Effect of Prolonged and Substantial Weight Loss on Incident Atrial Fibrillation: A Systematic Review and Meta-Analysis

Background. Overweight and obesity are associated with atrial fibrillation (AF), and bariatric surgery (BS), able to induce sustained and prolonged weight loss, might represent the ideal treatment in the prevention of AF. Previous studies could not definitely establish a role for weight loss and BS in preventing incident AF so far. During the last few years, several studies on the effect of bariatric surgery on cardiovascular diseases have been published, and we performed a systematic review and meta-analysis to evaluate the role of weight loss through BS in the prevention of incident AF in obesity. Methods. This meta-analysis followed the PRISMA guideline. Eligible studies were controlled trials evaluating the appearance of atrial fibrillation in patients undergoing weight loss through BS as compared with patients receiving medical treatment. Quality of studies was assessed according to the Newcastle-Ottawa Quality Assessment Scale, and risk-of-bias was evaluated employing the Egger’s test. All analyses were run by a random-effects model according to Hartung and Knapp and sensitivity analyses were performed. Heterogeneity was assessed through Q and I2 statistics for each comparison, and potential publication bias was formally investigated. Results. Ten studies were included in the meta-analysis, and the overall result was statistically significant [OR = 0.665 (0.475–0.929), p = 0.017], with significant heterogeneity (Q = 48.98, p < 0.001; I2 = 81.6%), but with no publication bias. In sensitivity analyses, the amount of weight loss, percentage of patients with diabetes and value of the Newcastle-Ottawa Quality Assessment Scale, were all associated with significance of effect. Since age was different in one study, a sensitivity analysis was performed by excluding this study; OR was similar [OR = 0.608 (0.454–0.814), p < 0.001]; heterogeneity was reduced but still significant (Q = 35.74, p < 0.001, I2 = 77.6%) and again no publication bias was detected. Conclusions. Bariatric surgery as compared to medical treatment is associated with reduced appearance of incident AF.


Introduction
Overweight and obesity are associated with higher prevalence of co-morbidities, including cardiovascular diseases (heart failure, ischemic heart disease, stroke), renal failure and cancer, and with reduced life expectancy and reduced quality of life [1]. Recent studies have highlighted the association between obesity and atrial fibrillation (AF), showing that two different kinds of AF (incident AF and persistent/recurrent AF) can occur in obesity. Various studies have shown that obesity is associated with an increased prevalence of incident AF and persistent/recurrent AF [2,3]; in addition, weight gain is associated with increased prevalence of incident AF [4], while no data are available for persistent/recurrent AF. In addition to being a risk factor for the development of well-known comorbidities,

Search Strategy and Inclusion Criteria
This meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline [24]. Eligible studies were controlled trials evaluating the appearance of atrial fibrillation in patients undergoing weight loss through bariatric surgery as compared with patients receiving medical treatment. Full reports, published in any language, were included. Two independent authors (AEP, LC) searched relevant literature in databases including PubMed, Embase, and Cochrane Library from inception until 30 October 2022. The following keyword was used for disease: atrial fibrillation. To define exposure, the following keywords were used: obesity, overweight, weight loss, weight reduction, loss of weight, decrease in weight, weight decrease, weight changes, changes in weight, bariatric surgery (laparoscopic adjustable gastric banding, gastric bypass, sleeve gastrectomy, biliopancreatic diversion, biliointestinal bypass). The title and abstracts written in English language were reviewed to recognize eligible studies on the impact of weight loss on the appearance of atrial fibrillation. Additional studies were also manually searched through the references cited in reviews. Cohort as well as case-control studies were included. If the results of one study were reported in more publications, only the most recent and complete data were included in analysis. When required, authors of the studies were also contacted by mail to yield more details. The following studies were excluded: descriptive studies, editorials, review articles, systematic reviews and metaanalyses, and studies that did not provide risk ratios or effect sizes. Decisions on trials to include were taken by the authors (AEP, LC, CLR, SM, ET, FF), and disagreements were resolved by discussion. The reason for exclusion of other trials was specified (lack of details, no controls, pre-existing AF, Figure 1). Ten studies [22,[25][26][27][28][29][30][31][32][33] fulfilled the inclusion criteria (Table 1 shows details of the 10 studies included in this meta-analysis). The protocol of the meta-analysis has been registered (Prospero, CRD42022374548). changes, changes in weight, bariatric surgery (laparoscopic adjustable gastric banding, gastric bypass, sleeve gastrectomy, biliopancreatic diversion, biliointestinal bypass). The title and abstracts written in English language were reviewed to recognize eligible studies on the impact of weight loss on the appearance of atrial fibrillation. Additional studies were also manually searched through the references cited in reviews. Cohort as well as case-control studies were included. If the results of one study were reported in more publications, only the most recent and complete data were included in analysis. When required, authors of the studies were also contacted by mail to yield more details. The following studies were excluded: descriptive studies, editorials, review articles, systematic reviews and meta-analyses, and studies that did not provide risk ratios or effect sizes. Decisions on trials to include were taken by the authors (AEP, LC, CLR, SM, ET, FF), and disagreements were resolved by discussion. The reason for exclusion of other trials was specified (lack of details, no controls, pre-existing AF, Figure 1). Ten studies [22,[25][26][27][28][29][30][31][32][33] fulfilled the inclusion criteria (Table 1 shows details of the 10 studies included in this metaanalysis). The protocol of the meta-analysis has been registered (Prospero, CRD42022374548).

Figure 1.
Flow-chart of the analysis performed according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA). After screening of literature according to search terms, most paper were excluded for reasons indicated in the squares. * Search terms: "atrial fibrillation" AND ("bariatric surgery" OR "biliopancreatic diversion" OR "gastric bypass" OR "roux-en-Y" OR "sleeve gastrectomy" OR "biliointestinal bypass" OR "adjustable gastric banding") ** AF atrial fibrillation.

Figure 1.
Flow-chart of the analysis performed according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA). After screening of literature according to search terms, most paper were excluded for reasons indicated in the squares. * Search terms: "atrial fibrillation" AND ("bariatric surgery" OR "biliopancreatic diversion" OR "gastric bypass" OR "rouxen-Y" OR "sleeve gastrectomy" OR "biliointestinal bypass" OR "adjustable gastric banding") ** AF atrial fibrillation.

Data Extraction
The following data were extracted: name of the first author, year of publication, type of study, type of bariatric surgery, number of patients who have undergone surgery and number of control patients. For both groups additional items were extracted: percentage of male and female patients, duration of follow-up, mean age, mean BMI (for the first group before surgery and for control at the start of follow-up), percentage of weight lost during follow-up, percentage of patients diagnosed with type 2 diabetes, with hypertension, with coronary artery disease and with heart failure. The details for each study are shown in Table 1. For the analysis, event rate was extracted per outcome parameter for each group and unadjusted and adjusted HRs with their 95% confidence intervals (CIs) for the association with outcome of interest.

Quality Assessment
Quality of reports was assessed independently by reviewers according to Newcastle-Ottawa Quality Assessment Scale (NOS) for Cohort Studies [34]. The NOS scale is based on a "star system" in which a study is judged from three broad perspectives: study group selection, group comparability, and ascertaining the outcome of interest. The variables considered are: risk of bias linked to the selection of participants, confounding variables, performance, detection and measurement of exposure, attrition and reporting biases. The length of follow-up was set at a minimum of 5 years to be evaluated as adequate. Disagreement for the quality assessment was resolved by discussion. A score was eventually built, classifying poor, intermediate, or good quality, based on the number of the above criteria available for each paper. The NOS score is reported in Table 2.

Statistical Analysis
Intervention effect (weight loss vs. controls) was expressed as odds ratio (OR), with 95% confidence intervals (CIs); all analyses were performed by a random-effects model according to Hartung and Knapp [35]. Heterogeneity was assessed through Q and I 2 statistics for each comparison, and potential sources of heterogeneity were discussed where appropriate [36]. Heterogeneity was considered statistically significant for a p value < 0.05. Through meta-regression, we also evaluated, the possible role of several patients' and study characteristics on the incidence of new cases of atrial fibrillation (AF). This was done independently of statistically significant heterogeneity. The dependent variable was the incidence of cases of AF. The role of each covariate in heterogeneity was expressed by Wald test estimated by the meta-regression. The following covariates were included in the meta-regression analysis: number of patients enrolled, age, kind of study (prospective or retrospective), percentage of persons with diabetes mellitus, NOS score, duration of follow-up, body mass index (BMI) of each study (weighted means of intervention and control patients), amount of weight loss and efficacy of treatment (vs. controls) in each study. Meta-regression was performed considering all studies together. In a secondary analysis, we also evaluated the existence of a potential publication bias, that means the tendency of authors and editors to publish studies in which the experimental results achieved statistical significance, more favorably than in studies in which the results were not significant, which would ultimately introduce bias into the overall published literature [37]. Funnel-plot asymmetry was evaluated by using the Egger's test for small study effects through the meta-bias routine [38]. Sensitivity analyses were also performed by considering the prevalent kind of BS employed in the studies [i.e., restrictive surgery (VGB = vertical banded gastrotomy; AGB = adjustable gastric banding), metabolic surgery (LSG = laparoscopic sleeve gastrectomy; RYGB = roux-en-y gastric bypass), or a mixture of procedures (GA = gastrectomy; GR = gastric resection; GIB = gastric intestinal bypass) plus AGB, LSG and RYGB], percentage of patients with diabetes mellitus, percent weight loss, and exclusion of studies showing substantial variation in basal conditions from other studies. To this end, the Baujat method was used to identify outliers in this model [39]. The Baujat method detects outliers by analyzing the change of the general effect by systematically leaving out one study at a time, i.e., the contribution of this study to the between-study heterogeneity statistic Q, and contribution to I 2 [40]. A leave-one-out sensitivity analysis was also conducted to assess each study's influence on the pooled estimate by omitting one study at a time and recalculating the combined estimates for the remaining studies.
No publication bias was detected. In summary, a fair part of the heterogeneity was due to one study [28]. Finally, also when other methods of analysis were employed, namely fixed model (RR) instead of random model (OR) or direct analysis of RR as reported by the authors of the studies (ES), the effect was always similarly significant (Supplementary  Table S2). Supplementary Table S3 reports the Newcastle-Ottawa Quality Assessment Scale (NOS).
Supplementary Table S4 shows a comparison of baseline clinical conditions of patients undergoing bariatric surgery (BS) and controls in the ten studies included in this metaanalysis. A total of 61,197 patients (22,831 patients undergoing BS and 38,366 control patients receiving medical treatment) was considered; from the Table it appears that the former and the latter patients were well matched as to age, BMI, duration of follow-up, percentage of patients with diabetes, with hypertension, with coronary heart disease, with heart failure; as expected, weight loss was different in BS and in control patients.

Discussion
Both incident AF and persistent/recurrent AF can occur in obesity. There are similarities and differences; for instance, obesity is associated with increased prevalence of both incident AF and persistent/recurrent AF [2,3], and weight gain is associated with incident AF [4] while no data are available for persistent/recurrent AF. The reasons why obesity is associated with AF are still largely unknown; obesity is a risk factor for the development of several comorbidities, namely diabetes, arterial hypertension and obstructive sleep apnea, dyslipidemia, that can lead to cardiovascular disease such as coronary heart disease and heart failure [41]; obesity may increase the risk of AF through several mechanisms, including structural and electrical remodeling, which contribute to the development of the arrhythmogenic milieux. This has been shown in experimental studies in sheep [5], in which increased pericardial fat volumes are associated with infiltration of epicardial fat into the myocardium, resulting in potential voltage abnormalities, conduction block, and increased AF vulnerability [6].
It is well established that bariatric surgery induces sustained and prolonged weight loss and thus might represent the ideal treatment to prevent AF through weight loss. Weight loss, especially sustained weight loss through bariatric surgery, can be beneficial in reducing the progress of incident AF to recurrent AF [18,19], but large studies and even meta-analyses failed to establish a role of weight loss in preventing incident AF [20,21].
The present meta-analysis shows that bariatric surgery, compared to medical treatment, is associated with reduced appearance of incident AF, similar to what has been seen for persistent/recurrent AF. The effect was statistically significant in six out of ten studies analyzed, and the overall effect was statistically significant. In addition, the effect was related to the percentage of weight lost and was more evident for cohorts not entirely consisting of patients with diabetes. Lastly, sensitivity analysis showed that after exclusion of one single study [32], the results were even more statistically significant. At metaregression, only age negatively correlated with effect, and at higher age the effect was lower. After exclusion of one study [32], in which age of the patients was higher than the rest of the studies, and contribution to heterogeneity was significant [39,40], only BMI correlated with effect, indicating a greater effect for higher BMI. All sensitivity analyses reported in the Supplementary File show that the effect was statistically significant under most conditions, and that most of the negative results were due to one single study that was different from others, mainly for the age of the patients. In addition, the effect was similar when calculated as a random model, fixed model, or when evaluated through risk ratios (RR)s as published by the authors of the papers considered in this meta-analysis.
This meta-analysis has several limitations. The first limitation is that, from the studies included, we do not have data on the duration of obesity, even though age of BS and control patients was not different. Secondly, we could only analyze obesity as expressed by body mass index, not through measures of fat distribution [42]. Although BMI is strongly correlated with percent body fat across populations, there are limitations in its predictive ability to estimate body fat for any given individual, in particular in regard of visceral adiposity. In fact, visceral adiposity as measured by WC, WHR, or detailed imaging methods has been shown to be a risk factor independently of BMI [43]. Thirdly, BS is associated with loss of fat-free mass together with fat mass, and muscle loss is an important component of fat free mass [44]. Under conditions of accelerated loss of fat and muscle mass, such as thyrotoxicosis, an increased frequency of accelerated heart rhythm and AF are observed [45]. With BS, no significant changes of thyroid hormones are observed [46], and the only involvement of heart muscle seems to be a remodeling of heart chambers, a situation that does not increase the risk of AF [47]. Finally, different metaanalyses led to different results regarding the effect of BS on incident AF. The main reason is probably the number of studies considered; the strength of our study is that 10 studies were included, while in two previous meta-analyses [20,21] only seven studies were considered. Differently from previous meta-analyses [2,4,[18][19][20][21], all analyses were performed by a random-effects model according to Hartung and Knapp [35], that is straightforward and outperforms the standard DerSimonian-Laird method [48], especially when less than ten studies are considered [49]; also, when other methods of analysis were employed, such as fixed model (RR) analysis, or direct analysis of RR as reported by the authors of the studies (ES), the effect was similarly significant; finally, one study [32] was excluded in this meta-analysis because of a clear difference of age of patients compared to other studies. All these differences deserve a comment; first, the role of age might be of importance in AF, as it is in other cardiovascular diseases. In keeping with this possibility, we have shown for the first time that bariatric surgery is associated with reduced all-cause mortality and Nutrients 2023, 15, 940 9 of 12 reduced cardiovascular mortality [7]; however, in a second meta-analysis, we were able to show that age can be of importance in affecting prevention of all-cause mortality and of mortality due to cancer and diabetes [8]. A different possibility is that the amount of weight loss is important, as indicated in this meta-analysis and as reported by Jones et al. [50], who showed that a weight loss of less than 5% is not able to prevent incident AF. The quality of studies assessed through the NOS score might also be of importance, as indicated by our sensitivity analysis; in fact, the effect was statistically significant for higher NOS score (i.e., above mean NOS), but was non-significant for NOS score below mean NOS score.
One should also remember that bariatric surgery does not only mean prolonged and substantial weight loss; bariatric surgery is associated with reduced incidence of diabetes, cardiovascular disease, cancer, and liver and kidney disease [9], effects that occur regardless of the patients' metabolic status [10].
In summary, bariatric surgery is associated with reduced incident AF in patients affected by obesity. This might be instrumental in reducing cardiovascular morbidity and mortality in people with obesity [41], but more studies are needed to determine the amount of weight loss needed to prevent AF. Moreover, a causal relationship still needs to be confirmed between weight loss and prevention of AF; bariatric surgery is associated, beyond weight loss, with other important changes, namely improved endothelial function [12] and reduced sub-inflammation [13], reduced prevalence of arterial hypertension [14] and of left ventricular hypertrophy, improved glucose tolerance and reduced insulin resistance [14,16], improved quality of life [15]. It remains to be established if these factors, not clearly defined in the studies considered, could also contribute independently from weight loss in preventing AF.

Conclusions
Bariatric surgery as compared to medical treatment is significantly associated with reduced incident AF, possibly mainly through prolonged and sustained weight loss. The present report is of great value, and consistent with previous data demonstrating that prolonged and sustained weight loss prevent progression of incident AF to recurrent AF.  Figure S1: Meta-analysis after exclusion of study [32]; Supplementary Table S3: The Newcastle Ottawa Scale [34]; Supplementary Table S4: Comparison of baseline conditions of patients undergoing bariatric surgery (BS) and controls in the ten studies included in this meta-analysis.

Author Contributions:
The study was a project of A.E.P., C.W.L.R. and F.F., A.E.P. and L.C. searched literature, E.T. performed meta-analyses. C.W.L.R. and S.M. improved and supervised the project. All authors contributed to writing the MS. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.

Institutional Review Board Statement:
This study was exempt from ethics approval as only data from previously published studies were retrieved and synthesized.

Informed Consent Statement: Not applicable.
Data Availability Statement: Data are available from the corresponding author on the basis of reasonable request.