Improving the Predictive Accuracy of Recanalization on Stroke Outcome in Patients Treated With Tissue Plasminogen Activator
Background and Purpose— Although early recanalization is a powerful predictor of stroke outcome after thrombolysis, some stroke patients remain disabled despite tissue plasminogen activator (tPA)–induced recanalization. Therefore, we sought to investigate whether the predictive accuracy of early recanalization on stroke outcome is improved when combined with clinical and radiological information.
Methods— We evaluated 177 patients with nonlacunar strokes in the middle cerebral artery (MCA) treated with intravenous tPA who were followed up during 3 months. Transcranial Doppler monitoring of recanalization was conducted during the first hours after tPA administration. The relative contribution of clinical, transcranial Doppler, and radiological information on stroke outcome was evaluated. We used logistic regression to derive a predictive model for good outcome (modified Rankin Scale score ≤2) after thrombolysis.
Results— Median National Institutes of Health Stroke Scale (NIHSS) score before tPA was 16. At 3 months, 87 patients (49.2%) became functionally independent (modified Rankin Scale score ≤2). In a logistic regression model, degree of recanalization within 300 minutes (P<0.001), proximal MCA occlusion (P<0.001), baseline NIHSS score (P=0.0013), systolic blood pressure (P=0.0116), and early ischemic changes on CT (P=0.0253) independently predicted outcome at 3 months. A 5-item score was developed on the basis of the factors significantly associated with stroke outcome in the logistic regression (total score range, 0 to 7). The likelihood of good outcome at 3 months was 0.82 (95% CI, 0.72 to 0.92) in patients who scored 0 to 2, 0.51 (95% CI, 0.36 to 0.66) in those who scored 3 to 4, and 0.15 (95% CI, 0.05 to 0.25) in those who scored 5 to 7 points.
Conclusions— The combination of clinical, radiological, and hemodynamic information predicts with a high accuracy long-term stroke outcome during or shortly after intravenous tPA administration.
The rationale for thrombolysis for acute ischemic stroke is recanalization of occluded arteries to reestablish brain function by saving tissue at risk. The speed of intracranial clot lysis has been shown to be strongly associated with early neurological improvement, reduced infarct size, and favorable prognosis.1–4 Although early recanalization has consistently been demonstrated as a powerful predictor of good long-term outcome after thrombolysis,1–6 some stroke patients experience little or no improvement and remain disabled despite tissue plasminogen activator (tPA)–induced recanalization. Several factors—including stroke severity, older age, systolic hypertension, extent of hypodensity or brain swelling on pretreatment CT, and admission hyperglycemia—have been shown to be predictors of poor outcome in stroke patients treated with tPA.7–11 The beneficial effect of early restoration of cerebral blood flow on stroke outcome may be hampered in part by such factors as extent of irreversible brain injury before recanalization, excessive glucose burden at the time of reperfusion, and blood pressure changes during thrombolysis. Therefore, a comprehensive multimodal approach is needed to better refine the predictive accuracy of tPA-induced recanalization on stroke prognosis.
See Editorial Comment, page 156
A tool that would allow accurate prediction of individual long-term outcome could be very useful in clinical research. A statistical model including all previously suggested independent outcome predictors could improve the design and analysis of clinical trials. However, a comprehensive prognostic algorithm for stroke recovery to be applied during or shortly after tPA administration has not previously been developed. Therefore, we sought to investigate whether the predictive accuracy of early recanalization on stroke outcome is improved when combined with clinical and radiological information.
Subjects and Methods
Our target group consisted of consecutive patients with acute ischemic stroke evaluated at 4 academic stroke centers who were monitored with transcranial Doppler (TCD) during standard treatment with systemic tPA. Data were prospectively collected and entered into a multicenter database. A total of 206 tPA-treated patients were identified for analysis. Patients with posterior circulation stroke (n=14) and lacunar strokes (n=15) were excluded. Final analysis included 177 patients with nonlacunar strokes involving the vascular territory of the middle cerebral artery (MCA) who were followed up during 3 months.
On arrival in the emergency room, patients underwent standard neurological examination, ECG, blood chemistry, and noncontrast CT before tPA administration. On the basis of standard definitions, the presence of vascular risk factors, including age, sex, history of hypertension, diabetes mellitus, and atrial fibrillation, was recorded. Time to treatment, pretreatment systolic and diastolic blood pressures, and total tPA dose were also recorded. The institutional ethics committee for each center approved the study protocol.
A standard TCD examination was performed in the emergency room on admission before tPA administration with 1-channel, 2-MHz equipment (TCD 100M, Spencer Technologies, DWL Multidop ×4, Multigon 500M, Neuroscan). A standard set of diagnostic criteria was applied to diagnose arterial occlusion. These criteria have been prospectively validated against angiography in patients with cerebral ischemia, and sensitivity for MCA occlusion was 93% with a specificity of >96%.12
Proximal MCA occlusion was defined as the absence of flow or the presence of minimal flow signal throughout the MCA at an insonation depth between 45 and 65 mm accompanied by flow diversion in the ipsilateral anterior and posterior cerebral arteries. Distal MCA occlusion was defined as the presence of abnormal flow signals at depths of <45 mm or a diffuse dampening of the mean flow velocity in the affected MCA of >30% compared with the unaffected MCA with signs of flow diversion. After the site of MCA occlusion was identified, continuous monitoring of the residual flow signals was performed with a Marc 500 head frame (Spencer Technologies) or DWL metal head frame to maintain tight transducer fixation and a constant angle of insonation. TCD monitoring of recanalization was conducted during the first hours after tPA administration. Recanalization on TCD was diagnosed as partial when blunted or dampened signals appeared in a previously demonstrated absent or minimal flow. Complete recanalization on TCD was diagnosed if the end-diastolic flow velocity improved to normal or elevated values (normal or stenotic signals).13 No change in the abnormal waveforms indicated that no recanalization had occurred. Changes on TCD were determined by the investigators using direct visual control of monitoring display. All investigators have taken a standard tutorial and multiple choice examination in grading flow signals using the Thrombolysis in Brain Ischemia (TIBI) flow grading system.14 Experienced TCD users at all participating sites showed >90% rate of correct answers during computerized examination.
The timing of arterial recanalization on TCD after symptom onset was determined as the time of earliest arrival of a normal or stenotic (low resistance) signals (complete recanalization) or a blunted or dampened signal (partial recanalization).
On admission, all patients underwent a CT scan before tPA administration. The extent of hypodensity or swelling resulting from acute ischemic edema on baseline CT was quantified with the methods described in the Alberta Stroke Program Early CT Score Study (ASPECTS).11 In this method, the MCA territory was divided into 10 standardized regions, and 1 point is subtracted for an area of early ischemic change such as focal swelling or parenchymal hypoattenuation for each of the defined regions. A normal CT scan has an ASPECTS value of 10 points. A score of 0 indicates diffuse ischemia throughout the MCA territory. CT scans were reviewed in each center by a local neurologist or neuroradiologist with extensive experience in acute stroke who was blinded to clinical, TCD, and outcome data.
Urgent CT scan was repeated in all patients who experienced clinical worsening (increase ≥4 points on the National Institutes of Health Stroke Scale [NIHSS] score). Symptomatic intracranial hemorrhage (SICH) was defined as a parenchymal hemorrhage on CT in relation to neurological worsening.
Clinical status at baseline and 24 hours after symptom onset was assessed with the NIHSS by a neurologist or a senior neurology resident who was video-trained and certified for application of the NIHSS.15 Early neurological deterioration or improvement was defined as an increase or decrease of ≥4 points in NIHSS score after 24 hours from baseline assessment.7 At the 3-month follow-up, patients were interviewed in the outpatient clinic (90%) or by telephone (10%) to evaluate clinical outcome in terms of level of independence in activities of daily living. If the patients were unable to answer the questions, information was obtained from a relative or caregiver. Modified Rankin Scale16 (mRS) was used to assess clinical outcome at 90 days. We defined good outcome as an mRS score ≤2.
The analysis was performed with SPSS 9.0 software (SPSS Inc). Statistical significance for differences between patients with good (mRS score ≤2) and poor (mRS >2) outcome were assessed by Pearson’s χ2 test for categorical variables and the Mann-Whitney U or Student’s t test for continuous variables. Continuous variables significantly associated with outcome were categorized to obtain the best discrimination between good and poor outcome. For this purpose, a receiver-operating characteristic (ROC) curve was applied for baseline glucose, time to recanalization, systolic blood pressure (SBP), and age. The probability of good outcome and independence at 3 months was assessed by forward stepwise logistic regression analysis. Variables with a value of P≤0.1 on univariate testing were included. The Hosmer and Lemeshow statistic was used to determine the sensitivity and specificity of the model and to assess goodness of fit. Model calibration was assessed with a calibration curve. A grading score was based on the factors significantly associated with stroke outcome in the logistic regression model. Categories for factors identified were scored according to their coefficient values (β coefficient) in the logistic regression. Sensitivity, specificity, and the positive and negative predictive values of all identified factors with respect to good outcome were calculated by use of ROC curves. Differences in the positive and negative predictive values were analyzed by comparing the area under the ROC curves derived from the grading score developed and each factor identified in the logistic regression. A level of P<0.05 was accepted as statistically significant.
We evaluated 177 patients (84 men, 93 women) with acute stroke resulting from MCA occlusion treated with intravenous tPA <3 hours after stroke onset. Demographic, risk factor profile, and baseline clinical findings are shown in Table 1. Mean age was 67.9±13.6 years (range, 31 to 93 years). The time elapsed between symptom onset to tPA bolus was 134.1±46.7 minutes.
Median baseline NIHSS score was 16 points (interquartile range, 13 to 20 points). Twenty-five patients (14.2%) had scores <10, 109 (61.9%) had scores of 11 to 20, and 43 (23.9%) had scores >20. No differences (P=0.437) were found among centers with regard to baseline NIHSS score.
The median ASPECTS value on baseline CT scan was 9 points (interquartile range, 7 to 10 points). Fifty patients (29%) had ASPECTS scores ≤7; 127 (71%) had scores >7 points.
On baseline TCD assessment, proximal MCA occlusion was detected in 100 patients (56.8%), and distal occlusion was seen in 77 (43.2%). TCD monitoring was started 131.2±65 minutes after stroke onset. Mean duration of TCD monitoring was 150.8±57 minutes. Recanalization during monitoring with TCD was found at a mean time of 164.45±49 minutes after stroke onset and at 37.3±35 minutes after tPA bolus. Complete recanalization was seen in 64 patients (36%), and partial recanalization was seen in 42 (24%). In 71 patients (40%), the MCA remained occluded at the end of the observation time. On TCD, 4 patients (2.8%) recanalized in <90 minutes, 46 (25.2%) between 90 to 180 minutes, 52 (28.7%) between 180 to 270 minutes, and 7 (4.7%) between 270 and 360 minutes after stroke onset.
Among the 168 patients with an NIHSS score at 24 hours, 98 (58.3%) improved, 17 (10.1%) worsened, and 53 (31.5%) remained stable during the first 24 hours. SICH occurred in 17 patients (9.6%).
At 3 months, the median mRS score was 3 (interquartile range, 1 to 5). Eighty-seven patients (49.2%) became functionally independent (mRS score ≤2). No differences (P=0.1) were found among centers regarding the proportion of patients who became independent at 3 months. The relative contribution of different variables to outcome and independence at 3 months is shown in Table 1. History of diabetes mellitus (P=0.002), time to treatment (P=0.045), time to recanalization (P=<0.001), baseline SBP (P=0.005), age (P=0.008), serum glucose (P=0.004), baseline NIHSS score (P<0.001), ASPECTS score (P=0.001), and proximal MCA occlusion (P=0.001) were factors significantly associated with long-term outcome. We categorized continuous variables that appeared significant on univariate analysis. An ROC curve provided cutoff points that better distinguished between patients with good and poor outcome: time to treatment, 137 minutes; serum glucose level, 138 mg/dL; age, 70 years; SBP, 150 mm Hg; and time to recanalization, 300 minutes. NIHSS score was categorized as <10, between 10 and 20, and >20 points. Categories for ASPECTS score were ≤7 and >7.
The derived logistic regression model revealed that recanalization <300 minutes (OR, 4.11; 95% CI, 2.42 to 6.95; P<0.001), proximal MCA occlusion (OR, 0.25; 95% CI, 0.10 to 0.61; P<0.001), baseline NIHSS score (OR, 0.35; 95% CI, 0.16 to 0.78; P=0.0013), SBP (OR, 0.32; 95% CI, 0.13 to 0.76; P=0.0116), and ASPECTS score (OR, 2.98; 95% CI, 1.13 to 7.85; P=0.0253) emerged as independent predictors of outcome at 3 months. Age, sex, history of hypertension, diabetes mellitus, time to treatment, and serum glucose were eliminated by forward stepwise variable selection.
A total of 72 descriptive categories of patients were obtained on the basis of the combination of the 5 significant variables from the logistic regression analysis. The sensitivity and specificity of the model for the prediction of good outcome were 0.82 and 0.71, respectively, with an overall accuracy of 0.76. The fit of the model to the data was good (P=0.56, Hosmer and Lemeshow statistic). A calibration curve revealed that the model was well calibrated.
Each of the 5 factors identified were categorized. Categories and point values of each category were assigned on the basis of their coefficient values in the logistic regression model (Table 2). Recanalization was scored as 0 (complete), 1 (partial), or 2 (no) points; NIHSS as 0 (≤10), 1 (between 10 and 20), or 2 (>20) points; SBP as 0 (≤150 mm Hg) or 1 (>150 mm Hg) point, and ASPECTS value as 0 (>7) or 1 (≤7) point. Total score of the resulting Multimodal Outcome Score for Stroke Thrombolysis (MOST) ranged from 0 to 7 (Table 3). An ROC curve provided a cut point of total MOST of 4 points (sensitivity, 77%; specificity, 82.1%) that better distinguishes between favorable and unfavorable outcome. As shown in Table 4, MOST showed a better efficiency in predicting good outcome at 3 months than the NIHSS score, ASPECTS values, occlusion location, SBP, and degree of early recanalization taken alone. Moreover, a total MOST >4 points predicted SICH after thrombolysis (sensitivity, 65%; specificity, 75.3%; P=0.032).
Table 5 shows the likelihood of good outcome and independence at 3 months according to total MOST. Fifty-seven patients (36.5%) were assigned to the high probability level of recovery (MOST, 0 to 2), 45 (28.8%) to the medium level (MOST, 3 to 4), and 54 (34.6%) to the low probability level of recovery (MOST, 5 to 7). The likelihood of good outcome at 3 months was 0.82 (95% CI, 0.72 to 0.92) in patients who scored 0 to 2, 0.51 (95% CI, 0.36 to 0.66) in those who scored 3 to 4, and 0.15 (95% CI, 0.05 to 0.25) in those who scored 5 to 7.
Our study demonstrates that the combination of 4 pretreatment items (NIHSS score, occlusion location, ASPECTS value, and SBP) and 1 posttreatment item (degree of recanalization) provides a more accurate stratification of the probability of recovery in stroke patients treated with intravenous tPA than any factor taken alone. The likelihood of good outcome and independence at 3 months was 82% in patients with a total MOST of 0 to 2 and only 15% in those who scored 5 to 7 points.
In our model, degree of early recanalization was a powerful predictor of good outcome after thrombolysis. General agreement exists that clinical benefit of tPA in ischemic stroke is linked to accelerated clot lysis and early recanalization.3 This beneficial effect of recanalization has been shown to be sustained at 3 months.2,5 However, despite the central role of early recanalization on neurological recovery, previous predictive models did not evaluate its contribution to the prediction of long-term outcome. Baird et al17 demonstrated that the combination of clinical and diffusion-weighted MR variables provides a better prediction of stroke recovery than any factor alone. However, predictive algorithms of stroke recovery in the setting of thrombolysis have not been previously developed. We have shown that the predictive accuracy of recanalization on stroke outcome is improved when combined with clinical and radiological information.
Our study confirms the independent contribution of various previously suggested predictors of stroke outcome. Stroke severity and high blood pressure have been shown to be powerful clinical predictors of poor outcome in both unselected stroke patients and those treated with tPA.7,8 Similarly, in patients with anterior circulation stroke, a global ASPECTS value on baseline CT of <7 points has been demonstrated to independently predict poor outcome and high risk of SICH after thrombolysis.11 Angiographic studies have also elucidated the relationship between location of arterial occlusion and outcome after intravenous thrombolysis.18
Increased blood pressure during the first few hours of acute ischemic stroke has been suggested to be reactive and indicative of a large infarction.19 In our model, baseline SBP of >150 mm Hg was independently associated with poor outcome. However, an aggressive blood pressure lowering below this value may decrease cerebral perfusion pressure, leading to neurological worsening. Therefore, further studies are required to elucidate the impact of blood pressure reduction on stroke outcome after thrombolysis. Although early recanalization has consistently demonstrated to be a powerful predictor of early neurological recovery and good outcome after thrombolysis,1–5 some stroke patients (20% in our series) experience little or no improvement and remain disabled despite tPA-induced recanalization. This is a limitation of recanalization as a predictor of stroke outcome that can be compensated by other variables of MOST. On the other hand, in our series, up to 25% of patients who did not recanalize at 300 minutes became functionally independent at 3 months. This may indicate that alternative mechanisms, including collateralization of flow, delayed but still nutritious reperfusion, and neuronal reorganization, may be responsible for this long-term benefit in the absence of early recanalization.20,21
MOST was constructed on the basis of the relative weight of independent predictors of stroke outcome in the logistic regression modeling. The model was well calibrated and internally valid. Although predictive models could potentially be used to aid management of individual stroke patients, they are more often used to predict outcome in groups of patients.22 MOST may also be useful in stratifying patients in randomized controlled trials of neuroprotective drugs to increase the likelihood of balance between the different treatment groups. For instance, patients who are likely to recover could be stratified or even excluded from some drug trials.
The present study has certain limitations. Dichotomization of ordinal scales decreases the power of these predictors, ie, NIHSS scale, in a combined model and places limitations on the tools of statistical analysis. Moreover, continuous TCD monitoring may have enhanced thrombolysis and led to a different outcome than in standard clinical practice.23 Furthermore, although our model was well calibrated and internally valid, an external validation of MOST needs to be done prospectively in a different patient group before it may be recommended for general use. Currently, this score should not dissuade physicians from treating patients according to accepted criteria.
In conclusion, the present study demonstrates that the predictive accuracy of early recanalization on stroke outcome is markedly improved when combined with clinical and radiological information. MOST is a 5-item score that predicts with a high accuracy long-term stroke outcome during or shortly after intravenous tPA administration.
Stroke fellows are supported by NIH Fellowship Training grant 1-T32-NS07412-O1A1 for the Stroke Program, University of Texas, Houston Medical School. Dr Alexandrov is supported by the NIH 1 K23 NS02229-01 Career Development Award. CLOTBUST Investigators: University of Houston, Houston, Tex: Chin-I Chen, MD; Oleg Chernyshev; John Y. Choi, MD; Sheila Ford, RN; Zsolt Garami, MD; James C. Grotta, MD; Marc Malkoff, MD; Sandi Shaw, RN. Hospital Vall d”Hebron, Barcelona, Spain: Joan Montaner, MD; Juan F. Arenillas, MD; Rafael Huertas, MD; Manuel Quintana; Marc Ribo, MD; Esteban Santamarina, MD; Gloria Dalmases, NR. University of Alberta, Edmonton, Alberta, Canada: Ashfaq Shuaib, MD.
- Received June 17, 2003.
- Revision received September 11, 2003.
- Accepted October 29, 2003.
Christou I, Alexandrov AV, Burgin WS, Wojner AW, Felberg RA, Malkoff M, Grotta JC. Timing of recanalization after tissue plasminogen activator therapy determined by transcranial Doppler correlates with clinical recovery from ischemic stroke. Stroke. 2000; 31: 1812–1816.
Molina CA, Montaner J, Abilleira S, Arenillas JF, Ribo M, Huertas R, Romero F, Alvarez-Sabin J. Time course of tissue plasminogen activator-induced recanalization in acute cardioembolic stroke: a case-control study. Stroke. 2001; 32: 2821–2827.
Alexandrov AV, Burgin SW, Demchuk AM, El-Mitwalli A, Grotta JC. Speed of intracranial clot lysis with intravenous tissue plasminogen activator therapy: sonographic classification and short-term improvement. Circulation. 2001; 103: 2897–2902.
Minematsu K, Yamaguchi T, Omae T. Spectacular shrinking deficit: rapid recovery from a major hemispheric syndrome by migration of an embolus. Neurology. 1992; 42: 157–162.
Labiche LA, Al-Senani F, Wojner AW, Grotta JC, Malkoff M, Alexandrov AV. Is the benefit of early recanalization sustained at 3 months? A prospective cohort study. Stroke. 2003; 34: 695–698.
Felberg RA, Okon NJ, El-Mitwalli A, Burgin WS, Grotta JC, Alexandrov AV. Early dramatic recovery during intravenous tissue plasminogen activator infusion: clinical pattern and outcome in acute middle cerebral artery stroke. Stroke. 2002; 33: 1301–1307.
Demchuk AM, Tanne D, Hill MD, Kasner SE, Hanson S, Grond M, Levine SR, for the Multicentre tPA Stroke Survey Group. Predictors of good outcome after intravenous tPA for acute ischemic stroke. Neurology. 2001; 57: 474–480.
Dávalos A, Toni D, Iweins F, Lesaffre F, Bastianello S, Castillo J. Neurological deterioration in acute ischemic stroke: potential predictors and associated factors in the European Cooperative Acute Stroke Study (ECASS) I. Stroke. 1999; 30: 2631–2636.
Burgin WS, Malkoff M, Demchuk AM, Felberg RA, Christou I, Grota JC, Alexandrov AV. Transcranial Doppler ultrasound criteria for recanalization after thrombolysis for middle cerebral artery stroke. Stroke. 2000; 31: 1128–1132.
Burgin WS, Wojner AW, Demchuk AM, Felberg RA, Barber PA, Hill MD, Malkoff M, El-Mitwalli A, Piriyawat P, Newcommon N, Button A. Validity and reliability of the Thrombolysis in Brain Infarction (TIBI) flow grades. Stroke. 2001; 32: 324. Abstract.
Lyden P, Brott T, Tilley B, Welch KMA, Mascha EJ, Levine S, Haley EC, Grotta J, Marler J, for the NINDS t-PA Stroke Study Group. Improved reliability of the NIH stroke scale using video training. Stroke. 1994; 25: 2220–2226.
Van Swieten JC, Koudstaal PJ, Visser MC, Schouten HJA, van Gijn J. Interobserver agreement for the assessment of handicap in stroke patients. Stroke. 1988; 19: 604–607.
Baird AE, Dambrosia J, Janket SJ, Eichbaum Q, Chaves C, Silver B, Barber PA, Parsons M, Darby D, Davis S, et al. A three-item scale for the early prediction of stroke recovery. Lancet. 2000; 357: 2095–2099.
Brott T, Lu M, Fagan SC, Frankel M, Grotta JC, Broderick J, Kwiatkowski T, Lewandowski C, Haley EC, Marler JR, Tilley BC. Hypertension and its treatment in the NINDS rt-PA Stroke Trial. Stroke. 1998; 29: 1504–1509.
Berrouschot J, Barthel H, Hesse S, Knapp WH, Schneider D, von Kummer R. Reperfusion and metabolic recovery of brain tissue and clinical outcome after ischemic stroke and thrombolytic therapy. Stroke. 2000; 31: 1545–1551.
FOOD Trial Collaboration. Performance of a statistical model to predict stroke outcome in the context of a large, simple, randomized, controlled trial of feeding. Stroke. 2003; 34: 127–133.
Alexandrov AV, Demchuk AM, Felberg RA, Christou I, Barber PA, Burgin WS, Malkoff M, Wojner AW, Grotta JC. High rate of complete recanalization and dramatic clinical recovery during tPA infusion when continuously monitored with 2-MHz transcranial Doppler monitoring. Stroke. 200; 31: 610–614.