Predictors of functional outcome vary by the hemisphere of involvement in major ischemic stroke treated with intra-arterial therapy: a retrospective cohort study

This retrospective cohort study examined 35 consecutive left-hemispheric and 35 consecutive right-hemispheric stroke patients with anterior circulation PAO who underwent IAT at Massachusetts General Hospital from June 2004 to August 2008. Inclusion criteria were: (1) acute occlusion of the intracranial ICA and/or proximal MCA, including stem (M1) and proximal branch lesions (M2), on CTA; (2) treatment with IAT; and (3) available clinical follow-up at three months. Medical records were reviewed for clinical data. Our institutional review board approved the study.

All subjects underwent NCCT followed by CTA of the head and neck performed on a 16- or 64-slice multi-detector helical scanner (GE Medical Systems, Waukesha, WI). NCCT was performed with 120-140kVp, 170mA, and 5-mm axial slice thickness. For CTA, nonionic low-osmolar iodinated contrast media (Isovue, 370 mg iodine/mL; Bracco Diagnostics, Princeton, NJ) was injected via a peripheral intravenous catheter at 3.5 cc/s for 100 cc, followed by 40 cc saline at 4.0 cc/s. Images were acquired from the vertex to the aortic arch after an 8-10s delay, using 140 kVp, auto-mA (300-715 mA), 0.5s rotation time, 1.25 mm helical thickness, 0.625 mm interval, 0.938:1 pitch, 9.37 mm/rotation speed, 22 cm FOV. CTA images were reconstructed in standard algorithm at 2.5 mm.

The ASPECTS grading system was applied to the pre-treatment NCCT and CTA-SI for each patient[25]. The ten brain regions (caudate, lentiform nucleus, internal capsule, insula, and six cortical regions) used to calculate the ASPECTS score were inspected using the entire set of images for each study. One point was deducted for each affected region from a total possible score of ten. Therefore, lower scores indicated more extensive ischemic change. Readers were blinded to all clinical information except for stroke side. Window width and level settings were selected to optimize contrast between normal and hypodense (ischemic) brain parenchyma. All imaging was evaluated independently by three neuroradiologists (AY, JR, PS), and differences were resolved by consensus.

Follow-up imaging from 12-72 hours after IAT was reviewed to evaluate for hemorrhagic transformation based on ECASS definitions: HI1 (small petechiae without mass effect), HI2 (more confluent petechiae without mass effect), PH1 (<30% of infarct with mild mass effect), PH2 (>30% of infarct with significant mass effect)[26].

The baseline NIHSS score was used to assess the neurologic status of stroke patients in the emergency department. Three-month mRS, the primary clinical endpoint, was assessed by one of the stroke neurologists or neurointerventionalists, who were not blinded to clinical information. An mRS ≤ 2 was considered a good outcome.

Univariate analysis of clinical, imaging and outcome data was performed comparing left- and right-hemispheric stroke patients. Subsequently, univariate analysis comparing good and poor outcome was performed for each hemisphere. The Fisher's exact test was used for categorical variables. The Student's t test was used for normally-distributed continuous variables. The Mann-Whitney U test was used for ordinal data. A stepwise multiple logistic regression was performed to identify independent predictors of outcome. Normality was tested with the Kolmogorov-Smirnov test. Statistical analysis was performed utilizing the SAS 9.1 software package (SAS Institute Inc., Cary, NC) and MedCalc^® 10.0 Software.

Table 1 lists the baseline and post-treatment characteristics of the entire study population. Thirty-eight of the 70 patients were females (54.3%); mean age was 67.1 ± 18.4 years; median NIHSS score was 18 (interquartile range IQR, 14-21). There were 21 terminal ICA occlusions, all extending into the proximal MCA. There were 49 MCA occlusions: 40 involved the M1 segment and 9 involved an M2 segment. The median NCCT ASPECTS was 8 (IQR, 6-9). The median CTA-SI ASPECTS was 5 (IQR, 3-7).

The median Mori reperfusion score was 2 (IQR, 1-3). The mean time to reperfusion or procedure termination was 411.2 ± 153.2 minutes. Nineteen patients (27.1%) had a good outcome (mRS ≤ 2) at three months. Twenty-five patients (35.7%) died. The median mRS was 4 (IQR, 2-6).

Independent predictors of good three-month outcome for the entire cohort were Mori reperfusion score (O.R. 11.4, 95% C.I. 2.84-45.5, p < 0.0001), age (O.R. 0.93, 95% C.I. 0.88-0.98, P = 0.0004), and NIHSS score (O.R. 0.79, 95% C.I. 0.64-0.98, p = 0.02). NCCT ASPECTS and CTA-SI ASPECTS were not statistically different between patients with good and poor outcome.

Of sixty-seven patients with post-treatment imaging, hemorrhagic transformation was detected in 39 patients (58.2%): 27 HI1, 8 HI2, 1 PH1 and 3 PH2. There was a 4.5% rate of significant hemorrhagic transformation (PH2), which is within reported rates for IAT[9, 11, 20]. Of three patients with PH2, one had a left M1 occlusion with mRS of 5, one had a right M1 occlusion with mRS of 3, and one had a right M1 occlusion and died.

Table 2 compares left- and right-hemispheric strokes for baseline variables, reperfusion and clinical outcome. Left-sided strokes had higher NIHSS scores (p = 0.02) and lower admission SBP (p = 0.009). The remaining baseline variables were not statistically different, including age, occlusion level, and ASPECTS.

There were no significant differences in the degree of reperfusion, clinical outcome and hemorrhage rate between the two hemispheres. The median Mori score for both sides was 2 (IQR, 1-3; P = 0.44). Nine of 35 patients (25.7%) with left-sided strokes versus 10 of 35 patients (28.6%) with right-sided strokes had a good outcome (p = 0.99). Hemorrhage occurred in 17/33 (51.5%) left-sided strokes and 22/35 (62.9%) right-sided strokes (p = 0.46).

For left-sided strokes, univariate predictors of outcome were lower NIHSS score (p = 0.0008) and greater reperfusion (p = 0.003, Table 3). There was a strong statistical trend for higher NCCT ASPECTS (p = 0.05) and CTA-SI ASPECTS (p = 0.06) in patients with good outcomes (Figure 1). Independent predictors of good outcome were NIHSS score (O.R. 0.51, 95% C.I. 0.28-0.92, p = 0.0002) and Mori reperfusion score (O.R. 21.3, 95% C.I. 1.13-500, p = 0.006).

For right-sided strokes, univariate predictors of outcome were younger age (p = 0.009) and greater reperfusion (p = 0.006, Table 4). There was no association of pre-treatment ASPECTS with outcome (Figure 2). Independent predictors of good outcome were age (O.R. 0.89, 95% C.I. 0.80-0.98, p = 0.002) and Mori score (O.R. 12.0, 95% C.I. 1.40-100, p = 0.003).

Numerous studies have shown that baseline NIHSS score predicts outcome in stroke patients who receive IV and/or IA therapies[13, 15, 19, 22]. In our study, it was an independent predictor of outcome for the entire cohort. However, when analyzing outcome by side, NIHSS score was an independent predictor of outcome only for left-sided strokes. By contrast, the VISTA analysis demonstrated a significant correlation between NIHSS and outcome for both hemispheres[3]. The likely explanation is that over a broad range of stroke severity (as seen in the VISTA database) NIHSS score is a reasonable surrogate for occlusion level, even for right-sided strokes. For a homogeneous population of large vessel occlusions (as in our study), NIHSS score may not adequately discriminate the amount of territory at risk in the right hemisphere. This last point is supported by numerous imaging studies that have demonstrated that patients with right-sided strokes may have low NIHSS scores despite substantial infarct volumes [4‐6].

Similarly, age was an independent predictor of outcome for right-sided but not for left-sided strokes. Age likely emerged as a predictor for the right hemisphere due to the absence of an NIHSS effect. As expected, reperfusion was an independent predictor of outcome for both hemispheres.

If verified in future studies, these findings have significant implications for treatment selection, acute prognostication, and clinical investigation of anterior circulation stroke patients undergoing IAT. Based on our results, using the NIHSS to stratify PAO patients by stroke severity should be done only for left-hemispheric strokes. Right-hemispheric strokes secondary to PAO should be stratified using a clinical scale tailored to the right hemisphere[4].

Neither NCCT nor CTA-SI ASPECTS predicted clinical outcome in the entire patient cohort. However, both ASPECTS scores appeared to predict outcome for left-sided strokes. Patients with left-sided strokes and good outcome had higher NCCT ASPECTS (8 vs. 7, p = 0.05) and CTA-SI ASPECTS (7 vs. 4.5, p = 0.06) scores on pre-treatment imaging than those with poor outcome (Figure 1). No such association between ASPECTS and outcome was seen in right-hemispheric strokes (Figure 2). The lack of a relationship between ischemic burden and outcome in the right hemisphere confirms the findings of a previous study which identified a subset of right-sided stroke patients who had good outcomes despite substantial lesion volumes[6]. No such group was found among the left-sided stroke patients in that study.

Published studies of IAT that have assessed predictors of outcome either have not mentioned the side of involvement[11, 14, 17‐19, 28], or have not analyzed left- and right-sided strokes separately [15, 16]. Only two studies present their outcomes by the involved hemisphere, although no analysis is performed comparing the two groups[2, 16]. This may explain why some studies that have assessed both pre-treatment imaging and outcome have found an association[15, 28] and some have not[2, 14, 16‐19]. The difference may depend on the proportion of left-versus right-sided strokes in a particular study.

One plausible explanation for the imaging findings is that language localizes to the left hemisphere in 88-96% of right-handed and 73-4% of left-handed people [29, 30]. Furthermore, many anatomic structures responsible for speech are localized to the peri-insular region, including Broca's area[31]. Because of the nature of the pial collateral circulation, the peri-insular region has the most severe reduction in blood flow and progresses to infarction most rapidly with ICA and proximal MCA occlusions. Therefore, in the usually dominant left hemisphere, even modest infarcts will likely have a negative impact on language and clinical outcome, which may explain why smaller infarct burden is associated with better outcome on the left.

While incompletely understood, poor outcomes in right-hemispheric strokes are thought to be related to cognitive deficits, most notably neglect[2, 32]. Previous studies of neglect have implicated the right superior temporal gyrus and inferior parietal lobule[33]. Because they lie distally in the MCA territory, an infarct has to be relatively large to involve these regions, and therefore, more infarct burden can be sustained before having a negative clinical effect.

These findings could have significant implications for imaging selection in acute stroke therapy. They highlight the weakness of volumetric-based analysis that does not account for eloquence[34].

Our patients with right-hemispheric stroke demonstrated significantly higher mean admission SBP than those with left-hemispheric stroke (168.4 mmHg vs 148.6 mmHg, p = 0.009). This finding was also demonstrated by Meyer et al, who showed that the increase in blood pressure for right-sided strokes was accompanied by significantly increased plasma catecholamine levels[35] and that these changes were mediated by infarction of the insular cortex, the area most at risk in our patients. The autonomic role of the right insular cortex is also supported by a study demonstrating that right insular infarction is associated with elevated serum cardiac troponin T levels[36], mediated by excessive sympathetic activation. It is important to note that there was no significant difference in the admission SBP between our patients with good versus poor outcome.

One limitation of this study is the small number of subjects, particularly those with good outcome. This reflects the highly selected nature of our study population, as well as the generally poor natural history of patients with proximal intracranial artery occlusion. This aspect of the study may have limited our ability to distinguish differences in the rates of good outcome between the two hemispheres. However, despite the small study population, we were able to demonstrate significant differences in predictors of outcome, particularly with respect to the NIHSS. Clearly, our findings need to be verified in larger studies. Additionally, the retrospective nature of this analysis introduces potential biases that should be addressed in prospective studies. While the left- and right-sided strokes in this study had similar stroke severities as reflected by the level of arterial occlusion and the extent of pre-treatment ischemic change on imaging, future studies should incorporate quantitative or semiquantitative assessments of the pial collateral circulation in their analyses.