4 (0.5) | 0.022 |
Magnesium, median (IQR), mg/dL | 2 (0.4) | 2 (0.5) | 2 (0.3) | 0.682 |
Complications | ||||
Stroke/TIA, n (%) | 15 (2.9) | 12 (7.1) | 3 (0.9) | <0.001 |
Pulmonary complications, n (%) | 30 (5.9) | 16 (9.5) | 14 (4.1) | 0.015 |
Renal dysfunction, n (%)a | 26 (5.1) | 13 (7.7) | 13 (3.8) | 0.060 |
ICU length of stay, median (IQR), days | 2 (2) | 3 (2) | 2 (2) | <0.001 |
Hospital length of stay, median (IQR), days | 7 (4) | 8 (4) | 6 (4) | <0.001 |
Mortality, n (%) | 24 (4.7) | 15 (8.9) | 9 (2.6) | 0.002 |
PoAF, postoperative atrial fibrillation; IQR, interquartile range; SD, standard deviation; RCA, right coronary artery; IABP, intra-aortic balloon pump; TIA, transient ischemic attack; ICU, intensive care unit.
aCreatinine level of >2 mg/dL.
Predictors of PoAF
Table 3 shows the findings of univariate and multivariate logistic regression analyses to identify predictors of PoAF. In the multivariate analysis excluding ACEF I and II scores, age, preoperative hemoglobin, emergency cases, preoperative CKD or dialysis requirement, angiotensin-converting enzyme inhibitor (ACE-I) or angiotensin receptor blocker (ARB) use, preoperative β-blocker use, inotrope requirement, postoperative stroke, and postoperative potassium levels were independent predictors of PoAF after isolated on-pump CABG (p < 0.05).
Logistic regression analysis for predictors of PoAF
. | Univariate analysis . | Multivariable analysis . | ||||||
---|---|---|---|---|---|---|---|---|
OR . | p value . | 95% CI of OR . | OR . | p value . | 95% CI of OR . | |||
Variables without risk scores | ||||||||
Age | 1.06 | <0.001 | 1.03 | 1.08 | 1.05 | <0.001 | 1.03 | 1.08 |
Gender (female) | 0.76 | 0.217 | 0.49 | 1.18 | 0.66 | 0.091 | 0.40 | 1.07 |
Preoperative hemoglobin | 0.85 | 0.006 | 0.76 | 0.95 | 0.86 | 0.020 | 0.75 | 0.98 |
Procedure priority | ||||||||
Elective (reference) | 1.00 | − | − | − | 1.00 | − | − | − |
Emergent | 1.41 | 0.122 | 0.91 | 2.19 | 1.42 | 0.175 | 0.86 | 2.35 |
Emergency | 18.39 | 0.006 | 2.27 | 148.73 | 13.71 | 0.018 | 1.56 | 120.63 |
Preoperative CKD or dialysis | 4.24 | 0.009 | 1.43 | 12.61 | 3.78 | 0.028 | 1.15 | 12.38 |
Preoperative ACE-I or ARB use | 1.87 | 0.001 | 1.28 | 2.73 | 1.97 | 0.003 | 1.26 | 3.09 |
Preoperative β blocker use | 0.78 | 0.189 | 0.54 | 1.13 | 0.56 | 0.010 | 0.36 | 0.87 |
Inotrope use | 1.80 | 0.002 | 1.24 | 2.63 | 1.57 | 0.045 | 1.01 | 2.43 |
Postoperative stroke | 8.64 | 0.001 | 2.40 | 31.06 | 4.51 | 0.032 | 1.14 | 17.89 |
Postoperative potassium | 0.61 | 0.007 | 0.43 | 0.87 | 0.53 | 0.002 | 0.36 | 0.80 |
Risk scores | ||||||||
ACEF I score | 4.82 | <0.001 | 2.71 | 8.57 | − | − | − | − |
ACEF II score | 2.16 | <0.001 | 1.60 | 2.92 | − | − | − | − |
. | Univariate analysis . | Multivariable analysis . | ||||||
---|---|---|---|---|---|---|---|---|
OR . | p value . | 95% CI of OR . | OR . | p value . | 95% CI of OR . | |||
Variables without risk scores | ||||||||
Age | 1.06 | <0.001 | 1.03 | 1.08 | 1.05 | <0.001 | 1.03 | 1.08 |
Gender (female) | 0.76 | 0.217 | 0.49 | 1.18 | 0.66 | 0.091 | 0.40 | 1.07 |
Preoperative hemoglobin | 0.85 | 0.006 | 0.76 | 0.95 | 0.86 | 0.020 | 0.75 | 0.98 |
Procedure priority | ||||||||
Elective (reference) | 1.00 | − | − | − | 1.00 | − | − | − |
Emergent | 1.41 | 0.122 | 0.91 | 2.19 | 1.42 | 0.175 | 0.86 | 2.35 |
Emergency | 18.39 | 0.006 | 2.27 | 148.73 | 13.71 | 0.018 | 1.56 | 120.63 |
Preoperative CKD or dialysis | 4.24 | 0.009 | 1.43 | 12.61 | 3.78 | 0.028 | 1.15 | 12.38 |
Preoperative ACE-I or ARB use | 1.87 | 0.001 | 1.28 | 2.73 | 1.97 | 0.003 | 1.26 | 3.09 |
Preoperative β blocker use | 0.78 | 0.189 | 0.54 | 1.13 | 0.56 | 0.010 | 0.36 | 0.87 |
Inotrope use | 1.80 | 0.002 | 1.24 | 2.63 | 1.57 | 0.045 | 1.01 | 2.43 |
Postoperative stroke | 8.64 | 0.001 | 2.40 | 31.06 | 4.51 | 0.032 | 1.14 | 17.89 |
Postoperative potassium | 0.61 | 0.007 | 0.43 | 0.87 | 0.53 | 0.002 | 0.36 | 0.80 |
Risk scores | ||||||||
ACEF I score | 4.82 | <0.001 | 2.71 | 8.57 | − | − | − | − |
ACEF II score | 2.16 | <0.001 | 1.60 | 2.92 | − | − | − | − |
“Stepwise” method used to identify predictive variables in the multivariate model. OR, odds ratio; 95% CI, 95% confidence interval; ACEF, age, creatinine, ejection fraction; ACE-I, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; CKD, chronic kidney disease.
p value of Hosmer and Lemeshow Test for multivariate model = 0.333.
Predictive Values of ACEF I and II Scores
The predictive values of ACEF I and II scores for PoAF are presented in Table 4. For ACEF I score, the identified cut-off was >1.12, with an AUC of 0.642 (95% confidence interval [CI]: 0.599–0.684, p < 0.001), sensitivity of 75.60%, and specificity of 48.24% (Fig. 1). For ACEF II scores, the identified cut-off was >1.14, with an AUC of 0.647 (95% CI: 0.603–0.688, p < 0.001), sensitivity of 79.76%, and specificity of 43.24% (Fig. 1). In addition, the z-test indicated no significant difference between the AUCs of both scores (z = 0.296, p = 0.787).
Predictive values of ACEF I and II scores
. | ACEF I . | ACEF II . |
---|---|---|
Cut-off value | >1.12 | >1.14 |
AUC (95% CI) | 0.642 (0.599–0.684) | 0.647 (0.603–0.688) |
Sensitivity, % (95% CI) | 75.60 (68.4–81.9) | 79.76 (72.9–85.6) |
Specificity, % (95% CI) | 48.24 (42.8–53.7) | 43.24 (37.9–48.7) |
Accuracy, % | 23.8 | 23 |
. | ACEF I . | ACEF II . |
---|---|---|
Cut-off value | >1.12 | >1.14 |
AUC (95% CI) | 0.642 (0.599–0.684) | 0.647 (0.603–0.688) |
Sensitivity, % (95% CI) | 75.60 (68.4–81.9) | 79.76 (72.9–85.6) |
Specificity, % (95% CI) | 48.24 (42.8–53.7) | 43.24 (37.9–48.7) |
Accuracy, % | 23.8 | 23 |
ACEF, age, creatinine, ejection fraction; AUC, area under the curve; CI, confidence interval.
Receiver operating characteristic (ROC) curve of predicting post-CABG atrial fibrillation for the mode and, ACEF I and II scores.
Receiver operating characteristic (ROC) curve of predicting post-CABG atrial fibrillation for the mode and, ACEF I and II scores.
The developed model showed a greater discriminative power, with an AUC of 0.742 (95% CI: 0.696–0.789, p < 0.001; Fig. 1). ACEF scores differed significantly from the model (both p < 0.001; Table 5).
Comparison of diagnostic performance of ACEF I, ACEF II, and model for predicting PoAF
Diagnostic element . | AUC . | 95% CI . | p value versus model . |
---|---|---|---|
Model | 0.742 | 0.695–0.788 | − |
ACEF I score | 0.642 | 0.599–0.684 | <0.001 |
ACEF II score | 0.647 | 0.603–0.688 | <0.001 |
Diagnostic element . | AUC . | 95% CI . | p value versus model . |
---|---|---|---|
Model | 0.742 | 0.695–0.788 | − |
ACEF I score | 0.642 | 0.599–0.684 | <0.001 |
ACEF II score | 0.647 | 0.603–0.688 | <0.001 |
ACEF, age, creatinine, ejection fraction; AUC, area under the curve; CI, confidence interval.
Nomogram Development
Based on parameters obtained from the logistic regression model, a nomogram was created for calculating PoAF probabilities after isolated on-pump CABG (Fig. 2). Each variable was assigned a score between 0 and 11, with a higher overall score indicating a greater risk of developing PoAF.
The nomogram model predicting PoAF after isolated on-pump CABG procedures.
Discussion
The ACEF I score was initially proposed to predict mortality risk after elective cardiac procedures. It has subsequently been applied in diverse clinical settings, including percutaneous coronary procedures [9], short and long outcomes in acute coronary syndrome [10, 11], and transcatheter aortic valve implantation [12]. The ACEF II score has also demonstrated a high predictive power in percutaneous coronary interventions and patients with aortic dissections [17, 18]. Both ACEF scores are significantly associated with an increased risk of subsequent adverse cardiovascular outcomes such as mortality. However, their predictive ability for PoAF in patients undergoing cardiac surgery has not been previously evaluated.
In our study, both ACEF I and II scores showed moderate discriminative ability, with AUCs of 0.642 and 0.647, respectively, indicating a fair level of accuracy in discriminating patients with higher and lower risks of developing PoAF. In addition, with the determined cut-offs, the ACEF II score demonstrated a slightly greater sensitivity (79.76%) than the ACEF I score (75.60%), suggesting that the ACEF II score is better at identifying patients who will develop PoAF. However, with the determined cut-offs, the ACEF I score demonstrated greater specificity (48.24%) than the ACEF II score (43.24%), indicating that the ACEF I score is better at accurately identifying patients who will not develop PoAF based on their preoperative laboratory parameters. Nevertheless, the overall accuracy of the ACEF I and II scores were moderate (23.8% and 23.0%, respectively), showing the inherent challenge of accurately predicting PoAF based only on preoperative parameters.
PoAF is likely related to a combination of perioperative factors [14, 19, 20]. Currently, there is no widely accepted risk score model to predict PoAF. Various models have been developed and validated to predict the likelihood of new-onset PoAF after cardiac surgery [20‒23]. Fan et al. [20] created a prediction model to quantify the risk of PoAF after isolated CABG using eight variables, including age, gender, hypertension, heart rate, LVEF <50%, left atrial diameter >40 mm, and on-pump procedures, which showed a moderate discriminative ability, with an AUC of 0.661. In addition, Gong et al. [23] developed a model based on seven variables: age, intraoperative operative time >4 h, preoperative insulin, statin use, elevated mean arterial pressure, body mass index >23 kg/m2, and left atrial diameter >40 mm. Their model demonstrated good performance, with an AUC of 0.727. In our study, we developed a model using a comprehensive set of preoperative, intraoperative, and postoperative variables, with the final model including ten variables. Our model showed a discriminative ability greater than those of the previously reported models, including the ACEF I and II scores, which were the primary comparators in our study, achieving an AUC of 0.742.
Previous studies have identified advanced age, impaired renal functions represented by creatinine levels, and impaired LVEF as significant risk factors for PoAF [14, 20]. In our study, age emerged as a notable factor, with each year increasing the risk of PoAF by 5.3% (odds ratio [OR] = 1.05). Similarly, CKD or preoperative dialysis requirement increased the risk of PoAF more than threefold (OR = 3.78). However, LVEF was not found to be an independent predictor of PoAF in our prediction model, which is consistent with another study [23]. Preoperative low hemoglobin levels are associated with an increased risk of developing PoAF, likely due to adrenergic activation and ischemic injury to atrial cells [24]. In our study, preoperative hemoglobin levels demonstrated a protective effect as higher levels were associated with reduced risk of PoAF (OR = 0.86). In addition, the urgency of the procedure also affected PoAF development, particularly in emergent cases (OR = 13.71). All these factors are key components of both ACEF scores, demonstrating their potential value as predictive scores for PoAF to guide preoperative risk assessments.
In addition to clinical risk factors, our findings emphasize the influence of perioperative management on PoAF development. Preoperative β-blockers use significantly reduced the risk of PoAF (OR = 0.56), consistent with prior studies that demonstrate their efficacy in controlling perioperative sympathetic stimulation [25, 26]. Conversely, inotropic support was found to increase the risk of developing PoAF (OR = 1.57), likely due to elevated adrenergic response that facilitates PoAF onset [27]. Preoperative ACE-I or ARB use nearly doubled the risk of PoAF (OR = 1.97), which could be explained by their effects on blood pressure regulation that may cause hemodynamic instability during surgery. Furthermore, they may increase susceptibility to arrhythmias by altering potassium and sodium levels, both of which are essential for cardiac function. Electrolyte abnormalities, particularly hypokalemia, are known to be important factors in the development of arrhythmias after cardiac surgery [27]. In our model, higher postoperative potassium levels were associated with a decreased risk of PoAF (OR = 0.53), indicating the importance of maintaining potassium homeostasis in preventing PoAF. Our findings shed light on the complex interaction of numerous risk factors that contribute to PoAF and emphasize the need to identify them to reduce the occurrence of this complication in patients after cardiac surgery.
Limitations and Strengths
Our study has several limitations and strengths. Its limitations include its retrospective design, which may introduce inherent biases due to reliance on existing data. In addition, AF diagnosis relied solely on ECG monitoring within the hospital setting, lacking follow-up post-discharge. Consequently, patients experiencing PoAF shortly after discharge might have been inadvertently classified in the non-PoAF group due to the absence of continuous monitoring. Moreover, including only patients undergoing on-pump CABG and strict exclusion criteria may limit the applicability of its findings to other patient populations. However, we excluded off-pump procedures to minimize potential confounding effects arising from differences in inflammatory response and myocardial injury. Its strength lies in its multicenter approach, which enhances the generalizability of its findings by including data from several institutions, increasing the diversity of patient groups and treatment approaches. In conclusion, our study presented a comprehensive predictive model for PoAF that demonstrates improved discriminative ability compared to the established ACEF I and II scores.
Conclusion
Our study revealed that ACEF I and II scores demonstrate moderate diagnostic performance for PoAF in patients after isolated on-pump CABG. It also presented a comprehensive predictive model for PoAF that demonstrates better discriminative performance than the ACEF I and II scores. While its findings are promising, it must be noted that the practical utility of both ACEF I and II scores and the developed nomogram in predicting PoAF requires further validation, particularly in diverse patient populations and surgical procedures, to confirm their predictive capability for clinical use.
Statement of Ethics
The study was approved by the Local Ethics Committee of Bozok University Hospital and performed in accordance with the Declaration of Helsinki (2024-GÖKAEK-242_24.04.2024_03). Patient consent was waived due to the retrospective nature of the analysis.
Conflict of Interest Statement
The authors have no conflicts of interest to declare.
Funding Sources
This study was not supported by any sponsor or funder.
Author Contributions
Sameh Alagha contributed to the concept and design of the methodology described in this article and analyzed the data. Sameh Alagha, Serkan Mola, Mehmet Çeber, and Alp Yıldırım enrolled patients. Serkan Mola and Mehmet Çeber wrote the first draft of the manuscript. All authors contributed to the final version of the manuscript and have read and confirmed that they meet the ICMJE criteria for authorship.
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author.