An Obesity Paradox in Atrial Fibrillation Recurrence after Cardioversion: Evidence From a Vietnamese Cohort

2.1. Study Design and Population

This single-center observational cohort study included consecutive patients with atrial fibrillation (AF) who underwent successful Direct-Current Cardioversion (DCCV) at Tam Duc Heart Hospital, Ho Chi Minh City, Vietnam, between December 2022 and April 2025. All participants were of Kinh ethnicity, with no participants from other ethnic groups. This period represents the eligible DCCV treatment period rather than the period of research-specific data collection. Clinical, echocardiographic, laboratory, treatment, and follow-up data were obtained from medical records generated as part of routine clinical care. Patients were assessed for AF recurrence during the 3 months following successful DCCV; consequently, follow-up of the last eligible patient extended through July 2025. No study-specific intervention was performed, and decisions regarding DCCV and subsequent clinical management were made independently by the treating physicians in accordance with the hospital's clinical guidelines.

No formal a priori sample-size calculation was performed because all consecutive eligible patients during the predefined study period were included. The resulting cohort of 96 patients is comparable in size to previous observational studies of AF recurrence after successful cardioversion that included approximately 76–99 patients [6, 8]. Eligible patients were aged ≥18 years and had documented AF requiring DCCV. Successful cardioversion was defined as restoration of sinus rhythm following the final shock with maintenance of sinus rhythm until hospital discharge. Patients were excluded if they had significant valvular heart disease, prior catheter ablation for AF, permanent AF, incomplete echocardiographic data, or insufficient follow-up information. DCCV was delivered using synchronized biphasic shocks under intravenous sedation, starting at 50 – 100 J and escalating to 150 J as needed. Anterolateral electrode placement was used, with one to two shocks delivered per procedure.

2.2. Data Collection and Baseline Assessment

Baseline demographic, clinical, and echocardiographic data were obtained from electronic medical records at the time of cardioversion. Collected variables included age, sex, Body Mass Index (BMI), AF type (paroxysmal, persistent, or long-standing persistent), CHA₂DS₂-VASc score, heart failure status, Left Ventricular Ejection Fraction (LVEF), and use of Antiarrhythmic Drugs (AADs) following cardioversion. Intravenous sedation was achieved with Propofol Lipuro 1% (titrated to 1–2 mg/kg) or thiopental 2.5% (initial dose, 2–3 mL), with adequate sedation confirmed by loss of verbal and pain responsiveness. Amiodarone was administered orally at 400 mg/day for 7 days, followed by 200 mg/day during the 3-month follow-up.

Blood samples for measurement of N-terminal pro–B-type natriuretic peptide (NT-proBNP) were obtained at baseline and at 3 months after cardioversion as part of routine clinical care using either the Elecsys proBNP II assay on Cobas e analyzers (Roche Diagnostics, F. Hoffmann-La Roche Ltd) or the Architect NT-proBNP assay on Architect/Alinity i platforms (Abbott Laboratories), according to routine laboratory procedures. Laboratory measurements were performed according to the hospital's routine standardized laboratory procedures.

2.3. Follow-up and Outcome Definition

Patients were followed for 3 months after confirmed successful cardioversion. The primary outcome was AF recurrence within 3 months after successful DCCV. We defined recurrence as any documented episode of atrial fibrillation lasting at least 30 seconds. These episodes could be captured on a 12-lead Electrocardiogram (ECG) or a 24-hour Holter ECG during the 3-month mark. We also counted recurrence at any point during follow-up if a patient felt symptoms like palpitations or an irregular heartbeat that suggested the arrhythmia had returned, confirmed by an ECG.

2.4. Statistical Analysis

Continuous variables are presented as mean (SD) for approximately normally distributed data and as median (IQR) for skewed distributions. Categorical variables are presented as counts and percentages. We used logistic regression to model the binary outcome of 3-month AF recurrence, as all patients successfully completed a 3-month follow-up without censoring and the time-to-event variation was negligible. Given 39 recurrence events, the multivariable model was restricted to four clinically relevant covariates (left atrial diameter, body mass index, age, and heart failure status) to reduce the risk of overfitting. For discrimination analysis, a parsimonious model comprising LAD, BMI, and age was retained because adding heart failure did not improve model fit or discrimination (likelihood-ratio p=0.239; DeLong p=0.954).

Model performance should therefore be considered exploratory. Results are reported as adjusted odds ratios (aORs) with 95% Confidence Intervals (CIs).

Post-DCCV Antiarrhythmic Drug (AAD) therapy was excluded from multivariable analysis because most patients were maintained on a homogeneous regimen of amiodarone (91.2% in the no-recurrence group vs. 87.2% in the recurrence group). Due to this lack of significant pharmacological variance between cohorts, AAD use was not considered a viable discrete predictor for the model.

Model discrimination was assessed using Receiver Operating Characteristic (ROC) curve analysis, and the Area Under the Curve (AUC) was calculated. Differences in AUC between nested models were compared using the DeLong test. The optimal LAD cutoff for identifying 3-month recurrence was determined by maximizing the Youden index; sensitivity and specificity at this threshold were reported. Calibration was assessed by comparing predicted and observed risks (including the Hosmer–Lemeshow goodness-of-fit test as a descriptive measure). Subgroup analyses were exploratory, and no formal adjustment for multiple comparisons was applied. Accordingly, subgroup findings were interpreted based primarily on interaction tests and should be considered hypothesis-generating. Effect modification was evaluated by including interaction terms in regression models, and p-values for interaction were reported. Change in NT-proBNP levels was analyzed using the Mann–Whitney U test and was considered exploratory.

All analyses were performed using R software (version 4.5.2). A two-sided p-value < 0.05 was considered statistically significant.

2.5. Ethics Statement

This cohort study was conducted at Tam Duc Heart Hospital. The study protocol was approved by the Institutional Review Board (IRB) of Tam Duc Heart Hospital (approval No. 30.24/GCN-BVTĐ; dated November 14, 2024). The study included routinely collected clinical records of patients who underwent DCCV between December 2022 and April 2025. December 31, 2022 represents the beginning of the eligible clinical-record period rather than the initiation of research-specific data collection. All clinical and follow-up information analyzed in this study was derived from medical records generated during routine clinical care. No study-specific intervention or modification of patient management was performed, and decisions regarding DCCV were made independently by the treating physicians in accordance with the hospital's clinical guidelines. This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline.

3.1. Baseline Characteristics

Of 124 patients assessed for eligibility, 16 were excluded (11 did not meet inclusion criteria; 5 declined to participate), and 12 experienced failed DCCV. The final study cohort comprised 96 patients who achieved successful restoration of sinus rhythm following DCCV. There was no missing data for the primary variables of interest. At the 3-month follow-up, 57 patients (59.4%) remained in the non-recurrence group, while 39 (40.6%) experienced atrial fibrillation recurrence (Fig. 1).

Baseline characteristics are summarized in Table 1. Patients who experienced AF recurrence were significantly younger than those without recurrence (59.2 {10.0} vs. 64.3 {12.6} years, p = 0.029). In addition, Left Atrial Diameter (LAD) was significantly greater in the recurrence group (42.6 {6.0} vs. 39.9 {4.9} mm, p = 0.021). No significant differences were observed between the two groups regarding sex distribution (female, 25.6 vs. 33.3%, p = 0.562), left ventricular ejection fraction (56.9 {11.5} vs. 59.5 {9.5} %, p = 0.251), body mass index (24.3 {3.5} vs. 25.3 {3.2} kg/m², p = 0.140), or NT-proBNP levels (475 {296–854} vs. 400 {189–805} pg/mL; p = 0.160). Persistent AF was the most prevalent type. Antiarrhythmic Drug (AAD) utilization, primarily amiodarone, was high across the cohort (89.6%) and showed no statistically significant difference between those who maintained sinus rhythm and those who experienced recurrence (p = 0.638).

3.2. Multivariable Predictors of AF Recurrence

In multivariable logistic regression adjusted for age, BMI, and heart failure status, LAD remained independently associated with AF recurrence. Each 1-mm increase in LAD was associated with a 13% increase in the odds of recurrence (aOR 1.13; 95% CI, 1.03–1.25; p = 0.011). Each 1 kg/m² increase in BMI was associated with a 15% reduction in the odds of recurrence (aOR per 1 kg/m² increase 0.85; 95% CI, 0.74–0.99; p = 0.038). Age demonstrated a borderline inverse association with recurrence (aOR 0.96; 95% CI, 0.93–1.00; p = 0.051). Heart failure status was not independently associated with recurrence (p = 0.239) (Table 2).

3.3. Discriminative Performance of the Reduced Model

The AUC for LAD alone was 0.62 (95% CI, 0.50–0.74). In multivariable analysis, heart failure was not independently associated with 3-month AF recurrence (aOR 1.86, 95% CI 0.66–5.21; p = 0.239). A reduced model comprising LAD, BMI, and age was therefore evaluated for discrimination. This three-variable model yielded an AUC of 0.73 (95% CI, 0.62–0.83), reflecting modestly improved discrimination compared with LAD alone (p = 0.03). Calibration assessment using the Hosmer–Lemeshow test did not indicate significant lack of fit. The LAD cutoff derived from Youden index optimization (41 mm) showed moderate sensitivity and specificity and should be considered exploratory (Fig. 2).

3.4. NT-proBNP changes after Cardioversion

Exploratory analysis demonstrated a greater reduction in NT-proBNP levels over 3 months among patients maintaining sinus rhythm compared with those with recurrence (p = 0.010) (Fig. 3). This analysis was exploratory, not adjusted for potential confounders, and was therefore not included in the multivariable model.

3.5. Subgroup Analysis

The association between LAD and recurrence was directionally consistent across subgroups including sex, age (<65 vs. ≥ 65 years), AF type (paroxysmal vs. persistent/long-standing persistent), and heart failure status. Formal interaction testing did not demonstrate statistically significant effect modification by sex (Pinteraction = 0.85), age group (Pinteraction = 0.91), AF type (Pinteraction = 0.52), or heart failure status (Pinteraction = 0.22) (Fig. 4). Given the modest sample size and limited number of recurrence events, these subgroup analyses should be considered exploratory and interpreted with caution.

4.1. Principal Findings

This study cohort reflects a middle-aged to older population with a high burden of cardiovascular comorbidities, consistent with the reported rise in AF-related metabolic risk factors in Vietnam. The recurrence rate of 40.6% observed in our study underscores a significant clinical challenge. Larger LAD and lower BMI were independently associated with increased odds of recurrence. An integrative model incorporating LAD, BMI, and age demonstrated modest improvement in discrimination compared with LAD alone.

These findings reinforce the relevance of structural atrial remodeling in early rhythm instability after cardioversion, consistent with prior reports showing substantial recurrence rates even after initially successful cardioversion [3, 8, 10].

4.2. Left Atrial Diameter and Structural Remodeling

Patients with AF recurrence were younger and had a significantly larger left atrial diameter than those without recurrence. While the younger age observed in this cohort warrants further investigation, the association between left atrial enlargement and AF recurrence is consistent with atrial structural remodeling. Left atrial enlargement reflects chronic pressure and volume overload, atrial fibrosis, and electrical remodeling [9]. Meta-analytic evidence further supports the role of left atrial volume index as a strong predictor of recurrence [4]. Advanced echocardiographic parameters, including left atrial strain and mechanical dysfunction, provide additional prognostic value [5, 9]. Importantly, LA diameter remains a practical surrogate, with validation studies demonstrating its correlation with left atrial volume [11].

Structural remodeling alters atrial conduction properties and facilitates re-entry mechanisms that promote AF recurrence even after successful restoration of sinus rhythm. AF recurrence is influenced by complex clinical phenotypes rather than single parameters, as demonstrated by cluster-based analyses [12]. Early failure of rhythm control strategies may indicate a more advanced arrhythmogenic substrate and is associated with a higher likelihood of subsequent recurrence [13]. In the present study, LA diameter maintained an independent association with recurrence after adjustment for relevant clinical covariates, suggesting its continued utility for risk stratification. These findings indicate a graded increase in recurrence risk associated with increasing LA diameter. Although our analysis suggests a threshold of 41.0 mm, this value must be viewed as exploratory; this anatomical measurement likely represents a transition toward an arrhythmogenic substrate, requiring validation in larger, prospective, multicenter cohorts before guiding definitive clinical practice.

4.3. Body Mass Index and Recurrence Risk

The inverse association between BMI and AF recurrence in our Vietnamese cohort warrants cautious interpretation. While obesity is a well-established risk factor for incident AF, its relationship with post-cardioversion outcomes appears heterogeneous [1]. Our findings, which contrast with Western reports linking higher BMI to increased recurrence [10], align with prior observations in Asian populations, where a U-shaped association between BMI and AF outcomes has been proposed [14]. These discrepancies likely stem from limitations inherent in BMI as a crude surrogate for cardiometabolic risk. BMI fails to account for cardiometabolic risk, which is profoundly influenced by central adiposity, dietary habits, and physical activity [15]. Moreover, underlying metabolic dysfunction is driven by insulin resistance and systemic inflammation, mediated by markers such as WISP-1 and Fetuin-A [16, 17]. Lifestyle interventions, including aerobic exercise combined with fasting, have been shown to significantly mitigate these pro-inflammatory pathways and optimize body composition [16]. Therefore, the inverse relationship between BMI and AF recurrence observed in our study is likely confounded by unmeasured variables, such as visceral fat distribution, metabolic reserve, and inflammatory status, rather than reflecting a genuine protective effect of excess body weight. These findings should be considered hypothesis-generating and indicate that conventional BMI thresholds derived from Western populations may not accurately represent the arrhythmogenic substrate in Vietnamese patients.

4.4. Model Performance and Clinical Implications

The addition of BMI and age to LAD modestly improved discrimination (AUC 0.73). Although this performance does not support definitive risk stratification for clinical decision-making, it suggests that combining structural and clinical variables may enhance recurrence risk estimation compared with structural assessment alone. AF recurrence is increasingly recognized as a multifactorial phenomenon driven by complex patient phenotypes [12]. Given the modest discrimination and limited sample size, these findings should not be used to guide therapeutic escalation but may inform future model development and validation efforts.

4.5. NT-proBNP Dynamics and Subgroup Consistency

Patients maintaining sinus rhythm exhibited greater reductions in NT-proBNP levels over follow-up. NT-proBNP reduction in patients maintaining sinus rhythm is consistent with reduced atrial wall stress. Prior studies demonstrate that natriuretic peptides improve prediction of AF recurrence [17]. While this observation is biologically plausible, reflecting reduced atrial wall stress, it was exploratory and not adjusted for confounding variables. Consequently, we cannot definitively establish the independent predictive value of NT-proBNP in this cohort. Future investigations must incorporate these critical clinical variables into multivariable models to determine whether NT-proBNP provides incremental prognostic information beyond established risk markers.

Subgroup analyses (Fig. 4) showed no evidence of effect modification of the association between LAD and 3-month AF recurrence by sex (P for interaction = 0.85), age group (P for interaction = 0.91), AF type (P for interaction = 0.52), or heart failure status (P for interaction = 0.22). Given the exploratory nature of these analyses, no adjustment for multiple comparisons was applied, and the subgroup findings should be interpreted cautiously.

4.6. Study Limitations and Future Directions

Several limitations merit consideration. First, the single-center observational design limits generalizability and precludes internal validation. Second, our multivariable model was constrained by the modest sample size (N = 96) and event rate (n = 39); to avoid overfitting, we restricted the number of covariates, which inherently limits our ability to comprehensively adjust for all latent confounders and necessitates that findings be interpreted as associative rather than causal. Third, intermittent ECG monitoring may have underestimated asymptomatic recurrences. Finally, given our cohort size, we were unable to perform robust, sex-specific analyses; future large-scale, prospective registries are required to evaluate whether these outcomes differ by sex. In addition, all participants were of Kinh ethnicity, which may limit the generalizability of the findings to other ethnic populations.

However, these limitations do not undermine the clinical relevance of our work. From a practical perspective, LAD remains an inexpensive and widely available echocardiographic parameter that may assist preliminary risk assessment, particularly in resource-constrained settings where advanced imaging is not routinely accessible. Future multi-center, prospective studies incorporating larger cohorts and longitudinal follow-ups are necessary to validate our proposed LAD cut-off values and to further explore the interplay between metabolic risk factors and atrial remodeling in Southeast Asian populations.