clinical outcome in patients with acute middle
cerebral artery stroke
van den Wijngaard IR Wermer MJH Boiten J Algra A Holswilder G Meijer FJA Dippel DWJ Velthuis BK Majoie CBLM van Walderveen MAA
Stroke 2016;47:762-7
8
510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard Processed on: 2-6-2017 Processed on: 2-6-2017 Processed on: 2-6-2017
Processed on: 2-6-2017 PDF page: 150PDF page: 150PDF page: 150PDF page: 150
150
ABSTRACT
Background and purpose Venous flow in the downstream territory of an occluded artery
may influence patient prognosis after ischemic stroke. Our aim was to study cortical venous filling (CVF) in a time-resolved manner with dynamic CT-angiography (CTA) and to assess the relationship with clinical outcome.
Methods Patients with a proximal middle cerebral artery occlusion underwent non-contrast
CT (NCCT) and whole brain CT perfusion/dynamic-CTA within 9 hours after stroke-onset. We defined poor outcome as a modified Rankin scale (mRS) score of ≥3. Association between the extent and velocity of CVF and poor outcome at 3 months was analyzed with Poisson- regression. Prognostic value of optimal CVF (maximum opacification of cortical veins) in addition to age, stroke-severity, treatment, Alberta Stroke Program Early CT (ASPECT) score, cerebral blood flow and collateral status was assessed with logistic regression and summarized with the area-under-the-curve (AUC).
Results Eighty-eight patients were included, with a mean age of 67 years. By combining the
extent and velocity of optimal CVF we observed a decreased risk of poor outcome in patients with good and fast optimal CVF, risk ratio of 0.5 (95%confidence interval 0.3-0.7). Extent and velocity of optimal CVF had additional prognostic value (AUC 0.88; 95%CI 0.77-0.98; p<0.02) compared with a model without CVF information.
Conclusion The combination of extent and velocity of optimal cortical venous filling, as
assessed with dynamic-CTA, is useful to identify patients with acute middle-cerebral-artery stroke at higher risk of poor clinical outcome at 3-months follow-up.
510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard Processed on: 2-6-2017 Processed on: 2-6-2017 Processed on: 2-6-2017
Processed on: 2-6-2017 PDF page: 151PDF page: 151PDF page: 151PDF page: 151
8
151
INTRODUCTION
Patients with acute ischemic stroke as a result of a proximal anterior circulation occlusion still have a substantial risk of poor clinical outcome despite recent advances in acute stroke treatment.1-4 Imaging markers, such as small infarct core and good collateral status, may
identify patients most likely to benefit from reperfusion therapies and have been used to extend time-windows for treatment.2,5 Multimodal CT-imaging, including NCCT, CTA and CT
perfusion (CTP) is most commonly used in clinical practice to evaluate these imaging markers in acute ischemic stroke patients because this modality is fast, non-invasive, inexpensive and widely available.
Poor cortical venous filling (CVF) in the downstream territory of an occluded artery is a promising prognostic feature, which may be defined by the extent of collaterals and resultant perfusion.6,7 CVF can provide an indirect assessment of perfusion through the
microcirculation.8,9 The Prognostic Evaluation based on Cortical vein score difference In
Stroke (PRECISE) score, as assessed with single-phase-CTA, predicted poor clinical outcome.7
From this observational study the investigators concluded that the extent of cortical venous filling influenced prognosis in the setting of a proximal artery occlusion.7 Measuring delay in
the opacification of cortical veins could have additional prognostic value in acute ischemic stroke, as has been suggested in a nonhuman primate model.6
Multiphase or dynamic-CTA is superior to conventional single-phase-CTA for assessment of vessel filling, since both extent and velocity of vessel filling can be taken into account.10-13
Similarly, CVF may be better defined using dynamic-CTA as compared with single-phase- CTA. To our knowledge, CVF in acute ischemic stroke has not yet been studied in a time- resolved manner. The aim of our study was to investigate whether extent and velocity of CVF as assessed with dynamic-CTA predicts clinical outcome at 3 months in patients with acute middle cerebral artery stroke.
Methods
Study Design
Patients from two university medical centers (Leiden and Radboud University Medical Centers) were selected from the Dutch Acute Stroke (DUST) Study and the Multicenter Randomized Clinical trial of Endovascular Treatment for Acute ischemic stroke in the Netherlands (MR CLEAN). Protocol details of these clinical studies with inclusion and exclusion criteria have been published before.14,15 Patients with an acute ischemic stroke
with a proximal middle cerebral artery (MCA) occlusion (M1 or M2 segments) were included, with or without occlusion of the internal carotid artery (ICA). Clinical data were retrieved from the study databases. The primary outcome measure was a poor clinical outcome defined as a modified Rankin Scale score of three or more at 90 days, indicating functional dependence or death.
510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard Processed on: 2-6-2017 Processed on: 2-6-2017 Processed on: 2-6-2017
Processed on: 2-6-2017 PDF page: 152PDF page: 152PDF page: 152PDF page: 152
152
CT Image Acquisition
For CT image acquisition, a 320-slice multidetector-CT (Aquilion-One Toshiba Medical Systems, Tokyo, Japan) was used, resulting in whole-brain CTP coverage. All patients underwent a standard scanning protocol at presentation, including NCCT, single-phase-CTA from the aortic arch to the vertex, and whole brain CT perfusion/dynamic-CTA. NCCT was evaluated for early ischemic changes (ASPECTS); CTP/CTA for extent of the arterial occlusion (clot burden score), collateral status, cerebral-blood-flow (CBF), cerebral-blood-volume (CBV), mean-transit-time (MTT) and time-to-peak (TTP) (supplemental methods). For whole brain CT perfusion/dynamic-CTA, a total of 19 volumes were obtained over a duration of 1 minute. Dynamic-CTAs were derived from the 320-slice CTP by subtracting the first unenhanced volume of the CTP-study from the subsequent contrast enhanced volumes to ensure that only vessels remained visible. Maximum intensity projections (MIPs) of all 19 volumes were created and displayed in time herewith creating angiography-like movies of time-resolved MIPs. The total radiation dose amounted to 8.4mSv with our current acquisition protocol. CVF was evaluated for the affected and non-affected hemisphere on these time-resolved MIPs.
Cortical venous filling
All cortical veins that drain into the superior sagittal sinus and anastomotic veins were assessed in both hemispheres.(Figure 1 and 2) By visual assessment, cortical venous contrast opacification and the number of cortical and anastomotic veins were evaluated in comparison with the contralateral hemisphere. We defined first CVF as the appearance of any cortical vein draining in the superior sagittal sinus. Optimal CVF was defined as maximum contrast opacification of all cortical veins. If discrepancies in opacification between different cortical veins were present, a choice was made for the moment when the majority of cortical veins showed highest contrast opacification. We defined the end of venous filling as the first moment that contrast medium in all of the cortical veins had completely disappeared. Time points of different venous filling parameters were indicated according to timing acquisition of each of the 19 volumes in seconds. Moreover, venous parameters were related to first contrast opacification of the ICA to correct for differences in time-delay of CTP/dynamic-CTA acquisition after contrast bolus injection between patients.
To assess the velocity of CVF we calculated for all venous time points the median differences in seconds between venous filling of the ipsilateral (affected) versus the contralateral (non- affected) hemisphere. Fast CVF was defined as time points smaller or equal to the median value and slow CVF as larger than the median value. Additionally for extent of CVF, we assessed at the time points of maximum venous opacification whether the number of cortical veins in the affected hemisphere was present in < 50% (poor extent of cortical venous filling) or ≥ 50% (good extent of cortical venous filling) in comparison with the non-affected hemisphere. Consensus reading for the extent of CVF was done in the cohort of the Leiden University Medical Center by a trained neurologist and neuroradiologist who were given information regarding the clini cal symptoms only. All CVF times were independently assessed in the same cohort and agreement between the two observers for velocity of CVF was assessed with multiple Bland-Altman plots. Reperfusion outcomes were independently assessed on digital
510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard Processed on: 2-6-2017 Processed on: 2-6-2017 Processed on: 2-6-2017
Processed on: 2-6-2017 PDF page: 153PDF page: 153PDF page: 153PDF page: 153
8
153
subtraction angiography after endovascular treatment with the modified thrombolysis in cerebral ischemia (mTICI) score and were provided by the MR CLEAN investigators.
Figure 1: Illustration of good versus poor extent of cortical venous filling. Dynamic CT-angiography
maximum intensity projections (MIPs) of a single volume in two patients. Both patients underwent multimodal CT imaging within 90 minutes after symptom-onset, which showed in both patients an occlusion of the M1 segment of the middle cerebral artery and a high ASPECT score. (A) shows good extent of optimal cortical venous filling in the non-affected (black arrows) as well as affected hemisphere (white arrows). Patient A was treated with intravenous thrombolysis and had a good clinical outcome.
(B) shows asymmetry in cortical venous filling between both hemispheres with poor extent of cortical
venous filling in the affected hemisphere (white arrows). Patient B was also treated with intravenous thrombolysis but had a poor clinical outcome at 3 month follow-up.
Statistical analysis
We analysed the association between clinical or radiological characteristics and poor clinical outcome (mRS≥3) after 3 months with univariable and multivariable Poisson regression analysis. The prognostic value of optimal CVF on clinical outcome in addition to characteristics including age, stroke severity, treatment, ASPECTS, CBF and collateral status was analysed with logistic regression models. In a first basic model (model 1) the variables age, baseline National Institute of Health Stroke Scale (NIHSS) score, treatment and ASPECTS were used. In model 2, CBF was added to model 1. In model 3 the collateral status was added to model 2. For the final model 4, we added a combination of extent and velocity of optimal CVF to model 3. Subsequently the area under the curve of the ROC-curves (AUC-ROC) was calculated for all models and potential improvements between the final models and the initial models were determined. Comparison of AUC-ROCs between hierarchical models was done with the likelihood-ratio-test.
510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard Processed on: 2-6-2017 Processed on: 2-6-2017 Processed on: 2-6-2017
Processed on: 2-6-2017 PDF page: 154PDF page: 154PDF page: 154PDF page: 154
154
Figure 2: Illustration of time-resolved assessment of cortical venous filling. Dynamic CT-angiography
maximum intensity projections (MIPs) of 4 volumes displayed chronologically in the same patient with right middle cerebral artery stroke illustrating the importance of time-resolved assessment of cortical venous filling. (A) shows the arterial-venous phase in which cortical veins are visible in the non-affected hemisphere (black arrows), but not yet visible in the affected hemisphere (white arrow). (B) shows optimal cortical venous filling in the non-affected hemisphere (black arrows), while optimal venous filling in the affected hemisphere is seen 5 seconds later as shown in (C) (white arrows). (D) shows a volume in the late venous phase in which cortical venous filling is more prominent in the affected (white arrow) than non-affected hemisphere (black arrow).
RESULTS
Patients
Between July 2010 and July 2014, a total of 96 patients were eligible for evaluation of CVF on dynamic-CTA. Eight patients were excluded because of insufficient quality of the post- processed images for assessment of venous filling. Forty-two patients were selected from the DUST and forty-six patients from the MR CLEAN trial. Sixty-five patients were selected from the Leiden University Medical Center and twenty-three patients from the Radboud University Medical Center. The mean age of the study participants was 67 years; median NIHSS was 15, and 44 participants (50%) were women. Median time from symptom onset to multimodal CT imaging was 77 minutes. Most patients were treated with IVT (59%) alone, whereas IAT was performed in 22 patients (25%) (supplemental table I). At 3 months follow-up, 54 patients (61%) had a poor clinical outcome (mRS 3-6), and 22 of 88 patients (25%) were dead. Bland-Altman analysis showed good agreement between the two observers for all venous time points (n=65, Supplemental Figure I). Therefore, average time points of the two observers
510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard Processed on: 2-6-2017 Processed on: 2-6-2017 Processed on: 2-6-2017
Processed on: 2-6-2017 PDF page: 155PDF page: 155PDF page: 155PDF page: 155
8
155
were used for the final analyses. For the entire study population the mean difference between the non-affected and affected hemisphere was 2.8 seconds for first CVF, 3.7 seconds for optimal CVF and 5.7 seconds for end of CVF (Supplemental Table II) The median difference of arterial contrast supply to the ipsi- and contralateral internal carotid arteries was zero, whereas the median difference for optimal CVF between the non-affected and affected hemisphere was 3 seconds (IQR 1-5). Therefore, we defined slow filling to be present if the difference in optimal CVF in the affected hemisphere was more than 3 seconds in comparison with optimal CVF in the non-affected hemisphere. Because of movement artefacts in few but relevant volumes, time points of optimal CVF could not be assessed with certainty in two patients and the end of CVF could not be assessed in three patients. In 37/86 patients (43%) the velocity of optimal CVF was slow and in 49/86 (57%) it was fast. With regards to the extent of CVF, 61 patients (69%) had a good CVF (i.e. ≥ 50% compared with the non-affected hemisphere), whereas 27 patients (31%) had a poor CVF status.
Univariable analyses
Venous predictors of poor clinical outcome at 3 months follow-up were: poor extent of optimal CVF (RR 1.8; 95%CI 1.4-2.4) and slow velocity of optimal CVF (derived from the differences in optimal CVF between the affected and non-affected hemisphere) (RR 1.6; 95%CI 1.1-2.2). (Table 1, risk ratios of other clinical and radiological variables are presented in supplementary table III). Total duration of CVF in the affected hemisphere was not related to clinical outcome at follow-up.
Combining the extent and velocity of CVF resulted in four subgroups (Table 1). Poor outcome was seen in 14 out of 38 patients (37%) with good extent and fast CVF, as opposed to those with either poor extent, slow CVF or both (39/48=81%; RR 0.5; 95%CI 0.3-0.7).
Multivariable analyses
Poisson analysis
Adjustment for separate clinical and radiological variables did not alter the relationship much between extent plus velocity of optimal CVF and poor clinical outcome. (supplemental table IV) Only when combining optimal CVF with the four most influential covariates (MTT, CBF, CBS and collateral status), the relative risk for poor clinical outcome decreased but remained statistically significant (RR 1.7; 95%CI 1.0-2.7, p<0.05).
ROC analysis
At ROC analysis (Supplemental Figure II), the area under the curve (AUC) was 0.80 (95%CI 0.68-0.92) for model 1 which included variables age, NIHSS, treatment and ASPECT score at admission to predict clinical outcome at 3 months follow-up. Adding CBF to model 1 resulted in model 2 with an AUC of 0.80 (95%CI 0.68-0.93). Adding collateral status to model 2 resulted in model 3 with an AUC of 0.84 (95%CI 0.74-0.95). Adding the extent and velocity of CVF to model 3 resulted in model 4 with an AUC of 0.88 (95%CI 0.77-0.98). The prediction of clinical outcome was better for model 4 as compared with model 3 (p<0.02).
510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard Processed on: 2-6-2017 Processed on: 2-6-2017 Processed on: 2-6-2017
Processed on: 2-6-2017 PDF page: 156PDF page: 156PDF page: 156PDF page: 156
156
Table 1: Poor outcome at follow-up (mRS 3-6) in relation to cortical venous filling(CVF) characteristics
dichotomised at median values (N=88).
Risk of poor clinical outcome at follow up (n/N %) Dichotomised CVF characteristics Poor outcome(n)/ Characteristic present(N) Poor outcome(n) / Characteristic absent(N) Risk Ratio (95%CI) p-value‡ Extent CVF in affected hemisphere Poor extent of CVF* 24/27(89%) 30/61(49%) 1.8(1.4-2.4) <0.001 Velocities CVF in affected hemisphere Slow first CVF† 28/37(76%) 26/51(51%) 1.5(1.1-2.1) 0.02 Slow optimal CVF† 26/36(72%) 27/50(54%) 1.4(1.0-1.9) Slow end of CVF† 24/36(67%) 29/49(59%) 1.1(0.8-1.6)
Slow first to optimal
CVF† 20/35(57%) 33/51(65%) 0.9(0.6-1.3)
Slow first to final CVF† 22/37(59%) 31/48(65%) 0.9(0.7-1.3)
Interhemispheric time differences (Δ)
Δ Internal carotid
artery > median Δ 16/26(62%) 28/50(56%) 1.1(0.7-1.6)
Δ First venous filling >
median Δ 21/29(72%) 33/59(56%) 1.3(0.9-1.8)
Δ Optimal venous filling
> median Δ 29/37(78%) 24/49(49%) 1.6(1.1-2.2) 0.006
Δ End of venous filling
> median Δ 17/29(59%) 36/56(64%) 0.9(0.6-1.3)
Extent and Delay of CVF
< 50% and slow 13/15(89%) 14/38(37%)§ 2.4(1.5-3.7) <0.001
< 50% and fast 10/11(91%) 14/38(37%)§ 2.5(1.6-3.9) <0.001
> 50% and slow 16/22(73%) 14/38(37%)§ 2.0(1.2-3.2) 0.006
> 50% and fast 14/38(37%) 39/48(81%) 0.5(0.3-0.7) <0.001
*poor extent venous filling was defined as less than 50% of cortical venous filling in the affected hemisphere compared with the non-affected hemisphere
†Slow defined as longer than median value in seconds of specific venous time parameter ‡p-values are given in case < .05
§ >50% extent and fast CVF was taken as reference
Numbers may not add up to a total population number because of missing values
Cortical venous filling and endovascular treatment
Twenty-two patients (25%) of our cohort underwent mechanical thrombectomy (in 18 patients preceded by IVT) , of whom 8 had good clinical outcome at follow-up. One patient was excluded from this subgroup analysis because optimal CVF could not be determined with certainty due to a movement artefact. After mechanical trombectomy with good reperfusion (mTICI 2b and 3), 4 out of 6 patients with good extent and fast CVF had good clinical outcome,
510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard 510560-L-sub01-bw-vdWijngaard Processed on: 2-6-2017 Processed on: 2-6-2017 Processed on: 2-6-2017
Processed on: 2-6-2017 PDF page: 157PDF page: 157PDF page: 157PDF page: 157
8
157
while 2 out of 7 patients with good reperfusion had good clinical outcome in case of poor extent and/or slow CVF. However, in patients with poor reperfusion (mTICI 0-2a) CVF status did not much alter clinical outcome, since none of 4 patients with good CVF and only 1 out of 4 patients with poor CVF had good clinical outcome. The risk ratio for good clinical outcome in the group with good CVF and good reperfusion status was 3.3 (95%CI 1.1-10.6, p<0.05) when compared with patient groups with poor CVF and/or poor reperfusion status combined. (See Supplemental Table V)
DISCUSSION
Our study showed that the assessment of extent and velocity of optimal cortical venous filling is useful to identify patients with acute middle cerebral artery stroke at higher risk of poor clinical outcome at 3-month follow-up. Previously, it was already demonstrated that extent of cortical venous drainage, as assessed with single-phase-CTA, predicted clinical outcome.7
By adding information about the velocity of optimal CVF, we showed that good extent of optimal CVF is a necessary but insufficient condition for good clinical outcome. A combined assessment of extent and velocity of CVF had additional prognostic value over established predictors of outcome.
Reduced venous drainage in acute ischemic stroke could be explained by multiple factors.16
First of all, the blood exiting the brain via the venous system matches the amount of blood entering the brain from the arterial system.17 Interestingly, the combined assessment
of extent and velocity of CVF had predictive value over the extent of arterial occlusion. Moreover, the differences in ICA filling times were small and had little influence on the interhemispheric difference in optimal CVF times. In experimental studies, additional mechanisms of narrowing of the venous lumen are active vein constriction,18 obstruction
by leukocyte–platelet aggregates19, 20 and compression of thin walled venules by edema.16, 17, 21 Poor venous outflow from the affected hemisphere is associated with poor arterial
collateral flow in animal studies.6 Our study confirms that venous assessment complements
the information obtained from arterial collateral assessment and plays an important role in