Revised iScore to Predict Outcomes after Acute Ischemic Stroke Sheng-Feng Sung, MD,*† Yu-Wei Chen, MD,‡x Ling-Chien Hung, MD,* and Huey-Juan Lin, MD, MPHjj{
The iScore is a validated tool to predict mortality and functional outcome after acute ischemic stroke. It incorporates stroke subtype according to the Trial of Org 10172 in Acute Stroke Treatment (TOAST) classification as one of its factors. However, the TOAST stroke subtype may not be easily determined without extensive investigations. We aimed to test if the stroke subtype can be substituted by the Oxfordshire Community Stroke Project (OCSP) classification. We applied the iScore and the revised iScore, in which the TOASTsubtype was replaced by the OCSP classification, to patients admitted to a single hospital for acute ischemic stroke. Outcome measures included poor functional status (modified Rankin scale score, 3-6) at discharge and 3 months. The performance between the iScore and the revised iScore was assessed by determining the discrimination and calibration of the scores. We studied 3196 patients at the acute stage, and among them 2349 patients were available for the 3-month assessment. The discrimination of the revised iScore was comparable with the iScore for poor outcome at discharge (area under the receiver operating characteristic curve, .767 versus .775; P 5.06) and at 3-month (.801 versus .810; P 5.06). The correlation between the observed and the expected outcomes was high for both the iScore (Pearson correlation coefficient, .993 at discharge and .995 at 3 months; both P , .0001) and the revised iScore (.985 and .993, respectively; both P , .0001). The revised iScore reliably predicts clinical outcomes at discharge and 3 months for patients with acute ischemic stroke. Key Words: Acute ischemic stroke—prediction— risk score—outcome—prognosis. Ó 2014 by National Stroke Association
Introduction Accurate estimation of stroke prognosis benefits clinical management and may help health care providers communicate effectively with patients and their families. From the *Division of Neurology, Department of Internal Medicine, Ditmanson Medical Foundation Chia-Yi Christian Hospital, Chiayi City; †Min-Hwei College of Health Care Management, Tainan; ‡Department of Neurology, Landseed Hospital, Tao-Yuan County; xDepartment of Neurology, National Taiwan University Hospital, Taipei; jjDepartment of Neurology, Chi-Mei Medical Center, Tainan; and {Department of Cosmetic Science, Chia Nan University of Pharmacy and Science, Tainan, Taiwan. Received December 21, 2013; revision received January 5, 2014; accepted January 9, 2014. Address correspondence to Huey-Juan Lin, MD, MPH, Department of Neurology, Chi-Mei Medical Center, 901 Chung-Hwa Road, Yong-Kang District, Tainan 710, Taiwan. E-mail: huikuan@mail. chimei.org.tw. 1052-3057/$ - see front matter Ó 2014 by National Stroke Association http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2014.01.010
However, clinicians with expertise in stroke care might perform inferior to a validated risk score in predicting stroke outcomes.1 The iScore (details listed in Table 1) originally aimed to estimate the risk of short- and longterm mortality after an acute ischemic stroke.2 It also reasonably predicts poor functional outcomes early after hospitalization and the effectiveness of thrombolytic therapy.3,4 Notably, the score has been externally validated in different populations.5-7 The iScore incorporates clinical parameters available during hospitalization, including demographic data, stroke severity, risk factors, stroke subtype, and comorbid conditions.2 However, the use of the Trial of Org 10172 in Acute Stroke Treatment (TOAST)8 for subtype classification requires extensive diagnostic laboratory tests and may be limited to availability and feasibility. On the other hand, the Oxfordshire Community Stroke Project (OCSP) classification9 is based on clinical syndromes alone and can be readily obtained on admission. An OCSP diagnosis of lacunar infarct (LACI) made early after stroke onset
Journal of Stroke and Cerebrovascular Diseases, Vol. -, No. - (---), 2014: pp 1-6
1
S.-F. SUNG ET AL.
2
Table 1. Risk scoring system derived from the 30-day mortality in the iScore Variable
Score
Age (y) Sex Female Male Stroke severity (using Canadian Neurologic Scale score) 0 #4 5-7 $8 Stroke subtype Lacunar Nonlacunar Undetermined etiology Risk factor Atrial fibrillation Congestive heart failure Comorbid condition Cancer Renal dialysis Preadmission disability Independent Dependent Glucose on admission ,135 mg/dL $135 mg/dL
1Age (y) 0 110 1105 165 140 0 0 130 135 110 110 110 135 0 115 0 115
could predict an etiology of small artery occlusion using the TOAST criteria.10 In addition, the prevalence of intracranial large arterial abnormalities is very low in patients classified as LACIs.11 OCSP seems a reasonable alternative for TOAST in the iScore. Therefore, we conducted this hospital-based cohort study to compare the performance between the iScore and the revised iScore, in which the TOAST subtype was replaced by the OCSP classification.
acute corresponding ischemic lesion(s) on diffusionweighted magnetic resonance imaging. In the present study, we included patients aged 18 years or older admitted between October 2007 and September 2013. Patients with in-hospital stroke and those without complete data for computation of the iScore were excluded. All patients underwent a thorough clinical assessment on admission including the OCSP classification. Stroke severity was determined by the National Institutes of Health Stroke Scale (NIHSS) score. At discharge, the cause of stroke was classified according to the TOAST criteria by the attending neurologists using all available diagnostic information. Functional status at discharge was measured with the modified Rankin Scale (mRS) score. For patients with written informed consent for follow-up evaluation, an mRS score at 3 months after stroke was obtained by the direct assessment or by telephone interview. The study protocol was approved by the Institutional Review Board of Chia-Yi Christian Hospital.
Risk Scores Variables pertaining to the iScore were documented in the registry except the history of cancer, which was collected by chart review. The baseline NIHSS score was transformed to the Canadian Neurologic Scale (CNS) score according to the following rules: a CNS score of 0 equals an NIHSS score of greater than 22, a CNS score of 1-4 equals an NIHSS score of 14-22, a CNS score of 57 equals an NIHSS score of 9-13, and a CNS score of 8 or more equals an NIHSS score of 8 or less.13 We calculated the 30-day iScore (Table 1) for each patient. To test our study hypothesis, we substituted the OCSP classification for the TOAST stroke subtype and modified the scoring accordingly to construct a revised iScore. Specifically, patients with LACI according to the OCSP classification were assigned 0 point, whereas all others were assigned 30 points.
Outcome Measures
Methods Patients We identified patients with acute ischemic stroke from a hospital-based registry, which prospectively registered all stroke patients admitted within 10 days of symptom onset to the Ditmanson Medical Foundation Chia-Yi Christian Hospital. The hospital where the study was conducted is a 1000-bed regional hospital serving a city and its adjoining rural area consisting of around 500,000 inhabitants. The registry has been set up conforming to the design of the nationwide Taiwan Stroke Registry.12 Ischemic stroke was defined as an acute onset of neurologic deficits persisting longer than 24 hours with no hemorrhage on the first brain computed tomography or with
The outcome of interest was poor functional outcome defined as having an mRS score of 3-6 at discharge from acute care hospitalization. The secondary outcome was poor functional status at 3 months, which excluded patients discharged alive but did not provide consent for follow-up.
Statistical Analysis Continuous variables were expressed as the mean 6 standard deviation or median (interquartile range), and categorical variables as counts and percentages. Chisquare tests or Fisher’s exact tests were used to compare categorical variables, and t tests or Mann–Whitney U tests for continuous variables whichever appropriate. The
REVISED ISCORE TO PREDICT STROKE OUTCOMES
relationship between the TOASTand the OCSP stroke subtype classification schemes was assessed using the phi coefficient. The TOAST subtype of ‘‘small artery occlusion (lacune)’’ was used as the reference standard in the calculation of the positive and negative predictive values, sensitivity, specificity, and accuracy of an OCSP diagnosis of LACI. Confidence intervals (CIs) were calculated by using the exact binomial procedure. We performed separate univariate logistic regressions for the iScore and the revised iScore to determine their performance for predicting the poor functional outcome at discharge and 3 months. The iScore or the revised iScore was entered in the models as a continuous variable. Using these models, we obtained the predicted probability of the outcome for each patient
3
according to his/her iScore or revised iScore. Then we divided patients into risk score groups, which were basically defined at 10-point intervals starting from 60 or less, 61-70, 71-80, ., and up to 251 or more. For each risk score group, we divided the sum of the predicted probabilities of patients by the total number of patients in that group to obtain the predicted outcome for that risk score group. The model discrimination was judged by the area under the receiver operating characteristic curve (AUC), and compared with the DeLong method.14 We assessed the model calibration by the Hosmer–Lemeshow goodness-of-fit statistic. Because the Hosmer–Lemeshow test is known to be oversensitive to small deviations from good fit in large samples, we also used Pearson
Table 2. Characteristics of the study patients Follow-up at 3 months Characteristic Demographics Age, mean (SD) (y) Female NIHSS score #8 9-13 14-22 .22 Median (IQR) TOAST subtype Small artery occlusion (lacune) Nonlacunar Undetermined etiology OCSP subtype Lacunar infarcts Others Risk factors Hypertension Diabetes mellitus Hyperlipidemia Atrial fibrillation Coronary artery disease Congestive heart failure Current smoker Comorbid conditions Cancer Renal dialysis Dependent before admission Glucose on admission $135 mg/dL Intravenous thrombolysis Rehabilitation Length of stay, median (IQR) iScore, median (IQR) Revised iScore, median (IQR)
All (n 5 3196)
Yes (n 5 2349)
No (n 5 847)
P value
69.3 (12.1) 1321 (41.3)
69.2 (12.1) 955 (40.7)
69.6 (12.0) 366 (43.2)
.435 .195 .244
2219 (69.4) 378 (11.8) 355 (11.1) 244 (7.6) 5 (3-10)
1647 (70.1) 269 (11.5) 249 (10.6) 184 (7.8) 5 (3-10)
572 (67.5) 109 (12.9) 106 (12.5) 60 (7.1) 5 (3-11)
1198 (37.5) 1251 (39.1) 747 (23.4)
865 (36.8) 933 (39.7) 551 (23.5)
333 (39.3) 318 (37.5) 196 (23.1)
1210 (37.9) 1986 (62.1)
877 (37.3) 1472 (62.7)
333 (39.3) 514 (60.7)
2580 (80.7) 1402 (43.9) 1804 (56.5) 551 (17.2) 412 (12.9) 181 (5.7) 740 (23.2)
1905 (81.1) 1026 (43.7) 1336 (56.9) 426 (18.1) 310 (13.2) 126 (5.4) 549 (23.4)
675 (79.7) 376 (44.4) 468 (55.3) 125 (14.8) 102 (12.0) 55 (6.5) 191 (22.6)
.374 .720 .415 .026 .390 .223 .627
182 (5.8) 54 (1.7) 392 (12.3) 1647 (51.5) 187 (5.9) 1933 (60.5) 7 (4-10) 117 (90-155) 115 (90-152)
142 (6.1) 43 (1.8) 291 (12.4) 1205 (51.3) 152 (6.5) 1408 (59.9) 7 (5-10) 118 (89-155) 115 (91-152)
40 (4.8) 11 (1.3) 101 (11.9) 442 (52.2) 35 (4.1) 525 (62.0) 7 (4-10) 115 (91-155) 115 (90-153)
.163 .303 .724 .658 .013 .297 .276 .940 .954
.610 .405
.308
Abbreviations: IQR, interquartile range; NIHSS, National Institutes of Health Stroke Scale; OCSP, Oxfordshire Community Stroke Project; SD, standard deviation; TOAST, Trial of Org 10172 in Acute Stroke Treatment. Data are numbers (percentage) unless specified otherwise.
S.-F. SUNG ET AL.
4
Table 3. Comparison of the TOAST and the OCSP classifications OCSP TOAST
LACIs
PACIs
TACIs
POCIs
Unknown
Total
Small artery occlusion (lacune) Large artery atherosclerosis Cardioembolism Other determined etiology Undetermined etiology Total
861 (71.2) 99 (8.2) 54 (4.5) 10 (.8) 186 (15.4) 1210 (100)
185 (18.2) 351 (34.6) 154 (15.2) 18 (1.8) 308 (30.3) 1016 (100)
4 (.9) 156 (36.2) 154 (35.7) 3 (.7) 114 (26.5) 431 (100)
146 (27.3) 189 (35.4) 52 (9.7) 11 (2.1) 136 (25.5) 534 (100)
2 (40.0) 0 (0) 0 (0) 0 (0) 3 (60.0) 5 (100)
1198 (37.5) 795 (24.9) 414 (13.0) 42 (1.3) 747 (23.4) 3196 (100)
Abbreviations: LACIs, lacunar infarcts; OCSP, Oxfordshire Community Stroke Project; PACIs, partial anterior circulation infarcts; POCIs, posterior–anterior circulation infarcts; TACIs, total anterior circulation infarcts; TOAST, Trial of Org 10172 in Acute Stroke Treatment. Data are numbers (percentage).
correlation coefficient to compare predicted versus observed outcomes at the risk score level. The observed and predicted rates were plotted as continuous function of the risk scores. Statistical analyses were performed using Stata 13.1 (StataCorp, College Station, Texas).
Results After excluding 96 patients with in-hospital stroke and 212 patients with missing baseline variables, we enrolled 3196 patients, of whom 78 died at discharge. Among the 2641 patients who consented to follow-up, 370 patients were lost at 3 months. Hence, the 3-month cohort comprised 2349 patients, 73% of the entire cohort (see Supplement Fig S1). The baseline characteristics between patients with and without 3-month follow-up were comparable except there were more patients with atrial fibrillation in the 3-month cohort (Table 2). Overall, 5.9% of patients received intravenous thrombolysis and 60.5% underwent rehabilitation. Patients in the 3-month cohort had a higher proportion of thrombolytic treatment. The median hospital length of stay was 7 (interquartile range, 4-10) days. The proportion of small artery occlusion according to the TOAST classification was similar to that of LACI by the OCSP classification, either in the entire or the 3-month cohort, around 37%-38%. In Table 3, comparison of the OCSP and TOAST classifications is pre-
sented. The positive predictive value of an OCSP diagnosis of LACI for a TOAST classification of small artery occlusion was 71.2% (95% CI, 68.5%-73.7%); negative predictive value, 83.0% (95% CI, 81.3%-84.7%); sensitivity, 71.9% (95% CI, 69.2%-74.4%); specificity, 82.5% (95% CI, 80.8%-84.2%); and accuracy, 78.5% (95% CI, 77.1%79.9%). Being classified as small artery occlusion (lacune) according to the TOAST classification was highly associated with being classified as LACI according to the OCSP classification (phi coefficient 5 .54). Poor functional outcome including death occurred in 1813 (56.8%) patients at the time of discharge and in 1041 (44.3%) patients of the follow-up cohort at 3 months after stroke onset. The discrimination of the revised iScore was comparable to the iScore for poor outcome at discharge (AUC, .767 versus .775; P 5 .06; Fig 1A) and at 3-month (AUC, .801 versus .810; P 5 .06; Fig 1B). The iScore was well calibrated for the 3-month outcome (Hosmer–Lemeshow statistic, P 5 .345) but was unsatisfactory for the outcome at discharge (P ,.01). The calibration of the revised iScore was diminished for both outcomes at discharge (P , .001) and 3 months (P 5.02). Therefore, we plotted observed versus predicted outcomes (Fig 2 and Supplement Tables S1-S4). The correlation between the observed and expected rates of poor functional outcomes was high for both the iScore (Pearson correlation coefficient, .993 for outcome at discharge and
Figure 1. Receiver operating characteristic curves for the iScore and the revised iScore to predict outcomes at discharge (A) and 3 months (B).
REVISED ISCORE TO PREDICT STROKE OUTCOMES
Figure 2.
5
Observed versus predicted outcomes at discharge and 3 months according to the iScore (A, B) and the revised iScore (C, D).
.995 for outcome at 3 months; both P , .0001) and the revised iScore (.985 and .993; both P , .0001).
Discussion We demonstrated that by replacing the TOAST subtypes with the OCSP classifications, both the iScore and the revised iScore are of predictive value for identifying patients at high risk of death or disability at discharge and 3 months after acute ischemic stroke. The iScore slightly outperformed the revised iScore in discriminative performance and model fit. The TOAST subtype of small artery occlusion predicts both functional outcomes independent of the baseline NIHSS,15 and the magnitude of early improvement in stroke severity.16 The iScore studies also showed the added value of the TOAST subtype in predicting early functional outcomes after stroke.2,3 Exclusion of stroke subtype from the iScore will lead to very different estimates of outcomes. Nevertheless, it is not easy to classify the TOAST subtype in the early hours of hospital arrival when extensive cerebrovascular investigations are not performed yet.10 In addition, etiologic classification of ischemic stroke is infeasible in settings, which lack adequate medical resources. Without complete study, the classification of the TOAST stroke subtype is likely to be inaccurate.17,18
On the contrary, the OCSP classification is simple for use in clinical practice even if results of detailed investigations are unavailable. The OCSP syndromes assessed within 6 hours of stroke correlate well with the pattern of ischemic change on computed tomography in patients with nonlacunar stroke.19 Moreover, the functional outcome and natural history of acute ischemic stroke have been found to be strikingly different between the groups according to the OCSP classification.9 As compared with LACI, posterior circulation infarcts have higher likelihood of death and handicap at 3 months.20 The OCSP classification has also been included in a validated stroke prognosis score to predict early mortality and the length of hospital stay in acute stroke.21,22 A previous study has shown that a diagnosis of LACI using the OCSP classification within 6 hours of stroke predicts 76% of a TOAST diagnosis of small artery occlusion and has a similar capability to estimate stroke outcome at 3 months.10 The 71% positive predictive value in our study was comparable. The inferior performance on non-LACIs (negative predictive value, 83%) relative to that in their study (93%) might be because of the higher prevalence of small artery occlusion (37% versus 16%) in our study. Using the OCSP diagnosis as a substitute for the TOAST definition of LACI causes an 11% increased estimate and an 11% decreased estimate of the poor functional outcome.
S.-F. SUNG ET AL.
6
Overall, the predictive performance of the revised iScore was similar to that of the iScore. Our study has limitations. First, this was a single hospital-based study, which might limit the generalizability of our findings. Second, the outcome measures in our study were different from those in the iScore study. However, our study results show that the performances of the iScore and the revised iScore were comparable for our designated outcomes. Third, we did not explore the effects of specific treatments (eg, thrombolytic therapy and rehabilitation) on the functional outcomes.
Conclusions An OCSP diagnosis of LACI made on admission provides a reasonable substitute for small artery occlusion based on the TOAST criteria. Using the OCSP classification, the revised iScore can be reliably applied to patients with acute ischemic stroke to predict clinical outcomes at discharge and 3 months.
Supplementary Data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2014. 01.010.
References 1. Saposnik G, Cote R, Mamdani M, et al. Accuracy of clinician vs risk score prediction of ischemic stroke outcomes. Neurology 2013;81:448-455. 2. Saposnik G, Kapral MK, Liu Y, et al. IScore: a risk score to predict death early after hospitalization for an acute ischemic stroke. Circulation 2011;123:739-749. 3. Saposnik G, Raptis S, Kapral MK, et al. The iScore predicts poor functional outcomes early after hospitalization for an acute ischemic stroke. Stroke 2011;42:3421-3428. 4. Saposnik G, Fang J, Kapral MK, et al. The iScore predicts effectiveness of thrombolytic therapy for acute ischemic stroke. Stroke 2012;43:1315-1322. 5. Park TH, Saposnik G, Bae HJ, et al. The iScore predicts functional outcome in Korean patients with ischemic stroke. Stroke 2013;44:1440-1442. 6. Zhang N, Liu G, Zhang G, et al. External validation of the iScore for predicting ischemic stroke mortality in patients in China. Stroke 2013;44:1924-1929. 7. Dragoumanos V, Tzirogiannis KN, Panoutsopoulos GI, et al. Evaluation of IScore validity in a Greek cohort of patients with type 2 diabetes. BMC Neurol 2013;13:121. 8. Adams HP, Bendixen BH, Kappelle LJ, et al. Classification of subtype of acute ischemic stroke. Definitions for use in a multicenter clinical trial. TOAST. Trial of Org 10172 in Acute Stroke Treatment. Stroke 1993;24:35-41.
9. Bamford J, Sandercock P, Dennis M, et al. Classification and natural history of clinically identifiable subtypes of cerebral infarction. Lancet 1991;337:1521-1526. 10. Phillips SJ, Dai D, Mitnitski A, et al. Clinical diagnosis of lacunar stroke in the first 6 hours after symptom onset: analysis of data from the glycine antagonist in neuroprotection (GAIN) Americas trial. Stroke 2007;38: 2706-2711. 11. Mead GE, Wardlaw JM, Dennis MS, et al. Relationship between pattern of intracranial artery abnormalities on transcranial doppler and Oxfordshire Community Stroke Project clinical classification of ischemic stroke. Stroke 2000;31:714-719. 12. Hsieh FI, Lien LM, Chen ST, et al. Get With the GuidelinesStroke performance indicators: surveillance of stroke care in the Taiwan Stroke Registry: Get With the GuidelinesStroke in Taiwan. Circulation 2010;122:1116-1123. 13. Nilanont Y, Komoltri C, Saposnik G, et al. The Canadian Neurological Scale and the NIHSS: development and validation of a simple conversion model. Cerebrovasc Dis 2010;30:120-126. 14. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44:837-845. 15. Adams HP, Davis PH, Leira EC, et al. Baseline NIH Stroke Scale score strongly predicts outcome after stroke: a report of the Trial of Org 10172 in Acute Stroke Treatment (TOAST). Neurology 1999;53:126-131. 16. Sprigg N, Gray LJ, Bath PMW, et al. Early recovery and functional outcome are related with causal stroke subtype: data from the tinzaparin in acute ischemic stroke trial. J Stroke Cerebrovasc Dis 2007;16:180-184. 17. Madden KP, Karanjia PN, Adams HP, et al. Accuracy of initial stroke subtype diagnosis in the TOAST study. Trial of ORG 10172 in Acute Stroke Treatment. Neurology 1995;45:1975-1979. 18. Fure B, Wyller TB, Thommessen B. TOAST criteria applied in acute ischemic stroke. Acta Neurol Scand 2005;112:254-258. 19. Kobayashi A, Wardlaw JM, Lindley RI, et al. Oxfordshire community stroke project clinical stroke syndrome and appearances of tissue and vascular lesions on pretreatment CT in hyperacute ischemic stroke among the first 510 patients in the Third International Stroke Trial (IST3). Stroke 2009;40:743-748. 20. Di Carlo A, Lamassa M, Baldereschi M, et al. Risk factors and outcome of subtypes of ischemic stroke. Data from a multicenter multinational hospital-based registry. The European Community Stroke Project. J Neurol Sci 2006; 244:143-150. 21. Myint PK, Clark AB, Kwok CS, et al. A simple 8-point score strongly predicts early outcomes in acute stroke. Int J Stroke 2013; http://dx.doi.org/10.1111/ijs.12088 [Epub ahead of print]. 22. Kwok CS, Potter JF, Dalton G, et al. The SOAR stroke score predicts inpatient and 7-day mortality in acute stroke. Stroke 2013;44:2010-2012.
The iScore is a validated tool to predict mortality and functional outcome after acute ischemic stroke. It incorporates stroke subtype according to the Trial of Org 10172 in Acute Stroke Treatment (TOAST) classification as one of its factors. However, the TOAST stroke subtype may not be easily determined without extensive investigations. We aimed to test if the stroke subtype can be substituted by the Oxfordshire Community Stroke Project (OCSP) classification. We applied the iScore and the revised iScore, in which the TOASTsubtype was replaced by the OCSP classification, to patients admitted to a single hospital for acute ischemic stroke. Outcome measures included poor functional status (modified Rankin scale score, 3-6) at discharge and 3 months. The performance between the iScore and the revised iScore was assessed by determining the discrimination and calibration of the scores. We studied 3196 patients at the acute stage, and among them 2349 patients were available for the 3-month assessment. The discrimination of the revised iScore was comparable with the iScore for poor outcome at discharge (area under the receiver operating characteristic curve, .767 versus .775; P 5.06) and at 3-month (.801 versus .810; P 5.06). The correlation between the observed and the expected outcomes was high for both the iScore (Pearson correlation coefficient, .993 at discharge and .995 at 3 months; both P , .0001) and the revised iScore (.985 and .993, respectively; both P , .0001). The revised iScore reliably predicts clinical outcomes at discharge and 3 months for patients with acute ischemic stroke. Key Words: Acute ischemic stroke—prediction— risk score—outcome—prognosis. Ó 2014 by National Stroke Association
Introduction Accurate estimation of stroke prognosis benefits clinical management and may help health care providers communicate effectively with patients and their families. From the *Division of Neurology, Department of Internal Medicine, Ditmanson Medical Foundation Chia-Yi Christian Hospital, Chiayi City; †Min-Hwei College of Health Care Management, Tainan; ‡Department of Neurology, Landseed Hospital, Tao-Yuan County; xDepartment of Neurology, National Taiwan University Hospital, Taipei; jjDepartment of Neurology, Chi-Mei Medical Center, Tainan; and {Department of Cosmetic Science, Chia Nan University of Pharmacy and Science, Tainan, Taiwan. Received December 21, 2013; revision received January 5, 2014; accepted January 9, 2014. Address correspondence to Huey-Juan Lin, MD, MPH, Department of Neurology, Chi-Mei Medical Center, 901 Chung-Hwa Road, Yong-Kang District, Tainan 710, Taiwan. E-mail: huikuan@mail. chimei.org.tw. 1052-3057/$ - see front matter Ó 2014 by National Stroke Association http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2014.01.010
However, clinicians with expertise in stroke care might perform inferior to a validated risk score in predicting stroke outcomes.1 The iScore (details listed in Table 1) originally aimed to estimate the risk of short- and longterm mortality after an acute ischemic stroke.2 It also reasonably predicts poor functional outcomes early after hospitalization and the effectiveness of thrombolytic therapy.3,4 Notably, the score has been externally validated in different populations.5-7 The iScore incorporates clinical parameters available during hospitalization, including demographic data, stroke severity, risk factors, stroke subtype, and comorbid conditions.2 However, the use of the Trial of Org 10172 in Acute Stroke Treatment (TOAST)8 for subtype classification requires extensive diagnostic laboratory tests and may be limited to availability and feasibility. On the other hand, the Oxfordshire Community Stroke Project (OCSP) classification9 is based on clinical syndromes alone and can be readily obtained on admission. An OCSP diagnosis of lacunar infarct (LACI) made early after stroke onset
Journal of Stroke and Cerebrovascular Diseases, Vol. -, No. - (---), 2014: pp 1-6
1
S.-F. SUNG ET AL.
2
Table 1. Risk scoring system derived from the 30-day mortality in the iScore Variable
Score
Age (y) Sex Female Male Stroke severity (using Canadian Neurologic Scale score) 0 #4 5-7 $8 Stroke subtype Lacunar Nonlacunar Undetermined etiology Risk factor Atrial fibrillation Congestive heart failure Comorbid condition Cancer Renal dialysis Preadmission disability Independent Dependent Glucose on admission ,135 mg/dL $135 mg/dL
1Age (y) 0 110 1105 165 140 0 0 130 135 110 110 110 135 0 115 0 115
could predict an etiology of small artery occlusion using the TOAST criteria.10 In addition, the prevalence of intracranial large arterial abnormalities is very low in patients classified as LACIs.11 OCSP seems a reasonable alternative for TOAST in the iScore. Therefore, we conducted this hospital-based cohort study to compare the performance between the iScore and the revised iScore, in which the TOAST subtype was replaced by the OCSP classification.
acute corresponding ischemic lesion(s) on diffusionweighted magnetic resonance imaging. In the present study, we included patients aged 18 years or older admitted between October 2007 and September 2013. Patients with in-hospital stroke and those without complete data for computation of the iScore were excluded. All patients underwent a thorough clinical assessment on admission including the OCSP classification. Stroke severity was determined by the National Institutes of Health Stroke Scale (NIHSS) score. At discharge, the cause of stroke was classified according to the TOAST criteria by the attending neurologists using all available diagnostic information. Functional status at discharge was measured with the modified Rankin Scale (mRS) score. For patients with written informed consent for follow-up evaluation, an mRS score at 3 months after stroke was obtained by the direct assessment or by telephone interview. The study protocol was approved by the Institutional Review Board of Chia-Yi Christian Hospital.
Risk Scores Variables pertaining to the iScore were documented in the registry except the history of cancer, which was collected by chart review. The baseline NIHSS score was transformed to the Canadian Neurologic Scale (CNS) score according to the following rules: a CNS score of 0 equals an NIHSS score of greater than 22, a CNS score of 1-4 equals an NIHSS score of 14-22, a CNS score of 57 equals an NIHSS score of 9-13, and a CNS score of 8 or more equals an NIHSS score of 8 or less.13 We calculated the 30-day iScore (Table 1) for each patient. To test our study hypothesis, we substituted the OCSP classification for the TOAST stroke subtype and modified the scoring accordingly to construct a revised iScore. Specifically, patients with LACI according to the OCSP classification were assigned 0 point, whereas all others were assigned 30 points.
Outcome Measures
Methods Patients We identified patients with acute ischemic stroke from a hospital-based registry, which prospectively registered all stroke patients admitted within 10 days of symptom onset to the Ditmanson Medical Foundation Chia-Yi Christian Hospital. The hospital where the study was conducted is a 1000-bed regional hospital serving a city and its adjoining rural area consisting of around 500,000 inhabitants. The registry has been set up conforming to the design of the nationwide Taiwan Stroke Registry.12 Ischemic stroke was defined as an acute onset of neurologic deficits persisting longer than 24 hours with no hemorrhage on the first brain computed tomography or with
The outcome of interest was poor functional outcome defined as having an mRS score of 3-6 at discharge from acute care hospitalization. The secondary outcome was poor functional status at 3 months, which excluded patients discharged alive but did not provide consent for follow-up.
Statistical Analysis Continuous variables were expressed as the mean 6 standard deviation or median (interquartile range), and categorical variables as counts and percentages. Chisquare tests or Fisher’s exact tests were used to compare categorical variables, and t tests or Mann–Whitney U tests for continuous variables whichever appropriate. The
REVISED ISCORE TO PREDICT STROKE OUTCOMES
relationship between the TOASTand the OCSP stroke subtype classification schemes was assessed using the phi coefficient. The TOAST subtype of ‘‘small artery occlusion (lacune)’’ was used as the reference standard in the calculation of the positive and negative predictive values, sensitivity, specificity, and accuracy of an OCSP diagnosis of LACI. Confidence intervals (CIs) were calculated by using the exact binomial procedure. We performed separate univariate logistic regressions for the iScore and the revised iScore to determine their performance for predicting the poor functional outcome at discharge and 3 months. The iScore or the revised iScore was entered in the models as a continuous variable. Using these models, we obtained the predicted probability of the outcome for each patient
3
according to his/her iScore or revised iScore. Then we divided patients into risk score groups, which were basically defined at 10-point intervals starting from 60 or less, 61-70, 71-80, ., and up to 251 or more. For each risk score group, we divided the sum of the predicted probabilities of patients by the total number of patients in that group to obtain the predicted outcome for that risk score group. The model discrimination was judged by the area under the receiver operating characteristic curve (AUC), and compared with the DeLong method.14 We assessed the model calibration by the Hosmer–Lemeshow goodness-of-fit statistic. Because the Hosmer–Lemeshow test is known to be oversensitive to small deviations from good fit in large samples, we also used Pearson
Table 2. Characteristics of the study patients Follow-up at 3 months Characteristic Demographics Age, mean (SD) (y) Female NIHSS score #8 9-13 14-22 .22 Median (IQR) TOAST subtype Small artery occlusion (lacune) Nonlacunar Undetermined etiology OCSP subtype Lacunar infarcts Others Risk factors Hypertension Diabetes mellitus Hyperlipidemia Atrial fibrillation Coronary artery disease Congestive heart failure Current smoker Comorbid conditions Cancer Renal dialysis Dependent before admission Glucose on admission $135 mg/dL Intravenous thrombolysis Rehabilitation Length of stay, median (IQR) iScore, median (IQR) Revised iScore, median (IQR)
All (n 5 3196)
Yes (n 5 2349)
No (n 5 847)
P value
69.3 (12.1) 1321 (41.3)
69.2 (12.1) 955 (40.7)
69.6 (12.0) 366 (43.2)
.435 .195 .244
2219 (69.4) 378 (11.8) 355 (11.1) 244 (7.6) 5 (3-10)
1647 (70.1) 269 (11.5) 249 (10.6) 184 (7.8) 5 (3-10)
572 (67.5) 109 (12.9) 106 (12.5) 60 (7.1) 5 (3-11)
1198 (37.5) 1251 (39.1) 747 (23.4)
865 (36.8) 933 (39.7) 551 (23.5)
333 (39.3) 318 (37.5) 196 (23.1)
1210 (37.9) 1986 (62.1)
877 (37.3) 1472 (62.7)
333 (39.3) 514 (60.7)
2580 (80.7) 1402 (43.9) 1804 (56.5) 551 (17.2) 412 (12.9) 181 (5.7) 740 (23.2)
1905 (81.1) 1026 (43.7) 1336 (56.9) 426 (18.1) 310 (13.2) 126 (5.4) 549 (23.4)
675 (79.7) 376 (44.4) 468 (55.3) 125 (14.8) 102 (12.0) 55 (6.5) 191 (22.6)
.374 .720 .415 .026 .390 .223 .627
182 (5.8) 54 (1.7) 392 (12.3) 1647 (51.5) 187 (5.9) 1933 (60.5) 7 (4-10) 117 (90-155) 115 (90-152)
142 (6.1) 43 (1.8) 291 (12.4) 1205 (51.3) 152 (6.5) 1408 (59.9) 7 (5-10) 118 (89-155) 115 (91-152)
40 (4.8) 11 (1.3) 101 (11.9) 442 (52.2) 35 (4.1) 525 (62.0) 7 (4-10) 115 (91-155) 115 (90-153)
.163 .303 .724 .658 .013 .297 .276 .940 .954
.610 .405
.308
Abbreviations: IQR, interquartile range; NIHSS, National Institutes of Health Stroke Scale; OCSP, Oxfordshire Community Stroke Project; SD, standard deviation; TOAST, Trial of Org 10172 in Acute Stroke Treatment. Data are numbers (percentage) unless specified otherwise.
S.-F. SUNG ET AL.
4
Table 3. Comparison of the TOAST and the OCSP classifications OCSP TOAST
LACIs
PACIs
TACIs
POCIs
Unknown
Total
Small artery occlusion (lacune) Large artery atherosclerosis Cardioembolism Other determined etiology Undetermined etiology Total
861 (71.2) 99 (8.2) 54 (4.5) 10 (.8) 186 (15.4) 1210 (100)
185 (18.2) 351 (34.6) 154 (15.2) 18 (1.8) 308 (30.3) 1016 (100)
4 (.9) 156 (36.2) 154 (35.7) 3 (.7) 114 (26.5) 431 (100)
146 (27.3) 189 (35.4) 52 (9.7) 11 (2.1) 136 (25.5) 534 (100)
2 (40.0) 0 (0) 0 (0) 0 (0) 3 (60.0) 5 (100)
1198 (37.5) 795 (24.9) 414 (13.0) 42 (1.3) 747 (23.4) 3196 (100)
Abbreviations: LACIs, lacunar infarcts; OCSP, Oxfordshire Community Stroke Project; PACIs, partial anterior circulation infarcts; POCIs, posterior–anterior circulation infarcts; TACIs, total anterior circulation infarcts; TOAST, Trial of Org 10172 in Acute Stroke Treatment. Data are numbers (percentage).
correlation coefficient to compare predicted versus observed outcomes at the risk score level. The observed and predicted rates were plotted as continuous function of the risk scores. Statistical analyses were performed using Stata 13.1 (StataCorp, College Station, Texas).
Results After excluding 96 patients with in-hospital stroke and 212 patients with missing baseline variables, we enrolled 3196 patients, of whom 78 died at discharge. Among the 2641 patients who consented to follow-up, 370 patients were lost at 3 months. Hence, the 3-month cohort comprised 2349 patients, 73% of the entire cohort (see Supplement Fig S1). The baseline characteristics between patients with and without 3-month follow-up were comparable except there were more patients with atrial fibrillation in the 3-month cohort (Table 2). Overall, 5.9% of patients received intravenous thrombolysis and 60.5% underwent rehabilitation. Patients in the 3-month cohort had a higher proportion of thrombolytic treatment. The median hospital length of stay was 7 (interquartile range, 4-10) days. The proportion of small artery occlusion according to the TOAST classification was similar to that of LACI by the OCSP classification, either in the entire or the 3-month cohort, around 37%-38%. In Table 3, comparison of the OCSP and TOAST classifications is pre-
sented. The positive predictive value of an OCSP diagnosis of LACI for a TOAST classification of small artery occlusion was 71.2% (95% CI, 68.5%-73.7%); negative predictive value, 83.0% (95% CI, 81.3%-84.7%); sensitivity, 71.9% (95% CI, 69.2%-74.4%); specificity, 82.5% (95% CI, 80.8%-84.2%); and accuracy, 78.5% (95% CI, 77.1%79.9%). Being classified as small artery occlusion (lacune) according to the TOAST classification was highly associated with being classified as LACI according to the OCSP classification (phi coefficient 5 .54). Poor functional outcome including death occurred in 1813 (56.8%) patients at the time of discharge and in 1041 (44.3%) patients of the follow-up cohort at 3 months after stroke onset. The discrimination of the revised iScore was comparable to the iScore for poor outcome at discharge (AUC, .767 versus .775; P 5 .06; Fig 1A) and at 3-month (AUC, .801 versus .810; P 5 .06; Fig 1B). The iScore was well calibrated for the 3-month outcome (Hosmer–Lemeshow statistic, P 5 .345) but was unsatisfactory for the outcome at discharge (P ,.01). The calibration of the revised iScore was diminished for both outcomes at discharge (P , .001) and 3 months (P 5.02). Therefore, we plotted observed versus predicted outcomes (Fig 2 and Supplement Tables S1-S4). The correlation between the observed and expected rates of poor functional outcomes was high for both the iScore (Pearson correlation coefficient, .993 for outcome at discharge and
Figure 1. Receiver operating characteristic curves for the iScore and the revised iScore to predict outcomes at discharge (A) and 3 months (B).
REVISED ISCORE TO PREDICT STROKE OUTCOMES
Figure 2.
5
Observed versus predicted outcomes at discharge and 3 months according to the iScore (A, B) and the revised iScore (C, D).
.995 for outcome at 3 months; both P , .0001) and the revised iScore (.985 and .993; both P , .0001).
Discussion We demonstrated that by replacing the TOAST subtypes with the OCSP classifications, both the iScore and the revised iScore are of predictive value for identifying patients at high risk of death or disability at discharge and 3 months after acute ischemic stroke. The iScore slightly outperformed the revised iScore in discriminative performance and model fit. The TOAST subtype of small artery occlusion predicts both functional outcomes independent of the baseline NIHSS,15 and the magnitude of early improvement in stroke severity.16 The iScore studies also showed the added value of the TOAST subtype in predicting early functional outcomes after stroke.2,3 Exclusion of stroke subtype from the iScore will lead to very different estimates of outcomes. Nevertheless, it is not easy to classify the TOAST subtype in the early hours of hospital arrival when extensive cerebrovascular investigations are not performed yet.10 In addition, etiologic classification of ischemic stroke is infeasible in settings, which lack adequate medical resources. Without complete study, the classification of the TOAST stroke subtype is likely to be inaccurate.17,18
On the contrary, the OCSP classification is simple for use in clinical practice even if results of detailed investigations are unavailable. The OCSP syndromes assessed within 6 hours of stroke correlate well with the pattern of ischemic change on computed tomography in patients with nonlacunar stroke.19 Moreover, the functional outcome and natural history of acute ischemic stroke have been found to be strikingly different between the groups according to the OCSP classification.9 As compared with LACI, posterior circulation infarcts have higher likelihood of death and handicap at 3 months.20 The OCSP classification has also been included in a validated stroke prognosis score to predict early mortality and the length of hospital stay in acute stroke.21,22 A previous study has shown that a diagnosis of LACI using the OCSP classification within 6 hours of stroke predicts 76% of a TOAST diagnosis of small artery occlusion and has a similar capability to estimate stroke outcome at 3 months.10 The 71% positive predictive value in our study was comparable. The inferior performance on non-LACIs (negative predictive value, 83%) relative to that in their study (93%) might be because of the higher prevalence of small artery occlusion (37% versus 16%) in our study. Using the OCSP diagnosis as a substitute for the TOAST definition of LACI causes an 11% increased estimate and an 11% decreased estimate of the poor functional outcome.
S.-F. SUNG ET AL.
6
Overall, the predictive performance of the revised iScore was similar to that of the iScore. Our study has limitations. First, this was a single hospital-based study, which might limit the generalizability of our findings. Second, the outcome measures in our study were different from those in the iScore study. However, our study results show that the performances of the iScore and the revised iScore were comparable for our designated outcomes. Third, we did not explore the effects of specific treatments (eg, thrombolytic therapy and rehabilitation) on the functional outcomes.
Conclusions An OCSP diagnosis of LACI made on admission provides a reasonable substitute for small artery occlusion based on the TOAST criteria. Using the OCSP classification, the revised iScore can be reliably applied to patients with acute ischemic stroke to predict clinical outcomes at discharge and 3 months.
Supplementary Data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2014. 01.010.
References 1. Saposnik G, Cote R, Mamdani M, et al. Accuracy of clinician vs risk score prediction of ischemic stroke outcomes. Neurology 2013;81:448-455. 2. Saposnik G, Kapral MK, Liu Y, et al. IScore: a risk score to predict death early after hospitalization for an acute ischemic stroke. Circulation 2011;123:739-749. 3. Saposnik G, Raptis S, Kapral MK, et al. The iScore predicts poor functional outcomes early after hospitalization for an acute ischemic stroke. Stroke 2011;42:3421-3428. 4. Saposnik G, Fang J, Kapral MK, et al. The iScore predicts effectiveness of thrombolytic therapy for acute ischemic stroke. Stroke 2012;43:1315-1322. 5. Park TH, Saposnik G, Bae HJ, et al. The iScore predicts functional outcome in Korean patients with ischemic stroke. Stroke 2013;44:1440-1442. 6. Zhang N, Liu G, Zhang G, et al. External validation of the iScore for predicting ischemic stroke mortality in patients in China. Stroke 2013;44:1924-1929. 7. Dragoumanos V, Tzirogiannis KN, Panoutsopoulos GI, et al. Evaluation of IScore validity in a Greek cohort of patients with type 2 diabetes. BMC Neurol 2013;13:121. 8. Adams HP, Bendixen BH, Kappelle LJ, et al. Classification of subtype of acute ischemic stroke. Definitions for use in a multicenter clinical trial. TOAST. Trial of Org 10172 in Acute Stroke Treatment. Stroke 1993;24:35-41.
9. Bamford J, Sandercock P, Dennis M, et al. Classification and natural history of clinically identifiable subtypes of cerebral infarction. Lancet 1991;337:1521-1526. 10. Phillips SJ, Dai D, Mitnitski A, et al. Clinical diagnosis of lacunar stroke in the first 6 hours after symptom onset: analysis of data from the glycine antagonist in neuroprotection (GAIN) Americas trial. Stroke 2007;38: 2706-2711. 11. Mead GE, Wardlaw JM, Dennis MS, et al. Relationship between pattern of intracranial artery abnormalities on transcranial doppler and Oxfordshire Community Stroke Project clinical classification of ischemic stroke. Stroke 2000;31:714-719. 12. Hsieh FI, Lien LM, Chen ST, et al. Get With the GuidelinesStroke performance indicators: surveillance of stroke care in the Taiwan Stroke Registry: Get With the GuidelinesStroke in Taiwan. Circulation 2010;122:1116-1123. 13. Nilanont Y, Komoltri C, Saposnik G, et al. The Canadian Neurological Scale and the NIHSS: development and validation of a simple conversion model. Cerebrovasc Dis 2010;30:120-126. 14. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44:837-845. 15. Adams HP, Davis PH, Leira EC, et al. Baseline NIH Stroke Scale score strongly predicts outcome after stroke: a report of the Trial of Org 10172 in Acute Stroke Treatment (TOAST). Neurology 1999;53:126-131. 16. Sprigg N, Gray LJ, Bath PMW, et al. Early recovery and functional outcome are related with causal stroke subtype: data from the tinzaparin in acute ischemic stroke trial. J Stroke Cerebrovasc Dis 2007;16:180-184. 17. Madden KP, Karanjia PN, Adams HP, et al. Accuracy of initial stroke subtype diagnosis in the TOAST study. Trial of ORG 10172 in Acute Stroke Treatment. Neurology 1995;45:1975-1979. 18. Fure B, Wyller TB, Thommessen B. TOAST criteria applied in acute ischemic stroke. Acta Neurol Scand 2005;112:254-258. 19. Kobayashi A, Wardlaw JM, Lindley RI, et al. Oxfordshire community stroke project clinical stroke syndrome and appearances of tissue and vascular lesions on pretreatment CT in hyperacute ischemic stroke among the first 510 patients in the Third International Stroke Trial (IST3). Stroke 2009;40:743-748. 20. Di Carlo A, Lamassa M, Baldereschi M, et al. Risk factors and outcome of subtypes of ischemic stroke. Data from a multicenter multinational hospital-based registry. The European Community Stroke Project. J Neurol Sci 2006; 244:143-150. 21. Myint PK, Clark AB, Kwok CS, et al. A simple 8-point score strongly predicts early outcomes in acute stroke. Int J Stroke 2013; http://dx.doi.org/10.1111/ijs.12088 [Epub ahead of print]. 22. Kwok CS, Potter JF, Dalton G, et al. The SOAR stroke score predicts inpatient and 7-day mortality in acute stroke. Stroke 2013;44:2010-2012.
