Brain edema after traumatic brain injury (TBI) is an important component in the management and outcome of head trauma patients, but there are no easily accessible assessments of edema in routine clinical practice. Specific gravity (SG), which is the ratio of the density of a material to the density of water, is useful for identifying variations of water content in tissues, especially in brain parenchyma. Radiological attenuation is linearly correlated with SG in human tissue (1,2). This property allows the measurement, by computed tomography (CT), of the volume and weight of any tissue, thus allowing the computation of SG by dividing the weight by the volume. However, as this method provides only an indirect assessment of SG, this value will be named “estimated SG” (eSG). This methodology has been validated for lung and heart (3,4). We recently validated it for the measurement of the SG of different solutes and used it for the study of cerebral SG in human TBI (5). In this latter study, we observed a diffuse increase of eSG in noncontused hemispheric areas, unrelated to subarachnoid hemorrhage. However, this first study was performed in only 15 patients, and eSG determination was based on the analysis of a CT performed 3 days after TBI, thus leaving the value of eSG unknown at the very early phase of TBI. Furthermore, many aspects of its validity were questioned in the accompanying editorial (6). To confirm and better understand our previous observations, we performed a new study, in a much larger series of patients, of eSG in TBI restricted to the early phase. The aim of the present study was thus to correlate the initial brain eSG of the noncontused hemispheric areas to the clinical symptoms, the therapeutic intensity level, and the outcome in human TBI. In addition, the change in eSG occurring over time was also assessed in a subgroup of patients.
Three hundred ten patients hospitalized in our unit for TBI between January 1, 2001 and March 31, 2004 were the eligible study population. Patients were included in this study if 1) their TBI was severe enough to have required mechanical ventilation within the first 24 h after trauma; 2) they had a CT performed at our hospital within the first 24 h after trauma; and 3) this CT was performed before any surgical intervention. These criteria were fulfilled in the 120 patients in the study group. None of these patients was included in our first report (5). The study was approved by our local ethical committee (Comité de Protection des Personnes Pitié-Salpêtrière, Paris, France). All patients were treated according to algorithms routinely used in our unit, as already described (5). Clinical characteristics of the patients, events that occurred at the scene of the accident and during stay at the intensive care unit (ICU), and outcome were retrospectively extracted from an existing clinical database devoted to TBI patients. CTs were extracted from our central neuroradiological storage system.
Cerebral CT scans performed for headache assessment in 40 patients (30 men and 10 women) matched for age and intracranial volume with TBI patients were analyzed. Each examination had to be considered completely normal by the neuroradiologist (AZ) to be included in the control series.
Two groups of patients were defined according to the eSG of their noncontused hemispheric areas: the normal eSG group (NSG) having a eSG less than 1.96 sd above controls (i.e., eSG < 1.0355 g/mL), and the increased eSG group (ISG) defined as a eSG higher than 1.96 sd above controls (i.e., eSG > 1.0355 g/mL).
The following items or events measured or occurring at the accident scene were analyzed: mean Glasgow coma score (GCS), mechanism and kinetic of the accident, necessity to tracheally intubate and ventilate the patient's lungs, presence of a mydriasis, hypotension, hypoxia, use of mannitol or hypertonic saline before CT (combined in the terminology “osmotherapy”), occurrence of initial diabetes insipidus, and simplified acute physiological score (SAPS 2) at admission. The following items or events measured or occurring during the ICU stay were also recorded: ICP monitoring, duration of intracranial pressure (ICP) monitoring and/or cerebrospinal fluid (CSF) drainage, use of barbiturates and/or propofol and/or neuromuscular blockers and/or hypothermia as second step therapies, occurrence of systemic infection, occurrence of acute lung injury or acute respiratory distress syndrome, duration of ventilation, and length of stay in the ICU. Glasgow outcome scale (GOS; 1 for death and 5 for no sequel) was assessed at ICU discharge by the physician in charge of the patient and at 1 yr by the surgeons or the investigators during a long-term follow-up consultation.
Each CT was acquired as 5-mm thick contiguous slices with a high-speed advantage CT (GE medical system, Milwaukee, WI). Complete calibration of the CT was performed every month by CT manufacturer technicians as a routine procedure. An internal checking was also performed once a week to determine the absence of drift.
Follow-up of overall intracranial content eSG overtime was assessed in a subgroup of 15 patients who had additional cerebral CTs performed within 2 and 11 days, 12 and 20 days, and after day 21.
CTs were analyzed using the software package BrainView as recently described (5). In a first step, weight, volume, and eSG of the overall intracranial content were determined after automatic delineation. In a second step, only the noncontused hemispheric areas were considered. To do so, only normally appearing brain hemispheric parenchyma was delineated after exclusion of contusions and of intraventricular and subarachnoid CSF and hemorrhage. A similar procedure was performed in controls. Each CT scan was classified according to Traumatic Coma Data Bank classification (7) by the radiologist blinded to data of the quantitative CT.
Assuming an α risk of 0.05 (including the Bonferroni correction) and a β risk of 0.10, according to the values observed in our previous study (5), and assuming that 50% of TBI patients have an increased eSG, we calculated that at least 38 patients would be required in each subgroup (controls, NSG, and ISG) to detect a difference in eSG of at least 0.0020 g/mL (NQuery Advisor® 3.0, Statistical Solution, Cork, Ireland).
Normality of eSG distribution was assessed by a Kolmogorov–Smirnov test. Comparisons between TBI patients and controls and within TBI patients were performed using the Student's t-test for unpaired data and the Fisher's exact method. Comparisons of eSG according to the Marshall score and the initial GCS were performed by analysis of variance followed by Newman–Keuls post hoc test. Influence of osmotherapy on eSG in the different Marshall groups was assessed by a two-way analysis of variance for two grouping factors. Change overtime of eSG was performed by a one way analysis of variance for repeated measures.
χ2 and Student's t-test were used to compare patients according to eSG (NSG versus ISG) and outcome (GOS 4–5 versus GOS 1–3) at ICU discharge and 6 mo followed by multivariate stepwise logistic regression analyses where all the variables suggested by the univariate analysis (P < 0.10) were entered into the model. Calibration and discrimination of the logistic models were assessed using the Hosmer–Lemeshow statistics and the ROC curves, respectively.
The results are expressed as mean ± sd. All tests were two-sided, and P values <0.05 were considered significant. The statistical analyses were performed using JMP IN 5.1 statistical software (SAS Institute, NC).
The mean value of the GCS at the scene of accident was 7 ± 3 (median = 7). The mechanism of injury was a motor vehicle accident in 60 cases, a fall in 33, a pedestrian accident in 16, and an assault in 11. CT was performed 5 ± 6 h after TBI. The 40 control subjects were matched for age (38 ± 12 vs 35 ± 15 yr in the TBI group) and intracerebral volume (1350 ± 142 vs 1343 ± 136 mL) with the TBI patients.
eSG of the overall intracranial content and of the noncontused hemispheric areas was significantly higher in TBI patients than in controls (Table 1). eSG in patients with TBI followed a normal distribution (Fig. 1). eSG of the overall intracranial content was highly correlated to the one the noncontused hemispheric areas (Fig. 2) with a mean difference of 0.0002 g/mL and a mean bias of 0.0011 g/mL.
eSG was significantly higher in patients classified as having a Marshall score of 3 or 4 than in patients having a Marshall score of 1 or 2 or a mass lesion (Fig. 3). These differences were independent of the administration of mannitol or hypertonic saline given at the scene or during transportation (Fig. 4). The distribution of the Marshall score was also different among the NSG and ISG patients, the largest proportion of patients having a score of 3 being in the ISG group (Table 2). eSG was significantly higher in patients with an initial GCS of 3 or 4 (Fig. 3).
Eighty-three patients (69%) were classified in the NSG group, and 37 patients (31%) in the ISG group. ISG patients were significantly older, had a lower GCS and higher SAPS 2 than NSG patients (Table 3). The occurrence of mydriasis at the scene and the use of osmotherapy before CT were significantly more frequent in the ISG group. Osmotherapy was given before CT because of mydriasis in 27 patients and because of transcranial Doppler abnormality (pulsatility index > 1.2) in 16 patients. Eleven patients with mydriasis did not receive osmotherapy. The incidence of tracheal intubation, hypoxemia, hypotension, and hypothermia as well as the mechanism and the velocity of the accident were similar between the two groups. Only the initial SAPS 2 (odds ratio: 1.08 per point increase; 95% CI: 1.02–1.14), the presence of mydriasis on scene (odds ratio: 3.4; 95% CI: 1.03–11.3), and the use of osmotherapy (odds ratio: 4.6; 95% CI: 1.4–15.0) remained significant predictors of eSG after multivariate analysis. Calibration and discrimination of the model were appropriate as shown by the Hosmer–Lemeshow statistics (7.67, P = 0.36) and the area under the ROC curve (0.87; 95% CI: 0.80–0.94).
The incidence of extraventricular drain insertion (and/or Codman® intraparenchymal catheter) for ICP monitoring and CSF drainage and the use of barbiturates for uncontrollable ICP were significantly more frequent in the ISG group (Table 4). At the opposite, all other variables tested regarding the therapeutic intensity level in the ICU were similar between the two groups.
A low initial GCS, a high initial SAPS 2, a low velocity injury, the presence of mydriasis at the scene, the use of osmotherapy, and a high eSG of the overall intracranial content or of the noncontused areas were predictors of a poor outcome at ICU discharge in monovariate analysis (Table 5). Similar results were observed at 1 yr with the exception of eSG that was not significant in the monovariate analysis. Only the SAPS 2 (odds ratio: 1.1 per point increase; 95% CI: 1.05–1.17 and 1.09; 95% CI: 1.04–1.14 at ICU discharge and 1 yr, respectively) and the use of osmotherapy (odds ratio: 4.1; 95% CI: 1.5–11.4 and 3.3; 95% CI: 1.2–9.0 at ICU discharge and 1 yr, respectively) remained significant predictors of outcome in multivariate analysis. GOS was significantly different in NSG and ISG patients at ICU discharge, but not at 1 yr (Table 6). eSG decreased with time. Among the 15 patients we followed, none had an increased eSG after day 21 (Fig. 5).
This study confirms our first observation of brain eSG in human TBI using the same CT methodology (5) but in a larger group of patients and at a different time-window after trauma. In addition to this initial observation, we classified patients into two groups according to their initial eSG. Patients with an increased eSG had a lower GCS and more often mydriasis at the scene of the accident, more frequently received osmotherapy at the initial phase, had an extraventricular drain more frequently implanted for ICP monitoring and CSF drainage, more frequently received barbiturates as a second-line therapy, and more frequently had a CT classified in the third category of the Marshall score.
These results raise two questions: Are they valid or could they be due to a bias or an artifact? What could be the potential implications for our understanding of the pathophysiology of brain edema in human TBI? Regarding the first question, our technique obviously provides an indirect assessment of SG and has been validated only ex vivo (5,8) although some previous in vivo studies have shown good correlations, in different clinical conditions, between radiological attenuation and SG (9–11). Even though the method by itself might be slightly inaccurate, because of the interposition of the skull for example, our study is based on comparisons between TBI patients and controls and among TBI patients, thus controlling this potential bias. The question of the influence of osmotherapy on our results is a concern, because the ISG group received osmotherapy more frequently than did the NSG group, and osmotherapy was an independent predictor of increased eSG in the multivariate analysis. Osmotherapy, by decreasing the brain water content, increases eSG and its use alone could theoretically explain our results. However, the main argument against this hypothesis is that eSG was increased in the Marshall 3 patients who did and who did not receive osmotherapy compared with Marshall 2 patients. In addition, one study suggests that osmotherapy affects eSG by 0.03% (12), whereas the difference observed in the present study between TBI patients and controls was in the order of 0.3%, a 10-fold larger increase. It is thus very likely that brain edema, linked to an increased eSG, was the cause for the occurrence of mydriasis and/or osmotherapy administration, and not the opposite. Could the increase in eSG that we observed be solely due to hyperemia? This is unlikely because hyperemia has been eliminated as the cause of early edema in TBI (13). Furthermore, the increase in eSG observed in our patients was a long-lasting phenomenon that seems incompatible with hyperemia.
Barzo et al. (14) demonstrated, in an experimental model of diffuse injury, that TBI is associated with a rapid and transient blood–brain barrier (BBB) opening that begins at the time of the trauma and lasts no more than 30 min. A similar time window was evidenced by Baldwin et al. (15). Human autopsy data have shown that a complete breakdown of the BBB with hemorrhage and extravasation of all types of plasma proteins occurs immediately after trauma (16). Similar observations have been made on cortical biopsies of TBI patients showing perivascular hemorrhages and extravasation of proteinaceous edema fluid (17). After this first opening, the BBB could seal within the first hour after trauma (18). It might then remain closed in the noncontused areas. This is in agreement with the observations of Lang et al. (19) and Baldwin et al. (15). After this rapid closure, proteins and cells might be trapped extracellularly, leading to an increase in brain eSG. A healing process might then be activated, allowing a progressive normalization of SG.
Our observation of an increased eSG after TBI is in sharp contrast with most of the experimental literature. A decrease in brain SG together with an increase in brain water content has been consistently observed in numerous studies using experimental models of brain injury, such as a cryogenic lesion (20,21) or controlled cortical impact injury (22,23). At this stage, only hypotheses can be made to explain these striking differences. For a similar increase in brain water content, brain SG might decrease in a case of predominant cellular edema, whereas vasogenic edema might have variable consequences on SG. A partial disruption of the BBB with a leak of water, electrolytes, and proteins should decrease SG, as the added volume has a SG [between 1.0245 and 1.0285 g/mL (24)] less than the hemispheres. Conversely, a major disruption of the BBB with a leak of water, electrolytes, proteins, and blood cells should increase the SG, as the added volume has a SG more than the hemispheres. This could explain the increase in hemispheric eSG observed in one-third of our patients. Indeed, a similar increase in SG has been described by Bullock et al. (10). These authors measured SG of the white matter in 39 patients early after TBI using density gradient columns. Twelve of these patients (31%) presented an increased SG of the white matter together with an increased radiological attenuation in the sampling area on CT.
Interestingly, neither the mechanism nor the kinetics of the TBI was different between the two groups. Although assessing the intensity of TBI a posteriori is difficult, this raises the possibility of an individual factor in the intensity of the eSG change, and likelihood of an initial BBB leakage after TBI in humans. These interindividual differences, well known in clinical practice, might depend on biochemical or genetic differences among patients. Another more mechanistic explanation would be that the initial BBB opening could result solely from the initial surge in arterial blood pressure induced by TBI. This would explain the differences among experimental models where brain injury is obtained in animals under general anesthesia, thus likely limiting this arterial blood pressure surge.
The initial GCS and SAPS 2, the velocity of the injury, the occurrence of mydriasis at the scene, and the use of osmotherapy were predictors of outcome at ICU discharge and at 1 yr. Only the SAPS 2 and the use of osmotherapy remained significant predictors in multivariate analysis. eSG of the overall intracranial content or of the noncontused areas was also a predictor of outcome, but only at ICU discharge and in the monovariate analysis. This indicates that long-term outcome might be independent of the intensity of the initial brain edema, providing that efficient therapies to treat increased ICP are applied rapidly (25). In this respect, it must be pointed out that the prevalence of increased eSG that we observed in this study could be underestimated by the fact that some of these patients with an increased eSG might have died at the scene of the accident because of brain herniation.
The clinical usefulness of the automatic determination of eSG of the overall intracranial content in human TBI will have to be addressed in a large prospective study.
The authors thank the nurses of the Neurosurgical Intensive Care Unit and the technicians of the Department of Neuroradiology for their active participation.
1. Phelps ME, Gado MH, Hoffman EJ. Correlation of effective atomic number and electron density with attenuation coefficients measured with polychromatic X-rays. Radiology 1975;117:585–8.
2. Mull RT. Mass estimates by computed tomography: physical density from CT numbers. Am J Radiol 1984;143:1101–4.
3. Malbouisson LM, Muller JC, Constantin JM, et al. Computed tomography assessment of positive end-expiratory pressure-induced alveolar recruitment in patients with acute respiratory distress syndrome. Am J Respir Crit Care Med 2001;163:1444–50.
4. Malbouisson LM, Busch CJ, Puybasset L, et al; CT Scan ARDS Study Group. Role of the heart in the loss of aeration characterizing lower lobes in acute respiratory distress syndrome. Am J Respir Crit Care Med 2000;161:2005–12.
5. Lescot T, Bonnet MP, Zouaoui A, et al. A quantitative computed tomography assessment of brain weight, volume, and specific gravity in severe head trauma. Intensive Care Med 2005;31: 1042–50.
6. Stocchetti N. Does the brain become heavier or lighter after trauma? The long story of brain water content and its direct or indirect measurement. Intensive Care Med 2005;31:1009–11.
7. Marshall LF, Marshall SB, Klauber MR, et al. The diagnosis of head injury requires a classification based on computed axial tomography. J Neurotrauma 1992;9 (Suppl 1):S287–92.
8. Unger E, Littlefield J, Gado M. Water content and water structure in CT and MR signal changes: possible influence in detection of early stroke. Am J Neuroradiol 1988;9:687–91.
9. Rieth KG, Fujiwara K, Di Chiro G, et al. Serial measurements of CT attenuation and specific gravity in experimental cerebral edema. Radiology 1980;135:343–8.
10. Bullock R, Smith R, Favier J, et al. Brain specific gravity and CT scan density measurements after human head injury. J Neurosurg 1985;63:64–8.
11. Takagi H, Shapiro K, Marmarou A, Wisoff H. Microgravimetric analysis of human brain tissue: correlation with computerized tomography scanning. J Neurosurg 1981;54:797–801.
12. Lescot T, Bonnet MP, Fetita C, et al. Effects of hypertonic saline solution (HSS) on cerebral contusion (CC) after traumatic brain injury: a tomodensitometric study [abstract]. J Neurosurg Anesthesiol 2003;15:A375.
13. Marmarou A, Fatouros PP, Barzo P, et al. Contribution of edema and cerebral blood volume to traumatic brain swelling in head-injured patients. J Neurosurg 2000;93:183–93.
14. Barzo P, Marmarou A, Fatouros P, et al. Magnetic resonance imaging-monitored acute blood-brain barrier changes in experimental traumatic brain injury. J Neurosurg 1996;85:1113–21.
15. Baldwin SA, Fugaccia I, Brown DR, et al. Blood-brain barrier breach following cortical contusion in the rat. J Neurosurg 1996;85:476–81.
16. Liu HM, Sturner WQ. Extravasation of plasma proteins in brain trauma. Forensic Sci Int 1988;38:285–95.
17. Castejon OJ, Castejon HV, Diaz M, Castellano A. Consecutive light microscopy, scanning-transmission electron microscopy and transmission electron microscopy of traumatic human brain oedema and ischaemic brain damage. Histol Histopathol 2001;16:1117–34.
18. van den Brink WA, Santos BO, Marmarou A, Avezaat CJ. Quantitative analysis of blood-brain barrier damage in two models of experimental head injury in the rat. Acta Neurochir Suppl (Wien) 1994;60:456–8.
19. Lang DA, Hadley DM, Teasdale GM, et al. Gadolinium DTPA enhanced magnetic resonance imaging in acute head injury. Acta Neurochir (Wien) 1991;109:5–11.
20. Ikeda Y, Toda S, Wang M, Nakazawa S. Early changes of blood-brain barrier and superoxide scavenging activity in rat cryogenic brain injury. Acta Neurochir Suppl (Wien) 1994;60:136–8.
21. Loberg EM, Torvik A. Uptake of plasma proteins into damaged neurons. An experimental study on cryogenic lesions in rats. Acta Neuropathol (Berl) 1991;81:479–85.
22. Esen F, Erdem T, Aktan D, et al. Effects of magnesium administration on brain edema and blood-brain barrier breakdown after experimental traumatic brain injury in rats. J Neurosurg Anesthesiol 2003;15:119–25.
23. Shapira Y, Shohami E, Sidi A, et al. Experimental closed head injury in rats: mechanical, pathophysiologic, and neurologic properties. Crit Care Med 1988;16:258–65.
24. Blacque-Belair A, Mathieu de Fossey B, Fourestier M. Dictionnaire des constantes biologiques et physiques, 4th edn. Paris: Maloine, 1965.
25. Aldrich EF, Eisenberg HM, Saydjari C, et al. Diffuse brain swelling in severely head-injured children. A report from the NIH Traumatic Coma Data Bank. J Neurosurg 1992;76:450–4.