Association of Cortical Vein Filling with Clot Location and Clinical Outcomes in Acute Ischaemic Stroke Patients

Delay in cortical vein filling during the late-venous phase (delayed-LCVF) is characterized by opacification of cerebral veins despite contrast clearance from contralateral veins on dynamic computed tomography angiography (dCTA) in acute ischemic stroke (AIS) patients. The aim of the study was to investigate the associations of delayed-LCVF with clot location, reperfusion status at 24 hours, and 90-days functional outcome in AIS patients who received reperfusion therapy. A prospective cohort of AIS patients treated with intravenous thrombolysis was studied. Groupwise comparison, univariate, and multivariate regression analyses were used to study the association of delayed-LCVF with clot location and clinical outcomes. Of 93 patients (mean age = 72 ± 12 years) with hemispheric AIS included in the study, 46 (49%) demonstrated delayed-LCVF. Patients with delayed-LCVF demonstrated a significantly higher proportion of proximal occlusion (72% vs 13%, P =< 0.0001), and poor reperfusion at 24 hours (41% vs 11%, P = 0.001). The proportion of poor functional outcome at 90 days was not significantly different (22/56 (48%) vs 17/61 (36%), P = 0.297). The appearance of delayed-LCVF on baseline dCTA may be a surrogate for large vessel occlusion, and an early marker for poor 24-hour angiographic reperfusion.

Baseline characteristics. Of the 154 patients, 93 (60.4%) patients (mean age = 71.6 ± 12.4 years; the number of females = 49 (52.7%)) with acute anterior circulation vessel occlusion who received IVT met the inclusion criteria (Table 1). Figure 3 shows the distribution of patients for different clot locations, reperfusion status at 24 hours and functional outcome at 90 days. Out of 93 patients with hemispheric ischaemic stroke included in the study, 46 (49.5%) patients showed delayed-LCVF. The median NIHSS score at admission and 24 hours were 14 (IQR = 8) and 7 (IQR = 9), respectively. Fifty-four (58%) patients demonstrated good functional outcome (mRS 0-2) at 90 days and 52.7% of patients demonstrated major reperfusion at 24 hours. The average OTT for this AIS cohort was 162.6 ( ± 83.9) minutes. The summary of clinical outcomes for all patients is shown in Table 2.
Associations with clot location. Results of univariate logistic regression analysis for association with proximal clot is shown in Supplementary Table 2. Independent variables (with P ≤ 0.1) and other important covariates (NIHSS at admission) were used for stepwise backward multivariate logistic regression analysis to study the association with incidence of the proximal clot (see Table 3; Model 3A). Finally, TPME, dyslipidemia, and delayed-LCVF were retained in the final multivariate logistic regression model and treated as potential confounders (Table 3; Model 3C). Higher rates of delayed-LCVF (OR = 106.62; 95% CI = [15,756]; P = < 0.0001) and dyslipidemia (OR = 5.8; 95% OR = [1. 6,21]; P = 0.007) were positively associated with incidence of proximal clot. Interestingly, each unit increase in TPME was negatively associated with presence of proximal clot (OR = 0.56; 95% CI = [0.35, 0.9]; P = 0.02). The model showed good discrimination ability with an area under the receiver operating characteristic (ROC) curve of 0.89 (sensitivity = 82%, specificity = 81.5%). We also compared the reduced models with and without the inclusion of delayed-LCVF (Model 3C (reduced Delayed-LCVF association with reperfusion status at 24 hours, and functional outcome at 90 days. Patients with delayed-LCVF demonstrated a significantly higher proportion of poor angiographic reperfusion at 24 hours (68% vs 31.8%, P = 0.001) ( Table 2). Moreover, bivariate logistic regression analysis also revealed a significant association of delayed-LCVF with overall angiographic reperfusion status at 24 hours (P = < 0.001). Delayed-LCVF was positively associated with poor reperfusion at 24 hours (OR = 4.4; 95% CI = [1.8, 10.6]; P = 0.001). Independent variables (with P ≤ 0.1; age, NIHSS at admission, acute core volume, collateral status, delayed-LCVF, clot location and hypertension) (see Supplementary Table 3) were used for stepwise backward multivariate logistic regression analysis to study the association with poor angiographic reperfusion (see Table 4; Model 4 A). In the reduced model (Table 4; Model 4 C), delayed-LCVF, clot location, and baseline core volume were retained. Higher rates of delayed-LCVF (OR = 3.7; 95% CI = [1.2, 11.28]; P = 0.021), and increasing acute core volume (OR = 1.02; 95% OR = [1, 1.04]; P = 0.021) were significantly associated with poor angiographic reperfusion at 24 hours. Comparison between the reduced models with and without inclusion of delayed-LCVF (Model 4C (reduced model with delayed-LCVF) vs Model 4B (reduced model without delayed-LCVF)) revealed that the addition of delayed-LCVF to the model revealed no added advantage on discriminative accuracy: BIC (BIC Model4C Table 2). To study the association with functional outcome at 90 days using backward stepwise multivariate regression analyses, independent variables (with P ≤ 0.1; NIHSS at admission, baseline core volume, penumbra, dyslipidaemia, and reperfusion status at 24 hours (see Supplementary Table 1) and other important covariates (clot location, delayed-LCVF) were included in the final multivariate regression model (Table 5; Model 5A). In the reduced multivariate regression model (Model 5C), delayed-LCVF was not significantly associated with functional outcome at 90 days, when adjusted for NIHSS at admission, acute core volume, clot location, and 24-hour reperfusion status (Table 3). Increasing acute core volume and reperfusion status at 24 hours were significantly associated with poor functional outcome at 90 days. Comparison of model characteristics between the two models, with and without delayed-LCVF, revealed addition of delayed-LCVF significantly improved the discriminative accuracy: BIC (BIC Model5C vs BIC Model5B = 92.5 vs 116), sensitivity (Sensitivity Model5C vs Sensitivity Model5B = 82% vs 69%), or discrimination accuracy (ROC Model5C vs ROC Model5B = 0.93 vs 0.80; Specificity Model5C vs Specificity Model5B = 88.7% vs 79.25%; PPV Model4C vs PPV Model4B = 84.21% vs 71%) ( Table 5).

Discussion
In this study, we sought to investigate associations of novel cortical vein filling pattern, observed during the late venous phase on time-resolved dCTA, with the acute core, tissue at risk, clot location, reperfusion status at 24 hours, and long-term functional outcomes in a cohort of AIS patients treated with IVT. In the current study, we found that the patients with proximal (thrombus in ICA and/or M1P) occlusion are at significantly higher risk of showing delayed-LCVF. We also noted a strong association of delayed-LCVF with poor reperfusion status at 24 hours. We could not demonstrate that the impact of delayed-LCVF on these acute outcomes translated into poor functional outcomes at 90 days, but this may be due to our small sample size and the fact that functional outcome is quite distal in the causal chain we are investigating. The inclusion of delayed-LCVF to the multivariate model significantly improved the predictive accuracy of poor functional outcome. We found no association of delayed-LCVF with ischemic infarct core or tissue at risk. Previously, we reported the presence of delayed cortical vein filling pattern in late-venous phase on dCTA in a cohort of AIS patients, where delayed-LCVF was found to be independently associated with poor baseline arterial collaterals and delay in maximised collateral enhancement 1 . There are limited studies on the role of cortical veins in stroke pathophysiology, and their associations with thrombus location and clinical outcome 13,17,[28][29][30][31] .
Our findings indicate that delayed-LCVF was significantly more common in patients with M1P and/or ICA occlusion. Delayed-LCVF showed significant improvement in discriminative accuracy when it was added to the multivariate regression model. As such, the appearance of delayed-LCVF on dCTA is a surrogate for proximal large vessel occlusion which is known to be associated with poor outcome. Previous studies have reported that AIS patients with thrombus in M1D, M2, and M3 segments are more likely to undergo recanalization than those with M1P and ICA occlusions 18,27 . Large vessel occlusions are less likely to be recanalised after IVT and are more likely to have poor clinical outcomes. The fact that delayed-LCVF is strongly associated with proximal thrombus may be used as an important parameter towards stroke prognostication and selection of patients for IVT. We postulate that proximal large vessel occlusions may lead to delayed-late cortical vein filling. In this study, we found that patients who showed delayed-LCVF on baseline dCTA demonstrated the significantly higher rate of poor angiographic reperfusion at 24 hours. This is clinically relevant suggesting that patients with the delayed-LCVF pattern on baseline dCTA will be poorly reperfused despite IVT. However, the addition of delayed-LCVF to multivariate regression model didn't improve the predictive accuracy of the model over and above clot location. Our results show that delayed-LCVF is a statistically significant prognostic indicator of early angiographic reperfusion (at 24 hours). From the current literature, we know that reperfusion status is a significant predictor of long-term (90 days) outcome 32,33 . In a multivariate logistic regression analysis, delayed-LCVF was not a significant covariate for predicting 90 days' functional outcome. Although, our point estimate supports an effect on 90 days mRS (OR = 1.6), we do not have sufficient power to demonstrate this at a statistically significant level. We hypothesize that patients with the delayed-LCVF pattern on baseline dCTA may show an unfavourable trajectory. Interestingly, the addition of delayed-LCVF significantly improved the predictive accuracy of functional outcome at 90 days. The presence of delayed-LCVF may aid in identifying patients at risk of 24-hour poor angiographic reperfusion. Other studies have also shown significant association of cerebral-venous flow with prognosis in stroke in both animals and humans 13,17,34 . Interestingly, animal studies focussing primarily on early and mid-venous phase have shown that the presence of cortical vein filling after ischaemic stroke was associated with decreased severity of hemiparesis and lower infarct volumes. This led to the hypothesis that the cortical venous flow may produce favourable outcomes as it would be more commonly prevalent in strokes with good collaterals 34 . Another study on humans also found that the cortical venous drainage was associated with good clinical outcomes 17 .
Imaging biomarkers towards the identification of patients who might benefit from early reperfusion therapy and guiding early intervention options to limit or even freeze infarct progression is crucial for strategies in acute stroke treatment 35,36 . The advent of cutting-edge next generation 320-detector row 640-slice multi-detector    CT (MDCT) scanners have facilitated the acquisition of whole-brain, sub-second, and volumetric acquisition of 4D-dCTA data 1,37,38 . CTP/CTA is not an invasive procedure compared to digital subtraction angiography (DSA), and is routinely obtained during clinical care of stroke patients 8,32,33,39 , has proven to be of added clinical utility in the early evaluation of stroke, facilitating precise localization of site of occlusion 40 , and identification of hypoperfused territory at risk of infarction 41 . The CTP was acquired simultaneously with the CTA with the use of same contrast bolus 42 . Dynamic CTA allows evaluation of intracranial vasculature and visualisation of contrast flow from its arterial to venous phases. Using appropriate reconstructions of dCTA using MIP algorithm, we investigated various stages of venous drainage and downstream venous dynamics; including the assessment of delayed-LCVF appearance in late venous phase. Assessment of impaired cortical venous drainage may provide valuable information over and above arterial collateral assessment, and the presence of delayed-LCVF could have a role in making informed decisions on patient management and prognosis. We understand that our study has several limitations, including small sample size and the variability in the cortical venous structures. Since the publication of MR CLEAN, REVASCAT 43,44 , EXTEND-IA, ESCAPE, and SWIFT PRIME, the standard practice now includes intravenous thrombolysis when possible, complementing thrombectomy [43][44][45][46] . However, endovascular treatment was not available at our centre at the time of the study. We tried to account for small sample size by using the Wilcoxon-Rank test, which would be conservative in this case. We acknowledge that additional occlusions distal to the M1 segment and variations in the prominences of M2, M3, and M4 trunks may have an impact on the assessment of delay in maximised enhancement. Moreover, it may also be influenced by the differences in the filling time of collaterals in different areas of the MCA territory.

Figure 3. Clinical outcome (reperfusion status at 24 hours, and 3-month modified Rankin score (mRS)) and clot location for presence and absence of delayed-LCVF.
Good clinical outcome was defined as mRS 0-2 at 3 months. To conclude, in this study, we sought to study the associations of delayed-LCVF with core volume, tissue at risk, clot location, and clinical outcome (vis a vis reperfusion at 24 hours, functional outcome (in terms of modified Rankin score (mRS)) at 90 days) in a prospective cohort of AIS patients who received intravenous thrombolytic therapy. Endovascular procedures or mechanical thrombectomy was not available at our centre at the time of the study. Delayed-LCVF patterns were more commonly seen in proximal thrombus occlusion in M1P and/ or ICA. Based on these findings, the appearance of delayed-LCVF on dCTA can be used as a surrogate for proximal thrombus or large vessel occlusion. It may also be useful in identifying patients at risk of poor angiographic reperfusion at 24 hours. Moreover, given the propensity of proximal thrombus towards poor clinical outcome after IVT 47 , and significant association of delayed-LCVF with both proximal thrombus and poor reperfusion at 24 hours, we postulate that the AIS patients with delayed-LCVF may progress unfavourably, and therefore alternate revascularisation strategies may be considered. We also found that addition of delayed-LCVF significantly improves predictive accuracy of 90 days' functional outcome. However, in the present cohort, we acknowledge that delayed-LCVF was not found to be a determinant factor in predicting functional outcome at 90 days. In light of the paucity of literature on the cerebral venous system and their role in stroke prognostication, we believe this study may be of clinical relevance towards understanding the role of cerebro-venous system, in particular, cortical vein, in the prognosis of AIS patients. We propose that delayed-LCVF is a marker that will allow clinicians to extract more prognostic information from imaging that is already routinely acquired. Also, in combination with NCCT, CTA/CTP can be rapidly obtained with minimal delay in treatment, and is widely available in emergency departments, and is well tolerated 41 . Dynamic CTA is a promising technique for the dynamic assessment of the cerebral vasculature. We caution that these results must be understood as preliminary and within the context of the study design. Further prospective studies are recommended to study the role of cortical veins in stroke prognostication.

Materials and Methods
Study design and patient selection. Consecutive acute ischaemic stroke patients admitted to the comprehensive stroke unit, Department of Neurology at our academic medical centre were prospectively studied provided they satisfied the following inclusion criteria: (a) aged 18 and above years, (b) acute anterior circulation vessel occlusion followed by IVT, (c) hemispheric stroke, and (d) dCTA data available at baseline and 24 hours. Patients without identifiable thrombus on the baseline dCTA were excluded. Patients received 0.9 mg/kg intravenous recombinant tissue plasminogen activator (rtPA). Baseline clinical characteristics included age, sex, and clinical risk factors (hypertension, diabetes, dyslipidaemia, history of smoking (past/present), atrial fibrillation Model 3A   (AF), depression and history of stroke and/or transient ischemic attack (TIA)). Clinical data were procured from the patient records. National Institutes of Health Stroke Scale (NIHSS) scores at the time of initiation of the rtPA and at 24 hours were obtained. The time delay between stroke onset and administration of tPA (time to tPA) or onset to treatment (OTT) was also recorded. Management of patients was in accordance with local guidelines and as per the discretion of the treating stroke physician. This study was approved by the Hunter New England Human Research Ethics Committee (HNEHREC, Newcastle, NSW) in accordance with the National Statement on Ethical Conduct in Human Research 2007. All methods were carried out in accordance with the approved guidelines. Informed consent was obtained from the patient in accordance with the Declaration of Helsinki. All the patients underwent non-contrast CT (NCCT), CT Perfusion (CTP) and CT angiography (CTA) at baseline and follow-up (24 h) NCCT, CTA, and magnetic resonance imaging (MRI), following our routine stroke imaging protocol 1 . Volumes of the acute perfusion lesion (relative delay time (DT) ≥ 3 seconds) and acute infarct core (relative CBF ≤ 30%) were calculated using previously validated thresholds 8,48 . Penumbra volume was defined as the volume of the perfusion lesion (DT threshold ≥ 3 seconds) minus the volume of the infarct core (relative CBF threshold < 30% within the DT ≥ 3 sec lesion). The threshold of DT ≥ 3 seconds was based on previous studies 48,49 .
Maximum  (M1D) 27 . M1 MCA was defined as a vessel extending from the ICA bifurcation to the origin of the first major branch in the Sylvian sulcus. Delayed-LCVF was identified by late venous phase opacification of cortical veins despite contrast clearance from contralateral cortical veins on maximum intensity projection (MIP) images from dCTA. The time to peak of maximum arterial enhancement (TPME) was also recorded 1 . Collateral grading was done to assess the morphological status using dCTA data based on the degree of reconstitution of the MCA up to the distal end of its occlusion. Collateral grading was classified as 'good' , 'reduced' or 'poor' using the Miteff scale 6,51 . Good collateral grading was assigned if the entire MCA distal to the occluded segment was reconstituted, i.e., if collaterals reconstituted vessels in the: (a) distal portion of the occluded vessel, or (b) proximal portion of the segment adjacent to the occluded vessel (e.g., if there was proximal M1 occlusion, the distal M1 or proximal M2 segments reconstituted) 6,51 . Collateral grading was assigned "poor" status if the reconstitution of the distal MCA was only partial, i.e., if collaterals reconstituted vessels in the: (a) distal portion of the segment adjacent to the occluded vessel, or (b) two segments distal to the occluded vessel, or (c) little or no significant reconstitution of the territory of the occluded vessel 6,51 .  Outcome measures. The modified Rankin Scale (mRS) was used to assess clinical outcome in terms of functional status at 3 months. Patient outcomes were dichotomized into good (mRS 0-2) versus poor/bad (mRS 3-6). Angiographic assessment of the degree of reperfusion was done by an independent blinded reviewer on a repeat CTA acquired at 24 hours using modified thrombolysis in cerebral infarction (mTICI) score 52 . An mTICI grade of 2a, defined as tissue reperfusion in < 50% of the occluded artery territory, was identified as partial reperfusion. Major reperfusion corresponded to tissue reperfusion in ≥ 50% of the occluded artery territory with grades of 2b or 3 on the mTICI scale. All patients with partial or nil angiographic reperfusion at 24 hours were lumped together into "poor reperfusion" category. Penumbral salvage was defined as the difference between the acute CTP lesion volume (PWI lesion) and the 24-hour DWI lesion volume 53,54 . We identified penumbral salvage in patients where 24-hour DWI lesion volume was smaller than the acute perfusion lesion volume. Percentage of penumbra salvaged was defined as (penumbral salvage volume/penumbra volume) × 100 55 .
Statistical analysis. All the statistical analyses were performed using STATA (Version 10, 2001; College Station, TX, USA). Numerical values given are the means (± standard deviation) or medians (interquartile range) for age, core and penumbra volumes, mRS scores, NIHSS at admission, NIHSS at 24 hours, and change in NIHSS scores as appropriate. For ordinal or continuous data, Mann-Whitney (Wilcoxon rank-sum) test was used.
Nominal data were analysed with the Pearson's chi-squared (χ 2 ) and the 2-tailed Fisher exact test. Groupwise comparison was made between the patients with and without delayed-LCVF. Group differences were considered significant at values of P < 0.05. To test the independent association of significant variables with delayed-LCVF, logistic regression models were fitted. Baseline infarct lesion volume and penumbra volume was dichotomized into small (≤ 25 mL) or large (> 25 mL) pertinent to the findings on 25 mL threshold of core volume that accurately predicted the presence of penumbra (tissue at risk), response to thrombolysis, and excellent outcome 56,57 . A stepwise backwards multivariate logistic regression analyses was used to study the association of delayed-LCVF with clot location, 24-hour angiographic reperfusion status, and 90 days functional outcome. Independent variables with P ≤ 0.1 (on univariate regression) and other important clinical covariates were included in the multivariate logistic regression model. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the overall rate of correct classification for the multivariate models were estimated. Finally, the receiver operating characteristic (ROC) curve for the regression model was plotted, and the area under the curve was computed to evaluate the discriminative ability.