Risk score for intracranial hemorrhage in patients with acute ischemic stroke treated with intravenous tissue-type plasminogen activator.
ABSTRACT There are few validated models for prediction of risk of symptomatic intracranial hemorrhage (sICH) after intravenous tissue-type plasminogen activator treatment for ischemic stroke. We used data from Get With The Guidelines-Stroke (GWTG-Stroke) to derive and validate a prediction tool for determining sICH risk.
The population consisted of 10 242 patients from 988 hospitals who received intravenous tissue-type plasminogen activator within 3 hours of symptom onset from January 2009 to June 2010. This sample was randomly divided into derivation (70%) and validation (30%) cohorts. Multivariable logistic regression identified predictors of intravenous tissue-type plasminogen activator-related sICH in the derivation sample; model β coefficients were used to assign point scores for prediction.
sICH within 36 hours was noted in 496 patients (4.8%). Multivariable adjusted independent predictors of sICH were increasing age (17 points), higher baseline National Institutes of Health Stroke Scale (42 points), higher systolic blood pressure (21 points), higher blood glucose (8 points), Asian race (9 points), and male sex (4 points). The C-statistic was 0.71 in the derivation sample and 0.70 in the independent internal validation sample. Plots of observed versus predicted sICH showed good model calibration in the derivation and validation cohorts. The model was externally validated in National Institute of Neurological Disorders and Stroke trial patients with a C-statistic of 0.68.
The GWTG-Stroke sICH risk "GRASPS" score provides clinicians with a validated method to determine the risk of sICH in patients treated with intravenous tissue-type plasminogen activator within 3 hours of stroke symptom onset.
- SourceAvailable from: Sheng-Feng Sung[Show abstract] [Hide abstract]
ABSTRACT: The Oxfordshire Community Stroke Project (OCSP) classification is a simple stroke classification system with value in predicting clinical outcomes. We investigated whether and how the addition of OCSP classification to the Safe Implementation of Thrombolysis in Stroke (SITS) symptomatic intracerebral hemorrhage (SICH) risk score improved the predictive performance. We constructed an extended risk score by adding an OCSP component, which assigns 3 points for total anterior circulation infarcts, 0 point for partial anterior circulation infarcts or lacunar infarcts. Patients with posterior circulation infarcts were assigned an extended risk score of zero. We analyzed prospectively collected data from 4 hospitals to compare the predictive performance between the original and the extended scores, using area under the receiver operating characteristic curve (AUC) and net reclassification improvement (NRI). In a total of 548 patients, the rates of SICH were 7.3% per the National Institute of Neurological Diseases and Stroke (NINDS) definition, 5.3% per the European-Australasian Cooperative Acute Stroke Study (ECASS) II, and 3.5% per the SITS-Monitoring Study (SITS-MOST). Both scores effectively predicted SICH across all three definitions. The extended score had a higher AUC for SICH per NINDS (0.704 versus 0.624, P = 0.015) and per ECASS II (0.703 versus 0.612, P = 0.016) compared with the SITS SICH risk score. NRI for the extended risk score was 22.3% (P = 0.011) for SICH per NINDS, 21.2% (P = 0.018) per ECASS II, and 24.5% (P = 0.024) per SITS-MOST. Incorporation of the OCSP classification into the SITS SICH risk score improves risk prediction for post-thrombolysis SICH.BMC Neurology 03/2014; 14(1):39. · 2.56 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Predicting functional outcome and mortality after stroke, with or without thrombolysis, is a critical role of neurologists. This article reviews the predictors of outcome after ischemic stroke. Several scores were recently designed to predict (1) mortality and poor functional outcome after ischemic stroke, (2) the functional outcome and risk of symptomatic intracranial hemorrhage (sICH) after thrombolysis, and (3) the risk of stroke following TIA. Validation of these prediction instruments is ongoing, and studies will be critical to determine the general applicability of these scores. Although several scores were developed to predict mortality and outcome after stroke, it may be premature to employ these prediction scores to determine individual patient outcome. Similarly, prediction scores should not be used to deny patients tissue plasminogen activator (tPA), even if the scores predict that the patient has a high likelihood of sICH or poor outcome after thrombolysis.Continuum (Minneapolis, Minn.). 04/2014; 20(2 Cerebrovascular Disease):412-28.
- [Show abstract] [Hide abstract]
ABSTRACT: Several prognostic scores have been developed to predict the risk of symptomatic intracranial hemorrhage (sICH) after ischemic stroke thrombolysis. We compared the performance of these scores in a multicenter cohort. We merged prospectively collected data of patients with consecutive ischemic stroke who received intravenous thrombolysis in 7 stroke centers. We identified and evaluated 6 scores that can provide an estimate of the risk of sICH in hyperacute settings: MSS (Multicenter Stroke Survey); HAT (Hemorrhage After Thrombolysis); SEDAN (blood sugar, early infarct signs, [hyper]dense cerebral artery sign, age, NIH Stroke Scale); GRASPS (glucose at presentation, race [Asian], age, sex [male], systolic blood pressure at presentation, and severity of stroke at presentation [NIH Stroke Scale]); SITS (Safe Implementation of Thrombolysis in Stroke); and SPAN (stroke prognostication using age and NIH Stroke Scale)-100 positive index. We included only patients with available variables for all scores. We calculated the area under the receiver operating characteristic curve (AUC-ROC) and also performed logistic regression and the Hosmer-Lemeshow test. The final cohort comprised 3012 eligible patients, of whom 221 (7.3%) had sICH per National Institute of Neurological Disorders and Stroke, 141 (4.7%) per European Cooperative Acute Stroke Study II, and 86 (2.9%) per Safe Implementation of Thrombolysis in Stroke criteria. The performance of the scores assessed with AUC-ROC for predicting European Cooperative Acute Stroke Study II sICH was: MSS, 0.63 (95% confidence interval, 0.58-0.68); HAT, 0.65 (0.60-0.70); SEDAN, 0.70 (0.66-0.73); GRASPS, 0.67 (0.62-0.72); SITS, 0.64 (0.59-0.69); and SPAN-100 positive index, 0.56 (0.50-0.61). SEDAN had significantly higher AUC-ROC values compared with all other scores, except for GRASPS where the difference was nonsignificant. SPAN-100 performed significantly worse compared with other scores. The discriminative ranking of the scores was the same for the National Institute of Neurological Disorders and Stroke, and Safe Implementation of Thrombolysis in Stroke definitions, with SEDAN performing best, GRASPS second, and SPAN-100 worst. SPAN-100 had the worst predictive power, and SEDAN constantly the highest predictive power. However, none of the scores had better than moderate performance.Stroke 01/2014; · 6.16 Impact Factor
Olson, Eric D. Peterson, Adrian F. Hernandez, Gregg C. Fonarow, Lee H. Schwamm and Eric
Bijoy K. Menon, Jeffrey L. Saver, Shyam Prabhakaran, Mathew Reeves, Li Liang, DaiWai M.
With Intravenous Tissue-Type Plasminogen Activator
Risk Score for Intracranial Hemorrhage in Patients With Acute Ischemic Stroke Treated
Print ISSN: 0039-2499. Online ISSN: 1524-4628
Copyright © 2012 American Heart Association, Inc. All rights reserved.
is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
2012;43:2293-2299; originally published online July 17, 2012;
World Wide Web at:
The online version of this article, along with updated information and services, is located on the
Data Supplement (unedited) at:
is online at:
Information about subscribing to
Information about reprints can be found online at:
document. Permissions and Rights Question and Answer process is available in the
Request Permissions in the middle column of the Web page under Services. Further information about this
Once the online version of the published article for which permission is being requested is located, click
can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial Office.
Requests for permissions to reproduce figures, tables, or portions of articles originally published
by guest on December 4, 2013 by guest on December 4, 2013 by guest on December 4, 2013 by guest on December 4, 2013http://stroke.ahajournals.org/ http://stroke.ahajournals.org/http://stroke.ahajournals.org/ http://stroke.ahajournals.org/Downloaded from Downloaded from Downloaded from Downloaded from
Risk Score for Intracranial Hemorrhage in Patients With
Acute Ischemic Stroke Treated With Intravenous
Tissue-Type Plasminogen Activator
Bijoy K. Menon, MD; Jeffrey L. Saver, MD; Shyam Prabhakaran, MD, MS; Mathew Reeves, PhD;
Li Liang, PhD; DaiWai M. Olson, RN, PhD; Eric D. Peterson, MD; Adrian F. Hernandez, MD;
Gregg C. Fonarow, MD; Lee H. Schwamm, MD; Eric E. Smith, MD, MPH
Background and Purpose—There are few validated models for prediction of risk of symptomatic intracranial hemorrhage
(sICH) after intravenous tissue-type plasminogen activator treatment for ischemic stroke. We used data from Get With
The Guidelines–Stroke (GWTG-Stroke) to derive and validate a prediction tool for determining sICH risk.
Methods—The population consisted of 10 242 patients from 988 hospitals who received intravenous tissue-type
plasminogen activator within 3 hours of symptom onset from January 2009 to June 2010. This sample was randomly
divided into derivation (70%) and validation (30%) cohorts. Multivariable logistic regression identified predictors of
intravenous tissue-type plasminogen activator-related sICH in the derivation sample; model ? coefficients were used to
assign point scores for prediction.
Results—sICH within 36 hours was noted in 496 patients (4.8%). Multivariable adjusted independent predictors of sICH
were increasing age (17 points), higher baseline National Institutes of Health Stroke Scale (42 points), higher systolic
blood pressure (21 points), higher blood glucose (8 points), Asian race (9 points), and male sex (4 points). The C-statistic
was 0.71 in the derivation sample and 0.70 in the independent internal validation sample. Plots of observed versus
predicted sICH showed good model calibration in the derivation and validation cohorts. The model was externally
validated in National Institute of Neurological Disorders and Stroke trial patients with a C-statistic of 0.68.
Conclusions—The GWTG-Stroke sICH risk “GRASPS” score provides clinicians with a validated method to
determine the risk of sICH in patients treated with intravenous tissue-type plasminogen activator within 3 hours
of stroke symptom onset. (Stroke. 2012;43:2293-2299.)
Key Words: acute ischemic stroke ? intravenous tPA ? risk score ? symptomatic intracranial hemorrhage
with an increased risk of symptomatic intracranial hemor-
rhage (sICH).1,2Validated prediction models could help a
clinician objectively identify patients at low, intermediate, or
high risk of sICH. Previous studies have identified risk
factors for sICH but have mostly been limited to a small
number of academic centers, were conducted before the
establishment of stroke systems of care including certification
programs for primary stroke centers, or were conducted
outside the US practice setting.3–6Prediction models for sICH
have been developed and have been validated in clinical trial
populations but not in more widespread academic and com-
ntravenous tissue-type plasminogen activator (IV tPA)
improves outcomes in ischemic stroke but is associated
munity practices.3,4To date there have been no large cohort
studies of sICH in contemporary US practice as have been
done in Canada or Europe.7–9
in US academic and nonacademic hospitals, we used data from
characteristics of patients with sICH and validate a clinical
prediction tool for determining risk of IV tPA-related sICH
among patients presenting to the emergency department and
treated with IV tPA within 3 hours of stroke symptom onset.
GWTG-Stroke is a voluntary quality improvement initiative
whereby participating hospitals use a web-based Patient Manage-
Received April 11, 2012; accepted May 7, 2012.
From the Calgary Stroke Program, Department of Clinical Neurosciences, Hotchkiss Brain Institute, University of Calgary, Calgary, Ontario, Canada
(B.K.M., E.E.S.); the Department of Neurology (J.L.S.) and Division of Cardiology (G.C.F.), University of California Los Angeles, Los Angeles, CA;
the Department of Neurology, Northwestern University-Feinberg School of Medicine, Chicago, IL (S.P.); the Department of Epidemiology, Michigan
State University, East Lansing, MI (M.R.); Duke Clinical Research Institute, Duke University, Durham, NC (L.L., D.M.O., E.D.P., A.F.H.); and the
Department of Neurology, Massachusetts General Hospital, Boston, MA (L.H.S.).
Joanna M. Wardlaw, MD, was the Guest Editor for this paper.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/
Correspondence to Bijoy K. Menon, MD, 1079 A, 29thStreet NW, Calgary, Alberta, Canada T3H4J2. E-mail Bijoy.Menon@Albertahealthservices.ca
© 2012 American Heart Association, Inc.
Stroke is available at http://stroke.ahajournals.orgDOI: 10.1161/STROKEAHA.112.660415
by guest on December 4, 2013 http://stroke.ahajournals.org/Downloaded from
ment Tool (Outcomes Sciences Inc, Cambridge, MA) to collect
clinical data, access decision support, and provide real-time online
reporting of performance on quality-of-care measures. Details of the
program have been described previously.10Information is abstracted
from the medical record and uploaded to the database with auto-
mated checks for data validity. Post-IV tPA sICH is defined as
neurological worsening within 36 hours of tPA administration that is
attributed to ICH verified by CT or MRI, as documented in the chart
by the treating physician. This definition is based on the criteria for
sICH in the 1995 National Institute of Neurological Disorders and
Stroke (NINDS) trial.1
Participating institutions are required to comply with local regu-
latory and privacy guidelines and to submit the GWTG protocol for
review and approval by their local Institutional Review Board.
Because data are used primarily at the local site for quality
improvement, sites are granted a waiver of informed consent under
the common rule. The Duke Clinical Research Institute served as the
data analysis center and had local Institutional Review Board
approval to conduct the study.
We included patient admissions from January 1, 2009. to June 30,
2010. During this time there were 13 481 patients with ischemic
stroke who presented to the emergency department within 3 hours
and were treated with IV tPA. By design, we did not include 3171
patients treated with IV tPA after 3 hours, 568 patients treated with
IV tPA for in-hospital acute ischemic stroke, and 243 patients treated
with tPA with unknown time from stroke symptom onset. Among the
13 481 patients presenting directly to the emergency department who
were treated within 3 hours, a total of 3239 (24%) were excluded for
the following reasons: (1) investigational or experimental protocol
for thrombolysis (n?117 [0.9%]); (2) transfer in from a different
hospital (n?732 [5.4%]); (3) transfer out to another hospital within
2 days thus precluding information on the presence or absence of
sICH within 36 hours (n?268 [2%]); (4) missing information on
discharge outcome (n?71 [0.5%]); (5) missing National Institutes of
Health Stroke Scale (NIHSS) score (a critical covariate, n?1342
[9.9%]); (6) international normalized ratio ?1.7 (above the cutoff
recommended by the US Food and Drug Administration drug
labeling, n?73 [0.5%]); and (7) treatment with intra-arterial thera-
pies (n?552 [4.1%]). Additionally, data from 13 hospitals providing
84 patients (0.6%) were excluded because ?75% of the data
submitted from these hospitals were complete, raising questions
regarding the accuracy of the data. The final study population of
10 242 patients was randomly divided into derivation (70%) and
validation (30%) cohorts (online-only Data Supplement Figure I).
The randomly selected derivation cohort and validation cohort were
well matched with respect to candidate predictor variables and
outcome (data not shown).
Candidate Predictor Variables
Potential predictor variables were determined based on prior knowl-
edge, causal relevance, and availability of data in the Patient
Management Tool. A complete list of candidate variables is available
as an online-only Data Supplement.
The rates of sICH were reported in each group for categorical
variables. The sICH rates were compared across the groups using
Pearson ?2test for categorical variables and Cochran-Mantel-
Haenszel test using row-mean score statistic for the continuous
variables. Multivariable logistic regression analysis was used to
identify independent predictors of sICH performed in the derivation
sample (70%). Within-hospital clustering was accounted for using a
generalized estimating equation approach with exchangeable work-
ing correlation matrix. Continuous covariates were evaluated for
appropriateness of the linearity assumption using plots displaying the
relationship of each variable with the log odds of sICH. When
appropriate, knots were determined to create splines (variables that
were modeled equivalent to piece-linear continuous variables). Age,
door-to-needle time, and international normalized ratio were in-
cluded as continuous variables because of a linear relationship with
log odds of sICH throughout their range. Baseline NIHSS, systolic
blood pressure, heart rate, and blood glucose showed ceiling effects
and thus were truncated. The remaining continuous variables were
modeled as splines with nonsignificant splints removed from the
reduced model. In addition, we tested for interactions between (1)
warfarin use on admission and international normalized ratio; and (2)
past medical history of diabetes mellitus and prior stroke/transient
The reduced model included only important predictors chosen based
on statistical significance in the multivariable model and clinical
relevance. The discriminatory ability of the model was assessed using
the C-statistic, which represents the probability that the model would
correctly predict that the risk was higher for a patient who had sICH
versus a patient who did not. Calibration of the model was assessed by
Hosmer-Lemeshow statistic and by the plots comparing predicted
versus observed probability of sICH. The reduced model was then
validated by assessing model performance, that is, discrimination and
calibration in the validation cohort (30%). The selected reduced model
was refit using the whole sample to develop a risk score by assigning a
weighted integer to each independent predictor based on the predictor’s
? coefficient. For each patient, the weighted integers were summed to
obtain a total risk score with a range of 0 to 101 points. Continuous
variables included in the reduced final model were categorized into
derive the risk score. This risk score was then assessed for discrimina-
tion with C-statistic and calibration by comparing predicted versus
observed sICH rates. The model was also externally validated in IV
tPA-treated patients from the 1995 NINDS trial (n?309 after excluding
2 patients with missing glucose values). All probability values are
2-sided with P?0.05 considered statistically significant. Analyses were
performed using SAS software (Version 9.2; SAS Institute, Cary, NC).
The final study population consisted of 10 242 patients from
988 hospitals. Mean age was 69.9?14.7 years, median
baseline NIHSS was 11 (interquartile range, 7–18), and
median time from stroke symptom onset to IV tPA treatment
was 135 minutes (interquartile range, 110–160 minutes).
Patients were treated at hospitals with median bed size 375
(interquartile range, 270–558), most patients were treated in
teaching hospitals (55.5%), and 34.5% patients were treated
in hospitals administering ?6 IV tPA cases a year.
sICH within 36 hours was noted in 496 (4.8%) patients.
Patients with sICH, compared with patients without sICH,
were older (mean age, 75.3 versus 69.6 years; P?0.0001),
had higher initial stroke severity (median NIHSS, 17 versus
11; P?0.0001), higher systolic blood pressure (median,
159 mm Hg versus 154 mm Hg; P?0.0001), and higher
blood glucose values (median, 126 mg/dL versus 118 mg/dL;
P?0.0001). Risk of sICH according to sex, medical comor-
bidities, and medication use is shown in Table 1. The
“in-hospital” mortality rate was 175 of 487 (35.9%) in
patients with sICH and 629 of 9604 (6.5%) in those without
sICH after excluding 151 patients transferred to other acute
Multivariable modeling in the derivation sample identified
increasing age, higher baseline NIHSS, higher systolic blood
pressure, higher blood glucose, Asian race, and male sex as
independent predictors of IV tPA-related sICH. C-statistic in
the derivation sample was 0.71 (95% CI, 0.68–0.73) and the
Hosmer-Lemeshow test probability value was 0.78. In the
independent validation sample, the C-statistic was 0.70 (95%
by guest on December 4, 2013 http://stroke.ahajournals.org/Downloaded from
Characteristics (N?10 242)
Rate of IV tPA-Related Symptomatic Intracranial Hemorrhage According to Patient and Hospital
Percent ICH After IV tPA
Native Hawaiian or Pacific Islander
American Indian or Alaska Native
Black or African American
Systolic blood pressure, mm Hg
Blood glucose, mg/dL
Serum creatinine, mg/dl 0.694
Anticoagulant before admission0.04
Antiplatelet drug before admission
Heart failure 0.005
Hospital type 0.46
IV tPA indicates intravenous tissue-type plasminogen activator; ICH, intracranial hemorrhage; NIHSS, National Institutes of Health Stroke Scale; INR,
international normalized ratio; TIA, transient ischemic attack; CAD, coronary artery disease.
Menon et alRisk Score for sICH With IV tPA
by guest on December 4, 2013http://stroke.ahajournals.org/Downloaded from
CI, 0.67–0.74) and the Hosmer-Lemeshow test probability
value was 0.18. Because good validation was demonstrated,
the model was then refitted using the whole study sample; the
C-statistic was 0.71 and the Hosmer-Lemeshow test proba-
bility value was 0.48. Model results are presented in Table 2.
Secondary analyses show no interaction among warfarin
use, increased international normalized ratio, and the risk of
sICH (P?0.42). No statistical interaction was observed be-
tween history of diabetes mellitus and history of prior stroke
in predicting risk of sICH (P?0.49).
A risk score was created by assigning a weighted integer to
each of the 6 independent predictors (Table 2) based on the
predictor’s ? coefficient and summing them to obtain a total
risk score with a range of 45 to 101 points (Figure 1). The risk
score demonstrated good discrimination (C-statistic 0.70) and
good calibration from a plot of predicted versus observed
sICH (P?0.35 for Hosmer-Lemeshow test). There was good
correlation between predicted and observed risk of sICH
across a wide range of predicted risk in the derivation and
validation cohorts (Figure 2A). Likewise, a plot of observed
versus predicted sICH in the validation cohort showed strong
correlation (Figure 2B). The C-statistic for the model in the
NINDS trial patients was 0.68 (Figure 3).
In this study, we identified risk factors for sICH after IV tPA
and used this information to derive and validate a risk score
for predicting risk of symptomatic ICH. This “GRASPS”
clinical risk score has 6 predictor variables: Glucose at
presentation, Race (Asian), Age, Sex (male), systolic blood
Sample (n?10 242)
Refitted Multivariate Model Using the Entire Study
CI of OR
CI of OR
NIHSS per 1 unit up to
Age per 10 y
Systolic BP per 10 mm Hg
up to maximum 180
Blood glucose per 10 mg/dL
up to maximum 150
Female versus male
Asian versus non-Asian
1.06 1.02 1.11 0.003
C-statistic 0.71 and Hosmer-Lemeshow test P value 0.48.
NIHSS indicates National Institutes of Health Stroke Scale; BP, blood
Figure 1. Risk score for symptomatic intracranial hemorrhage after IV tPA. IV tPA indicates intravenous tissue-type plasminogen activator.
by guest on December 4, 2013http://stroke.ahajournals.org/Downloaded from
Pressure at presentation, and Severity of stroke at presen-
tation (NIHSS). This clinical risk prediction tool (www.
strokeassociation.org/GWTGsICHcalculator) uses clinical
variables readily available at the time of presentation with
acute ischemic stroke. The c-statistic of 0.71 indicates good
discriminatory ability in predicting sICH in patients who
present within 3 hours of stroke symptom onset and are
treated with IV tPA. The score also has good calibration
indicating how closely the predicted risk agrees with the
actual risk of sICH.
Other risk scores for post IV tPA sICH have been pub-
lished. The Hemorrhage After Thrombolysis (HAT) score
includes stroke severity at presentation (NIHSS), presence of
hyperglycemia or diabetes, and hypodensity on baseline CT.3
The c-statistic for the validation cohort was 0.74, similar to
ours. The HAT score has also been externally validated in
clinical trial populations but not always with the same
discriminatory ability: the C-statistic was 0.72 in a subse-
quent validation using data from the Combined Lysis of
Thrombus in Brain Ischemia Using Transcranial Ultrasound
and Systemic tPA (CLOTBUST) and Transcranial Ultra-
sound in Clinical Sonothrombolysis (TUCSON) studies but
only 0.61 in the Stroke-Acute Ischemic NXY Treatment
(SAINT)-I and SAINT-II trials.3,11,12The Multicenter Re-
combinant Tissue Plasminogen Activator Stroke Survey
Group developed a 4-point risk score that included age ?60
years, NIHSS ?10, glucose ?8.325 mmol/L, and platelet
count ?150 000/mm3.4This score had a C-statistic of 0.69 in
their data set and 0.61 in subsequent validation in the
SAINT-I and SAINT-II trials.4,11Both of these studies
derived the risk scores from relatively small cohorts and did
not report on calibration.3,4The C-statistic is dependent on
case mix, disease severity, and prevalence of risk factors in
the cohort. Model discrimination should therefore be better
when the cohort is large and representative rather than highly
selected, like in a clinical trial. Calibration should also be
reported as a measure of model performance.13Our model
calibration and discrimination were good and validated well
externally in data from the NINDS trial. Additional studies
will be needed to determine the external validity of our risk
score and to compare our risk score with previously published
Figure 2. A, Calibration of the prediction tool in
the derivation and validation samples according to
observed versus predicted rate of sICH post-IV
tPA in 6 prespecified risk categories. B, Plot of
observed versus predicted rate of sICH in the
independent validation sample (n?3071) according
to decile of predicted risk. sICH indicates symp-
tomatic intracranial hemorrhage; IV tPA, intrave-
nous tissue-type plasminogen activator.
Figure 3. Observed versus predicted rate of sICH according to
quartile of predicted risk in the NINDS IV tPA data set, n?309.
sICH indicates symptomatic intracranial hemorrhage; NINDS,
National Institutes of Neurological Disorders and Stroke; IV tPA,
intravenous tissue-type plasminogen activator.
Menon et al Risk Score for sICH With IV tPA
by guest on December 4, 2013 http://stroke.ahajournals.org/ Downloaded from
Our study confirms known associations with higher stroke
severity, higher blood glucose, and higher blood pres-
sure.3,9,14–16Our results, however, suggest that there is in-
creasing risk with increasing systolic blood pressure even
when the systolic blood pressure is ?185 mm Hg. A limita-
tion of our study is that only initial blood pressure and
glucose were recorded without longitudinal assessments.
Therefore, we are unable to determine the extent to which
acute treatment of elevated blood pressure or glucose might
have mitigated the risk of subsequent tPA-related sICH.
Ours is the first study to report on male sex and Asian race
as independent predictors of IV tPA-related sICH, although
they account for a relatively small amount of variation in
sICH risk. It is important to note that the increased risk of
ICH associated with male sex and Asian race is modest and
does not appear sufficient to justify withholding IV tPA from
an otherwise eligible patient.
We failed to find an association between IV tPA-related
sICH and elevated international normalized ratio or warfarin
use in contrast to 2 previous single-center studies17,18but
consistent with findings from the Registry of the Canadian
Stroke Network.19We also failed to find an association
between sICH and the combination of previous stroke and
diabetes, which was an exclusion criterion in the European
Cooperative Acute Stroke Study (ECASS) III study.2We
failed to find an association between prestroke antiplatelet
drug use and sICH in patients receiving IV tPA within 3
hours. In an analysis of data from the European Cooperative
Acute Stroke Study (SITS-MOST) registry,9prestroke aspirin
use was associated with an increased risk of sICH using the
SITS-MOST definition of ICH but not when using the
definition of ICH used in the NINDS trial.1However, we did
not have data on the number, type or dose of agent, and
cannot exclude an increased risk in patients on combined
aspirin and clopidogrel as has been suggested in some
Our study has some limitations. Patients and hospitals may
not be entirely representative because hospital participation in
GWTG-Stroke is voluntary. Nonetheless, the demographics
of the GWTG-Stroke patient population are quite similar to
the overall demographics of all US patients with stroke.10We
used a definition of sICH based on the 1995 NINDS trial
definition that attributes clinical worsening to ICH docu-
mented on CT or MRI based on the investigator’s judgment.
This definition is different from the ECASS II definition in
which clinical deterioration is defined as a 4-point increase in
NIHSS or the SITS-MOST definition in which, in addition,
only parenchymal hemorrhages Type 2 can be considered
symptomatic.8,21We are not able to test our prediction score
predicting sICH by these other definitions because we do not
have information on hemorrhage type or posthemorrhage
NIHSS. sICH events were determined by the treating physi-
cians and not centrally reviewed; however, the very large
difference in mortality between patients with and without
sICH suggests that the sICH events were clinically relevant.
Our risk score performed nearly as well in the NINDS trial in
which the sICH events were prospectively recorded and
centrally reviewed. Neuroimaging data were not available in
the database. We were, however, able to predict the risk of IV
tPA-related sICH with similar discrimination as risk scores
that rely on neuroimaging findings.3We chose to analyze data
from patients treated ?3 hours after symptom onset, consis-
tent with current US Food and Drug Administration labeling
for IV tPA for acute ischemic stroke and therefore cannot be
certain whether the same risk factors for sICH are relevant to
patients treated beyond 3 hours. We also cannot be certain
that the same risk factors predict in-hospital stroke, because
patients with in-hospital stroke were excluded. Finally, we
excluded 11.1% of patients because of missing information
that could have introduced some selection bias.
In summary, the “GRASPS score” is a well-validated,
evidence-based determination of the risk of sICH in patients
treated with IV tPA within 3 hours of stroke symptom onset
that provides clinicians, patients, and families an objective
understanding of the risks involved with this treatment. The
score should not be used to infer which patients would derive
the most or least benefit from IV tPA because neither the
“GRASPS score” nor its components have been shown to
modify the effect of IV tPA treatment. The score could be used
to facilitate quality improvement by allowing hospitals to deter-
mine whether their rate of sICH exceeds the expected rate
predicted by the score. Calculation of the score can be facilitated
by use of applications on handheld devices or by a score
calculator accessed on the Internet. An application for calculat-
ing the score for individuals or groups of patients is available on
the web-based patient management tool for GWTG-Stroke
hospitals to calculate predicted risk for groups of patients.
Sources of Funding
The American Heart Association and the American Stroke Associ-
ation fund Get With the Guidelines–Stroke. The program is also
supported in part by unrestricted educational grants to the American
Heart Association by Pfizer, Inc, New York, NY, and the Merck-
Schering Plough Partnership (North Wales, PA), who did not
participate in the design, analysis, article preparation, or approval.
Dr Schwamm is a member of the Steering Committee for the
Desmoteplase in Acute Ischemic Stroke Trial (DIAS)-3 trial funded
by Lundbeck and is employed by Massachusetts General Hospital,
which has received IV tPA from Genentech as part of a National
Institutes of Health-sponsored clinical trial.
1. Tissue plasminogen activator for acute ischemic stroke. The National
Institute of Neurological Disorders and Stroke rtPA Stroke Study Group.
N Engl J Med. 1995;333:1581–1587.
2. Hacke W, Kaste M, Bluhmki E, Brozman M, Davalos A, Guidetti D, et
al. Thrombolysis with alteplase 3 to 4.5 hours after acute ischemic stroke.
N Engl J Med. 2008;359:1317–1329.
3. Lou M, Safdar A, Mehdiratta M, Kumar S, Schlaug G, Caplan L, et al.
The HAT score: a simple grading scale for predicting hemorrhage after
thrombolysis. Neurology. 2008;71:1417–1423.
4. Cucchiara B, Tanne D, Levine SR, Demchuk AM, Kasner S. A risk score
to predict intracranial hemorrhage after recombinant tissue plasminogen
activator for acute ischemic stroke. J Stroke Cerebrovasc Dis. 2008;17:
5. Cucchiara B, Kasner SE, Tanne D, Levine SR, Demchuk A, Messe SR,
et al. Factors associated with intracerebral hemorrhage after thrombolytic
therapy for ischemic stroke: pooled analysis of placebo data from the
by guest on December 4, 2013 http://stroke.ahajournals.org/Downloaded from
Stroke-Acute Ischemic NXY Treatment (SAINT) I and SAINT II trials.
6. Demchuk AM, Morgenstern LB, Krieger DW, Linda Chi T, Hu W, Wein
TH, et al. Serum glucose level and diabetes predict tissue plasminogen
activator-related intracerebral hemorrhage in acute ischemic stroke.
7. Nadeau JO, Shi S, Fang J, Kapral MK, Richards JA, Silver FL, et al. tPA
use for stroke in the registry of the Canadian Stroke Network. Can
J Neurol Sci. 2005;32:433–439.
8. Wahlgren N, Ahmed N, Davalos A, Ford GA, Grond M, Hacke W, et al.
Thrombolysis with alteplase for acute ischaemic stroke in the Safe Imple-
mentation of Thrombolysis in Stroke-Monitoring Study (SITS-MOST):
an observational study. Lancet. 2007;369:275–282.
9. Wahlgren N, Ahmed N, Eriksson N, Aichner F, Bluhmki E, Davalos A,
et al. Multivariable analysis of outcome predictors and adjustment of
main outcome results to baseline data profile in randomized controlled
trials: Safe Implementation of Thrombolysis in Stroke-Monitoring Study
(SITS-MOST). Stroke. 2008;39:3316–3322.
10. Reeves MJ, Fonarow GC, Smith EE, Pan W, Olson D, Hernandez AF, et
al. Representativeness of the Get With The Guidelines–Stroke registry:
comparison of patient and hospital characteristics among Medicare ben-
eficiaries hospitalized with ischemic stroke. Stroke. 2012;43:44–49.
11. Cucchiara B, Kasner S, Tanne D, Levine S, Demchuk A, Messe S, et al.
Validation assessment of risk scores to predict postthrombolysis intrace-
rebral haemorrhage. Int J Stroke. 2011;6:109–111.
12. Tsivgoulis G, Saqqur M, Barreto A, Demchuk AM, Ribo M, Rubiera M, et
al. Validity of hat score for predicting symptomatic intracranial hemorrhage
of sonothrombolysis. Cerebrovasc Dis. 2011;31:471–476.
13. Cook NR. Use and misuse of the receiver operating characteristic curve
in risk prediction. Circulation. 2007;115:928–935.
14. Sylaja PN, Cote R, Buchan AM, Hill MD. Thrombolysis in patients
older than 80 years with acute ischaemic stroke: Canadian alteplase
for stroke effectiveness study. J Neurol Neurosurg Psychiatry. 2006;
15. Tsivgoulis G, Frey JL, Flaster M, Sharma VK, Lao AY, Hoover SL, et al.
Pre-tissue plasminogen activator blood pressure levels and risk of symp-
tomatic intracerebral hemorrhage. Stroke. 2009;40:3631–3634.
16. Ahmed N, Wahlgren N, Brainin M, Castillo J, Ford GA, Kaste M, et al.
Relationship of blood pressure, antihypertensive therapy, and outcome in
ischemic stroke treated with intravenous thrombolysis: Retrospective
analysis from Safe Implementation of Thrombolysis in Stroke-
International Stroke Thrombolysis Register (SITS-ISTR). Stroke. 2009;
17. Prabhakaran S, Rivolta J, Vieira JR, Rincon F, Stillman J, Marshall RS.
Symptomatic intracerebral hemorrhage among eligible warfarin-treated
patients receiving intravenous tissue plasminogen activator for acute
ischemic stroke. Arch Neurol. 2010;67:559–563.
18. Seet RC, Zhang Y, Moore SA, Wijdicks EF, Rabinstein AA. Subthera-
peutic international normalized ratio in warfarin-treated patients increases
the risk for symptomatic intracerebral hemorrhage after intravenous
thrombolysis. Stroke. 2011;42:2333–2335.
19. Vergouwen MD, Casaubon LK, Swartz RH, Fang J, Stamplecoski M,
Kapral MK. Subtherapeutic warfarin is not associated with increased
hemorrhage rates in ischemic strokes treated with tissue plasminogen
activator. Stroke. 2011;42:1041–1045.
20. Hermann A, Dzialowski I, Koch R, Gahn G. Combined anti-platelet
therapy with aspirin and clopidogrel: risk factor for thrombolysis-related
intracerebral hemorrhage in acute ischemic stroke? J Neurol Sci. 2009;
21. Hacke W, Kaste M, Fieschi C, von Kummer R, Davalos A, Meier D, et
al. Randomised double-blind placebo-controlled trial of thrombolytic
therapy with intravenous alteplase in acute ischaemic stroke (ECASS II).
Second European-Australasian Acute Stroke Study investigators. Lancet.
Menon et al Risk Score for sICH With IV tPA
by guest on December 4, 2013http://stroke.ahajournals.org/Downloaded from
Risk Score for Intracranial Hemorrhage in Patients with Acute Ischemic Stroke Treated
With Intravenous t-PA.
Candidate Predictor Variables: Patient specific variables included demographics (age,
gender, race, baseline stroke severity by NIHSS), past medical history (prior stroke/TIA,
CAD, prior MI, prosthetic heart valve, hypertension, diabetes mellitus, atrial fibrillation,
smoking, dyslipidemia, heart failure, COPD/asthma, peripheral vascular disease, renal
insufficiency, on dialysis), laboratory parameters including blood glucose, Creatinine and
INR, vital signs including systolic blood pressure and heart rate on admission, interval times
and prior medications. Hospital level variables included number of beds, geographic region,
teaching status and number of acute ischemic strokes admitted and administered IV t-PA per
year. Less than 5% of data were missing for most candidate variables except blood glucose
(21%). Based on prior studies suggesting that hyperglycemia is a risk factor for sICH, we
retained serum glucose in the models and imputed missing values to the median value for
diabetics or non-diabetics, depending on whether the patient had a history of diabetes.
Otherwise, for categorical variables missing data were imputed to the most common category,
and for continuous variables to the median value.
We performed a sensitivity analysis and re-ran the final prediction model in patients with
complete information that met all study inclusion criteria. This study sample has 7938 (77.5%)
patients, and there are 366 sICH after IV tPA in this sample (4.61%). The model c statistic was
the same in patients with complete information (c statistic 0.71, Hosmer-Lemeshow test p
value=0.25) compared to the c statistic in the overall model including all patients (c statistic
0.71, Hosmer-Lemeshow test p value=0.48), showing that the prediction score works similarly
well in patients with complete information.
upplemental Figure 1:Study population criteria
1286531 patients from 1563 hospitals (1 April 2003 to 30 June 2010)
Records entered prior to 1 January
2009 (prior to requirement for
complete laboratory information to
be entered) (N=842562)
13481 ischemic stroke patients
who presenting to the ED
within 3 hours and were treated
with IV t-PA
1) Investigational or experimental
protocol for thrombolysis (n=117)
2) Transfer-in from a different hospital
3) Transfer-out to another hospital within
2 days thus precluding information on
the presence or absence of sICH within
36 hours (n=268)
4) Treatment with intra-arterial therapies
5) INR >1.7 (above the cut-off
recommended by the U.S. F.D.A. drug
6) Missing information (11.1%):
a) Missing discharge outcome (n=71)
b) Missing NIH stroke scale score (a
critical covariate, n=1342)
c) Patients from hospitals with <75%
data completion (n=84 from 13
443,969 ischemic stroke
patients (1 January 2009 to 30
1) Non ischemic strokes (N=175780)
2) Did not receive IV tPA (N=250726)
3) Unknown time of stroke onset
4) Onset to needle time>3 hours
5) In hospital stroke (N=568)
Final Study Population