Low factor XIII levels after intravenous thrombolysis predict short-term mortality in ischemic stroke patients

In this observational study we investigated whether levels of factor XIII (FXIII) and its major polymorphisms affect the outcome of thrombolysis by recombinant tissue plasminogen activator (rtPA) in acute ischemic stroke (AIS) patients. Study cohort included 132 consecutive AIS patients undergoing i.v. thrombolysis within 4.5 h of symptom onset. Blood samples taken on admission, immediately after and 24 h after therapy were analyzed for FXIII activity and antigen levels. FXIII-A p.Val34Leu, p.Tyr204Phe, FXIII-B p.His95Arg and intron K(IVS11 + 144) polymorphisms were genotyped. Neurological deficit was assessed using the National Institutes of Health Stroke Scale. Intracranial hemorrhage was classified according to ECASSII criteria. Long-term functional outcome was defined at 3 months post-event by the modified Rankin scale. FXIII levels showed a gradual decrease immediately after thrombolysis and 24 h later, which was not related to therapy-associated bleeding. In a multiple logistic regression model, a FXIII level in the lowest quartile 24 h post-lysis proved to be an independent predictor of mortality by 14 days post-event (OR:4.95, 95% CI:1.31–18.68, p < 0.05). No association was found between the investigated FXIII polymorphisms and therapeutic outcomes. In conclusion, our findings indicate that FXIII levels 24 h after thrombolysis might help to identify patients at increased risk for short-term mortality.

A total of 132 consecutive AIS patients undergoing thrombolysis were included in the study. Mean age was 69.0 ± 12.2 years, and 58.3% were men. The median NIHSS score on admission was 8 (interquartile range: [5][6][7][8][9][10][11][12][13][14]. According to the TOAST criteria, most patients suffered a large vessel thrombosis (n = 49, 37.1%). Average time from symptom onset to treatment with rtPA was less than 3 h in the cohort and the duration of thrombolysis was approximately one hour for each patient. In case of 7 patients intravenous thrombolytic therapy was supplemented with intraarterial thrombolysis according to the standard protocol; the final dose of rtPA and the duration of thrombolysis was not significantly different for these patients as compared to the rest of the study group.
The influence of thrombolysis on FXIII levels. Mean FXIII activity and antigen levels before and during the course of thrombolysis are shown in Table 2. On admission (before thrombolysis) a considerable number of patients (n = 39, 29.5%; data not shown in Table) had FXIII levels above the upper limit of the reference interval (above 143% or 28 mg/l). FXIII levels showed a gradual decrease after thrombolysis; at 24 h post-lysis significantly lower FXIII levels were detected as compared to initial values. Strong correlation was observed between FXIII activity and FXIII antigen levels at all investigated occasions (time points A, B and C: Pearson r = 0.915, p < 0.001; r = 0.919, p < 0.001 and r = 0.917, p < 0.001, respectively; data not shown in Table).

Association of FXIII levels on admission with stroke characteristics. FXIII levels on admission
showed no association with the severity of the stroke (Table 3) and no correlation was observed between FXIII levels and NIHSS scores on admission (Pearson r = −0.09, p = 0.28). FXIII activity was significantly higher in case of atherothrombotic stroke as compared to strokes of cardioembolic origin (Table 3). FXIII activity and antigen levels showed a weak negative association with the age of the patients (Pearson r = −0.299, p < 0.001 and r = −0.286, p < 0.001, respectively; data not shown in Table). FXIII antigen levels were significantly higher in active smokers vs. never-smokers (24.60 mg/l vs. 21.15 ± 0.9 mg/l, respectively, p < 0.05; data not shown in Table). No correlation was observed between FXIII levels and any measured routine clinical chemistry, hemostasis and hematology parameters. No correlation was observed between FXIII levels measured at any time points and data obtained from CT imaging analysis (ASPECTS; data not shown). No correlation was observed between FXIII activity/antigen levels measured on admission and the elapsed time between symptom onset to thrombolysis treatment (r = 0.081, p = 0.186 and r = 0.061, p = 0.251, respectively; data not shown in Table).
Association of FXIII levels during thrombolysis and therapy outcomes. FXIII levels before the initiation of the therapy (time point A) or immediately after thrombolysis (time point B) showed no association with short-term functional outcomes (Table 4). On the contrary, at 24 h after thrombolysis (time point C), significantly lower FXIII levels were detected in those patients who died within the first week following treatment. FXIII levels were remarkably low in these patients, approximately 50% as compared to values measured in patients with other outcomes. Mortality and low FXIII levels were not associated with bleeding complications. On the other hand, no difference was observed in the FXIII levels of patients with any other outcomes except for mortality (i.e. favorable outcome vs. no change or unfavorable outcome; data not shown). Patients experiencing bleeding after therapy were separately handled during the analysis due to different assumed underlying pathomechanisms.
Low FXIII levels were associated with mortality by day 14 post-event as well. When studying the baseline characteristics of patients grouped according to mortality by the end of the 2 nd week, it was found that besides the NIHSS on admission, only FXIII levels measured before and 24 h after thrombolysis were significantly different in the two groups (  functional outcomes at 3 months post-event ( Fig. 1). FXIII levels 24 h after lysis were significantly lower in those patients who died by the end of the 3 rd month after the event (mRS 6). In order to test whether a low FXIII level measured 24 h after thrombolysis is an independent predictor of short-and long-term mortality of patients, backward multiple regression analysis for mortality was performed including all potentially relevant risk factors and confounders (Table 6). NIHSS score on admission was found to be an independent predictor of mortality by 14 days FXIII levels are not associated with therapy-associated symptomatic or asymptomatic intracerebral hemorrhage. In patients suffering therapy-associated intracerebral bleeding complications (n = 13,  Table 4. Association of factor XIII (FXIII) levels during thrombolysis with short-term functional outcomes. Short-term functional outcomes were assessed at 7 days post-event. Results are expressed as mean ± standard deviation. n, number; favorable outcome: a decrease in NIHSS score by at least 4 points or to 0 by day 7; no change: a change in NIHSS score less than 4 points by day 7; unfavorable outcome: an increase in NIHSS score by at least 4 points by day 7; *stroke-associated death excluding death due to intracerebral hemorrhage; † ANOVA with Bonferroni post hoc test; ‡ adjusted to age and sex in the statistical model.  Table). None of the investigated factor XIII polymorphisms were associated significantly with stroke severity, unfavorable short-term or long-term outcomes of therapy, therapy-associated symptomatic intracranial hemorrhage and mortality (Table 7). In case of FXIII-A p.Val34Leu polymorphism, a trend could be observed showing that carriers of the Leu allele might be protected against unfavorable short-term outcomes (OR: 0.33; 95% CI: 0.09-1.10), but the association was not statistically significant (p = 0.072).

Discussion
Despite the fact that activated FXIII plays a prominent role in the protection against fibrinolysis by plasmin, this is the first comprehensive study investigating FXIII activity and antigen levels during thrombolysis in AIS patients and the association of FXIII levels and polymorphisms with clinical outcomes. Here we show that FXIII activity and antigen levels decrease gradually during thrombolysis. The mechanism involved in such reduction of FXIII     levels is not entirely clear. In theory, this phenomenon might be attributed to consumption and/or degradation of FXIII. A likely explanation of the considerable diminution of FXIII levels post-stroke is its continuous incorporation into the growing thrombus, leading to consumption due to ongoing activity of the coagulation system. This hypothesis was suggested by an earlier pilot study, where FXIII subunit levels (but not FXIII activity) were measured in a group of AIS patients 14 . In that cohort, 41 patients received thrombolysis by rtPA or urokinase and their results did not differ from AIS patients not receiving thrombolytic therapy (n = 23). Moreover, in another study that investigated the dynamics of FXIII levels during the early phases of acute myocardial infarction in 350 patients, a significant, transient drop of FXIII levels was observed in the majority of patients during the first five days post-event 19 . Similarly to our findings, the probability of early death was increased in those having the highest drop in FXIII levels (below 59.5%). As these patients did not receive thrombolytic agents, this phenomenon was attributed solely to consumption of FXIII associated with the evolution of infarction. An additional mechanism for the decrease in FXIII levels observed in our study might be the proteolytic degradation of FXIIIa in the thrombus by plasmin or other proteases. Recently, it has been shown that activated FXIII, but not the zymogen form of FXIII is cleaved and inactivated by plasmin in vitro 20 . Less is known about this effect of plasmin in vivo. FXIII activity and antigen measurements performed in our study reflect circulating zymogen FXIII levels, which are not expected to be cleaved by plasmin. In fact, in our study FXIII activity and antigen levels measured before and immediately after thrombolysis were not significantly different, suggesting that plasmin has negligible effect on zymogen FXIII in vivo. On the other hand, incorporated FXIIIa might become degraded and inactivated by plasmin (and/or other proteases, e.g. proteases released from polymorphonuclear cells 21,22 ) in the thrombus, attributing to the significant decrease observed in FXIII levels at 24 h post-lysis. Given the fact that the ELISA method used for FXIII-A 2 B 2 determination in our study detects the intact FXIII tetramer complex only and no degradation products, we could not test the latter hypothesis and this must be investigated experimentally in future studies.
Using a logistic regression model including all potentially relevant risk factors, we found that a low FXIII level 24 h post-event is an independent predictor of short-term mortality (by 14 days). This result suggests that the reduction in FXIII levels directly relates to the pathomechanism of fatal stroke. This effect was found to be independent of the severity of stroke as measured by the NIHSS. Remarkably, those patients who died within the first week after stroke had unusually low FXIII levels the day after thrombolysis; with FXIII levels reaching only approximately 50% as compared to patients with better outcomes. Low FXIII levels as measured 24 h post-lysis were found to be associated not only with short-term but with long-term mortality (by the end of the 3 rd month post-event) as well. However, in the logistic regression model, low FXIII levels 24 h post-lysis did not prove to be  Table 7. Factor XIII (FXIII) polymorphisms, stroke severity and the outcome of thrombolytic therapy. FXIII-A, factor XIII A subunit; FXIII-B, factor XIII B subunit; OD, odds ratio; CI, confidence interval; NIHSS, National Institute of Health Stroke Scale; ICH, intracranial hemorrhage (asymptomatic and symptomatic intracranial hemorrhage); SICH, symptomatic intracranial hemorrhage; ΔNIHSS, difference in NIHSS score by day 7; mRS, modified Rankin score. Short-term and long-term outcomes were assessed at 7 days and by the end of the an independent predictor of long-term mortality. Post-stroke mortality at a long-term is influenced by a number of factors including age, co-morbidities, etc. which could explain the loss of significance in this case. Although it is biologically plausible, low FXIII levels were not associated with therapy-associated intracranial bleeding in our study, indicating that hemorrhagic complications must be associated with an etiology other than that linked to FXIII. This finding is in compliance with the few previously published reports 12,13 . FXIII levels were not associated with stroke severity in our patient cohort. As it has been found earlier 11 , we also showed that FXIII activity was significantly higher in case of atherothrombotic stroke as compared to strokes of cardioembolic origin.
In this study no association was found between any of the investigated major FXIII polymorphisms with stroke severity, unfavorable outcomes of therapy, therapy-associated symptomatic intracranial hemorrhage and mortality. In case of the FXIII-A p.Val34Leu polymorphism, an interesting trend was observed showing a protective effect against unfavorable short term outcomes in carriers of the Leu allele, but results did not reach statistical significance. No gender-specific differences with regard to the outcome of therapy were observed in case of any of the investigated polymorphisms (data not shown). Among the polymorphisms investigated in the study, only FXIII-B intron K nt29756 G allele was associated with lower FXIII levels, but after adjustment to confounders differences between carriers and non-carriers were not statistically significant. In a most recent study investigating patients with coronary sclerosis (CS) and/or myocardial infarction (MI), carriers of the FXIII-B intron K nt29756 G allele had significantly lower FXIII levels, and the presence of the allele provided significant protection against CS and MI in patients with fibrinogen in the upper tertile, which prevailed only in the presence of FXIII-A Leu34 allele 18 . In our study, the presence of FXIII-B intron K nt29756 G allele did not seem to have any impact on therapeutic outcomes, although due to the limited number of patients, it was impossible to perform subgroup analysis to seek for any synergistic effect between FXIII-A Leu34 and the FXIII-B intron K nt29756 G allele.
The study has limitations. Firstly, the study was designed to look for changes in FXIII levels during thrombolysis and to compare it with outcomes and was not designed to be a case-control study. Theoretically, we could compare FXIII levels in AIS patients receiving and not receiving thrombolysis. However, while the group of patients receiving rtPA is a highly selected patient group with strict inclusion criteria (e.g. time-frame from symptom onset to treatment, age limit, absence of effective anticoagulation, etc.), the group not receiving rtPA is a group not meeting the inclusion criteria by definition. The important baseline differences of the two groups and the heterogeneity of the group not receiving thrombolysis suggest that it might not be adequate to draw conclusions from the comparison of these groups.
Second, the sample size in this study is limited but representative for a single-center study. In fact, when comparing our study with previously published literature on this subject, our study included the largest cohort of patients [13][14][15][16] . Although the number of patients with symptomatic hemorrhage was small in this cohort, the lack of association of hemorrhage with FXIII levels was evident. It has to be noted, that although the single-centered design of the study has limitations regarding patient numbers, it gave us the advantage of assured unified sample handling, allowing reliable and reproducible results.
In conclusion, a low FXIII level 24 h after thrombolysis in AIS patients is an independent predictor of mortality by 14 days post-event. Testing FXIII levels 24 h post-lysis might help to identify patients who are at high risk of mortality and need altered therapeutic approach. Further studies including large number of patients are required to find out whether early selection of such patients could help to improve outcomes by providing better treatment strategies.

Methods
Study population. The study was designed as an observational study. Patients were enrolled between March 2011 and January 2013 in a single Stroke Center (Department of Neurology, University of Debrecen, Hungary). Study population included 132 consecutive patients admitted with AIS within 4.5 h of their symptom onset. All patients underwent intravenous (i.v.) thrombolytic therapy using recombinant tissue plasminogen activator (rtPA, Alteplase, Boehringer Ingelheim, Germany) according to the European Stroke Organization guidelines 5 . Patients or their relatives had been informed about the study and gave written informed consent. Ethical approval was obtained by the Ethics Committee of University of Debrecen, Debrecen, Hungary. All investigations and methods were performed in accordance with the relevant guidelines and regulations.
Blood sampling and laboratory measurements. Peripheral blood samples were drawn from patients into vacutainer tubes. Tubes with no anticoagulant were used for routine clinical chemistry tests, tubes anticoagulated with K 3 -EDTA for complete blood count, tubes containing 0.109 M sodium citrate or CTAD (buffered citrate, theophylline, adenosine and dypiridamole) for hemostasis tests (Becton Dickinson, Franklin Lakes, NJ). Blood samples were collected on three different occasions: upon hospital admission (before thrombolysis) (time point A), immediately after the administration of rtPA infusion (64 ± 10 min, time point B) and approximately 24 h after the administration of thrombolytic therapy (time point C). Routine blood count and chemistry tests were performed only from samples collected at time point A. Chemistry tests included the measurements of serum ions, glucose levels, basic kidney function tests, liver function tests and high sensivity C-reactive protein (CRP) measurements (Roche Diagnostics, Mannheim, Germany). Prothrombin time (PT) and activated partial thromboplastin time (APTT) were determined at time point A while thrombin time (TT) was measured on all three occasions by routine methods (Siemens AG, Erlangen, Germany). Plasma from CTAD tubes was separated by centrifugation (1500 g, 20 min) and samples were stored at −70 °C until further analysis.
Clinical data. Neurological deficit of patients was determined by the National Institute of Health Stroke Scale (NIHSS) score 27 on admission (before thrombolysis), 2 h, 24 h, and 7 days after thrombolytic therapy. Upon admission, all patients underwent computer tomography brain scan (CT) and computer tomography angiography (CTA) to ensure the diagnosis of acute ischemic stroke. Stroke etiology was classified according to the Trial of Org 10172 in Acute Stroke Treatment TOAST criteria 28 . A follow-up CT was performed for every patient 24 h after rtPA infusion. Hemorrhagic events on the follow-up CT scan were classified as symptomatic (SICH) or asymptomatic intracranial hemorrhage (aSICH) according to the European Cooperative Acute Stroke Study (ECASS) II criteria 13,29,30 . CT images taken before and 24 h after thrombolysis were analyzed simultaneously by 4 different investigators and the ASPECT scores were calculated 31,32 . For each patient the time of symptom onset, demographic and clinical characteristics (i.e.: cardiovascular risk factors, previous medication, neurological status, smoking habits) were recorded upon admission. Patients were followed and at 3 months after the event the modified Rankin score (mRS) 33 was determined.
The following outcomes were investigated: 1/short-term functional outcome (by day 7 post-event), 2/ long-term functional outcome (by the end of the 3 rd month post-event), 3/stroke-associated mortality by day 7, day 14 and by the end of the 3 rd month post-event, 4/the presence of therapy-associated symptomatic or asymptomatic intracranial bleeding (according to ECASS II criteria). Unfavorable short-term outcome was defined as an increase in NIHSS score by at least 4 points by day 7. Unfavorable long-term functional outcome was defined as an mRS score greater than 1 at 3 months post-event.
Statistical analysis. Statistical analysis was performed using the Statistical Package for Social Sciences (SPSS, Release 22.0, Chicago, IL). Distribution of continuous variables was tested by the Kolmogorov-Smirnov test. Parametric variables were expressed as mean ± SD, non-parametric variables were presented as median (interquartile range) unless otherwise stated. In case of two-group analyses Student's t-test or in case of non-parametric data Mann-Whitney U test was applied. Strength of association between variables was tested using Pearson's correlation test. Multiple linear regression analysis was performed to adjust for parameters independently associated with FXIII levels. ANOVA using the Bonferroni correction was applied for multiple comparisons. Differences between categorical variables were assessed by the χ 2 test. Logistic regression model was used to analyze the effect of FXIII levels and FXIII polymorphisms on various outcomes, results were expressed as odds ratio (OR) and 95% confidence interval (95% CI). When testing for independent risk factors of mortality, all potentially relevant risk factors (e.g. age, gender, stroke severity, smoking, CRP levels, NIHSS score on admission) were included in a multivariable backward conditional stepwise logistic regression model. Results were considered statistically significant when p < 0.05.