This was a prospective observational cohort study (clinical trial registration number: ChiCTR-OCH-13003479). Between January 1, 2014, and December 31, 2015, 200 consecutive elderly patients (all aged over 75 years) with non-valvular AF at the General Hospital of the Chinese People's Liberation Army were enrolled (average age: 87.6 ± 7.7 years). The study was approved by the Medical Ethics Committee of the General Hospital of the People's Liberation Army (approval number: S2013-064-02).

The inclusion criteria for patients in the AF group (including new onset, paroxysmal, persistent, and permanent AF, as confirmed by electrocardiogram or 24 h Holter monitor) were age ≥75 years and at least one AF attack recorded in the year prior to the study. Patients or patients' relatives provided written consent. The exclusion criteria were age ≤75 years, rheumatic heart disease, biological or mechanical valve replacement or mitral valve repair, serious uncontrolled infection, and an AF attack recorded more than 1 year before the study.

Clinical data were collected at the baseline and during follow-up by experienced attending cardiologists. The baseline clinical data of enrolled patients were as follows: name, gender, age, type of AF (new onset, paroxysmal, persistent, or permanent), systolic blood pressure, diastolic blood pressure, smoking and drinking history, disease history (such as ischemic stroke, hypertension, diabetes, congestive heart failure, peripheral vascular disease, and renal/liver dysfunction), and medication status [including the use of antiplatelet drugs (aspirin and clopidogrel) and anticoagulation drugs (warfarin and novel oral anti-coagulants)]. Follow-up data including medication status and clinical endpoints were collected through medical records, telephone interviews, and questionnaires.

Two hundred elderly patients with AF were followed up for 2 years to observe thromboembolic (both arterial and venous thrombosis) and bleeding events. AF-related thrombotic events included ischemic stroke, lower extremity deep venous thrombosis, and thrombosis of other systems (including mesenteric thrombosis and venous catheter-related thrombosis). Arterial thrombosis was defined as acute myocardial infarction and unstable angina. Clinically relevant major bleeding events included a decrease in hemoglobin of more than 2 g/L, the need for a blood transfusion or hospitalization, bleeding in major organs, and fatal bleeding. Non-clinically relevant major bleeding events included chronic bleeding and a decrease in hemoglobin of <2 g/L. Minor bleeding events included subcutaneous hematoma, gastrointestinal bleeding (below the standard for major bleeding), and blood sputum (20). Data collection was completed for 188 patients; 12 patients were lost to follow-up, the loss to follow-up rate is 6%.

Two milliliters of fasting blood was drawn from all patients in the early morning on the day of enrollment. Blood samples were collected using EDTA anticoagulant vessels and then centrifuged at 4,000 rpm for 8 min (room temperature). Plasma was separated and stored at −80°C for batch analysis. The Aimplex method (Beijing Kuang Bo Biotechnology Limited by Share Ltd., Beijing, China) was used to detect 27 chemokines of the chemokine family, including neutrophil activating protein-2 (NAP2), P-selectin, E-selectin, intercellular cell adhesion molecule-1 (ICAM-1), vascular cell adhesion molecule-1 (VCAM-1), platelet factor 4 (PF4), monokine induced by interferon-r (MIG), B-lymphocyte chemoattractant 1 (BCA-1), regulated on activation normal T cell expressed and secreted (RANTES), monocyte chemoattractant protein (MIP)-3beta, MIP-1beta, tumor necrosis factor (TNF)-alpha, interleukin (IL)-1beta, IL-2, IL-4, IL-6, granulocyte chemotactic protein 2 (GCP2), IL-10, epithelial neutrophil activating peptide 78 (ENA-78), interferon-inducible protein-10 (IP-10), monocyte chemoattractant protein-1 (MCP-1), thymus and activation-regulated chemokine (TARC), chemokine C-C-motif ligand (CCL19), stromal cell-derived factor 1 [(SDF1; also known as C-X-C motif chemokine 12 (CXCL12)], fractalkine, and interferon-inducible T cell alpha chemoattractant (I-TAC). For each inflammatory chemokine, the intragroup error was <6%, while the intergroup error was <8%.

Continuous variables were expressed as the median [interquartile range (IQR)] if the data were non-normally distributed, or as the mean ± standard deviation (SD; ± SD) if the data were normally distributed (Kolmogorov-Smirnov criteria). The Kruskal-Wallis test was used to compare non-parametric data. Two-way ANOVA was used to compare data with a normal distribution. Qualitative data were expressed as percentages. A p < 0.05 was considered statistically significant. Statistical analysis was performed using IBM SPSS Statistics version 19.0 (SPSS, Inc., Chicago, IL, USA).

The clinical characteristics of 188 patients with AF are presented in Table 1. This cohort was predominantly elderly (aged 75–98 years, average age: 87.6 ± 7.7 years) male patients (92.6%). This cohort of elderly patients with AF was at high risk of thromboembolism [CHA₂DS₂-VASc (mean ± SD): 5.1 ± 1.6] and bleeding [HAS-BLED (mean ± SD): 3.4 + 1.2].

Among 188 elderly patients with AF, the number of patients on warfarin, rivaroxaban, clopidegrel, and aspirin were 10 (5.32%), 25 (13.30%), 78 (41.49%), and 74 (39.36%) respectively.

During the 2-year follow-up, among the 188 patients, there were 32 cases (17.0%) of AF-related thrombosis (including ischemic stroke, deep-vein thrombosis, and thrombosis of other systems), 36 cases (19.1%) of arterial thrombosis,35 cases (18.6%) of major bleeding events (including hemorrhagic stroke and other clinical major bleeding events), and 36 cases (19.2%) of minor bleeding. There were 54 cases (28.7%) of all-cause death (Table 2).

Thirty high-risk patients with AF but not exhibiting clinical adverse events (control group), 23 with arterial thrombosis (unstable angina and acute myocardial infarction), 15 with AF-related venous thrombosis, and 12 with major bleeding were selected and matched randomly to analyze the 27 chemokines. The baseline levels of IL-6, IL-10, VCAM-1, CCL19, BCA-1, IL-4, E-selectin, fractalkine, CXCL12, and GCP2 were found to differ statistically among the control, arterial thrombosis, AF-related venous thrombosis, and major bleeding groups (p < 0.05). Compared with those in the control group, the levels of IL-4 in patients with AF-related venous thrombosis, arterial thrombosis, or major bleeding events increased by 53- (0.53 vs. 0.01 pg/ml), 17- (0.17 vs. 0.01 pg/ml) and 19-fold (0.19 vs. 0.01 pg/ml), respectively. Compared with those in the control group, the levels of IL-6 in patients with arterial thrombosis increased by six-fold (39.78 vs. 4.98 pg/ml; Table 3).

Accumulating data indicate that endothelial dysfunction, platelet activation, and coagulation factor activation caused by inflammation might cause patients with AF to enter a prothrombotic state (21–23). The upregulation of inflammatory factors such as IL-6 and IL-8 are related to endothelial dysfunction, platelet activation, and active thrombosis (21, 22). Experimental studies have demonstrated that the IL-6-induced expression of tissue factor and von Willebrand factor is involved in platelet activation, possibly driven by platelet-monocyte interactions (24, 25). Patients with increased levels of IL-6 are more likely to have thrombus formation in the left atrial appendage (21, 22). Moreover, Pinto et al. (26) found that elevated levels of IL-6, von Willebrand factor, and TNF-alpha are significant predictors of stroke, based on a 3-year follow-up of 373 patients with AF. IL-6 is also an independent predictor of the composite endpoint of stroke or death in patients with AF (27) and is associated with worse outcomes after ischemic stroke (28).

The patients with AF in our study were at high risk of both thromboembolism and bleeding and were much older than the patients enrolled in these previous studies. We found that the levels of IL-6 increased significantly in patients with arterial thrombosis and AF-related thrombosis during the 2-year follow-up; our results were consistent with the previous findings. Interestingly, we also found that the expression of IL-6 was significantly increased in patients with AF with major bleeding. This result was consistent with the findings of the RE-LY (Randomized Evaluation of Long-Term Anticoagulation Therapy Study) trial, in which an independent relationship was identified between IL-6 and a higher risk of stroke and major bleeding independent of clinical risk factors, even after adjusting for cardiovascular biomarkers to attenuate the prognostic value (29).

Under normal physiological conditions, soluble E-selectin (sE-sel) is not expressed on the surface of endothelial cells (30). Under pathological conditions, increased expression of E-selectin on the surface of endothelial cells can mediate the activation of endothelial cells and the adhesion of leukocytes, which may play a role in the process of thrombosis (31). However, data regarding whether sE-sel contributes to the risk of thromboembolism in patients with AF remains limited and confusing. For instance, Freestone et al. (32) found that compared to patients with sinus rhythm, patients with AF have elevated levels of sE-sel, and low baseline sE-sel levels predict the successful maintenance of sinus rhythm at 6 months of follow-up (33). In contrast, another study conducted by Freestone et al. (34) indicated that in patients with systolic heart failure, the differences between plasma sE-sel levels in patients with AF and those in patients with normal sinus rhythm were not significant.

Similar to the first study by Freestone et al., a study by Krishnamoorthy et al. (35) of 423 patients with non-valvular AF followed up for 19 months showed that elevated levels of sE-sel were associated with an increased risk of clinical adverse events (acute myocardial infarction, ischemic stroke, and all-cause mortality). However, Roldán et al. (36) found no association between sE-sel levels and inflammation or abnormal thrombogenesis in 191 consecutive patients with chronic non-rheumatic AF who were not receiving anticoagulation therapy. However, plasma levels of sE-sel were significantly decreased (p < 0.01) after oral anticoagulation therapy.

In the Bruneck study (37), a prospective population-based cohort study with a 20-year follow-up (n = 909), the levels of soluble VCAM-1 were significantly associated with incident AF (hazard ratio: 1.49; 95% confidence interval, CI: 1.26–1.78, p < 0.001 with a Bonferroni correction for both values); even after adjusting for age and sex, the association was still significant. However, the association with E-selectin levels was insignificant (after a Bonferroni correction in unadjusted and age- and sex-adjusted analyses).

Our results showed that the levels of E-selectin were significantly increased in the major bleeding and AF-related thrombosis groups, compared with those in patients without thromboembolic and bleeding events. Furthermore, the levels of VCAM-1 were different significantly among control group vs. arterial thrombosis, venous thrombosis, and major bleeding groups (p < 0.05). Further in vivo and in vitro studies are needed to investigate whether increased E-selectin levels, which reflect endothelial activation and platelet aggregation, initiate AF and AF-related thrombosis or are simply a consequence of AF-accompanied cardiovascular disease and adverse clinical events.

IL-4 and IL-10, secreted by type II T helper lymphocytes (Th2 lymphocytes), can inhibit the synthesis and expression of proinflammatory markers such as IL-6, IL-8, and TNF by inhibiting type I T helper lymphocytes and stimulating B and Th2 lymphocytes, thus reducing the inflammatory response (38). The IL-10-592A/C polymorphism is associated with AF, and people of Han Chinese descent with the A allele are at an increased risk of AF (39). Henningsen et al. (30) did not find significant differences in either the genotypic or allelic frequency of IL-10-1082 between patients with AF and the control group, although the expression of inflammation and cell adhesion molecules in a rat model of venous thrombosis was inhibited by viral IL-10 gene transfer (40), and the IL-10-1082A/G polymorphism is associated with the risk of deep venous thrombosis (41). A C582T polymorphism in the IL-4 gene [odds ratio, OR ¼: 1.40, 95% CI: 1.13–1.73, p ¼: 0.003)] was found to be an independent predictor of thromboembolic stroke (42).

In our study, we did observe significantly higher baseline levels of IL-4 and IL-10 in patients with thromboembolic or bleeding events during the 2-year follow-up, compared to those in patients without these adverse events. Especially, compared to those in the patients without adverse events, the baseline levels of IL-4 in patients with venous thrombosis during follow-up increased by 53-fold. It suggested that IL-4 and IL-10 may highly be associated with the progression of AF-related adverse events. However, IL-4 levels could be associated with some medical setting, such as scleroderma, multiple sclerosis, chronic sinusitis, inflammatory bowel diseases, bronchial asthma, atopic dermatitis (43–45). Although we excluded patients with rheumatic heart disease and serious uncontrolled infection, other disease didn't be excluded, which needed to be explored further.

There is a close relationship between inflammation, platelet activation and AF. Inflammatory cytokines play a certain role in the maintenance and prognosis of AF (46). Inflammation leads to cardiac endothelial damage or dysfunction, and promotes the aggregation of platelets, and the activation fibrinolytic protein and clotting factors (47). At the same time, the reconstruction of atrial anatomy cause by AF provides the environment for platelet aggregation and blood coagulation which promotes thrombosis (48). Chemokines, to a certain extent, is involved in inflammation, angiogenesis and necrosis, which is a key mediator of inflammation-related cell migration (49). Studies suggests that pro-inflammatory cytokines, such as IL-1β, IL-6, and IL-8, promote a pro-coagulant state through the induction of tissue factor. Anti-inflammatory cytokines, including IL-2, IL-4, and IL-10, act to reduce coagulation induced by other pro-inflammatory factors (50). Additionally, the cytokines IL-4, IL-6, IL-8, IL-10, and IL-1β were all significantly elevated in massive PE patients in comparison to those with low-risk and submassive PE, indicating the overwhelming inflammatory response with larger embolism (51). It is found that the polymorphisms of IL-4−589C/T and IL-13 intron 3C/T were associated with the increased risk of VTE among women, while IL-6 174C/G was found to make men susceptible to the VTE event (51). When compared with patients with lacunar infarct, those with cardioembolic stroke were shown to have significantly increased levels of proinflammatory cytokines such as IL-1β and IL-6 (51). The mechanisms of chemokines with increased levels in patients that experienced thrombosis and bleeding events during the 2-year follow-up in our study should be further studied in vivo and in vitro.