The relationship between vitamin D and risk of atrial fibrillation: a dose-response analysis of observational studies

Background The relationship between serum vitamin D and atrial fibrillation (AF) or postoperative atrial fibrillation (POAF) in patients undergoing coronary artery bypass graft (CABG) is still debated. It is also unclear whether there is a dose-response relationship between circulating vitamin D and the risk of AF or POAF. Methods The Cochrane Library, PubMed, and Embase databases were searched for relevant studies. We used a “one-stage approach” with a restricted cubic spline model to summarize the dose-specific relationships between serum vitamin D and AF. Relative risk (RR) was used to measure the effects in this meta-analysis. Results In total, 13 studies were included with a total of 6519 cases of AF among 74,885 participants. Vitamin D deficiency (< 20 ng/ml) was associated with increased risks of AF (RR: 1.23, 95% CI: 1.05–1.43). In the dose-response analysis, the summary RR for a 10 ng/ml increased in vitamin D was 0.88 (95% CI: 0.78–0.98) and there was no evidence of a non-linear association, Pnon-linearity = 0.86. In the age subgroup, high vitamin D (per 10 ng/ml increase) reduced the risk of AF in the older group (> 65 years) (RR = 0.68, 95% CI = 0.52–0.89) but not among young individuals (< 65 years) (RR = 0.87, 95% CI = 0.72–1.06). In addition, a strong association was found between a 10 ng/ml increased in vitamin D and POAF incident in the patient after CABG (RR: 0.44, 95% CI: 0.24–0.82). Conclusion Our dose-response meta-analysis suggested serum vitamin D deficiency was associated with an increased risk of AF in the general population and POAF in patients after CABG. Further studies are needed to explore the age difference in the association between serum vitamin D level and the risk of AF and whether vitamin D supplements will prevent AF. Trial registration This study has been registered with PROSPERO (International prospective register of systematic reviews)-registration number-CRD42019119258.


Introduction
Historically, vitamin D is known for its important role in skeletal disease, [1][2][3]. The focus in recent decades has been on the risks of vitamin D and non-skeletal diseases, such as cardiovascular disease or atrial fibrillation (AF) [4]. AF is the most common cardiac arrhythmia in clinical practice and is associated with increased risk of morbidity. Vitamin D deficiency is common in many countries [5], only 23% of people would reach serum vitamin D concentration above 30 ng/ml [6]. A causeand-effect relation between low vitamin D status and AF incident would be of considerable benefit to public health. However, unlike for skeletal disease, the evidence for serum vitamin D deficiency (< 20 ng/ml) and the risk of AF has been inconclusive [7,8]. To date, results from several observational studies have suggested that patients with vitamin D deficiency were approximately twice as likely to have AF than patients with normal levels (> 30 ng/ml) [8][9][10][11]. Conversely, several prospective studies did not find this association [7,[12][13][14]. Thereafter, several articles have reviewed published studies and yielded conflicting results [7,15,16]. Alonso et al. did not find a clinically relevant association of circulating vitamin D per 1 standard deviation (8.5 ng/ml) decreased with AF risk [7]. In contrast, a meta-analysis concluded a weak but positive association between vitamin D deficiency and AF [15]. However, there are several limitations in the previous metaanalyses. For example, vitamin D levels were analyzed as either a categorical or continuous variable in the individual studies, so they could not pool all of the studies together. In addition, the shape of the doseresponse association between vitamin D and AF had been explored. Moreover, several new research articles reported higher serum 25(OH) D is associated with new-onset AF after coronary artery bypass grafting (CABG) surgery [17][18][19][20]. Therefore, we performed a comprehensive meta-analysis to evaluate the shape of the dose-response relation between circulating 25(OH) D concentration and the risk of AF and post-operation AF (POAF) after CABG.

Methods
This work has been performed according to PRISMA guidelines (http://www.prisma-statement.org; Additional file 1: Table S1) [21]. We systematically searched the PubMed, Embase databases and Cochrane Library up to March 10, 2019. Additional file 1: Table S2 provides a detailed description of the search strategy. Two researchers independently worked in the whole process of this meta-analysis from the literature search and selection to data analysis. Both randomized controlled trials and observational studies, reporting data about serum vitamin D level and AF were considered eligible for this metaanalysis. All discrepancies were resolved through discussion by the two authors. We used the robust error meta-regression method (REMR) for the doseresponse analysis of the vitamin D level and AF [22,23]. All statistical analyses were done by using Review Manager (RevMan) version 5.3 (The Cochrane Collaboration 2014; Nordic Cochrane Center Copenhagen, Denmark) and Stata software (Version 14.0, Stata Corp LP, College Station, Texas, USA). We used the Newcastle-Ottawa quality assessment scale (NOS) to evaluate the quality for all included studies [24], a NOS score of ≥6 stars was regarded as high-quality, otherwise, as low-quality studies [25,26]. Full details of the literature search strategy, study selection criteria, quality assessment, and statistical analysis have been reported in the Supplement Methods (Additional file 1). This study has been registered with PROS-PERO (International prospective register of systematic reviews)-registration number-CRD42019119258.

Study selection
We identified 1484 studies in our initial database search. After removing duplicates and studies with inadequate information on vitamin D and AF, 20 studies were reviewed in more detail. Of these 20 studies, 5 were excluded for the following reasons: a) they were focused on recurrence AF (n = 1) [27]; b) they were reviews or case reports (n = 2) [15,28]; or c) cross-section study (n = 1, 20]. Finally, 13 studies (14 reports) were included in this meta-analysis (Fig. 1).
Then, we conducted a non-linear dose-response by using restricted cubic model and found an inverse relationship between vitamin D and AF (Fig. 4).
In subgroup and sensitivity analysis, the positive association between vitamin D and risk AF persisted in almost all subgroup analyses defined by the duration of follow-up, geographic location, number of cases, study quality and adjustment for confounding and potential intermediate factors (Table 3).

Publication bias
There was some indication of publication bias with Egger's test, p = 0.07, or with Begg's test, p = 0.03 or by inspection of the funnel plot (Additional file 1: Figure S1-S3). Thus, we used the "trim-and-fill" method for the adjustment of publication bias.
However, the results showed "no trimming performed and data unchanged", which demonstrated that our results were stable. The publication bias of POAF was not conducted as limited studies (N < 10) according to the guideline [30].

Discussion
To the best of our knowledge, this is the first metaanalysis to evaluate the dose-response association between vitamin D and AF. Our results by dose-response  analysis suggested that vitamin D deficiency is a moderate predictor of AF. We found vitamin D deficiency (< 20 ng/ml) or vitamin inadequate (< 30 ng/ml) increased the risk of AF by 23% or 14%, respectively. Interestingly, the present dose-response analysis first showed vitamin D deficiency is also a moderate predictor for POAF, and novelty showed a 12% (AF) or 56% (AF post-CABG) increase in the RR per 10-unit increase in vitamin D with evidence of linear association. Formerly, the role of circulating vitamin D in AF remains unclear. A meta-analysis found a vitamin D deficiency is associated with a 31% increased risk for AF [15]. However, another study which only pooled cohort studies did not support this association [7]. Our results showed there is an inverse association between vitamin D and risk of AF, both in cohort and case-control studies. However, significant heterogeneity between the subgroup of study design (cohort and case-control study) was observed. The association in cohorts was significantly weaker than that in case-control studies. Besides the recall bias, the heterogeneity in the study population appeared to be one of the potential reasons. The cohort studies were general population-based and the casecontrol studies mainly focused on patients with cardiovascular disease (CVDs). Recent studies also have shown that vitamin D was a potential connection with CVDs (eg. ischemic heart disease) and diabetes [15,31]. Therefore, it might be reasonable to speculate that the association between vitamin D deficiency and AF may be amplified in patients with CVDs or at high risk of developing CVDs. However, considering the limited sample size, the role of vitamin D in patients with CVDs or at high CVDs risk need to be further studied.
The association between vitamin D deficiency and AF has several potential pathophysiological mechanisms. Inflammation has a crucial role in the pathogenesis of AF [32]. For example, C-reactive protein (CRP), the most robust and reproducible marker of vascular inflammation, could increase the risk of AF by up to two-fold [33]. Noticeable, low vitamin D status could directly or indirectly increase the synthesis of CRP [34]. Another important mechanism might be the activation of the renin-angiotensinaldosterone system (RAAS). RASS plays an  important role in both structural and electrical remodeling of the atrium. Studies in experimental animals showed vitamin D could inhibit the RAAS system [35]. Clinical researches showed that the use of ACEIs was associated with less atrial fibrosis, and the blockade of angiotensin II has been shown to have beneficial effects on electrical remodeling in human atrial tissue [36,37]. Moreover, a previous meta-analysis also reported that inhibition of RASS might reduce the risk of developing new-onset atrial fibrillation [38]. Therefore, low vitamin D level might increase the AF risk secondary to its negative regulatory property of the RAAS. It is not surprised that we found vitamin D deficiency is associated with increased POAF. Previous studies have suggested that deficiency of total  vitamin D is associated with increased prevalence of electrocardiographic abnormalities (e.g. prolonging the duration of action potentials) [39]. Moreover, in a recent prospective cohort study of patients undergoing cardiac surgery, low total 25(OH) D levels were independently associated with the risk of major cardiac and cerebrovascular events [39]. Another study also showed an inverse relationship between serum 25(OH) D level and left atrial or AF recurrence in patients after undergoing catheter ablation [40]. However, considering the small sample size and short-term follow-up, the relationship between vitamin D and POAF need to be further confirmed in larger, well-designed studies. We also studied the role of age in the present meta-analysis, we found low 25(OH) D level increased the risk of AF in the older individuals (age ≥ 65 years) but not young group (< 65 years). This result should be with caution in elder individuals. However, there was significant heterogeneity in the results, which might come from the study population or study design. When we excluded the casecontrol studies, both the older (RR: 0.96, 95%CI: 0.93-1.00) and young people (RR:0.94, 95%CI: 0.89-1.00) showed a weak association between vitamin D deficiency and risk of AF. Of note, these results were incosistent with the recent analysis of ARIC study [7], which showed that low vitamin D was a stronger indicator of AF in the youngest group (< 54 years) but not in the oldest (> 60 years), with an intermediate association in those aged 54-59 years. Therefore, based on current evidence, the age difference in the relationship between vitamin D and risk of AF is still unclear. Further prospective cohort studies are needed to clarify the age difference.

Study limitations
The present meta-analysis has several limitations. First, this was a meta-analysis of observational studies, which cannot chiefly prove causation, and the unmeasured and insufficiently measured variables (e.g. seasonal variation in vitamin D) would result in the possibility of residual confoundings. However, most of our studies were performed during the winter or spring months, which could reduce this confounding factor. Second, due to data restriction, the impact of vitamin D supplements on AF was not analyzed and need to be further investigated as we previously discussed. Third, some studies suggested that available vitamin D may be a more reliable marker of vitamin D status than total 25 (OH) D. However, none of the included studies measured available vitamin D.

Conclusion
Our dose-response suggested serum vitamin D deficiency was associated with an increased risk of AF in the general population and POAF in patients after CABG. Further studies are needed to explore if