We enrolled 243 consecutive adult patients with documented non-valvular AF according to European Society of Cardiology (ESC) Guidelines [16], who were admitted to John Paul II Hospital in Cracow, Poland, between June 2014 and July 2016. The study population was presented in detail previously [14]. Patients with acute myocardial infarction (MI), VTE within last 12 months, end-stage chronic kidney disease, severe liver injury, active cancer, pregnancy, symptoms of acute infection and C-reactive protein (CRP) > 10 mg/L were excluded. The risk of stroke was evaluated by CHA2DS2-VASc score [16]. The AF subtypes were assessed according to the 2012 ESC guidelines [16]. The diagnosis of HF and prior MI was established based on ESC guidelines [17, 18]. Reduced left ventricular ejection fraction (LVEF) on transthoracic echocardiography was defined as ≤ 40% [17]. Ischemic stroke was recognized based on typical symptoms confirmed on computed tomography and magnetic resonance imaging [19]. Fasting glycemia ≥ 7 mmol/L on 2 separate occasions or the hypoglycemic therapy were the criteria for diabetes mellitus (DM). The Local Ethic Committee of Jagiellonian University approved the study protocol, and the study was conducted in accordance with the Declaration of Helsinki. All participants gave their written informed consent. The study adhered to the STROBE reporting guidelines.

Regarding patients on direct oral anticoagulants (DOACs), the fasting blood was collected 24–28 h since the administration of the last dose of rivaroxaban and 12–18 h since the administration of apixaban or dabigatran. Patients on vitamin K antagonists (VKA) had blood drawn at least 24 h after the last dose of low-molecular-weight heparin administered as bridging therapy. Blood was drawn between 8 and 10 AM from the antecubital vein to tubes containing (9:1) 3.2% trisodium citrate which were centrifuged at 2500 g at 20 °C for 20 min whereas serum tubes were centrifuged at 1600 g at 4 °C for 10 min. Aliquots were stored at -80 °C. Routine laboratory investigations, including N-terminal B-type natriuretic peptide (NT-proBNP) were assessed by routine hospital techniques. Plasminogen activator inhibitor-1 (PAI-1) antigen and thrombin activatable fibrinolysis inhibitor (TAFI) activity (both Hyphen-Biomed, Neuville-Sur-Oise, France) were assayed using ELISA kits. Fibrinogen was determined by the Clauss method. The quantitative determination of the plasminogen and α2-antiplasmin activity were assessed in citrated plasma by the synthetic chromogenic substrate method with the use of a commercial calibration standard and values were expressed as a percentage of normal (both STA-Stachrom, Diagnostica Stago, Asnières-sur-Seine, France). In this assay, plasma was incubated with the streptokinase reagent which consist of streptokinase, human albumin, and buffer in the presence of fibrinogen. Then the plasminogen activity was quantified by the plasminogen-streptokinase complexes action on the synthetic chromogenic substrate, as the amount of p-nitroaniline release, measured at 405 nm. The Von Willebrand factor (vWF) antigen was evaluated by latex immunoassay (Diagnostica Stago, Asnières-sur-Seine, France). All results have been presented according to the manufacturer’s instructions.

Carbonyl contents were assessed using the reaction of 2,4-dinitrophenylhydrazine (DNPH) with PC, which leads to the formation of a Schiff base, and subsequently, the corresponding hydrazone which could be analyzed spectrophotometrically, according to Becatti et al. [5]. Briefly, 100 μL of plasma was incubated with 400 μL of DNPH and then precipitated with trichloracetic acid. Subsequently, the pellet was washed several times with a 1:1 mixture of ethanol/ethyl acetate and resuspended in 500 μL of guanidine hydrochloride. Absorbance at 370 nm was measured using spectrophotometer (Tecan, Sunrise). Carbonyls contents were evaluated by using a molar extinction coefficient of 22,000 mol/L−1 cm−1 and expressed as nmol/mL of PC per 1 mg of protein. The in-house reference range for healthy subjects is 0.54–2.03 nmol/mg [3]. The intra-assay variability was 5.9%, whereas inter-assay variability equaled 7.8%.

Calibrated automated thrombogram was assessed as described [20], and performed according to manufacturer’s instructions (Thrombinoscope BV, Maastricht, Netherlands). Briefly, the assay was performed in a 96-well plate fluorometer (Ascent Reader, Thermolabsystems OY, Helsinki, Finland). To 80 μL platelet-poor plasma 20 μL of (TF)-based activator (PPP Reagent; final TF concentration, 5 pM) and FluCa solution (both Diagnostica Stago) were added. Fluorescence readings began immediately (at 390 nm excitation and 460 nm emission wavelengths, at 37 °C) was followed over a 60 min period. Each plasma sample was analyzed in duplicate. Endogenous thrombin potential (ETP), calculated as the area under the curve of thrombin formed in time, was used to measure thrombin generation capacity. Inter-assay coefficients of variation were < 7%.

Fibrin clot permeability and fibrinolysis capacity determined as clot lysis time (CLT) were assessed as described [21]. Briefly, to assess clot permeability CaCl2 (20 mM) and human thrombin (1 U/mL; Sigma-Aldrich, St Louis, USA) were mixed with citrated plasma. The permeation coefficient (Ks) reflecting the average size of pores formed in the fibrin network. Ks was calculated as follows: Ks = Q × L × η / t × A × Δp. Q is the flow rate in percolating time (t), L is the length of a fibrin gel, η is the viscosity of liquid, A is the cross-sectional area, and Δp is a differential pressure. The interassay and interassay coefficients of variation were < 7%. CLT was defined as the time from the midpoint of the clear-to-maximum-turbid transition, representing clot formation, to the midpoint of the maximum-turbid-to-clear transition representing clot lysis. 12 μmol/L phospholipid vesicles, 15 mmol/L CaCl2, 0.6 pmol/L tissue factor (TF) (Innovin, Siemens) and 60 ng/mL recombinant tissue plasminogen activator (tPA) (Boehringer Ingelheim, Ingelheim, Germany) were mixed with citrated plasma to evaluate CLT. The turbidity was measured at 405 nm. The intra-assay variability was 8%.

The long-term outcomes were assessed by telephone or clinical visit for at least twice a year. The primary endpoint was ischemic stroke defined as shown above. The secondary endpoints were death and major bleeding defined according to the ISTH bleeding tool assessment [22].

The study was powered to have a 90% chance of detecting a 20% difference with 20% standard deviation in PC using a P value of 0.05 between AF patients with and without stroke, assuming stroke event rate for 8–9%. Based on the data in the published articles [4], to demonstrate such a difference or greater, 14 patients with and 138 without stroke were required in each group. For a P value of 0.01, 20 and 196 patients were required in respective groups.

Statistical analyses were performed using the SPSS Statistics software (Version 29.0.0.0, IBM Corp., Armonk, NY, USA). Continuous variables were expressed as a median (interquartile range [IQR]), whereas categorical variables were shown as a number (percentage). Normal distribution was assessed by the Shapiro–Wilk test. Intergroup differences were evaluated by Student’s t-test when normally distributed or by the Mann–Whitney U test for non-normally distributed variables. Analysis of variance followed by a post hoc Bonferroni test was used to compare differences of single measurements in more than two groups with normally distributed data whereas non-normally distributed data were analyzed by Kruskal–Wallis test and differences between groups were identified using a test for multiple comparisons of mean ranks. Categorical variables were compared by Fisher’s exact test. The Pearson or Spearman rank correlation coefficients were calculated to test the association between two variables having a normal or non-normal distribution, respectively. All independent variables potentially associated with both the exposure and outcome, and which lacked significant correlation with other independent variables, were included in the Cox proportional hazard regression to determine predictors of stroke as well as major bleeding or included in the linear multivariable analysis to find parameters independently associated with fibrin clot properties. A two-tailed P < 0.05 was considered statistically significant.