Abstract
Objective: To identify risk factors for refractory fever after subarachnoid hemorrhage (SAH), and to determine the impact of temperature elevation on outcome.
Methods: We studied a consecutive cohort of 353 patients with SAH with a maximum daily temperature (Tmax) recorded on at least 7 days between SAH days 0 and 10. Fever (>38.3 °C) was routinely treated with acetaminophen and conventional water-circulating cooling blankets. We calculated daily Tmax above 37.0 °C, and defined extreme Tmax as daily excess above 38.3 °C. Global outcome at 90 days was evaluated with the modified Rankin Scale (mRS), instrumental activities of daily living (IADLs) with the Lawton scale, and cognitive functioning with the Telephone Interview of Cognitive Status. Mixed-effects models were used to identify predictors of Tmax, and logistic regression models to evaluate the impact of Tmax on outcome.
Results: Average daily Tmax was 1.15 °C (range 0.04 to 2.74 °C). The strongest predictors of fever were poor Hunt-Hess grade and intraventricular hemorrhage (IVH) (both p < 0.001). After controlling for baseline outcome predictors, daily Tmax was associated with an increased risk of death or severe disability (mRS ≥ 4, adjusted OR 3.0 per °C, 95% CI 1.6 to 5.8), loss of independence in IADLs (OR 2.6, 95% CI 1.2 to 5.6), and cognitive impairment (OR 2.5, 95% CI 1.2 to 5.1, all p ≤ 0.02). These associations were even stronger when extreme Tmax was analyzed.
Conclusion: Treatment-refractory fever during the first 10 days after subarachnoid hemorrhage (SAH) is predicted by poor clinical grade and intraventricular hemorrhage, and is associated with increased mortality and more functional disability and cognitive impairment among survivors. Clinical trials are needed to evaluate the impact of prophylactic fever control on outcome after SAH.
Fever affects approximately 70% of patients with subarachnoid hemorrhage (SAH), and in some, body temperature can be highly elevated for days.1,2 Abnormal thermoregulation in patients with SAH has been attributed to hypothalamic or brainstem injury resulting from the toxic effects of adjacent blood, ischemic injury, or other causes.2
A large body of experimental evidence indicates that even small degrees of hyperthermia can exacerbate ischemic brain injury,3–6 which can occur during the acute phase of SAH7 or as a consequence of vasospasm. Brain temperature elevations have also been associated with hyperemia, exacerbation of cerebral edema, and elevated intracranial pressure after SAH.8,9 Although several studies have linked fever to the development of vasospasm after SAH,10,11 only one has examined the impact of fever on clinical outcome in a multivariate analysis.12 In that study, fever (defined as body temperature exceeding 38.3 °C for 2 consecutive days) was associated with symptomatic vasospasm, and was an independent predictor of death or severe disability at discharge.
New cooling devices, including adhesive external cooling systems and endovascular heat exchange catheters, may enable clinicians to control fever much more effectively than conventional measures had previously allowed.13,14 These devices might allow patients with SAH to be “clamped” to normothermia, thus minimizing fever burden. However, the long-term impact of fever on survival and functional outcome is unknown, and it is unclear whether higher-level aspects of recovery such as cognitive function and quality of life (QOL) are affected by fever. It is also unknown whether admission characteristics can identify patients with SAH who are most likely to experience sustained fever. In this study we sought to identify predictors of fever after SAH, and to evaluate its impact on outcome.
Methods.
Study population.
A total of 580 patients with SAH consecutively admitted to the Neurologic Intensive Care Unit (NICU) of Columbia University Medical Center between July 1, 1996, and May 1, 2002, were prospectively enrolled in the Columbia University SAH Outcomes Project. The study was approved by the hospital Institutional Review Board, and in all cases written informed consent was obtained from the patient or a surrogate with the exception of two patients who declined to participate. The diagnosis of SAH was established on the basis of an admission CT scan or by xanthochromia of the CSF. Patients with secondary SAH from trauma, arteriovenous malformation rupture, or other causes, and those <18 years old, were excluded.
Inclusion criteria.
Inclusion criteria for our study included recording of the first daily maximum temperature on or before SAH day 2 (day 0 refers to the calendar day of SAH) and a minimum of seven daily maximum temperature measurements between SAH day 0 and 10.
Clinical management.
Body temperature was measured hourly with an infrared tympanic thermometer (Genius 3000A, Sherwood Medical, St. Louis, MO). Tympanic thermometers were regularly calibrated and checked for accuracy against bladder temperatures. Treatment for fever (≥38.3 °C) included acetaminophen 650 mg every 4 hours, followed by placement of a water-circulating cooling blanket (SubZero Blanketrol Model 420, Cincinnati Sub-Zero, Inc., Cincinnati, OH) above or below the patient for acetaminophen-refractory fever. Thirty-six patients were also treated with an air-circulating cooling blanket (Polar Air Model 600 Air Cooling System, Augustine Medical, Inc., Minneapolis, MN) prior to application of a water-circulating cooling blanket as part of a clinical trial.15 All patients with persistent fever were treated with broad-spectrum antibiotics, either empirically or on the basis of culture results. The criteria used to define infections in our patient population have been previously described.2
Measures of fever.
On each calendar day we calculated both the extent of fever exceeding normal body temperature and the extent of extreme temperature elevation. Tmax was calculated by subtracting 37 °C from each patient's maximal daily temperature; when the maximal daily temperature was below 37.0 °C, a score of 0 was assigned. We also calculated an extreme Tmax by subtracting 38.3 °C from each patient's maximal daily temperature, with values below this threshold assigned a score of 0. Admission Tmax was calculated as the extent to which each patient's maximal daily temperature exceeded 37 °C on the calendar day of admission. The cutoff of 38.3 °C was selected because it conforms with published practice guidelines for initiating antipyretic therapy and a diagnostic evaluation for infection.16
Clinical and radiographic variables.
Demographic variables, clinical status, radiographic findings, complications, and events were prospectively collected throughout the hospital stay as previously described.17,18 Admission neurologic status was assessed with the Hunt-Hess scale,19 and CT scans were evaluated for the amount and location of SAH and IVH.20,21 The reliability of these evaluations has previously been found to be good to excellent.17 Transcranial Doppler ultrasonography was performed every 1 to 2 days throughout the study period. Complications were adjudicated in weekly meetings of the study team according to prespecified criteria.17,18
Outcome variables.
Three months after SAH each subject or his or her proxy was asked to complete a 45-minute telephone or in-person interview. Global outcome was assessed with the modified Rankin Scale (mRS; scored death = 6, full recovery = 0).22 When 3-month mRS scores were unavailable the 14-day mRS was carried forward. Disability in instrumental activities of daily living (IADLs, i.e., using a phone, housekeeping) was assessed with the Lawton IADL scale (scored 8 = independent, 30 = completely dependent), with poor outcome defined as a score of ≥9.23 Global cognitive functioning was evaluated with the Telephone Interview of Cognitive Status (TICS) and is scored from 0 (worst) to 51 (best).24 TICS scores of 30 or less were coded as impaired.25 Health-related QOL was assessed with the Sickness Impact Profile (SIP),26 with poor QOL defined as a score below the median.27 All assessment instruments were administered in the participant's native language (English or Spanish).
Statistical analysis.
Data were analyzed with SPSS (version 12.0, SPSS Inc.). We used the t-test, the χ2 test, and the Mann-Whitney U test to compare included vs excluded patients. Mixed-effects analysis was used to identify admission predictors of Tmax, with the variable “patient” as a random effect, and the fixed linear, and fixed quadratic effect “SAH day” nested within each patient. The daily Tmax values were assumed to be correlated within each patient, and were modeled using the autoregressive process AR.1 We modeled the main effect and the linear and quadratic interactions with SAH day for each predictor. To test for significant associations between complications and Tmax, we added complications, one at a time, to the multifactor admission prediction model.
To assess the impact of fever on outcome, demographic and admission clinical and radiographic variables were used to construct baseline logistic regression models for each outcome measure. After these baseline models were constructed, admission Tmax, mean Tmax, and mean extreme Tmax values were added individually to calculate adjusted ORs for poor outcome. Tests for interactions were performed for all significant variables retained in the baseline multivariate models. Significance was judged at p < 0.05.
Results.
Study population.
A total of 353 patients (61% of all patients admitted for SAH) were included in the study (figure 1). Compared to the 223 patients who were excluded, the study patients were less often Hunt-Hess grade 1 or 2 (alert and oriented) or grade 5 (comatose) on admission, had more SAH and IVH on CT, were less likely to have an aneurysm ≥10 mm, were more often female or nonwhite, and had a longer ICU length of stay (table 1).
Figure 1. Patient flow diagram. Subarachnoid hemorrhage (SAH) day 0 is defined as the calendar day of onset. A total of 151 patients were excluded because they did not receive medical attention at our institution until after SAH day 2. Another 72 patients were excluded for not having a minimum of seven daily Tmax within SAH days 0 to 10 due to early death (n = 47), early hospital discharge (n = 11), or incomplete temperature records (n = 14). In 75 patients (21% of the total patients) the day 14 modified Rankin score was carried forward because a day 90 score was not available.
Table 1 Comparison of study patients and excluded patients
Timing and magnitude of fever.
At least 10 Tmax values were available in 62% of the study patients. Average admission Tmax was 0.64 ± 0.58 °C (range 0 to 2.71), average daily Tmax was 1.15 ± 0.50 °C (range 0.04 to 2.74), and average daily extreme Tmax was 0.12 °C (range 0 to 1.51). At least one daily maximum temperature above 38.3 °C was observed in 72% of patients (n = 254). Mixed-effects analysis identified a significant quadratic interaction between clinical grade on admission and time, reflecting a sharper rise and fall in Tmax over the first 10 days after SAH onset among poor-grade patients (table 2, figure 2).
Table 2 Admission predictors of fever
Figure 2. Mean daily Tmax during the 10-day study period stratified by admission Hunt-Hess grade (A) and 3-month outcome according to the modified Rankin Scale (B). A modified Rankin Scale of 4 to 6 indicates death or moderate to severe disability.
Admission predictors of fever.
Multifactor mixed-effects analysis identified Hunt-Hess grade, presence of IVH, SAH sum score, large aneurysm size, and loss of consciousness at ictus as significant admission predictors of Tmax (table 2).
Complications associated with fever.
After controlling for the admission predictors of Tmax, eight complications were significantly associated with Tmax in multifactor mixed-effects analysis. Four of these complications were related to secondary neurologic injury, one to infection, and three to general medical complications (table 3).
Table 3 Complications associated with fever after adjustment for admission predictors using mixed-effects analysis
Impact of fever on outcome.
According to the mRS, 14% (52/353) of patients were dead at 90 days (score of 6), and 21% (73/353) were classified as moderately to severely disabled (score of 4 to 5). Poor Hunt-Hess grade (p < 0.001), aneurysm size ≥10 mm (p < 0.001), and older age (p < 0.001) were identified as significant baseline predictors of death as well as death or moderate to severe disability. After controlling for these baseline predictors, mean Tmax and mean extreme Tmax were significantly associated with both death and death or moderate to severe disability (table 4). Among survivors, 54% (139/257) of evaluable patients had limitations in instrumental ADLs, and 30% (63/208) had cognitive impairment according to the TICS. After controlling for baseline predictors for each of these outcomes, mean Tmax and mean extreme Tmax were significantly associated with poor outcome in these aspects of recovery as well (table 4). The magnitude of association with poor recovery was consistently greater for mean extreme Tmax than for mean Tmax for all aspects of outcome. Admission Tmax was not significantly associated with any aspect of outcome, and no significant associations were found between extent of temperature elevation and QOL.
Table 4 Association of fever with 3-month outcome after adjustment for baseline predictors
Discussion.
In this study of 353 patients with SAH, fever was predicted by the severity of initial hemorrhage and was associated with an increased risk of death, disability, and cognitive impairment at 3 months. The extent of fever exceeding 38.3 °C (101.0 °F) further increased risk for poor outcome in these aspects of recovery, indicating that extreme temperature elevations may be particularly deleterious after SAH.
We found a considerable burden of fever in our study—average daily Tmax was 1.15 °C above normal—which is consistent with other studies reporting a high frequency of fever after SAH.1,2,10–12 This reflects the therapeutic challenge of managing temperature in patients with SAH, as well as the relative lack of efficacy of conventional measures (antipyretics and water-circulating cooling blankets) to control fever in brain-injured patients.14,28
Poor Hunt-Hess grade, presence of IVH, SAH sum score, large aneurysm size, and loss of consciousness at ictus were identified as significant admission predictors of fever using mixed-effects analysis (table 2). IVH has been linked to unexplained fever in unselected neuro-ICU patients,2 increases the risk of fever after intracerebral hemorrhage,29 and causes spontaneous temperature elevation when induced experimentally, suggesting a link with neurogenic or central fever.30 Most notably, patients with poor clinical grade on admission had a significantly higher temperature curve than good-grade patients (figure 2). These findings suggest that the burden of fever after SAH is strongly associated with the severity of the initial bleeding event.
Fever has been attributed to infection in 50 to 70% of febrile patients with SAH in prior studies2,10; however, pneumonia was the only infectious complication significantly associated with fever in our study. Secondary causes of neurologic deterioration (namely brain stem herniation, symptomatic vasospasm, treated hydrocephalus, and cerebral infarction) were associated more frequently than infections, strongly implicating central mechanisms triggered by brain injury in the pathogenesis of post-SAH fever. Several investigators have noted a temporal association between the development of vasospasm and fever after SAH,31 as did we (figure 2). Transcranial Doppler evidence of vasospasm was not associated with the development of fever in our study, indicating that tissue ischemia may be a more important determinant of fever than the extent of arterial spasm per se. Hypotheses regarding human thermoregulatory disturbances after brain hemorrhage derived from animal models have proposed damage to the hypothalamus, midbrain, or pons by enhanced sympathetic activity, acute or delayed ischemia from vasospasm, toxic blood metabolites, and physical distortion as potential causes of hyperthermia.32,33
Interestingly, we found a significant association of anemia as well as hyperglycemia with temperature elevation. This link between deranged physiologic variables may be a reflection of a generalized disturbance of the brain's ability to maintain physiologic homeostasis, and to our knowledge, represents the first time that these variables have been linked to elevated temperature in a multifactor analysis. Respiratory failure was also found to be associated to Tmax, which could be explained by an increased risk for ventilator-associated pneumonia in this group of patients.
The most important finding of our study is the association of increased temperature with mortality, functional disability, and cognitive impairment (table 4). After controlling for admission predictors, each 1 °C increase in mean Tmax was associated with an eightfold increase in the odds of death, and a threefold increase in the likelihood of death or severe disability. Among survivors, mean Tmax increased the odds of cognitive impairment or disability in instrumental activities of daily living approximately 2.5-fold. Each of these associations became even stronger when elevations exceeding 38.3 °C were analyzed. We did not find an association between elevated temperature and QOL as assessed with the SIP. This may reflect the fact that QOL is a multidimensional assessment of the patient's perceived physical, social, and emotional health status and as such is less dependent on the severity of neurologic injury.
Admission body temperature predicts mortality in patients with ischemic stroke,34,35 and in a study of 173 patients with SAH, mortality was 60% among those with admission temperatures exceeding 37.5 °C compared to 35% among those who were afebrile.11 In our study, admission Tmax was not significantly associated with outcome. The association of admission body temperature with outcome after cerebral infarction is understandable, since the majority of ischemic injury is completed within 6 hours of onset. After SAH, a greater burden of temperature-modifiable brain injury probably occurs during the first 10 days after hemorrhage.
Several limitations of our study deserve mention. First, because all patients were routinely treated with standard antipyretic and cooling blankets, our study essentially reflects the impact of treatment-refractory fever on outcome, rather than pure untreated fever. The intensity of fever was calculated based on daily Tmax and extreme Tmax values, whereas other authors have analyzed temperature hourly.13 A continuously monitored method of recording temperature might have been even more informative, and might have yielded more accurate associations with outcome. We measured tympanic temperature, which may not be as accurate as bladder, rectal, or esophageal temperatures for reflecting core and brain temperature.36,37 Compared to those who were excluded, our study patients tended to have larger amounts of blood, smaller aneurysms, and a longer ICU length of stay. Hence, patients who were in good condition (Hunt-Hess grade 1 or 2) or very poor condition (Hunt-Hess grade 5) were relatively underrepresented. Thus, our findings cannot be readily generalized to good grade nonaneurysmal or extremely poor grade patients. Due to incomplete 90-day follow-up we carried forward 14-day mRS scores in 21% of patients, and in 15 to 30% of patients TICS, Lawton IADL, and SIP data were not available. This may have reduced our power to detect associations. Since we only evaluated outcome at 90 days, it remains unclear if fever is associated with poor functional and cognitive outcome over a longer period of time. Finally, it remains unclear whether increased temperature actually causes neurologic injury, or is merely a marker of more severe neurologic injury. Clinical trials assessing the impact of prophylactic fever control on outcome after SAH are warranted.
Footnotes
Editorial, see page 973
This article was previously published in electronic format as an Expedited E-Pub at www.neurology.org.
Supported by a Grant-in-Aid from the American Heart Association to Dr. Mayer (#9750432N).
Disclosure: Dr. Connolly has received research support from Innercool Therapies. Dr. Mayer has received research support, consulting fees, speaking honoraria, and stock options from Medivance, Inc. and stock options from Radiant Medical. The remaining authors have nothing to disclose.
Received March 10, 2006. Accepted in final form September 19, 2006.
References
Letters: Rapid online correspondence
REQUIREMENTS
You must ensure that your Disclosures have been updated within the previous six months. Please go to our Submission Site to add or update your Disclosure information.
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
You May Also be Interested in
Alert Me
Recommended articles
Global and domain-specific cognitive impairment and outcome after subarachnoid hemorrhage
Haptoglobin and the development of cerebral artery vasospasm after subarachnoid hemorrhage
Predictors and clinical impact of epilepsy after subarachnoid hemorrhage
Endovascular treatment of cerebral vasospasm after subarachnoid hemorrhage