MEDLINE, Embase and the Cochrane Central Register of Controlled Trials databases were searched for RCTs that compared catheter ablation with AADs for the treatment of AF, without language restrictions (last search update, May 1, 2014). In addition, reference lists from initially identified articles were retrieved in order to avoid the exclusion of any relevant studies. The following medical subject heading terms were used: ‘Atrial fibrillation’, ‘catheter ablation’ and ‘randomized controlled trials’.

Studies were included if they satisfied the following criteria: i) Study design was a RCT; ii) study population consisted of human participants with paroxysmal, persistent or long-standing persistent AF; iii) interventions included pulmonary vein isolation, no matter which technique was used; and iv) follow-up was ≥12 months.

The methodological quality of each eligible study (risk of bias) was evaluated as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (version 5.1.0) (20). The risk of bias for each trial was assessed on the basis of the prime endpoint of recurrence of AF. The following criteria were evaluated and assigned a value of ‘high’, ‘low’ or ‘unclear’ by two authors. With regard to selection bias, the authors aimed to determine whether the method of randomization was adequate and whether the treatment allocation was concealed. In addition, performance and detection biases were assessed by determining whether the participants and personnel were blinded to the intervention, and whether the outcome assessor was blinded to the intervention. With regard to attrition bias, the authors determined whether any incomplete outcome data were sufficiently assessed and handled, while reporting bias was assessed by determining whether selective outcome reporting had been identified. Finally, the existence of any additional sources of bias was analyzed.

Two authors independently extracted relevant data from the included trials, and disagreements were resolved by discussion and consensus. The following data were extracted from each RCT: Name of the first author, year of publication, number of patients (intervention vs. control), age of populations, composition of gender, time of follow-up, definition of primary outcome and other important clinical information. In the case of a trial being reported in multiple publications, the most complete study or the article with the longest follow-up time was selected. The primary endpoint was the recurrence of AF. Secondary endpoints included all-cause mortality, stroke/transient ischemic attack (TIA) and quality of life (QoL).

Data analysis was performed based on an intention-to-treat analysis. Relative risks (RRs) with 95% confidence intervals (CIs) were selected to compare the differences for dichotomous outcomes, while weighted mean differences (WMDs) with 95% CIs were used to compare continuous outcomes. Heterogeneity among studies was analyzed using a χ²-based Q test and I² statistic. If significant heterogeneity was identified (P<0.05 and I² >50%), a random-effects model was selected, whilst a fixed-effects model was selected in all other cases.

Begg's funnel plot and Egger's test were used to evaluate the significance of the publication bias. The meta-analysis was performed using Stata software, version 11.0 (StataCorp, College Station, TX, USA) and Revman 5.1 software (The Cochrane Collaboration). A two-tailed P-value of <0.05 was considered to indicate a statistically significant difference. Finally, the quality of the evidence was evaluated using the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) system (21).

A flow diagram of the literature search is presented in Fig. 1. Briefly, 1,917 potentially relevant articles were identified following electronic and additional manual searches. Among the 1,917 publications yielded, 900 articles were excluded following screening of the titles and abstracts. Subsequently, 26 full-text articles were assessed for eligibility. Among these, 15 articles were further excluded for a variety of reasons, as described in Fig. 1. Finally, 11 RCTs involving 1,763 patients with AF were included in the meta-analysis (9–19).

Primary characteristics of the 11 RCTs included in the meta-analysis are shown in Table I. The 11 trials were published between 2003 and 2013. Among them, four trials enrolled only patients with paroxysmal AF (13,15,16,18), two trials enrolled patients with only persistent AF (11,19), and the remaining five trials enrolled patients with paroxysmal and persistent AF (9,10,12,14,17). Three trials enrolled patients to receive pulmonary vein ablation as the first-line therapy (10,17,18), whereas the remaining studies included patients that had failed at least one AAD treatment protocol or were intolerant of AADs. The majority of the trials compared catheter ablation with AADs, with the exception of one trial that compared catheter ablation plus AADs with single AAD administration (12). All the AF patients assigned to catheter ablation underwent a single ablation procedure in four of the studies (9,10,12,14). In the other trials, patients that received catheter ablation therapy underwent two or more ablation procedures in the blanking period when required.

A risk of bias assessment was performed for each trial, and the results are presented in Fig. 2A. Of the 11 trials, six trials adopted appropriate methods to generate the random sequence (10–12,14,15,18). One study reported allocation concealment using sealed envelopes (15); however, the methods of concealment in other trials were not mentioned. Blinding of the outcome assessors was reported in three trials (11,12,18). The overall risk of bias is presented in Fig. 2B.

The majority of the included trials considered the recurrence of AF and/or atrial tachyarrhythmia as their primary end point. Therefore, the overall effect of catheter ablation against AADs for the recurrence of AF was assessed. The results indicated that catheter ablation was able to significantly reduce the recurrence of AF, as compared with AADs (RR, 0.47; 95% CI, 0.38–0.58; P<0.001; Fig. 3). Significant heterogeneity was detected among the trials (Q=26.31; I²=62%; P=0.003).

Three trials enrolled drug-naive patients and the result was similar to the overall effect (RR, 0.50; 95% CI, 0.27–0.92; P=0.03). Sensitivity analysis indicated that no single study significantly altered the combined effect, which ranged between 0.45 (95% CI, 0.37–0.55) and 0.41 (95% CI, 0.43–0.62).

A total of 18 mortalities were reported in six studies, among which eight cases had received catheter ablation and 10 cases had received AADs. No statistically significant difference in the mortality rate was detected between the catheter ablation and AAD treatment groups (RR, 0.87; 95% CI, 0.37–2.06; P=0.76; Fig. 4), with no evidence for significant heterogeneity (Q=2.12, I²=0%, P=0.83).

A total of 17 strokes/TIA were reported, among which 10 events occurred in catheter ablation patients and seven events occurred in AAD patients. However, no statistically significant difference was detected between the catheter ablation and AAD therapies (RR, 1.83; 95% CI, 0.73–4.55; P=0.20; Fig. 5), and there was no evident heterogeneity (Q=0.39; I²=0%; P=0.98).

Four studies included results with regard to the differences in the QoL outcome, including the physical component summary (PCS) and the mental component summary (MCS). When compared with the baseline observations, the catheter ablation and AAD treatment groups exhibited a significantly improved QoL at the end of the study. However, catheter ablation was shown to result in improved QoL outcomes compared with AADs (PCS: WMD, 2.23; 95% CI, 0.24–4.21; P=0.03; MCS: WMD, 2.69; 95% CI, 0.04–5.35; P=0.05; Fig. 6).

The quality of the evidence was evaluated using the GRADE system. As shown in Table II, the evidence quality of the outcomes ranged between low and high. Potential publication bias was assessed using Begg's funnel plot and Egger's test, and the results indicated that there was no potential publication bias (Fig. 7; Egger's test, P=0.066).