Association of Thromboelastography With Hematoma Expansion and Disability in Spontaneous Intracerebral Hemorrhage

Abstract

Background and Objectives

Hematoma expansion (HE), an increase in bleeding into the brain, is a modifiable cause of disability and death after intracerebral hemorrhage (ICH). Some specialized hemostatic biomarkers are associated with HE. Widely available hemostatic biomarkers could improve our understanding of HE and lead to targeted treatments.

Methods

We conducted a multicenter, prospective, cohort study of patients with spontaneous ICH between 2019 and 2023 at 6 medical centers across the United States. Patients with an initial CT scan within 12 hours of symptom onset were enrolled. All patients had initial and follow-up CT scans for HE calculation and a blood draw that occurred before follow-up imaging. Patients who received desmopressin were excluded from analysis. Correlation between continuous numerical variables (i.e., HE and hemostatic biomarkers) was calculated with Spearman correlation. Correlations between the 3-month modified Rankin Scale (mRS) and biomarkers were calculated using Kendall correlation.

Results

We enrolled 82 patients with spontaneous ICH (34.1% women) with a mean age of 61 ± 13.3 years. Symptom onset occurred within a median time of 1.8 (1.15–4.48) hours before first CT scan imaging and 5.8 (4.45–7.57) hours before blood draw. Seventeen (21%) used aspirin, 4 (5%) used clopidogrel, and none used anticoagulants before ICH. The Thromboelastography (TEG) Coagulation Index, a composite calculation from individual parameters, was correlated with subsequent HE (r = −0.26, p = 0.01, 90% Conf. Int. [−1.0 to −0.11]). TEG K, a measure of fibrinogen-dependent clot strength, also correlated with HE (rho = 0.19, p = 0.048, 90% Conf. Int. [0.01–37.0]). TEG Maximum Amplitude (tau = −0.32, 90% Conf. Int [–0.57 to 0.0], p = 0.02), a measure of clot strength dependent on platelets, and HE (tau = 0.35, 90% Conf Int [0.1–0.6], p = 0.01) were associated with the 3-month mRS.

Discussion

Hemostatic biomarkers from TEG, particularly those related to clot strength and fibrin generation, were associated with subsequent HE and the mRS at 3 months. These results suggest that hemostatic biomarkers have potential to predict HE and functional outcomes in a broad swath of patients with spontaneous ICH.

Introduction

Intracerebral hemorrhage (ICH) is a deadly form of stroke.¹ Hematoma expansion (HE) is characterized by an increase in volume of the hematoma on interval CT scans and is a modifiable predictor of patient outcomes.^2-4 Most HE occurs during the first few hours after ICH symptom onset. More HE predicts worse patient outcomes.² Acute treatment to reduce the likelihood of HE has been the focus of several clinical trials, with varying degrees of success.^5-10 Additional methods are needed to identify patients most likely to have subsequent HE, who are most likely to benefit from hemostatic treatment.

Several hemostatic biomarkers predict HE and identify specific mechanisms of hemostasis for therapeutic intervention.¹¹ The international normalized ratio (INR) is a hemostatic biomarker that predicts more HE in patients known to take warfarin. Faster normalization of the INR predicts less HE and likely less death.¹² However, the INR is limited in scope to patients taking warfarin or with hepatic failure. Other hemostatic biomarkers may be useful in a broader swath of patients. Thromboelastography (TEG) is a non-invasive, point-of-care test that quantifies viscoelastic characteristics of whole blood samples. TEG has been reported to predict later in a larger, multicenter cohort, potentially identifying patients for targeted treatment.

Methods

Study Design

We conducted an observational cohort study of patients with spontaneous ICH between 2019 and 2023. The study was designed with 80% power to detect small effect sizes (r = 0.2). Patient recruitment was initially delayed by the COVID-19 pandemic and bounded by the end date of the award.

Institutions

We consecutively enrolled patients at 6 medical centers across the United States. Northwestern Memorial Hospital is a Comprehensive Stroke Center in downtown Chicago, IL, and recruited 12 patients. Northwestern Medical Central DuPage Hospital (4 patients) is a comprehensive stroke center in Winfield, IL, about 40 miles west of Chicago. Rush University Medical Center (1 patient) is in west Chicago, IL. University of Chicago Medical Center (13 patients) is a referral center for acute neurologic emergencies on Chicago’s south side. The University of Iowa (21 patients) is in Iowa City and is the only academic medical center in the state. Memorial Hermann Hospital (31 patients) is in Houston, TX, with a citywide catchment. Both Central DuPage and Memorial Hermann also have mobile stroke units.

Study Population

Patients with spontaneous ICH were consecutively recruited. Patient ages ranged from 35 to 90 years. Patients with a history of malignancy, vascular malformation on imaging (e.g., aneurysm), sepsis (e.g., concern for septic emboli), hemoglobin <7.5 g/dL (because further blood draws might precipitate packed red blood cell transfusion), or a condition other than ICH that was likely to be disabling or fatal (e.g., metastatic cancer, hepatic liver failure) were excluded. In addition, pregnant women were excluded due to potential anemia for this subgroup.

In this report, we excluded patients who underwent neurosurgery before a follow-up CT scan because it precludes measuring HE. We excluded patients who received an intervention to reduce the likelihood of HE (e.g., desmopressin) before blood draw because interventions alter hemostatic biomarkers and, potentially, HE.¹³

Hemostatic Biomarkers

Whole blood samples were collected. Contextual information (e.g., use of antiplatelet medication before ICH) was prospectively recorded in a database. The TEG (Haemonetics Co., Braintree, MA) Hemostasis Analyzer System was used to measure global hemostasis of blood samples. TEG is a non-invasive, point-of-care assay that assesses coagulation factors, fibrinogen, platelets, and fibrinolysis in a single wave form.² The device measures clot viscoelasticity using a resonance method, in which the blood sample is exposed to a vibration frequency and resonance of the clot is converted to the clot strength readout.

The TEG Global Hemostasis cartridge provides 4 main parameters: Reaction time (R time), Kinetic time (K time), α angle, and Maximum Amplitude (MA). The R time accounts for the time to initial fibrin formation and reflects coagulation factor activity. The K time calculates the amount of time to reach clot strength and reflects the amplification phase of coagulation that is dependent primarily on fibrinogen with some platelet contribution. The α angle is the slope of the line between R and K and reflects the rate of fibrin accumulation and cross-linking during the propagation phase. MA represents the strength and stability of fibrin clot.

We used a summary score of TEG, the Coagulation Index (CI), previously used in the context of intracranial hemorrhage, ischemic hemorrhage, and trauma.^14-17 The TEG CI is a mathematical construct that combines various TEG parameters involving fibrinolysis, platelet activity, and activation of coagulation, underscoring that the CI is a summary of a complex hemostatic system. The CI was calculated from 4 individual TEG parameters using the following equation:

$$\begin{gathered} CoagulationIndex= \\ 0.1227(R)+0.0092(K)+0.1655(MA)-0.02421(\alpha )-5.0220 \end{gathered}$$

Normal ranges for the TEG CI occur between −3 and +3, where values above +3 are considered “hypercoagulable.”^14-17 Additional hemostasis measurements included the Clauss Fibrinogen assay was used to measure fibrinogen concentration (mg/dL).

Hematoma Volumes

AnalyzeDirect (Overland Park, KS) was used to measure hematoma volumes for initial and follow-up CTs. The hematoma is characterized by density (30–80 Hounsfield units) relative to the surrounding, less-dense brain tissue. HE was calculated as the absolute difference in hematoma volume between interval CT scans.

Patient Outcomes

Disability or death at 3-month follow-up was determined by the modified Rankin Scale (mRS) using a validated questionnaire.^6,17

Study Variables

The primary response variables of this study were volume of HE (mL) and mRS at the 3-month follow-up. We collected demographic information (age, sex, race, ethnicity) and hemostatic biomarkers (R, K, MA, α angle, the summary TEG CI, and fibrinogen concentration). The following clinical variables were also recorded: pre-ICH medication history, systolic blood pressure at admission, diabetes status, hypertension history, ICH location, INR, initial and final hematoma volume, Glascow Coma Scale (GCS) score on admission, NIH Stroke Scales (NIHSSs) score on admission, and the ICH score.

Statistical Analysis

Normally distributed data are presented as mean ± SD, and non-normally distributed data as median (Q1–Q3). Group means were compared with the Wilcoxon rank-sum test. All correlation tests were one-sided (α = 0.05), which corresponds to a 90% twosided CI (1–2α). Correlation between continuous HE and the normalized TEG CI were calculated with Pearson correlation. Correlations between continuous numerical variables (HE and hemostatic biomarkers) were calculated using Spearman rank correlation. A Jackknife Euclidean–based resampling method was used to calculate Spearman rho and the 90% CI with the SpearmanCI package in R studio.¹⁸ This method provides CIs but does not define p values. To complement these likelihood-based intervals, we used a paired bootstrap procedure in which subjects were resampled with replacement and Spearman rho was recalculated for each bootstrap sample for significance calculations. Similarly, we used resampling techniques to calculate correlations between the mRS and hemostatic biomarkers using Kendall’s Tau-B for ordinal variables. To prevent inconsistencies due to ties and the small sample size of mRS results (n = 25), we obtained a one-sided permutation p value (10,000 permutations under H₀: τ = 0) and a 90% bootstrap CI. Univariate linear regression analyses were conducted with each study variable and continuous, absolute HE. Significant variables, or p < 0.1, were included in multivariable regression models with HE to test independent association. The final multivariable linear regression model assessed the association between HE and TEG CI or the individual TEG parameters with ICH score as a confounding variable. Post hoc sensitivity analysis was conducted in R using the “pwr” package with a significance level of 0.05 and sample size of 82 for each Pearson or Spearman hypothesis test.¹⁹

Standard Protocol Approvals, Registrations, and Patient Consents

This work was supported by the National Institute of Neurologic Disorders and Stroke, National Institutes of Health (NIH), under award numbers R01NS110779 and R01NS117608. This manuscript is the result of funding in whole or in part by the NIH. It is subject to the NIH Public Access Policy. Through acceptance of this federal funding, NIH has been given a right to make this manuscript publicly available in PubMed Central upon the Official Date of Publication, as defined by NIH. All data collection was approved by the Northwestern University Institutional Review Board (IRB ID: STU00208910). The study followed Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. All participants were at least 21 years with signed consent from the patient or a legally authorized representative.

Data Availability

All data will be shared per NIH regulations, and investigators may obtain access to deidentified data by request.

Results

Initially, 140 patients with spontaneous ICH were screened. Of those, 82 patients met the requirements for statistical analysis (Figure). There was no significant difference in study variables for those excluded from statistical analysis. The median hours between ICH symptom onset and first CT scan were 1.8 (1.15–4.48) hours and 5.8 (4.45–7.57) hours from ICH symptom onset to blood draw. The median initial hematoma volume was 12.3 (6.7–27.9) mL with a median HE of 1.5 (0.4–3.9) mL. Demographics, medical history, and clinical assessment results at admission are presented in Table 1. Seventeen (21%) patients used aspirin, 4 (5%) used clopidogrel, and none used anticoagulants before ICH.

Figure.

Flowchart of Spontaneous ICH Patients Meeting the Requirements to be Included in the Statistical Analysis

Table 1: Demographics, Medical History, and Baseline Clinical Variables for the Spontaneous Intracerebral Hemorrhage Cohort.

Demographic information includes age, race, ethnicity, and sex for the spontaneous ICH cohort (n = 82). Relevant medical history was recorded including hypertension history and pre-ICH antiplatelet medication use, including aspirin, non-steroidal anti-inflammatories (NSAID), and Clopidogrel. Baseline clinical variables were recorded upon hospital admission including systolic blood pressure (BP), location of ICH in the brain, and the internal normalized ratio (INR). Results are presented by number of patients and percentage (N (%)), mean ± standard deviation (SD), or median and interquartile range [Q1 – Q3] as appropriate.

Variable	N (%), mean ± SD or median [Q1 – Q3]
Age (years)	61 ± 13.3
Race
White	54 (65.7%)
Black or African American	21 (25.6%)
Asian	3 (3.7%)
American Indian/Native Alaskan	1 (1.2%)
Other/Unknown	3 (3.6%)
Ethnicity
Hispanic	16 (19.5%)
Women	28 (34.1%)
History of Hypertension	55 (67.0%)
Diabetes	15 (18.3%)
Aspirin Dose
0	65 (79.3%)
81 mg/day	13 (15.9%)
>81 mg/day	4 (4.9%)
NSAID Use	7 (8.5%)
Clopidogrel Use	4 (4.9%)
Systolic BP on Admit, mm Hg	171 ± 41.2
ICH Location
Brainstem	7 (8.5%)
Caudate	3 (3.7%)
Cerebellar	2 (2.4%)
Lentiform Nucleus	6 (7.3%)
Lobar	21 (25.7%)
Thalamus	14 (17.1%)
Other / Indeterminablea	29 (35.4%)
International Normalized Ratio	1.0 [0.97 – 1.1]
Median time (hours) between…
Symptom onset and follow-up CT	16.4 [9.67 – 26.83]
Initial and follow-up CT	13.2 [7.28 – 22.71]
Hematoma Volume (mL)
Initial	12.3 [6.7 – 27.9]
Hematoma Expansion	1.5 [0.4 – 3.9]

^aMay span more than one location without a clear origin

TEG results (R, K, MA, an angle, and summary CI) for the cohort were generally normal (Table 2). Thirty (39.5%) patients had a TEG CI in the hypercoagulable range >3. No patient was hypocoagulable (i.e., TEG CI < −3).

Table 2: Median Results for Hemostatic Biomarkers of Spontaneous ICH Cohort.

Median and interquartile range (IQR) for the hemostatic biomarkers for spontaneous ICH patients. The expected normal range for each variable are listed in the far-right column.

Variable	Median [IQR]	Normal Range
Thromboelastography
R (min)	4.4 [3.2 - 5.7]	3.8 - 9.8
K (min)	1.3 [1.0 - 1.6]	1.1 - 3.5
a angle (degrees)	72 [68.1 - 75.6]	48 - 78
MA (mm)	64 [61.1 - 67.2]	55 - 73
Coagulation Index	2.8 [2.2 - 3.6]	−3.0 - 3.0
Clauss Fibrinogen (mg/dL)	360 [297 – 420]	200 - 400
Admit NIHSS	14 [6-18]	0: No symptoms 1-4: Minor Stroke 5-15: Moderate Stroke 16- 20: Moderate to severe stroke 21-42: Severe Stroke
Glascow Coma Scale	14 [11-15]	3-8: Severe Stroke 9-12: Moderate Stroke 13-15: Mild Stroke
ICH Score	1 [0 – 2]	30-day mortality 0 points: 0% 1 point: 13% 2 points: 26% 3 points: 72% 4 points: 97% 5 points: 100% 6 points: 100%

TEG CI was correlated with subsequent HE (r = −0.26, p = 0.01, 90% Conf. Int. [−1.0 to −0.11], Power = 0.77). The association of the TEG CI with HE was driven by K (rho = 0.19, p = 0.048, 90% Conf. Int. [0.01–37.0], Power = 0.53). Because HE is associated with disability outcomes at 3 months (tau = 0.35, 90% Conf Int [0.1–0.6], p = 0.01, Power = 0.53), we tested if individual components of TEG were associated with the mRS at 3 months (Table 3). In addition to its association with HE, K was positively associated with the mRS, although we present this association cautiously. There was a similar negative association between TEG MA and the mRS. We looked for typical confounders of the associated TEG with later HE and outcomes. INR was normal in the cohort 1.0 (0.97–1.1), and no patients in this cohort took anticoagulants before ICH. Fibrinogen levels were normal (Table 2). Neither INR, fibrinogen, nor GCS were correlated with HE or the mRS at follow-up. The NIHSS on admission was moderately associated with HE (rho = 0.32, 90% Conf. Int [0.13–0.49], p = 0.03, Power = 0.90) and the mRS at 3 months (tau = 0.28, 90% Conf. Int [0.0–0.56], p = 0.047, Power = 0.38). ICH score was moderately associated with subsequent HE (rho = 0.22, 90% Conf. Int [0.03–0.40], p = 0.03, Power = 0.64), but not the mRS at 3 months (Table 3).

Table 3: Hemostatic Biomarker Correlation with the modified Rankin Score.

Hematoma expansion, hemostasis biomarker, and standardized stroke scores correlations with disability at 3-month follow-up measured by the modified Rankin score (mRS). Kendall’s Tau-B correlation coefficients were calculated with one-sided permutation p-value (10,000 permutations under H₀ : = 0) and a 90% bootstrap confidence interval.

Hemostatic Biomarkers		τ 90% Confidence Interval)	P - value
Hematoma Expansion (mL)		0.35 (0.1, 0.6)	0.01
Thromboelastography
R (min)	Time to initial fibrin formation; reflects coagulation factor activity	−0.13 (−0.4, 0.2)	0.2
K (min)	Time to reach clot strength; reflects the amplification phase of coagulation dependent primarily on fibrinogen	0.28 (0.0, 0.56)	0.04
α angle (degrees)	Slope of the line between R and K; reflects the rate of fibrin accumulation	−0.06 (−0.35, 0.2)	0.3
MA (mm)	Highest point of TEG curve; reflects the strength of clot dependent on platelets	−0.32 (−0.57, 0.0)	0.02
Coagulation Index	Mathematical summary of R, K, α, and MA; reflects overall coagulation status	−0.21 (−0.47, 0.1)	0.1
Clauss Fibrinogen (mg/dL)		−0.17 (−0.76, 0.4)	0.3
Admit NIHSS		0.28 (0.0, 0.56)	0.047
Admit GCS		−0.09 (−0.37, 0.22)	0.7
ICH Score		0.19 (−0.11, 0.48)	0.15

Individual TEG components were not associated with HE in a multivariable regression model with ICH Score; however, the composite TEG CI had significant association with HE (β = −1.6±0.70, p = 0.03), whereas ICH score did not (β = 0.99±1.1, p = 0.36) when forced into the same regression model. The overall model was not statistically significant (R² = 0.04, F (8.8, 72) = 2.6, p = 0.08).

Discussion

In a multicenter study in 3 different metropolitan areas, we found that hemostatic biomarkers of clot strength and platelet-mediated fibrin generation were associated with subsequent HE and functional disability at 3 months after ICH. Findings were similar across hemostatic biomarkers. There was neither hypofibrinogenemia nor pre-ICH anticoagulant use, additional confounders of the association between hemostatic biomarkers and HE. These results suggest hemostatic biomarkers are associated with subsequent HE and outcomes in a broad swath of patients with spontaneous ICH, even when the results are in the normal range.

These results add to previous work showing hemostatic biomarkers predict HE and later disability.^11,15,20 These results imply treatments that promote activation of the coagulation cascade to generate fibrin would be more likely to reduce HE and improve functional outcomes in selected patients.⁵ However, this is speculative, absent hemostatic biomarker data from clinical trials.

Hemostatic biomarkers in this cohort were generally within the normal range, despite the associations with HE and subsequent functional outcomes. No patient had low fibrinogen or a “hypocoagulable” TEG CI. The composite TEG CI was associated with HE in a multivariable regression model controlling for ICH score. ICH score is a single variable calculated from multiple predictors of HE (age, initial hematoma volume, intraventricular hemorrhage status, and infratentorial origin) and was chosen as a confounding variable to prevent overfitting of the model with a relatively small sample size.

Clot formation at the lower side of the normal range may be insufficient to mount an adequate response in the setting of acute ICH to minimize HE. HE typically occurs within the first several hours after ICH symptom onset. However, not every patient identified within a few hours of symptom onset will have HE. In the Stopping Intracerebral Haemorrhage with Tranexamic Acid for Hyperacute Onset Presentation including Mobile Stroke Units (STOP-MSU) trial, fewer than half of patients had HE despite a 2-hour time window from ICH symptom onset.⁶ Rapid hemostatic biomarkers are a potential method to identify patients most likely to have subsequent HE, even within stringent time windows for treatment.

Pre-ICH anticoagulation medications change hemostatic biomarkers (e.g., the INR) and predict more subsequent HE.¹² We did not have patients on anticoagulation in our cohort and thus cannot comment on the use of the TEG to identify patients on pre-ICH anticoagulants, although this is intriguing. This cohort had too few patients on pre-ICH antiplatelet medications for a meaningful analysis. We have previously noted associations between pre-ICH antiplatelet medications and specific biomarkers of platelet activity.^20-22 Compared with previous work, these hemostatic biomarkers are less specific. The coagulation cascade is sufficiently complex that no single measure is sufficient to capture all potential predictors of later HE in every context. A panel of hemostatic biomarkers might be more appropriate and could be evaluated in future research.

There are potential limitations to this work. We measured HE, which requires 2 CT scans, therefore, patients without follow-up CT scans due to neurosurgery or death were not included in this investigation. A strength of this study was the ability to quickly identify ICH patients with a median time between symptom onset and first CT scan of less than 2 hours. Blood draws did not occur until a few hours later, potentially limiting our ability to detect HE, despite finding significant results. We attribute that delay to obtaining informed consent for the extra blood draws and study procedures. We report a relatively small sample size, particularly for the mRS due to loss of participation at the 3-month follow-up, and we were unable to perform a multivariable logistic regression to test confounding variables for the association between TEG and disability outcomes. Recruitment was hampered by the COVID-19 pandemic, which broadly reduced stroke admission in addition to a hold on recruitment for several months.²³ In addition, a significant number of patients were excluded. This resulted in slightly underpowered calculations for small effect sizes that may limit the generalizability of our findings; however, previous literature using TEG from a single center¹¹ and biological plausibility support our results. A major strength is this multicenter study was conducted at 6 medical centers across the United States that encompassed a variety of patient populations, suggesting that our results are likely to be generalizable.

In summary, we found that hemostatic biomarkers of coagulation, particularly related to clot strength and fibrin generation, were associated with subsequent HE and functional outcomes at follow-up. Future investigations might use hemostatic biomarkers to select patients for hemostatic treatment to reduce HE and improve patient outcomes.

Glossary

GCS

Glascow Coma Scale

HE

Hematoma Expansion

ICH

intracerebral hemorrhage

INR

international normalized ratio

MA

Maximum Amplitude

mRS

modified Rankin Scale

NIHSS

NIH Stroke Scale