Noncontrast Computed Tomography Markers as Predictors of Revised Hematoma Expansion in Acute Intracerebral Hemorrhage.

Background Noncontrast computed tomography (NCCT) markers are the emerging predictors of hematoma expansion in intracerebral hemorrhage. However, the relationship between NCCT markers and the dynamic change of hematoma in parenchymal tissues and the ventricular system remains unclear. Methods and Results We included 314 consecutive patients with intracerebral hemorrhage admitted to our hospital from July 2011 to May 2017. The intracerebral hemorrhage volumes and intraventricular hemorrhage (IVH) volumes were measured using a semiautomated, computer-assisted technique. Revised hematoma expansion (RHE) was defined by incorporating the original definition of hematoma expansion into IVH growth. Receiver operating characteristic curve analysis was used to compare the performance of the NCCT markers in predicting the IVH growth and RHE. Of 314 patients in our study, 61 (19.4%) had IVH growth and 93 (23.9%) had RHE. After adjustment for potential confounding variables, blend sign, black hole sign, island sign, and expansion-prone hematoma could independently predict IVH growth and RHE in the multivariate logistic regression analysis. Expansion-prone hematoma had a higher predictive performance of RHE than any single marker. The diagnostic accuracy of RHE in predicting poor prognosis was significantly higher than that of hematoma expansion. Conclusions The NCCT markers are independently associated with IVH growth and RHE. Furthermore, the expansion-prone hematoma has a higher predictive accuracy for prediction of RHE and poor outcome than any single NCCT marker. These findings may assist in risk stratification of NCCT signs for predicting active bleeding.

Antihypertensive Treatment of Acute Cerebral Hemorrhage 2

expansion‐prone hematoma

noncontrast computed tomography

revised hematoma expansion

Clinical Perspective

What Is New?

We have investigated whether the noncontrast computed tomography markers could predict the intraventricular hemorrhage growth and revised hematoma expansion criteria, which has not been reported in previous studies.

We tested the diagnostic performance of noncontrast computed tomography markers in predicting intraventricular hemorrhage growth and revised hematoma expansion criteria, and found that noncontrast computed tomography markers are independently associated with intraventricular hemorrhage growth and revised hematoma expansion.

The expansion‐prone hematoma has a higher predictive accuracy for predicting revised hematoma expansion and poor outcome than any single noncontrast computed tomography marker.

What Are the Clinical Implications?

Our findings may assist in the risk stratification of active bleeding, and be helpful for providing more accurate prognostic information for clinical decision‐making.

Our results may help clinicians to select patients for antiexpansion treatment in future clinical trials.

Intracerebral hemorrhage (ICH) accounts for 8% to 27% of all strokes and leads to high mortality and morbidity worldwide. The mortality of patients with ICH ranges from 35% at 7 days to 59% at 1 year, and >60% of survivors are left with severe functional disability. Hematoma expansion (HE) occurs in ≈30% of patients with ICH , and is considered as a potentially modifiable predictor target for antiexpansion treatment in many clinical trials. , , , Although several trials have curbed the growth of hematoma, the outcomes of patients have not been improved accordingly. , ,

More recently, many investigators have shifted their focus to the dynamic changes of intraventricular hemorrhage (IVH). Delayed IVH, which accounted for 8% to 10% of ICH, is an independent predictor of poor outcome. , IVH volume of >2 mL could also independently predict the poor outcome of ICH. Some investigators reported that IVH volume growth as small as 1 mL could be an optimal threshold for predicting unfavorable outcomes. Li et al have found that combined delayed IVH and IVH expansion of >1 mL is closely correlated with poor outcome. Therefore, the conventional definition of HE may be one part of the active bleeding of ICH. Extending the current HE definition to the dynamic changes of hematoma volume in intraparenchymal tissues and ventricular system may be necessary. Furthermore, it is important to predict the dynamic process of ICH for antiexpansion treatment.

The computed tomographic angiography spot sign is a promising indicator to stratify the risk of HE and poor outcome. , , Dowlatshahi et al have innovatively discussed the association of computed tomographic angiography spot sign with revised HE (RHE). However, seeing that the computed tomographic angiography spot sign is not widely available in most hospitals, the association of the noncontrast computed tomography (NCCT) markers may be the alternatives to the computed tomographic angiography spot sign to establish simple models for predicting the HE. , , , , Meantime, whether NCCT markers could predict IVH growth and RHE remains unclear.

In this study, we aim to investigate whether the NCCT markers could predict the IVH growth and RHE criteria. We further tested this diagnostic performance of NCCT markers in predicting IVH growth and RHE criteria.

Methods

Study Population

All patients with ICH admitted to the First Affiliated Hospital of Chongqing Medical University were included in our ongoing prospective study. More than 1000 patients with acute stroke were treated in this tertiary referral hospital over a year period. Patients aged >18 years were enrolled from July 2011 to May 2017. Patients with an axial computed tomography (CT) scan performed within 6 hours after ICH onset and a follow‐up CT scan performed within 36 hours after the initial CT scan were included. We excluded patients with anticoagulant‐associated ICH, primary IVH, hemorrhagic transformation after cerebral infarction, and secondary ICH as a result of tumor or trauma.

Standard Protocol Approval and Patient Consent

All study procedures and protocols involving human participants comply with the ethical standards of the Declaration of Helsinki. The study was approved by the Ethics Committee of the First Affiliated Hospital of Chongqing Medical University. All patients (or their legal representatives) provided written informed consent.

Clinical Data Collection and Image Analysis

The clinical and imaging data, including age, sex, medical history, prior medication use, admission Glasgow Coma Scale score, time from ICH onset to initial CT scans, and baseline blood pressure (BP), were recorded. Modified Rankin Scale (mRS) score was assessed at 90 days through telephone by trained neurologists. All CT scans were performed without intravenous contrast injection. All CT images were saved as Digital Imaging and Communications in Medicine format and further reviewed independently by 2 experienced readers (Q.L. and W.S.Y.) who were blinded to clinical and outcome data.

Hematoma location was classified as basal ganglia, thalamus, lobar, and infratentorial. The ICH volumes and IVH volumes were measured using a semiautomated computer‐assisted software (Mimics Software, version 20.0; Materialise NV, Leuven, Belgium). Briefly, predefined maximum and minimum Hounsfield units were set for building a mask. The segmentation accuracy of ICH and IVH hematomas was confirmed by a region growing algorithm with a visual inspection. We manually segmented the ICH hematomas and IVH hematomas if they were connected.

HE was defined as relative intraparenchymal hematoma growth >33% or absolute hematoma growth of >6 mL of the baseline hematoma volume. IVH growth was defined as either a newly occurring IVH on follow‐up CT without the presence of IVH on initial CT (delayed IVH) or an absolute growth of IVH volume >1 mL from initial CT scan to follow‐up CT scan, as previously described. , We defined the RHE criteria as HE or IVH growth, including 4 types, as follows: ≥6 mL or >33% or IVH expansion ≥1 mL or delayed IVH. Active bleeding refers to the definition of HE, IVH growth, or RHE.

Several NCCT markers, including blend sign, black hole sign, and island sign, were defined, as previously described (Figure S1). , , In brief, we defined the NCCT blend sign as follows: (1) blending of the relatively hyperattenuating area with an adjacent hypoattenuating region within a hematoma with an easily recognized border; (2) there was at least an 18–Hounsfield unit difference between the 2 density area in a hematoma, and the relatively high‐density area cannot encapsulate the low‐density area. The NCCT black hole sign was defined as the low‐density area wrapped in the high‐density area with an identifiable border, and the difference between the density regions was >28 Hounsfield units. The NCCT island sign was defined as >3 scattered small round or oval hematomas separated from the main hematoma or >4 small hematoma parts or all of which may link with the main hematoma. Expansion‐prone hematoma was defined as the presence of ≥1 of the above‐mentioned NCCT markers. Our primary outcome was poor outcome, defined as 90‐day mRS score of 4 to 6. , The poor clinical outcome of mRS score of 3 to 6 was considered as a secondary outcome.

Statistical Analysis

All statistical analyses were performed using SPSS 21.0 (SPSS, Chicago, IL) and MedCalc version 11.4.2. The performances of blend sign, black hole sign, island sign, and expansion‐prone hematoma for predicting IVH growth and RHE were evaluated using the receiver operating characteristic curve analysis. The area under the curve (AUC), sensitivity, specificity, positive predictive value, and negative predictive value were calculated to evaluate the predictive performance. All categorical variables, such as sex, alcohol consumption, smoking, diabetes mellitus, history of hypertension, presence of IVH on initial CT, NCCT markers, expansion‐prone hematoma, poor outcome, and ICH location, were compared using the χ2 test or the Fisher exact test, where appropriate. The continuous variables, including age, baseline ICH volume, baseline IVH volume, baseline Glasgow Coma Scale score, time from onset to initial CT, admission systolic BP, and admission diastolic BP, were compared using the Student t test or the Mann‐Whitney U test. Interobserver agreement on categorical variables was calculated using the Cohen κ interagreement test. Multivariate logistic regression analyses were performed by including all variables with P≤0.1 in the univariate analysis. The level of significance was set to a P<0.05.

The data that support the findings of this study are available from the corresponding author on reasonable request.

Results

We finally included 314 patients with ICH in the analysis (Figure 1). The mean age of these patients was 59.7±12.3 years. The median time from ICH onset to initial CT scans was 2 hours (interquartile range, 1–4 hours), and the interval time from symptom onset to follow‐up CT scan was 20 hours (interquartile range, 14–26 hours). Hematomas on the initial CT scan were located at basal ganglia (173 [55.1%]), thalamus (78 [24.8%]), cerebral lobes (40 [12.7%]), and infratentorial area (23 [7.3%]). There were 48 (15.3%) patients with blend sign, 43 (13.7%) with black hole sign, 47 (15.0%) with island sign, and 96 (30.6%) with expansion‐prone hematoma. Interobserver agreement was excellent for evaluation of the presence of blend sign (κ=0.84 [95% CI, 0.74–0.92]), black hole sign (κ=0.87 [95% CI, 0.77–0.94]), and island sign (κ=0.91 [95% CI, 0.84–0.96]).

Figure 1

Cohort selection flowchart.

CT indicates computed tomography; ICH, intracerebral hemorrhage; and IVH, intraventricular hemorrhage.

The baseline demographics, clinical, radiological characteristics, and functional outcome between patients with and without IVH growth or RHE were shown in Table 1. Of 314 patients, 75 (23.9%) had HE, 61 (19.4%) had IVH growth, and 93 (29.6%) had RHE. Of 61 patients with IVH growth, there were 18 (29.5%) without HE. Among these 18 patients, 14 (77.8%) had the primary outcome (mRS score 4–6) and 16 (88.9%) had the secondary outcome (mRS score 3–6). Of 93 patients with RHE, there were 29 (31.2%) with blend sign, 26 (28.0%) with black hole sign, 34 (36.6%) with island sign, and 54 (58.1%) with expansion‐prone hematoma (Table 1). Different active bleeding definition, stratified by HE‐defining characteristics, was shown in Table S1. Among these 25 patients with delayed IVH, 23 (92%) had HE.

Table 1

Univariate Analysis for IVH Growth or RHE

	IVH Growth (n=61; 19.4%)	No IVH Growth (n=253; 80.6%)	P Value	RHE (n=93; 29.6%)	No RHE (n=221; 70.4%)	P Value
Demographic
Age, mean (SD), y	63.3 (10.9)	58.8 (12.5)	0.009*	62.0 (11.6)	58.7 (12.5)	0.029*
Sex, male, n (%)	39 (63.9)	167 (66.0)	0.760	64 (68.8)	142 (64.3)	0.437
Clinical characteristics
Alcohol consumption, n (%)	25 (41.0)	108 (42.7)	0.809	41 (44.1)	92 (41.6)	0.687
Smoking, n (%)	26 (42.6)	119 (47.0)	0.535	45 (48.4)	100 (45.2)	0.611
Diabetes mellitus, n (%)	7 (11.5)	29 (11.5)	0.998	13 (14.0)	23 (10.4)	0.364
History of hypertension, n (%)	45 (73.8)	177 (70.0)	0.557	67 (72.0)	155 (70.1)	0.735
Admission SBP, mean (SD), mm Hg	178.8 (32.8)	169.1 (26.8)	0.016*	177.6 (30.4)	168.1 (27.0)	0.007*
Admission DBP, mean (SD), mm Hg	102.3 (19.1)	98.3 (17.7)	0.124	101.8 (18.6)	98.0 (17.7)	0.082*
Admission GCS score, median (IQR)	10 (6–14)	14 (12–15)	<0.001*	12 (7.5–14)	14 (12–15)	<0.001*
Imaging features
Time from onset to CT, median (IQR), h	1 (1–2)	2 (1–4)	<0.001*	1.5 (1–2.75)	2 (1–4)	0.001*
Presence of IVH on initial CT, n (%)	36 (59.0)	63 (24.9)	<0.001*	38 (40.9)	61 (27.6)	0.021*
Baseline ICH volume, median (IQR), mL	15.9 (9.7–28.9)	11.7 (7.1–20.9)	0.001*	16.8 (9.2–30.1)	11.4 (6.7–19.3)	<0.001*
Baseline IVH volume, median (IQR), mL	1.1 (0–7.8)	0 (0–0.12)	<0.001*	0 (0–4.2)	0 (0–1.1)	0.046*
Blend sign, n (%)	17 (27.9)	31 (12.3)	0.002*	29 (31.2)	19 (8.6)	<0.001*
Black hole sign, n (%)	19 (31.1)	24 (9.5)	<0.001*	26 (28.0)	17 (7.7)	<0.001*
Island sign, n (%)	25 (41.0)	22 (8.7)	<0.001*	34 (36.6)	13 (5.9)	<0.001*
Expansion‐prone hematoma, n (%)	35 (57.4)	61 (24.1)	<0.001*	54 (58.1)	42 (19.0)	<0.001*
ICH locations, n (%)
Basal ganglia hemorrhage	26 (42.6)	147 (58.1)	0.029*	46 (49.5)	127 (57.5)	0.193
Thalamic hemorrhage	23 (37.7)	55 (21.7)	0.010*	24 (25.8)	54 (24.4)	0.797
Lobar hemorrhage	8 (13.1)	32 (12.6)	0.922	17 (18.3)	23 (10.4)	0.056*
Infratentorial hemorrhage	4 (6.6)	19 (7.5)	1.000	6 (6.5)	17 (7.7)	0.700
Outcome
90‐d mRS score of 4–6, n (%)	51 (83.6)	69 (27.3)	<0.001*	65 (69.9)	55 (24.9)	<0.001*
90‐d mRS score of 3–6, n (%)	56 (91.8)	103 (40.7)	<0.001*	74 (79.6)	85 (38.5)	<0.001*
90‐d mRS score, median (IQR)	6 (4–6)	2 (1–4)	<0.001*	5 (3–6)	2 (1–3.5)	<0.001*

CT indicates computed tomography; DBP, diastolic blood pressure; GCS, Glasgow Coma Scale; ICH, intracerebral hemorrhage; IQR, interquartile range; IVH, intraventricular hemorrhage; mRS, modified Rankin Scale; RHE, revised hematoma expansion; and SBP, systolic blood pressure.

P≤0.1.

After adjustment for age, baseline Glasgow Coma Scale score, time from onset to initial CT, baseline hematoma volume, baseline IVH volume, presence of IVH on initial CT, systolic BP, and ICH location, we found that blend sign, black hole sign, island sign, and expansion‐prone hematoma could predict IVH growth in the multivariate logistic regression analysis, individually (Table 2). Moreover, blend sign, black hole sign, island sign, and expansion‐prone hematoma could independently predict RHE in the multivariate logistic regression analysis after adjusting the potential confounding factors (Table 2).

Table 2

Multivariable Logistic Regression Models of NCCT Markers for Predicting IVH Growth and RHE

Variables	Adjusted Odds Ratio	95% CI	P Value
IVH growth*
Blend sign	5.08	2.00–12.91	0.001
Black hole sign	2.81	1.13–6.98	0.026
Island sign	6.94	2.78–17.33	<0.001
Expansion‐prone hematoma	7.38	3.03–17.99	<0.001
RHE ‡
Blend sign	5.58	2.56–12.18	<0.001
Black hole sign	2.57	1.14–5.83	0.023
Island sign	6.55	2.88–14.87	<0.001
Expansion‐prone hematoma	6.11	3.13–11.93	<0.001

IVH indicates intraventricular hemorrhage; NCCT, noncontrast computed tomography; and RHE, revised hematoma expansion.

Adjusted for age, baseline Glasgow Coma Scale score, time from onset to initial computed tomography, baseline hematoma volume, baseline IVH volume, presence of IVH on initial computed tomography, admission systolic blood pressure, and intracerebral hemorrhage location.

Adjusted for age, baseline Glasgow Coma Scale score, time from onset to initial computed tomography, baseline hematoma volume, baseline IVH volume, presence of IVH on initial computed tomography, admission systolic blood pressure, admission diastolic blood pressure, and intracerebral hemorrhage location.

The diagnostic performances of the NCCT markers in predicting the active bleeding of ICH were illustrated in Table 3. Briefly, the model of expansion‐prone hematoma (AUC=0.70) had higher predictive performance for prediction of RHE than blend sign (AUC=0.61), black hole sign (AUC=0.60), and island sign (AUC=0.65; Table 3). Univariate predictors of primary and secondary outcome were shown in Table S2. Multivariable logistic regression models of NCCT markers and the definitions of active bleeding for primary and secondary outcome were shown in Table S3. The NCCT markers and the definitions of active bleeding for predicting primary and secondary outcome were illustrated in Figure 2. Expansion‐prone hematoma had a higher predictive value than blend sign or island sign in predicting the primary outcome (P<0.05; Figure 2A). Expansion‐prone hematoma had higher predictive value than any single sign in predicting secondary outcome (P<0.05; Figure 2D). The diagnostic accuracy of RHE in predicting primary and secondary outcome was significantly higher than that of HE (P<0.05; Figure 2B and 2E). In addition, the diagnostic performances between expansion‐prone hematoma and RHE were not significantly different in predicting primary and secondary outcome (Figure 2C and 2F).

Table 3

NCCT Markers Associated With IVH Growth and RHE

Outcome Points	Sensitivity, %	Specificity, %	PPV, %	NPV, %	AUC (95% CI)
IVH growth
Blend sign	27.9	87.7	35.0	83.0	0.58 (0.50–0.66)
Black hole sign	31.1	90.5	44.0	85.0	0.61 (0.52–0.69)
Island sign	41.0	91.3	53.0	87.0	0.66 (0.58–0.75)
Expansion‐prone hematoma	57.4	75.9	36.0	88.0	0.67 (0.59–0.75)
RHE
Blend sign	31.2	91.4	60.0	76.0	0.61 (0.54–0.69)
Black hole sign	28.0	92.3	60.0	75.0	0.60 (0.53–0.67)
Island sign	36.6	94.1	72.0	78.0	0.65 (0.58–0.73)
Expansion‐prone hematoma	58.1	81.0	56.0	82.0	0.70 (0.63–0.76)

AUC indicates area under the curve; IVH, intraventricular hemorrhage; NCCT, noncontrast computed tomography; NPV, negative predictive value; PPV, positive predictive value; and RHE, revised hematoma expansion.

Figure 2

The receiver operating characteristic curves of noncontrast computed tomography (NCCT) markers and active bleeding for predicting primary outcome (modified Rankin Scale [mRS] score 4–6) and secondary outcome (mRS score 3–6).

A, The NCCT markers for predicting primary outcome. B, Definitions of active bleeding for predicting primary outcome. C, Comparison of the expansion‐prone hematoma (EPH) and revised hematoma expansion (RHE) for predicting primary outcome. D, The NCCT markers for predicting secondary outcome. E, Definitions of active bleeding for predicting secondary outcome. F, Comparison of the EPH and RHE for predicting secondary outcome. AUC indicates area under the curve; and IVH, intraventricular hemorrhage.

Discussion

In this study, we found that blend sign, black hole sign, and island sign could independently predict IVH growth and RHE. Furthermore, we found that the model of expansion‐prone hematoma showed higher performance for predicting RHE and poor outcome than any single NCCT maker. The predictive accuracy among expansion‐prone hematoma and RHE was similar in predicting poor outcome.

In previous studies, heterogeneous or irregularly shaped hematoma in the intraparenchymal tissue may reflect the active bleeding of HE. , Many NCCT markers, such as blend sign, black hole sign, CT hypodensities, and island sign, are associated with HE and have been validated in several studies. , Recently, the standards for detecting NCCT markers of HE from the International NCCT ICH Study Group have summarized practical standards for detecting NCCT markers and encourage that future clinical investigators may include the NCCT markers in the studies on HE.

It is generally known that HE is strongly related to unfavorable outcome. What is more, the dynamic changes of IVH volume increase >1 mL or delayed IVH also could predict poor outcomes. , , In our study, of 61 patients with IVH growth, 18 (29.5%) patients had not undergone HE. Among these patients, 14 (77.8%) had poor outcomes, which were traditionally defined as nonexpanders. Moreover, we found that IVH growth possessed a higher AUC value and odds ratio for prediction of poor outcome than HE. Thus, in clinical practice, it is important to predict the dynamic process of IVH growth together with conventional definitions of HE in the short‐term stage of ICH. In this study, the NCCT markers, including blend sign, black hole sign, and island sign, were considered as the independent factors for predicting IVH growth, which would be expected to improve the ability to predict clinical functional outcomes.

Recently, Yogendrakumar et al have pointed out that the inclusion of IVH expansion into the current definition of HE could provide superior diagnostic accuracy for predicting poor outcome compared with the conventional definition of HE, which is confirmed in our results. Moreover, blend sign, black hole sign, and island sign could independently predict the revised definition of HE by incorporating IVH growth and the current definition of HE. From a clinical standpoint, although the specificity of the single sign in predicting RHE is high, its sensitivity still needs to be improved. A prediction model composed of NCCT markers is expected to improve the risk prognostication.

Recently, a predictive model with ≥1 of the blend sign, black hole sign, or island sign was considered as expansion‐prone hematoma, which has higher performance for predicting HE than any single sign. In this study, when we extend the conventional definition of HE by incorporating the IVH growth, expansion‐prone hematoma could be a better model for predicting RHE than any single NCCT maker. In our former study, we have defined a subgroup of small hematomas without NCCT markers as benign ICH. Patients with benign ICH are at low risk of HE and unfavorable outcome. It indicates that the 3 NCCT signs could represent patients with ICH with a high risk of active bleeding and poor outcome. In addition, expansion‐prone hematoma could be a better predictor of poor outcome when compared with any single NCCT sign, which is similar to a prior study.

In several retrospective studies, patients with NCCT markers did not benefit from tranexamic acid and intensive BP reduction. , However, a recent subsequent analysis of the ATACH‐2 (Antihypertensive Treatment of Acute Cerebral Hemorrhage 2) trial found that intensive BP reduction in the ultraearly phase of ICH could reduce the rate of HE and further improved the patients’ outcome. In this study, our results support that NCCT markers could predict the IVH growth and RHE, and help researchers identify patients with a high risk of the extending definition of HE. Concurrently, we found that the model of expansion‐prone hematoma has better accuracy for predicting the RHE and unfavorable functional outcome of ICH than any single sign, which may be beneficial to the antiexpansion treatment, such as BP reduction in the ultraearly stage.

Our findings have several clinical implications. First, NCCT markers are closely associated with IVH growth and RHE. This finding has not been reported in previous studies. Second, combining blend sign, black hole sign, and island sign into a prediction model can better predict RHE, which is conducive to the risk stratification of active bleeding. Furthermore, the model of expansion‐prone hematoma may be helpful for risk prognostication and selecting patients for antiexpansion treatment in clinical trials and provide more accurate prognostic information for clinical decision‐making in clinical practice.

Our research has some limitations. First, this study was a single‐center study, which may limit the generalizability to other populations. Second, the sample size in this research is relatively small, which has limited the accuracy of our subgroup analysis to some degree. Therefore, further replication of our findings in a large‐scale multicenter cohort is warranted.

Conclusions

We reported that NCCT markers are closely associated with IVH growth and RHE. Moreover, expansion‐prone hematoma showed higher predictive accuracy for the prediction of RHE and poor outcome than any single NCCT marker. These findings may assist in risk stratification of active bleeding and clinical decision‐making in patients with acute ICH.

Sources of Funding

This study was supported by grants from the National Key Research and Development Program of China (Nos. 2018YFC1312200 and 2018YFC1312203), the China Association of Science and Technology Young Talent Project (No. 2017QNRC001), and Chongqing High‐End Young Investigator Project (No. 2019GDRC005).

Disclosures

Dr Dowlatshahi has received consulting and research contracts from Ottawa Hospital Research Institute. The remaining authors have no disclosures to report.

Tables S1–S3

Figure S1

Click here for additional data file.