Prognostic Value and Determinants of First-Day Mean Arterial Pressure in Spontaneous Supratentorial Intracerebral Hemorrhage
Background and Purpose The onset of spontaneous intracerebral hemorrhage (ICH) is often accompanied by transient blood pressure (BP) elevation. The prognostic value and the determinants of this BP reaction have not entirely been solved, and the present study was focused on these questions.
Methods From 1985 to 1991 in Central Finland (population, 246 000), a total of 425 patients had first-ever ICH verified by CT or necropsy. The hematoma was supratentorial in 337 patients. Of the 306 patients with supratentorial ICH who had CT, 282 had the BP measured at least once within 24 hours of onset, and they formed the study population. The case notes and CT films were reviewed, and mean arterial pressure (MAP) was calculated from the highest BP reading.
Results The fatality rate was high; 43% of the patients were dead within 28 days of onset. Six independent predictors of the 28-day survival were identified by multiple logistic regression; these predictors were consciousness on admission, first-day MAP, subarachnoid spread of the bleed, lateral shift of hemispheral midline structures, admission blood glucose, and vomiting. The MAPs varied between 66.7 and 203.3 mm Hg, and the cutoff points of the MAP quartiles were 118, 132, and 145 mm Hg. Patients in the first three MAP quartiles had relatively fair outcome, with 71%, 65%, and 60%, respectively, alive 28 days after onset. This was in sharp contrast to the fourth quartile, with only 33% surviving the first 28 days (log-rank, P<.0001 to P=.0010). Patients unconscious/comatose on admission had significantly higher MAPs than did those who were alert or somnolent/disoriented (ANOVA, P=.0079). However, at all levels of consciousness, the 28-day fatality rate increased from the first to the fourth MAP quartile: 69% in the alert, 186% in the somnolent/disoriented, and 45% in the unconscious/comatose patients. Stepwise multiple regression analysis gave four independent predictors of the first-day MAP: hypertension, age (in an inverse fashion), admission blood glucose, and hematoma volume.
Conclusions The most important predictor of the 28-day survival was the level of consciousness on admission, followed by first-day MAP. Hypertension was the most important predictor of the first-day MAP, followed by age, which had an inverse effect on the MAP level. At all levels of consciousness, high first-day MAP (especially if >145 mm Hg) worsened the 28-day survival rate.
Transient BP increases are common after the onset of spontaneous ICH.1 2 3 4 However, the prognostic value and the pathophysiological mechanisms behind this BP reaction are not entirely known. High levels of BP measured on admission and during the first day of admission have usually,5 6 7 8 but not always,9 10 predicted a poor prognosis. In one report, high initial MAP did not predict the 30-day survival, but a combination with impaired consciousness increased the fatality rate significantly.11 High BP at stroke onset has been associated with elevated urinary catecholamine output,12 with large brain lesions, or with lesions of the cerebellum,13 and finally, high BP has been explained as a response to mental stress caused by hospital admission.3
The present study was focused on the prognostic value and the determinants of high MAP measured during the first 24 hours of ICH onset. The patients included in the study had first-ever supratentorial ICH, and the diagnosis was confirmed in all cases by CT.
Subjects and Methods
The patients were traced from the discharge lists of the Central Hospital of Central Finland (population of catchment area, 246 000) and from the medical and medicolegal necropsies of the same population. All medical and autopsy reports were perused, and all relevant data were collected on standard forms. During the time interval from September 1985 through December 1991, a total of 425 patients had first-ever ICH. In 337 patients, the hematoma was supratentorial, 306 had the diagnosis confirmed by CT, and 282 of them had the BP measured within 24 hours of onset. These 282 patients, 131 men (mean age, 65.4 years) and 151 women (mean age, 69.3 years), were included in the present study. The sex and age distribution of the 55 excluded patients, 29 men (mean age, 63.5 years) and 26 women (mean age, 71.6 years), corresponded with those included. The most important reason for exclusion was early death, most often on the day of onset, which resulted either in nonadmission (31%) or missing CT examination (26%). Late admission (>24 hours of onset) was the cause for exclusion in 40%.
The first BP was most often measured by a trained nurse using a calibrated sphygmomanometer immediately after the patient’s arrival in the emergency room. In almost all patients, several BP readings during the first 24 hours after onset were available; of these, the highest one was included in the present analysis. In 10 patients admitted to the Central Hospital after 24 hours of onset, the BP readings at the referring health center were taken. MAP was calculated by adding 1/3 of the pulse pressure (systolic minus diastolic) to the diastolic pressure. The patient’s prior medication was continued, and additional antihypertensive medication was considered only if the BP exceeded 220/125 to 220/130 mm Hg. In 60 patients, this was the case, and 24 patients received clonidine, calcium blockers, sodium nitroprusside, or furosemide. The mean±SD BP of the treated patients (229±29/131±19 mm Hg) was nonsignificantly higher than that of the untreated patients (227±26/123±14 mm Hg), as was the 28-day fatality rate, 79% and 75%, respectively. A standard 12-lead electrocardiogram and blood glucose levels (Glucometer, Medisense) were taken in the emergency room.
The median delay from symptom onset to admission to the Central Hospital was 3 hours: 68% were admitted with 6 hours; 80%, within 12 hours; and 97%, within 24 hours. There was a nonsignificant trend: patients with larger hematomas were mostly admitted during the first 6 hours compared with the later 6-hour periods (Kruskal-Wallis, P=.228). CT examination was performed on 47% of the patients during the first 24 hours and on 82% within 48 hours of stroke onset. Conservative treatment of the hematoma was the rule; only 4 patients had neurosurgical evacuation.
Hypertension (doctor-diagnosed, with the patient being on antihypertensive medication) was the most prevalent (37%) comorbid disease in the medical history, followed by cerebrovascular disease, usually ischemic brain infarction (23%) and congestive heart failure (19%). Ten percent of the patients had diabetes treated with diet or antidiabetic medication, and 12% were for various reasons on anticoagulant treatment.
In classifying the hematomas by location, we used the CT atlas of Kretschmann and Weinrich.14 One-hundred sixty-seven (59%) of the patients had deep hematomas, and 115 (41%) had lobar hematomas. The location of the deep hematomas was as follows: thalamus (63), putamen (34), internal or external capsule (15), and caudate nucleus (5). In an additional 50 large deep hematomas the exact origin of the bleeding could not be determined. The lobar hematomas were located as follows: temporal (36), frontal (33), parietal (30), and occipital (16).
Hematoma volume (in milliliters) was calculated using a computer program. In each CT slice, the area of the hematoma was measured planimetrically and multiplied by the slice thickness. Summing these subvolumes gave the total hematoma volume. The lateral shift (in millimeters) of the hemispheral midline structures was measured at the level of the third ventricle. Hematoma volumes varied from 1 to 208 mL (median, 35.5 mL; interquartile distance, 12 to 71 mL), and the lateral shift of the midline structures varied from 0 to 24 mm (median, 4 mm; interquartile distance, 1 to 9 mm). Intraventricular and subarachnoid spread of the bleed as well as hydrocephalic enlargement of the lateral ventricles were qualitatively noted (no/yes). Intraventricular extension was present in 160 (57%), hydrocephalic enlargement of the ventricles in 78 (28%), and subarachnoid spread in 54 (19%) of the patients. In 18 of the 54 patients with subarachnoid blood, the hematoma was located in the frontal or temporal lobe, and in these locations a ruptured arterial aneurysm can be the cause. This possibility was excluded in 8 patients by autopsy or carotid angiogram, and 7 of the remaining 10 patients also had an intraventricular extension, which can be the route to the subarachnoid space. As a conclusion, the number of patients with ruptured arterial aneurysm as a possible cause of the intracerebral hematoma remained minimal.
Univariate analysis was performed using the χ2 test in nominal and ordinal data and the t test or Mann-Whitney U test in continuous data. Pearson’s product moment and Spearman’s rank correlation were used in estimating the correlation between two data sets, and parametric and nonparametric ANOVA was used in comparing the means and medians of three or more data sets. The significance of survival differences was tested by the log-rank test. Multiple logistic regression (StatPac Gold, version 3.2, Walonick Associates Inc) was used to find factors that simultaneously predicted the 28-day survival, and stepwise multiple regression was used to find factors that explained the first-day MAP level. Dummy variables were used to represent nominal and ordinal scale variables.
The early fatality rate was high: 91 patients (32%) were dead within 7 days, and 120 patients (43%) were dead within 28 days after onset.
On admission, 95 (34%) of the patients were alert, 110 (39%) were somnolent/disoriented, and 77 (27%) were unconscious/comatose. In 31 of the unconscious/comatose patients, the presence or absence of hemiparesis could not be evaluated, but hemiparesis of various degrees was present in 79% of the remaining 251 patients. The admission MAP (mean±SD) of the 104 hypertensive patients (138.4±21.4 mm Hg) was significantly higher than that of the 178 normotensive subjects (129.9±20.9 mm Hg).
The first-day MAPs varied from 66.7 to 203.3 mm Hg, and the cutoff points of MAP quartiles were 118, 132, and 145 mm Hg. Patients who belonged to the first three MAP quartiles had a relatively fair outcome, with 71%, 65%, and 60% alive 28 days after onset (Figure⇓). This was in contrast to the fourth quartile, with only 33% of the patients surviving the first 28 days of onset. The differences between the first three and the fourth quartile were highly significant (log-rank, from P<.0001 to P=.001). The level of consciousness on admission had an important bearing on the MAP. The MAP of unconscious/comatose patients (139.4±23.8 mm Hg) was significantly higher than that of patients who were alert (130.6±19.2 mm Hg) or somnolent/disoriented (130.3 ±21.0 mm Hg) (ANOVA, P=.0079). However, regardless the level of consciousness, the proportion of patients dying during the first 28 days of onset increased from the first to the fourth MAP quartile: in alert patients, from 0.13 to 0.22 (+69%); in somnolent/disoriented patients, from 0.22 to 0.63 (+186%); and in unconscious/comatose patients, from 0.67 to 0.97 (+45%).
The 14 factors that differed significantly in univariate analysis between the deceased and the survivors (Table 1⇓) were included in multiple logistic regression to find covariates that significantly increased the predictability of 28-day survival. Hypertension in the medical history was not included because the prevalences among the deceased (40%) and the survivors (35%) did not differ statistically significantly (P=.35). After removal of variables that did not contribute substantially to the prediction and selecting the most easily available of two highly correlated variables, eg, shift of midline structures instead of hematoma volume (rs=.751) and intraventricular extension instead of enlargement of the ventricles (rs=.699), six independent variables were left in the equation (Table 2⇓): the level of consciousness, first-day MAP, subarachnoid spread, shift of midline structures, admission blood glucose, and vomiting on admission. The significance level of the equation for the χ2 statistic was <.0001. When the probability of .5 of surviving was taken as the cutoff point, the sensitivity and specificity were 78% and 90%, and the positive and negative predictive values 87% and 85%, respectively.
To find the predictors of the first-day MAP, all 12 variables that differed in univariate analysis significantly between patients who had MAPs of ≤145 mm Hg and those with MAPs of >145 mm Hg (Table 3⇓) were included in stepwise multiple regression. Four of these variables met the significance level and were included in the equation (Table 4⇓). We will stress that the level of consciousness did not appear in the equation even when the F value to enter was set as low as 2.5. The inverse effect of age on the MAP level depended for the most part on men, who showed a significant negative correlation between age and MAP (r=−.206, P=.018); in women, this correlation was nonsignificant (r=−.111, P=.175). There was also a trend of a negative correlation between age and hematoma volume in men (rs=−.122, P=.164) but not in women (rs=−.03, P=.693). On the other hand, both hematoma volume and lateral shift of the midline structures were significantly correlated with the first-day MAP: rs=.153 (P=.001) and rs=.147 (P=.014), respectively.
The crude annual incidence of 27.4/100 000 (95% CI, 24.9 to 30.2) calculated on the basis of all 425 ICH patients traced for the present study from the population of whole Central Finland corresponds well with the incidence of 31/100 000 (95% CI, 26.4 to 36.3) obtained in an epidemiological study including the Jyväskylä Region, Central Finland, 1985 to 1989.15 Thus, the patient material of the present study may well be considered as representative.
Because of the large variation of the delays (in hours) from symptom onset to admission, the highest BP reading during the first 24 hours of onset was chosen instead of admission BP. Increase of the hematoma volume during the first day of onset has been observed in some ICH patients,16 17 and because our approach gives more time for a possible volume increase, it may give BP readings that predict the outcome better than one reading on admission.
Logistic regression analysis gave an equation with six independent variables (Table 2⇑) that predicted the 28-day case fatality with a 78% sensitivity and 90% specificity. The level of consciousness on admission and first-day MAP, separately, were the two most important variables. Admission BP has been included in several studies in which multivariate analysis has been used in the search of independent predictors of outcome after supratentorial ICH.5 6 9 10 11 18 19 20 The results in these studies have, however, not been in unison. In univariate analysis, high admission BP has been associated with poor survival5 6 ; there has been no difference in the mean BP level between patients with good or poor functional outcome or between patients dying or surviving the first months after onset9 10 19 ; and finally, the 30-day survivors have had even higher (nonsignificantly) admission BP than the deceased.20 High admission BP expressed as pulse pressure remained a significant independent predictor of poor prognosis in one6 but not in two other studies.11 18 Contrary to our results, in the Swedish study11 the admission MAP did not differ between patients dead or alive 30 days after onset or between patients alert or with impaired consciousness. However, a combination of impaired consciousness and high admission MAP predicted a poor 30-day survival. Patient selection (eg, exclusion of seriously ill patients from the Swedish study, with only 27% dying within 30 days [43% died during 28 days in the present study]) and the relatively small number of patients may explain the differing results.
Of the other four variables in the equation predicting the 28-day survival, the magnitude of lateral shift of hemispheral midline structures is very closely related with the hematoma volume. In addition, large hematomas are apt to increase the intracranial pressure and, as a result of the lateral shift of the midline structures, to distort the brain stem, resulting in vomiting (and disturbed consciousness).
High admission blood glucose levels in ICH, especially in the nondiabetic patients, have been associated with high early fatality.21 Blood glucose was included in the external validation study22 of the results based on patients from the Pilot Stroke Data Bank.6 Hyperglycemia (blood glucose of >180 mg/dL) was more common among patients dying within 30 days of onset but did not contribute significant predictive information in a logistic regression model.22 Whether high blood glucose at stroke onset is a stress reaction or only a marker of an undiagnosed or latent diabetes remains to be determined. In the present study, the mean admission blood glucose level among nondiabetic patients was significantly higher in those dying within 28 days than in the survivors and among patients in the fourth MAP quartile than in those with lower MAP levels. Similar, although nonsignificant, differences were observed among the 27 diabetic patients. We think that our results point to a stroke-induced stress reaction with possibly the same pathophysiological mechanism involved in the BP elevation.
Stepwise multiple regression revealed four variables that predicted the MAP level: hypertension, age, admission blood glucose, and hematoma volume. Hypertension has generally been considered to be an important predictor of high BP after stroke onset, but the inverse effect of age has not been previously documented. The inverse effect of aging may be explained by the increasing intracranial reserve space due to decrease of the cerebral and increase of the ventricular volumes.23 Therefore, intracerebral hematomas may not increase the intracranial pressure as easily as in younger persons, resulting in a milder Cushing response. The association of intracerebral hematomas with high BP may, apart from the Cushing response, depend on other mechanisms, because a lesion at any site in the central nervous system can potentially produce acute elevation of the BP.24 Hematomas near the hypothalamus may increase the BP, either by increasing the sympathetic activity or by adrenal cortical or medullary hypersecretion. On the other hand, lateral dislocation of the hemispheral midline structures may distort the brain stem, which may damage both the ascending reticular activating system and the nuclei of tractus solitarius. This can result in disturbed consciousness and elevation of BP caused by sympathetic activity disinhibition. A similar BP reaction caused by vasopressin release may follow damage of the area A1 noradrenergic neurons in the ventral brain stem. Mental stress on hospital admission3 as a cause of high BP cannot be totally discarded, but in unconscious patients, its role must be minimal.
Selected Abbreviations and Acronyms
|MAP||=||mean arterial pressure|
- Received February 4, 1997.
- Revision received May 5, 1997.
- Accepted May 6, 1997.
- Copyright © 1997 by American Heart Association
Britton M, Carlsson A, de Faire U. Blood pressure course in patients with acute stroke and matched controls. Stroke. 1986;17:861-864.
Carlberg B, Asplund K, Hägg E. Factors influencing admission blood pressure levels in patients with acute stroke. Stroke. 1991;22:527-530.
Harper G, Castleden CM, Potter JF. Factors affecting changes in blood pressure after acute stroke. Stroke. 1994;25:1726-1729.
Portenoy RK, Lipton RB, Berger AR, Lesser ML, Lantos G. Intracerebral haemorrhage: a model for the prediction of outcome. J Neurol Neurosurg Psychiatry. 1987;50:976-979.
Dandapani BK, Suzuki S, Kelley RE, Reyes-Iglesias Y, Duncan RC. Relation between blood pressure and outcome in intracerebral hemorrhage. Stroke. 1995;26:21-24.
Passero S, Burgalassi L. Blood pressure and short-term outcome in primary intracerebral hemorrhage. Cerebrovasc Dis. 1996;6(suppl 2):110. Abstract.
Broderick JP, Brott TG, Duldner JE, Tomsick T, Huster G. Volume of intracerebral hemorrhage: a powerful and easy-to-use predictor of 30-day mortality. Stroke. 1993;24:987-993.
Carlberg B, Asplund K, Hägg E. The prognostic value of admission blood pressure in patients with acute stroke. Stroke. 1993;24:1372-1375.
Kretschmann HJ, Weinrich W. Neuroanatomy and Cranial Computed Tomography. New York, NY: Thieme; 1986.
Fogelholm R, Nuutila M, Vuorela A-L. Primary intracerebral haemorrhage in the Jyväskylä region, Central Finland, 1985-89: incidence, case fatality rate, and functional outcome. J Neurol Neurosurg Psychiatry. 1992;55:546-552.
Chen ST, Chen SD, Hsu CY, Hogan EL. Progression of hypertensive intracerebral hemorrhage. Neurology. 1989;39:1509-1514.
Franke CL, van Swieten JC, Algra A, van Gijn J. Prognostic factors in patients with intracerebral haematoma. J Neurol Neurosurg Psychiatry. 1992;55:653-657.
Lisk DR, Pasteur W, Rhoades H, Putnam RD, Grotta JC. Early presentation of hemispheric intracerebral hemorrhage: prediction of outcome and guidelines for treatment allocation. Neurology. 1994;44:133-139.
Woo E, Ma JTC, Robinson JD, Yu YL. Hyperglycemia is a stress response in acute stroke. Stroke. 1988;19:1359-1364.