Introduction

Traumatic brain injury (TBI) affects 2 million people per year and is a leading cause of death and disability among young people in the United States. Clinical and laboratory studies demonstrate that poor outcome is associated with ongoing (secondary) injury that evolves over time after the initial (primary) injury. Intensive care management of TBI patients therefore involves monitoring to identify secondary injury. Intracranial pressure (ICP) is the physiological parameter most frequently monitored. Other physiological parameters may be monitored continuously at the bedside, including partial pressure of brain tissue oxygen (PbtO2), cerebral metabolites (via microdialysis catheters), and regional cerebral blood flow among others. [15].

The most recent Guidelines for Severe Traumatic Brain Injury include the use of PbtO2 monitors but do not provide guidance about how PbtO2 should be managed although a threshold for treatment is suggested [6, 7]. Reduced PbtO2 is common after severe TBI [8] and may occur despite a normal ICP and CPP [9]. Furthermore, there is an association between brain hypoxia and both poor outcome and mortality after severe TBI [10]. We therefore sought to identify the medical therapies used to treat reduced PbtO2 and to assess their efficacy.

Methods

Patient Population

Subjects were identified retrospectively from a prospective observational database (Brain Oxygen Monitoring Outcome study) of patients with severe TBI treated in the NeuroIntensive Care Unit at Hospital of the University of Pennsylvania (HUP), an academic Level 1 Trauma Center. We selected patients treated between November 2001 and September 2004, a time period when we still were evaluating different treatments for compromised PbtO2. All patients treated during the study period were evaluated for inclusion in the study. During that period, we initiated treatment when PbtO2 was <25 mmHg, whereas our current treatment threshold is PbtO2 <20 mmHg. Inclusion criteria were as follows: admission within 3 h of injury; an admission Glasgow Coma Score (GCS) ≤8; and an admission injury severity score (ISS [11]) ≥16. Patients with cranial gunshot wounds or other penetrating cranial injuries were excluded from analysis. While survival data were previously published for this cohort, the relationship between medical interventions to treat brain oxygen, the response rate, and survival has not been examined [7]. The Institutional Review Board approved this study.

Intracranial and Physiological Monitoring

Patients were cared for in the NeuroIntensive Care Unit and were monitored continuously using ICP, PbtO2, and brain tissue temperature using commercially available devices (LICOX CMP Triple Lumen Monitoring System, Integra NeuroSciences, Plainsboro, NJ). Probes were inserted at the bedside through a burr-hole into the frontal lobe and secured with a triple lumen bolt. The ICP and PbtO2 monitors were placed into white matter that appeared normal on admission head CT and on the side of maximal pathology or swelling. To allow for probe equilibration, data from the first 3 h were discarded. All patients were monitored for at least 72 h unless care was withdrawn first. The monitors were removed when ICP had been normal for 24 h without specific treatment (except sedation for ventilator care) or when the patient was able to consistently follow commands.

Each patient had an indwelling arterial (usually radial artery) catheter and mean arterial pressure (MAP) was recorded continuously in all patients. Cerebral perfusion pressure (CPP) was calculated from measured parameters (CPP = MAP − ICP). Heart rate and arterial oxygen saturation (SpO2) were recorded in all patients.

General Management

All patients were resuscitated according to Advanced Trauma Life Support (ATLS) guidelines and were managed according to a local algorithm based on the Brain Trauma Foundation Guidelines for Severe TBI and published recommendations for severe TBI and ICU care [25, 1216]. This included: (1) early identification and evacuation of traumatic hematomas, (2) intubation and ventilation with low-volume pressure-limited ventilation to maintain PaO2 between 80 and 100 mmHg, and SaO2 >93%, and PaCO2 between 35 and 40 mmHg, (3) sedation using propofol during the first 24 h followed by sedation (lorazepam) and analgesia using morphine or fentanyl, (4) bedrest with head elevation initially of ≥30°, (5) normothermia ~35–37°C, (6) euvolemia, (7) anticonvulsant prophylaxis with phenytoin for 1 week (longer if seizures occurred), and (8) packed red blood cell transfusion if Hgb was <10.

ICP and CPP Management

Therapies were instituted to maintain ICP <20 mmHg and CPP >60 mmHg, according to our local protocol (Fig. 1). Elevated ICP >20 mmHg for >2 min was initially treated with adjustment of head position, sedation, and analgesia. If ICP remained >20 mmHg for more than 10 min despite these initial measures osmotherapy was administered using repeated boluses of mannitol (0.5–1 g/kg, 25% solution). After adequate fluid resuscitation phenylephrine (10–300 mcg/min) was administered when CPP ≤60 mmHg for greater than 15 min. Thereafter optimized hyperventilation PaCO2 ~30 mmHg, ventriculostomy placement, additional propofol or a decompressive hemicraniectomy (DCH) were used as second-tier therapies for refractory intracranial hypertension (>20 mmHg for >15 min continuously despite therapy). Induced hypothermia and hypertonic saline were not used to manage ICP in the patients included in this study.

Fig. 1
figure 1

ICP management algorithm

PbtO2 Management

Patients received cause-directed therapy to maintain PbtO2 ≥25 mmHg [17]. When PbtO2 was <25 mmHg and there was intracranial hypertension (ICP >20 mmHg), measures were taken to reduce ICP as described above. An oxygen challenge (100% FiO2 for 2 min) was used as a temporary measure to restore PbtO2 and to verify probe function. If ICP <20 mmHg or ICP reduction did not increase PbtO2, then CPP was increased (usually with phenylephrine). If the cause of low PbtO2 was thought to be due to lung dysfunction (based on assessment of arterial blood gas, chest X-ray, and ventilator requirements) then pulmonary function was optimized (e.g., by increased FiO2 and/or positive end-expiratory pressure, tracheal suction or pulmonary toilette). If excess metabolic demand was suspected (e.g., pain, agitation, fever, or seizures) then analgesic, sedative, or antiepileptic medications were administered. If these measures failed and hemoglobin was <10 mg/dl, then a blood transfusion was administered.

Data Collection, Analysis, and Definitions

Clinical and radiological variables of this study included age, admission (post-resuscitation) Glasgow Coma Scale (GCS), mode of injury, brain oxygen (PbtO2), intracranial pressure, brain temperature, core body temperature, blood pressure, heart rate, inspired fraction of oxygen (FiO2), serum, hemoglobin level, serum sodium, serum glucose. All physiologic and intracranial variables were continuously recorded using a bedside critical care monitor system (Component Monitoring System M1046-9090C, Hewlett Packard, Andover, MA) and were recorded in the ICU flow-sheet usually every 15 min and at least every 30 min. Marshall and Rotterdam CT classifications were determined by consensus of two neurosurgeons (LM and LEB) who evaluated the images independently and who were blinded to patient clinical information at the time of their assessment.

Increased ICP was defined as ICP >20 mmHg for >2 min. Refractory intracranial hypertension was defined as ICP >20 mmHg for >15 min in a 1-h period despite therapy. Compromised brain oxygen was defined as PbtO2 <25 mmHg for >10 min. Brain hypoxia was defined as PbtO2 <15 mmHg for >10 min. A successful response to PbtO2 therapy was defined as a return of PbtO2 to normal value (>25 mmHg) sustained over >10 min. In-hospital mortality was collected from hospital discharge data.

Statistical Analysis

Statistics were performed in GraphPad Prism 4.0 (GraphPad Inc., La Jolla, CA). Data are expressed as the mean ± standard deviation of the mean (SD) or as the median where the data is not normally distributed. A P value <0.05 was considered statistically significant. Mann–Whitney U tests were used to compare two groups of observations with non-parametric distributions. When the data were normally distributed a Student’s T test was used to compare groups. The Kruskal–Wallis test was used to look for correlation between categorical variables (such as the Marshall Score) and outcome.

Results

Patient Population

Forty-nine patients including 40 were male and 9 were females (mean age 42 ± 19 years) who had episodes of compromised PbtO2 (PbtO2 <25 mmHg for >5 min) that required treatment were identified. Thirty-four (69%) of these patients had a GCS of 3 on admission. Forty-one (84%) patients had a Marshall Score on their admission CT scan of 3 or more. Patient characteristics at presentation are presented in Table 1.

Table 1 Patient characteristics

Physiologic Variables

Data were analyzed from a total of 260 days of continuous PbtO2 monitoring (mean 5.2 ± 3.5 days). The daily mean PbtO2 among all patients was 33.8 ± 11.8 mmHg. A total of 564 episodes of compromised PbtO2 (PbtO2 <25 mmHg) and 127 episodes of brain hypoxia (PbtO2 <15 mmHg) were detected during this period of monitoring.

Medical Management of Compromised PbtO2

Of the 564 episodes of compromised PbtO2 (PbtO2 <25 mmHg), 379 episodes in 49 patients were treated with goal-directed therapy. Forty-three (88%) of these patients had more than one episode of compromised PbtO2. Thirty-eight patients (78%) had at least 1 episode of brain hypoxia (PbtO2 <15 mmHg for >10 min), 28 patients (57%) had 2 or more episodes. Medical management corrected 72% of the episodes of compromised PbtO2.

Ventilator manipulation, CPP augmentation, and sedation were the most frequent interventions. Increasing FiO2 was implemented as a treatment 186 times and restored normal PbtO2 77% of the time. Cerebral perfusion pressure augmentation was achieved by ICP reduction (n = 24) or increasing MAP (n = 69) in 93 episodes of reduced PbtO2 and was effective 62% of the time. When ICP was increased, treating it restored PbtO2 in 54% of episodes. Mannitol, the most commonly used agent, was successful 9 (75%) of 12 times it was used. When ICP was not considered a cause of reduced PbtO2, vasopressors were used 55 times to elevate CPP to ≥60 mmHg and restored PbtO2 75% of the time. Phenylephrine was the most frequently administered vasopressor and was effective 35 of the 46 times (76%) it was administered. Norepinephrine was effective 5 of the 7 times (71%) it was administered. Dopamine was effective the one time it was used whereas epinephrine did not improve PbtO2 following its single administration. Titrating patient sedation was used to improve patient PbtO2 56 times and was effective in 70% of treatments. Additional interventions including intravenous fluid bolus (n = 2), head repositioning (n = 6), airway suctioning (n = 8), and blood transfusions (n = 12) were effective in 50, 83, 88 and 50% of treated episodes of compromised brain oxygen, respectively. See Fig. 2.

Fig. 2
figure 2

Medical interventions for compromised PbtO2 (use/response rate)

Treatment Response Rates

Our data suggest similar response rates among various treatments and the variability in overall response to therapy among the patients. This may in part be associated with the “cause-directed” approach we used to correct reduced PbtO2. We compared the response rate observed for each intervention to the overall response rate for the patients who received that treatment. The intervention-specific to overall response rate ratios were calculated and are presented in Fig. 3. The interventions with a better than expected response rate (a ratio >1) were increasing FiO2, pressors (both neosynephrine and norepinephrine), airway suctioning, and benzodiazepines (midazolam and/or lorazepam). On the other hand, propofol, fentanyl, mannitol, and blood transfusion had ratios less than 1; these interventions were less effective than the average intervention for the patients in whom they were used.

Fig. 3
figure 3

Marginal patient-specific effectiveness of interventions

PbtO2 Treatment Response and Mortality

Eighteen patients (37%) died and 31 (63%) were alive at hospital discharge. The overall response rate (defined as total responses/total interventions) to medical treatment of compromised brain PbtO2 was associated with outcome. Survivors (n = 38) had a 71% response rate to therapy for abnormal PbtO2, whereas non-survivors (n = 11) had only a 44% response rate (P = 0.01). Mean daily PbtO2 (31.55 ± 20.64 mmHg) was significantly less in the 11 patients who died than in those who survived (37.43 ± 16.54 mmHg; P < 0.05). Non-survivors had more daily episodes (1 ± 0.8 vs. 0.5 ± 0.6, P = 0.03) and longer cumulative duration (273 ± 178 vs. 132 ± 159, P = 0.002) of brain hypoxia (PbtO2 <15 mmHg). Non-survivors also had a longer cumulative duration of brain hypoxia (461.8 ± 584.7 vs. 264 ± 494.8 min; P = 0.03) than survivors (Table 2). Survivors had more interventions for compromised PbtO2 than non-survivors (8.5 vs. 4.9, P = 0.15), however, this trend was non-significant and reflected the longer number of monitored days (5.7 vs. 3.4, P = 0.06) for survivors compared to non-survivors. The number of interventions per monitor day for survivors (1.48) was similar to that for non-survivors (1.45).

Table 2 Comparison of PbtO2 in survivors versus non-survivors

Older age and worse GCS are associated with worse outcome in patients with TBI. We asked if these two variables affected treatment response. Age less than 40 was associated with a significant increased response rate to PbtO2-directed therapy (76 ± 6.4% vs. 56 ± 6.7%, P = 0.04) and showed a trend toward improved outcome (survival rate at discharge 87 ± 7.2% vs. 69 ± 9.2%, P = 0.14). We did not find a relationship between GCS and response to therapy or mortality (Table 3). There was a trend for the Marshall CT score to be associated with outcome: 8 of 8 (100%) of patients with a Marshall score of 2 survived, whereas 7 of 15 (47%) patients with Marshall score of 5 survived (Kruskal–Wallis test, P = 0.07). The Marshall score was not associated with response rate to therapy (data not shown, Kruskal–Wallis test, P = 0.21).

Table 3 Differences in response rate to PbtO2-directed interventions and outcome by age and GCS

Discussion

In this study, we examined 49 patients who received medical management for reduced PbtO2. Our findings can be summarized as follows: (1) the commonest therapies to correct reduced PbtO2 were manipulation of pulmonary function, CPP augmentation, and sedation; (2) three quarters of episodes of compromised PbtO2 responded to medical management; (3) younger age was associated with a better response to therapy; and (4) responders had a lower mortality. These data suggest that PbtO2-based care when combined with ICP and CPP management may reduce mortality in some patients with severe TBI and provide useful information to design PbtO2 treatment strategies in severe TBI.

Study Limitations

Our study has several potential limitations. First, the study was performed on patients treated at a single institution so it may lack external validity. Second, the data were examined retrospectively and this may bias our results. Third, while management of compromised PbtO2 was protocol-driven, we do not have data on protocol compliance. Fourth, there are several reported thresholds for brain hypoxia or when to initiate therapy for brain hypoxia. In this study, we initiated therapy when PbtO2 was <25 mmHg and defined brain hypoxia as <15 mmHg. Fifth, therapies were often administered in parallel or in rapid sequence. Therefore, the efficacy of each separate therapeutic intervention may be confounded by combination treatments and less frequently used interventions may have lower response rates simply because they are second-tier therapies used in patients who have not responded to more commonly used interventions. In addition, it is conceivable that an intervention could have occurred for reasons other than to correct PbtO2 deficits or the effect of therapies applied as part of general care, e.g., positioning, rather than when a specific abnormality occurred may have been underestimated. Instead our data provide an expectation of what may happen when compromised PbtO2 is treated. Sixth, we do not have reliable data to exclude any deleterious effects of PbtO2-based interventions. While we think this is unlikely, it may be important since many of the therapies we used (e.g. increased FiO2, CPP augmentation, blood transfusion) are known to have adverse effects. This question is to be addressed in a Phase II trial of PbtO2-based care currently underway at our and other institutions. Seventh, our primary patient outcome measure was in-hospital mortality since it was included in the prospective database. While this is useful and accepted in ICU studies, it may not adequately describe TBI outcome where more long-term functional measures are important. Finally, since our sample size is small and the study was not designed to determine whether PbtO2-based therapy may improve severe TBI outcome, as suggested by some but not all studies our results only suggest but do not prove the value of PbtO2-based care [7, 1823]. Large clinical trials are needed to verify this issue. The data in this study provide useful information to help plan for such a trial.

Management of Reduced PbtO2

The use of PbtO2 monitors recently was incorporated into the Guidelines for Severe Traumatic Brain Injury [22]. However, PbtO2 monitors should not to be used alone but instead used with other monitors in particular an ICP monitor and standard ICU monitors. Our data show that compromised PbtO2 or brain hypoxia may not always be associated with ICP or CPP abnormalities. Medical interventions to improve ICP and CPP, however, often but not always can help improve abnormal PbtO2. Indeed, one-quarter of interventions with mannitol to reduce ICP and with phenylephrine to increase CPP did not correct brain hypoxia. We found that the most commonly used therapies to increase PbtO2 were increased FiO2, CPP augmentation usually with vasopressor administration, and sedation. However, overall only three-quarters of the episodes of compromised PbtO2 responded to medical therapies.

The response of abnormal PbtO2 to the most commonly used interventions was similar with a response rate of about 70%. This may reflect how we used the various therapies, i.e. in a cause directed manner. When we examined relative response rates of various interventions (Fig. 3), the most commonly used interventions (FiO2 supplementation, CPP augmentation with vasopressors, and reduction of brain metabolic demand with benzodiazepines) appeared to have better relative response rates when compared to other interventions in the same patients. For example, blood transfusion was successful in 50% of the instances it was used to correct abnormal PbtO2. However, the relative efficacy of various treatments must be interpreted with caution since those that appear to be less effective often were used as later tier therapies, i.e., they appear less effective since the abnormal PbtO2 was “refractory” to other interventions. Patient-specific factors also may be important: in our series younger age was associated with a better response rate to PbtO2-directed therapies.

There are few other studies that have addressed an overall strategy how best to manage compromised PbtO2 or brain hypoxia. We have observed that decompressive craniectomy can improve PbtO2 and reduce the therapeutic intensity level in severe TBI patients with medically refractory intracranial hypertension [24, 25]. In the present study, we chose only to examine medical therapies for reduced PbtO2. Previous studies that have addressed this question have tended to focus on one particular therapeutic intervention. For example, Kiening et al. in 21 patients observed that a CPP increase CPP from 32 ± 2 to 67 ± 4 mmHg improved PbtO2 by 62%. However, a CPP increase >68 mmHg did not lead to further improvement in PbtO2 [26]. Other reports have detailed the effects of individual therapies on PbtO2 in TBI or subarachnoid hemorrhage such as transfusion [2730], body position [31], barbiturates [32], hypertension [33], hyperventilation [34], CPP changes [3537], or hypertonic saline [38, 39] among others.

The Potential Importance of Correcting Reduced PbtO2

How the information provided by various monitors used in severe TBI patients is interpreted and how abnormalities are treated has the potential to provide a therapeutic benefit. It is conceivable that the information provided by a PbtO2 monitor may be used to establish the optimal ICP or CPP in individual patients (i.e., target patient and pathology specific factors) and to individualize therapy. Further work will be necessary to examine whether other factors such as cerebral autoregulation (pressure or metabolic) or PbtO2 reactivity can further target therapy. Our data, however, imply that among patients who receive standard ICP and CPP management and who develop reduced PbtO2 those who respond to PbtO2 therapy have a reduced mortality. In addition, a longer cumulative duration of brain hypoxia was associated with increased mortality. Whether the effect of PbtO2-directed therapy is an independent factor associated with mortality remains uncertain. However, this finding is consistent with other observational data that suggests that reduced PbtO2 is associated with mortality and worse outcome [10, 4042].

Conclusions

Our data provide information about the efficacy of various medical therapies for reduced PbtO2 and suggest that successful treatment of compromised PbtO2 may be associated with reduced mortality. Our data also show that there are medical interventions other than those to improve ICP and CPP that can correct compromised PbtO2. Our data, however, do not prove a benefit to correction of PbtO2—this important question requires a rigorous clinical trial. These data may be used to aid the design of such a trial.