Intracranial thrombus extent predicts clinical outcome, final infarct size and hemorrhagic transformation in ischemic stroke: the clot burden score

Abstract

Background In ischemic stroke, functional outcomes vary depending on site of intracranial occlusion. We tested the prognostic value of a semiquantitative computed tomography angiography-based clot burden score.

Methods Clot burden score allots major anterior circulation arteries 10 points for presence of contrast opacification on computed tomography angiography. Two points each are subtracted for thrombus preventing contrast opacification in the proximal M1, distal M1 or supraclinoid internal carotid artery and one point each for M2 branches, A1 and infraclinoid internal carotid artery. We retrospectively studied patients with disabling neurological deficits (National Institute of Health Stroke Scale score ≥5) and computed tomography angiography within 24-hours from symptom onset. We analyzed percentages independent functional outcome (modified Rankin Scale score ≤2), final infarct Alberta Stroke Program Early Computed Tomography Score and parenchymal hematoma rates across categorized clot burden score groups and performed multivariable analysis.

Results We identified 263 patients (median age 73-years, National Institute of Health Stroke Scale score 10, onset-to-computed tomography angiography time 165 min). Clot burden score<10 was associated with reduced odds of independent functional outcome (odds ratio 0·09 for clot burden score≤5; odds ratio 0·22 for clot burden score 6–7; odds ratio 0·48 for clot burden score 8–9; all versus clot burden score 10; P<0·02 for all). Lower clot burden scores were associated with lower follow-up Alberta Stroke Program Early CT Scores (P<0·001) and higher parenchymal hematoma rates (P=0·008). Inter-rater reliability for clot burden score was 0·87 (lower 95% confidence interval 0·71) and intra-rater reliability 0·96 (lower 95% confidence interval 0·92).

Conclusion The quantification of intracranial thrombus extent with the clot burden score predicts functional outcome, final infarct size and parenchymal hematoma risk acutely. The score needs external validation and could be useful for patient stratification in stroke trials.

Introduction

From current intravenous (i.v.) thrombolysis trials in acute ischemic stroke, only i.v. thrombolysis with tissue plasminogen activator (tPA) within 3 h from symptom onset compared with placebo demonstrated an effect on improved functional outcome (1). Vascular imaging was not a prerequisite in any of these trials. However, spontaneous or treatment induced recanalization until 24 h from symptom onset was associated with improved functional outcomes in recent studies (2–4). Patients with intracranial occlusion may therefore represent a target population for thrombolysis in later time windows or for more aggressive treatment protocols like intraarterial (i.a.) thrombolysis (5).

Vascular recanalization rates vary depending on thrombus location with lower rates in proximal versus distal arteries (2, 4, 6–9). Consequently, in anterior circulation stroke, response to thrombolysis and clinical outcomes have been best in patients with distal middle cerebral artery (MCA) occlusion and worst in terminal internal carotid artery (ICA) occlusion (4, 8, 10). Suspected mechanisms include larger thrombus burden and impaired collateral blood flow (2, 6, 11).

Quantification of thrombus burden may therefore allow homogenization of patient cohorts who might expect a differential treatment response based on site and extent of intracranial occlusion (9, 12). Computed tomography angiography (CTA) is now widely available in emergency departments and can reliably, rapidly and safely diagnose occlusion of major intracranial arteries (13, 14). Still, the impact of CTA-defined occlusion on patient management, outcome and treatment response is debated (9, 12, 15).

The aim of our study was to assess the prognostic value and reliability of a semiquantitative CTA grading system, the clot burden score (CBS), in acute anterior circulation ischemic stroke. We hypothesized that by quantification of intracranial thrombus extent, outcome in terms of functional independence, mortality and final infarct size could be predicted.

Subjects and methods

Patients

We retrospectively studied patients with a final diagnosis of acute anterior circulation ischemic stroke admitted to Foothills Medical Center in Calgary from 04/2002 to 01/2007. Non-contrast CT (NCCT) is the default imaging modality in patients with suspected acute ischemic stroke in our centre. Non-contrast CT is followed by CTA in most patients. Decision to proceed to CTA is at the discretion of the treating stroke neurologist.

For this study, patients were included if they presented with acute disabling neurological deficits defined as a National Institute of Health Stroke Scale (NIHSS) score ≥5 and if CTA was performed within 24 h from symptom onset. Exclusion criteria were pre-morbid modified Rankin scale (mRS) score >2 and final diagnosis of transient ischemic attack (TIA), posterior circulation stroke or nonischemic etiology of symptoms. Clinical baseline variables including NIHSS score and mRS score are routinely recorded prospectively in the patient record. In cases where these scores were unavailable, they were derived retrospectively (16). We classified stroke etiology by Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria (17).

Functional outcome is recorded prospectively at 3 months in a stroke follow-up clinic. Missing functional outcome data were imputed from the discharge mRS using the last-score-carried-forward principle. Functional independence was defined as mRS scores ≤2, death as mRS score 6. This study was approved under a waiver of consent by the local institutional ethics committee.

Imaging

Standard non-helical NCCT was performed on a multislice scanner (GE Medical Systems, Fairfield, CT, USA or Siemens Medical Solutions, Erlangen, Germany) using 120 kV, 170 mAs with 5-mm slice thickness. Coverage was from skull base to vertex with continuous axial slices parallel to the orbitomeatal line. Non-contrast CT was followed by CTA with a helical scan technique. Acquisitions were obtained after a single bolus intravenous contrast injection of 90–120 ml nonionic contrast media into an antecubital vein at 3–5 ml/s. Imaging was auto-triggered by the appearance of contrast material in the ascending aorta. Minimum coverage was from foramen magnum to centrum semiovale with 0·6–1·25 mm slice thickness. Multiplanar volume-reformatted images were immediately created by the CT technologist with 2·5–4·0 mm slice thickness in axial, sagittal and coronal planes. This image reformatting was typically completed within minutes and available for review.

Follow-up NCCT imaging was performed routinely and follow-up magnetic resonance imaging (MRI) as indicated by the treating stroke neurologist. Follow-up imaging had to be performed between days 1 and 7 after symptom onset to be assessed for final infarct extension.

Assessment of thrombus burden

We developed a CBS for the anterior circulation to quantify the extent of ipsilateral intracranial thrombus, allotting major arteries 10 points for the presence of contrast opacification on CTA. Two points each were subtracted for absence of contrast opacification in the complete cross-section of any part of the proximal M1 segment, distal M1 segment or supraclinoid ICA and 1 point each for M2 branches, A1 segment and infraclinoid ICA (Figs. 1 and 2). Partial filling defects suggesting stenosis or non-occlusive thrombus were rated as patent. A score of 10 indicates absence of a visible occlusion on CTA, a score of 0 indicates occlusion of all major intracranial anterior circulation arteries.

Figure 1.  Ten-point clot burden score (CBS): one or two points each (as indicated) are subtracted for absent contrast opacification on computed tomography angiography (CTA) in the infraclinoid internal carotid artery (ICA) (1), supraclinoid ICA (2), proximal M1 segment (2), distal M1 segment (2), M2 branches (one each) and A1 segment (1). The CBS applies only to the symptomatic hemisphere.

Figure 2.  Coronally reconstructed computed tomography angiography. Arrows indicate absence of contrast opacification in the left infraclinoid internal carotid artery (ICA) (one point; part A), supraclinoid ICA (two points; lower arrow part B) and proximal M1 segment (two points; upper arrow part B). The clot burden score is 5.

The CBS was summed based upon the neuroradiologists' CTA report. In cases where the report was not clear about thrombus location or thrombus extension, CTAs were reviewed by a stroke fellow with knowledge of symptom side but blinded to all further clinical data. In addition, reliability for CBS was assessed within and between observers. Three physician groups: neuroradiologists (WYH, MEH), stroke neurologists (MDH, AMD) and stroke fellows (PNS, SS), performed interpretation on a sample of 20 randomly selected CTAs with knowledge of the affected hemisphere. Source images and multiplanar reconstructions were available for evaluation. Each rater repeated reading of the same CTAs in a different but random order at least 1 month apart.

Image analysis

We prospectively and independently applied the Alberta Stroke Program Early CT score (ASPECTS) to all baseline NCCT and available follow-up scans by three-reader consensus (one stroke neurologist, and two stroke fellows). Readers were blinded to clinical information except for symptom side. We assigned the ASPECTS as previously described (18). Signs of ischemia on an NCCT scan were defined as parenchymal X-ray hypoattenuation and/or effacement of cortical sulci. ASPECTS is only scored for acute changes, a score of 10 indicates a normal scan and a score of 0 complete MCA infarction. On follow-up NCCT scans, we applied the ASPECTS to regions of subacute brain infarction. On follow-up MRI scans we similarly scored hyperintense regions on diffusion-weighted sequences (19). Hemorrhagic transformation (HT) on follow-up scan was allocated parenchymal hematoma (PH) or hemorrhagic infarction (HI) by ECASS criteria (20).

Statistical analysis

Data are reported using standard descriptive statistics. The CBS was divided into four groups (≤5, 6–7, 8–9, 10) arbitrarily. The primary clinical outcome was independent functional outcome defined as a modified Rankin scale score ≤2. Secondary outcomes were death, final-infarct size applying ASPECTS and HT on follow-up imaging categorized as HI and PH.

We developed multivariable logistic regression models to adjust for age, gender, baseline clinical variables, use of thrombolysis and onset-to-CTA time in predicting clinical outcome. Model development was based upon clinical variable selection and step-wise manual elimination of variables that were not predictive of outcome in bivariable analysis (P<0·25). For the final parsimonious model, we included only variables that were predictive of outcome in bivariable analysis or terms of special interest considered relevant a priori.

A potential confounding factor must not be affected by the exposure or the disease. In particular, it cannot be an intermediate step in the causal path between the exposure (CBS) and the disease (clinical outcome) (21, 22). In developing the multivariable model, we assumed the following causal chain: an intracranial arterial occlusion (CBS) caused an ischemic stroke syndrome (NIHSS score), which was manifest by areas of hypodensity on NCCT (ASPECTS). We therefore did not consider baseline NIHSS score and ASPECTS as potential confounders on the CBS in the prediction of final clinical outcome.

We evaluated logistic regression model performance using the Hosmer–Lemeshow goodness of fit statistic (measuring model calibration) and the c-statistic (measuring model discrimination). We tested for a potential CBS by thrombolysis interaction using the Wald statistic.

Results

Baseline data

We identified 657 consecutive patients who received CTA for clinically suspected acute ischemic stroke within 24 h from symptom onset. Among these, 276 patients had final diagnosis anterior circulation ischemic stroke and a baseline NIHSS score ≥5. Another 13 patients were excluded as their baseline mRS score was >2. Therefore, this report describes 263 patients with a median age of 73 years [interquartile range (IQR) 59–79] of whom 146 (55%) were men. Clinical data and CBS were available for all patients, baseline ASPECTS for 254 patients and follow-up ASPECTS for 214 patients.

Overall, median baseline NIHSS score was 10 (IQR 7–17), onset-to-CTA-time 165 min (IQR 98–347), baseline ASPECTS 8 (IQR 6–10). One hundred thirty-one patients (49·8%) received i.v. thrombolysis and/or thrombolytic or mechanical i.a. therapy (101 i.v. tPA only, 22 combined i.v.–i.a. treatment, 8 i.a. treatment only). One hundred fifty-two patients (58%) had a visible intracranial occlusion on CTA. The overall distribution of CBS values was skewed with a median CBS value of 9 (IQR 7–10).

Baseline clinical data according to categorized CBS groups are summarized in Table 1. Patients with lower CBS (i.e. higher thrombus burden) had higher baseline NIHSS scores and lower baseline ASPECTS (Table 1 and Fig. 3). There was a significant correlation of lower CBS with lower ASPECTS (Spearman's ρ=0·45, P<0·001) and higher NIHSS scores (Spearman's ρ=−0·50, P<0·001).

Table 1.   Baseline characteristics according to categorized clot burden score (CBS) groups

N	CBS ≤5	CBS 6–7	CBS 8–9	CBS 10	P
	25	41	86	111
IQR, interquartile range; CTA, computed tomography angiography; NIHSS, National Institute of Health Stroke Scale; ASPECTS, Alberta Stroke Program Early CT Score; i.v., intravenous; i.a., intra-arterial; tPA, tissue plasmogen activator; TOAST, Trial of Org 10172 in Acute Stroke Treatment.
Age, median (IQR)	71 (44–80)	74 (60–79)	73 (58–81)	73 (64–70)	0.798
Gender female, % (n)	28 (7)	54 (22)	57 (49)	43 (48)	0.228
Onset-to-CTA time, median (IQR)	209 (120–586)	141 (90–270)	145.5 (90–347)	174 (112–330)	0.155
Baseline NIHSS, median (IQR)	18 (15–20)	15 (11–19)	12 (8–18)	7 (6–11)	<0.001
Baseline ASPECTS, median (IQR)	5 (3–8.5)	7 (5–9)	7 (5–9)	10 (8–10)	<0.001
Any thrombolysis, % (n)	68 (17)	80 (33)	47 (40)	37 (41)	0.004
I.v. tPA only	44 (11)	46 (19)	35 (30)	37 (41)
I.a. therapy only	8 (2)	10 (4)	2 (2)	0 (0)
Combined i.v.–i.a. therapy	16 (4)	24 (10)	9 (8)	0 (0)
Risk factors, % (n)
Arterial hypertension	52 (13)	54 (22)	63 (54)	65 (72)	0.451
Diabetes	16 (4)	15 (6)	15 (13)	14 (16)	0.993
Hypercholesterolemia	32 (8)	15 (6)	28 (24)	28 (31)	0.288
Smoking (current)	32 (8)	20 (8)	21 (18)	20 (22)	0.585
Coronary artery disease	40 (10)	17 (7)	22 (19)	27 (30)	0.184
Atrial fibrillation	12 (3)	29 (12)	31 (27)	23 (26)	0.194
History of TIA or stroke	16 (4)	17 (7)	21 (18)	19 (21)	0.956
TOAST, % (n)					<0.001
Large artery	32 (8)	15 (6)	14 (12)	23 (26)
Cardioembolic	24 (6)	54 (22)	55 (47)	34 (38)
Small vessel	0 (0)	0 (0)	0 (0)	15 (17)
Other determined	12 (3)	10 (4)	5 (4)	4 (4)
Undetermined	32 (8)	22 (9)	27 (23)	23 (26)

Figure 3.  Baseline ASPECTS (bASPECTS) score and baseline NIHSS (bNIHSS) score for categorized CBS groups where 10 is best and 0 is worst. With lower clot burden score, patients had lower bASPECTS (range 0–10) and higher bNIHSS scores (range 0–42), respectively.

Clinical outcome

At 3 months, 119 patients (45·2%) were functionally independent, 107 patients (40·7%) were functionally dependent and 37 patients (14·1%) were deceased. Figure 4 shows the distribution of functional outcomes across categorized CBS groups. Patients with higher CBS were more likely to have an independent functional outcome (P<0·001; Table 2). In contrast, mortality increased in patients with low CBS compared with patients with high CBS (P<0·001).

Figure 4.  Functional outcome across categorized CBS groups where 10 is best and 0 is worst.

Table 2.   Clinical and imaging outcomes for categorized Clot Burden Score (CBS) groups

Clinical outcome, N	CBS ≤5	CBS 6–7	CBS 8–9	CBS 10	P
	25	41	86	111
F/U denotes follow-up; HT, hemorrhagic transformation; PH, parenchymal hematoma; HI, hemorrhagic infarction; IQR, interquartile range; mRS, modified Rankin scale; ASPECTS, Alberta Stroke Program Early CT Score.
mRS ≤2, % (n)	16 (4)	29 (12)	44 (38)	59 (65)	<0.001
Death, % (n)	44 (11)	17 (7)	12 (10)	8 (9)	<0.001
Imaging outcome, N	20	35	71	88
F/U ASPECTS, median (IQR)	2.5 (0.5–5.5)	4 (3–8)	6 (4–7)	8 (7–9.5)	<0.001
HT, % (n)	45 (9)	40 (14)	23 (16)	15 (13)	0.003
PH, % (n)	25 (5)	11 (4)	7 (5)	3 (3)	0.008
HI, % (n)	20 (4)	29 (10)	15 (11)	11 (10)	0.128

In logistic regression analysis, CBS was an independent predictor of good functional outcome in a graded fashion after adjustment for age, gender and presence vs. absence of thrombolytic therapy. A CBS<10 was associated with reduced odds of good functional outcome (Table 3). Other clinical variables (vascular risk factors, onset-to-CTA time) were not independent predictors of outcome and so were dropped from the final model. Particularly inclusion of history of diabetes mellitus and hypertension did not substantially influence the results (data not shown). Apart from CBS, age was the only significant additional predictor (Table 3). In testing the performance of our final regression equations, we showed adequate goodness of fit and good discriminative ability (Table 3).

Table 3.   Multivariable analysis for independent functional outcome and death [odds ratios for clot burden score (CBS) categories compared to CBS 10]

Parameter	mRS 0–2^*	P	Death^†	P
	Odds ratio (CI95)		Odds ratio (CI95)
^* c-statistic=0.720; Hosmer–Lemeshow goodness of fit test (20 groups), χ^2 statistic P=0.821. ^† c-statistic=0.762; Hosmer–Lemeshow goodness of fit test (20 groups), χ^2 statistic P=0.198. mRS, modified Rankin scale; CI, confidence interval.
CBS 10	1.00	–	1.00	–
CBS 8–9	0.48 (0.27–0.88)	0.018	1.54 (0.59–4.03)	0.377
CBS 6–7	0.22 (0.10–0.53)	0.001	2.41 (0.78–7.42)	0.125
CBS ≤5	0.09 (0.03–0.31)	<0.001	11.02 (3.60–33.76)	<0.001
Age (per year increase)	0.96 (0.94–0.98)	<0.001	1.04 (1.01–1.08)	0.012
Male gender	0.94 (0.54–1.61)	0.812	1.04 (0.47–2.27)	0.929
Thrombolysis present	1.22 (0.70–2.13)	0.475	1.10 (0.50–2.43)	0.805

We assessed an interaction term to assess the effect of i.v. and/or i.a. thrombolysis by the CBS and found no evidence of interaction, implying that in this sample, we could not find a differential effect of thrombolysis on independent functional outcome according to the baseline CBS (P=0·095, Wald test).

Similarly, a CBS≤5 (i.e. high thrombus burden) predicted death (Table 3). Again, age was the only significant additional predictor. CBS values 6–7 and 8–9 both indicated a non-significant trend towards higher mortality. Assessment of test performance showed adequate goodness of fit and good discriminative ability of the final regression equations (Table 3).

Imaging outcome

Final infarct size was larger in patients with low compared with patients with high CBS (P<0·001; Table 2). Furthermore, patients with lower CBS were more likely to have hemorrhagic infarct transformation (P=0·003) and parenchymal hematoma (P=0·008) on follow-up scans (Table 2). Differences in HI rates were not statistically significant (P=0·128).

Reliability testing

In a substudy of 20 randomly selected cases, overall interrater reliability among six readers (two stroke neurologists, two stroke fellows, two neuroradiologists) for the exact CBS, assessed by the intraclass correlation coefficient was 0·87 (lower CI₉₅ 0·71) and intrarater reliability was 0·96 (lower CI₉₅ 0·83). The minimal detectable difference between raters was two points (23).

Discussion

We have shown that a CT-angiography based CBS predicted several clinical and imaging outcome parameters in a large cohort of patients presenting with acute anterior circulation ischemic stroke. With increasing CBS (i.e. less thrombus burden), patients were significantly more likely to have an independent functional outcome and less likely to die. For example, in a patient with a CBS of 5 or below, the odds of having an independent functional outcome was 10 times lower and the odds of death more than 10 times higher compared with a patient with a CBS of 10. Accordingly, patients with lower CBS had significantly larger final infarcts and were significantly more likely to have parenchymal hematoma formation.

There is considerable evidence that the site of intracranial arterial occlusion is an important predictor of spontaneous or treatment induced recanalization and subsequently good clinical outcome in anterior circulation ischemic stroke (2, 4, 8, 9). In order to compare treatment efficacies of different treatment regimens in randomized stroke trials, objective methods are required to homogenize patient cohorts based on intracranial occlusion status. A recently proposed angiographic grading scheme, based on occlusion site and collateral supply, was associated with initial stroke severity, recanalization rate with i.a. thrombolysis and short term clinical outcome (24, 25). Our results support the concept that not only the occlusion site but the resulting amount of thrombus burden in different vascular segments is a major determinant of stroke severity and outcome in anterior circulation stroke. As an advantage over angiographic scores, CBS does not rely on invasive vascular imaging and is therefore readily available when trying to decide on the best treatment approach in individual patients.

Because of low recanalization rates with i.v. thrombolysis, more aggressive treatment protocols like immediate i.v.–i.a. treatment have been proposed for patients with distal ICA or proximal MCA occlusion (9). This treatment concept is currently tested in the Interventional Management of Stroke (IMS) 3 trial (26). As we did not perform repeat angiographic studies in most patients, we cannot comment on the relationship between thrombus burden and recanalization rates. This may be possible in ongoing trials, which apply CT angiography for patient stratification. A European trial comparing i.v. thrombolysis with i.a. therapy in patients with ICA or MCA occlusion documented on CTA or MRA has recently been proposed (27). Quantification of thrombus burden with the CBS may be helpful to assess whether these more aggressive treatment approaches (such as i.a. therapy) preferentially benefit based on the extent of intracranial thrombus burden (28).

The CBS was associated with the HT rate and particularly parenchymal hematoma rate in our study. Presence of arterial vascular occlusion is commonplace in ischemic stroke patients who have hemorrhagic infarct transformation with or without thrombolysis (12, 29). Potentially, the likelihood of hemorrhagic conversion was triggered by more severe ischemia in patients with lower CBS as indicated by higher baseline NIHSS scores and lower baseline ASPECTS in our study. Furthermore, higher thrombus burden may cause more severe focal hypoperfusion, which was predictive of HT in consecutive stroke patients studied with single-photon emission tomography (29). It needs to be determined if the severity of ischemia, particularly the extent of early ischemic changes on NCCT, and the amount of thrombus burden have an additive or multiplicative joint effect in predicting hemorrhagic complications (30).

Our study has limitations. Clinical outcomes were identified retrospectively based on the chart review of discharge outcome (36%) or in the 3 month follow-up clinic (64%). The CBS was derived from the CTA report. However, subsequent reliability testing revealed excellent inter- and intrarater reliability for CBS among physicians with different levels of training. The CT angiogram may not fully define the thrombus extent because lack of contrast opacification in distal arteries may represent delayed filling (i.e. the scan level being ahead of the dye) rather than intraluminal thrombus. However, this may be an advantage of the score as it may indicate a patient's collateral status which was a predictor of functional outcome in previous angiographic studies (31, 32). Finally, we have shown that the CBS predicted functional outcome in our study population. However, its predictive value may not be present outside this data set and thus needs to be validated in a separate group of patients before it can be considered a complete prediction model.

This study describes a simple and reliable CT-angiography-based CBS, which predicted independent functional outcome and death, and was closely linked to the final infarct size and HT rate in a large cohort of patients with acute anterior circulation stroke. Our results need external validation. Application of CBS to interventional stroke trials comparing i.v. thrombolysis with i.v.–i.a. or immediate i.a. therapy could determine whether more aggressive treatment paradigms may preferentially benefit based on the extent of intracranial thrombus burden.

Acknowledgements

Michael D. Hill was funded by the Canadian Institutes for Health Research and the Heart & Stroke Foundation of Alberta, NWT and Nunavut. Philip A. Barber was funded by the Canadian Institutes for Health Research of Alberta, the Heritage Foundation for Medical Research, and the Heart and Stroke foundation of Canada. Andrew M. Demchuk received salary support from Alberta Heritage Foundation for Medical Research. The study was supported from funding from the Heart and Stroke Foundation of Alberta, NWT and Nunavut.

Conflicts of interest disclosure: The author reports no conflict of interest.