Introduction
Atrial fibrillation (AF), one of the most common supraventricular arrhythmias characterized by the rapid and irregular rhythm of the atria, currently affects more than 33 million people globally, and its prevalence is expected to more than double in the next 40 years [1]. Untimely and non-standard AF treatment significantly affects patients’ quality of life, including progressive heart failure, stroke, sudden cardiac death, and non-cardiovascular death [2]. Therefore, prompt intervention is crucial to avoid disease progression in patients with AF.
Despite the increasing types of surgical procedures available for AF ablation, surgery remains the most effective therapeutic strategy for AF, and the cornerstone of ablation is pulmonary venous isolation [3, 4]. The postoperative recurrence rate of PsAF is significantly higher than that of paroxysmal AF, and patients are more likely to experience cardiac dysfunction and stroke [5]. Debate among experts persists regarding whether persistent AF might cause severe atrial muscle fibrosis and atrial dysfunction, which would activate atrial reversion [6–8].
Only a few predictive models for postoperative recurrence have been developed on the basis of these factors, although numerous studies have analyzed the causes of postoperative recurrence, including the linear ablation method and site [9], atrial scar [10], sex [11], duration [6], sleep apnea syndrome [12], and left atrial diameter (LAD) [13]. However, nomograms are a commonly used diagnostic tool that can predict the probability of clinical events by incorporating potential risk factors, and can also be used to determine the clinical prognosis regarding disease outcomes. The objective of the study was to retrospectively analyze the risk factors for recurrence in patients with persistent atrial fibrillation. An additional aim was to analyze and establish an appropriate predictive model for postoperative recurrence, to improve the success rates of surgery and monitoring of patients at high risk, and to enhance patient quality of life and prognosis.
Methods and Patients
Patients
Patients with PsAF who were admitted to the cardiovascular medicine department at the Affiliated Changzhou No.2 People’s Hospital of Nanjing Medical University between June 2013 and June 2021 were examined in this retrospective study. The clinical characteristics of patients with PsAF were thoroughly reviewed and analyzed. The inclusion criteria were patients with 1) persistent atrial fibrillation lasting more than 7 days (consistent with the 2014 AHA guidelines) [14]; 2) no prior AF surgical history; and 3) electrocardiogram abnormalities corresponding to atrial fibrillation.
The exclusion criteria were 1) patients with severe pulmonary, cardiovascular, hepatic, or renal diseases; 2) patients with coagulation dysfunction, left atrium and left atrial appendage thrombosis; 3) patients with hypertensive heart disease, congenital heart disease, or valvular heart disease; 4) patients with rheumatic valvular diseases, immune-associated diseases, or malignant tumors; and 5) patients with missing information or who were unreachable.
Postoperative recurrence was defined by signs and symptoms of atrial tachycardia, atrial fibrillation, and atrial flutter 3 months after AF ablation (as confirmed by a definite electrocardiogram or dynamic electrocardiogram). The Ethics Review Committee of the Affiliated Changzhou No.2 People’s Hospital of Nanjing Medical University approved this study. Because this was a retrospective study, informed consent from eligible patients was not required.
Clinical characteristics
The electronic medical record system at the hospital was used to extract the fundamental clinical characteristics, laboratory test results, treatments, and outcomes of individuals. Sex, smoking status, alcohol consumption, age, body weight, and body mass index (BMI) were among the basic characteristics analyzed. Medical history included complications (such as hypertension, diabetes, hyperthyroidism, pulmonary hypertension, sleep apnea syndrome, acute myocardial infarction, coronary heart disease, stroke, syncope, and heart failure); calculated CHA2DS2-VASc Scores (two points each were assigned for age ≥75 years and for history of stroke, TIA, or thromboembolism, and 1 point was assigned for each of the following items: congestive heart failure, hypertension, diabetes mellitus, age 65–75 years, vascular disease (defined as previous myocardial infarction, complex aortic plaque, carotid stenosis, and peripheral artery disease), and female sex category); types of anticoagulants taken (warfarin, dabigatran, rivaroxaban, and others); whether antiarrhythmic medication (amiodarone) was taken; and whether mechanical electrocardiography was performed during the operation (if sinus rhythm was not restored at the completion of radiofrequency ablation, electrical cardioversion was performed on the operating table). Left atrial anterior and posterior diameter, ejection fraction (EF), left ventricular end-diastolic diameter (LVEDD), ventricular septal thickness (IVS), left ventricular diameter shortening fraction, and other cardiac ultrasound parameters were among the clinical and examination information included. Preoperative venous testing was conducted for laboratory indicators such as white blood cells, neutrophil ratio, red blood cell distribution width, creatine kinase, troponin, c-reactive protein, blood creatinine, brain natriuretic peptide (BNP), cholesterol, triglycerides, low-density lipoprotein cholesterol, thyroid-stimulating hormone, and three free iodine thyroid-related.
Catheter ablation procedure
After insertion of three catheters through the right femoral vein, a ten-pole catheter into the coronary sinus, two long shears into the atrial septum twice, a beacon electrode (PentaRay), and an ablation catheter (Johnson’s ST head), the procedure was completed. An electrode (PentaRay) for mapping was used to create the left atrium model. The site of radiofrequency ablation was isolated at a distance of 4 mm from the pulmonary vein entrance. The complete isolation of both bilateral pulmonary veins was considered the hallmark of successful surgery, and auxiliary ablation lines such as the left atrium, posterior wall, atrial low-voltage scar area, and even mitral isthmus were added when required. Electrical cardioversion was used to ensure sinus rhythm after surgery if the sinus rhythm of patients with PsAF was not restored after isolation. Anticoagulants were administered preoperatively for at least 3 weeks, with esophageal ultrasound, CT, or intracardiac ultrasound, to ensure the absence of thrombi in the atria. Anticoagulants were taken postoperatively for at least 3 months until reevaluation.
Patient follow-up and study endpoint
In addition to outpatient visits, all patients were followed up by telephone. Ambulatory follow-up was conducted every 3 months during the first year, and follow-up and examinations were conducted every 3–6 months thereafter. All patients were followed up for 1 year. The definition of recurrence was as stated above, but was supported by clear ECG or ambulatory ECG evidence.
Statistical analysis
Data analysis was conducted in SPSS version 22.0 (IBM, Chicago, IL, USA) and R (version 4.0.3), and a two-sided P-value <0.05 was considered statistically significant. Categorical variables were examined with the Pearson chi-square or Fisher exact test, and are expressed as frequency and percentage. Univariate analysis was used to compare the baseline characteristics of the recurrence group and non-recurrence group to assess independent predictors of recurrence. Multivariate logistic regression analysis was then used to further analyze significantly different variables. The final included variables were obtained with the stepwise forward method, and the variance expansion factor was calculated to determine whether each factor displayed multicollinearity. Finally, with these variables, a predictive model was developed.
To generate forest plots, we performed univariate and multivariate analyses with the “Forest Plot” software package in R Software (version 4.0.3). The original data were resampled with the bootstrap method, and the nomogram model was used to forecast recurrence after resampling. With the “rms” package, nomogram analysis for predicting recurrence in patients with PsAF integrated duration of AF episodes, LAD, BMI, CKMB, and alcohol consumption. Receiver operating characteristic curve analysis was performed with the “pROC” package software in R to evaluate the ability of the combined model to predict postoperative recurrence in patients with PsAF. Subsequently, the calibration curve and decision curve were plotted with the “ggDCA” package to verify our nomogram’s predictive accuracy and reliability.
Results
Patient Characteristics
Retrospective analysis was performed on all clinical baseline characteristics of the 238 patients with PsAF. Among them, 29 patients were excluded on the basis of the exclusion criteria, and 209 patients were included in this study (Figure 1). Table 1 summarizes the clinical data for the two groups of patients. After radiofrequency catheter ablation surgery, 42 (20.10%) patients were unable to maintain sinus rhythm continuously. The recurrence group consisted of 27 men and 15 women, with a median age of 66.42 years. The average body weight of patients in the recurrence group was 75.4±11.1 kg, and echocardiography revealed an average LAD of 49.4±4.6 mm; the group included 15 patients (35.7%) who experienced sustained atrial fibrillation for 3–6 years, 8 patients (19%) for more than 6 years; 29 patients (69%) who had hypertension, 21 patients (50%) who had a long history of drinking, and 31 patients (73.8%) who underwent electrical conversion during surgery.
Recurrence (n=42) | Non-recurrence (n=167) | P value | |
---|---|---|---|
Age, years | 66 (57–69) | 67 (57–72) | 0.77 |
Male, n(%) | 27 (64.3) | 107 (64.1) | 0.98 |
BMI (kg/m2) | 27.3 (24.7–29.1) | 23.6 (21.9–26.1) | <0.001 |
COPD, n(%) | 1 (2.4) | 4 (2.4) | 0.1 |
PAH, n(%) | 0 (0) | 2 (1.2) | 0.48 |
SAS, n(%) | 2 (4.8) | 5 (3) | 0.57 |
Smoke, n(%) | 17 (40.5) | 46 (27.7) | 0.1 |
Syncope, n(%) | 0 (0) | 1 (0.6) | 0.62 |
Drink, n(%) | 21 (50.0) | 44 (26.5) | 0.003 |
HBP, n(%) | 29 (69.0) | 90 (54.2) | 0.08 |
CHD, n(%) | 4 (9.5) | 20 (12) | 0.66 |
OMI, n(%) | 0 (0) | 3 (1.8) | 0.38 |
DM, n(%) | 8 (19.1) | 30 (17.9) | 0.87 |
Stroke, n(%) | 0 (0) | 8 (4.8) | 0.15 |
Hyperlipidemia, n(%) | 10 (23.8) | 33 (19.8) | 0.56 |
HF, n(%) | 9 (21.4) | 35 (21.0) | 0.82 |
Hyperthyroidism, n(%) | 2 (4.8) | 2 (1.2) | 0.13 |
Hypothyroidism, n(%) | 0 (0) | 1 (0.6) | 0.62 |
Prior stent implantation, n(%) | 1 (2.4) | 1 (0.6) | 0.3 |
CKMB (U/L) | 14.6 (12.4–16.5) | 13.1 (11.0–15.6) | 0.014 |
TNI (μg/L) | 0.01 (0.01–0.01) | 0.01 (0.01–0.01) | 0.92 |
BNP (pg/mL) | 546 (190–985) | 782 (374–1390) | 0.38 |
CR (μmol/L) | 85.9 (66.0–106.5) | 71.0 (63.0–86.0) | 0.01 |
TCH (mmol/L) | 3.7 (3.4–4.7) | 3.9 (3.5–4.6) | 0.32 |
TG (mmol/L) | 1.3 (1.0–2.1) | 1.3 (1.1–2.0) | 0.84 |
LDL (mmol/L) | 2.1±0.6 | 2.2±0.7 | 0.26 |
LVEDD (mm) | 51.0 (49.5–54.5) | 50.0 (47.0–54.0) | 0.05 |
LVPWT (mm) | 10.0 (9.3–10.0) | 9.0 (9.0–10.0) | 0.05 |
LVESD (mm) | 36.0 (34.5–37.5) | 35.0 (32.0–37.0) | 0.04 |
LAD (mm) | 49.4±4.6 | 42.1±5.3 | <0.001 |
LDFS (%) | 29.0 (27.5–31.5) | 31.0 (28.0–33.0) | 0.24 |
EF (%) | 55.0 (52.5–58.0) | 58.0 (54.0–61.3) | 0.1 |
IVS (mm) | 10.0 (10.0–11.0) | 10.0 (9.0–10.0) | 0.019 |
Electrical cardioversion, n(%) | 29 (70.7) | 88 (52.7) | 0.08 |
CHA2DS2-VASc Scores | 2.0(2.0–3.0) | 2.0(2.0–3.0) | 0.69 |
Antiarrhythmic drugs (Amiodarone), n(%) | 17(40.5) | 55(33.1) | 0.37 |
Duration grouping, n | 42 | 167 | <0.001 |
<1 year, n(%) | 13 (31.0) | 69 (41.6) | |
1–3 year, n(%) | 6 (14.3) | 72 (43.4) | |
3–6 year, n(%) | 15 (35.7) | 20 (12.0) | |
>6 year, n(%) | 8 (19.0) | 5 (3.0) | |
Anticoagulants, n | 42 | 167 | 0.04 |
Warfarin, n(%) | 14 (33.3) | 41 (24.6) | |
Dabigatran, n(%) | 14 (33.3) | 31 (18.6) | |
Rivaroxaban, n(%) | 11 (26.2) | 81 (48.5) | |
Others, n(%) | 3 (7.1) | 14 (8.4) |
BMI, body mass index; BNP, brain natriuretic peptide; CHD, coronary heart disease; CKMB, creatine kinase-mb; COPD, chronic obstructive pulmonary disease; CR, creatinine; DM, diabetes mellitus; EF, ejection fraction; HBP, high blood pressure; HF, heart failure; IVS, thickness of the interval; LAD, left atrial diameter; LDFS, the left indoor diameter shortens the fraction; LDL, low density lipoprotein; LVEDD, left ventricular end diastolic diameter; LVESD, left ventricular end contractile diameter; LVPWT, posterior wall thickness of the left ventricle; OMI, old myocardial infarction; PAH, pulmonary arterial hypertension; SAS, sleep apnea syndrome; CH, total cholesterol; TG, triglyceride.
Evaluation of Independent Prognostic Factors
Basic demographic factors, vital signs, and laboratory test results in the primary cohort were further analyzed with the univariate logistic model to identify potential indicators (Table 2). Univariate regression analysis (P<0.05) demonstrated that variables such as weight, BMI, TT, CKMB, LVESD, LAD, IVS, drinking history, CR, type of anticoagulant, and duration of AF episodes were potential predictors. Five prognostic factors – BMI, CKMB, LAD, drinking history, and duration of AF episodes – were included in the final predictive model (each P<0.05) after further analysis of all candidate factors with a multivariable logistic regression model (Table 2 and Figure 2). The forward stepwise variable selection method was used, wherein the variables were sequentially entered into the model, as shown in Figure 3.
Univariate model | Multivariable model | |||||
---|---|---|---|---|---|---|
OR | 95% CI | P value | OR | 95% CI | P value | |
Weight | 1.075 | 1.040–1.110 | <0.001 | 0.977 | 0.905–1.055 | 0.554 |
TT | 1.008 | 0.998–1.017 | 0.112 | |||
BMI | 1.327 | 1.174–1.500 | <0.001 | 1.181 | 1.006–1.386 | 0.043 |
CR | 1.021 | 1.003–1.040 | 0.023 | 1.023 | 0.995–1.052 | 0.113 |
CKMB | 1.093 | 1.018–1.174 | 0.014 | 1.098 | 1.006–1.386 | 0.037 |
LVESD | 1.045 | 0.970–1.126 | 0.242 | |||
LAD | 1.251 | 1.156–1.355 | <0.001 | 1.261 | 1.141–1.394 | <0.001 |
IVS | 1.416 | 1.017–1.970 | 0.039 | 1.012 | 0.908–1.128 | 0.832 |
Drink | 0.004 | 0.032 | ||||
No | 1 | 1 | ||||
Yes | 0.358 | 0.178–0.717 | 0.004 | 0.354 | 0.136–0.934 | 0.032 |
Duration grouping | <0.001 | <0.001 | ||||
<1 year | 1 | 1 | ||||
1–3 year | 0.118 | 0.033–0.417 | 0.001 | 0.073 | 0.015–0.362 | 0.001 |
3–6 year | 0.052 | 0.013–0.210 | <0.001 | 0.04 | 0.007–0.222 | <0.001 |
>6 year | 0.127 | 0.127–1.274 | 0.254 | 0.354 | 0.354–0.069 | 0.214 |
Anticoagulants | 0.05 | |||||
Warfarin | 1 | |||||
Dabigatran | 1.593 | 0.398–6.377 | 0.51 | |||
Rivaroxaban | 2.108 | 0.521–8.526 | 0.296 | |||
Others | 0.634 | 0.157–2.562 | 0.522 |
BMI, body mass index; CKMB, creatine kinase-mb; CR, creatinine; IVS, thickness of the interval; LAD, left atrial diameter; LVESD, left ventricular end contractile diameter; TT, thrombin time.
The inclusion of clinical variables improved the AUC from 0.814 to 0.906. The final model, consisting of LAD, duration of AF episodes, BMI, drinking history, and CKMB level, was a good predictor of postoperative recurrence in patients with PsAF.
Construction and Performance Evaluation of the Prognostic Nomogram
The duration of AF episodes, LAD, BMI, CKMB, and drinking history were used to construct a nomogram for predicting recurrence in patients with PsAF, as shown in Figure 4. Each independent prognostic parameter was given a weighted score to create the nomogram, and each predictor was then connected by a straight line to the points axis to represent a particular point. A higher recurrence probability was represented by a higher total score, which was calculated by summation of the points allotted to the nomogram’s prognostic factors.
To evaluate the nomogram’s discrimination ability, calibration, and clinical practicability, we performed receiver operating characteristic curve, C-index, calibration curve, and decision curve analyses with 1000 bootstrap resamplings. The C-index of the prognostic nomogram was 0.906, and the calibration curves for probabilities in recurrence prediction showed a good fit of the nomogram for recruitment prediction. The AUC was 0.895, and the sensitivity and specificity were 93.3% and 71.4%, respectively, thus implying that the nomogram model had good discrimination ability. Moreover, the nomogram predictive model was valuable in making effective judgments, and the decision curve analysis curve demonstrated its net benefit across a wide range of threshold probabilities (Figure 5).
Discussion
Catheter ablation is the mainstay therapy for PsAF. However, this procedure is limited by the high rate of long-term recurrence. The clinical data for 209 patients with PsAF were retrieved after database screening for this single-center retrospective investigation. To create a prognostic nomogram, we examined factors that significantly differed between the groups with and without recurrence. This novel nomographic chart showed satisfactory performance according to C-index, AUC, calibration curve, and decision curve analyses. Thus, the nomographic chart can be effectively implemented clinically.
We followed patients for 1 year after surgery: 42 of the 209 patients had AF recurrence, based on ECG or Holter data. In the analysis of recurrent PsAF, we identified five risk factors on the basis of statistics: BMI, CKMB, LAD, duration of AF episodes, and alcohol consumption history, in agreement with previous observations. Higher BMI has previously been shown to increase the risk of atrial fibrillation by at least 50%, and BMI is used as a proxy for obesity [15], possibly because of its effects on cardiac muscle structure by increasing oxidative stress or fat cell apoptosis [16, 17]. Furthermore, our statistical results revealed that high BMI was an independent risk factor and was more prevalent in the recurrence group. Moreover, most patients with AF were clinically obese. Atrial structural remodeling [18] is therefore a key etiology for persistent AF [1], thus leading to a series of physiological and pathological changes, because the complexity of atrial muscle conduction determines the electrophysiological basis of AF [19]. Furthermore, atrial electrical remodeling [1] and atrial fibrosis [18] affect redox signaling [20]. Studies on the effects of atrial diameter and AF duration on the recurrence rate in patients with AF have been conducted in centers worldwide. Both factors have been shown to affect cardiac function recovery and mortality in patients with AF and heart failure [21–23]. We observed that recurrence was more prevalent in patients with larger atrial size and longer duration of AF episodes, thus corroborating previous results. We divided the patients into four groups in our study, on the basis of AF episode duration. The recurrence rate was significantly higher in patients at 1–3 years and 3–6 years. Thus, our findings provide a solid basis and a useful tool for follow-up before and after surgery.
CKMB is a marker of myocardial injury and can also cause atrial inflammation [24]. Moreover, CKMB has been found to be associated with early postoperative recurrence [25] and to specifically predict the incidence of atrial fibrillation in patients with acute coronary syndrome within 3–5 years [26]. However, insufficient in-depth studies have been performed on this mechanism. Numerous large population-based meta-analyses have demonstrated that drinking history – whether light to moderate or heavy – is an AF risk factor [27, 28], whereas abstinence from alcohol significantly decreases the risk of AF recurrence [29]. Alcohol consumption has complex pathological effects on AF and significantly correlates with the duration and amount of alcohol consumed [30], probably because of electrical remodeling, which affects the autonomic nervous response and alters excitation-contraction coupling [31, 32].
Here, we built a strong prognostic nomogram including five significantly distinct variables in the predictive model for postoperative recurrence. However, our nomogram did not include several previously reported predictors, such as BNP, hypertension, diabetes, white blood cells, and red blood cell distribution width. The following factors may account for errors in research: 1) Most previous reports have argued that BNP is associated with recurrence in patients with AF with heart failure complicated by insufficient EF [33, 34]. Although many of our patients had cardiac dysfunction, most showed normal ejection fractions. The variation might have been due to the different patient populations examined. 2) Hypertension, diabetes, and other comorbidities are risk factors for AF. However, because of the current disease detection rate, and because the disease risk factors did not accurately reflect the risk factors for postoperative recurrence of the disease, most of our patients had either both or only one of these comorbidities [35]. Thus, further testing in larger sample-size studies might be warranted. 3) Differences in instruments and techniques might have influenced certain test indicators, such as the neutrophil/lymphocyte ratio and RBC distribution width. Although some studies have linked these factors to relapse [36, 37], most of these variables were not included in the predictive models developed and validated in a large prospective cohort. 4) Patients with PsAF made up most of the study population, and alternative study topics might have produced different conclusions.
Nevertheless, our study has several limitations. First, this was a single-center retrospective study with a small sample size. Second, our analysis did not quantify the indicators, such as alcohol consumption, whose specific dose was not determined; there are individual differences in how much each person drinks and how much each person defines as habitual drinking. Additionally, the study period commenced in 2013, and surgical indications and technology have continually been updated, thus resulting in a non-standardized patient baseline. Finally, our nomogram has not been validated in other hospitals. Therefore, additional study is required to evaluate the reliability of our proposed nomogram model.
Conclusion
In conclusion, we developed a nomogram with five baseline predictors that are easy to use, then performed internal validation. To improve outcome management, this tool would allow physicians to assess and predict postoperative recurrence in patients with PsAF and identify high-risk individuals. According to the findings, our nomogram predictive model has good clinical translational value. We recommend further validation with external cohorts with large sample sizes.