Literature
Home医源资料库在线期刊中风学杂志2006年第37卷第4期

Predicting Long-Term Independency in Activities of Daily Living After Middle Cerebral Artery Stroke

来源:中风学杂志
摘要:Predictingdisabilityinstroke-acriticalreviewoftheliterature。Predictingfunctionaloutcomeandsurvivalafteracuteischemicstroke。Predictingfunctionaloutcomeinacutestroke:comparisonofasimplesixvariablemodelwithotherpredictivesystemsandinformalclinicalprediction。...

点击显示 收起

    the Center of Excellence for Rehabilitation Medicine (S.K.S., G.K., M.W.M.P., A.J.H.P.), Rehabilitation Center De Hoogstraat Utrecht, the Netherlands
    Rudolf Magnus Institute of Neuroscience (S.K.S., G.K., M.W.M.P., L.J.K., A.J.H.P.), University Medical Center Utrecht, the Netherlands
    Institute for Rehabilitation Research (M.W.M.P.), Hoensbroek, the Netherlands
    Department of Neurology (L.J.K.), University Medical Center Utrecht, the Netherlands.

    Abstract

    Background and Purpose— To investigate whether neuroimaging information has added predictive value compared with clinical information for independency in activities of daily living (ADL) 1 year after stroke.

    Methods— Seventy-five first-ever middle cerebral artery stroke survivors were evaluated in logistic regression analyses. Model 1 was derived on the basis of clinical variables; for model 2, neuroimaging variables were added to model 1. Independent variables were stroke severity (National Institutes of Health Stroke Scale), consciousness (Glasgow Coma Scale), urinary continence, demographic variables (age, gender, relationship, educational level), hospital of admission, and clinical instruments: sitting balance (trunk control test), motor functioning (Motricity Index), and ADL (Barthel Index). Neuroimaging variables, determined on conventional MRI scans, included: number of days to scanning, lesion volume, lesion localization (cortex/subcortex), hemisphere, and the presence of white matter lesions. ADL independency was defined as 19 and 20 points on Barthel Index. Differences in accuracy of prediction of ADL independence between models 1 and 2 were analyzed by comparing areas under the curve (AUC) in a receiver operating characteristic analysis.

    Results— Model 1 contained as significant predictors: age and ADL (AUC 0.84), correctly predicting 77%. In model 2, number of days to scanning, hemisphere, and lesion volume were added to model 1, increasing the AUC from 0.84 to 0.87, accurately predicting 83% of the surviving patients.

    Conclusions— Clinical variables in the second week after stroke are good predictors for independency in ADL 1 year after stroke. Neuroimaging variables on conventional MRI scans do not have added value in long-term prediction of ADL.

    Key Words: activities of daily living  magnetic resonance imaging  stroke

    Introduction

    Stroke is the most common cause of dependency in activities of daily living (ADL) among elderly.1 Prediction of ADL at an early stage enables clinicians to set realistic therapeutic goals and to inform patients.2 In recent years, a number of prognostic studies have evaluated demographic, radiological, and neurological determinants, which either individually or in combination claim to predict functional outcome in stroke.2–9 One of these determinants is lesion volume, which is found to be significantly associated with functional outcome after stroke.6–8 However, studies on predictive values of lesion volume used different follow-up times (days5; 1 month;7 or 3 to 6 months6–8), whereas functional improvement may extent beyond 6 months10 after stroke. In addition, the heterogeneity of stroke patients and methodological inadequacies in previous studies2 are important factors that may preclude comparison between studies3 and confound outcomes.

    To date, only 3 studies examined the combination of clinical and imaging information attempting to predict 3-month outcome4,11 and 6-month functional outcome12 in ischemic stroke patients. Although lesion volume was significantly associated with functional outcome,4,11,12 these studies were not unanimous in the added value of neuroimaging as an independent factor for improving prognostic accuracy. Only Baird4 found that imaging in combination with National Institutes of Health Stroke Scale (NIHSS) was more accurate in predicting ADL at 3 months after stroke than NIHSS alone. However, none of these studies investigated whether information derived from brain imaging with MRI has added value in predicting ADL 1 year after stroke compared with information derived from clinical assessment only.

    The aim of the present study was to examine the added value of neuroimaging information from MRI scans obtained in the second week after stroke by comparing infarct volume and lesion localization with clinical variables in predicting independency in ADL of stroke survivors with middle cerebral artery (MCA) infarctions at 1 year after stroke.

    Methods

    Patients admitted to stroke units of 6 hospitals in the Netherlands in 1999 to 2001 were asked to participate. After informed consent, an MRI scan was obtained at a mean of 11 days (SD 3.5) after stroke. Data about stroke severity (NIHSS13 and Glasgow Coma Scale 14) on hospital admission were retrospectively retrieved from medical records,15 and demographic variables (age, gender, relationship) were collected. Other measures of functional status (Barthel Index [BI]16; trunk control test;17 Motricity Index, MI18) were part of a physical examination. An experienced researcher tested patients &6 days after stroke (mean 5.6 days; SD 2.6; range 1 to 13 days), blinded for the neuroimaging findings. The research protocol was approved by the medical ethics committee of University Medical Center Utrecht.

    Patients

    Patients included had a nonlacunar first-ever ischemic MCA infarction, were between 18 and 85 years of age, had a premorbid BI 18, and a stable neurological condition 1 week after stroke. All patients had a visible lesion on the MRI scan. Patients with other lesions, infarctions other than MCA infarctions such as infratentorial infarctions, and patients with border-zone infarctions, multiple infarctions, and lacunar infarctions were excluded. Lacunar infarctions were defined as infarctions with a diameter ranging from 3 to 4 mm to 15 to 20 mm and located at the site of the basal ganglia, internal capsule, or corona radiate.19,20 Patients with premorbid cognitive limitations or comorbidity influencing functional outcome were excluded. Patients who died or had another stroke during the follow-up period were excluded so that ADL of stroke survivors 1 year after stroke could be predicted.

    Dependent Variable

    Level of independency with the BI was assessed face to face by the same researcher 1 year after stroke (mean 377 days after stroke; SD 22). Persons with BI scores 19 and 20 were considered independent.9

    Independent Variables

    Clinical Variables

    The NIHSS13 was normally distributed in our study group and used as a continuous variable. For the GCS,14 a cutoff point of 15 (<15 and 15) was used because patients with a GCS of 15 are conscious, showing adequate response. Binary determinants such as gender (male/female), relationship (Yes/No [Y/N]), and urinary incontinence (Y/N) were collected. Age was normally distributed in our study population and used as a continuous variable. Educational level was dichotomized: low education (no education to 10 years of education) and high education (completed high school education).21 Sitting balance (Y/N) was assessed using the trunk control test.17 Patient’s ADL was measured with the BI16 and dichotomized into patients with severe to very severe disability (9 points) and moderate or mild disability (>9 points).22 The MI18 for upper and lower paretic limb, ranging from 0 to 200 points, was dichotomized into scores <100 points and higher, using the split-median approach. Because patients were included from 6 different hospitals, we used hospital as an independent variable in our model.

    Neuroimaging Variables

    All scans were performed with a standard scanning protocol using 0.5, 1.0, or 1.5-T MRI scans, with number of slices, thickness, and gap identical. Scans were digitally stored and further analyzed using a Philips Easy Vision Workstation. Areas of abnormal hyperintensity typical for recent cerebral infarction were traced on each slice of the T2-weighted or fluid-attenuated inversion recovery images. Surfaces of areas of abnormal hyperintensity were summed and multiplied with slice thickness (6 mm) and interslice gap (1.2 mm) to calculate infarct volumes. Measurement of infarct volume23,24 was performed, de-identified, and blinded for the patients’ clinical status by a research fellow (S.K.S.) in cooperation with an experienced neuroradiologist (ThD Witkamp). For logistic reasons, scans from 1 hospital were delivered on hard copies. These copies were digitized and analyzed.24 Because we used both digital analysis and analysis of digitized hard copies, an intermethod analysis was performed for 24 MRI scans that were delivered digitally and on hard copies. This gave an intraclass correlation coefficient of 0.97 (95% CI, 0.94 to 0.99). Because of this high agreement, all scans were used for statistical analysis. Lesions affecting cortical gray matter were described as "cortical lesions." If cortex was not affected, lesions were described as "purely subcortical." In addition, lesions were divided into left and right hemispheric lesions.

    Because lesion volume was not normally distributed and no optimal cutoff point could be assessed with receiver operating characteristic (ROC) analysis, it was defined into small and large infarctions with a cutoff point of 22 mL as used previously.8 The presence of white matter lesions25 was used as candidate determinant in the model because this is known to influence stroke outcome. In addition, number of days to scanning after stroke was used as an independent determinant for predicting functional outcome, knowing that time of scanning is associated with lesion volume visible on MRI.26 The number of days to scanning was normally distributed in our population so a continuous variable was used.

    Data Analysis

    Clinical variables shown to be significantly associated with outcome of disability2 were used as independent candidate variables for model 1, whereas neuroimaging variables were added to these clinical variables in a second model (model 2).

    Based on bivariate logistic regression analysis, significant determinants were selected for the development of a multivariate logistic model necessary for the prediction of ADL independency 1 year after stroke. Odds ratios and 95% CIs were calculated. Candidate determinants for model development were allowed if odds ratio reached a liberal level of significance (P0.15). To prevent overfitting of the derived model in relation to the number of positive events per variable27 and improve stability of the regression model, determinants were removed if collinearity diagnostics showed correlation coefficients between candidate determinants of 0.7.28 Because of the large number of variables compared with number of patients involved, the maximum likelihood estimation of parameters in the multivariate clinical model (model 1) was based on a conditional "stepwise forward" approach. Probability for entry of a variable was set at 0.05, for removal at 0.10. Each hypothesis was 2-tailed tested with a significance level of 0.05.

    To evaluate the added value of neuroimaging variables, neuroimaging variables significant after bivariate analysis were added at once ("enter" approach) to the clinical variables of model 1 in a second logistic regression analysis (model 2).

    Subsequently, surplus value of model 2 compared with model 1 was investigated by comparing the area under the curve (AUC) in an ROC curve for both models. An area of 1 implies perfect sensitivity and specificity, whereas an area of 0.5 implies that the predictions of the model are no better than would be obtained by chance. The trapezoid rule was used to calculate the AUC for each model.29,30 The ROC curves of both models were graphically displayed and tested if the AUC of model 2 was significantly different from that of model 1. Models were significantly different if z 1.96. The paired z score between model 2 and model 1 was calculated by the equation: z=AUC1–AUC2/(SE21+SE22–2rSE1SE2), with r as the Pearson product moment correlation coefficient between model 1 and model 2.31 All analyses were performed using SPSS (version 12.0.1 for Windows).

    Results

    Of 115 patients scanned, 21 were excluded because of MRI findings: other lesions (4 patients); infratentorial lesions (10 patients); multiple infarctions (left and right hemisphere; 5 patients); border-zone infarction (1 patient); and in 1 patient, the MRI scan could not be evaluated because of movement artifacts. At 1 year after stroke, 19 of these 94 patients were excluded from further analysis: 9 patients died; 4 patients had recurrent stroke; 2 patients developed comorbidity seriously affecting functional outcome; and 4 patients refused further examination. The 13 patients who died or had another stroke incident were analyzed and compared with the remaining 75 patients with MCA strokes. Both groups did not differ significantly in terms of age (P=0.19); gender (P=0.63); stroke severity (NIHSS; P=0.71); lesion location (cortex/subcortex, P=0.89; hemisphere, P=0.83), and mean lesion volume (P=0.87).

    Measurements were available for 75 stroke patients. Table 1 presents patient characteristics and neuroimaging variables. Eighty-four percent of the patients were admitted to the hospital at the day of stroke onset. Median GCS at admission was 15 (interquartile range , 25% to 75%; 15 to 15), and median stroke severity at admittance was 12 (NIHSS; IQR, 7 to 15), indicating that most patients had moderate neurological deficits. Median BI at 6 days after stroke was 8 of 20 points (IQR, 4 to 17). At 1 year after stroke, 33 patients (56%) were independent in ADL (BI, 19 or 20). Discharge status of patients after hospital stay was: 35% home; 37% rehabilitation center; 4% geriatric home; 23% nursing home; and 1% elsewhere.

    Bivariate and Multivariate Analyses

    Table 2 shows odds ratios and 95% CIs from bivariate logistic regression analyses for independent variables. Of 16 variables, 12 were significantly related to the probability of independency at 1 year after stroke. MI and urinary continence were removed because both determinants showed collinearity with the BI. The remaining 10 variables were used as candidate determinants for developing multivariate logistic models (models 1 and 2).

    Table 3 shows regression coefficients of included clinical determinants and the derived logistic regression model of probability for regaining ADL independency 1 year after stroke in model 1 (sensitivity 77%; specificity 83%; false-positives 21%; false-negatives 19%). Model 2 shows the logistic regression model with added neuroimaging variables (sensitivity 83%; specificity 82%; false-positives 24%; false-negatives 12%). The Figure shows the AUC in ROC analysis for models 1 and 2.

    Receiver operating curves of model 1 (clinical) and model 2 (clinical imaging).

    The AUC was 0.84 for model 1 (SE 0.05; P=0.00; 95% CI, 0.75 to 0.94) and 0.87 for model 2 (SE 0.04; P=0.00; 95% CI, 0.79 to 0.95). Comparison of the 2 derived ROC curves shows that the AUC was not significantly different between models 1 and 2 (R=0.95; z=1.85).

    Discussion

    The aim of the present study was to evaluate whether neuroimaging variables combined with clinical variables (model 2) have added value in predicting functional outcome compared with a model based on clinical determinants alone (model 1). Model 1 contained only 2 clinical variables (age and BI) and was already highly predictive of ADL independency measured at 1 year after stroke. Adding neuroimaging variables (model 2) did not significantly increase the accuracy of predicting long-term outcome of ADL.

    The clinical predicting variables in model 1 were recognized before as important factors that independently predict outcome of ADL.2,3,32 This finding further underpins the prognostic value of age and ADL in the acute phase of stroke (BI) as factors that are highly associated with outcome of ADL.

    Our results agree with findings from the study of Johnston et al.11 They showed that the NIHSS at 1 week after stroke was highly predictive of excellent or devastating outcomes using the NIHSS, BI, and GCS at 3 months after ischemic stroke, and that a combination of clinical information and infarct volume assessed on computed tomography scans 1 week after stroke did not improve predictive value. Our findings also agree with those of Wardlaw et al.12 In their logistic regression analysis, lesion volume was not found to be an independent predictor for 6-month outcome above age and stroke severity, although lesion volume was found to be bivariately associated with functional outcome (P=0.009).

    However, our results disagree with those from Baird et al.4 They concluded that imaging predictors assessed with diffusion-weighted magnetic resonance (time in hours of scanning and lesion volume) in combination with neurological predictors (NIHSS) were more accurate in predicting 3-month BI than either clinical or imaging factors alone.

    Found differences between the above mentioned studies may be attributable to differences in: (1) moment of measuring functional outcome (3, 6, or 12 months after stroke); (2) clinical variables used as candidate determinants for regression modeling; (3) outcome assessment instruments, and (4) criteria applied for selecting patients. For example, the present study included patients with MCA stroke, whereas Wardlaw et al12 included patients with stroke in anterior and posterior cerebral circulation using Rankin scores 6 months after stroke, whereas Baird et al4 restricted to patients with anterior circulation infarctions using the BI at 3 months after stroke.

    Our study adds to the literature by providing a long-term follow-up of a homogenous group of stroke survivors with MCA infarcts in which functional outcome was predicted. However, our study had some limitations. First, patients were included within the first week after admission in a stroke unit leading to a selection of patients for our study population. Second, the number of determinants included in both derived regression models was restricted by the limited number of events per variable. Third, the T2-weighted MRI slices used for lesion segmentation were relatively thick (6 mm), which could have resulted in an underestimation of lesion volume, particularly when lesions were small. However, a study in patients with multiple sclerosis showed that estimations of computed lesion volumes did not increase substantially if lesion slice thickness of 6 mm was further minimized.33 It appears unlikely that minimization of lesion slice thickness would have led to different results in our study.

    Our study implicates that patient characteristics and clinical instruments obtained in the early phase after stroke are sufficient to predict long-term independency in ADL of the stroke survivor after MCA stroke. Neuroimaging information does not have added predictive value compared with clinical information.

    Acknowledgments

    The Scientific Foundation of Rehabilitation Center De Hoogstraat financially supported this project. S.K.S. was supported by a grant from the Netherlands Organization for Health Research and Development (ZON/MW). We thank ThD Witkamp, MD, neuroradiologist, for his assistance in analyzing the MRI scans. We thank the participating neurologists and radiologists. We also thank Ms DCWM Vos-Vromans, research assistant, for contacting and including patients.

    References

    Stineman MG, Maislin G, Fiedler RC, Granger CV. A prediction model for functional recovery in stroke. Stroke. 1997; 28: 550–556.

    Kwakkel G, Wagenaar RC, Kollen BJ, Lankhorst GJ. Predicting disability in stroke-a critical review of the literature. Age Ageing. 1996; 25: 479–489.

    Meijer R, Ihnenfeldt DS, de Groot IJ, van Limbeek J, Vermeulen M, de Haan RJ. Prognostic factors for ambulation and activities of daily living in the subacute phase after stroke. A systematic review of the literature. Clin Rehabil. 2003; 17: 119–129.

    Baird AE, Dambrosia J, Janket S, Eichbaum Q, Chaves C, Silver B, Barber PA, Parsons M, Darby D, Davis S, Caplan LR, Edelman RE, Warach S. A three-item scale for the early prediction of stroke recovery. Lancet. 2001; 357: 2095–2099.

    Chamorro A, Vila N, Ascaso C, Saiz A, Montalvo J, Alonso P, Tolosa E. Early prediction of stroke severity. Role of the erythrocyte sedimentation rate. Stroke. 1995; 26: 573–576.

    Johnston KC, Connors AF Jr, Wagner DP, Knaus WA, Wang X, Haley EC Jr. A predictive risk model for outcomes of ischemic stroke. Stroke. 2000; 31: 448–455.

    Thijs VN, Lansberg MG, Beaulieu C, Marks MP, Moseley ME, Albers GW. Is early ischemic lesion volume on diffusion-weighted imaging an independent predictor of stroke outcome A multivariable analysis. Stroke. 2000; 31: 2597–2602.

    van Everdingen KJ, van der Grond J, Kappelle LJ, Ramos LM, Mali WP. Diffusion-weighted magnetic resonance imaging in acute stroke. Stroke. 1998; 29: 1783–1790.

    Weimar C, Ziegler A, Konig IR, Diener HC. Predicting functional outcome and survival after acute ischemic stroke. J Neurol. 2002; 249: 888–895.

    Ferrucci L, Bandinelli S, Guralnik JM, Lamponi M, Bertini C, Falchini M, Baroni A. Recovery of functional status after stroke. A postrehabilitation follow-up study. Stroke. 1993; 24: 200–205.

    Johnston KC, Wagner DP, Haley EC Jr, Connors AF Jr; RANTTAS Investigators. Randomized Trial of Tirilazad Mesylate in Acute Stroke.Combined clinical and imaging information as an early stroke outcome measure. Stroke. 2002; 33: 466–472.

    Wardlaw JM, Keir SL, Bastin ME, Armitage PA, Rana AK. Is diffusion imaging appearance an independent predictor of outcome after ischemic stroke Neurology. 2002; 59: 1381–1387.

    Brott T, Adams HP Jr, Olinger CP, Marler JR, Barsan WG, Biller J, Spilker J, Holleran R, Eberle R, Hertzberg V, Rorick M, Moomaw CJ, Walker M. Measurements of acute cerebral infarction: a clinical examination scale. Stroke. 1989; 20: 864–870.

    Rocca B, Martin C, Viviand X, Bidet PF, Saint-Gilles HL, Chevalier A. Comparison of four severity scores in patients with head trauma. J Trauma. 1989; 29: 299–305.

    Kasner SE, Chalela JA, Luciano JM, Cucchiara BL, Raps EC, McGarvey ML, Conroy MB, Localio AR. Reliability and validity of estimating the NIH stroke scale score from medical records. Stroke. 1999; 30: 1534–1537.

    Mahoney FI, Barthel DW. Functional evaluation: the Barthel Index. Md Med J. 1965; 14: 61–65.

    Duarte E, Marco E, Muniesa JM, Belmonte R, Diaz P, Tejero M, Escalada F. Trunk control test as a functional predictor in stroke patients. J Rehabil Med. 2002; 34: 267–272.

    Collen FM, Wade DT, Bradshaw CM. Mobility after stroke: reliability of measures of impairment and disability. Int Disabil Stud. 1990; 12: 6–9.

    Bamford JM, Warlow CP. Evolution and testing of the lacunar hypothesis. Stroke. 1988; 19: 1074–1082.

    Fisher CM. Lacunar strokes and infarcts: a review. Neurology. 1982; 32: 871–876.

    Verhage F. Intelligence and Age: Study With Dutch People From age 12 to 77. Assen: van Gorcum; 1964. The Netherlands.

    Wade DT. Measurement in Neurological Rehabilitation. New York, NY: Oxford University Press Inc; 1992.

    Saunders DE, Clifton AG, Brown MM. Measurement of infarct size using MRI predicts prognosis in middle cerebral artery infarction. Stroke. 1995; 26: 2272–2276.

    van der Worp HB, Claus SP, Bar PR, Ramos LM, Algra A, van Gijn J, Kappelle LJ. Reproducibility of measurements of cerebral infarct volume on CT scans. Stroke. 2001; 32: 424–430.

    Leys D, Englund E, Del Ser T, Inzitari D, Fazekas F, Bornstein N, Erkinjuntti T, Bowler JV, Pantoni L, Parnetti L, De Reuck J, Ferro J, Bogousslavsky J. White matter changes in stroke patients. Relationship with stroke subtype and outcome. Eur Neurol. 1999; 42: 67–75.

    Beaulieu C, de Crespigny A, Tong DC, Moseley ME, Albers GW, Marks MP. Longitudinal magnetic resonance imaging study of perfusion and diffusion in stroke: evolution of lesion volume and correlation with clinical outcome. Ann Neurol. 1999; 46: 568–578.

    Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol. 1996; 49: 1373–1379.

    Kwakkel G, Kollen BJ, van der Grond J, Prevo AJ. Probability of regaining dexterity in the flaccid upper limb: impact of severity of paresis and time since onset in acute stroke. Stroke. 2003; 34: 2181–2186.

    Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982; 143: 29–36.

    Metz CE. Basic principles of ROC analysis. Semin Nucl Med. 1978; 8: 283–298.

    Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology. 1983; 148: 839–843.

    Counsell C, Dennis M, McDowall M. Predicting functional outcome in acute stroke: comparison of a simple six variable model with other predictive systems and informal clinical prediction. J Neurol Neurosurg Psychiatry. 2004; 75: 401–405.

    Filippi M, Horsfield MA, Campi A, Mammi S, Pereira C, Comi G. Resolution-dependent estimates of lesion volumes in magnetic resonance imaging studies of the brain in multiple sclerosis. Ann Neurol. 1995; 38: 749–754.

作者: Sven K. Schiemanck, MD; Gert Kwakkel, PhD; Marcel 2007-5-14
医学百科App—中西医基础知识学习工具
  • 相关内容
  • 近期更新
  • 热文榜
  • 医学百科App—健康测试工具