Introduction
Intracerebral haemorrhage (ICH) occurs in c.5/100,000 individuals, is associated with substantial mortality [1], and carries an up to 15% risk of recurrence [2]. Apart from patients in whom the cause of bleeding is identified, up to 40% of patients classify as those with ICH of unknown cause [2, 3]. We have recently demonstrated that young adults with ICH of unknown cause are characterised by prohaemorrhagic fibrin clot phenotype, along with lower factor (F) II, lower FVII, and higher antithrombin (AT) activity [4]. Little is known about the role of natural anticoagulants in the pathogenesis of ICH, despite the fact that elevated levels of several natural anticoagulants, such as activated protein C, thrombomodulin or tissue factor pathway inhibitor (TFPI), have been demonstrated in haemorrhages of unknown cause [5].
Tissue factor pathway inhibitor (TFPI), a serine protease inhibitor occurring in two isoforms, TFPIα and TFPIβ, and synthesised mainly by endothelial cells, inhibits tissue factor (TF)–FVIIa complexes, whereas TFPIα additionally blocks the early forms of prothrombinase (complex of FXa and FVa) [5–8]. Up to 80% of the TFPIα isoform is bound to the endothelium and the remaining 20% circulates in the plasma, of which two thirds is associated with lipoproteins (mainly low–density lipoprotein [LDL]) and C–terminally degraded [6]. The remaining 20% that circulates in the plasma occurs in either free form i.e. full–length (10%, the active form) or carboxy–terminal truncated form (10%) [9].
It has been demonstrated that free TFPIα (fTFPIα) is increased in plasma obtained from patients with mild-to-moderate bleeding disorders such as epistaxis, easy bruising or menorrhagia, in particular in those with bleeding disorders of unknown cause and with platelet function disorders [10]. Interestingly, fTFPIα levels in such patients have been positively correlated with the lag time of the thrombin generation curve [10]. Lower levels of fTFPIα have been (albeit inconsistently) reported to increase the risk of thrombosis [11, 12]. In the context of intracerebral haemorrhage, total TFPI levels have been reported as unaffected in adults in the acute phase of a subarachnoid haemorrhage [13] and in acute ICH in children with haemophilia compared to control subjects [14].
Clinical rationale for study
To the best of our knowledge, elevated fTFPIα in ICH of unknown cause has not been previously investigated. We hypothesised that, as in bleeding of unknown cause in other locations, patients following ICH of unknown cause have elevated levels of this inhibitor. Therefore the aim of this study was to assess plasma fTFPIα levels and its associations with coagulation factors, fibrin clot properties and lysis in patients with ICH of unknown cause below the age of 50.
Material and methods
Patients
We recruited 44 consecutive patients who had suffered ICH of unknown cause at least three months prior to referral to the Centre for Coagulation Disorders, Krakow, Poland between 2013 and 2019. This patient group, and a control group matched for age, sex, body mass index (BMI), and hypertension, have been described in detail previously [4]. Briefly, the inclusion criteria were age 18–50 years and a diagnosis of ICH of unknown cause based on clinical symptoms, computed tomography scan, and according to the SMASH–U classification [15]. The key exclusion criteria were: known malignancy, kidney disease (acute up to stage G 3b and chronic up to stage G5), advanced liver injury (classes B and C on the Child–Pugh Score scale), diagnosed coagulation factor deficiencies, von Willebrand disease, thrombocytopenia (< 100,000/µl), brain aneurysm, arteriovenous malformation, and trauma. The patients did not show any clinical signs or symptoms of infection or deep venous thrombosis.
We collected data on demographics, comorbidities, current smoking, alcohol use and medications. The severity of neurological deficit was measured on admission using the National Institutes of Health Stroke Scale, and stroke outcome was assessed at discharge using the modified Rankin Scale. Definitions of all the comorbidities were as defined previously [16]. All participants gave their written informed consent, and the study was approved by the local Ethics Committee.
Laboratory investigations
Fasting blood samples were obtained from an antecubital vein, between the hours of 8am and 10am. Routine laboratory investigations included blood cell counts, glucose, creatinine, C–reactive protein, D–dimer, international normalised ratio, and activated partial thromboplastin time. Additionally, fibrinogen (von Clauss assay), FII, FV, FVII, FVIII, FIX, FX and FXI, AT activity, plasminogen activator inhibitor–1 antigen (PAI–1; ELISA, Hyphen, Neuville–sur–Oise, France) and prothrombin fragments 1 + 2 (F 1.2; ELISA, Siemens, Marburg, Germany) were assayed as previously described [4]. fTFPIα was determined with a commercially available ELISA kit (Diagnostica Stago, Asnieres, France). In our lab, the reference values for healthy individuals are 4.0–11.5 ng/mL.
Analysis of plasma fibrin clot variables was carried out as previously described [3]. Briefly, fibrin clot permeability (Ks) was measured using a hydrostatic pressure-driven system based on the volume of a percolating buffer using the formula: Ks = Q x L x η/t x A x Δp, where Q is the flow rate in time, L is the length of a fibrin gel, η is the viscosity of liquid (in poise), t is the percolating time, A is the cross–sectional area (in cm2), and Δp is the differential pressure (in dyne/cm2).
To measure fibrin clot turbidity, polymerisation was initiated by mixing plasma citrated samples 2:1 with a Tris buffer containing 0.6 U/mL human thrombin (Sigma–Aldrich, St. Louis, MO, USA) and 50 mmol/L calcium chloride. Using a Perkin–Elmer Lambda 4B spectrophotometer (Molecular Devices, San Jose, CA, USA), absorbance was read at 405 nm, and the lag phase of the turbidity curve, as well as the maximum absorbance at the plateau phase (ΔAbs), were recorded. The lag phase denotes the time required for initial protofibril formation, whereas ΔAbs indicates the number of protofibrils per fibre.
Fibrinolysis capacity was assessed in three assays. In the first, the turbidity method was used to determine clot lysis time (CLT), 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. In this assay, the citrated plasma was mixed with calcium chloride (final concentration 15 mmol/L), recombinant human tissue factor (final concentration 0.6 pmol/L; Innovin, Siemens, Marburg, Germany), phospholipid vesicles (final concentration 12 μmol/L), and recombinant tissue–type plasminogen activator (rtPA, final concentration 60 ng/mL; Boehringer Ingelheim, Ingelheim, Germany). The second marker of fibrinolysis was the time required for a 50% decrease in clot turbidity (t50%). Here, 100 μL of citrated plasma was diluted with 100 μL of a Tris buffer containing 20 mM calcium chloride, 1 U/mL human thrombin (Sigma–Aldrich), and 14 μM rtPA (Boehringer Ingelheim). In the third assay, the lysis rate of the fibrin clots formed as described above and perfused with buffer containing a relatively high final concentration of rtPA i.e. 0.2 μmol/l (Boehringer Ingelheim) was determined by measuring the D–dimer concentrations (Abcam, Waltham, MA, US) every 15 min. in the effluent. The maximum rate of D–dimer increase (D–Drate) and maximum D–dimer concentrations (D–Dmax) were recorded.
Statistical analysis
Data was expressed as mean (standard deviation, SD) or median (interquartile range, IQR), according to its distribution assessed by the Shapiro–Wilk test. Differences in variables between the ICH group and controls were analysed using a Student t–test, U–Mann Whitney test, Chi2 test or Fisher’s exact test, as appropriate. Correlations were assessed using Pearson’s correlation or Spearman’s rank correlation coefficient, separately for ICH group and controls. Univariate and multivariate logistic regression were performed to assess the association between fTFPIα levels and the occurrence of ICH. In the multivariate regression, the model was adjusted for age, sex, hypertension, and platelet count. Two–sided p values of < 0.05 were considered statistically significant. Analysis was performed using the STATISTICA 12.0 software package (Stat Soft Inc., Tulsa, OK, USA, 2011).
Results
The ICH group comprised 44 patients with a median age of 41 (IQR 27–47) years, of whom 20 (45.5%) were female. As many as 23 (52.3%) were obese, 16 (36.4%) had hypertension, and 14 (31.8%) were current smokers. They did not differ from the controls (n = 47) in terms of demographics, comorbidities or medications, as shown previously [4]. Baseline patient characteristics are set out in Table 1.
Variable |
fTFPIα > 11.5 ng/mL (n = 9) |
fTFPIα ≤ 11.5 ng/mL (n = 35) |
P–value |
ICH group (n = 44) |
Controls (n = 47) |
P-value |
Age (years) |
44.8 (2.5) |
37.8 (7.4) |
0.008 |
41.0 (27.0–47.0) |
40.0 (32.0–44.0) |
0.46 |
Female sex, n [%] |
0 (0) |
20 (57.1) |
0.007 |
20 (45.5) |
22 (46.8) |
0.90 |
BMI, kg/m2 |
25.4 (3.4) |
25.5 (4.1) |
0.99 |
25.5 (3.9) |
25.7 (4.3) |
0.81 |
Medical history |
||||||
Hypertension, n [%] |
2 (22.2) |
14 (40.0) |
0.55 |
16 (36.4) |
21 (44.7) |
0.42 |
Diabetes mellitus, n [%] |
2 (22.2) |
4 (11.4) |
0.77 |
6 (13.6) |
4 (8.5) |
0.43 |
Coronary artery disease, n [%] |
1 (11.1) |
2 (5.7) |
0.87 |
3 (6.8) |
1 (2.1) |
0.28 |
Previous myocardial infarction, n [%] |
1 (11.1) |
1 (2.9) |
0.87 |
2 (4.5) |
0 (0) |
0.14 |
Current smoking, n [%] |
3 (33.3) |
11 (31.4) |
0.77 |
14 (31.8) |
14 (29.8) |
0.83 |
Medications |
||||||
ACEI, n [%] |
3 (33.3) |
9 (25.7) |
0.97 |
12 (27.3) |
18 (38.3) |
0.26 |
β–blockers, n [%] |
2 (22.2) |
7 (20.0) |
0.75 |
9 (20.5) |
9 (19.1) |
0.88 |
Calcium channel blocker, n [%] |
1 (11.1) |
4 (11.4) |
0.57 |
5 (11.4) |
7 (14.9) |
0.62 |
Diuretics, n [%] |
1 (11.1) |
5 (14.3) |
0.77 |
6 (13.6) |
13 (27.7) |
0.10 |
Statins, n [%] |
3 (33.3) |
5 (14.3) |
0.40 |
8 (18.2) |
11 (23.4) |
0.54 |
Laboratory investigations |
||||||
Haemoglobin, g/dL |
13.6 (0.9) |
13.8 (1.0) |
0.73 |
13.7 (1.0) |
13.9 (1.3) |
0.43 |
White blood cells, 109/L |
6.7 (6.1–7.1) |
6.8 (5.7–8.0) |
0.80 |
7.1 (6.4–8.1) |
6.2 (5.5–7.5) |
0.007 |
Platelets, 109/L |
181.0 (156.0–205.0) |
232.0 (189.0–289.0) |
0.016 |
214.5 (179.5–257.5) |
248.0 (211.0–298.0) |
0.02 |
APTT, s |
32.1 (30.8–33.1) |
30.7 (29.2–33.0) |
0.68 |
31.3 (29.3–33.0) |
29.7 (27.2–32.3) |
0.14 |
ALT, U/L |
18.0 (14.0–29.0) |
22.0 (17.0–30.0) |
0.38 |
22.0 (17.0–29.5) |
25.0 (19.0–30.0) |
0.39 |
Creatinine, μM |
87.6 (68.5–98.0) |
73.0 (65.3–81.4) |
0.36 |
74.1 (65.4–88.9) |
73.0 (67.0–81.0) |
0.73 |
C–reactive protein, mg/L |
2.4 (2.1–4.4) |
2.4 (1.6–3.4) |
0.48 |
2.4 (1.8–3.8) |
1.9 (1.2–3.4) |
0.15 |
LDL cholesterol, mM |
2.6 (2.5–3.1) |
3.2 (2.5–3.9) |
0.25 |
3.1 (2.5–3.5) |
3.0 (2.4–3.5) |
0.35 |
Coagulation variables |
||||||
Fibrinogen, g/L |
2.4 (2.2–2.8) |
2.7 (2.5–3.3) |
0.06 |
2.7 (2.4–3.1) |
3.0 (2.3–3.5) |
0.46 |
D–Dimer, ng/mL |
346.1 (110.0) |
333.8 (116.6) |
0.77 |
333.0 (218.0–422.5) |
293.0 (218.0–398.0) |
0.22 |
F1.2, nmol/L |
121.0 (119.0–125.0) |
128.0 (110.0–149.0) |
0.78 |
124.0 (113.0–148.5) |
119.0 (108.0–152.0) |
0.61 |
Factor II, [%] |
99.6 (6.3) |
97.2 (10.4) |
0.50 |
98.9 (90.2–104.2) |
108.0 (98.0–120.0) |
0.0001 |
Factor V, [%] |
99.4 (10.8) |
98.8 (9.0) |
0.89 |
99.9 (95.2–104.1) |
100.0 (93.0–114.0) |
0.06 |
Factor VII, [%] |
90.9 (89.4–107.1) |
93.2 (87.9–103.1) |
0.59 |
92.6 (88.2–104.6) |
103.0 (95.0–114.0) |
0.0003 |
Factor VIII, [%] |
121.2 (91.6–126.9) |
106.8 (86.5–126.3) |
0.62 |
108.8 (87.6–126.6) |
116.0 (102.0–134.0) |
0.066 |
Factor IX, [%] |
99.2 (12.6) |
97.1 (11.8) |
0.65 |
97.5 (11.8) |
102.2 (11.8) |
0.06 |
Factor X, [%] |
100.6 (87.9–110.1) |
99.4 (84.2–109.0) |
0.73 |
99.5 (85.4–109.6) |
101.0 (95.0–109.0) |
0.14 |
Antithrombin, [%] |
101.3 (16.1) |
107.3 (11.9) |
0.21 |
106.1 (12.9) |
97.0 (10.9) |
0.0004 |
fTFPIα, ng/mL |
13.0 (11.9–13.3) |
8.0 (7.5–8.6) |
< 0.001 |
8.3 (7.6–9.5) |
7.4 (6.9–8.5) |
0.006 |
Patients following ICH of unknown cause had 10.8% higher median fTFPIα levels than controls [8.3 (7.6–9.5) vs. 7.4 (6.9–8.5) ng/mL; p = 0.006; Fig. 1]. fTFPIα correlated with age both in the ICH group (r = 0.38; p = 0.01) and controls (r = 0.31, p = 0.03) and was higher in males than in females both in the ICH group (10.0 ± 2.5 vs. 7.8 ± 0.7 ng/mL; p = 0.0004) and in the controls (8.1 ± 1.4 vs. 7.1 ± 1.1 ng/mL; p = 0.007). However, fTFPIα was not related to any comorbidities, medications or routine laboratory investigations, including inflammatory markers or D–Dimer. In the ICH subjects, fTFPIα levels negatively correlated with fibrinogen, PAI–1 antigen and ΔAbs (Fig. 2 A, B, and C, respectively), but not Ks, CLT, t50%, D–Drate or D–Dmax, coagulation factors or antithrombin. However, ΔAbs positively correlated with fibrinogen (r = 0.68, p < 0.0001), and inversely correlated with Ks (r = –0.57, p = 0.0001), while correlation with t50% was of borderline significance (r = 0.30, p = 0.05). PAI–1 antigen demonstrated correlations with CLT (r = 0.54, p = 0.0001) and t50% (r = 0.48, p = 0.0009). In the control group, fTFPIα was not associated with any variable apart from age.
In the ICH group, fTFPIα level >11.5 ng/mL was found in nine patients (20.5%). These individuals were all males, older and with lower platelet counts than the remaining ICH subjects (Tab. 1). Interestingly, they were also characterised by a longer lag phase of the turbidity curve and lower ΔAbs (Tab. 2). Analysis of the ICH subjects with fTFPIα in the top quartile (> 9.4 ng/mL, 11 patients) versus the remainder showed similar results. In the control group, none of the subjects had a fTFPIα level above the upper limit of the reference range (> 11.5 ng/mL).
Variable |
fTFPIα > 11.5 ng/mL (n = 9) |
fTFPIα ≤ 11.5 ng/mL (n = 35) |
P-value |
Ks, 10–9 cm2 |
9.5 (9.0–10.1) |
9.0 (8.2–9.6) |
0.16 |
Lag phase, s |
49.1 (5.0) |
44.9 (4.8) |
0.023 |
ΔAbs |
0.70 (0.12) |
0.76 (0.07) |
0.042 |
CLT, min |
61.0 (56.0–75.0) |
57.0 (57.0–82.0) |
0.37 |
PAI–1:Ag., ng/mL |
8.8 (7.7–12.0) |
7.8 (7.8–14.7) |
0.19 |
t1/2, min |
8.0 (1.1) |
7.9 (0.9) |
0.89 |
D–Drate, mg/L/min |
0.082 (0.007) |
0.079 (0.006) |
0.28 |
D–Dmax, mg/L |
3.5 (3.4–3.7) |
3.4 (3.4–3.9) |
0.31 |
In univariate analysis, a 1 ng/mL increase in fTFPIα was associated with a 61% greater chance of ICH (OR 1.61, 95% CI 1.19–2.18). After adjusting for potential confounders, this association remained significant, with area under the curve (AUC) for the full model (age, sex, hypertension, platelet count, fTFPIα) of 0.74, 95% CI 0.64–0.84, p = 0.021.
Discussion
To the best of our knowledge, the present study is the first to show that adult patients with a history of ICH of unknown cause under 50 years of age demonstrate elevated levels of fTFPIα, the main physiological regulator of the initiation of blood coagulation. Increasing concentrations of fTFPIα were associated with impaired fibrin clot formation, decreased clot density, and impaired inhibition of fibrinolysis. Our findings suggest a previously unreported mechanism that may contribute to the occurrence of ICH in young adults. Given recent advances in targeting TFPI with monoclonal antibodies [17], our findings might have therapeutic implications if validated in future studies and could help reduce the risk of ICH recurrence.
In the present study, the detected levels of fTFPIα were generally concordant with the literature, with higher levels of fTFPIα in males and older subjects [10. 18]. We did not observe the positive correlations with BMI that have been reported both in patients with mild bleeding and in controls [10]. In the studied ICH group, the levels of fTFPIα were mildly elevated, which was similar to the results in patients with mild and moderate bleeding disorders [10]. Since it has been shown that males experience ICH more frequently and at a younger age than women [21], we speculate that elevated fTFPIα in male ICH survivors at least in part explains this observation.
Other potential factors that affect haemorrhage occurrence deserve comment. The prevalence of hypertension in the current sample was similar to other ICH cohorts [2]. Of note, in our patients it was mild and not considered to be a cause of the index event. More importantly, it was not associated with fTFPIα concentration. The rate of smokers and diabetics did not differ from other studies [22, 23]. LDL cholesterol levels and statin use among the current ICH group did not differ from the controls and was not associated with fTFPIα, and therefore it is unlikely to have influenced the results. Another potential contributor to ICH occurrence is thrombocytopenia, which has been also observed in patients with COVID–19 [24, 25]. The current ICH group had lower platelet count than controls (albeit within the normal range). However, after adjusing for platelet count, higher fTFPIα concentrations were still associated with a greater chance of ICH.
We have shown that fTFPIα levels negatively correlate with fibrinogen, which has not been described previously. Fibrinogen is the key determinant of fibrin clot structure and function [26, 27]. In our study, lower fibrinogen levels were in line with lower fibrin clot maximum absorbance in turbidimetry, which reflects the decreased density of the fibrin clot [28]. Interestingly, lower clot density was associated with higher fTFPIα, increased clot porosity and a tendency to clot lysis. ICH subjects with fTFPIα levels above the upper limit of the reference range also exhibited prolonged lag time of the turbidity curve. In patients with mild and moderate bleeding disorders, fTFPIα correlated positively with thrombin generation parameters: prolonged lag time and increased time to peak [10]. In the present study, we did not measure thrombin generation [4].
An important finding is the decreasing concentration of PAI–1 antigen with elevated levels of fTFPIα. PAI–1 is of key importance in regulating fibrinolysis by binding active tPA molecules, forming an inactive complex and preventing plasminogen activation. Its deficiency can cause hyperfibrinolytic bleeding [29]. Although in our subjects PAI–1 concentration was within reference limits, it could still contribute to bleeding [30]. PAI–1 also has an impact on the results of fibrinolysis assays [31]. In the present study, it strongly correlated with CLT and t50%, meaning that it might be another factor associated with fTFPIα that potentiates the lysis of the fibrin clot.
Our study has several limitations. Firstly, the number of participants was restricted, although the number of patients in the ICH group was similar to the subgroups with ICH of undetermined aetiology in young adults in other studies [2 ,22, 32]. Secondly, the results do not necessarily demonstrate a cause and effect relationship, and are not generalisable to the most severe ICH patients. The impact of clinical factors such as resistant hypertension [33] and alcohol abuse [34] cannot be excluded. We did not examine coagulation parameters in the acute period of ICH, although it has been shown that fTFPIα is unchanged in the acute phase of ICH [13, 14]. The fTFPIα assay is currently for research use only; perhaps further steps should be made towards its approval in clinical practice. The coagulation parameters were evaluated a few months after the index ICH; future studies could investigate fTFPIα levels as a prognostic factor for ICH.
To conclude, young adults who suffer from ICH demonstrate higher levels of a natural anticoagulant, fTFPIα, which is associated with prolonged fibrin clot formation, decreased clot density, and impaired inhibition of fibrinolysis.
Clinical implications/future directions
Our findings contribute to the understanding of the pathophysiology of ICH of unknown cause, and may form the foundations for future large-cohort studies of patients with ICH with long term follow–up.
Article information
Acknowledgements: None.
Conflicts of interest: None.
Funding: This study was supported by the Jagiellonian University Medical College, Krakow, Poland (grant number N41/DBS/000682, to A.U.) and by the science fund of the St. John Paul II Hospital, Krakow, Poland (no. FN/15/2024 to M.B.).