Incidence and prognostic significance of fever following intracerebral hemorrhage

Patient selection.

We reviewed the charts and CT scans of 251 consecutive patients with the diagnosis of spontaneous supratentorial ICH who were admitted within 24 hours after onset of symptoms from 1992 to 1996 to the University of Heidelberg Neurological Hospital. Patients who were seen in our emergency rooms only or with incomplete data (n = 17) were not included in this study. Only patients with spontaneous hemorrhage were included. All patients with trauma, intracranial tumors, or primary subarachnoidal hemorrhage were excluded. The presumed cause of hemorrhage was assessed according to patients’ history and clinical, radiologic, or intraoperative findings. Cerebral amyloid angiopathy was diagnosed in patients older than 60 years with cortical or subcortical lobar hemorrhage, especially in the presence of multiple hemorrhages or a history of previous hemorrhages, after the exclusion of other causes of ICH with angiography and MRI. Patient data were reviewed without patient identification, as is required by the local ethics committee standards.

Initial clinical evaluation.

The following features were extracted from the routine medical and neurologic evaluation on admission in the emergency room: age, sex, Glasgow Coma Scale (GCS14) score, body temperature, mean arterial pressure (MAP), and serum glucose levels. For further analysis, the data were categorized using cut-off points commonly used in clinical practice as diagnostic criteria for therapeutic decisions.

The GCS score was classified into 1) 3 to 7, 2) 8 to 13, and 3) more than 13. For patients who arrived already intubated or sedated, the initial GCS score, assessed by the ambulance team, was used.

Temperature was measured orally or rectally. The temperatures were categorized into 1) hypothermic (<36.5 °C), 2) normal (36.5 to 37.4 °C), 3) subfebrile (37.5 to 38.4 °C), and 4) febrile (≥38.5 °C).

Blood pressure (BP) was measured noninvasively using a conventional BP cuff. MAP was calculated from the diastolic and systolic values (MAP = P_diast + 1/3 × [P_syst − P_diast]). The MAP values were dichotomized into 1) normotensive and hypotensive (<120 mm Hg) and 2) hypertensive (≥120 mm Hg).

Blood glucose levels were determined using routine laboratory methods and, according to diagnostic criteria for diabetes, grouped into 1) low (<5.5 mmol/L), 2) normal (5.5 to 11.0 mmol/L), and 3) increased (>11.0 mmol/L).

According to the Glasgow Outcome Scale (GOS15), outcome at hospital discharge was determined as 1) death, 2) vegetative state, 3) severely disabled, 4) moderate recovery, and 5) good recovery. For statistical analysis, outcome was dichotomized into 1) poor (GOS 1 or 2), and 2) moderate or favorable (GOS 3 to 5). Patients’ ages were categorized into 1) less than 55 years, 2) 55 to 75 years, and 3) older than 75 years.

Neuroradiologic data.

All patients’ diagnoses were confirmed by CT scanning immediately after hospital admission. The hematoma site was classified as lobar, ganglionic (subclassified into thalamic, putaminous, lateral ganglionic, and caudate), or purely ventricular. Ventricular extension was determined by assessing CT scans for the presence of blood in the ventricles. ICH volume was calculated with the formula for ellipsoids (4/3πabc; a, b, and c represent the respective radii in three dimensions). This method has been demonstrated to estimate the hematoma volume reliably.16 ICH volume was classified as 1) small (<25 cm³), 2) moderate (25 to 60 cm³), or 3) massive (>60 cm³). In patients with atypical ICH, angiography or cranial MRI was performed. To detect hematoma enlargement, secondary CT evaluation was performed if clinical deterioration was evident or routinely after 3 days.

Medical treatment and monitoring.

All patients received standardized medical treatment according to an institutional protocol for ICH. In the initial phase of the disease, patients were monitored in the neurologic critical care unit.

Body core temperature was measured on admission and during the further course orally or rectally at 2-hour intervals. The time intervals were shortened in patients with increased temperatures. Subfebrile temperatures of up to 38.0 °C were tolerated if a patient felt no discomfort. Patients with fever of more than 38.0 °C were initially treated with acetaminophen (1000 mg rectally). If this was not successful, external cooling with cold blankets was started. Fever resistant to these measures was treated with parenteral metamizole (250 to 500 mg) or chlorpromazine (10 to 25 mg).

Routine chest radiography and urine analyses were performed on all patients on the day of admission. A search for infectious causes was started in all patients presenting with fever or newly developing fever. Routine measures consisted of a physical examination, repeat chest radiography, urine sediment and culture, change of all catheters, and laboratory tests for systemic infection including complete blood count and C-reactive protein. The presence of urinary tract infection was defined as significant bacteriuria (>10⁵ cells/μL) or leukocyturia (>20 cells per field of view). The diagnosis of pulmonary infection was established according to clinical and radiologic findings.

BP was closely monitored. In normotensive patients, BP was measured initially in 30-minute intervals. This interval was extended in patients who were persistently normotensive over several hours. In hypertensive patients (diastolic BP >95 mm Hg or systolic BP >160 mm Hg), BP was monitored continuously with an intra-arterial radial or femoral catheter. BP was adjusted to systolic values between 140 and 170 mm Hg. If the systolic BP was persistently higher than 170 mm Hg, antihypertensive treatment was started in most patients with uradipil (an alpha-blocking agent) or clonidine. Hypotensive patients with large space-occupying hematomas were treated with fluids and vasopressor drugs.

Blood glucose levels were monitored quantitatively using bedside tests. The time intervals between blood glucose measurements varied according to baseline values or a history of diabetes. At least three measurements per day were taken, even among patients who were always normoglycemic. Hyperglycemia (blood glucose level >11 mmol/L) was treated with insulin.

Corticosteroids were not used. Hyperosmolar agents, barbiturates, and hyperventilation were not routinely used. Coagulation disorders were reversed as soon as possible using coagulation factors, plasma, or platelets, according to the underlying cause. Electrolytes and fluid levels were closely monitored and adjusted to normal values.

Ventricular drainage was considered a part of medical treatment and was carried out in all patients with evidence of enlarged ventricles on CT scanning. Ventriculostomy was also used for intracranial pressure monitoring. Among patients who did not require ventriculostomy, intracranial pressure monitoring devices were not routinely inserted.

The duration of increased temperature, blood glucose levels, and arterial BP were determined from the single measurements during the observation period of 72 hours and categorized for each variable into 1) always within normal limits, 2) abnormal for less than 24 hours, 3) abnormal for 24 to 48 hours, and 4) abnormal for more than 48 hours.

Surgical treatment.

Indications for hematoma evacuation were made on an individual basis. Patients over 80 years of age and with hematoma volume of less than 10 cm³ were not considered for surgery. In patients with large left-hemispheric hematomas or unfavorable clinical status on hospital admission, the indications for surgery were made restrictively. Hematoma evacuation was performed by osteoplastic trepanation by cortical incision and subsequent gentle suction of the clots using an operating microscope.

Statistical analysis.

Statistical analysis was performed on a personal computer using statistical software (Statview 5.0, SAS, Cary, NC). Chi-square and Fisher’s exact tests were used to determine associations between the variables categorized. To examine the significance of possible prognostic variables, two multiple logistic regression models were fitted to the data set with the dichotomized GOS score (GOS 1 or 2 versus GOS 3 to 5) as the response variable. The first model included all patients, using the following categorized clinical and radiologic features obtained on hospital admission as possible explanatory variables: sex, age, GCS score, MAP, blood glucose level, body temperature, location, hematoma volume, presence of ventricular hemorrhage, and presence of a coagulation disorder. The second model was applied to analyze predictive factors within the first 72 hours. In addition to these variables, the presence of infection; hematoma enlargement or secondary hemorrhage; hematoma evacuation; and the duration (categorized into <24 hours, 24 to 48 hours, and >48 hours) of increased temperature, blood glucose level, and arterial hypertension were included in this model as possible explanatory variables. To achieve statistical comparability, we excluded all patients who died within the first 72 hours in the second model. In both models, the possible explanatory variables were tested in reverse order, starting with all independent variables and then stepwise removing the variable that was least explanatory until all variables remaining in the model contributed with p values of less than 0.05. The level of significance (α) chosen was p < 0.05 for all statistical analyses.

Clinical and radiologic findings on hospital admission.

Table 1 shows the clinical characteristics of all 251 patients, as used for further statistical analysis. Table 2 shows the radiologic findings and the origin of the hemorrhage. The mean patient age was 63.5 years (SD 12.5 years). The mean length of hospital stay was 16 days (SD 12 days). A total of 209 patients (83%) were transferred to rehabilitation units, primary care centers, or long-term care facilities for additional care. On discharge, clinical outcome was death (GOS 1) in 55 (22%), vegetative state (GOS 2) in 5 (2%), severe disability (GOS 3) in 104 (41%), moderate disability (GOS 4) in 64 (26%), and no remaining symptoms (GOS 5) in 23 patients (9%). The cause of death was herniation in 45 (82%), cardiac complications in 4 (7%), infection in 5 (9%), and unexplained in 1 patient (2%).

Univariate analysis with the dichotomized outcome as the response variable revealed a significant association with age, hematoma volume, initial GCS score, blood glucose level, body core temperature, ventricular hemorrhage, and origin of the hemorrhage (classified as presence or absence of coagulation disorders with the inclusion of anticoagulant therapy).

A total of 35 patients (14%) presented with hypothermic temperatures (<36.5 °C). Most patients (n = 169; 67%) had normal temperatures (36.5 to 37.4 °C); 45 patients (18%) had subfebrile body temperatures (37.5 to 38.4 °C). The incidence of fever of 38.5 °C or more was low (n = 2; 1%). The relationship between initial body temperature and outcome is presented in figure 1. Thirty-three percent of patients had poor outcomes (GOS 1 or 2), but no patients who completely recovered (GOS 5) had an increased body temperature (≥37.5 °C) on admission. The mortality rate was lowest in the group of patients presenting with low normal or hypothermic temperatures of less than 36.5 °C (11% versus 22% in the overall sample).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Initial temperature and outcome (Glasgow Outcome Scale [GOS] score 1 [death] through 5 [no symptoms]) in all 251 patients, p < 0.005. The absolute number of patients is indicated above the bars, which indicate the relative frequencies for each outcome group. Shaded bars, <37.5 °C; hatched bars, ≥37.5 °C.

Increased temperatures on admission were associated with the presence of intraventricular blood. Thirty-four of 129 patients (26%) had intraventricular hemorrhage, but only 13 of 122 patients (10%) without intraventricular hemorrhage had abnormal temperatures of 37.5 °C or more (p = 0.001). No association was found between the location of the hemorrhage and increased temperatures. Patients with thalamic hemorrhages did not show a higher incidence of fever.

Prognostic factors retained from the multiple logistic regression model were initial GCS score of 7 or less, age of more than 75 years, hematoma volume of more than 60 cm³, ventricular hemorrhage, and presence of a coagulation disorder (table 3). The initial body temperature was not accepted as an independent explanatory factor in this model. After adding temperature to the “final” model selected, temperature was still not accepted as an explanatory variable.

Findings during the first 72 hours.

For 196 patients, follow-up data were obtained for 72 hours after admission. Fifty-five patients were not included in this analysis because of death (n = 30) or discharge (n = 16) within the first 72 hours or incomplete data (n = 9). In the subgroup of 196 patients, outcome was death (GOS 1) in 24 (12%), vegetative state (GOS 2) in 5 (3%), severe disability (GOS 3) in 88 (45%), moderate disability (GOS 4) in 57 (29%), and no remaining symptoms (GOS 5) in 22 (11%) patients.

Most patients were treated medically. Surgical hematoma evacuation was performed in 37 patients (19%); 47 patients (24%) were treated with a ventriculostomy because of hydrocephalus.

The incidence of increased body temperatures was high. Only 18 patients (9%) had normal temperatures throughout the observation period (<37.5 °C). In 178 patients (91%), temperature was 37.5 °C or more at least once during the observation period, and in 83 patients (42%) 38.5 °C or more at least once during the observation period. The duration of increased temperatures (≥37.5 °C) during the 72-hour observation period was less than 24 hours in 66 (34%), 24 to 48 hours in 70 (36%), and more than 48 hours in 42 patients (21%).

The duration of increased blood glucose levels (>11 mmol/L) during the 72-hour observation period was less than 24 hours in 64 (33%), 24 to 48 hours in 20 (10%), and more than 48 hours in 12 patients (6%). Blood glucose level was less than 11 mmol/L throughout the observation period in 100 patients (51%).

Arterial hypertension (MAP > 120 mm Hg) was present for less than 24 hours in 98 (50%), 24 to 48 hours in 62 (32%), and more than 48 hours in 27 patients (14%). Only 9 patients (5%) had normal BP throughout the observation period.

Pulmonary infection was diagnosed within the first 72 hours in 84 patients (43%); urinary tract infection, in 69 patients (35%). The neurologic complications encountered were hematoma enlargement or secondary hemorrhage in 24 patients (12%) and symptomatic epilepsy in 25 patients (13%).

The multiple regression analysis for the 196 patients with 72-hour follow-up with the dichotomized GOS score as the dependent variable revealed hematoma enlargement or secondary hemorrhage, ventricular hemorrhage, initial GCS score of 7 or less, hematoma volume of more than 60 cm³, duration of increased BP of more than 48 hours, duration of increased blood glucose level of more than 48 hours, and duration of fever of 37.5 °C or more as independent prognostic factors (table 4). In contrast to the first model calculated for all patients, age and coagulation disorders were not significant factors, probably because of the higher early mortality rate of the very elderly patients and of patients with coagulation disorders. Presence of infection, hematoma location, and hematoma evacuation were also rejected as possible independent prognostic factors. After adding infection to the final model selected, infection still was not accepted as an explanatory variable, whereas duration of fever remained significant.

The relationship between duration of temperature of 37.5 °C or more and outcome is depicted in figure 2. All patients with a poor outcome (GOS 1 or 2) had increased body temperatures at least once. Forty-four percent of the patients with favorable outcomes (GOS 4 or 5), but 83% of the patients with poor outcomes (GOS 1 or 2), had increased temperatures for more than 24 hours.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Duration of fever (body core temperature ≥37.5 °C) during the first 72 hours from 191 patients who survived this period of time and outcome (Glasgow Outcome Scale [GOS] score 1 [death] through 5 [no symptoms]), p < 0.0001. The absolute number of patients is indicated above the bars, which represent the relative frequencies for each outcome group. Dotted bars, 0 hours; gray bars, 0 to 24 hours; hatched bars, 24 to 48 hours; black bars, >48 hours.