ABC | Volume 112, Nº5, May 2019

Original Article Correia et al LA stiffness and AF recurrence - a meta-analysis Arq Bras Cardiol. 2019; 112(5):501-508 until March 2018. We searched for a combination of English terms and Medical Subject Headings (MeSH) descriptors, consisting of seven keywords [("left atrial" OR "left atrium") AND ("stiff" OR "stiffness" OR “compliance”) AND (“ablation” OR “pulmonary vein isolation”)]. A manual search of references was also used to identify possible studies for inclusion. If necessary, an English translation of the retrieved articles would be obtained. Each title and abstract were independently analysed by the two investigators, who selected the articles which would be relevant to the review. After that, the full texts of the remaining articles were reviewed to select which would be included in the qualitative or quantitative analysis. In case of disagreement, the decision was made by discussion and consensus of the authors. Inclusion criteria We included observational studies (with prospective or retrospective nature) in humans, whose objective was to study the association between LA stiffness and recurrence of AF after the first RFCA. For qualitative analysis, studieswith the following characteristics were included: 1) The study evaluated AF recurrence after the first RFCA in human subjects; 2) Retrospective or prospective observational studies; 3) The mean follow-up period was longer than 6 months; 4) The study included more than 20 subjects. For the quantitative analysis, we included studies that fulfilled all the previous criteria and reported hazard ratio (HR) and 95% confidence intervals (CI) of LA stiffness as predictors of AF recurrence. Quality assessment The risk of bias in the studies was evaluated using the National Heart, Lung and Blood Institute Quality Assessment Tool for Case Series Studies. 7 The evaluation was done independently by two raters (ETOC, LMSB), and in case of disagreement the decision was made by consensus of the raters. The following characteristics were assessed: 1) Was the study question or objective clearly stated?; 2) Was the study population clearly and fully described, including a case definition?; 3) Were the cases consecutive?; 4) Were the subjects comparable?; 5) Was the intervention clearly described?; 6) Were the outcome measures clearly defined, valid, reliable, and implemented consistently across all study participants?; 7) Was the length of follow-up adequate?; 8) Were the statistical methods well-described?; 9) Were the results well-described? After these characteristics were assessed, the authors gave the studies one of the quality ratings (good, fair or poor). Studies were rated as ‘poor’ if they met less than three criteria, ‘fair’ if they met three to five criteria, and ‘good’ if they met more than five criteria. All four articles selected met almost all the criteria and received a good quality rating by the two raters. The quality assessment of the four studies is reported in Table 1. Data extraction Data extraction was performed using a standard form by two investigators (ETOC, OMPS) and cross-verified by a third (ETM). Extracted data included: 1) First author’s last name, publication year; 2) Characteristics of included studies: number of patients, region of the study, study design, ablation strategy, method of LA stiffness measurement, method of AF detection, length of follow-up period, length of blanking period and main findings; 4) Outcome results: HR and 95%CI of LA stiffness as a predictor of AF recurrence in multivariate analysis. Statistical analysis The association between AF recurrence and LA stiffness following RFCA was measured by HR with 95% CI. Adjusted HRs were used, since all the studies included in the quantitative analysis employed multivariate analysis by Cox proportional hazard model to adjust for potential confounders. Log of the HR was obtained by calculating their natural logarithms. Then, standard errors were determined from the logarithmic scale and corresponding 95% CIs. The inverse variance method was used to weigh studies for the combined overall statistics. Statistical significance was defined at p‑values < 0.05. Heterogeneity between studies was assessed using the Cochran’s Q test and I² statistics and then evaluated by I² values. I² values less than 30% were defined as low heterogeneity; less than 60% were considered moderate heterogeneity; and above 60% defined as high heterogeneity. 8 The random-effects model was chosen because of the different methods of LA stiffness measurements, what could lead to heterogeneity. Sensitivity analysis was done by leaving out studies and checking the consistency of the overall effect estimate. A meta-regression was not done because of the small number of studies included. The results are presented in a forest plot with 95% CI. Publication bias was verified by a funnel plot, although only 3 studies were included, which made the interpretation more difficult. All analyses were done using Review Manager 5.3 software. Results Study selection Initially, a total of 62 studies were identified in the databases, 57 in PubMed and 5 in the Cochrane Central Register of Controlled Trials. In the duplicate analysis, we identified 2 duplicates, which were then excluded. After a careful reading of the titles and abstracts, 57 of 62 studies were excluded because they were not related to the present review. The full texts of the five studies were analysed, and 4 of them included in the qualitative analysis. The study excluded, by Marino et al. 9 analysed only 20 patients and the mean follow‑up period was shorter than 6months. For the quantitative analysis, one full‑text article was excluded because it did not report HR and 95% CI of LA stiffness as predictors of AF recurrence. 10 Finally, four studies were included in the qualitative analysis and three in the quantitative analysis. The flow diagram of the study selection is depicted in Figure 1. Characteristics of the included studies Four studies were included in this review, 10-13 all of them prospective single centre case series studies (Table 1). The study of Machino-Ohtsuka et al. 11 included 155 patients, and in the 502