Early Prediction of Stroke Severity
Role of the Erythrocyte Sedimentation Rate
Background and Purpose Early predictors of functional outcome after stroke are necessary for better planning of treatment and care.
Methods We evaluated prospectively early clinical predictors of short-term functional outcome in a group of patients with ischemic cerebral infarction and explored whether the intensity of the acute-phase response provided further information concerning the short-term functional outcome. We evaluated a group of 208 ischemic stroke patients using the Mathew scale at entry. All patients had neuroimaging studies and routine blood tests, including erythrocyte sedimentation rate (ESR), within 72 hours from clinical onset. At discharge, functional outcome was graded according to a Stroke Outcome Scale.
Results Larger infarcts, more embolic infarcts, and fewer lacunar infarcts were observed in the poor-outcome group. Vascular risk factors, radiological findings not related to the index stroke, time to admission, and treatment were similar in the two outcome groups. Variables with statistically significant differences between outcome groups included the following: age >65 years, female sex, admission Mathew score <75, worsening at clinical presentation, infarct volume >6 cm3, complicating infections, fasting glucose >110 mg, nonfasting glucose >130 mg, and elevated ESR. With stepwise logistic regression analysis, Mathew score on admission, infarct volume, mode of clinical presentation, and ESR remained in the predictive model of stroke outcome, with a sensitivity and specificity of 89.91% and 85.71%, respectively. After removing the computed tomographic information from the model the same variables remained, with a sensitivity and specificity of 83.05% and 94.29%, respectively.
Conclusions Infarct size and clinical severity on admission are the stronger predictors of short-term functional outcome. Mode of clinical presentation, clinical evolution during the first day of stroke, and ESR are also independent predictors of short-term stroke outcome. These findings might be indicative of an inadequate collateral profile and/or a more pronounced prothrombotic state.
Although an efficacious treatment against ischemic stroke is yet to be demonstrated, it is increasingly accepted that therapeutic measures should be administered as soon as possible after stroke onset to increase their chance of being effective.1 2 This need for prompt therapeutic decisions emphasizes the importance of early predictors of functional outcome. A better knowledge of clinical and laboratory markers of functional outcome would result in a more individualized and possibly improved approach to management of the patient with stroke.
It is well accepted that infarct size as detected on neuroimaging studies constitutes a strong predictor of clinical outcome.3 However, computed tomographic (CT) scan or brain magnetic resonance imaging (MRI) information concerning the extent of a cerebral infarction is usually available too late after clinical onset to be of help in the decision-making phase. The predictive value of neuroimaging is then limited to the long-term phase of stroke. As a result, prediction of stroke outcome basically relies on clinical findings. Among them, the severity of the clinical deficit on admission is considered to be the major determinant of functional outcome.4 5
To our knowledge, no previous studies have prospectively assessed for prognostic purposes the acute-phase response associated with brain ischemia using well-defined clinical and radiological criteria. In patients with myocardial infarction an elevated acute-phase response correlates well with the extent of tissue damage.6 In patients with cerebral infarction this acute-phase response is very poorly delineated.7
After our preliminary data showed that erythrocyte sedimentation rate (ESR) might be of help in the early prediction of stroke outcome,8 the main purpose of the present study was to quantify this parameter in a larger cohort of patients with acute stroke and to evaluate its independent contribution in the prediction of functional outcome. A better knowledge of the acute-phase response associated with ischemic stroke would contribute to a better understanding of the rheological abnormalities described in these patients.9
Subjects and Methods
From October 1992 to October 1993, 273 consecutive patients presenting with ischemic stroke or transient ischemic attack were admitted to the Stroke Unit of the Neurology Service, including 208 patients with ischemic stroke admitted within 72 hours from clinical onset. Fifty-three patients were excluded from analysis for any of the following reasons: nonischemic stroke, time from stroke onset to hospital admission >72 hours, or incomplete workup. Twelve additional patients were also excluded from analysis because of potential interference with the acute-phase response. The list of these conditions included the following: recent (<1 month) history of myocardial infarction, sepsis, bacterial infection, malignancy, pregnancy, severe renal or hepatic failure, collagen disease, thrombocytopenia, thrombocytosis, or deep vein thrombosis. We scored the neurological condition of the patients on the day of admission using the scale of Mathew et al,10 which was later dichotomized as <75 versus ≥75 of a maximum of 100.
All patients had at least a brain CT scan, and 150 patients (72%) also had a brain MRI. Given that the first CT scan was usually performed too soon after stroke onset, imaging studies were repeated as necessary for the better provision of infarct size and vascular topography. Vascular territories were mapped with the CT atlas of Damasio,11 and infarct volume was measured according to the atlas of Gado, Hanaway, and Frank.12 Infarcts were further classified into small and large (≤6 or >6 cm3 in volume, respectively), and the coexistence of old lesions, silent infarcts, or white matter disease was also taken into account. The most likely stroke mechanism classified the infarcts into cardioembolic, atherosclerotic, lacunar, and infarcts of undetermined cause, according to the criteria used by the Stroke Data Bank.13 The mode of presentation of symptoms and the clinical evolution experienced by patients during the first 24 hours were recorded by either history or direct clinical observation. This information was dichotomized as “improvement” or “no improvement,” the latter referring to symptoms that increased, fluctuated, or were maximal from onset and did not undergo significant changes during the first 24 hours. Invasive and noninvasive tests aimed at defining the stroke subtype were done as appropriate, and therapeutic decisions were made by the responsible physicians. On admission, we performed routine blood tests in all patients. The appearance of complicating infections was evaluated with daily clinical examinations, frequent leukocyte counts, urine tests, chest x-rays, or other tests (eg, echography) when appropriate. Laboratory variables of interest included nonfasting glucose (<130 versus ≥130 mg), fasting glucose (<110 versus ≥110 mg), hematocrit, and ESR (in millimeters per hour) measured by the Westergren method within 72 hours from clinical onset.14 According to the normal ESR values established in our laboratory, we defined a “low ESR group” in men with ESR ≤13 mm/h or women with ESR ≤20 mm/h and a “high ESR group” in men with ESR >13 mm/h or women with ESR >20 mm/h. Other variables of interest included admission systolic and diastolic blood pressure, complicating infections, and type of treatment (antiaggregants versus anticoagulants).
Functional outcome was measured on the day of hospital discharge (mean, 10±3 days) following a Stroke Outcome Scale (SOS) graded from 1 to 6. The category “good functional outcome” included scores between grade 1 (patients with neurological symptoms but no signs and no impairment in activity) and grade 3 (symptoms and signs, with no impairment in activity); “poor functional outcome” included scores between grade 4 (impairment in one limb or impairment of speech) and grade 6 (impairment in two limbs and of speech, or death). In this scale, arm impairment refers to inability to feed, dress, or use the toilet without the assistance of a mechanical device or other person; leg impairment refers to inability to walk without help from another person or mechanical aid; and speech impairment refers to inability to communicate in unfamiliar situations or situations in which the patient would have to cope alone.
For statistical analysis, continuous data were expressed as mean±SD. We performed univariate analyses using the χ2 test with continuity correction and the t test for continuous variables. This enabled us to establish categories for the continuous variables as follows: age (<65 versus ≥65 years), history of hypertension (treated or >160 mm Hg systolic or >90 mm Hg diastolic), diabetes (treated or fasting glucose >110 mg), and hypercholesterolemia (treated or >240 mg). Other epidemiological variables considered of interest included sex, history of smoking, history of myocardial infarction, angina, atrial fibrillation, congestive heart failure, and previous transient ischemic attack or stroke. Stepwise logistic regression was also used to assess the independent contribution of variables statistically significant on univariate analysis in the prediction of functional outcome. Values of P<.05 were considered statistically significant.
One hundred thirty men and 78 women (mean age, 67±12 years; range, 20 to 94 years) were admitted within 23±18 hours from their stroke onset (16% within 6 hours, 54% within 24 hours, and 82% within 48 hours). Mathew score at entry was 78.4±19.5, Mathew score at hospital discharge was 80.1±28.4, and mean infarct volume was 33±54 cm3. Mode of clinical presentation and initial evolution showed that 35% of the patients improved during the first 24 hours, 25% deteriorated, 36% had maximal deficits from onset, 4% fluctuated, and in the remaining 1% this clinical information was missing. Antiaggregants were used in 48% of the patients, anticoagulants in 51%, and no antithrombotic agents in 1%. Complicating infections appeared in 16% of the patients. Patients without infections had an admission Mathew score of 82.4±16 compared with 56.8±21.1 in patients with complicating infections (P<.0001).
A good functional outcome (SOS 1 to 3) was observed in 76 patients (37%), whereas 132 patients (63%) had a poor functional outcome (SOS 4 to 6). In the latter group, SOS subgroups showed that 63 patients (30%) had grade 4, 34 patients (16%) had grade 5, and 35 patients (17%), including 16 deaths during hospitalization, had grade 6. Vascular risk factors, as well as the radiological findings not directly related to the index stroke, did not differ between the outcome groups (Table 1⇓). Time to admission to the hospital (24±18 versus 22±18 hours), type of treatment, and time to its initiation (29±39 versus 35±50 hours) did not explain the differences between the outcome groups. On the contrary, in regard to stroke mechanism and infarct size, larger infarcts, more cardioembolic strokes, and fewer lacunar strokes were observed in the poor-outcome group (Table 1⇓).
At univariate analysis, variables that discriminated between outcome groups included age >65 years, female sex, Mathew score at entry <75, worsening at clinical presentation, infarct volume >6 cm3, complicating infections, fasting glucose >110 mg, nonfasting glucose >130 mg, and elevated ESR (Table 2⇓).
The independent contribution of variables found statistically significant in the prediction of functional outcome was evaluated by means of logistic regression. Clinical severity on admission, type of clinical presentation, infarct volume, and elevated ESR remained in the model (Table 3⇓), with a sensitivity and specificity of 89.91% and 85.71%, respectively. To evaluate how much prediction was improved when ESR was added to the other predictors, a stepwise logistic regression model was used. The sensitivity of the model increased from 85.32% to 89.91%, the specificity ranged from 90.4% to 85.71%, and the total predictive value increased from 87.21% to 88.37%. Thus, by adding ESR to the model more patients with poor outcome were correctly classified.
Given that the information concerning infarct size is usually obtained after the therapeutic decisions must be made, a second analysis was done after infarct size had been removed from the model (Table 3⇑). The same variables remained in the model, which had a sensitivity and specificity of 83.05% and 94.29%, respectively.
Our results confirm the hypothesis that in addition to infarct size15 and clinical severity at entry,16 ESR is an independent predictor of short-term stroke outcome. In agreement with previous studies,17 we also found that the CT scan is of limited use for early prognostic purposes. A CT scan is mandatory to rule out alternative diagnosis, but according to our findings functional outcome is more specifically forecasted without the aid of the CT scan. Although the inclusion of the CT scan in the predictive model of functional outcome served to increase the sensitivity to appropriately diagnose those patients who would have a poor functional outcome, it also decreased the specificity of the model given that early normal CT scans do not preclude the appearance of large infarcts in subsequent studies.
Age has also been reported as a significant predictor of stroke outcome.18 However, we concur with previous studies that have suggested that this association may be better explained by the coexistence of other factors, such as severity of neurological impairment on admission.17 In the present series another explanation for the interaction between age and prognosis would be that older patients showed significantly larger infarcts (data not shown).
Previous studies have established that men have a worse prognosis for recovery.5 Although at univariate analysis we found that women fared worse, at multivariate analysis sex did not remain in the model. Older age, lower Mathew score on admission, and larger lesion size in the female group explained the sex differences that we found in functional outcome.
Recently the acute-phase response was found to be unrelated to the extent of cerebral infarction but associated with bacterial infections during the month preceding cerebral infarction.19 Our results dispute these findings, given that patients with recent or current infections were excluded and that blood samples were drawn before most poststroke infections usually tend to occur. Although we found higher ESR values in patients who later experienced complicating infections, suggesting that in some instances undiagnosed infections might explain some of these early elevations, neither infections nor the interaction between infections and ESR remained as independent predictors of stroke outcome when entered in a multivariate analysis. The association that we found between complicating infections and a poor outcome on univariate analysis was better explained by the fact that patients in whom infections developed had a significantly worse Mathew score on the day of admission. Patients with worse neurological deficits are more likely to have supervening complications, including infection.
A remarkable finding in the present series is the association between admission ESR, mode of clinical presentation, initial evolution, and stroke outcome. Thus, we found a significant association between higher ESR on admission and clinical deterioration during the first 24 hours of stroke onset. Since clinical worsening at such an early phase of stroke is most frequently due to thrombosis, it might be speculated that our measurement of ESR is an indirect marker of thrombus formation. This hypothesis is further supported by the finding that higher ESR levels were found in older patients with larger lesions and more severe deficits at entry; all these factors suggest a less developed capacity for collateral circulation, which would facilitate blood stasis and secondary thrombus progression.
Recently a close relationship among ESR, fibrinogen, and cerebral blood flow was established.20 21 22 Accordingly, an elevated ESR in the acute phase of stroke might indicate a greater increase in the fibrinogen concentration and a more pronounced reduction in cerebral blood flow, which would result in larger lesions. The inverse relationship between ESR and stroke outcome might also indicate the displacement of the normal concentration of natural anticoagulants. Recently a strong correlation was shown between ESR and protein C4b-BP, a high-molecular-weight protein that possesses a single binding site for vitamin K–dependent protein S, an inhibitor of coagulation.23 Although protein C4b-BP may behave as an acute-phase reactant, it is known that when protein S is bound to C4b-BP, the former has no anticoagulant activity.24 Thus, acute changes in protein C4b-BP could predispose to thrombosis. Nevertheless, further studies are needed to evaluate the potential relationship between ESR, protein C4b-BP, and protein S in patients with ischemic stroke.
Although the design of this study does not allow us to elucidate whether hematologic abnormalities antedated stroke onset or merely represent nonspecific postischemic changes, we may conclude that an early elevation of ESR is a good marker of poor short-term stroke outcome. When the information gathered from this simple test is viewed together with the severity of stroke on admission and the mode of clinical onset, a highly specific and sensitive predictive model of short-term functional outcome can be obtained. Inclusion of ESR measurement in the standard admission orders for patients with acute stroke provides clinicians with a low-cost and useful test at a time when radiological tests usually underestimate the real extent of brain infarctions and decrease the specificity of the models predicting functional outcome after ischemic stroke. With a more precise knowledge of early outcome predictors, the inclusion of patients in therapeutic trials could be stratified according to their odds of having an adverse clinical evolution.
- Received June 6, 1994.
- Revision received July 19, 1994.
- Accepted December 16, 1994.
- Copyright © 1995 by American Heart Association
Biller J, Love BB. Nihilism and stroke therapy. Stroke. 1991;22:1105-1107.
Astrup J, Siesjö K, Symon L. Thresholds in cerebral ischemia: the ischemic penumbra. Stroke. 1981;12:723-725.
Chambers BR, Norris J, Shurvell BR, Hachinski VC. Prognosis of acute stroke. Neurology. 1987;37:221-225.
Allen CMC. Predicting the outcome of acute stroke: a prognostic score. J Neurol Neurosurg Psychiatry. 1984;47:475-480.
de Beer FC, Hind CRK, Fox KM, Allan RM, Maseri A, Pepys MB. Measurement of C-reactive protein concentration in myocardial ischaemia and infarction. Br Heart J. 1982;47:239-243.
Chamorro A, Vila N, Montalvo J, Alday M. Role of erythrocyte sedimentation rate in the prediction of stroke severity. Ann Neurol. 1993;34:291. Abstract.
Fisher M, Meiselman HJ. Hemorheological factors in cerebral ischemia. Stroke. 1991;22:1164-1169.
Polis AT, Foulkes MA, Mohr JP, Hier DB, Price TR, Wolf PA. Representing computerized tomographic scans using the GAMP procedure in SAS·Graph. Proc SAS Users Group Int Conf. 1989;14:1162-1165.
Foulkes MA, Wolf PA, Price TR, Mohr JP, Hier DB. The Stroke Data Bank: design, methods, and baseline characteristics. Stroke. 1988;19:547-554.
Westergren A. Studies on the suspension stability of the blood in pulmonary tuberculosis. Acta Med Scand. 1921;54:247-282.
Berger L, Hakim AM. The association of hyperglycemia with cerebral edema in stroke. Stroke. 1986;17:865-871.
Censori B, Camerlingo M, Casto L, Ferraro B, Gazzaniga GC, Cesana B, Mamoli A. Prognostic factors in first-ever stroke in the carotid artery territory seen within 6 hours after onset. Stroke. 1993;24:532-535.
Kushner M, Reivich M, Fieschi C, Silver F, Chawluk MD, Rosen M, Greenberg J, Burke A, Alavi A. Metabolic and clinical correlates of acute ischemic infarction. Neurology. 1987;37:1103-1110.
Jongbloed L. Prognosis of function after stroke: a critical review. Stroke. 1986;17:765-776.
Syrjanen J, Teppo AM, Valtonen VV, Iivanainen M, Maury CPJ. Acute phase response in cerebral infarction. J Clin Pathol. 1989;42:63-68.
Gudmundsson M, Bjelle A. Plasma, serum and whole-blood viscosity variations with age, sex, and smoking habits. Angiology. 1993;44:384-391.
Ameriso SF, Paganini-Hill A, Meiselman HJ, Fisher M. Correlates of middle cerebral artery blood velocity in the elderly. Stroke. 1990;21:1579-1583.
Boerger LM, Morris PC, Thurnau GR, Esmon CT, Comp PC. Oral contraceptives and gender affect protein S status. Blood. 1987;69:692-694.