Fever in subarachnoid hemorrhage

Patients and methods.

We prospectively studied consecutive patients with nontraumatic SAH who were admitted between August 1998 and June 1999 to the neurologic intensive care unit of a major urban hospital. Patients were referred for treatment of SAH from local community hospitals or were directly admitted through the emergency department. Surgical clipping or endovascular coiling was accomplished in the first 24 hours of admission in almost all cases. As part of standard care, patients were monitored daily with transcranial Doppler ultrasound (TCD) through a transtemporal window for the presence of VSP, and treated with hemodynamic augmentation therapy (“triple-H”) if necessary. When present, fever was treated promptly in all patients with antipyretic drugs and cooling blankets. Patients with VSP refractory to clinical management underwent angiography with intent to perform angioplasty or papaverine infusion on an individualized basis.

Data were obtained prospectively, on admission: Hunt–Hess grade, Fisher group, and Glasgow Coma Score; on a daily basis: maximum tympanic temperature; presence of ventriculostomy, intubation, and infections. Intubation was always recorded except when electively performed for aneurysm treatment and the patient remained intubated less than 24 hours. Tympanic temperature was measured every 2 hours in all patients by using infrared thermometers (LighTouch, Exergen Corp., Watertown, MA). Data collection was complete. Rankin scores at hospital discharge were estimated from review of both physical therapy and physician notes. History of hypertension, diabetes, ethanol abuse, daily TCD velocities, microbiological culture data, and chest radiography reports were obtained from chart reviews.

Fever was defined as any tympanic temperature greater than 38.3 °C. To compare various causes of fever, patients were separated into two groups: febrile (temperature > 38.3 °C for 2 or more consecutive days) and afebrile (no fever or isolated temperatures > 38.3 °C). The total “number of days febrile” was considered the number of days during hospitalization that a patient had any temperature > 38.3 °C. For this parameter, no distinction was made between patients with consecutive or nonconsecutive days with fever. VSP was considered present if middle cerebral artery, internal carotid, anterior cerebral artery, or posterior cerebral artery mean velocities were greater than 120 cm/s by TCD, or if present by angiographic criteria. Symptomatic VSP was defined by the presence of TCD or angiographic criteria for VSP in addition to a new focal neurologic deficit. To be identified as having no infection, febrile patients required normal chest films and negative cultures on blood, CSF, sputum, catheter, and urine samples. Afebrile patients with partial investigation showing no infection (e.g., negative chest film and urine culture but no sputum or catheter tip available) were also considered not to have an infection. Pneumonia was defined by a new focal infiltrate on chest film or a diffuse infiltrate with abnormal sputum (abundant polymorphonuclear cells or pathogenic organism).17 Definitions of “possible” and “definite” pneumonia categories were adapted from previously published criteria.18 Upper airway infection was defined as a normal chest film with abnormal sputum. Urinary infections were defined by a positive culture and a urine sample with pyuria. Catheter-related infection was considered present if blood or catheter tip cultures grew pathogenic organisms. CSF infections were defined by positive cultures with pathogenic organisms.

Definitions of fever, infections, and VSP were prespecified before data collection and analysis. Based on previous work on normal daily temperature variations, we use a tympanic temperature greater than 38.3 °C (101 °F) as a cutoff for investigating possible infections.19 Two or more consecutive days with fever was chosen to maximize available culture data as well as to exclude other confounding causes of isolated febrile episodes, such as atelectasis or brain surgery, etc. Discharge outcomes were dichotomized into good (Rankin score 0 to 2) or poor (Rankin score 3 to 6) with a validated measure of functional independence.20

Univariate analysis was performed by using Fisher’s exact test for categorical variables, and Mann–Whitney U test for continuous variables. A two-step logistic regression model including all patients was used for both fever and clinical outcome prediction: all variables with a possible univariate association (p < 0.1) were initially included, then variables with definite lack of association (p > 0.3) excluded from the final logistic regression model. Spearman’s rank correlation coefficient (r_s) was used to examine the relationship between fever and clinical outcome data.

Results.

Ninety-two patients with nontraumatic SAH were admitted between August 1998 and June 1999. Mean time from onset of symptoms to admission was 2.4 ± 2.9 days. Baseline characteristics are displayed in table 1. Most patients were women (71%) with anterior circulation aneurysms (63%). No aneurysm was found in 10 (11%) patients. Most patients had good clinical grades on admission, as expressed by low (1 to 3) Hunt–Hess grades (79%) and high (13 to 15) Glasgow Coma Scores (58%). Seventy-two of 82 (88%) patients with aneurysms had either surgical clipping or endovascular coiling within 24 hours of admission.

Thirty-eight patients developed fever for at least 2 consecutive days. Evidence for infection was found in 28 of 38 (74%) febrile patients and in 27 of 54 (50%) afebrile patients. Investigation of infectious source included blood and urine cultures in all febrile and most (57% and 89%, respectively) afebrile patients. Chest films were available for all but three patients (one febrile and two afebrile). Sputum, CSF, and catheter tip cultures were available in most febrile patients (89%, 84%, and 76%, respectively). Most common sites of infection were lung (22 patients with possible, 15 with definite infections), urine (14 patients), catheter-related (11 patients), upper airways (9 patients), and brain/meninges (5 patients). Gram-negative organisms, followed by Staphylococcus aureus and S. pneumoniae, caused pneumonia in most instances, whereas in 10 (25%) patients, respiratory flora was polymicrobial. No infectious source was found in 10 of 38 (26%) patients with fever. No difference was found among patients with infectious vs noninfectious fever with regard to age, medical history, maximum temperature, or peripheral white blood cell count.

Table 2 shows the univariate predictors of fever. Patients with fever were more severely ill, both at baseline and during the hospital stay. Adverse Hunt–Hess grades, Fisher groups, and Glasgow Coma Scores on admission, and in-hospital events (VSP, symptomatic VSP, papaverine or angioplasty treatment, infection, ventriculostomy, intubation) correlated with more frequent fever (p < 0.05). Age correlated with fever with borderline significance (p = 0.054). Fever was not associated with aneurysm location, whether anterior vs posterior circulation, or anterior communicating artery (Acomm aneurysm) vs other locations. Patients undergoing surgical clipping or endovascular coiling had the same incidence of fever. In the multivariate analysis, three variables remained significant predictors of fever ( table 3): presence of ventriculostomy (OR = 8.5; 95% CI = 2.4 to 29.7), symptomatic VSP (OR = 5.0; 95% CI = 1.03 to 24.5), and age (OR = 1.75 per 10 years; 95% CI = 1.02 to 3.0). Both VSP (irrespective of being symptomatic or not) and Fisher Group 3 correlated with fever with borderline significance (0.05 < p < 0.1).

Forty-six of 92 (50%) patients remained functionally dependent on discharge (Rankin score > 2). Outcome was related to disease severity ( table 4). Adverse clinical grades on admission, older age, infection, fever, presence of ventriculostomy or intubation, symptomatic VSP—but not any VSP—were all associated with poor outcome (p < 0.01). Number of days febrile correlated with Rankin scores on discharge (r_s = 0.46, p < 0.001) ( figure). In the multivariate model, three variables predicted poor outcome ( table 5): number of days with fever (OR = 1.4 per day; 95% CI = 1.11 to 1.88), older age (OR = 1.64 per 10 years; 95% CI = 1.04 to 2.58), and presence of intubation (OR = 21.8; 95% CI = 5.6 to 84.5). Fisher Group 3 predicted poor outcome with borderline significance (p = 0.064). Mean length of hospital stay was 12.6 ± 7.4 days for afebrile patients and 24.2 ± 12.2 days for febrile patients.

• Download figure
• Open in new tab
• Download powerpoint

Figure. Box plots of the number of days febrile by discharge Rankin score in 92 patients with subarachnoid hemorrhage (r_s = 0.46, p < 0.001). The lower and upper bounds of the box indicate the first and third quartiles, and the thicker line indicates the median number of days febrile. Circles indicate minimum and maximum, whereas X indicates the mean number of days febrile.

Discussion.

Fever is a frequent concern in critically ill patients with SAH.2 Despite its common occurrence, we found only four case series in the published literature examining fever in SAH. In a careful retrospective analysis of 312 patients with SAH, fever was mentioned, without any statistical scrutiny, as a poor prognostic sign.14 Two separate case series suggested that a “plateau-type” fever pattern was more common in patients who developed VSP.12,16⇓ Finally, a study of 173 patients showed a univariate association between fever, VSP, and increased mortality.15

There were two main findings in our study: First, patients with symptomatic VSP had an increased risk (OR = 5.0) of developing fever independent of disease severity or presence of infection (p < 0.05). This finding implies that either VSP may be a cause of fever or both VSP and fever may have a common pathogenesis. Hypothalamic injury is reported in 61% of SAH patients undergoing autopsy21; specific tests of hypothalamic function indicated abnormalities in 14 of 18 selected patients in another study22; and a causal relationship was speculated in a case of SAH with diencephalic infarction and fever of noninfectious origin.13 Common pathogenetic mechanisms also have been studied in VSP and fever in SAH. In experimental models, hemoglobin injected into the CSF causes fever, probably mediated by release of prostaglandins E₂ or F_2α.11 VSP pathogenesis is thought to be mediated through various blood degradation products and substances released around the clot, and prostaglandin F_2α has been shown both to be increased in the CSF of SAH patients with VSP and to fluctuate with neurologic symptoms.23 The exact pathogenesis relating fever and VSP cannot be ascertained by this study and thus deserves further investigation.

The second main result in our study was an increased risk of poor outcome for each day that a patient has fever (OR = 1.4). This relationship was independent of disease severity, VSP, or infection. To our knowledge, this independent association has not been previously shown. In stroke patients, the association between fever and poor outcome is well established—even as small increments as 1 °C increase the risk of poor outcome 2.2-fold.8 Fever may exert its adverse effects by increasing cerebral metabolism, glutamate release, edema formation, or blood–brain barrier disruption.4-7,24⇓⇓⇓⇓ Therefore, one possibility is that the association between fever and poor outcome in SAH could be related to the worsened brain injury associated with increased brain temperature.

As expected, the causes of fever in these critically ill patients were varied. Presence of a ventriculostomy increased the risk of becoming febrile in this population more than eightfold. The cause of this association is unknown but may be related to common effects of blood in the CSF causing hydrocephalus, as well as fever and VSP. Older patients were also at risk becoming febrile. A relationship between age and increased risk of intraventricular hemorrhage, hydrocephalus, and VSP has been demonstrated in another study and may partially explain our findings.25 Lastly, Fisher Group 3 had a univariate association with fever with a trend in the multivariate analysis (p = 0.085). In Fisher’s original studies, patients with thick layers of clot in the basal cisterns (Group 3) had a higher incidence of symptomatic VSP than any other groups.26,27⇓ The possible association between Fisher group and fever may be explained by the direct effects of more blood in the CSF or substances released around the clot causing both VSP and fever.

Numerous studies have investigated predictors of outcome in SAH.25,28-31⇓⇓⇓⇓ In our study, only two factors besides fever were significantly associated with poor outcome: older age and presence of intubation. Older age has been consistently associated with worse outcome1,31,32⇓⇓; this association has been attributed to greater rates of concomitant illnesses, lessened ability to recover neurologic function after brain injury, and in-hospital complications.32 Intubation as a marker of outcome, to our knowledge, has never been investigated specifically in SAH. One study of 230 patients with ischemic or hemorrhagic stroke showed that intubated patients fared significantly worse.33 The association found in our study was striking, with a much higher proportion of poor outcome in intubated patients (OR = 21.8). Because intubation is a marker of disease severity, this strong association may explain why clinical grades on admission were not significant predictors of outcome in the multivariate analysis. When intubation was excluded from the multivariate model, more severe Hunt–Hess grade, Fisher group, and Glasgow coma scores were all associated with poor outcome (p < 0.05, data not shown).

This study has important practical applications. Fever in patients with SAH should be thoroughly investigated, pneumonia being the most frequent cause. Instituting treatment with empiric antibiotics may be appropriate to cover the most common offending pathogens. However, our data suggest that more than one quarter of patients have noninfectious causes of fever. In these patients, drug-induced and central fever are possibilities; a trial of switching medications and investigating the patient for signs of VSP in this setting is more appropriate than continuing empiric antibiotic therapy.

The prespecified definitions used in this study were chosen based on standard practice. Definition of fever was based on previous work on normal daily temperature variations,19 even though smaller temperature increases have been associated with poor outcome in stroke. Although we had a low threshold to define presence of infection, we cannot definitely exclude that some patients with infections might have gone undiagnosed, because no culture method is 100% sensitive. This study did not address whether treating fever directly impacts outcome. Ideally, future studies with randomized, controlled setting would answer whether aggressive temperature lowering is beneficial in SAH. Finally, because this study was performed in a single referral center, it is possible that the incidence and contribution of fever to outcome may vary in other populations.