A Simple Score to Predict New-onset Atrial Fibrillation After Ablation of Typical Atrial Flutter

Patient population

This is a multicenter prospective cohort study. From January 2013 to December 2017, participants from the following three Chinese centers were consecutively enrolled: The First Affiliated Hospital of Nanjing Medical University, Affiliated Hospital of Nantong University, and Jiangsu Taizhou People’s Hospital. The inclusion criteria were as follows: (1) aged ≥18 years; (2) receiving CCW-AFL catheter ablation. Exclusion criteria were as follows: (1) any documented AF episodes before ablation; (2) abnormal thyroid function; (3) previous cardiac surgery history; (4) previous ablation of atrial tachycardia or atrial flutter; (5) life expectancy <12 months; (6) refused to participate. Informed consents were obtained and the study protocols were approved by the Ethics Committees of the above three centers.

Clinical data collection and ECG parameters measurement

Before the index procedure, patients’ baseline data were obtained, including gender, age, BMI, comorbidities, echocardiography, and 12-lead ECG both under sinus rhythm (SR) and CCW-AFL (25 mm/s, 10 mm/mV). If ECG under SR or CCW-AFL was unobtainable after admission, the most recent corresponding ECG free from any antiarrhythmic drugs (AADs) influence (administration of AADs within 1 month before ECG acquisition) was collected. The ECG parameters reflecting electrical activity of the left atrium under SR and CCW-AFL, associated with atrial electrical remodeling as proven by our previous study,¹⁵ were measured by two independent cardiologists to provide an average value of three different beats using Adobe Acrobat XI Pro (Adobe, California, USA) (Figure S1 in the Data Supplement). These included: (1) duration of the negative component of flutter wave in lead II (D_FNII); (2) amplitude of the negative component of flutter wave in lead II (A_FNII); (3) amplitude of negative component of the P wave in lead V1 under SR (A_PNV1); (4) duration of negative component of the P wave in lead V1 under SR (D_PNV1); and (5) P wave duration in lead II under SR (D_PII). The values obtained from the two cardiologists were averaged. Then, P wave terminal force in Lead V1 (P_tfV1) under SR was calculated using the formula A_PNV1*D_PNV1 (mm*ms).

Electrophysiological study and catheter ablation

Before the procedure, AADs were discontinued for at least 5 half-lives. Oral anticoagulants were administered for at least three weeks prior to ablation. Transesophageal echocardiography was performed to exclude intra-cardiac thrombosis before ablation. Under routine local anesthesia, a quadripolar and a decapolar catheter were introduced to the right ventricle and coronary sinus via femoral approach, respectively. Activation mapping using the three-dimensional mapping system (Carto, Ensite, or Rhythmia) and entrainment mapping were applied to confirm the diagnosis of CCW-AFL. Subsequently, linear ablation between the tricuspid annulus and the inferior vena cava was performed with irrigated catheters under the power setting of 35-40 W and flow rate of 17-30 ml/min. Bidirectional block across the cavotricuspid isthmus line was defined as procedural endpoint.

Follow-up of patients

AADs were discontinued after a successful ablation. Anticoagulation was continued for at least 3 months postoperatively and discontinued at the physicians’ discretion, depending on the occurrence of NeAF and CHA2DS2-VASc score. Patients were followed up every three months within the first year, and every six months thereafter. ECG and 24-hour Holter monitoring were performed during each follow-up visit. Additional ECGs and 24-hour Holter monitoring were obtained in the event of arrhythmic symptoms. CCW-AFL recurrence was defined as documented CCW-AFL episode lasting >30s after a 3-month blanking period postoperatively. NeAF was defined as any AF episode longer than 30s postoperatively.¹⁶

Statistical analysis

Continuous variables were described as mean ± standard derivation (SD). Unpaired Student’s t-test was used in normally distributed variables. Friedman test was used in non-normally distributed variables. Categorical variables were presented as counts with percentages and compared using Chi-square tests or Fisher’s exact tests. Two-sided P<0.05 was considered statistically significant.

Patients from two centers were assigned into derivation cohort, and patients from the other center were assigned into validation cohort. A score system was derived from the derivation cohort as follows. All patients in the derivation cohort were assigned to new-AF or non-AF groups according to the presence or absence of NeAF during follow-up. Univariate and multivariate logistic regression analyses were conducted to identify the predictors of NeAF. For ease of clinical use, continuous variables that were statistically different were dichotomized with ROC curves to identify optimal cut-points for discrimination, which were rounded to the nearest clinically significant integer when applicable. Variables that were significant in univariate analysis were included into the model, with attention paid to avoid clinically relevant collinearity. A prediction score was created with the variables and strength of association by β coefficients rounded to the nearest integer.¹⁷

Subsequently, its diagnostic performance was evaluated with the area under the receiver operating characteristics curve (AUC) values. To evaluate incremental discrimination beyond existing criteria, we used Delong’s test to compare the AUC values from our derived score system with those from HATCH and CHA2DS2-VASc score systems. The Hosmer-Lemeshow goodness of fit test (P≥0.05) was evaluated to access calibration for model validation. The diagnostic performance of this score system was validated in a separate validation cohort. All statistical analyses were performed using SPSS software version 26.0.

Baseline characteristics and follow-up results

A total of 424 patients were assessed for eligibility, and 145 were excluded. Eight patients (2.9%) were lost to follow-up. Thus, 271 patients (mean age, 59.7±13.6 years; 205 male) were included in the final analysis (Figure 1). Ablation procedural endpoint was reached in all subjects. Major periprocedural complications were observed in 5 patients (1.8%), all of whom had vascular access complications. During a follow-up period of 73.0±6.5 months, 11 patients (4.1%) had CCW-AFL recurrence and underwent a second ablation. 107 cases (39.5%) had NeAF, with 48, 37 and 22 cases in the first year, first to third years, and after 3 years after ablation, respectively. Additionally, due to new-onset symptomatic AF, 57 cases underwent AF ablation and 48 were prescribed with AADs. Patients from the First Affiliated Hospital of Nanjing Medical University and Affiliated Hospital of Nantong University were assigned into derivation cohort (n=190, 70.1%), and patients from Jiangsu Taizhou People’s Hospital were assigned into validation cohort (n=81, 29.9%), which shared similar clinical characteristics (Table 1).

• Download figure
• Open in new tab

Figure 1 The algorithm of patient enrollment.

AF indicates atrial fibrillation; AFL, atrial flutter; AT, atrial tachycardia; CCW-AFL, counterclockwise atrial flutter.

Univariate and multivariate analyses

In the derivation cohort of 190 patients, 76 (40.0%) had NeAF. Statistical significant differences were found between the groups with and without NeAF in terms of age, hypertension, CHA2DS2-VASc score, HATCH score, left atrial diameter (LAD), D_FNII and D_PII, based on univariate analysis (Table 2). For clinical ease of use, we converted the continuous variables, such as age, LAD, D_FNII, and D_PII into categorical variables (Table S1 in the Data Supplement). By conducting a multivariate analysis on these individual parameters with a predictive value, it was found that hypertension, age ≥70 years, LAD ≥42 mm, D_PII ≥120 ms, and D_FNII ≥120 ms were independent predictors of NeAF (Table 3).

Derivation of the score system

The HAD-AF score, ranging from 0 to 9, was developed to predict NeAF based on the β coefficients of the above 5 variables (Table 3), with one point for hypertension and age ≥70 years, two points for LAD ≥42 mm and D_PII ≥120 ms, and three points for D_FNII ≥120 ms (Figure 2). The probability of NeAF increased along with HAD-AF score (Figure 2). Model-based probabilities closely matched the observed prevalence for each given score value, indicating good calibration (Figure 3). The AUC values of HAD-AF, HATCH and CHA2DS2-VASc score systems were 0.938 (95% CI 0.902-0.974), 0.651 (95% CI 0.571-0.730), and 0.679 (95% CI 0.600-0.757) (P<0.001), respectively (Figure 4, panel A). Additionally, a HAD-AF score >4 could predict NeAF with a sensitivity of 0.829 and a specificity of 0.939.

• Download figure
• Open in new tab

Figure 2 Description of the HAD-AF score.

Description of the HAD-AF score and point allocations for each clinical characteristic (top), with associated probability of development of new-onset atrial fibrillation based on the total score as estimated from the model (bottom).

• Download figure
• Open in new tab

Figure 3 Calibration of the HAD-AF score.

The Hosmer-Lemeshow goodness-of-fit test results using deciles of predicted probabilities were P=0.518, 0.199, and 0.640 for the derivation, validation, and overall sample, respectively, indicating support for a properly calibrated model. AF indicates atrial fibrillation.

• Download figure
• Open in new tab

Figure 4 Prediction value of different score systems in the derivation and validation cohorts.

Panel A, in the derivation cohort, AUC= 0.938, 0.651, and 0.679 for HAD-AF, HATCH, and CHA2DS2-VASc score systems, respectively. Panel B, in the validation cohort, AUC= 0.912, 0.610, and 0.635 for HAD-AF, HATCH, and CHA2DS2-VASc score systems, respectively. AUC indicates area under the receiver operating characteristic curve.

Validation of HAD-AF score

Among 81 patients assigned to the validation cohort, 31 (38.3%) experienced NeAF. Results of univariate regression analysis in the validation cohort were similar to those in the derivation cohort (Table S2 in the Data supplement). Model-based probabilities remained closely matched with the observed prevalence for each given score value (Figure 3). The AUC of the HAD-AF score to predict NeAF was 0.912 (95% CI 0.846-0.978), higher than that of the HATCH (0.610, 95% CI 0.480-0.741) and CHA2DS2-VAS_C score systems (0.635, 95% CI 0.506-0.764) (P<0.001), validating a higher predictive value of HAD-AF score (Figure 4, panel B).

Study limitation

Firstly, due to practical constraints²⁶ and early initiation of this study, the absence of long term Holter monitoring and implanted loop recorder might lower the detection rate of preoperative and postoperative AF. However, the preoperative management and postoperative follow-up flow in this study were still considered as a standardized screening protocol in such patients.²⁷ Secondly, atrial fibrosis could also be assessed by cardiac magnetic resonance imaging or possibly biomarkers. However, to simplify the HAD-AF risk score, magnetic resonance imaging parameters or biomarkers were not routinely measured in this study. Thirdly, as AF episode of over 30s was set as the endpoint, a larger multicenter prospective study was needed to test the value of HAD-AF score to improve clinical outcomes.