Theoretical models suggest that small differences only exist between brain and body temperature in health. Once the brain is injured, brain temperature is generally regarded to rise above body temperature. However, since reports of the magnitude of the temperature gradient between brain and body vary, it is still not clear whether conventional body temperature monitoring accurately predicts brain temperature at all times. In this prospective, descriptive study, 20 adults with severe primary brain trauma were studied during their stay in the neurointensive care unit. Brain temperature ranged from 33.4 to 39.9 °C. Comparisons between paired brain and rectal temperature measurements revealed no evidence of a systematic difference [mean difference −0.04 °C (range −0.13 to 0.05 °C, 95% CI), p = 0.39]. Contrary to popular belief, brain temperature did not exceed systemic temperature in this relatively homogeneous patient series. The mean values masked inconsistent and unpredictable individual brain–rectal temperature differences (range 1.8 to −2.9 °C) and reversal of the brain-body temperature gradient occurred in some patients. Brain temperature could not be predicted from body temperature at all times.
Brain temperature is seldom measured during routine neurointensive care but there is a long-held assumption that since the temperatures of healthy internal organs differ only slightly , temperature measurements of the rectum, bladder, pulmonary artery and tympanum can be used as surrogates for brain temperature. Using these conventional body sites, temperatures greater than 38.5 °C are commonly encountered during neurocritical care [2, 3] and cause concern. In animal models of cerebral ischaemia [4–7] and trauma [8, 9] a rise in body core temperature in excess of 38 °C is associated with increased neuronal damage. In stroke patients a rise in body temperature independently predicts poor outcome  and increased mortality [11–14] but when the human brain is injured by trauma, the evidence for a relationship between raised body temperature and worse neurological outcome is not as clear [15, 16]. However, it is assumed that the deleterious metabolic, inflammatory and biochemical mechanisms associated with raised body temperature in animal models of stroke and trauma may operate similarly in the brain injured human . This assumption underpins current opinion that even a small increase in body temperature (1–2 °C) above normal (37 °C) accelerates ischaemic damage and increases the size of the primary brain lesion [9, 18] and should therefore be prevented .
Little is known about human brain temperature, but theoretical models suggest that during normothermia the temperature of the brain is close to that of internal carotid arterial blood before the blood enters the circle of Willis . In man, direct measurement of brain temperature has been undertaken predominantly in neurosurgical patients where study populations have included a mixture of cases (brain trauma, subarachnoid haemorrhage, haemorrhagic stroke or brain tumour), with measurements made at different intracranial sites (intraventricular, intraparenchymal or subdural) using different sensor brands. A review of this mixed evidence suggests that following acute neurological disease and injury, the human brain is at a higher temperature than body tissues . However, variations in the patients' thermal state appear to have an impact on the brain-body temperature gradient; the greatest difference occurring during hypothermia with average brain temperature over 1 °C lower than core body temperature . These findings are consistent with models of heat transfer from brain to body core during hypothermia .
Information regarding the magnitude of the temperature gradient between brain and body appears to vary in different studies and may be related to the heterogeneity of the neurosurgical patients recruited. To help clarify whether conventional temperature monitoring accurately predicts brain temperature during neurointensive care, the aim of this study was to determine whether clinically important temperature differences exist between brain and body core tissues.
The study was approved by the Local Research Ethics Committee. Patients were admitted as direct referrals from the Emergency Department of our institution or as tertiary referrals from Emergency Departments of other hospitals within the Greater Manchester Region of the North-west of England. At this institution over 200 head-injured patients are admitted to our ICU annually. In this study, consecutive adult patients (≥ 16 years of age) with acute primary brain injury (with or without systemic injury) requiring acute medical management for their injuries were eligible for inclusion in the study. Patients requiring emergency neurosurgery and pregnant females were excluded. On arrival to ICU, patients were sedated, intubated and mechanically ventilated and admitted primarily for intracranial pressure monitoring and medical management. They were treated in accordance with local neurointensive care guidelines for maintaining cerebral perfusion pressure (CPP) at 60 mmHg or above and intracranial pressure (ICP) below 20 mmHg. Management of raised ICP was by use of appropriate positioning (30° head up), sedation, analgesia and neuromuscular blockade, and osmotherapy with mannitol (0.5 g.kg−1). Barbiturate coma was used in the event of an ICP refractory to the standard treatments. Informed assent from the next of kin was obtained for each patient prior to recruitment.
Since all patients in this study required ICP monitoring as part of routine medical management, the use of the Camino 110–4BT, 4 Fr, fibre-optic, transducer tipped sensor (Integra Neurosciences, Andover, UK) provided the additional temperature parameter required for the study. Thus in all cases, brain parenchymal temperature (Tbrain) was measured simultaneously with ICP using the single, multiparameter sensor. The sensor was inserted at the bedside under aseptic conditions, via a standard frontal burr hole through the dura, approximately 3–4 cm into brain parenchyma. The scalp incision was closed with a stitch. Measurements continued until removal of the sensor.
Brain temperature measurements were displayed continuously and recorded directly from the Camino MPM-1 monitor. Rectal temperature was measured using a Mon-a-therm 400 series, 9-Fr, thermistor (Mallinckrodt Medical, Tyco Healthcare, Gosport, UK) inserted into the rectum, 10 cm from the anus and displayed continuously via the bedside monitoring system (Marquette Electronics, Milwaukee, WI, USA). Simultaneous brain and rectal measurements were documented at 60-min intervals throughout the study.
In the event of a rise in rectal temperature, a series of conventional, routine treatments, following a ‘step-up’, four level protocol, was used to achieve a target rectal temperature, nominally 36.5 °C. Briefly, the step-up protocol for body cooling at our institution incorporates
• Level 1: administration of paracetamol (acetaminophen) 1 g (qds);
• Level 2: whole body surface cooling with ‘hand warm’ wet sheets applied to the trunk;
• Level 3: whole body surface cooling with ‘hand warm’ wet sheets to trunk plus neuromuscular blockade;
• Level 4: and intragastric iced water lavage.
At the end of the monitoring period, intracranial sensors are usually discarded. However, for the purposes of this study, catheters were retained whenever practical. Seven catheters were retained and the thermistors checked at 1 °C intervals (using a water bath temperature of 33°−40 °C) against a mercury-in-glass thermometer. In each case the Camino 110 4BT sensor (thermistor) was within the manufacturer's reported accuracy (± 0.3 °C) for the range of brain temperatures recorded in the study.
The manufacturer's reported accuracy of the YSI 400 series general purpose thermistor is ± 0.2 °C. Ten probes were selected at random from a batch of 100 and placed into a water bath at temperatures between 33° and 40 °C. The change in water temperature was measured using a mercury-in-glass thermometer and a single thermistor, and repeated for each of the 10 thermistors. Each thermistor agreed with the mercury thermometer to within ± 0.1 °C.
Injury severity for each patient was scored using the Abbreviated Injury Scale (AIS), 1990 Revision . The AIS is a consensus-derived system for classifying individual injuries by body region (nine regions in total) on a 6-point ordinal severity scale ranging from AIS 1 (minor) to AIS 6 (currently untreatable). With respect to the head (brain and cranium), AIS 3 is a critical head injury and AIS 5 is the most severe of life-threatening injuries. For overall severity of all injuries (head plus all other), the Injury Severity Scale (ISS)  was used, which is the sum of the squares of the highest AIS in three different body regions. By convention, ISS 1–8 is a minor injury, ISS 9–15 moderate injury and ISS = 16 a severe injury.
This study was designed to have a 90% power at the 5% significance level to identify the scale of difference between brain and rectum of 0.8 °C as reported by Rumana et al. . The number of patients required (n = 19) was rounded to 20 patients. Repeated measures analysis of variance (SPSS version 10, SPSS, Chicago, IL, USA) was used to assess differences between brain and rectal temperatures. Readings for patients were carried out at varying times after injury and for varying durations. Only paired values of brain and rectal temperature readings were used to calculate a mean difference, 95% confidence interval (mean, 95% CI) for each patient.
An overall mean was calculated using a random effects meta-analysis (STATA v.7, Stata Corp., College Station, TX, USA). The ‘random effects’ approach recognises that the individual patient means are themselves variable and allows for this additional variation with wider confidence limits on the overall average.
Twenty patients (18 male) aged 19–70 (median 37.5) years were studied for 17–259 h (median 118 h). Eight patients died during their ICU stay. Patients were admitted within 1–24 (median 7.5) h of injury. The baseline of the Glasgow coma scale (GCS) ranged from 3 to 13. Thirteen patients had an isolated head injury, with an AIS for the head between 3 and 5 (median 5). Six of the seven patients with head plus systemic injury had an AIS (head) of 5. ISS in patients with head injury only ranged from 10 to 29 (median 26). Patients with head and systemic trauma (n = 7) had the highest ISS, range 14–45 (median 35). At the time of recruitment, patients did not require emergency neurosurgery but in five, emergency craniotomy and evacuation of haematoma was performed as part of subsequent neurointensive care. In 10 of the 20 patients a second CT scan was performed after insertion of the ICP/temperature sensor. In these 10 patients, the position of the sensor tip could be confirmed radiologically and was shown to be between 2 and 4 cm into brain tissue (within the white matter). There was no radiological evidence that the sensor tip was within the lesion. There were no reported complications attributable to sensor placement.
The rectal temperature of all the 20 patients ranged from 33.5° to 39.9 °C, median 37.2 °C (interquartile range, IQR 36.5–37.7 °C). The brain temperature of all 20 patients ranged from 33.4° to 39.9 °C, median 37.1 °C, interquartile range 36.3–37.7 °C. Therefore, 25% of brain temperature values were in excess of 37.7 °C despite active body cooling (Fig. 1).
Calculation of the mean difference between Tbrain and Trectum pairs for all 20 patients is shown in Fig. 2. In this figure the mean values, with the 95% CI, are given for each patient. In 19 patients the mean difference was not more than 0.4 °C in either direction.
The exception is an outlying case where a mean difference between brain and rectal temperature of −1.4 °C. is evident from Fig. 2. With the exception of this outlier, the mean differences were not large and the overall difference (average for the group, including the outlying patient data), was even smaller, −0.09 °C (−0.182 to 0.001 °C, 95% CI) p = 0.053). However, this overall result may have been influenced unduly by the data from the single, outlying case. This patient had a wide temperature gradient (low Tbrain relative to Trectum) from admission to the time of death, 44 h after injury. To check whether inclusion of the data from this patient in the analyses influenced the overall average difference between brain and rectal temperature we re-analysed the data after excluding the outlying case. The mean difference for the remaining 19 patients was −0.04 °C (−0.13 to 0.05 °C, 95% CI), p = 0.39. This sensitivity analysis confirmed that inclusion of the outlier made no substantive difference to our conclusions. Unpredictable and inconsistent differences between brain and rectum do exist, however, in almost all individuals (range 1.8 to −2.9 °C, n = 1638 pairs). An example of this is shown in Fig. 3. In this fatal case, differences between brain and rectal temperature were negligible throughout the first 130 h of monitoring. However, shortly after this time, the small gradient between the two temperatures became dissociated such that brain temperature fell to a maximum of 3 °C below rectal temperature. This rapid dissociation in temperature was accompanied by a concomitant reversal in the pattern of ICP and CPP.
We have studied a discrete group of patients with severe brain injury during neurointensive care and found no evidence of a systematic difference between brain and rectal temperature. The average mean differences in 19 of the 20 patients were not more than 0.4 °C, suggesting that in our patient group, brain temperature on average did not exceed systemic temperature. However, these results mask the presence of brain–rectal temperature differences ranging from 1.8 to −2.9 °C during neurointensive care which appeared to be inconsistent and unpredictable. The results suggest that during neurointensive care, brain temperature at any moment can not necessarily be predicted from rectal (systemic) temperature.
Brain and rectal temperatures have been compared in a number of studies , although most have involved mixed groups of neurointensive care patients. We have identified only one previous study, by Rumana and colleagues , where recruitment was exclusively patients with severe head injury and, like our own group, where brain parenchyma and rectal temperature monitoring was conducted in all patients. Our present study is broadly similar to the Rumana et al.'s study, but our results appear to be contradictory. Rumana et al.'s study suggested that brain temperature was an average of 1 °C higher than rectal (and jugular bulb) temperature in head-injured patients treated during neurointensive care. We are unable to assess whether the severity of head injuries is comparable because Rumana et al. report only the Glasgow Coma Scale and do not include the AIS for the head. It is therefore not possible to assess the severity of brain injury between the two studies, which could be of fundamental importance. In patients with severe brain damage as evidenced by a low GCS (the extent of systemic injury is often unclear), brain temperature is typically reported to be higher than other body sites . On the other hand, there is also some evidence to suggest that patients with a consistently higher brain temperature have a better outcome, implying that higher brain temperatures are seen in patients with less severe brain injury [26, 27].
It is possible, therefore, that the differences between our own study and that of Rumana's group relate to differences in severity of the primary brain injury. Rumana et al.'s results, like our own, suggest that temperature differences between brain and rectum are unpredictable. Unable to identify any clinical finding predictive of brain to systemic temperature differences, Rumana et al. noted a significant narrowing of the temperature gradient between brain and rectum in patients with low (< 20 mmHg) CPP who subsequently died. We also noted changes in the temperature gradient in fatal cases, but with a widening, rather than a narrowing, of differences, as illustrated in Fig. 3. Here the rapid fall in brain temperature below rectal temperature was associated with a catastrophic increase in ICP (maximum 130 mmHg) and a concomitant fall in CPP. CT angiography in this patient confirmed cessation of cerebral blood flow. Our results support recent reports [27–30] that a fall in brain temperature relative to body temperature after brain damage is associated with poor cerebral blood flow and outcome.
One potential source of error in these types of studies in critical care is the validity of the measurement techniques employed. In the UK a variety of different conventional ‘body’ sites are used for routine temperature monitoring , some of which are known to be unreliable . It could be argued that changes in temperature within the rectum ‘lag’ changes at other core sites but in this study, as in previous publications by our group , this seldom occurred. In a small number of patients in whom simultaneous pulmonary artery blood and rectal temperature recordings were available, there was no rectal temperature lag. Indeed, Rumana et al. report little difference between rectal and core temperature (measured at the jugular bulb). This is in agreement with LeFrant et al. , adding further support for the use of the rectum as a site for core temperature measurement. To avoid measurement error the rectal probe was inserted 10 cm into the anus. Like many others who have used the rectum to estimate body core temperature, we cannot guarantee the position of the sensor tip at all times, but if probe displacement had occurred, the consequence of measurement error would be a fall in rectal temperature such that brain temperature would be increased relative to rectal temperature.
With the published information available to us from previous studies, it is difficult to understand why we did not observe a systematic difference between brain and systemic temperature showing that brain temperature exceeds body temperature. We have considered the issue of thermistor accuracy (±0.3 °C) as well as the question of the validity of single site measurements to represent global brain temperature, but these issues are common to clinical monitoring and are not unique to this study. Whether practical problems such as sensor positioning within brain or rectum may account for contradictory findings should be considered, as should the impact of the severity of the primary lesion and ensuing pathophysiological events during the monitoring period. Certainly it would seem that a fall in CPP to 20 mmHg has a marked effect on the temperature differences between the two sites; by a significant narrowing as shown by Rumana's group  or by reversal of the gradient as we have indicated (see Fig. 3). These results support the case for continued investigation of brain temperature monitoring and further efforts to determine the potential pathophysiological significance of the variations in brain–systemic temperature differences during neurointensive care.
We have shown in neurointensive care patients with severe traumatic brain injury that there is no systematic difference between brain and rectal (systemic) temperature during routine critical care. However, in our study individual mean brain temperature values mask large, clinically important, and unpredictable discordance between temperatures at the two sites. The potential pathophysiological significance of these variations is uncertain, but can be associated with important events such as inadequate cerebral blood flow. From the results of this study we would advise against yielding to the temptation to apply ‘correction’ factors to rectal temperature readings as a method for estimating brain temperature as this could lead to inappropriate interpretation of brain temperature as well as inappropriate treatment.
The authors thank Dr N. Hoggard, Consultant Neuroradiologist, for reporting the CT scans; T. Rainey, University of Manchester, for technical assistance, and M. Ogden, Medical Physics Department, for assistance with data extraction. This work was supported by the Research and Development Department and the Department of Intensive Care Medicine, Hope Hospital, and by the Brain and Spinal Injury Charity (BASIC).