This research was conducted in accordance with the guidelines of the 2020 Systematic Review and Meta-Analysis Preferred Reporting Project (PRISMA 2020)(19) (Supplementary Table 1) .The protocol has been registered in PROSPERO (Prospective Registration for International System Evaluation.https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=273682. CRD4202173680)
PubMed, Cochrane Library, Embase database and conference articles (American Heart Association: https://www.ahajournals.org/journal/circ, American College of Cardiology: https://www.jacc.org/ and European Society of Cardiology: https://www.escardio.org/) were searched by using the following MeSH to retrieve articles up to August 17, 2021, includes full text and conference abstracts, without language restrictions.
For patients: “atrial fibrillation”, “atrial flutter”, “atrial tachycardia”, and “ablation”
For exposure/intervention: “weight loss”, “weight reduction”
For outcomes: we did not apply any keywords for outcomes because all reported outcomes related to AF ablation were included, such as AF recurrence, AF severity, or
quality of life.
Supplemental Table 2 provided the detailed search strategy.
We used Endnote X8 database, a reference management software, to organize all studies. All titles and abstracts were reviewed to consider eligible for inclusion. And a full-text evaluation was presented after initial identification.
Eligible studies had to fulfil the following criteria: 1) Clinical trials or observational studies; 2) Study on the relationship between weight loss and outcomes after AF ablation; 3) The patients in this study were adults (age > 18 years), diagnosed with AF, and undergo catheter ablation with weight management;4) reported the relationship between weight loss and AF recurrence and other outcomes (AF severity, quality, symptoms); 5) The literature reported odds' ratio (OR), RR, hazard ratio (HR), and the 95% CI provided available data to calculate the estimation effect for the AF recurrence.
Additionally, we excluded studies with:
1) For multiple reports based on the same data source, we excluded studies with the shorter follow-up time or smaller sample size.
2) Case-control design due to the potential bias.
Data extraction and quality assessment
Studies were reviewed by two independent authors (X.Z-L and X-L) according to the above inclusion and exclusion criteria. Disagreements were resolved by consensus. Data were extracted by 2 investigators (X.Z-L and X-L), including first author, publication year, country, follow-up time, demographic characteristics (sample size, average age, gender, body mass index, left atrial diameter, AF type, history of diabetes, history of hypertension, high history of lipemia), study design, data source, methods of weight loss and AF diagnosis, outcomes, corresponding 95% CI and estimate effect, and adjustments.
The quality of the included studies was assessed according to the Newcastle-Ottawa
evaluation scale (NOS). Scores range from 0 to 9, with NOS scores greater than 7 being considered high quality.(20)
Statistical analysis and bias risk assessment
Review Manager (Version 5.1., The Nordic Cochrane Center, The Cochrane Collaboration, Copenhagen, Denmark, 2011.) and Stata 16.0 (Stata Corp LP, College Station, TX, USA) was used for statistical analysis. RR were used to combine the estimated effects of random effects models. We estimate the adjusted RR's by calculating the natural logarithm of RR (log [RR]) and its standard error (SE log [RR]). For those studies that did not provide RRs, we calculated crude RR by event and total number. In addition, we performed subgroup analysis stratified by gender, methods of weight loss, types of AF.
For dose-response analysis, we computed summary RRs and 95% CIs for a 5% in weight loss 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 weight loss by using the method of Greenland and Longnecker11. We performed the non-linear dose-response analysis by using the robust error meta-regression method described by Xu et al.(12 )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 weight loss and RRs with variance estimates for at least two quantitative exposure categories.(12 )If the median or mean weight loss was not provided and reported in ranges, we estimated the midpoint of each category by averaging the lower and upper boundaries of that category. If the highest or lowest category was open-ended, we assumed that the open-ended interval length was the same as the adjacent interval.14 We used I2 statistics to estimate heterogeneity between studies. I2 < 50%, I2 at 50–75%, I2 > 75% were regard as low heterogeneity, moderate heterogeneity and high heterogeneity,(21) respectively. Egger’s, Begg’s, or Funnel plot were used to detect publication bias. P < 0.05 with two tails is considered statistically significant.