See the corresponding editorial in this issue, pp 891–892.

J Neurosurg 120:893–900, 2014

©AANS, 2014

Patient-specific thresholds of intracranial pressure in severe

traumatic brain injury

Clinical article
Christos Lazaridis, M.D.,1,2 Stacia M. DeSantis, Ph.D., 3 Peter Smielewski, Ph.D.,1
David K. Menon, M.D., Ph.D., F.Med.Sci., 4 Peter Hutchinson, F.R.C.S.(SN), Ph.D.,1
John D. Pickard, F.R.C.S., M.Chir., F.Med.Sci.,1 and Marek Czosnyka, Ph.D.1
1
Academic Neurosurgical Unit, University of Cambridge Clinical School, Cambridge; 2Department of
Neurology, Divisions of Neurocritical Care and Vascular Neurology, Baylor College of Medicine, Houston;
3
Division of Biostatistics, School of Public Health at Houston, University of Texas, Houston, Texas; and
4
University Department of Anaesthesia, Addenbrooke’s Hospital, University of Cambridge, United Kingdom

Object. Based on continuous monitoring of the pressure reactivity index (PRx), the authors defined individual-
ized intracranial pressure (ICP) thresholds by graphing the relationship between ICP and PRx. These investigators
hypothesized that an “ICP dose” based on individually assessed ICP thresholds would correlate more closely with
the 6-month outcome when compared with ICP doses derived by the recommended universal thresholds of 20 and
25 mm Hg.
Methods. This study was a retrospective analysis of prospectively collected data from 327 patients with severe
traumatic brain injury.
Results. Individualized thresholds were visually identified from graphs of PRx versus ICP; PRx > 0.2 was the
cutoff. Intracranial pressure doses were then computed as the cumulative area under the curve above the defined
thresholds in graphing ICP versus time. The term “Dose 20” (D20) was used to refer to an ICP threshold of 20 mm
Hg; the markers D25 and DPRx were calculated similarly. Separate logistic regression models were fit with death as
the outcome and each dose as the predictor, both alone and adjusted for covariates. The discriminative ability of each
dose for mortality was assessed by receiver operating characteristic AUC analysis in which 5-fold cross-validation
was used. A clearly identifiable PRx-based threshold was possible in 224 patients (68%). The DPRx (AUC 0.81, 95%
CI 0.74–0.87) was found to have the highest area under the curve (AUC) over both D20 (0.75, 95% CI 0.68–0.81)
and D25 (0.77, 95% CI 0.70–0.83); in the cross-validation model, DPRx remained the best discriminator of mortality
(DPRx: AUC 0.77 [95% CI 0.68–0.89]; D20: 0.72 [95% CI 0.66–0.81]; and D25: 0.65 [95% CI 0.56–0.73]).
Conclusions. The authors explored the importance of different ICP thresholds for outcome by calculating pa-
tient-specific ICP doses based on the continuous monitoring of cerebrovascular pressure reactivity. They found that
these individualized doses of intracranial hypertension were stronger predictors of death than doses derived from the
universal thresholds of 20 and 25 mm Hg. The PRx could offer a method that can be directed toward individualizing
the ICP threshold.
(http://thejns.org/doi/abs/10.3171/2014.1.JNS131292)

Key Words      •      intracranial pressure      •      cerebrovascular pressure reactivity      •     

neuromonitoring      •      clinical outcome      •      traumatic brain injury

I
ntracranial hypertension has been closely linked to
adverse outcomes after traumatic brain injury (TBI);
nevertheless there have been no large randomized tri-
als that directly compare intracranial pressure (ICP) treat-
Abbreviations used in this paper: ABP = arterial blood pressure;
AUC = area under the curve; BTF = Brain Trauma Foundation;
CPP = cerebral perfusion pressure; DECRA = decompressive cra-
niectomy; D20, D25, DPRx = ICP doses based on thresholds of 20
mm Hg, 25 mm Hg, and PRx > 0.2; GCS = Glasgow Coma Scale;
GOS = Glasgow Outcome Scale; ICP = intracranial pressure; MAP
= mean arterial pressure; PRx = pressure reactivity index; ROC =
receiver operating characteristic; TBI = traumatic brain injury.
ment thresholds. Data from observational studies and
noncontrolled series have suggested thresholds ranging
from 15 to 25 mm Hg.2,11,20,25,28 The Brain Trauma Foun-
dation’s (BTF’s) latest guideline5 has identified a lack
of Level I evidence and offered, as a Level II, the treat-
ment threshold of 20 mm Hg, mainly based on the larg-
est available prospective observational study, which was
published by Marmarou et al.20 It is also recognized in
the BTF guideline that rather than accepting a generic,
This article contains some figures that are displayed in color
on­line but in black-and-white in the print edition.

J Neurosurg / Volume 120 / April 2014 893


absolute ICP threshold, an attempt should be made to
individualize thresholds based on patient characteristics,
other critical parameters, and on a risk-benefit consider-
ation of treating ICP values.
Cerebrovascular pressure reactivity is defined as the
ability of vascular smooth muscle to respond to chang-
es in transmural pressure, and represents one of the key
mechanisms responsible for autoregulation of cerebral
blood flow.22 Pressure reactivity can be determined by
observing the response of ICP to changes in mean arte-
rial pressure (MAP); if intact, a rise in MAP will lead
within 5–15 seconds to vasoconstriction with reduction
of cerebral blood volume, and ICP will decrease; if de-
fective, cerebral blood volume will increase passively
and ICP will rise. The opposite applies to a reduction in
MAP.24 A computer-aided method has been developed at
Cambridge to calculate and monitor the moving coher-
ence/correlation index between spontaneous slow waves
(20–200 seconds) of MAP and ICP.9,27
This method derives a pressure reactivity index (PRx)
that has values in the range between -1 and +1. A negative
or zero value reflects a normally reactive vascular bed,
whereas positive values reflect passive, nonreactive ves-
sels. Previous studies have established a significant cor-
relation between PRx and outcome after head injury, in-
cluding a time-dependent element: if PRx persisted above
0.2 for more than 6 hours, this was usually associated with
a fatal outcome.9,10,29 We defined patient-specific, pressure
reactivity–guided ICP thresholds by graphing the rela-
tionship between ICP and PRx over the total monitoring
time. We hypothesized that an “ICP dose” based on a dis-
turbed pressure-reactivity ICP threshold would correlate
more closely with clinical outcome when compared with
an ICP dose calculated using the generic, recommended
thresholds of 20 and 25 mm Hg. Our aim was to explore
the predictive ability of individualized ICP thresholds
based on PRx, compared with ICP insults derived from
universal thresholds of 20 and 25 mm Hg.
Methods
Patient Population
Monitoring of arterial blood pressure (ABP) and ICP
in patients following TBI has been an integral part of rou-
tine clinical management. The computerized data storage
protocol was reviewed and approved by the local ethics
committee of Addenbrooke’s Hospital, Cambridge Univer-
sity, and the neurocritical care unit user’s group. All data
were prospectively collected and stored. With individual
consent from patient representatives, we retrospectively
analyzed anonymized digital recordings of ABP and ICP
waveforms obtained in 327 consecutive patients with se-
vere TBI who were admitted to the neurocritical care unit
at Addenbrooke’s Hospital between 2003 and 2009.
All patients were sedated, intubated, and received me-
chanical ventilation during the recording period. A cere-
bral perfusion pressure (CPP)- and ICP-oriented protocol
for management of head injury was used, with CPP main-
tained at > 60 mm Hg and ICP < 20–25 mm Hg.21 Baseline
neurological status of each patient was determined using
894
C. Lazaridis et al.
the Glasgow Coma Scale (GCS) score. The postresuscita-
tion GCS score was used in patients in whom sedation was
discontinued immediately following hospital admission. In
patients who were deemed too unstable to undergo formal
neurological assessment on admission, the GCS score col-
lected on scene was used. The clinical outcome was as-
sessed at 6 months by using the Glasgow Outcome Scale
(GOS).16 All monitoring modalities recorded in the study
were part of standard clinical care.
Data Acquisition and Processing
The ABP was monitored invasively through the radial
or femoral artery with the aid of a standard pressure moni-
toring kit (Baxter Healthcare, CardioVascular Group) and
was zeroed at the level of the right atrium. The ICP was
monitored using an intraparenchymal probe (Codman ICP
MicroSensor, Codman & Shurtleff) inserted into the fron-
tal cortex. All signals were digitized using an analog-to-
digital converter (DT9801, Data Translation), sampled at a
frequency of 100 Hz, and recorded using a laptop computer
with ICM+ software (University of Cambridge, Cambridge
Enterprise, Cambridge, UK, http://www.neurosurg.cam.
ac.uk/icmplus). The same software was later used for the
retrospective analysis of all stored signals. Time-averaged
values of ICP, ABP, and CPP (CPP = ABP - ICP) were cal-
culated using waveform time integration over 60-second
intervals. The PRx was calculated as a short-term moving
Pearson correlation coefficient between changes in 30 con-
secutive 10-second averages of ABP and corresponding
ICP signals (with 80% overlap of data). Averaging over 10
seconds was used to suppress the influence of pulse and
respiratory waves.
The ICP Thresholds: Definitions and “Dose” Calculations
Based on the continuous measurement and monitor-
ing of PRx we defined patient-specific, individualized
ICP thresholds. These thresholds were visually identified
from graphs of PRx versus ICP over the total monitor-
ing time for each patient individually. The cutoff used
was PRx > 0.2; this was chosen based on previous work
from the group demonstrating that at that level there is
significant disturbance of pressure reactivity and an in-
creased mortality rate.9,29 The value for the ICP threshold
was only accepted if the graph showed a distinct change
of PRx values from < 0.2 to consistently > 0.2 (Fig. 1
shows 2 examples of individualized ICP threshold deter-
mination). To quantify the physiological insult from in-
tracranial hypertension we used a method that accounts
for the cumulative extent and duration of these episodes.
The method computes a “dose” of secondary brain injury
as the cumulative area under the curve above a defined
threshold. The trapezoidal method was used to calculate
doses from graphs of ICP versus time; the ICP “dose” is
measured in mm Hg*hour.32 For an ICP threshold of 20
mm Hg we named the dose D20. Using the same method-
ology we calculated D25 and DPRx. For identification of
ICP thresholds and calculation of doses the investigators
were blinded to clinical outcome.
Statistical Analysis
The data are presented as mean or median values
J Neurosurg / Volume 120 / April 2014
Pressure reactivity-based individualized ICP thresholds

Fig. 1.  Examples of individualized ICP threshold determination based on PRx > 0.2 for 2 patients.  A: A distinct change is
shown from PRx < 0.2 to PRx > 0.2 at an ICP of 12 mm Hg.  B: Tracings from a different patient in whom the ICP threshold
is 56 mm Hg (PRx changes from < 0.2 to > 0.2 at a range of 54–58 mm Hg). Both patients subsequently developed refractory
intracranial hypertension.

with their standard deviation or range, where appropri-
ate. The nonparametric Wilcoxon rank-sum test was used
to compare each measure by GOS status. Nonparametric
Spearman correlations and associated p values were used
to correlate ICP and PRx. For outcome analysis, 1 mean
value of the variables PRx, CPP, and ICP was calculated
for each patient. The predictive ability of each marker on
death was assessed. Logistic regression was performed
for each binary outcome and for each marker in the study.
For each analysis, receiver operating characteristic (ROC)
curves were calculated and the area under the curve
(AUC) was used as a measure of discriminative ability
and after adjusting for baseline GCS score, age, and sex.
Because observed AUCs are “over fit” to the data, to de-
termine how well the ICP doses would perform in terms
of prediction, 5-fold cross-validation of each covariate-
adjusted model was performed. For this procedure, the
logistic regression model was iteratively fit to four-fifths
of the data, and a predicted AUC was calculated on the
remaining one-fifth of the data; the 5 AUCs were then av-
eraged. Cross-validated results are presented as cross-val-
idated AUC and the 5th and 95th percentiles, and provide
an estimate of how well the different ICP doses would
predict death in a new data set. Statistical analyses were
performed using SAS version 9.3 and R 2.14.2 software.
Results
Descriptive Statistics
The database population included 327 patients (246
males, 75%), ranging in age from 15 to 87 years (median
age 36 years). The median baseline GCS score was 6 and
ranged from 3 to 15; 25% of patients had a baseline GCS
score > 8, but their condition subsequently deteriorated,
requiring critical care and warranting invasive monitor-
ing. The GOS score was missing in 5 patients (1.5%).
Outcomes were grouped as follows: good recovery in 51
(16%); moderate disability in 82 (25%); severe disability
in 105 (32%); persistent vegetative state in 9 (3%); and
death in 75 (23%). Patients in persistent vegetative state
were excluded from further outcome analysis because

J Neurosurg / Volume 120 / April 2014 895

 C. Lazaridis et al.

this group is disproportionally smaller than other out-

come groups, leaving 313 patients for outcome analysis.
Duration of data recording ranged between 1 and 28 days
per patient, with a mean of 4 days. Descriptive statistics
and p values for GCS score, age, MAP, ICP, CPP, and PRx
according to GOS group are reported in Table 1. Age,
GCS score, ICP, CPP, and PRx were significantly differ-
ent based on GOS status (p < 0.05).
The ICP Thresholds and ROC Analysis
Patient-specific ICP thresholds were visually identi-
fied from graphs of PRx versus mean ICP. A clearly iden-
tifiable threshold based on the set criteria was possible
in 224 patients (68%). The mean, median, interquartile
range, and SD values for the ICP threshold based on PRx
were 25, 24, 20–32, and 10, respectively. Figure 2 depicts
the ICP threshold distribution for PRx > 0.2 for the whole
cohort. Separate logistic regression models with death as
the outcome and dose as the predictor (both alone and
adjusted for the covariates GCS score, age, and sex) were
fit. To assess the discriminative ability of each dose for
mortality, an ROC analysis was performed and the AUC
was reported (Table 2).
In the covariate-adjusted logistic regression model,
all the doses calculated were significantly associated with
death (p < 0.0001 for D20, D25, and DPRx). Furthermore,
DPRx (0.81, 95% CI 0.74–0.87) was found to have the
highest AUC over both D20 (0.75, 95% CI 0.68–0.81) and
D25 (0.77, 95% CI 0.70–0.83), indicating that it has the best
discriminative ability. Cross-validation confirmed the re-
sults of the observed AUCs; in the cross-validated model,
DPRx was still the best predictor of death (DPRx AUC
0.77 [95% CI 0.68–0.89], D20 0.72 [95% CI 0.66–0.81],
and D25 0.65 [95% CI 0.56–0.73]). Figure 3 depicts the
resulting observed ROC curves for DPRx versus D20, both
unadjusted and adjusted for covariates. The larger area un-
der the ROC curve indicates that inclusion of these covari-
ates improves the discriminative ability of the marker.
Discussion
Raised ICP is detrimental across different neurologi-
cal conditions. Different management thresholds apply
depending on the clinical scenario. Patients with chronic
hydrocephalus withstand, without adverse effects, an
ICP rise up to 40 mm Hg during infusion studies.18 In
pseudotumor cerebri, patients often have chronically el-
TABLE 1: Patient demographics, clinical variables, and outcome in 322 patients with severe TBI*
Fig. 2.  The x axis shows ICP in mm Hg. The ICP threshold distribu-
tion for the whole cohort, as defined by patient-specific, individualized
thresholds for PRx > 0.2. The median value of 24 mm Hg is depicted.
The mean, interquartile range, and SD were 25, 20–32, and 10, re-
spectively.
evated ICP > 30 mm Hg without immediate neurological
compromise.4 In contrast, patients with severe TBI are
thought to have an increased mortality rate after an ICP
threshold that lies between 20 and 25 mm Hg. This can
be attributed to an acute, unaccounted for impairment of
free CSF communication; a mechanism serving to equili-
brate any detrimental pressure gradients within the cra-
niospinal space (Pascal’s law).
We explored the predictive ability of individual-
ized ICP thresholds based on the PRx, as compared with
“standard” fixed ICP thresholds. We found that the ICP
doses derived from an index describing the status of cere-
brovascular pressure reactivity were stronger predictors
of the 6-month mortality rate compared with doses calcu-
lated based on the “suggested” ICP threshold of 20 mm
Hg and also from a second fixed threshold of 25 mm Hg.
The most recent version of the BTF guidelines has rec-
ognized the lack of Level I evidence as it pertains to ICP
thresholds.5 The guidelines concluded with proposals, as
key issues for future investigation, to explore methods for
identifying critical ICP or “herniation pressure,” and for
identifying physiological parameters that would comple-
ment measurements of absolute ICP, in an effort to better

GOS Group No. of Pts Age (yrs)† M/F GCS Score† MAP ICP† CPP† PRx†
death 75 45 ± 18 60:15 6±3 95 ± 9 22 ± 11 73 ± 12 0.09 ± 0.2
PVS 9 — — — — — — —
severe disability 105 38 ± 15 78:27 6±3 94 ± 6 16 ± 4 78 ± 6 0.02 ± 0.16
moderate disability 82 32 ± 14 67:15 8±4 93 ± 7 16 ± 5 78 ± 7 0.01 ± 0.14
good recovery 51 34 ± 17 32:19 7±3 94 ± 6 15 ± 4 78 ± 7 −0.01 ± 0.13
total 322 237:85 17.5 ± 7 0.03 ± 0.17

*  Values are expressed as the mean ± SD. The MAP, ICP, CPP, and PRx values were averaged in each patient over the whole monitoring period. Pts =
patients; PVS = persistent vegetative state; — = excluded from further analysis.
†  Significant difference (p < 0.05, ANOVA) between GOS groups.

896 J Neurosurg / Volume 120 / April 2014

Pressure reactivity-based individualized ICP thresholds
TABLE 2: Logistic regression for ICP doses and ROC analysis of predictive power for mortality in patients with severe TBI
Univariate Covariate-
ICP Dose p Value AUC 95% CI Adjusted p Value*
D20 <0.0011 0.63 0.55–0.70 <0.0001
D25 0.0001 0.65 0.57–0.73 <0.0001
DPRx <0.0001 0.72 0.63–0.80 <0.0001
*  Adjusted for baseline GCS score, age, and sex.
quantify the impact of intracranial hypertension on sec-
ondary brain injury and neurological outcome. We found
highly significant correlations of increased mean ICP and
ICP dose with mortality rates. Intracranial hypertension
is commonly encountered in patients with severe TBI
and, although not universally, is widely accepted to cor-
relate strongly with neurological outcome.12,23
Nevertheless, recent publications have challenged the
traditional understanding on monitoring and treatment
of high ICP. Shafi et al.26 analyzed the National Trauma
Data Bank for the period 1994–2001, comparing patients
who underwent ICP monitoring with those who were
not monitored. After adjustment for multiple confound-
ing factors, including admission GCS score, age, blood
pressure, and injury severity score, ICP monitoring was
associated with a 45% lower rate of survival. The authors
interpreted this result as a failure of BTF criteria to iden-
tify the patients who were most likely to benefit from ICP
monitoring. Further possible explanations include the
idea that intracranial hypertension is a surrogate for se-
verity of brain injury, and that treatment interventions are
either ineffective or even potentially harmful.
In discussing interventions, the recent decompressive
craniectomy (DECRA) trial showed that this intervention,
despite effectively reducing ICP, did not translate into im-
proved neurological outcomes.8 It is of relevance that the
issue of an appropriate ICP threshold was highlighted as
one of the critiques and main differentiating features of the
DECRA trial from the ongoing RESCUEicp (Randomized
Evaluation of Surgery with Craniectomy for Uncontrolla-
ble Elevation of ICP) study.15 Our findings become further
pertinent in view of the recent publication of the random-
Fig. 3.  The ROC curves for DPRx versus D20, both unadjusted and
adjusted for covariates. The larger area under the ROC curve indicates
superiority of DPRx over D20; also, inclusion of covariates improves the
discriminative ability of the markers.
J Neurosurg / Volume 120 / April 2014
Covariate-Adjusted
AUC 95% CI Cross-Validated AUC* 95% CI
0.75 0.68–0.81 0.72 0.66–0.81
0.77 0.70–0.83 0.65 0.56–0.73
0.81 0.74–0.87 0.77 0.68–0.89
ized controlled trial of ICP monitoring in severe TBI by
Chesnut et al.7 This was the first such trial to compare man-
agement of intracranial hypertension based on monitoring
and treatment of ICP above the fixed threshold of 20 mm
Hg, versus a protocol based on clinical examination and
neuroimaging. This trial found no benefit of one protocol
over the other. It is beyond our scope to discuss this study
in detail; nevertheless and in conjunction with the afore-
mentioned DECRA trial, we believe that an important as-
pect in interpreting the results should be the limitation of
using fixed, universal ICP thresholds and thus disregarding
patient-specific injury patterns, individual pathophysiol-
ogy, and response to treatment interventions.
We chose to quantify secondary brain injury caused
by intracranial hypertension by using a method that ac-
counts for the cumulative extent and duration of these
episodes. This method computes a “dose” of intracrani-
al hypertension as the cumulative area under the curve
above a defined threshold. This approach may more accu-
rately reflect the impact of secondary brain insults on out-
come than did previous methods (for example, tradition-
ally taking into account only the mean ICP or the time
spent above a given threshold), because it accounts for
both the degree and the duration of ICP elevation.3,13,17,32
An additional advantage, as pointed out by Vik et al.,32 is
that the predictive power of doses for different thresholds
can be explored.
In this study we explored different thresholds by cal-
culating doses based on pressure reactivity and compar-
ing them against doses derived from the conventionally
accepted threshold of 20 mm Hg, and from a second fixed
threshold of 25 mm Hg, because this is the recommended
range in the BTF guidelines. To our knowledge this is
the first report attempting the determination of individu-
alized patient-specific ICP thresholds in patients with se-
vere TBI, using these thresholds to quantify ICP dose per
patient, and comparing these doses to the ones derived
from the currently accepted generic thresholds of 20–25
mm Hg. Although not undertaken in the present work,
similar methodology could be applied in the evaluation
of different CPP thresholds; furthermore, the relative
contribution to secondary brain injury of ICP versus CPP
insults could be explored. Cerebrovascular pressure re-
activity is an intrinsic underlying mechanism of cerebral
blood flow regulation. The status of pressure reactivity
has been shown to be important in bridging CPP with
ICP-targeted approaches.14 In the current analysis, as in
prior data from our group, the PRx is strongly correlated
with death (Table 1).
Steiner et al.29 and more recently Aries et al.1 have
897

demonstrated the value of using the PRx in identifying an
optimal CPP, under and above which outcome worsens.
In fact, this work has been recognized and incorporated
in the BTF guidelines as a method of optimizing cerebral
perfusion.6,19 In addition, we have observed that continu-
ous bedside monitoring of PRx in certain circumstances
can serve as a physiological alarm before the occurrence
of refractory intracranial hypertension (Fig. 4, recording
from a patient in our database). Finally, we report here
that an ICP dose as calculated by a patient-specific ICP
threshold corresponding to PRx > 0.2 was the best dis-
criminator of mortality in our cohort and as compared
with doses derived from “fixed” thresholds of 20 and 25
mm Hg.
Limitations of the Study
Intracranial pressure thresholds were identified ret-
rospectively from recordings of the whole period of ICP
monitoring. The findings, therefore, are not necessarily
applicable to real-time data collected in the ICU. Thresh-
old determination was done by visual inspection of ICP
versus PRx graphs; the next step to improve this meth-
odology would include an algorithm for objective and
automated identification of a patient-specific threshold.
To limit bias, both threshold identification and dose cal-
culation were done by investigators who were blinded to
clinical outcome.
We were able to identify a PRx-based ICP threshold in
two-thirds of our patients; frequent reasons that prevented
us from identifying a specific threshold are depicted in Fig.
5. Three patterns emerge: one of disturbed PRx throughout
the ICP range; a second reverse pattern where PRx remains
consistently preserved; and a third pattern where PRx var-
ies randomly with ICP. In these situations, PRx cannot be
Fig. 4.  An episode of refractory intracranial hypertension is depicted. At time point A, PRx is noted to increase progressively;
shortly thereafter ICP is noted to rise. Time point B is characterized by a precipitous ICP increase.
898
C. Lazaridis et al.
used in assessing ICP thresholds with the current meth-
odology. Possible explanations for these patterns relate to
an insufficient number of data sets either because of in-
adequate recording time or, less likely, because of a high
rate of artifacts (artifacts were minimized mainly because
patients were under heavy sedation and neuromuscular
blockade). It should be also kept in mind that the calcu-
lation of PRx requires the presence of spontaneous fluc-
tuations in ABP. Although such fluctuations are present in
the majority of patients following TBI, the magnitude may
be insufficient to produce significant ICP changes. In such
cases the value of PRx will be unreliable.
Apart from representing technical limitations, these
patterns could be physiologically interpreted as states of
dissociation between cerebrovascular pressure reactivity
and mean ICP. One should also consider that the influ-
ence of treatment targets on the relationship between ICP
thresholds and neurological outcomes is unknown. How-
ever, this bias is inevitable because this is a retrospective
review of patients from 2003 to 2009, who were treated
according to best available contemporary evidence, and
institutional and international guidelines. Last, in this
study we did not investigate markers of brain tissue oxy-
genation and metabolism that have been used in attempts
to characterize the physiological burden of intracranial
hypertension and to correlate these markers to neurologi-
cal outcomes.30,31 The interplay between cerebrovascular
pressure reactivity, tissue oxygenation, tissue metabolism,
and ICP is complex and not fully understood. It is plau-
sible that information from multimodality monitoring
will provide a more complete description of the physi-
ological consequences of ICP insults at the tissue level. In
addition, these multimodality data could potentially serve
in tailoring treatment thresholds and interventions away
J Neurosurg / Volume 120 / April 2014
Pressure reactivity-based individualized ICP thresholds
Fig. 5.  Examples in which PRx cannot be used in assessing ICP thresholds. Three patterns emerge: one of disturbed PRx
(> 0.2) throughout the ICP range (A); a second reverse pattern in which PRx remains consistently preserved throughout the ICP
range (B); and a third pattern in which PRx varies randomly with ICP (C).
from fixed, generic ICP values, and toward dynamic and
patient-specific cerebral physiology.
Conclusions
The predictive ability of individualized ICP thresh-
olds based on the continuous monitoring of cerebrovascu-
lar pressure reactivity was stronger than fixed thresholds
of 20 and 25 mm Hg, in a large single-center database of
patients with severe TBI. Monitoring of the PRx could
supplement ICP monitoring by offering patient-specific
pathophysiological information.
Acknowledgments
We acknowledge great support from Addenbrooke’s Hospital
Neurocritical Care Unit staff and Academic Neurosurgical Unit
registrars, without whose help the collection of computerized data
would never have succeeded. We also thank Emma Vought, M.S.,
for assistance with figure reproduction.
Disclosure
The software for brain monitoring ICM+ (http://www.neuro
surg.cam.ac.uk/icmplus) is licensed by the University of Cambridge
(Cambridge Enterprise Ltd.). Mr. Czosnyka and Mr. Smielewski
have a financial interest in a part of the licensing fee. Mr. Hutchinson
is supported by an Academy of Medical Sciences/Health Foundation
Senior Scientist Fellowship and grants from the Medical Research
Council (MRC)/National Institute for Health Research (NIHR). He
has also received a lecture honorarium from Codman and is a pat-
ent holder in Technicam. Mr. Menon is supported by funding from
the MRC, the NIHR Cambridge Biomedical Centre, and an NIHR
J Neurosurg / Volume 120 / April 2014
Senior Investigator award. Mr. Pickard is an NIHR senior investiga-
tor awardee and a principal investigator within the NIHR Biomedical
Research Centre (Cambridge University Hospital Foundation Trust)
and lead principal investigator for the MRC “Acute Brain Injury
Programme” grant. The remaining authors have not disclosed any
potential conflict of interest.
Author contributions to the study and manuscript preparation
include the following. Conception and design: Lazaridis, Czosnyka.
Analysis and interpretation of data: Lazaridis, DeSantis, Smielewski,
Czosnyka. Drafting the article: Lazaridis, Czosnyka. Critically revis-
ing the article: Lazaridis, Smielewski, Menon, Hutchinson, Pickard,
Czosnyka. Reviewed submitted version of manuscript: Smielewski,
Menon, Hutchinson, Pickard, Czosnyka. Approved the final ver-
sion of the manuscript on behalf of all authors: Lazaridis. Statistical
analysis: DeSantis. Study supervision: Czosnyka.
References
  1.  Aries MJ, Czosnyka M, Budohoski KP, Steiner LA, Lavinio
A, Kolias AG, et al: Continuous determination of optimal ce-
rebral perfusion pressure in traumatic brain injury. Crit Care
Med 40:2456–2463, 2012
  2.  Balestreri M, Czosnyka M, Hutchinson P, Steiner LA, Hiler
M, Smielewski P, et al: Impact of intracranial pressure and
cerebral perfusion pressure on severe disability and mortality
after head injury. Neurocrit Care 4:8–13, 2006
  3.  Barton CW, Hemphill JC, Morabito D, Manley G: A novel
method of evaluating the impact of secondary brain insults on
functional outcomes in traumatic brain-injured patients. Acad
Emerg Med 12:1–6, 2005
  4.  Biousse V, Bruce BB, Newman NJ: Update on the pathophysi-
ology and management of idiopathic intracranial hyperten-
sion. J Neurol Neurosurg Psychiatry 83:488–494, 2012
  5.  Bratton SL, Chestnut RM, Ghajar J, McConnell Hammond
FF, Harris OA, Hartl R, et al: Guidelines for the management
899

of severe traumatic brain injury. VIII. Intracranial pressure
thresholds. J Neurotrauma 24 (Suppl 1):S55–S58, 2007 (Er-
ratum in J Neurotrauma 25:276–278, 2008)
  6.  Bratton SL, Chestnut RM, Ghajar J, McConnell Hammond FF,
Harris OA, Hartl R, et al: Guidelines for the management of
severe traumatic brain injury. IX. Cerebral perfusion thresh-
olds. J Neurotrauma 24 (Suppl 1):S59–S64, 2007 (Erratum
in J Neurotrauma 25:276–278, 2008)
  7.  Chesnut RM, Temkin N, Carney N, Dikmen S, Rondina C,
Videtta W, et al: A trial of intracranial-pressure monitoring in
traumatic brain injury. N Engl J Med 367:2471–2481, 2012
  8.  Cooper DJ, Rosenfeld JV, Murray L, Arabi YM, Davies AR,
D’Urso P, et al: Decompressive craniectomy in diffuse trau-
matic brain injury. N Engl J Med 364:1493–1502, 2011
  9.  Czosnyka M, Smielewski P, Kirkpatrick P, Laing RJ, Menon
D, Pickard JD: Continuous assessment of the cerebral vaso-
motor reactivity in head injury. Neurosurgery 41:11–19, 1997
10. Czosnyka M, Smielewski P, Kirkpatrick P, Piechnik S, La-
ing R, Pickard JD: Continuous monitoring of cerebrovascu-
lar pressure-reactivity in head injury. Acta Neurochir Suppl
71:74–77, 1998
11. Eisenberg HM, Frankowski RF, Contant CF, Marshall LF,
Walker MD: High-dose barbiturate control of elevated intra-
cranial pressure in patients with severe head injury. J Neuro-
surg 69:15–23, 1988
12.  Elf K, Nilsson P, Ronne-Engström E, Howells T, Enblad P:
Cerebral perfusion pressure between 50 and 60 mm Hg may
be beneficial in head-injured patients: a computerized second-
ary insult monitoring study. Neurosurgery 56:962–971, 2005
13.  Hemphill JC III, Barton CW, Morabito D, Manley GT: Influ-
ence of data resolution and interpolation method on assess-
ment of secondary brain insults in neurocritical care. Physiol
Meas 26:373–386, 2005
14.  Howells T, Elf K, Jones PA, Ronne-Engström E, Piper I, Nils-
son P, et al: Pressure reactivity as a guide in the treatment of
cerebral perfusion pressure in patients with brain trauma. J
Neurosurg 102:311–317, 2005
15.  Hutchinson PJ, Kirkpatrick PJ: Craniectomy in diffuse trau-
matic brain injury. N Engl J Med 365:375–376, 2011 (Letter)
16.  Jennett B, Bond M: Assessment of outcome after severe brain
damage. Lancet 1:480–484, 1975
17.  Kahraman S, Dutton RP, Hu P, Xiao Y, Aarabi B, Stein DM, et
al: Automated measurement of “pressure times time dose” of
intracranial hypertension best predicts outcome after severe
traumatic brain injury. J Trauma 69:110–118, 2010
18.  Kim DJ, Czosnyka Z, Keong N, Radolovich DK, Smielewski
P, Sutcliffe MP, et al: Index of cerebrospinal compensatory
reserve in hydrocephalus. Neurosurgery 64:494–502, 2009
19.  Lazaridis C, Smielewski P, Steiner LA, Brady KM, Hutchin-
son P, Pickard JD, et al: Optimal cerebral perfusion pressure:
are we ready for it? Neurol Res 35:138–148, 2013
20.  Marmarou A, Anderson RL, Ward JD, Choi SC, Young HF,
Eisenberg HM, et al: Impact of ICP instability and hypoten-
sion on outcome in patients with severe head trauma. J Neu-
rosurg 75 (Suppl 1):S59–S66, 1991
900
C. Lazaridis et al.
21. Patel HC, Menon DK, Tebbs S, Hawker R, Hutchinson PJ,
Kirkpatrick PJ: Specialist neurocritical care and outcome
from head injury. Intensive Care Med 28:547–553, 2002
22.  Paulson OB, Strandgaard S, Edvinsson L: Cerebral autoregu-
lation. Cerebrovasc Brain Metab Rev 2:161–192, 1990
23.  Resnick DK, Marion DW, Carlier P: Outcome analysis of pa-
tients with severe head injuries and prolonged intracranial hy-
pertension. J Trauma 42:1108–1111, 1997
24.  Rosner MJ, Becker DP: Origin and evolution of plateau waves.
Experimental observations and a theoretical model. J Neuro-
surg 60:312–324, 1984
25.  Schreiber MA, Aoki N, Scott BG, Beck JR: Determinants of
mortality in patients with severe blunt head injury. Arch Surg
137:285–290, 2002
26.  Shafi S, Diaz-Arrastia R, Madden C, Gentilello L: Intracra-
nial pressure monitoring in brain-injured patients is associ-
ated with worsening of survival. J Trauma 64:335–340, 2008
27.  Smielewski P, Lavinio A, Timofeev I, Radolovich D, Perkes I,
Pickard JD, et al: ICM+, a flexible platform for investigations
of cerebrospinal dynamics in clinical practice. Acta Neuro-
chir Suppl 102:145–151, 2008
28. Sorrentino E, Diedler J, Kasprowicz M, Budohoski KP,
Haubrich C, Smielewski P, et al: Critical thresholds for cere-
brovascular reactivity after traumatic brain injury. Neurocrit
Care 16:258–266, 2012
29.  Steiner LA, Czosnyka M, Piechnik SK, Smielewski P, Chat-
field D, Menon DK, et al: Continuous monitoring of cerebro-
vascular pressure reactivity allows determination of optimal
cerebral perfusion pressure in patients with traumatic brain
injury. Crit Care Med 30:733–738, 2002
30.  Stiefel MF, Udoetuk JD, Spiotta AM, Gracias VH, Goldberg
A, Maloney-Wilensky E, et al: Conventional neurocritical
care and cerebral oxygenation after traumatic brain injury. J
Neurosurg 105:568–575, 2006
31.  Timofeev I, Carpenter KL, Nortje J, Al-Rawi PG, O’Connell
MT, Czosnyka M, et al: Cerebral extracellular chemistry and
outcome following traumatic brain injury: a microdialysis
study of 223 patients. Brain 134:484–494, 2011
32.  Vik A, Nag T, Fredriksli OA, Skandsen T, Moen KG, Schirm-
er-Mikalsen K, et al: Relationship of “dose” of intracranial
hypertension to outcome in severe traumatic brain injury.
Clinical article. J Neurosurg 109:678–684, 2008
Manuscript submitted June 20, 2013.
Accepted January 7, 2014.
Please include this information when citing this paper: published
online February 7, 2014; DOI: 10.3171/2014.1.JNS131292.
Address correspondence to: Christos Lazaridis, M.D., Depart-
ment of Neurology, Divisions of Neurocritical Care and Vascular
Neurology, Baylor College of Medicine, 6501 Fannin St., Mail slot
NB320, Houston, TX 77030. email: lazaridi@bcm.edu.
J Neurosurg / Volume 120 / April 2014