A 12-Year Audit of Neurological Outcomes Associated with Core Body Temperature >37.5°C in Children with Severe Traumatic Brain Injury

Abstract

Hyperthermia exposure is associated with poor neurological outcomes in patients with severe traumatic brain injury (TBI). Our aim was to describe fever in children with severe TBI admitted to a pediatric intensive care unit (PICU) for at least 72 hours and to evaluate associations between fever using a novel approach to describe thermal exposure and neurobehavioral outcomes. The cohort included children from birth to 17 years of age admitted to the PICU between 2000 and 2012 for at least 72 hours who had severe TBI with intracranial pressure monitoring. Patients with non-accidental TBI or pre-existing developmental delays were excluded. Hyperthermia was defined as a core temperature >37.5°C. Hourly temperature measurements were used to calculate the area under the curve (AUC) using the linear trapezoidal rule. Each participant was followed up at the Brain Injury Clinic 6 and 18 months postinjury. Neurobehavioral outcome scores were analyzed against AUC using standard statistical methods. Ninety-eight patients admitted with severe TBI met the study inclusion criteria. Hyperthermia/fever was common (89.7%), and patients remained hyperthermic for a median of 9.4 hours. However, no statistically significant association was found between AUC and abnormal neurological outcomes. The follow-up rates were good at 6 (86.7%) and 18 months (83.7%). The neurological deficit improved with time, with “favorable outcomes” increasing from 72% to 94% at the respective follow-up months. Our study used a novel method to describe patients' fever, providing a different indicator of thermal exposure than that previously reported. In addition, the AUC was well correlated with the maximum temperature recorded and the proportion of time >37.5°C, indicating that it is a good surrogate for thermal exposure. Interestingly, the neurological disabilities of the patients improved over time.

Background

Traumatic brain injury (TBI) is a global health issue affecting individuals and their immediate families (Andelic et al, 2018). Young people seem disproportionately affected (AIWH, 2020), impacting their development, education, and possible future career aspirations (Mealings et al, 2012; Ryan et al, 2016). One of the treatment strategies for TBI is to manipulate the core body temperature by focusing on cooling, which is known as therapeutic hypothermia. However, there is a gap in the literature regarding the consequences of fever in pediatric patients with TBI (Kochanek et al, 2019). Understanding how fever affects pediatric patients may help develop strategies to improve the long-term consequences of this devastating injury.

Clinical teams managing patients with TBI have focused on the impact of increased intracranial pressure (ICP) on neurological outcomes (Dunn, 2002). Interestingly, high core body temperature and ICP were also associated with poor neurological outcomes at 1-year follow-up (Jiang et al, 2002). Consequently, meticulous ICU management is vital to prevent increased ICP and to minimize further neurological injury (Unterberg et al, 1993). Clinical teams have protocols and guidelines (Kochanek et al, 2019) developed by international expert groups to help manage such patients. Implementation of these protocols has resulted in better neurological outcomes, reduced mortality rates, and increased home discharge rates after pediatric TBI (O'Lynnger et al, 2016).

TBI affects >3 million children globally per year, with ∼200,000 deaths (Dewan et al, 2016). For the past 20 years, clinicians have made a few inroads in changing the prognosis of severe TBI, with disability and death remaining stubbornly at the same rate (Taylor, 2004). The relationship between fever, increased ICP, and poor neurological outcomes has been discussed in the literature, indicating that fever and raised ICP can be used to predict a negative neurological outcome (Bonds et al, 2015). Therefore, to understand how much thermal exposure is required to affect neurological outcomes in children with severe TBI, we used a novel approach and calculated the area under the curve (AUC) for core temperature for the first 72 hours after admission and correlated this to neurobehavioral outcomes. We performed a retrospective observational study of 130 patients whose admissions occurred over 12 years to assess whether hyperthermia is detrimental to neurobehavioral outcomes.

Methods

We conducted a retrospective study at a university-affiliated hospital in a large metropolitan city. The participants were children admitted to the pediatric intensive care unit (PICU) with severe TBI between 2000 and 2012. From 2006 to 2010, the unit participated in a hypothermia in traumatic brain injury pilot RCT (HiTBIC) (Beca et al, 2015). This study randomized patients into a hypothermia or normothermia arm for 72 hours. Before the participation in HiTBIC the unit would manage core temperature at clinicians' discretion with the use of cooling blankets. The inclusion criteria were neonates, infants, and children up to 17 years of age with severe TBI requiring intubation, ventilation, placement of an ICP monitoring device, and admission to the PICU for at least 72 hours. Demographic data collected included age, sex, Glasgow Coma Score (GCS), and the mechanism of injury. We excluded patients with proven non-accidental TBI and those with pre-existing developmental delays.

We extracted each patient's hourly temperature measurements during the first 72 hours of PICU admission using a computerized clinical information system (Sunrise Critical Care™; Eclipsys Corporation). Our unit's standard practice measures “core body temperature” using rectal/esophageal Philips™ probes. Hyperthermia has no consensus definition; therefore, we defined it as a core temperature >37.5°C. This temperature was selected because it is the threshold for clinicians to consider changes in body temperature. The fever/hyperthermia dose was described as the AUC for temperatures >37.5°C.

Temperatures >37.5°C were plotted against time for each study participant (Fig. 1). A linear trapezoidal rule was used to calculate the AUC for temperatures >37.5°C. The maximum or “peak” temperatures and the proportion of time the core temperature was >37.5°C were calculated for each patient. These two variables were compared with the AUC using Spearman's correlation coefficient.

FIG. 1.

Graphical representation of temperature >37.5°C versus time for a patient. *The shaded area in the graph is the calculated AUC *On the y-axis, the reference point 0.0 is temperature of 37.5°C, and each mark represents a unique temperature above the reference point. (e.g., 0.5 equals the patient temperature of 38°C). AUC, area under the curve.

Each participant was examined at the Brain Injury Clinic 6 and 18 months postinjury. Clinical notes were reviewed, and information was extracted from formal clinical and neuropsychological assessments. The interviews included neurobehavioral assessments, motor deficit assessments, and developmental milestones. The Glasgow Outcome Score (GOS) (Rhine et al, 2012) was calculated based on the clinical assessments.

The collected data were subjected to statistical analysis using the SAS^® 9.3 software (SAS Institute, Inc., Cary, NC). The data are summarized as mean, standard deviation, median, and interquartile range for continuous variables. For categorical variables, frequencies and percentages were used to describe the patient characteristics and study outcomes. Associations between temperature summaries and study outcomes were assessed using Spearman correlation and Wald chi-square tests, and were considered significant if the probability was ≤0.05. This study was approved by the Sydney Children's Hospital Network Human Research Ethics Committee (HREC reference: LNR/14/SCHN/66).

Results

Study population

A total of 130 patients with TBI were admitted to the PICU between 2000 and 2012 (Fig. 2). Of these, 98 (75.4%) met the study inclusion criteria, and their characteristics are shown in Table 1. Of the excluded patients, 26 had an ICU stay of <72 hours, 3 had non-accidental injuries, and 3 had pre-existing medical conditions, temporal lobe epilepsy, x-linked adrenoleukodystrophy, and moderate speech and language delay.

FIG. 2.

Flow diagram showing screening, inclusion, and exclusion in the study.

Table 1.

Patient Characteristics in the Study

Characteristic	TBI patients (n = 98)
Age, median (IQR), years	8.12 (4.01–11.83)
Male sex, n (%)	67 (68.4%)
GCS, mean (IQR)	4 (3–5)
Mechanical ventilation, mean (IQR), days	9 (0.5–30.5)
ICP monitoring, mean (IQR), days	6.5 (1–22)
PICU LOS, mean (IQR), days	11.5 (3–36.5)
Hospital LOS, mean (IQR), days	58.9 (7–260)
Mechanism of injury
Motor vehicle accident (passenger), n (%)	31 (32%)
Car vs. pedestrian, n (%)	24 (25%)
Fall from height, n (%)	18 (18%)
Fall from moving object (bike/horse), n (%)	17 (17%)
Car vs. bike, n (%)	6 (6%)
Penetrating injury/assault, n (%)	2 (2%)

GCS, Glasgow Coma Score; ICP, intracranial pressure; IQR, interquartile range; LOS, length of stay; PICU, pediatric intensive care unit; TBI, traumatic brain injury.

Fewer females (31.6%) were admitted to the PICU with TBI than males (68.4%). The mean age of the patients was 8 (range 0.5–17 years). The mean (IQR) GCS score on first ICU admission was 4 (range 3–5). Mortality was 2.1% (two patients) at 30 and 90 days, with both deaths occurring in the intensive care unit.

The leading mechanism of injury for the cohort was motor vehicle accidents (Passenger) (31.6%), followed by motor vehicle accidents and pedestrian or cyclist accidents (24.5%) (Fig. 3).

FIG. 3.

Mechanism of injury of the 98 patients who had a severe traumatic brain injury and were included in the study. MVA, motor vehicle accidents.

Temperature analysis

A summary of the descriptive statistics for temperature >37.5°C is shown in Table 2. The median AUC was 4.3 (maximum 54.3). The larger the AUC, the longer and more severe is the temperature dose for a particular participant. The incidence of hyperthermia/fever >37.5°C was 89.7%. The median proportion of time with a temperature >37.5°C was 9.4 hours (maximum 63.3 hours), and the median maximum or peak temperature was 38.4°C (maximum 40.6°C).

Table 2.

Descriptive Statistics for Temperature (Area Under the Curve, Proportion of Time with Temperature >37.5°C, and Maximum Temperatures)

Temperature >37.5°C	AUC	Proportion of time (hours)	Maximum temperature (°C)
Median	4.3	9.4	38.4
IQR	1.35–9.67	3.6–21.6	38–38.7
Maximum	54.3	63.3	40.6

AUC, area under the curve; IQR, interquartile range.

We used Spearman's correlation coefficient and compared the AUC with the maximum or peak temperatures, and the proportion of time with a temperature >37.5°C. We found a strong positive correlation between AUC, maximum temperature (r = +0.845, p < 0.0001), and the proportion of time with a temperature >37.5°C (r = +0.940, p < 0.0001) (Fig. 4).

FIG. 4.

Correlation of AUC versus maximum temperature (r = +0.845, p < 0.0001) and proportion of time >37.5°C for each patient (r = +0.940, p < 0.0001).

Follow-up

The follow-up rates in the brain injury clinic for this cohort were 86.7% (n = 85) and 83.7% (n = 81) at 6 and 18 months, respectively. Unfortunately, 2 of our cohort died, 10 were transferred to other hospitals, and 1 relocated overseas. By the 18-month time point, four of our cohorts were deemed fully recovered and discharged from the clinic.

Looking at the early outcome (6 months) versus the late outcome (18 months) postinjury, the percentage with “favorable outcome” {GOS 4 and 5 (good outcome and moderate disability)} increased from 72% to 94%, and those with severe disability significantly reduced from 27.6% to 6.1% (Fig. 5). Thus, there was an appreciable improvement in the neurobehavioral and motor domains when comparing the early versus late outcomes. However, this was not replicated in the milestone domain and the overall results, where the outcomes were very similar.

FIG. 5.

Outcome of patients that at 6 and 18 months follow-up. The outcome score used is “Glasgow Outcome Score.”

There was no statistically significant association between temperature variables (AUC, proportion of time with temperature >37.5°C and maximum temperature) and abnormal neurological outcomes. The odds ratio (OR) for AUC; OR: 0.96, confidence interval (95% CI) 0.90–1.01 (p = 0.131), proportion of time temperature >37.5°C; OR: 0.27, 95% CI 0.02–4.68 (p = 0.368) and maximum/peak temperature; OR: 0.71, 95% CI 0.30–1.67 (p = 0.433).

The PICU length of stay (LOS) was associated with abnormal neurological outcomes (OR: 1.16; 95% CI 1.003–1.344; p = 0.045). Using Spearman's correlation coefficient, a very weak negative correlation was observed between the AUC and PICU LOS (r = −0.17, p = 0.087) and duration of ICP monitoring (r = −0.21, p = 0.044).

Discussion

Our study evaluated the possible association between fever in the first 72 hours after admission to the PICU and neurological outcomes in children with severe TBI. Our results did not demonstrate any association between hyperthermia and poor neurological outcomes (mortality, abnormal neurology, functional deficits, and GOS). Thus, our results contrast with those reported in previous observational TBI studies, in which hyperthermia was associated with unfavorable outcomes (Bonds et al, 2015; Jiang et al, 2002; Natale et al, 2000; Saxena et al, 2015). Most of the published literature is on adult patients with TBI, and there is limited evidence on the impact of fever on neurological outcomes in children.

Therefore, we speculate that our results represent a biological process that protects the brain from the negative effects of fever. A possible biological explanation could be the neuroplasticity of the pediatric brain. However, the argument that neuroplasticity protects a child's brain from severe TBI is disputed by Anderson et al (2012) and Crowe et al (2012), who reported that children demonstrate lower cognitive function 3 years after severe TBI. It is well recognized that severe TBI can cause significant neurological deficits, including cognitive, behavioral, and social deficits that may result in lifelong disability (Paget et al, 2012). Despite evidence of neuroplasticity, some children with severe TBI may never recover their neurological function, particularly when compared with their peers.

An alternative explanation for the lack of association between fever and neurological outcomes could be the temperature threshold selected to define fever. With no clear consensus on fever, thresholds of 36.5°C or 38.5°C could have been chosen and deemed biologically plausible. A different threshold would have significantly impacted the calculated “area under the curve” (AUC). A larger or smaller AUC could have created a more sensitive marker for the impact of fever on neurological outcomes. We recognize that the incidence of fever (core temperature >37.5°C) was high in our study population (89.7%), which is consistent with other clinical experiences (Jiang et al, 2002; Natale et al, 2000). However, the amount of time a patient was hyperthermic was relatively short, with a median of 9.4 hours out of 72 hours of the study period. This short exposure to hyperthermia could be explained by evolving clinical practice and the use of cool devices during the study period to manage refractory ICPs.

Finally, the lack of association between fever exposure and neurological outcomes could be due to the lack of sensitivity of the neurological assessment tool. The GOS is important for describing large neurological deficits, from no deficit to severe deficit and death. As intensive care teams become more successful in saving lives, more sophisticated tools are needed to better understand the impact of our management strategies. The assessment of neuropsychological outcomes in children post-TBI is complex and many tools are available to measure outcomes. GOS was used in this study because it is a validated, simple, and widely accepted tool. Using the traditional (Agrawal et al, 2012; Bahloul et al, 2011) grouping of favorable (GOS 4 and 5) and unfavorable (GOS 1, 2, and 3) cohorts, the number of patients with favorable outcomes increased from 72% at 6 months to 94% at 18 months postinjury.

Although the GOS is good at categorizing outcomes, it can be considered a blunt tool and is limited in its ability to truly predict neurocognitive deficits, especially for survivors with clear neuropsychiatric deficits on close examination (Ramezani et al, 2018). The functional deficit profile and GOS suggested that more patients improved over time. This is consistent with what has been documented in other studies. Thakker et al (2014), Shaklai et al (2014), and Andruszkow et al (1997) reported improved functional and GOS outcomes in follow-up studies of children with moderate-to-severe TBI. In this cohort of patients, the follow-up rate was very good; however, there was no follow-up for patients who were transferred to other institutions.

We found an association between the length of the PICU stay and abnormal neurological outcomes. This relationship between prolonged PICU admission and worse neurological outcomes is consistent with the results of previous reports. For example, Rhine et al (2012) reported that children with inflicted brain injury who spent 10 or more days in the PICU were ∼21 times more likely to have a poor GOS score. However, that study assessed children with inflicted injuries, which we excluded because they had significantly worse outcomes than those with similar injury severity (Paul and Adamo, 2014).

The overall mortality rate of this group was very low, with 30- and 90-day mortality rates of 2.1% resulting from complications from the initial TBI. The low mortality in our cohort may be due to the strict inclusion criteria, as all patients discharged before the 72 hours were excluded, including possible deaths that occurred in the first 72 hours of admission to the PICU. Other observational studies in children with severe TBI have reported an 18–30% mortality rate, with most deaths occurring within the first 48–72 hours (Ducrocq et al, 2006; Thakker et al, 1997).

Our study used a novel method to quantify the thermal exposure experienced by patients. We defined normal core body temperature as ≤37.5°C and acknowledged the lack of consensus in the literature, with clinical guidelines such as the Brain Trauma Foundation (Kochanek et al, 2019) failing to define normothermia. Thus, the AUC is a unique way of representing the cumulative temperature over time instead of using peaks and troughs or consecutive hours, as described in other observational studies (Jiang et al, 2002; Natale et al, 2000).

To the best of our knowledge, this is the first time that the AUC has been used as a surrogate marker for temperature. Although it is good at showing the cumulative temperature dose over a period, it requires mathematical calculations. In addition, they do not readily provide the peak or frequency of temperature dose. However, when the AUC is compared with the maximum/peak temperatures and the proportion of time with temperatures >37.5°C there is a strong positive correlation, suggesting that the AUC is a good surrogate marker for thermal exposure.

We do recognize that our study has some limitations. First, because this was a retrospective study, it had an inherent bias. Therefore, the findings of this study need to be re-evaluated in a prospective observational clinical study. Furthermore, this was a single-center study, so we acknowledge that the results are more hypothesis-generating. A further limitation of this study was that the definition of severe TBI was based on clinical profile rather than radiological evidence of pathology, mechanism of injury, and GCS at the scene. Finally, we defined fever as a core body temperature >37.5°C, considering that the brain temperature may be up to 2.5°C higher.

This is an arbitrary line that can significantly affect the statistical analysis. However, with no consensus on the definition of fever, researchers will continue to try and choose the best cutoff point. Despite this limitation, our study has several strengths. One strength of this study was that our follow-up rate of >85% at 18 months was good, reducing attrition bias. Second, our neurological assessors were from a different team than the PICU team; therefore, they were less likely to be influenced by the management provided in the PICU. Finally, we believe that using AUC as a thermal exposure marker is a better way to describe the clinical journey.

Conclusion

In our study, hyperthermia in the first 72 hours after admission to a PICU was not associated with significantly worse neurological outcomes in children with TBI at the monthly follow-up. Although there were no differences in severe neurological deficits, this study did not answer whether hyperthermia affects children's cognitive outcomes. A larger multicenter randomized controlled trial examining the functional and cognitive outcomes of children with severe TBI is the next step toward understanding this area.

Authors' Contributions

E.C. and T.B. contributed to the design of the study, T.B. analyzed the results, and E.C. contributed to the writing of the article.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

No funding was received for this article.