Posterior circulation collaterals as predictors of outcome in basilar artery occlusion: a sub-analysis of the BASICS randomized trial

Introduction and purpose Basilar artery occlusion (BAO) is still one of the most devastating neurological conditions associated with high morbidity and mortality. In the present study, we aimed to assess the role of posterior circulation collaterals as predictors of outcome in the BASICS trial and to compare two grading systems (BATMAN score and PC-CS) in terms of prognostic value. Methods We performed a sub-analysis of the BASICS trial. Baseline clinical and imaging variables were analyzed. For the imaging analysis, baseline CT and CTA were analyzed by a central core lab. Only those patients with good or moderate quality of baseline CTA and with confirmed BAO were included. Multivariable binary logistic regression analysis was used to test the independent association of clinical and imaging characteristics with a favorable outcome at 3 months (defined as a modified Rankin Score of ≤3). ROC curve analysis was used to assess and compare accuracy between the two collateral grading systems. Results The mean age was 67.0 (±12.5) years, 196 (65.3%) patients were males and the median NIHSS was 21.5 (IQR 11–35). Median NCCT pc-ASPECTS was 10 (IQR10-10) and median collateral scores for BATMAN and PC-CS were 8 (IQR 7–9) and 7 (IQR 6–8) respectively. Collateral scores were associated with favorable outcome at 3 months for both BATMAN and PC-CS but only with a modest accuracy on ROC curve analysis (AUC 0.62, 95% CI [0.55–0.69] and 0.67, 95% CI [0.60–0.74] respectively). Age (OR 0.97, 95% CI [0.95–1.00]), NIHSS (OR 0.91, 95% CI [0.89–0.94]) and collateral score (PC-CS – OR 1.2495% CI [1.02–1.51]) were independently associated with clinical outcome. Conclusion The two collateral grading systems presented modest prognostic accuracy. Only the PC-CS was independently associated with a favorable outcome at 3 months.


Introduction
Basilar artery occlusion (BAO) accounts for approximately 10% of strokes due to large vessel occlusion (LVO) and is present in about 8% of patients with symptomatic vertebrobasilar territory ischemia (1)(2)(3).Technical developments during the past few years have improved the diagnosis and management of BAO significantly, even so it is still one of the most devastating neurological conditions associated with high morbidity and mortality even in those treated with mechanical thrombectomy (4)(5)(6)(7).
Several clinical and imaging predictors of outcome after BAO, such as the National Institutes of Health Stroke Scale (NIHSS) score at admission, age, PC-ASPECTS, and recanalization have been identified in earlier studies (3,8).Collaterals have been recognized to influence recanalization, reperfusion, hemorrhagic transformation, and subsequent neurological outcomes, and are increasingly assessed in daily clinical practice, in patients with anterior circulation strokes (9).In contrast, the presence of posterior circulation collaterals in BAO patients is a feature that is yet far from being regularly used in clinical practice.
The noninvasive nature of CTA and its wide and rapid availability makes it the preferred diagnostic modality to assess collateral status in the acute stage of BAO.Two main collateral grading system exist for the posterior circulation, the posterior circulation collateral score (PC-CS) and the Basilar Artery on Computed Tomography Angiography (BATMAN) score (10,11).Only a few studies have examined the role of posterior circulation collaterals as a predictor of outcome and systematically compared the two grading systems (12,13).
In the present study, we aimed to assess the role of posterior circulation collaterals as predictors of outcome in the BASICS trial and to compare the two systems in terms of prognostic significance.

Methods
We performed a sub-analysis of the BASICS trial, a multicenter, open-label, international, randomized, controlled trial with blinded outcome assessment to test the efficacy and safety of endovascular therapy initiated within 6 h after the estimated time of basilar-artery occlusion with best medical therapy (14).The design of the trial has been described in detail elsewhere (15).
Baseline clinical variables included: age, sex, cardiovascular risk factors, baseline NIHSS, glycemia, blood pressure and IV rt-PA use.Stroke subtyping was done according to the TOAST classification and further dichotomized into cardioembolic and other (grouping large vessel atherosclerosis, other causes and uncertain).Estimated time of BAO, defined as time consistent with the clinical diagnosis of BAO on the judgment of the treating physician, was used to compute several time metrics including symptom onset-BAO, symptom onset-IV rtPA, symptom onset-EVT and BAO-recanalization.The primary outcome analyzed was a favorable clinical outcome, defined as modified Rankin score (mRS) ≤3 at 90 days.
For the imaging analysis baseline CT and CTA were analyzed by a central core lab.Patients with low quality baseline CTA were excluded from the analysis.Reperfusion was defined as a TICI2b-3 adjudicate by a central core lab.The pc-ASPECTS on NCCT and the pc-ASPECTS on CTA were quantified (16).All occluded vessel segments (including absence of posterior communicating arteries -Pcomms) were identified.Collaterals were graded according to two grading systems -the Basilar Artery on Computed Tomography Angiography (BATMAN) score and Posterior Circulation Collateral score (PC-CS).
The BATMAN Score is a 10-point semiquantitative scoring system based on CTA that was graded as follows: one point for each of the other segments giving a maximum score of 10: 1 point if either intracranial vertebral artery was patent; 1 point for each patent segment of the basilar artery, the proximal segment, extending from the vertebrobasilar junction to the origin of AICAs, the middle segment from the origins of AICAs to the origin of SCAs, and the rostral segment from the origin of SCAs to its rostral end; and 1 point for each patent P1 segment of PCA (10).
The PC-CS ranges from 0 to 10 and was graded as follows: each patent PICA, AICA, and SCA is allocated with 1 point.Identification of a PCoA is allocated with 1 point if its diameter is smaller than the ipsilateral P1 segment and 2 points if its diameter is equal or larger than the ipsilateral P1 segment.Fetal variants of the PCA, defined as a PCA arising from the anterior circulation and absence of P1 segments, are not included (11).
Both PC-PCs and BATMAN scores were dichotomized in good vs. poor collaterals as previously reported (10,11).For PC-CS good collaterals was defined as a score ≥ 6.For the BATMAN score good collaterals was defined as a score ≥ 7. We compared PC-ASPECTS (on both NCCT and CTA) at baseline and day 1 between patient with good and poor collaterals.

Statistical analysis
Categorical variables were reported as absolute numbers and percentages.Continuous variables were reported as mean ± standard ROC curve analysis was used to assess and compare accuracy between the two collateral grading systems.The area under the curve and 95% CI was calculated and used to compare the two curves.Binary logistic regression analysis was used to test the independent association of clinical and imaging characteristics with the primary outcome (mRS at 90 days).The variables in the univariable analysis associated with the primary outcome with a p-value ≤0.2 were included in the first model.Variables with known association with the outcome (age, NIHSS, IV rt-PA and endovascular treatment) and the predictors of interest (collaterals score) were selected a priori to be included in the final model.To avoid overfitting, only those variables with a p value ≤0.1 in the first model were kept in the final model.To test if posterior circulation collaterals have an effect modification of mechanical thrombectomy on the outcome, an interaction term (collateral score x mechanical thrombectomy) was added in the regression models.A p-value <0.05 was considered statistically significant.IBM SPSS version 25.0 software was used to perform all the statistical analyses.

Discussion
In the present study, both collateral scores were associated with outcomes at 3 months in patients with acute BAO.The two grading systems presented modest prognostic accuracy.Only the PC-CS was independently associated with a favorable outcome at 3 months.Although the BATMAN score did not reach statistical significance, both collateral grading systems performed numerically similar in terms of predicting outcome.
BAO is a disease with high morbidity and mortality.In the present study, patients presented with high NIHSS score (median 21.5, IQR 11-35) and most patients (N = 123, 59%) had an unfavorable outcome at 3 months.A similar proportion was observed in previous reports (4,17).In meta analyses of large series of patients with BAO, the main independent predictors of outcome were the following: stroke severity (assessed with the National Institutes of Health stroke scale), age, location and length of the occlusion, time-to-treatment, recanalization; and collaterals as seen on angiography (4,(18)(19)(20).Similarly, in the present study, all these variables, except for time-to-treatment and recanalization, were associated with outcomes at 3 months in the univariable analysis.However, after multivariable analysis, only age, NIHSS and collateral status (PC-CS) remained independently associated with the outcome.NCCT pc-ASPECTS showed a strong trend to be associated with the outcome.
Collateral status is strongly recognized as an important prognostic factor in anterior circulation large vessel occlusion strokes (9,21).The presence of posterior circulation collaterals in BAO patients is a feature that is still not being regularly used in clinical practice.This is probably related to the more recently proven benefit of mechanical thrombectomy in BAO and also the presence of two different grading systems that involve a more laborious evaluation when compared to the evaluation of leptomeningeal collaterals in the anterior circulation (5,6).
Retrograde blood flow from the anterior circulation through the posterior communicating arteries (PCoAs) may serve as a primary collateral pathway, potentially providing instant diversion of blood flow to the ischemic occipital and infratentorial regions (22).Secondary collateral pathways comprise arteriolar anastomoses from the inferior and posterior divisions of the middle cerebral artery (MCA), potentially supplying the posterior cerebral artery (PCA) and leptomeningeal anastomoses between the branches of the intradural segments of the vertebral arteries and branches of the basilar artery (23).
The noninvasive nature of CTA and its wide and rapid availability make it an attractive diagnostic modality to assess vascular anatomy in the acute stage and potentially to determine collateral status.The PC-CS allocate more weight on the availability of collateral pathways such as the PICA, AICA, SCA and the PcoA.On the other hand, BATMAN assigns more weight on the clot burden counting the occluded segments of the basilar and vertebral arteries and PCA, also evaluating collateral flow from the PcoA.Previous studies comparing the two grading systems have yielded conflicting results.The PC-CS and BATMAN scores showed a significant association with outcome in previous studies.Alemseged et al. reported the BATMAN score performing better in ROC curve analysis (BATMAN -AUC 0.8; 95% CI [0.7-0.9], vs. PC-CS -AUC, 0.63; 95% CI [0.5-0.7];p = 0.04).Additionally, the PC-CS was not independently associated with outcomes after adjusting for stroke severity (10).
However, in a similar study, Dias et al. reported the PC-CS performing better than BATMAN (AUC 0.80, 95% CI [0.62-0.98]and  (13).Different from previous reports, our results show both grading systems to be associated with a better outcome but performing only modestly on ROC curve analysis (AUC 0.67, 95% CI [0.60-0.74]and 0.62 95% CI [0.55-0.69]for the PC-CS and BATMAN score respectively).Novel BAO scores which quantify the amount of thrombus burden in posterior circulation arteries may have greater prognostic accuracy in patients with BAO as it has been shown for patients with anterior circulation stroke with the Clot Burden Score (24).
Our study has limitations.Poor quality admission CTA (34-11.7%)and uncertain/absent BAO (48-16%) were present in a substantial proportion of patients and these were excluded from the analysis.This might have limited statistical power of the analysis.Still, a large number of patients were included achieving a larger sample size than similar previous studies.Even with the use of multivariable analysis, there is still a possibility of residual confounding.However, the PC-CS grading system still predicted a favorable outcome after adjusting for important prognostic factors in BAO (IV rt-PA, endovascular treatment and pc-ASPECTS) including age and NIHSS which are the most important predictors of outcome in stroke.Also, although being applicable to a large proportion of patients, the two grading systems might not perform well enough in certain circumstances such as anatomic variations including fetal variants of the PCA and other persistent carotid-basilar anastomosis.

Conclusion
Collateral score assessment was significantly associated with outcomes in BAO even when adjusting for important clinical and imaging variables.This emphasizes the importance of collateral assessment in the posterior circulation, analogous to the more established role of collateral assessment in the anterior circulation.
The studies involving humans were approved by Conselho Nacional de Ética em Pesquisa (CONEP).CAAE: 66471717.6.2003.5040.The studies were conducted in accordance with the local legislation and institutional requirements.The participants provided their written informed consent to participate in this study.ROC curve comparing the BATMAN score and the PC-CS.
SD) or median ± interquartile range (IQR).Differences between proportions were assessed by the chi-square test or the Fisher's exact test as appropriate.Differences in continuous variables were assessed by the independent-samples T-test or the Mann-Whitney U test in case of non-normally distributed data.Univariable binary logistic regression was used to generate OR and 95% CI for the analysis of continuous variables.Odds ratios (OR) and 95% confidence intervals (CI) are reported in all tests.The Wilcoxon signed rank test was used to compared PC-ASPECTS (for both NCCT and CTA) at baseline and day 1 between patients with good and poor collaterals.

TABLE 2
Baseline imaging characteristics and outcome at 3 months.