Role of a thrombin generation assay in the prediction of infection severity

Thrombin plays a central role in sepsis pathophysiology. The correlation of thrombin generation (TG) assays with infection severity and prognosis, and whether it can be used as a clinical tool, have been poorly explored and are the subjects of our research. We recruited 130 patients with systemic infection between 2016 and 2019. Patients were divided according to infection severity by using the sequential organ failure assessment (SOFA) and quickSOFA (qSOFA) scores. The hemostatic state was analyzed by Calibrated Automated Thrombogram. The primary end points were TG values and the secondary end point was in-hospital mortality. Patients with qSOFA ≥ 2 had a longer lag time (5.6 vs. 4.6 min) and time to peak (8 vs. 6.9 min) than those with lower scores (p = 0.014 and 0.01, respectively). SOFA ≥ 2 had a longer lag time (5.2 vs. 4.3 min), time to peak (7.5 vs. 6.7 min) and lower endogenous thrombin potential (ETP) (1834 vs. 2015 nM*min), p = 0.008, 0.019, and 0.048, respectively. Patients who died (11) had lower ETP (1648 vs. 1928 nM*min) and peak height (284 vs. 345 nM), p = 0.034 and 0.012, respectively. In conclusion TG assays may be a valuable tool in predicting infection severity and prognosis.

www.nature.com/scientificreports/ Study population. Eligible patients included patients ≥ 18 years old with newly diagnosed systemic infectious disease (in the preceding 24 h) and suspected sepsis who were willing to sign an informed consent form.
Patients who could not sign were evaluated and approved for recruitment by an unbiased physician not associated with the research. Patients with systemic infectious disease were suspected of sepsis if they had increased inflammatory markers (white blood cell count, C reactive protein), compromised hemodynamics measurements (systolic BP < 90 or diastolic BP < 60, tachycardia > 90 beats/minute, fever > 38 or < 36 °C, respiratory rate > 20, decreased mental state) or instant positive bacteremia. Patients were recruited from internal wards, the intensive care unit and emergency room department of the Emek Medical Center, in Afula, Israel. Exclusion criteria included patients known to have thrombophilia, coagulopathy, active oncological disease, those known to be current users of anticoagulation medication and pregnancy.
Thrombin generation. TG evaluation was conducted with the use of a CAT assay (Thrombinoscope B.V, Maastricht, the Netherlands) for analysis of platelet-poor plasma (PPP). Six milliliters of blood were collected in two tubes containing sodium citrate (3 ml, 9NC Coagulation sodium citrate 3.2%).
To obtain PPP the samples were centrifuged twice, first for 15 min at 2500 revolutions per minute (RPM) and then for 10 min at 2000 RPM. The plasma was stored at − 70 °C for later analysis. According to the manufacturer's instructions, measurements were conducted on 80 μl of PPP triggered by 20 μl PPP-reagent (4 μmol phospholipid and 5 pmol tissue factor). Measurements were calibrated with 20 μl Thrombin Calibrator, Fluorogenic substrate (20 μl) was added to sample. The thrombogram generated four parameters: lag time (in minutes), time to peak (TTpeak) (in minutes), peak height (nM) and endogenous thrombin potential (ETP) (in nM*minute), representing the area under the TG curve 6,7 . An example of CAT assay output is shown in Fig. 1. Infection severity scores. The SOFA score grades the dysfunction of six organ systems using laboratory variables (partial pressure of oxygen, fraction of inspired oxygen, platelets, bilirubin, creatinine) and clinical signs and symptoms (mechanical ventilation, Glasgow coma score, mean arterial pressure, use of vasoactive agents, urine output). The qSOFA score includes the following clinical criteria: respiratory rate ≥ 22/min, altered mentation, systolic blood pressure (BP) ≤ 100 mm Hg. For each patient, a qSOFA score (values between 0 and 3) and a SOFA score (0-24) were calculated (www. medca lc. com). Laboratory variables including complete blood count, chemistry tests, arterial blood gases and blood cultures were taken and analyzed by the Emek Medical Center laboratory within the first 24 h of infectious disease diagnosis. Clinical signs and symptoms were collected from electronic records.
End points. Primary: TG parameters-lag time, time to peak, peak height and ETP.
Secondary: in-hospital mortality.
Statistical analysis. The study population was divided into two groups by their SOFA and qSOFA scores (high vs. low). Cut off values were chosen reflecting high vs. low mortality risk 2 . Based on former research 8 , to see a mean difference in ETP of 300 nM*min between the low-and high-score groups with standard deviation (SD) of 480 for both groups and a 1:4 group ratio, we needed to recruit a total of 130 patients for a power of 80% and alpha value of 5% (two-sided test). Categorical variables (i.e., gender) were represented by frequency and percentage, and continuous variables (i.e., ETP) were represented by standard distribution measurement (i.e., mean, SD). Low and high SOFA and qSOFA scores were compared to CAT values by using a t-test. For categorical variables, a chi-square test or Fisher test were used. The associations between categorical or continuous variables and each of the CAT values were examined with multivariate general linear model. The multivariate analysis was performed with the stepwise algorithm (the final model included only variables with P < 0.05). In addition, the correlation between the qSOFA and SOFA scores (as an ordinal variable) and CAT values was tested by using Spearman correlation. The accuracy of SOFA, qSOFA and CAT values were assessed by comparing receiver operating characteristic (ROC) curves using the test proposed by DeLong et al. 9 . Statistical analysis was conducted by SAS 9.4 software, and statistical significance was considered when p < 0.05.

Results
Between 2016 and 2019, one hundred and thirty patients with systemic infection were recruited. Eight of these patients did not have a fully calculated qSOFA or SOFA scores, or CAT values due to technical problems and were dropped out. Demographics and comorbidities varied; patients with higher SOFA/qSOFA scores were older and sicker (with higher rates of stroke, hypertension, and heart failure). Higher SOFA scores were noted in men and in patients with chronic renal failure (Table 1).
Patients with severe infection, as determined by qSOFA ≥ 2, had a longer lag time (5.6 vs. 4.6 min) and TTpeak (8 vs. 6.9 min) than those with lower qSOFA scores, p = 0.014 and p = 0.01, respectively ( Table 2). These significant differences were maintained after adjustment for risk factors and co-morbidities (Table 3). Furthermore, the qSOFA score showed a positive correlation of 0.23 with lag time and 0.24 with TTpeak, p = 0.01 and 0.008, respectively. A negative correlation was observed between ETP and qSOFA (− 0.13) and peak height and qSOFA (− 0.17), p = 0.16 and 0.06 respectively (Table 4).
With regard to infection severity, as determined by the SOFA score, patients with higher SOFA scores had a longer lag time (5.2 vs. 4.3 min), lower ETP value (1834 vs. 2015 nM*min) and longer TTpeak (7.5 vs. 6.7 min), with p = 0.008, 0.048, and 0.019, respectively (Table 2). After adjustment for risk factors and co-morbidities, lag time difference maintained statistical significance (Table 3). A positive correlation between lag time and  Table 4). The predictive abilities of TG parameters and the high qSOFA and SOFA scores were investigated using ROC curves and are presented in Fig. 2. For qSOFA, the ROC curve showed an area under the curve (AUC) of 64% for lag time and 65% for TTpeak. For SOFA ≥ 2, the ROC curve showed an AUC of 65% for lag time, 63% for ETP and 63% for TTpeak.
The ROC curves for TG and mortality showed an AUC of 67.9% for ETP, 73.4% for peak height, 90.1% for qSOFA and 86.4% for SOFA (Fig. 3). Combining the SOFA score and TG values did not result in a significantly better predictive ability, with an AUC of 0.868 for SOFA and ETP, and 0.878 for SOFA and peak height.

Discussion
In our study, which involved patients with systemic infection, the results demonstrated that in both infection severity scores, qSOFA and SOFA, the clotting time values (lag time and TTpeak) were longer as infection severity increased. Additionally, the clotting time values exhibited relatively good discriminative abilities for infection severity with area under the ROC curve of approximately 65%. A trend of an inverse relationship between both peak height and ETP values with infection severity was noticed in both scores. Overall, we concluded that as infection severity increases, the potential ability for TG production in-vitro decreases. Although systemic infection in general, and sepsis specifically, are considered to be a hypercoagulative states, past research results are inconclusive. Similar to our results, some other studies have also correlated sepsis with a decreased potential ability to generate thrombin. Mihajlovic 11 . Other, former observations have also shown reduced TG values in patients with sepsis in general and DIC in particular, yet these interesting studies did not include patients from the whole sepsis spectrum and compared values to healthy controls, not distributing according to sepsis severity [12][13][14] . Other studies published conflicting results [15][16][17] .
With regards to TG and mortality, patients with lower TG values had higher mortality rates. Most noticeable are the ETP and peak height values, which portray a good discriminatively ability. Consistent with our results, studies by Seo et al. and Massion et al. correlated low TG values with mortality in patients with sepsis 12,18 .
Thus, our findings strengthen past observations and, perhaps, resolve conflicting and confusing results, showing a correlation between increased infection severity and decreased potential ability to generate thrombin. Uniquely, our study evaluated infection severity by means of validated, routinely used scores, namely, SOFA and qSOFA, which their association with TG have not been examined thoroughly, if at all, to date. These results could have two possible explanations. As infection severity increases, a hypocoagulative state is formed. Or, conversely, Table 3. Thrombin generation and qSOFA/SOFA scores, adjustment for confounders. β; the standardized regression coefficients, ETP (endogenous thrombin potential). *Adjusted for age, Sex, BMI, diabetes mellitus, ischemic heart disease, stroke/TIA, COPD, hypertension, heart failure, chronic kidney disease. www.nature.com/scientificreports/ and perhaps more reasonably, as infection severity increases extremely, it causes the consumption of clotting factors, which in turn, leads to low TG values. These results imply that TG assays, particularly CAT, could be an effective and easily accessible tool in hospital settings to assess the severity and prognosis in these complex patients. Data concerning reference values is scarce and has been gathered only recently 19 , further studies are needed to determine reference values and to explore the possibility of TG assays in impacting and guiding the decision-making process in these complex patients and complex scenarios.
Our study had some limitations. Most of our patients had low qSOFA and SOFA scores. Perhaps the results would have been more conclusive if a greater number of patients had had higher score representations. The low mortality rate raises concern about the power of this patient population to draw conclusions. Moreover, we calculated qSOFA and SOFA after recruitment, therefore a formal diagnosis of sepsis could not have been done during recruitment. However, this enrollment allows for studying patients from the whole infection severity spectrum. Lastly, there is very little data regarding reference values for the CAT at the moment, thus we encourage further research in this field.  . ROC receiver operating characteristic. ETP endogenous thrombin potential. ttPeak time to peak. SOFA sequential organ failure assessment. qSOFA quick sequential organ failure assessment. AUC, area under the curve; 95%CI 95% confidence interval. SOFA sequential organ failure assessment.

Conclusion
TG assays may be a valuable tool in predicting infection severity and prognosis. Moreover, TG values suggest that, in severe infection, there is a reduced potential of generating thrombin.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.