Angiographic Assessment of Pial Collaterals as a Prognostic Indicator Following Intra-arterial Thrombolysis for Acute Ischemic Stroke

Patients and Techniques

Clinical information, arteriograms, and CT scans collected prospectively from 65 consecutive patients who underwent intra-arterial thrombolysis between November 1994 and April 2002 at the Ohio State University Medical Center were retrospectively reviewed to determine the extent of pial collateral formation and infarct volume. Patients gave written informed consent as part of a protocol that included screening angiography and thrombolytic treatment. The protocol was reviewed and approved by the Ohio State University Hospital Institutional Human Subjects Review Board, and the procedures followed were in accordance with institutional guidelines. Patients who were seen within 6 hours of symptoms of a carotid or vertebral basilar stroke and were considered thrombolytic candidates according to clinical, laboratory, CT, and angiographic criteria similar to PROACT underwent screening cerebral angiography (4). Unlike the PROACT study, the present study also included patients with ischemic stroke involving territories other than the middle cerebral artery (MCA), and patients >85 years of age. All patients seen between 0 and 3 hours were offered intravenous tissue plasminogen activator (tPA) as an alternative. Presentation and discharge National Institutes of Health Stroke Scale (NIHSS) score, discharge modified Rankin scale score, times from symptom onset until cerebral angiography, and demographic information were recorded. Some patients who met inclusion and exclusion criteria before angiography did not receive treatment either because the angiogram was negative or their symptoms resolved in the interim between clinical assessment and angiography.

Additional exclusion criteria for retrospective review of angiographic data included the following:

1. Posttreatment cross-sectional imaging study of the brain was not available.

2. Angiographic evaluation for pial collateral pathways did not include a delayed venous phase.

3. The internal carotid artery (ICA) was occluded and there was antegrade filling of the M1 segment via either posterior communicating artery (PCOMA), ophthalmic artery or anterior communicating artery (ACOMA) without anterior cerebral artery (ACA) or MCA occlusion or these pathways were not evaluated. Only patients whose collateral blood flow depended on pial vessels were studied.

4. A complication compromising pial collateral flow occurred such as extracranial dissection of the parent vessel or intraprocedural thromboembolic event.

Patients underwent angiography and thrombolytic treatment by using local anesthesia, aseptic technique, and digital fluoroscopic control in a biplanar angiographic suite. Sedatives were avoided where possible. Methods of thrombolytic therapy for patients included continuous infusion or pulse spray. Microcatheter positioning relative to the thrombus was determined by the operator but was typically in the thrombus. Mechanical methods for clot disruption such as balloon angioplasty or clot retrieval were not used. Thrombolytic agents infused in this study included urokinase (up to one million units), recombinant prourokinase (up to 6 mg), and tPA (up to 90 mg). Other end points for thrombolytic administration included: onset of intracranial hemorrhage, and recanalization or physician to terminate the procedure because of technical difficulties for further treatment or technical complication.

Angiograms of all patients were reviewed by an interventional neuroradiologist (G.A.C.) who was blinded to all clinical information during this review. Occlusion site was scored as follows: occlusions at the internal carotid artery were assigned a score of 1; occlusions at the M1, P1, or A1 segment were assigned a score of 2; occlusions at the M2, A2, or P2 segments were assigned a score of 3; occlusions at the P3, M3, or A3 segment were assigned a score of 4; and more distal occlusions were assigned a score of 5. Cerebral arterial segments are defined elsewhere (13). Scores of 1 and 2 were considered proximal and scores 3–5 distal.

Complete recanalization was defined as no visible occlusion on angiography following thrombolysis. Partial recanalization included patients with branch occlusion following thrombolysis and reconstitution of flow in at least one branch. No recanalization included patients with lack of reconstitution in flow.

Evaluation of pial collateral formation was based on retrograde contrast opacification of vessels within the occluded territory on delayed angiographic images. Collaterals were scored (Fig 1) as follows:

1. A score of 1 was assigned if collaterals reconstituted the distal portion of the occluded vessel segment (ie, if there was M1 segment occlusion the M1 segment distal to the occlusion reconstituted; Fig 2).

2. A score of 2 was assigned if collaterals reconstituted vessels in the proximal portion of the segment adjacent to the occluded vessel (ie, if there was M1 segment occlusion with reconstitution to the proximal M2 vessel segments).

3. A score of 3 was assigned if collaterals reconstituted vessels in the distal portion of the segment adjacent to the occluded vessel (ie, if there was M1 segment occlusion with reconstitution to the distal portion of the M2 vessel segments).

4. A score of 4 was assigned if collaterals reconstituted vessels two segments distal to the occluded vessel (ie, if there was M1 segment occlusion with reconstitution up to the M3 segment branches).

5. A score of 5 was assigned if there was little or no significant reconstitution of the territory of the occluded vessel.

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

Scoring of the anatomic extent of pial collateral blood flow from the ACA territory to the MCA territory during occlusion of the M1 segment. Scoring corresponds to the angiographically visible retrograde reconstitution of the MCA segments on the delayed venous phase. Each color is depicts the furthest extent of retrograde opacification depicted on anteroposterior cerebral angiograms for each pial collateral score.

• Download figure
• Open in new tab
• Download powerpoint

Fig 2.

Anteroposterior (A and C) and lateral (B and C) images from a left internal carotid artery angiogram obtained during the early (A and B) and delayed angiographic phases (C and D) in a patient with acute ischemic stroke due to occlusion at the M1 segment (arrow). Note the retrograde opacification of the MCA branches via pial collateral vessels extending from the ACA (arrowheads). Because there is reverse opacification of the MCA extending to the distal M1 segment, a pial collateral formation score of 1 was assigned.

For occlusion sites at the ICA bifurcation, the occluded vessel segment was assumed to include the proximal M1 and A1 segments. Therefore, scoring here was similar to an M1 occlusion where a score of 1 is assigned if there was reconstitution of the distal M1 segment. There is no possibility to reconstitute the distal ICA in an ICA bifurcation occlusion.

Acute infarction on 24–48-hour CT was defined as a new hypoattenuated region relative to pretreatment CT. Infarcted regions were traced out on axial cross-sectional images by a CAQ (certificate of added qualification)–certified neuroradiologist (G.A.C.), and areas were calculated by using IMPAX image analysis software (Agfa Corporation, Ridgefield Park, NJ). Infarct volume was calculated by summing the cross-sectional areas multiplied by section thickness.

Pial collaterals were scored for all angiograms, and infarct volumes for the first 25 patients were determined at two different interpretation sessions spaced at least 30 days to test for reliability. Intraobserver agreement was calculated, and in cases of pial score disagreement a third review determined scoring. The interpreter was blinded to all clinical and CT data during the interpretation sessions.

Statistical analyses were performed by using JMP5 software (SAS Institute, Cary, NC). A discharge NIHSS score of 42 was assigned to patients who died. Contingency analysis with two-tail Fisher exact test was used to determine whether there was significant differences in NIHSS score between patients with higher versus lower pial collateral scores. A Wilcoxon rank sum test was used to determine whether there was a difference in infarct volume between patients who had complete recanalization versus no or partial recanalization, dependent upon pial collateral score and occlusion site. Significant differences in discharge modified Rankin scale score among patients with complete recanalization depending on whether they had higher versus lower pial collateral scores were determined with a contingency analysis by using a two-tail Fisher exact test. A similar analysis was also attempted for the group of patients with no or partial recanalization. Differences in infarct volume and pial collateral score relative to time to treatment were analyzed by using Wilcoxon rank sum test to determine whether time to treatment was dependent upon the outcomes of infarct volume and pial collateral score. The influence of age and sex on pial collateral formation was also analyzed with Wilcoxon rank sum test and a two-tail Fisher exact test, respectively. Linear regression analyses were performed separately by using infarct volume and then NIHSS score as the outcome variable. To facilitate linear regression analysis by using discharge NIHSS score, NIHSS score was considered a continuous variable. Clinical outcome at the time of discharge was also analyzed with a multivariate logistic regression analysis by using modified Rankin scale score as the outcome variable. For all regression analyses the variables included age, sex, side of infarction (left or right), time to treatment, presence of hemorrhage on 24–48-hour CT, occlusion site, pial collateral score, and recanalization. Backward selection was then performed. The regression analyses were repeated after rejecting the variables with estimates having a probability of proving the null hypothesis >0.1 based on the Wald test. First, the variable with the largest P value for the Wald test was removed from the model, and a regression was run on the remaining variables. Then, a likelihood ratio test was performed, and the smaller model was tested for predictive power. This was continued with all of the remaining variables with α ≥ 0.1, until the smallest model was determined. During backward selection interaction between variables was evaluated. For all regression analyses, lack of multicollinearity among the independent variables was confirmed by using simple χ² tests.

Results

Of the initial 65 patients who underwent intra-arterial thrombolysis for acute thromboembolic ischemic stroke at our institution, 34 were given intra-arterial tPA, 28 urokinase, and three recombinant prourokinase. None of the patients chose to undergo intravenous tPA as an alternative to intra-arterial tPA. Twelve of 65 consecutive patients were excluded (five ICA, five MCA, one ACA, and one basilar artery). In four patients, no angiogram was available for review. There was ICA occlusion with antegrade filling of the MCA via the ACOMA and PCOMA in one patient, one patient had a technical complication (dissection of the vertebral artery), in three patients with ICA occlusion potential source of collateral formation via ACOMA or PCOMA was not evaluated, and in two patients delayed venous phase was not acquired. One patient with an angiogram had no CT available for review.

Fifty-three patients with both adequate angiographic data and infarct volume measurements available were entered into the study, 29 female and 24 male. The mean age was 65.5 years (σ [standard deviation] = 15.1 years; range, 33–90 years). Mean time to delivery of thrombolytic agent (symptom onset to initiation of thrombolytic agent delivery) was 262.2 minutes (σ = 89.5 minutes). Median NIHSS score at presentation was 16 (range, 4–30). Median NIHSS score at discharge was 10 (range, 0–31; four patients died). Median Rankin score before symptom onset was estimated to be 0 (range, 0–4); the median discharge Rankin score was 4 (range, 0–6). The mean infarct volume was 79.7 cm³ (σ = 103.8 cm³; range, 0–526.6 cm³).

Intraobserver difference in infarct volumes of 25 randomly selected patients was 7.8% (σ = 4.65%) with an intraobserver variation of 8.6%. Intraobserver assessment of pial collateral score was identical in 46 of 53 patients and different by 1 score point in seven patients (κ = 0.81; SE = 0.065).

Mean infarct volumes divided according to pial collateral score and recanalization are described in Table 1. It is clear that more favorable collateral scores resulted in lower infarct volumes. If one considers only the patients with no infarction, all had pial collateral scores of 1, despite no recanalization. It is also clear that infarct volumes were lower in patients with complete recanalization, regardless of pial collateral grades.

Median presentation NIHSS scores, discharge NIHSS scores, changes in NIHSS score, and discharge-modified Rankin scale scores relative to pial collateral score and recanalization rate are presented on Table 2. Significantly more patients had an NIHSS score <15 on presentation among the group with pial collateral score of 1 or 2 versus those with pial collateral scores of 3–5 (P = .0217; two-tail Fisher exact test). Complete recanalization is shown to be necessary for a better outcome, as seen in the 21 patients with no or partial recanalization with a pial collateral score of 1 or 2, all of whom had poor discharge scores. Among patients with complete recanalization, those who had a pial collateral score of 1 or 2 were more likely to have a discharge-modified Rankin score <3 compared with patients with pial collateral scores of 3–5 (P = .0131; two-tail Fisher exact test). No patients with pial collateral scores of 4 or 5 had modified discharge Rankin scores better than 4, regardless of recanalization. Infarct volumes were smaller in patients with better pial collateral scores (1 or 2) relative to those with poorer pial collateral scores (3–5) when occlusion site and recanalization rate were taken into account, as shown in Table 3.

Mean infarct volume for the 10 patients whose treatment began <200 minutes after symptom onset was 24.03 cm³ (σ = 28.4 cm³) compared with 92.6 cm³ (σ = 110.8 cm³) in the other 43 patients whose treatment began after 200 minutes (P = .0210; Wilcoxon rank sum test). Regression analyses, however, did not identify a statistically significant association (P > .05) between time to treatment and change infarct volume or clinical outcome when tested against other predictors of outcome, such as recanalization or pial collateral score.

Extent of pial collaterals increased linearly with age (P = .0214, Wilcoxon rank sums test). The mean age of those patients with pial collateral scores of 1 or 2 was 64.7 years (σ = 13.9), whereas the mean age of patients with pial collateral score of 3–5 was 67.2 years (σ = 17.6). This was not statistically significant (P = .33). Percentage of female sex increased from pial collateral grade 1 to 5 as follows: grade 1, 48.2%; grade 2, 55.6%; grade 3, 60.0%; grade 4, 66.7%; and grade 5, 75.0%. Women with pial collateral scores of 1 or 2 accounted for 50.0%, whereas female patients with pial collateral scores of 3–5 totaled 64.5%, which was not statistically significant (P = .38, two-tailed Fisher exact test). The average age was 65.5 years for women and 65.5 years for men, and the median presenting NIHSS scores were 16.0 for women and 11.0 for men.

Linear regression analysis for infarct volumes is shown in Table 4, and demonstrates that age per year, high collateral score, distal occlusion site, and complete recanalization are all significant predictors of infarct volume. Age per year and high collateral score (3–5) are shown to increase infarct volume, whereas distal occlusion site and complete recanalization decrease it significantly. Linear regression analysis also identified age, high pial collateral score, and complete recanalization as significant predictors of the outcome variable, discharge NIHSS, as seen in Table 5. Estimates indicated that high collateral score (3–5) increased discharge NIHSS score, whereas complete recanalization decreased discharge NIHSS score. Logistic regression analysis for poor outcome at the time of patient discharge on the basis of modified Rankin scale score >2 is presented in Table 6. Predictive factors for outcome, based on the modified Rankin score, are age per year, high collateral score, and complete recanalization. High collateral score and complete recanalization improves outcome, whereas age per year slightly worsens outcome.

Discussion

Data presented here indicate that an anatomically based angiographic scoring method that evaluates the extent of pial collateral formation can help predict infarct volume and clinical outcome in patients undergoing thrombolytic treatment for acute stroke. It is evident from this study that patients with favorable pial collateral scores are more likely to have a good outcome at the time of discharge. Modified Rankin discharge scores were the best among patients with both recanalization as well as good pial collateral score (Table 2). All patients with modified Rankin scores of 0 at discharge had good pial collateral scores, and most of these patients also had complete recanalization. A modified Rankin discharge score is, however, shown to be worse among the patients with poorer pial collateral score (4 and 5), regardless of recanalization. It is true, though, that there is a clear benefit to intra-arterial thrombolysis because infarct volume decreases with recanalization even with unfavorable pial collateral scores (Table 1).

Estimates derived from regression analysis models indicate that extent of pial collateral formation has at least as much influence on infarct size and clinical outcome as recanalization does. Furthermore, the presence of good pial collaterals and complete recanalization has a synergistic effect. Other predictors examined in this study did not affect outcomes as significantly as these two variables. This suggests that treatments enhancing collateral formation during acute stroke can reduce infarct volume, whereas treatments restricting collateral vasculature hypothetically enlarge infarct size. Clinical outcome was substantially better in patients with complete recanalization and good pial collateral formation (ie, 1 or 2). It therefore follows that the assessment of pial collateral blood flow on angiography can provide a time efficient, reproducible method to assess the ischemic penumbra at the time of intra-arterial thrombolysis.

Conclusion

Findings presented here underscore the importance of evaluating the anatomic extent of pial collateral formation on angiography when assessing the expected outcome in patients undergoing intra-arterial thrombolytic treatment for acute ischemic stroke. Angiographic assessment of the anatomic extent of pial collateral formation can provide reliable ordinal data in a time-efficient method predictive of outcome in patients undergoing intra-arterial thrombolysis for acute ischemic stroke independent of other predictors. This information is especially useful when cross-sectional perfusion studies are not available. Relative to other predictive factors, the extent of pial collateral formation has a significant impact on stroke volume and immediate clinical outcome. In particular, this study defines pial collaterals in terms of anatomic extent on angiography rather than a more subjective assessment of pial collaterals as “good” or “bad” as done in previous studies. Evaluation of pial collaterals may play a dominant role when developing an outcomes grading scheme in acute stroke. Results presented here indicate that good pial collateral formation is a predictor of outcome independent of other factors including recanalization, site of occlusion, age, or time to treatment. It is therefore no surprise that thrombolysis should act to improve clinical outcome at the time of hospital discharge and infarct volume, whether collaterals are poor or good. It is also evident that patients with excellent pial collaterals will have better outcomes than patients with poor pial collaterals, regardless of recanalization.