Abstract
This study aims to establish and validate a nomogram as a predictive model in patients with new-onset atrial fibrillation (AF) after dual-chamber cardiac implantable electronic device (pacemaker) implantation.
A total of 1120 Chinese patients with new-onset AF after pacemaker implantation were included in this retrospective study. Patients had AF of at least 180/minute lasting 5 minutes or longer, detected by atrial lead and recorded at least 3 months after implantation. Patients with previous atrial tachyarrhythmias before device implantation were excluded. A total of 276 patients were ultimately enrolled, with 51 patients in the AF group and 225 patients in the non-AF group. Least absolute shrinkage and selection operator (LASSO) method was used to determine the best predictors. Through multivariate logistic regression analysis, a nomogram was drawn as a predictive model. Concordance index, calibration plot, and decision curve analyses were applied to evaluate model discrimination, calibration, and clinical applicability. Internal verification was performed using a bootstrap method.
The LASSO method regression analysis found that variables including peripheral arterial disease, atrial pacing-ventricular pacing of at least 50%, atrial sense–ventricular sense of at least 50%, increased left atrium diameter, and age were important predictors of developing AF. In multivariate logistic regression, peripheral arterial disease, atrial pacing-ventricular pacing of at least 50%, and age were found to be independent predictors of new-onset AF.
This nomogram may help physicians identify patients at high risk of new-onset AF after pacemaker implantation at an early stage in a Chinese population.
Abbreviations and Acronyms
- ABI
ankle-brachial index
- AF
atrial fibrillation
- AP
atrial pacing
- CIED
cardiac implantable electronic device
- CIED-AF
cardiac implantable electronic device–detected atrial fibrillation
- C-index
concordance index
- HF
heart failure
- LA
left atrial
- LASSO
least absolute shrinkage and selection operator
- OR
odds ratio
- PAD
peripheral arterial disease
- SCAF
subclinical atrial fibrillation
- VP
ventricular pacing
- VS
ventricular sense
Introduction
Atrial fibrillation (AF) is the most common arrythmia in adults older than 40 years.1 Stroke is the most serious complication of AF and can triple the mortality rate in patients with AF. Each year, 16.9 million people worldwide experience stroke. The causes remain unexplained for approximately 20% to 40% of strokes.2 Between 10% and 30% of these unexplained stroke cases may be the result of a missed AF diagnosis.3–5 Cardiac implantable electronic device–detected AF (CIED-AF), which is a type of subclinical atrial fibrillation (SCAF), has attracted much attention in recent years, as it can increase the detection rate of AF episodes. Although different from clinical AF, SCAF, which often manifests as asymptomatic AF, has a strong predictive effect on clinical AF.6 Subclinical AF occurs frequently in older people. Cardiac implantable electronic devices help detect CIED-AF.7 The detected increase in AF load may lead to increased ischemic stroke, heart failure (HF), and mortality.8–10 Identifying patients with CIEDs at high risk of new-onset AF is important to prevent adverse clinical outcomes. The current diagnosis of AF relies mainly on clinical symptoms, electrocardiography, and other detection methods. However, AF is often paroxysmal and asymptomatic, which makes early effective and accurate detection of AF difficult. Therefore, establishment of an SCAF prediction model may help clinicians make effective interventions and have good therapeutic effect.
Age is a frequently reported independent predictor of CIED-AF.7,11 Some studies have also found that conditions such as hypertension, HF, and cerebral microvascular disease; CHADS2 (congestive HF, hypertension, age ≥75 years, and diabetes mellitus, stroke) and CHA2DS2-VASc (congestive HF, hypertension, age ≥75 years, diabetes mellitus, stroke or transient ischemic attack, vascular disease, age 65–74 years, sex category) scores; echocardiographic indicators including left atrial (LA) diameter and LA volume index; and pacemaker parameters such as ventricular pacing ratio may be predictors of CIED-AF.10–14 However, their effectiveness in predicting the risk of CIED-AF is still controversial. Currently, there are no predictive tools to integrate and analyze the risk factors affecting CIED-AF.
Nomogram is a prognostic model that describes statistics in a graphical manner through biological clinical variables. It is a tool for prediction based on statistical data obtained from patients with diseases with identical characteristics. Each variable in the nomogram is given a value that represents its prognostic significance. Individualized risk can be estimated according to the patients' different clinical characteristics. The cumulative scores of all variables are matched with the result scale, and a value is obtained to estimate the specific results.15–17 Compared with other predictive tools, nomogram is most accurate and has superior discriminative capability in predicting results.18 As an important tool for guiding modern medical decisions, nomogram has been widely used in the fields such as surgery, cancer, and myocardial infarction.19–21 No previous research to date has reported use of nomogram as a predictive model for prediction of CIED-AF occurrence.
The purpose of this study was to identify independent predictors and establish a nomogram based on parameter integration to predict the probability of new-onset AF in patients after dual-chamber pacemaker implantation. The goal of this model is to help clinicians identify high-risk patients, control risk factors, and select reasonable treatment options for these patients.
Patients and Methods
Patient Selection
To evaluate the predictive ability and discrimination performance of nomograms, bootstrap verification (1000 bootstrap samples) was performed to calculate the concordance index (C-index). Patients with new-onset AF after CIED implantation7 at the cardiovascular department of Sir Run Run Shaw Hospital in Hangzhou, China, from January 1, 2017, to December 31, 2018, were included in the study. Patients were excluded if they had a history of AF or atrial flutter or other persistent supraventricular tachycardia lasting more than 5 minutes as confirmed by previous examination or if they had a history of cardiac surgery for congenital heart disease, valvular heart disease, or myocardial infarction. Patients were also excluded if they had an implanted single-chamber pacemaker, implantable cardioverter-defibrillator, or cardiac resynchronization therapy-defibrillator/pacemaker or if they had undergone replacement of a CIED. Patients were divided into 2 groups based on whether new-onset AF was detected after pacemaker implantation (new-onset AF and non-AF groups). The inclusion and exclusion criteria are shown in Table I.22
Data Collection
The following medical data were obtained from electronic medical records based on relevant demographic, clinical, and laboratory features. These features consist of general information, history, auxiliary examination, echocardiographic parameters, medication administration record, and pacemaker parameters. General information included age, sex, body mass index, and smoking and drinking status. Medical history included congestive HF, hypertension, coronary heart disease, peripheral arterial disease (PAD), diabetes, hyperlipidemia, cerebrovascular disease, chronic kidney disease, and chronic obstructive pulmonary disease. Peripheral arterial disease was previously diagnosis according to ankle-brachial index (ABI) of 0.9 or less, ABI of 1.4 or greater, or imaging evidence combined with clinical manifestations. Congestive HF was defined by moderate to severe systolic left ventricular dysfunction, left ventricular ejection fraction of 40% or less, or receiving medication to relieve symptoms of HF. Hypertension was defined by systolic blood pressure of 140 mm Hg or greater and/or diastolic blood pressure of 90 mm Hg or greater as measured 3 times in the same patient on different days or if the patient had a history of using antihypertensive drugs to control blood pressure. Diabetes was diagnosed according to fasting venous plasma glucose concentration of at least 7.0 mmol/L (126 mg/dL), random vein plasma glucose concentration of at least 11.1 mmol/L (200 mg/dL), or positive glucose tolerance test vein plasma glucose concentration of at least 11.1 mmol/L 2 hours after oral administration of 75 g of glucose and/or the use of insulin or oral hypoglycemic drugs. If a patient met any 1 of the 3 criteria, all 3 measurements were duplicated the following day.
Coronary heart disease included patients with known but untreated disease and those being treated with anti-platelet aggregation drugs and lipid-lowering drugs. Diagnosis of cerebrovascular disease was made according to the corresponding neurological symptoms and signs and from imaging evidence. The attack lasted more than 24 hours. Chronic kidney disease was diagnosed as basic kidney disease lasting more than 3 months with an estimated glomerular filtration rate less than or equal to 60 mL/min/1.73 m2. Chronic obstructive pulmonary disease was diagnosed according to a history of more than 2 years of cough, expectoration, asthma, and other respiratory symptoms, as well as lung function test airflow limitation (forced expiratory volume in 1 second/forced vital capacity <70%). Auxiliary examination included N-terminal prohormone of brain natriuretic peptide, estimated glomerular filtration rate, blood urea nitrogen, and high-sensitivity C-reactive protein. Echocardiographic parameters included left ventricular ejection fraction and LA diameter. Increased LA diameter was defined as 40 mm or greater for men and 38 mm or greater for women by echocardiography.23 Prior medication administration records included angiotensin-converting enzyme inhibitors/angiotensin II receptor antagonists, b-blockers, calcium channel blockers, lipid-lowering drugs, diuretics, and antiplatelet drugs. Pacemaker parameters included pacemaker manufacturer, type of electrode, and mode setting. CHADS2 and CHA2DS2-VASc scores were also included. The nominal pacemaker setting for paced atrioventricular intervals/sensed atrioventricular intervals delay is 180/150 ms with the atrioventricular interval optimization function on.
The primary outcome was new-onset AF (atrial rate ≥180 bpm detected by atrial lead) with duration of at least 5 minutes occurring at least 3 months after CIED implantation. Confirmation of AF diagnosis must have been made by at least 1 clinician, and CIED-AF was diagnosed according to guidelines from the European Society of Cardiology.24 Patients were followed up regularly at 1, 3, 6, 9, and 12 months after implantation. During follow-up, pacemaker parameters were recorded.
All participants provided written consent before entering the study. The study was approved by the medical ethics committees of Sir Run Run Shaw Hospital, affiliated with the Medical School of Zhejiang University, with the ethics approval No. keyan20180106-10.
Statistical Analysis
Data were analyzed using SPSS 25.0 software and R software version 3.6.0 (http://www.r-project.org). Categorical variables were compared by χ2 test and expressed as a number and percentage. Continuous variables with a normal distribution were compared by t test and expressed as mean (SD). Continuous variables with non-normal distribution were compared by Mann-Whitney U test and expressed as median (IQR).
The least absolute shrinkage and selection operator (LASSO) regression method for reducing high-dimensional data was applied to select the best predictor for new-onset AF in patients after dual-chamber pacemaker implantation. The LASSO regression was performed using the “glmnet” package of R software. Cross-validation was used to determine the adjustment parameter (λ) suitable for LASSO regression. The principle is that the sum of the absolute values of the negative log likelihood ratios must be smaller than the adjustment parameter λ. If λ is large, it has no effect on the estimated regression parameters, but as λ decreases, some coefficients may decrease to 0. Therefore, λ was chosen with the smallest cross-validation error. Finally, the model was refitted by using all the observed values and the selected λ. Most coefficients were reduced to 0, and the remaining nonzero coefficient variables were risk factors selected by LASSO. Subsequently, after combined consideration with the clinical significance of the variables, LASSO regression results, and sample size, final variables included in the model were determined.
The binary classification logistic regression was used to construct a prediction model. In logistic regression, the association between independent variables and clinical events was determined by 95% CI, odds ratio (OR), and P value. P < .05 was considered statistically significant.
Based on binary classification logistic regression, a nomogram was established to predict the probability of new-onset AF in patients with dual-chamber pacemakers. To evaluate the accuracy and discriminative ability of the nomogram, the C-index was calculated and a calibration curve constructed. The “rmda” package decision curve analysis chart in R software was used to quantify the net benefit under different threshold probabilities in patients with new-onset AF after dual-chamber pacemaker implantation to determine the clinical utility of the nomogram. Lift verification (1000 bootstrap samples) was applied for internal verification of the model.
Results
Clinical Characteristics of the Study Cohort
The primary cohort included 1120 patients who underwent surgery for electronic device implantation in the cardiology department of Sir Run Run Shaw Hospital between January 2017 and December 2018. After screening according to the inclusion and exclusion criteria, 276 patients were ultimately enrolled, including 51 patients with new-onset AF and 225 patients without AF. Nineteen patients had incomplete medical records, 60 patients had irregular outpatient visits and lack of follow-up data, and 3 patients died within 1 year of follow-up, for a total of 82 patients who were excluded because of incomplete data or loss to follow-up (Fig. 1).
Patients were followed up for a mean period of approximately 23.44 months. New-onset AF was detected in 51 patients (18.7%) after double-chamber pacemaker implantation, and 225 patients did not develop AF. The follow-up time for patients with new-onset AF was longer (23.5 months vs 21.0 months). The clinical characteristics between the 2 groups were compared, and the results showed a significant difference in age (P = .012) and history of PAD (P < .001). The average age of patients with new-onset AF was 72.9 years, which was significantly older than the non-AF group (68.8 years old). The incidence of PAD was significantly higher in the new-onset AF group than in the non-AF group (17.6% vs 4%). There were also significant differences in the pacing pattern between the 2 groups. In the group of patients with new-onset AF, atrial pacing (AP)–ventricular pacing (VP) ≥50% accounted for 19.6% of cases, which was higher than 8.0% in the group without AF (P = .014). Atrial pacing ≥50% and atrial sense (AS)–ventricular sense (VS) ≥50% were also statistically different between the 2 groups, with P values of .048 and .037, respectively. The percentage of AS-VS ≥50% in the new-onset AF group was 17.6%, and in the group without new-onset AF, the percentage was 32.4%. Thus, it was found that the population with less AS-VS ≥50% (which indicated a greater percentage of AP and/or VP) was more prone to new-onset AF. There was a negative correlation between AS-VS ≥50% and risk of new-onset AF. There were no significant differences between the 2 groups in sex, body mass index, or history of hypertension, diabetes, coronary heart disease, HF, and drug use. Descriptive statistics of the patients in this study are shown in Table II.
Feature Selection
The 28 independent variables assessed in this study were selected by the LASSO regression algorithm. When the most suitable tuning parameter λ was 0.036, the partial likelihood deviation reached the minimum value (Fig. 2A). At that time, 5 nonzero coefficient variables were retained in the LASSO analysis (Fig. 2B). The variable with the most predictive accuracy was PAD, followed by AP-VP ≥50%, AS-VS ≥50%, LA diameter, and age.
Nomogram Construction and Performance
After combined analysis of the variables' clinical significance and LASSO regression results and after further assessment of whether the variables were colinear and crossover, 4 variables were used for logistic regression as initial independent variables: PAD, AP-VP ≥50%, increased LA diameter, and age. Logistic regression results showed that PAD (OR, 1.812 [95% CI, 0.917–3.585]; P < .001), AP-VP ≥50% (OR, 3.006 [95% CI, 1.242–7.279]; P = .015), and age (OR, 1.037 [95% CI, 1.004–1.072]; P = .029) were independent predictors (Table III). The predictive factors obtained above were integrated into a model through R language, and the results were visualized with a nomogram (Fig. 3).
Nomogram Example
Taking a 65-year-old male patient with a dual-chamber pacemaker and PAD as an example (Fig. 3), 65 years in the age line corresponds to a score of 50 in the points line, and the presence of PAD corresponds to a score of 70. If the patient is admitted to the hospital with an echocardiogram indicating an increased LA diameter of 42 mm, the corresponding score is 20. If AP-VP <50% is detected during pacemaker program-control follow-up, the corresponding score is 0. The above-mentioned scores are cumulative, so the total score for this patient is 140 points (50 + 70 + 20 + 0). Therefore, when a vertical line is drawn from the total score axis to the risk of new-onset AF axis, the result shows that the risk of new-onset AF is 50% (Fig. 3).
Verification of the Nomogram
The C-index for nomogram prediction was 0.726 (95% CI, 0.717–0.735) for the cohort (Fig. 4A), or 0.727 (95% CI, 0.651–0.800) through bootstrapping validation. The calibration curve of the nomogram for the risk of new-onset AF in patients with a dual-chamber pacemaker indicated that the predictions were highly consistent with the observed results. The above-described data showed that the model had good predictive ability (Fig. 4B).
Clinical Value of the Nomogram
The clinical decision curve is a new method to evaluate the clinical application value of the predictive model. The decision curve analysis of the nomogram is presented in Figure 5. The nomogram predictive model established in this study had a higher clinical net benefit than the single-factor model of AP-VP ≥50% or increase in LA diameter. Therefore, the decision curve analysis confirmed the net benefits of the predictive nomogram.
Discussion
Cardiac implantable electronic device–detected AF is reported to be closely associated with stroke and HF.8,25 Whether the risk of CIED-AF is the same as for clinically observed AF is still controversial. Currently, there are no published guidelines or clinical trials to help clinicians make an evidence-based clinical decision for the prevention of CIED-AF.26 Also, there is no generally accepted predictive model of new-onset AF for patients with an implanted pacemaker. In this study, the authors aimed to establish and validate a predictive nomogram for predicting the risk of developing new-onset AF in patients with CIEDs. Considering that pacemaker implantation may induce surgery-related AF, combined with previous literature, AF recorded by pacemaker at least 3 months after implantation was chosen as the outcome event of this study.
It was reported that approximately 13% to 35% of patients with CIEDs may experience new-onset AF.13,25,27 In this study, the incidence of new-onset AF after dual-chamber pacemaker implantation was 18.7%, which was consistent with previous studies. The differences in the incidence of AF from various studies may be the result of differences in research backgrounds and methods of diagnosing AF. In addition, some studies did not exclude patients with a history of AF.
In this study, multivariate logistic regression analysis results revealed that PAD, AP-VP ≥50%, and age were independent predictors of new-onset AF after dual-chamber pacemaker implantation. Although LA diameter was not an independent predictor, LA diameter did play an important role in the progress of new-onset AF. This study used the LASSO method, a data dimensionality reduction method, which is suitable for the regression of large groups of clinical factors, selects variables for the sample data, and extracts the most important predictors by compressing the original coefficients to avoid overfitting.
Atrial fibrillation is more common in patients with PAD than in the general population.28,29 Data from the Reduction of Atherothrombosis for Continued Health (REACH) Registry indicated a high prevalence of PAD coexisting with AF.30 The 2 diseases have common risk factors and epidemiological characteristics, and the coexistence of these conditions significantly increases the risk of major adverse events (including stroke, myocardial infarction, cardiovascular death, and all-cause death).22,31 In an observational study based on global risk assessment of AF in older women, Perez et al32 found that PAD was independently associated with a high incidence of AF. The Multi-Ethnic Study of Atherosclerosis33 and Cardiovascular Health Study34 also confirmed this relationship. The authors found that PAD remained an independent risk factor for AF after adjusting for factors such as age, sex, ethnicity, and cardiovascular disease. In addition, studies have shown that the risk of AF increases as the value of the ABI decreased.33,35 Bekwelem et al36 found that the relationship between PAD and AF may be mainly related to traditional atherosclerotic risk factors. Atherosclerosis and AF have several similar risk factors and have both been associated with elevated levels of proinflammatory cytokines, endothelial dysfunction, and platelet-mediated thrombosis.37–39 Theses common risk factors may explain why PAD was observed as a risk factor of CIED-AF in this study.
Previous studies have shown a linear increase in the cumulative percentage of VP and the risk of AF.40,41 Cheung et al13 studied patients with different indications for dual-chamber pacemaker implantation and found that in patients with sinus node dysfunction and a dual-chamber pacemaker, cumulative VP ≥50% is related to the elevated occurrence risk of AF by 2-fold. In combination with previous studies, it was speculated that AP-VP pacing may enhance myocardial cell extension, atrial fibrosis, and destruction of intercellular coupling, thereby promoting the remodeling of the atrium. In addition, right ventricular pacing may lead to mechanical desynchronization of the heart. This desynchronization will impair left ventricular function, change hemodynamics, gradually increase LA pressure, and further promote atrial myocyte electrical remodeling, which may cause CIED-AF.42–45
Consistent with previous studies, the findings of this study revealed that advanced age could contribute to the development of new-onset AF.7 Several possible mechanisms may explain the relationship between AF and age. Atrial fibrosis, which is an important structural basis for AF, increases with ages.46,47 This may be one of the causes of AF in older adult patients with dual-chamber pacemakers. In addition, 1 study found that aging may be the most important risk factor affecting the stability of the cardiovascular internal environment.48
This study also indicated that increased LA diameter could contribute to the development of new-onset AF.49 The following mechanisms may explain the relationship between increased LA diameter and new-onset AF. On one hand, CIED implantation may directly induce cardiac fibrosis and atrial expansion, which lead to cardiac remodeling and reentry loop formation and, eventually, to AF. On the other hand, CIED implantation may trigger tachycardia, which may promote atrial contraction dysfunction and induce atrial dilation.50 Atrial dilation promotes atrial remodeling and atrial fibrosis by increasing the diastolic length of atrial muscle cells, which triggers AF.51
To the authors' knowledge, this is the first study to establish a nomogram-illustrated model to evaluate the risk of new-onset AF after CIED implantation in a Chinese population. This nomogram may be a convenient tool for clinicians to identify patients at high risk for new-onset AF early after dual-chamber pacemaker implantation. Early recognition of patients at high risk for new-onset AF may be helpful for AF prevention management and prognosis improvement. By applying the nomogram, the risk factors can be controlled, thus reducing the incidence of new-onset AF. The probabilities derived from the nomogram can also be used prospectively to evaluate the effects and outcomes of long-term primary prevention for patients at high risk for developing AF after dual-chamber pacemaker implantation. In addition, some scholars believe that SCAF may be the direct cause of stroke, whereas others consider SCAF to be only a risk factor for stroke. Being able to assess the risk of AF through the nomogram may allow clinicians to enhance the frequency of programmed follow-up for high-risk patients to reduce the rate of missed diagnosis. More timely information on the occurrence of SCAF and its complications (eg, stroke and HF) can be provided, and potential mechanisms can be further revealed.
This study had several limitations. First, the data were collected retrospectively in a single center, and the sample size was small. The included independent variables cannot completely cover all AF risk factors. As a result, the predictive power of the model was limited. Second, categorical variables were used to build a line chart to ensure that the nomogram is an easy-to-use clinical tool. Compared with the use of continuous variables, this simplicity may partially sacrifice the effectiveness of prognostic risk prediction.52 Third, this study used self-service internal verification. Although this is an acceptable method of negotiating chart development and verification, external validation is still needed to test its versatility. Fourth, there may be inadequate detection conditions for the variables involved in this study. For example, not every patient was tested for ABI after admission, which may lead to a missed PAD diagnosis in some patients. In addition, LA examination only involved LA diameter. Including LA volume and LA function measurements may make the predictive model more accurate. Therefore, further research is needed to supplement the follow-up information, increase the sample size, and select more independent variables to reflect the risk of new-onset AF after pacemaker implantation. This will help to improve the predictive ability of the model.
Conclusion
In conclusion, the authors established and validated an original nomogram that can be conveniently used to identify the risk of new-onset AF in patients after dual-chamber pacemaker implantation in a Chinese population. The proposed nomogram incorporating PAD, AP-VP ≥50%, age, and increased LA diameter may be a valuable tool for clinicians in making decisions and implementing interventions with clinical practical utility.
Author Contributions: Y.Y. and G.F. drafted the study protocol and design. J.L., Z.S., and C.S. performed the data sorting and input and data collection. F.X., J.T., and J.J. contributed to refinement of the data collection tools and design of follow-up procedures. J.L., C.X., and Z.S. performed statistical analysis and were responsible for daily research management. Y.Y. and J.L. prepared and edited the manuscript. Y.Y. is the principal investigator. All authors read and approved the final manuscript.
Conflict of Interest Disclosures: The authors declare that they have no competing interests.
Funding/Support: This work was supported by the Medical Health Science and Technology program of Zhejiang Province (grants 2018KY492 and 2020KY165), the project supported by Zhejiang Provincial Natural Science Foundation of China (grant No. LY21H020006), and the Clinical Research Fund Project of Zhejiang Medical Association (funding for cardiovascular and cerebrovascular diseases, grant No. 2018ZYC-A13).
Availability of Data and Materials: The data sets used during the current study are available from the corresponding author upon reasonable request.