Elevated intracranial pressure (ICP) is associated with herniation, significant morbidity, and mortality in patients with severe traumatic brain injury (TBI) [1–3]. The Brain Trauma Foundation guidelines recommend lowering ICP below 20 mmHg but state that identification of the critical value of ICP is a major unanswered question [4]. Because patients may have considerable variance in the duration and magnitude of elevated ICP, appropriate treatment targets are difficult to identify, as was recently demonstrated in a major trial of surgical decompression for refractory ICP [5, 6].

Prior studies that have investigated an association between elevated ICP and poor outcome may have been limited by their ability to detect only individual episodes [7]. Episodic measurements are unable to provide information regarding depth and duration of insult; whereas more continuous, high resolution capture of elevated ICP over time provides an area under the curve or “pressure time dose” (PTD) of intracranial hypertension. In addition to serving as a potentially more intuitive assessment of elevated ICP, high resolution collection may also be more sensitive to elevated ICP detection when compared to manual measurements [8, 9]. A recent NIH sponsored conference on research and technology in neuro-critical care recommended, as a future research application, the development of “advanced processing of ICP to predict brain swelling and use as a biomarker outcome variable [10].”

There have been few studies that have evaluated ICP PTD using high resolution automated data capture in patients with severe TBI [8, 11–13]. The aim of this study was to evaluate the association of ICP PTD with mortality and functional outcome in patients with severe TBI.

Patients and Methods

Study subjects were admitted to R. Adams Cowley Shock Trauma, a level I academic tertiary medical center, with a dedicated neurotrauma intensive care unit. After study approval from the Institutional Review Board (IRB), data were prospectively collected, reviewed, and analyzed from patients with severe TBI, defined by Glasgow Coma Scale score of <9 who were older than 14 years of age, admitted within 6 h of injury onset between 2007 and 2008, and required ICP monitoring as part of routine clinical care. All patients had evidence of TBI confirmed on computed tomography (CT). Patients were excluded if there was significant non-cerebral injury (abbreviated injury score >3), had delayed initiation of ICP monitoring (after 24 h), or were immediately considered unlikely to survive based on urgent neurosurgical consultation.

Routine clinical characteristics including patient demographics, mechanism of injury, routine vital signs, type of ICP monitoring, and need for surgical decompression were recorded. The Marshall CT classification was made by a neuro-intensivist (KNS) based on the admission head CT [14]. Outcome measures included in-hospital mortality and the extended Glasgow Outcome Scale (GOSE) to analyze functional outcome at 3 and 6 months. The GOSE was obtained by a trained research coordinator who was not involved in collection or aware of ICP data using a structured telephone interview [15].

ICP Monitoring and Continuous Data Collection

ICP was monitored by either an intraventricular catheter or by parenchymal monitor (Camino; Integra Neurosciences). In patients in whom multiple monitors were in place, parenchymal monitoring was used for analysis. Blood pressure was monitored invasively by arterial line, and cerebral perfusion pressure (CPP) was calculated by the monitor’s proprietary algorithm.

Automated, continuous real-time vital sign data were captured from bedside monitors (GE-Marquette-Solar-7000/8000) through a vital sign data recorder (VSDR) as previously described [8]. Vital sign data were captured throughout the trauma center at 6 s intervals. In order to account for periodic drainage for ICP monitored through an external ventricular drain, a piecewise cubic Hermite interpolation method was used (Matlab 7.7 R2008b; Mathworks, Natick, MA) [16]. After any CSF drainage, any first peak was discarded unless subsequent data was within 10 % of the range of observed values. This cleaning procedure discarded less than 1 % of data points. Continuous data were then compressed and transferred via secured server for further filtering and analysis. Pre-determined extreme ranges were removed from data analysis (ICP < 0, ICP > 100, CPP < 0, CPP > 250, MAP < 0, MAP > 250 all in mmHg). Hourly values for ICP were recorded as cumulative area under the curve or ″Pressure times Time Dose″ (PTD; mmHg/h) to describe both the cumulative amplitude and duration of episodes above and below an ICP threshold of 20 mmHg for each of the 12 h time periods and over 7 days. Previous researchers have shown that PTD is superior to mean, minimum, maximum, and % time above and below thresholds for a number of physiological parameters following TBI [8, 11]. Using the continuous data capture, the number of 5 min episodes of ICP above 20 mmHg was also obtained (Fig. 1).

Fig. 1
figure 1

Sample graph of continuous ICP data collection. Using an ICP threshold of 20, there are two episodes of elevated ICP during this period

Full size image

Patient Management

Patients with severe TBI admitted to our institution are admitted to a dedicated neuro-trauma unit, and clinical teams utilize a standardized protocol for the early and intensive care management, based on the Brain Trauma Foundation guidelines [17, 18] (Table 1). Patients were intubated and received mechanical ventilation, with a goal of eucarbia (pCO2 35–40 mmHg). ICP management includes elevated height of bed and ICP monitoring, either with a parenchymal monitor and/or external ventricular drain. Clinicians target euvolemia and euthermia, blood pressure algorithms to maintain ICP and CPP goals of less than 20 mmHg and greater than 60 mmHg, respectively. Stepwise treatment includes the use of hypertonic saline as the first tier response for any elevation in ICP, sedation and pain medication, and surgical decompression for mass lesions or refractory ICP. Therapeutic hypothermia and barbiturate coma are seldom used.

Table 1 Tiered options for protocol driven management of elevated ICP. In tier two, medical therapies are used as a bridge to decompressive craniectomy
Full size table

Statistical Analysis

Baseline clinical characteristics of the study population are presented as percentages for categorical variables and medians and ranges for continuous factors. Initial scatter plots indicated a non-normal distribution of PTD measurements for both ICP and CPP. Thus, PTD values were transformed using the natural logarithm and the mean value constants were added to each variable before transforming to overcome undefined values of log 0. Separate logistic regression analyses were then conducted to ascertain the unadjusted effect on outcome of the natural log of PTD for [1] ICP > 20 mmHg and [2] CPP < 60 mmHg during three different intervals (i.e., 24, 48, and 72 h following admission). Outcome measures included in-hospital surgical decompression, 30 day mortality, and poor functional outcome as determined by the GOSE at both 3 and 6 months. GOSE was modeled both as a binary outcome measure (GOSE 1–4 [poor] vs. GOSE 5–8 [good]) and as an ordinal outcome measure (GOSE 1–2 [very poor], GOSE 3–4 [poor], GOSE 5–6 [good], GOSE 7–8 [very good]), the latter created to take into account the natural ordering of functional categories. Further analysis involved development of additional logistic regression models to determine if the effect of transformed PTD on outcome changed after adjustment by age and admission GCS score. Because the log transformation represents a percent change rather than an absolute difference between numbers, results reflect the effect of a 10 % increase in PTD on the risk of a poor outcome for easier interpretation.

Similar regression analysis was conducted to determine the unadjusted and adjusted effect of a unit increase in the number of 5 min episodes of ICP > 20 mmHg on the same outcome measures. Episode data, however, were normally distributed and were therefore not transformed before examination in a regression model.

P values and odds ratios (OR), along with their 95 % confidence intervals (CI), are reported for the estimated regression coefficient to reflect the result of a 10 % increase in the PTD variable within each model. A p value below 0.05 is indicative of statistical significance.

Results

Subject Characteristics

Table 2 contains a summary of the demographic and clinical characteristics of the severe TBI population. The majority of patients were male and under the age of 40. Approximately 20 % of patients in our group had an admission GCS that was above eight, but subsequently suffered neurological deterioration which warranted ICP monitoring. Hypotension defined by a systolic blood pressure less than 90 was only present in 5 % of patients. The predominant mechanism of injury was fall and motorcycle or vehicle collision.

Table 2 Clinical characteristics and demographics in 60 patients with traumatic brain injury
Full size table

ICP Monitoring

The median start of ICP monitoring was 4 h and 42 min after admission, and admission was defined by the time any vital sign was first captured (e.g., heart rate). The median monitoring time was 4.67 days. Three patients had no episodes of elevated ICP.

Marshall CT Scores and Surgical Decompression

Patients were evaluated by Marshall CT classification according to their initial CT scan. The majority of patients had Marshall CT category 2 (55 %). Twenty-three (38 %) patients were treated with surgical decompression. Surgery and outcome by Marshall Score are presented in Table 3. No patients in this study received pentobarbital.

Table 3 Craniectomy and outcome by Marshall Score
Full size table

There was no signification association between 72 h PTD and surgery (p = 0.51) or increased number of refractory episodes at 72 h and surgery (p = 0.59).

ICP Dose, ICP Episodes, and Outcome

Eight (13 %) patients died before discharge. At 3 months, 58 % (30/52) of patients had a poor outcome, while 47 % (25/53) of patients had poor outcome at 6 months. ICP PTD measurements were analyzed in blocks of 24 h intervals (Table 4). No significant association was found between PTD and decompressive craniectomy at any of the three time points (24, 48, and 72 h; data not shown). However, unadjusted PTD values from 48 and 72 h of monitoring were significantly associated with 30 day mortality, with the 95 % confidence interval just covering 1.00, at 24 h. The associations with mortality were even more evident at 24, 48, and 72 h after adjustment for age and admission GCS scores. With regard to functional outcome at 3 months, 72 h of unadjusted PTD was significantly associated with GOSE 1–4, with slightly wider non-significant confidence intervals for shorter intervals of monitoring. This association between PTD and poor GOSE continued at 6 months as well, becoming stronger at all three intervals of monitoring when corrected for age and GCS. In addition to a binary outcome, the effect of ICP PTD was ascertained when GOSE was examined as an ordinal outcome measure. Across the ordinal analysis, a 10 % increase in ICP PTD predicted a poorer GOSE outcome (e.g., 72 h of PTD: 3 months, OR 1.05, p = 0.001; 6 months, OR 1.04, p = 0.006) (Table 5). Furthermore, the strength and significance of these associations remained when analyzing the effect of ICP > 25 mmHg.

Table 4 Unadjusted and adjusted effect of 10 % Increase in PTD for ICP > 20 at 24, 48, and 72 h on the risk of mortality and poor GOSE
Full size table
Table 5 Unadjusted and adjusted effect of 10 % increase in PTD for ICP > 20 at 24, 48, and 72 h on cumulative poor GOSE
Full size table

When individual 5 min episodes of intracranial hypertension were examined, there was no significant association with decompressive craniectomy, mortality or functional outcome at any time point (Table 6).

Table 6 Unadjusted and adjusted effect of unit increase in number of 5 min episodes for ICP > 20 at 24, 48, and 72 hours on mortality and GOSE
Full size table

CPP Dose, CPP Episodes, and Outcome

PTD measurements of CPP < 60 mmHg were also analyzed in blocks of 24 h intervals. Contrary to findings related to ICP, however, no significant associations were evident between CPP PTD and outcome (30 day mortality or ordinal GOSE at 3 or 6 months), whether unadjusted or adjusted for age and admission GCS scores, at 24, 48, or 72 h (Table 7).

Table 7 Unadjusted and adjusted effect of 10 % increase in PTD for CPP < 60 at 24, 48, and 72 h on the risk of mortality and cumulative poor GOSE
Full size table

Discussion

This is a prospective observational study examining the relationship between physiologic area measurements of ICP, episodes of intracranial hypertension, and outcome. We observed an association between ICP PTD and poor outcome, as defined by mortality and long-term functional outcome. These high resolution measurements were feasible and more useful than episodic detection of raised ICP.

Raised ICP is a focus of management in patients with head injury. While several studies have demonstrated an association with raised ICP and poor outcome, this result has not been consistent across reports [1, 19, 20]. This observation is likely due to differences in monitoring criteria, patient selection, and varying parameters of measurement [21]. Furthermore, definitive proof that lowering ICP improves outcome which has been elusive, in part, because many of the therapies for lowering ICP also alleviate the damage from tissue shifts and cerebral edema, even when the ICP is normal [22]. Recently, a large randomized trial evaluating decompressive craniectomy for refractory ICP in head injury failed to demonstrate effectiveness for surgery, raising questions about how clinicians detect and treat ICP in head injured patients. Three possibilities may explain this finding: (1) the treatment for ICP lowering was ineffective, (2) elevated ICP may be a biomarker with limited utility as a treatment target, (3) the methodology for ICP detection did not sufficiently exploit the mechanistic relationship between abnormal ICP and outcome.

In this study, we examined the PTD as an area under the curve measurement of ICP. The underlying hypothesis was that a dose calculation of elevated ICP, which would incorporate both magnitude and time, would be most relevant to the injured and at risk brain. Importantly, not only did PTD demonstrate an association with mortality, but this association was also present for functional outcome and showed durability at 6 months follow up, even after controlling for age and admission GCS. In our cohort, there was an “ordered effect”, where increasing duration of observation increased the strength of our association. It is possible that PTD would predict poor outcome at earlier time points; our study may have been underpowered to detect this association, or it is possible that 24 h following admission is just too soon to detect an association. However, the number of patients in our cohort is significantly less than prior studies that have used number of episodes of ICP to make establish an association between ICP and outcome [23, 24]. Sample size calculations in therapeutic studies of ICP should consider the timing of the intervention and the expected outcome based on ICP PTD data at that time point.

In this study, we analyzed the result as a 10 % increase in PTD. The PTD for ICP > 20 at 72 h ranged from 0 to 9,000 with a median of 511. A 10 % increase then yields an absolute increase of 51. A smaller percentage cutoff would be possible but likely less informative, since the maximum, low incidence values at the extremes of several thousand are likely the result of values not excluded by our protocol. The relationships we have described are consistent, only with different effect sizes, at different thresholds. Further study identifying the natural history of PTD range and clinically relevant cut points are required.

With regard to surgical evacuation of hematoma or decompression, there was no association between ICP PTD and outcome. Several patients in our cohort underwent surgery because of mass occupying lesions and neurological deterioration rather than ICP elevation refractory to medical management.

Our group has previously demonstrated that automated measurements of ICP are superior to routine hourly measurements with regard to detection of ICP elevation, even if episodes of elevated ICP are brief [8, 25]. Most of the patients in our cohort had evidence of mass lesions; however, similar results have been seen in cohorts of TBI patients with greater number of Marshall CT category III [11]. As in the current cohort, ICP PTD in this study demonstrated an association with functional outcome at 6 months.

Elevated ICP is likely both a marker of the severity of brain injury as well an intermediary in the pathogenesis of cerebral ischemia and neuronal death. It is also an indirect measure of compliance and likely exerts it effect in a dose-dependent manner. Therefore, ICP PTD may be a candidate physiological biomarker for patients with TBI. In this context, ICP PTD may serve as an intermediate endpoint to better evaluate the effectiveness, or lack thereof, of existing and future treatments. While clinicians may need to respond to sudden elevations in increased ICP at the bedside, PTD analysis may be a useful method to observe ICP physiologically meaningful trends over time in any individual patient.

Measuring PTD using an automated, high resolution system is feasible and reliable. Individual episodes of elevated ICP, using a common clinically used definition of 5 min of sustained rise, did not predict patients who might have poor prognosis. In contrast, a 72 h assessment using PTD had robust predictive capabilities. Prior studies have examined other similar physiological relationships important in acute brain injury, and a similar association with poor outcome has been demonstrated, despite using a manual collection system [8].

PTD analysis may identify patients at risk for the detrimental sequelae of raised ICP; however, several questions still remain. Our cohort was heavily weighted toward patients with severe TBI, and these results should be replicated in patients with moderate brain injury. In this way, continuous PTD monitoring may be able to identify those patients at highest risk for neurological deterioration and potentially as an early warning system. In addition, the relationship between PTD and outcome in other forms of acute brain injury remains to be clarified. Finally, because some patients with TBI, such as those with primary brainstem injury, may have poor outcome independent of ICP, PTD guided decision-making may not be as beneficial in these patients. Importantly, we did not observe an association between low CPP and outcome, despite the presence of an association between high ICP and poor outcome. Our cohort may have been insufficiently powered to detect any true association. Alternatively, the impact of increased pressure or reduced compliance may be more detrimental than decreased perfusion.

Future analyses should examine other physiological variables that affect critically ill neurological patients, such as temperature, blood pressure, and brain tissue oxygenation. The identification of optimal ICP lowering therapies is a significant unanswered question in TBI. Because ICP PTD is able to serve as a potential marker of ongoing brain injury in small numbers of patients, clinical trials of various ICP lowering agents should consider using ICP PTD as a biomarker efficacy endpoint.