Association of Plasma Transforming Growth Factor-β1 Levels and the Risk of Atrial Fibrillation: A Meta-Analysis

Introduction Numerous studies have demonstrated that plasma transforming growth factor-β1 (TGF-β1) may be involved in the pathogenesis of atrial fibrillation (AF), but some discrepancy remained. We performed a meta-analysis to evaluate the association between the plasma level of TGF-β1 and the risk of AF. Methods Published clinical studies evaluating the association between the plasma level of TGF-β1 and the risk of AF were retrieved from PubMed and EMBASE databases. Two reviewers independently evaluated the quality of the included studies and extracted study data. Subgroup analysis and sensitivity analysis were performed to evaluate for heterogeneity between studies. Results Of the 395 studies identified initially, 13 studies were included into our analysis, with a total of 3354 patients. Higher plasma level of TGF-β1 was associated with increased risk of AF when evaluated as both a continuous variable (SMD 0.67; 95%CI 0.29–1.05) and a categorical variable (OR 1.01, 95% CI 1.01–1.02). Conclusions This meta-analysis suggests an association between elevated plasma TGF-β1 and new onset AF. Additional studies with larger sample sizes are needed to further investigate the relationship between plasma TGF-β1 and the occurrence of AF.


Introduction
Atrial fibrillation (AF) is the most common sustained arrhythmia with debilitating consequences such as stroke and heart failure. It is also associated with an increase in overall mortality [1,2,3]. Animal models and studies on patients with AF have confirmed that the development of AF is associated with both structural and electrical remodeling of the atria [4]. Patients with chronic AF have significant myocardial interstitial fibrosis which contributes to the occurrence and perpetuation of AF [5,6]. Transforming growth factor-β1 (TGF-β1) is an important factor in fibrosis [7]. It is involved in the process of cell proliferation, apoptosis and migration. It promotes the differentiation of cardiac fibroblasts and production of extracellular matrix such as collagen, fibronectin, and protein polysaccharide which leads to cardiac fibrosis [8]. In transgenic mouse models, the activation of TGF-β1 promotes atrial fibrosis and the development of AF [9]. On the other hand, the inhibition of TGF-β1 by pirfenidone (PFD) can significantly reduce the extent of atrial fibrosis [10]. These findings have prompted clinical studies on the relationship between plasma TGF-β1 levels and the development of AF in humans. However, the results generated have been inconsistent. Therefore, we conducted a comprehensive metaanalysis to evaluate the available evidence of whether high plasma TGF-β1 levels are related to the risk of having AF.

Search strategy
Articles were identified by searching PubMed and Embase online databases for articles published up until November 2015. The key terms used are 'TGF-β1', 'transforming growth factor-β1', 'transforming growth factor-beta1', 'transforming growth factor' and 'atrial fibrillation'. We manually searched the bibliographies of original papers and abstracts of the scientific sessions of the past 3 years. In addition, we sought the assistance from potential experts in the field to assess the quality of included articles. We evaluated the titles, abstracts and reference lists of all articles to identify potentially relevant studies.

Trial selection and inclusion criteria
Two reviewers (J. L. and Y. Y.) evaluated the titles and abstracts of all eligible studies. The full text of relevant studies was retrieved and assessed accordingly based on the inclusion criteria. Any disagreements on whether to include any study between the two investigators were resolved through joint review and discussions.
For inclusion, eligible trials should meet the following criteria: (1) the study design was case-control, prospective or retrospective cohort studies; (2) human subjects; (3) included the characteristics of study patients; (4) clearly defined endpoint events, such as AF occurrence or recurrence; (5) evaluated the plasma TGF-β1 levels of AF patients and non-AF patients; (6) reported the plasma level of TGF-β1 using [mean ± standard deviation (SD)] and odds ratio (OR) or hazard ratio (HR) of AF incidence and the corresponding 95% confidence interval (CI) for TGF-β1 levels.

Data extraction
Two independent reviewers (J. L. and Y. Y.) extracted data from included studies using a standard data extraction form. Information on authors and published journals were removed and then independently evaluated according to the described inclusion criteria. Relevant data were extracted from the manuscripts. We extracted and analyzed the plasma concentration of TGF-β1 expressed as mean ± SD from each primary study. Adjusted OR values were selected for the analysis. Additional data collected included study characteristics (first author's last name, publication year, study design, sample size, AF definition, follow-up duration, end-point events)  and patients baseline characteristics (age, sex, BMI, smoking, mean left atria diameter, left ventricular ejection fraction, the presence of CAD, hypertension, diabetes and medication).

Quality assessment
The two investigators (J. L. and Y. Y.) independently evaluated the quality of the eligible studies based on the guidelines by the Evidence-Based Medicine Working Group [11] and the United States Preventive Task Force [12]. Each study was judged in accordance to the 10-item STROBE checklist. We appraised the quality of studies according to the following characteristics: (1) the inclusion and exclusion criteria are clearly defined; (2) sample selection is clearly described; (3) involved population is representative of study sample; (4) the patients' follow-up period is adequate; (5) reports loss of follow-up; (6) clinical and demographic variables are complete; (7) the definition of AF is clearly defined; (8) the outcomes and outcome assessment are clearly defined; (9) temporality (evaluation of plasma TGF-β1 levels at baseline) and (10) adjustment of possible confounders on the multivariate analysis, especially for categorical variable. If any of the characteristics was not described, we assumed that it had not been performed.

Statistical analysis
All continuous variables were presented as (mean ± SD). The standard mean difference (SMD) was used to analyze the results in our meta-analysis. SMD method was used as different unit of measurements were presented for TGF-β1 levels. As the studies included in this meta-analysis may have used either a continuous or categorical variable for TGF-β1 levels, we performed a separate meta-analysis for both types of variables to evaluate the association between TGF-β1 levels and the occurrence of AF. The HR values in multivariate Cox proportional hazards model in each primary study were directly considered as OR values. I 2 derived from the chisquare test was used to evaluate the heterogeneity across the studies included. I 2 of 50% indicates that there was no significant heterogeneity [13]. A fixed effects model was used if no significant heterogeneity was found. When pooled effect resulted in significant heterogeneity, the random effects model was used. We conducted random effects meta-analysis using the inverse variance heterogeneity method. In addition, we also performed subgroup analysis based on the patients' age (50y or >50y), study design (cohort study or case control study), duration of follow-up (<12 months or 12 months), sample size (<100 or 100) and left ventricular ejection fraction (LVEF) (<50% or 50%). Sensitivity analysis was performed by sequentially removing each individual study. We assessed for publication bias by constructing a funnel plot. Twotailed p value of <0.05 was considered statistically significant. All statistical analyses were performed with Review Manager Version 5.3.

Search results
Data retrieval and study selection was shown in the flow chart (Fig 1). A total of 395 studies were found using our search criteria. After reviewing title and abstract of each study, we excluded 356 articles because they were either unrelated, review articles or basic science research papers. Then, we evaluated the remaining 39 studies in detail. Of these 39 studies, we excluded 26 studies because: 1 had duplicate data, 17 did not provide the plasma TGF-β1 levels, 6 did not provide (mean ± SD) data of TGF-β1 or OR/HR values, 1 did not provide baseline characteristics of patients and 1 had no control group. Finally, the remaining 13 studies were included into our meta-analysis.

Sensitivity and subgroup analysis
A subgroup analysis was performed based on the type of AF on 5 studies and it showed a positive correlation between high TGF-β1 plasma levels and the risk of new-onset AF (SMD 1.07; 95%CI 0.26-1.89) with significant heterogeneity (I² = 95%, P<0.00001) across studies. However, there was no clear relationship between plasma TGF-β1 levels and the risk of recurrent AF (SMD 0.38; 95%CI (-0.05-0.81) with significant heterogeneity (I² = 83%, P<0.00001) across studies (Fig 2).   Transforming Growth Factor-β1 and Atrial Fibrillation A predefined subgroup analysis was performed to investigate the origin of the heterogeneity between studies. In the subgroups with follow-up <12 months, LVEF <50% and sample size 100, there were no significant heterogeneity between studies. Therefore, the follow-up duration, LVEF and sample size are likely the origin of the significant heterogeneity in our metaanalysis (Table 4).
Finally, we performed a sensitivity analysis and found that there was no significant difference on the overall heterogeneity regardless of which study was removed. The result of the funnel plot for TGF-β1 in AF patients was asymmetrical, indicating the potential for publication bias (Fig 5). After removing the study with the highest levels of TGF-β1, the result of the funnel plot was symmetrical (Fig 6).  Transforming Growth Factor-β1 and Atrial Fibrillation

Discussion
The main finding of this comprehensive meta-analysis is that there is an association between high plasma TGF-β1 levels and risk of AF, especially for new-onset AF. To the best of our knowledge, this is the first comprehensive meta-analysis performed to investigate their relationship.
TGF-β1 is a major factor promoting collagen production in cardiac fibroblasts [27]. It is also considered to be a key factor in the signal cascade reaction during the process of tissue fibrosis [28,29]. Multiple studies have found increased atrial fibrosis on biopsy or autopsy specimens of patients with AF and concurrently elevated plasma level of TGF-β1 [30,31,32]. Previous studies have only investigated the plasma level of TGF-β1 in specific subgroups, such as the patients who developed recurrent AF (following surgical maze procedure, electrical cardioversion and catheter ablation) or patients with new-onset AF after cardiac surgery. In our meta-analysis, different groups were analyzed together to identify the potential role of TGF-β1 in promoting AF. We demonstrated that there is a positive correlation between higher plasma TGF-β1 levels and the development of new onset AF and the overall occurrence of AF. However, it is worth noting that there was no clear relationship between plasma TGF-β1 levels and recurrent AF in the subgroup analysis. This finding could be due to the heterogeneity in study population and AF management strategy between studies.
4 studies included information on the type of AF, whether they were persistent or paroxysmal AF. Persistent AF was defined as AF lasting for more than 7 days while paroxysmal AF  Table 4. Subgroup analyses of the association between the TGF-β plasma levels and incidence of AF. was defined as AF with spontaneous termination in less than 7 days. To investigate the relationship between TGF-β1 and the type of AF, we analyzed these 4 [14,[16][17][18]] studies which included the plasma level of TGF-β1 in both persistent AF and paroxysmal AF patients. TGF-β1 levels were found to be higher in patients with persistent AF compared to paroxysmal AF. The finding was expected as atrial fibrosis is more extensive with longer duration of AF. It also leads us to the hypothesis that TGF-β1 as an index of atrial fibrosis may inform us of the chronicity of AF. Despite being the most common sustained arrhythmia, the mechanism of AF is poorly understood. Systemic inflammatory response appear to be the contributing factor to the occurrence and recurrence of AF. A meta-analysis performed by Wu et al [33] demonstrated that high levels of circulating inflammatory factors especially CRP and IL-6 are associated with greater risk of AF in the general population, occurrence of AF after coronary artery bypass grafting and AF recurrence after electrical cardioversion or catheter ablation. In our analysis, we found that elevated levels of plasma TGF-β1 was associated with the occurrence of new onset AF. Our findings provide important insight into the mechanisms of AF.

Study Limitations
The present meta-analysis has several limitations. First, AF duration and methods of AF detection were different among the studies which account for the heterogeneity between the individual studies. Second, the sample size of the meta-analysis was relatively small. Finally, the asymmetrical funnel plot suggests that there may be publication bias.

Conclusions
In conclusion, this meta-analysis suggests an association between high plasma TGF-β1 and the occurrence of new onset AF. Additional studies with larger sample sizes are needed to further investigate the relationship between plasma TGF-β1 and the occurrence of AF.