4.3 Cause of Heterogeneity
In this study, obvious heterogeneity was reported in two meta-analyses of comparing EAT volumes and EAT thickness in two groups with and without AF occurrence. Given the obtained heterogeneity, the random effect model was used. However, f the classification of studies into three categories of LA-EAT volume, total EAT volume, and EAT thickness reduced heterogeneity. Because of small sample sizes, a subgroup analysis or meta regression could not find the possible sources among various factors, such as female gender, cardiac surgery types, EAT evaluation methods as well as quality of studies based on STROBE Checklist. Moreover, BMI and age, as well as other classical AF risk factors were additional possible sources of heterogeneity. According to the sensitivity analysis (Table 2-a), it is suggested that the study of Ertugrul et al. 21 be one of the sources of heterogeneity. The endpoint of other studies was AF occurrence, which was defined according to the guidelines for the management of atrial fibrillation of European Society of Cardiology. But in this study, AFACS was defined as arrhythmic episodes that deteriorate hemodynamics, lasts more than 20 minutes per episode, or lasts more than 60 minutes during a 24-hour follow up, which underestimates the occurrence of AFACS. The study by Ertugrul et al. omitted, the sensitivity analysis results showed there was no heterogeneity resource (Table 2-b).
4.4 Study Limitations
The studies observational nature limits the study that by itself had no high position in the Evidence-Based Medicine Pyramid. And lacking access to patient detailed data, it didn’t allow us to adjust for other covariates that may influence EAT including age, gender, and AF risk factors of hypertension, obstructive sleep apnea hypopnea syndrome, BMI and so on. Included studies were single-center trials with a small sample size, which might result in unreliable data. The inclusion of only 2 studies evaluating EAT thickness was limited because of the potential lack of power and the inability to draw firm conclusions. Finally, there was a significant degree of heterogeneity in the study, which might be a limitation that has been demonstrated in other published EAT meta-analyses with I2 values >90% 51-53.