Two authors (FL and XL) independently systematically searched the Cochrane Library, PubMed, and Embase databases for eligible studies until October 5, 2021. Three groups of keywords (linked to BMI, AF, radiofrequency ablation, respectively) were combined using the Boolean operator “and.” In addition, we searched the reference lists of three previous published meta-analyses (12, 13, 16) or other relevant publications to identify further studies. All discrepancies were resolved through discussion with each other. No language restrictions were applied in the whole literature search. The detail search strategy was provided in Supplementary Table 2.

According to the PICOS (population, intervention, comparison, outcome, and study design), the selection criteria were as follows: (i) population: patients with AF undergoing radiofrequency ablation; (ii) and exposure and comparison: high vs. low BMI level; (iii) outcomes: reported the association between of BMI on AF recurrence. No blank period was pre-defined: (iv) study design: cohort, nested case-control, or clinical trials. For multiple publications/reports created from the same data, the studies with the longest follow-up period or the largest number of AF cases were included. We only included studies with multivariate analysis. Case-control and univariate analysis studies were excluded considering their larger bias. Certain publication types (e.g., reviews, editorials, and animal studies), or studies with insufficient data were excluded from this study. The details reasons of excluded studies were listed in Supplementary Table 3.

Basic information of each study, including the authors, publication year, region, study design, participants (sex, age), follow-up time, adjustments for confounders, categories of BMI and adjusted risk ratios (RRs) with its 95% confidence intervals (CIs) for each BMI category were extracted. If multi-adjusted RRs were reported in one study, we extracted the most completely adjusted one.

The quality of all included observational studies were assessed by using Newcastle-Ottawa quality assessment scale (NOS) (19). The validated NOS items with a total of 9 stars involved three aspects including the selection of studies, the comparability of cohorts, and the assessment of the outcome. A NOS score of ≥7 stars was considered as acceptable quality, otherwise, as low-quality studies (20).

Summary RRs and 95% CIs for a 5-unit increment in BMI were using a random effects model. Study-specific slopes (linear trends) and 95% CIs from the natural logs of the reported RRs and CIs across categories of BMI were calculated by using the method of Greenland and Longnecker (21). Non-linear dose-response analysis were performed by using the robust error meta-regression method described by Xu and Sar (22). This method is based on a “one-stage approach” which treating each study as a cluster of the whole sample and considering the within study correlations by clustered robust error. It requires known levels of BMI and RRs with variance estimates for at least two quantitative exposure categories. For studies that did not set the lowest BMI group as a reference, data were transformed using a method described by Hamling et al. (23) which requires the number of cases and participants in each category. If these data could not be obtained from an article, the evidence was not pooled. If the median or mean BMI was not provided and reported in ranges, we estimated the midpoint of each category by averaging the lower and upper boundaries of that category (20). If the highest or lowest category was open-ended, we assumed that the open-ended interval length was the same as the adjacent interval. To assess the heterogeneity of RRs across studies, the I² (95% CI) statistic was calculated with the following interpretation: low heterogeneity, defined as I² < 50%; moderate heterogeneity, defined as I² 50–75%; and high heterogeneity, defined as I² > 75% (24, 25). Publication bias was explored by Egger’s test, Begg’s test, and funnel plot. Sensitivity analysis was performed by excluding a bank period < 3 months and leave one-out methods to confirm the robustness of primary analysis. P-value < 0.05 was considered statistically significant.

As shown in Figure 1, we identified 1,723 articles through an initial database search (Cochrane Library = 30, PubMed = 332, and EMBASE = 1,361) (Figure 1). After removing duplicate articles (n = 801), 922 studies remained. We further excluded 863 records by quickly screening the titles and abstracts, and 59 articles were reviewed in the detailed evaluation. Of the 59 records, 33 were excluded after the full-text review for the following reasons: (1) Reviews or meta-analyses (n = 2); (2) without target populations (n = 1); (3) without target exposure (n = 11); without target outcome (n = 3); (4) publications with insufficient data (n = 2); (5) studies based on duplicate population (n = 3); (6) case-control studies (n = 2); (7) univariate analysis (n = 9). Supplementary Table 3 provides the detailed reasons for exclusion following the full-text review. Finally, 26 (6–11, 26–45) studies were included in present study.

Table 1 shows the baseline characteristics of the patients included in the study. Overall, the publication years ranged from 2008 to 2021, with ten conducted in North America (the USA and Canada), eight in Asia, and eight in Europe. The age range of patients in the included studies was 52–65 years, sample size range was 72–3,333, and follow-up time range was 3–42 months. Among the 26 articles, 14 and 12 was prospective and retrospective cohort studies, respectively. AF recurrence in the included studies was detected by using 12-lead ECG or 24 h Holter monitoring.

The quality of the articles was acceptable, with a majority (n = 23) of studies scores Newcastle-Ottawa Scale ≥ 7 points (Supplementary Table 4).

Twenty-six (6–11, 26–45) studies including 7,878 cases/26,450 individuals were included in the dose-response analysis. The summary RR for a 5-unit increase in BMI was 1.15 (95% CI: 1.08–1.22, I² = 53%), with moderate heterogeneity (Figure 2). There was little evidence of heterogeneity (I² = 34%) after excluding studies from Cai et al. (29) and Kang et al. (7), but a positive association remained (RR = 1.12, 95% CI: 1.07–1.18). In the non-linear analysis, we found a borderline positive linear association between BMI and AF recurrence after ablation (P_{non–linearity} = 0.12) (Figure 3). Supplementary Table 5 displays the RR estimates and 95% CIs from the specific BMI values derived from the dose-response figures.

There may be a stronger association between BMI and AF recurrence among groups with larger AF recurrence cases (P_{heterogeneity} = 0.006). The positive association between BMI and AF recurrence persisted in almost all subgroups stratified by age, time of follow-up, region, and sample size after adjusting for confoundings (e.g., OSA), and there was little evidence of heterogeneity across these subgroups (Table 2).

In the sensitivity analysis, the pooled results were stable when omitting one study at a time, with an RR range from 1.13 (95% CI: 1.07–1.20, I² = 47%) to 1.16 (95% CI: 1.09–1.25, I² = 53%). The result was also stable by excluding studies with a blank period < 3 months (RR: 1.15, 95% CI: 1.08–1.23) (Supplementary Figure 1). Publication bias was assessed using Egger’s test (p = 0.004), Begg’s test (p = 0.019), or funnel plots (Supplementary Figure 2). However, metatrim analysis showed that the addition of six missing studies did not significantly change the results, with a decreased magnitude of the pooled effect size (RR: 1.11; 95% CI: 1.03–1.18) (Supplementary Figure 3).

Based on available cohort studies, we showed a borderline positive linear dose-response relationship between baseline BMI and AF recurrence in patients after ablation. We found a 15% increase in AF recurrence per 5-unit increase in BMI. These results were confirmed in all subgroups stratified by age, sex, study design, region, follow-up duration, sample size, and adjustment. Overall, our results showed that overweigh (BMI > 28 kg/m²) or obesity was independently positively associated with AF recurrence after ablation.

Several well-known risk factors, including age, hypertension, diabetes, and enlarged LAD, have been linked to the recurrence of AF after ablation. Moreover, these comorbidities were common in patients with obesity. Therefore, whether the role of obesity is independent of these comorbidities is still unclear. For example, Jongnarangsin et al. showed that obesity was not associated AF recurrence post AF ablation after adjustments (26), suggesting obesity may simply be a phenotypic marker of profibrillatory comorbidities and not the key predictor of ablation failure. Moreover, although several previous meta-analyses (12, 13, 16) showed that obesity was associated with an increased risk of AF, none of them conducted a stratified analysis according to these confounding factors; therefore, this relationship is still under debate. We showed that the positive association of BMI and AF recurrence after ablation was still positive in all subgroups after adjustments, including those for hypertension, diabetes, age, OSA, and LAD (Table 2), which suggested that obesity was a factor contributing to AF recurrence, independent of conventional risk factors.

Although we found a linear association between BMI and AF (p for non-linearity = 0.12), we should explain these results with caution. As shown in Figure 3, we should figure out that the dose-response curve is much steeper at the BMI > 35. Specifically, a change in BMI of 25–30 kg/m² increases the RR from 1.08 to 1.38, while RR increased sharply (1.95–2.78) between the 35 and 40 kg/m² BMI group, somewhat reflecting a non-linear trend at higher BMI values. This observation reinforces the previous findings, showing severe obesity was a strong risk factor for cardiovascular diseases, including incident AF. A previous meta-analysis (14) also examined the dose-response analysis of BMI and risk of AF recurrence, but it failed to demonstrate the existence of a linear dose-response relationship. Although we demonstrated a borderline linear relationship between BMI and AF recurrence, the study demonstrated a dramatic increase in the risk of AF recurrence when BMI was greater than 35 kg/m², so we would carefully interpret the borderline linear association. We should include more relevant studies. The non-linear p value might be significant if more population with high BMI were include. Overall, we should point out that grade II obesity substantially increased the risk of AF relapse after ablation. Weight loss intervention might be encouraged for these patients with severe obesity to reduce rate of AF relapse. 2020 European Society of Cardiology guidelines did not set a BMI threshold value for weight loss in obese patients who plan to receive AF ablation, while our results provide a valuable information for this.

In the context of AF ablation, obesity was not a new topic. Several reviews have shown that obesity increases the risk of AF relapse after ablation. Three showed that the obese group had a significantly increased risk of AF recurrence (13, 15, 16). In another excellent review by Wong et al., for every 5-unit increase in BMI, there was a 13% greater excess risk of post AF ablation (12). A recent meta-analysis included 12 studies reported a positive non-linear dose-response relationship between BMI and AF ablation (14). Interestingly, with the inclusion of more studies (N = 26), our results were opposite to theirs, showing a borderline positive linear dose-response relationship (P = 0.16), revealing a threshold effect between BMI and the risk of AF recurrence after ablation. In addition, the five published meta-analyses did not take into account confounding factors such as age, hypertension, AF category, diabetes, OSA, and LAD and did not demonstrate whether obesity could independently increase the risk of AF recurrence. We showed that the positive association of BMI and AF recurrence after ablation was still positive in all subgroups after adjustments, which suggested that obesity was a factor contributing to AF recurrence, independent of conventional risk factors.

Several potential mechanisms could explain this association. First, obesity could result in many changes involving various domains, such as hemodynamics, neurohumoral factors, inflammatory factors, metabolic factors, adipokines, and autonomics. For example, obesity leads to a high cardiac output state and the presence of left ventricular hypertrophy (eccentric or concentric) in association with left ventricular diastolic dysfunction (46). The hemodynamic changes associated with the elevation in left atrial and systolic blood pressure and left ventricular diastolic dysfunction contribute to atrial stretch and “trigger” AF (47). Furthermore, obesity is considered a state of chronic low-grade inflammation (48, 49). Studies have shown that patients with a higher BMI exhibit increased levels of several inflammatory and oxidative stress markers, including high-sensitivity C-reactive protein, serum creatinine, fibrinogen, and uric acid (50). Numerous studies have also demonstrated the implication of inflammation and oxidative stress in the pathophysiology of AF (51, 52). Therefore, it is reasonable to assume that obesity-associated inflammation may contribute to AF recurrence after ablation. Finally, it must be noted that the pathophysiological mechanisms linking obesity and AF are highly complex and remain incompletely understood. It is likely that a combination of multiple factors previously mentioned contributes to the recurrence of AF after ablation.

Some authors proposed that obesity was a phenotypic marker of OSA that is associated with AF recurrence. Obesity is an important risk factor for the incidence of OSA (53), and the association between OSA and AF has been explored extensively (28). A meta-analysis reported that patients with OSA had a 25% greater risk of AF recurrence after ablation than those without OSA (53). The present study showed that the positive association was significant regardless of adjustment for OSA, which suggested that obesity increased the risk of AF recurrence independently of OSA. Therefore, we hypothesized that OSA may not be a mechanism by which obesity results in AF recurrence. Furthermore, a large retrospective cohort study also showed that obesity remains a powerful predictor of incident AF in patients without OSA based on the gold standard diagnostic test (54). This finding was consistent with our results, which suggested that obesity contributes to AF recurrence after ablation independent of OSA. However, considering the limited studies in the subgroup (N = 7), further prospective studies are needed to confirm our results.

The association between underweight and AF relapse is still inconclusive. Based on a retrospective cohort study in China, Deng et al. showed that underweight (BMI < 18.5 kg/m²) significantly increased the risk of AF recurrence (HR 1.85, P = 0.02) (11). In contrast, by using a different definition of underweight (BMI < 20 kg/m²), Bunch et al. did not find an elevated risk of AF recurrence of underweight in the US (33). However, the scarcity of studies precludes us from performing a dose-response analysis, which may be attributed to the high prevalence of obesity in Western countries. However, evidence from several population-based studies has shown that underweight is a risk factor for AF. Therefore, although the results regarding underweight and AF relapse were inconclusive, we supposed there may be a positive association between underweight and the risk of AF recurrence in patients undergoing ablation. More studies are warranted to address this issue.

In addition, although BMI is widely adopted because it is convenient to apply and inexpensive, BMI is not perfect as a measure of obesity. Because BMI is an indirect measure of obesity, it cannot distinguish between fat and lean tissue and can produce a certain amount of error (55). The problem may lead to misclassification of the experimental population in terms of body fat and may introduce bias. The impact of other parameters of body fat and body composition (e.g., waist circumference and waist-hip ratio) on outcomes among patients undergoing ablation should be further assessed by future studies.

As shown in our results, the likelihood of AF recurrence after ablation increasing with BMI. In patients with grade II obesity (BMI > 35 kg/m²), the risk of AF recurrence increased by 95%. Our findings provide valuable insight for the clinical prevention of AF recurrence after ablation. For patients with severe obesity (BMI > 35 kg/m²), cardiologists should consider that their likelihood of AF relapse is much higher than that of individuals with a normal BMI. Therefore, for those patients strongly considering AF ablation, the risks and benefits should be carefully evaluated.

Weight loss is associated with improvements in risk factors (e.g., hypertension, OSA, and glycemic control), decreased morbidity and mortality, and a decreased risk of AF. Evidence from observational studies, such as the Long-Term Effect of Goal-Directed Weight Management in an Atrial Fibrillation (LEGACY) study, showed that long-term weight loss (bariatric surgery or weight management) could reduce AF risk and burden in people with AF (56). Moreover, the benefit of weight loss from the prevention of AF was dose-dependent. A 10% weight loss with accompanied risk factor modification reduced the risk among arrhythmia-free patients by 45% compared with that associated with 3% weight loss (56). Therefore, current guidelines set the weight loss target at > 10% weight reduction with lifestyle modifications for BMI < 27 kg/m² (18) to improve AF ablation. However, weight loss, either through lifestyle or bariatric surgery, is typically a long process. It is well-described that increasing the duration of AF has negative effects on ablation success rates. The time required for weight loss would have resulted in a delay of ablation of at least 6 months, which would have led to AF progression with worsening of success rates. Therefore, it is not recommended for all patients with overweight or obesity to lose weight; the benefits and risks should be carefully considered.

The strengths and limitations of this study merit careful consideration. This is the first dose-response meta-analysis that demonstrated a positive linear association between BMI and AF recurrence after ablation with a large sample size and robustness of the findings based on multiple subgroups (e.g., age, sex, duration of follow-up and adjustment for confounding and potential intermediate factors). However, it was limited by the inclusion of observational studies, which cannot definitively prove causation. Although most studies were adjusted for other comorbidities, it is impossible to fully take into account all confounding factors. For example, the dose of antiarrhythmic medication may differ between studies, and studies have shown that LA volume is more accurate than diameter. Second, several studies have shown that underweight and AF patients have worse outcomes (11). However, due to data restrictions, the impact of underweight (BMI < 18.5) on AF recurrence after ablation was not analyzed and needs to be further investigated. Third, BMI may not accurately indicate the degree of body adiposity, and the impact of other parameters of body fat and body composition (e.g., waist circumference and waist-hip ratio) on outcomes among patients undergoing ablation should be further assessed by future studies. Fourth, ablation techniques across studies may also influence our results. Overall, further well-designed, larger clinical trials are needed to confirm our results.