Recent analysis of cerebral physiology in adult traumatic brain injury (TBI) has suggested a potential role of individualized treatment regimens based on advanced monitoring of cerebrovascular reactivity and the derivation of individualized cerebral perfusion pressure (CPP) targets, termed optimal CPP (CPPopt).1,2 This represents a shift toward more individualized medicine in the care for moderate/severe TBI patients. Data from initial studies suggests stronger outcome associations with individualized CPP targets, compared with applying the same target range to all patients.1–3
Aside from individualized CPP targets, individualized epidemiologic intracranial pressure (ICP) thresholds have been suggested based on a single center retrospective study in adult TBI.4–6 Using the relationship between continuously monitored cerebrovascular reactivity with the pressure reactivity index (PRx) and ICP, one can identify the ICP threshold where all subsequent higher ICP values yield PRx measures consistently above +0.20,4 a threshold value for PRx known to be associated with impaired cerebrovascular reactivity and global outcome in adult TBI.7–10 This has been termed the patient-specific or individualized ICP threshold, identifiable in ∼68% of patients.4 Prior retrospective analysis supports a potentially stronger association between the dose of ICP above individual epidemiologic thresholds, compared with the Brain Trauma Foundation (BTF) guideline defined threshold of 20 mm Hg, with global outcome in TBI.4 However, this has not been replicated in any other group of patients or outside of this single center.
The goal of this study is to utilize the multicenter Collaborative European Neuro Trauma Effectiveness Research in TBI (CENTER-TBI) study11 high-resolution intensive care unit (ICU) cohort data set, to evaluate the ability to derive individualized ICP epidemiological thresholds using a semiautomated algorithm, and compare the association between dose above individual ICP threshold and BTF guideline thresholds (ie, 20 and 22 mm Hg) with global patient outcome.
All patients from the multicenter CENTER-TBI high resolution ICU cohort were included in this study. These patients were prospectively recruited between January 2015 and December 2017. A total of 21 centers in the European Union (EU) contributed. All patients were admitted to ICU for TBI management during the course of the study, with high frequency digital signals recorded from the ICU monitors during the course of their ICU stay. All patients suffered predominantly from moderate to severe TBI (moderate=Glasgow Coma Score [GCS] 9 to 12, and severe=GCS of 8 or less). A minority of patients suffered from mild TBI (GCS 13 to 15), with subsequent early deterioration leading to ICU admission for care and monitoring. All patients in this cohort had invasive ICP monitoring conducted in accordance with the BTF guidelines.12
Data used in these analyses were collected as part of the CENTER-TBI study which has individual national or local regulatory approval; the UK Ethics approval is provided as an exemplar: IRAS No: 150943; REC 14/SC/1370). The CENTER-TBI study (EC grant 602150) has been conducted in accordance with all relevant EU laws if directly applicable or of direct effect, and all relevant laws of the country where the recruiting sites were located, including but not limited to, the relevant privacy and data protection laws and regulations (the “Privacy Law”), the relevant laws and regulations on the use of human materials, and all relevant guidance relating to clinical studies from time to time in force including, but not limited to, the ICH Harmonised Tripartite Guideline for Good Clinical Practice (CPMP/ICH/135/95) (ICH GCP) and the World Medical Association Declaration of Helsinki entitled “Ethical Principles for Medical Research Involving Human Subjects.” Informed Consent by the patients and/or the legal representative/next of kin was obtained, according to local legislations, for all patients recruited into the Core Dataset of CENTER-TBI, and documented in the e-CRF.
As part of recruitment to the multicenter high resolution ICU cohort of CENTER-TBI,11 all patients demographics were prospectively recorded. Similarly, all patients had high frequency digital signals from ICU monitoring recorded throughout their ICU stay, with the goal of initiating recording within 24 hours of ICU admission. All digital ICU signals were further processed (see the Signal acquisition/signal processing section). For the purpose of this study, the following admission demographic variables were collected: age, sex, admission GCS total and motor score, and admission pupillary response (bilaterally reactive, unilateral reactive, bilateral unreactive). We focused on the use of entirely nonimputed raw data, as final imputation of the entire CENTER-TBI data set is an ongoing process and will be part of subsequent publications and analysis. CENTER-TBI data were accessed/extracted using Opal database software,13 accessed on September 16, 2018.
Signal Acquisition and Processing
Signal acquisition and processing was conducted in an identical manner to previous CENTER-TBI high resolution ICU substudy publications. Further details can be found in the Supplementary Digital Content (Appendix A, Supplemental Digital Content 1, http://links.lww.com/JNA/A162), and in previous publications from this cohort.14,15 PRx was derived via the moving correlation coefficient between 30 consecutive 10 second mean windows of the parent signals (ICP and mean arterial pressure), updated every minute.16
Individual Patient-specific ICP Threshold Determination
For each patient, the relationship between PRx and ICP for the entire recording period was utilized to determine their individual ICP epidemiologic threshold. On the basis of the methodology outlined in the previous publications,4 the ICP value where PRx is +0.20, and all higher ICP values have PRx values persistently above +0.20, was considered the individual ICP threshold. Previous publications employed manual direct observation of the relationship between PRx and ICP, via error bar plotting, to determine the individual ICP threshold.4,5 It must be acknowledged that these individual thresholds for ICP do not represent therapeutic targets, but an individualized epidemiological threshold derived from the relationship between cerebrovascular reactivity values associated with global long-term outcomes. Thus the derived individual thresholds quoted within this manuscript should not be considered as therapeutic in nature, but in the context of purely preliminary exploratory work into personalized ICP thresholds in TBI. Further, the method for determination requires the use of the entire recording period, limiting this current technique to purely retrospective analyses.
In this study, we developed a semiautomated algorithmic method using R statistical computing software (R Core Team . R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. www.R-project.org/). First, an error bar plot of PRx versus ICP, using 2.5 mm Hg bins of ICP, was constructed for every patient. This was smoothed using locally weighted scatterplot smoothing (LOESS) functions for each patient. Second, using these LOESS fitted values we subsequently identified the lowest ICP value for which PRx was between +0.19 and +0.21 (ie, the lowest ICP values for intersection between the fitted LOESS function and the line “y”=+0.20 (ie, PRx=+0.20). This ICP value was selected as the patient’s individual ICP threshold. These thresholds were then assessed for validity by manual inspection of each patient’s error bar and LOESS function plots of PRx versus ICP. Any discrepancies between the algorithm-derived individual ICP threshold and the manually inspected ICP threshold were then corrected by hand, if present (hence “semiautomated”). Figure 1 displays 2 patient examples of the error bar and LOESS function plots, with the individual ICP threshold identification.
Grand mean values of all physiological variables were calculated per patient. In addition, post-ICM+ processing of physiological data occurred in R. Dose above ICP threshold was determined for the BTF defined ICP thresholds of 20 and 22 mm Hg, as well as for the patient’s individual ICP threshold. Dose was calculated in the following manner for each min-by-min observation: if ICP >ICP threshold, then dose=ICP−ICP threshold, otherwise generate no value. We then summated the dose over the entire recording period, and subsequently divided this value by the total duration of recording (in hours) to generate the mean hourly dose above threshold for thresholds of: 20, 22 mm Hg and the patient’s individual ICP threshold.
All statistical analysis was conducted using R (R Core Team . R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. www.R-project.org/) and XLSTAT (Addinsoft, New York, NY; www.xlstat.com/en/) add-on package to Microsoft Excel (Microsoft Office 15, Version 16.0.7369.1323). Normality of continuous variables was assessed via Shapiro-Wilks test. For all testing described within, the α was set at 0.05 for significance. All continuous variables were found to be nonparametrically distributed.
Despite GOSE being collected at both 6 and 12 months postinjury in this cohort of patients, there were missing data in both categories of outcome, as described in previous publications from the CENTER-TBI high-resolution ICU cohort. Thus, we combined GOSE scores from both 6 and 12 months in order to provide a “6 to 12 Month” GOSE. For patients where GOSE was reported for both 6 and 12 months, the last (ie, latest or 12 mo) GOSE score was selected for analysis.
GOSE was then dichotomized into the following categories: (A) alive (GOSE 2 to 8) versus dead (GOSE 1); and (B) favorable (GOSE 5 to 8) versus unfavorable (GOSE 4 or less) outcomes. Demographics and physiological variables were compared between each dichotomized group using Mann-Whitney U and χ2 testing where appropriate. Box plots were created for variables of interest comparing between dichotomized groups.
Univariate logistic regression (ULR) and bivariate logistic regression was conducted, comparing variables to both dichotomized GOSE outcomes, assessing superiority via area under the receiver operating curve (AUC), Akaike Information Criterion (AIC), and Delong test. Only ULR and bivariate logistic regression was conducted as this is only the second set of data for which individual ICP thresholds have been assessed, and we were only interested in testing a new algorithm for detection and validate the previous single-center results. Bivariate models composed of hourly dose above ICP of 20 mm Hg and mean PRx, and hourly dose above ICP of 22 mm Hg and mean PRx, were both created to assess association with both dichotomized outcomes. These models were compare to the univariate models which assessed the association between hourly dose above individual ICP threshold and the dichotomized outcomes.
Finally, the results from the ULR analysis were confirmed through multivariable logistic regression, by controlling for standard International Mission for Prognosis and Analysis of Clinical Trials in TBI Core (IMPACT-Core) admission variables: age, GCS motor subscore and pupillary response (as measured through an ordinal scale: bilaterally reactive, unilateral reactive, bilateral unreactive).17 Not all patients had a complete data set for this analysis, so we focused only on those with complete IMPACT-Core variables and identifiable individual ICP thresholds (ie, n=127).
A total of 196 patients from the CENTER-TBI high-resolution ICU cohort with high-frequency physiological signals and demographic variables were included in this study. This particular cohort has been described in detail in previous publications.14 The mean age was 46.6±19.7 years, with 150 (76.5%) being male. Median admission GCS was 8 (interquartile range: 5 to 13), and mean duration of physiological monitoring was 159.3±115.1 hours.
Using the semiautomated algorithm described to determine individual ICP thresholds, a total of 128 of 196 (65.3%) had an identifiable individual ICP epidemiologic threshold, in keeping with a previous single center study on the topic,4 with mean individual ICP threshold of 23.0±11.8 mm Hg (interquartile range: 14.9 to 29.8 mm Hg), and 73 of the 128 patients with an identifiable individual ICP threshold displaying individual thresholds above that defined by the BTF. Our semiautomated algorithm correctly identified the presence or absence of an individual ICP threshold in 162 (83.2%) of 196 patients. Thirty-four patients had either an incorrectly identified individual ICP threshold when one was not present (n=20), or no individual ICP threshold was identified when one was present (n=14). These 34 discrepancies were identified through manual inspection of both the error bar and LOESS function plots of PRx versus ICP, and subsequently corrected.
Patient demographics for those patients with an individual ICP threshold and those without an identifiable individual ICP threshold are shown in Table 1, with comparison of demographic and physiological factors between the 2 groups of patients. Of note is the higher mean ICP (P=0.041) and PRx (P<0.0001) in the patients without an identifiable individual ICP threshold, as identified via Mann-Whitney U testing, with a sustained higher PRx value in keeping without being able to identify an ICP threshold using the described methodology.
TABLE 1 -
Patient Demographics, Physiology and Outcome—Patients With and Without Defined Individual ICP Threshold
||Mean/Median (±SD or IQR)
||Population With Identifiable Individual ICP Threshold
||Population With No Identifiable Individual ICP Threshold
P (Mann U, t Test, χ2)
|Mean age (y)
|Sex (n [%])
|Median admission GCS (total)
|Median admission GCS motor
|Admission pupil response
|Mean duration of high frequency physiological recording (h)
|Mean ICP (mm Hg)
|Mean CPP (mm Hg)
|MAP (mm Hg)
|Mean individual ICP threshold (mm Hg)
|Mean PRx (a.u.)
|Mean hourly dose of ICP above 20 mm Hg
|Mean hourly dose of ICP above 22 mm Hg
|Mean hourly dose of ICP above individual threshold
|Number alive—6-12 mo
|Number dead—6-12 mo
|Number favorable outcome—6-12 mo (GOSE 5 to 8)
|Number unfavorable outcome—6-12 mo (GOSE 1-4)
Bolded P-values are those reaching statistical significance.
*No statistically significant difference between various admission pupillary categories, comparing patients with identifiable individual ICP threshold to those without one.
CPP indicates cerebral perfusion pressure; GCS, Glasgow Coma Scale; GOSE, Glasgow Outcome Score—Extended; ICP, intracranial pressure; IQR, interquartile range; MAP, mean arterial blood pressure; PRx, pressure reactivity index (correlation between ICP and MAP).
Comparing demographics and physiological variables between dichotomized outcome groups for the patents with an identifiable individual ICP threshold, we find that only mean PRx (P<0.001 for alive/dead, and P=0.005 for favorable/unfavorable outcomes) and mean hourly dose above the patient’s individual ICP threshold (P=0.010 for alive/dead, and P=0.020 for favorable/unfavorable outcomes) are significantly different (ie, higher), via Mann-Whitney U testing, in those who died or demonstrated unfavorable outcome at 6 to 12 months. Mean hourly dose of ICP above 20 and 22 mm Hg failed to display any significant difference between the dichotomized groups (Appendix B, Supplementary Digital Content 2, http://links.lww.com/JNA/A163). Figure 2 displays box plots of the mean hourly dose above each ICP threshold across both dichotomized outcomes.
Mean Hourly Dose of ICP Above Threshold and Outcome—Univariate Analysis
ULR was performed for each demographic and mean hourly dose of ICP above threshold with both 6 to 12 months dichotomized outcomes. The results of the ULR analysis with AUC’s, AIC, and P-values tabulated for each variable are shown in Table 2. Age was noted to be statistically associated with both alive/dead (AUC=0.820; 95% confidence interval, 0.736-0.904; P<0.0001) and favorable/unfavorable (AUC=0.708; 95% CI 0.618-0.799; P<0.0001) outcomes. Higher mean PRx was also noted to be associated with mortality and unfavorable outcome, in keeping with the previous larger single-center studies on cerebrovascular reactivity in adult TBI.7,10,16
TABLE 2 -
Univariate/Bivariate Logistic Regression Analysis for IMPACT Core and Physiological Variables—Patients With Identifiable Individual ICP Threshold
||A/D AUC (95% CI)
||F/U AUC (95% CI)
|Admission GCS motor
|Admission pupil reactivity
|Mean hourly dose of ICP >20 mm Hg
|Mean hourly dose of ICP >22 mm Hg
|Mean hourly dose of ICP above individual threshold
Mean hourly dose of ICP >20 mm Hg+mean PRx
| Mean hourly dose of ICP >22 mm Hg+mean PRx
Bolded P-values are those reaching significance (ie, P<0.05).
A/D indicates alive/dead; AIC, Akaike Information Criterion; AUC, area under the receiver operating curve; CI, confidence interval; F/U, favorable/unfavorable outcome (ie, favorable, Glasgow Outcome Scale of 5 to 8; unfavorable, Glasgow Outcome Scale of 1 to 4); ICP, intracranial pressure; IMPACT, International Mission for Prognosis and Analysis of Clinical Trials; MAP, mean arterial blood pressure; PRx, pressure reactivity index (correlation between ICP and MAP).
The mean hourly dose of ICP above the patient’s individual threshold displayed the highest AUC and lowest AIC values for association with both dichotomized outcomes (AUC=0.678, P=0.029 for alive/dead, and AUC=0.610, P=0.060 for favorable/unfavorable), with higher dose associated with mortality and unfavorable 6- to 12-month outcome. This was in comparison to the mean hourly dose of ICP above the BTF-based treatment thresholds of 20 and 22 mm Hg,12 as well as bivariate models including mean hourly ICP dose above 20/22 mm Hg and mean PRx. The association with mortality was statistically much stronger than unfavorable outcome, also in keeping with previous studies assessing the association between ICP and global outcome in adult TBI.7,12
Comparing AUC’s using Delong test indicated a significant difference between the AUC for mean hourly dose of ICP above individual threshold and both mean hourly dose of ICP above 20 and 22 mm Hg for alive/dead outcome (P=0.047 and 0.044, respectively). However, no significant difference was noted between the AUC’s of the 3 hourly dosing variables when outcomes where dichotomized as favorable/unfavorable.
Finally, comparing the bivariate models with mean hourly dose of ICP above 20/22 mm Hg and mean PRx, to the univariate model with mean hourly dose of ICP above individual threshold, for alive/dead outcome, the univariate models with mean hourly dose of ICP above individual threshold displayed statistically significant higher AUC’s compared with the bivariate models (P<0.05 for all; Delong test). There was no difference in AUC when comparing the bivariate models to the univariate individual threshold model for favorable/unfavorable outcome. Figure 3 displays the univariate receiver operating curves for mean hourly dose of ICP above 20 mm Hg, above 22 mm Hg, and above individual ICP threshold.
Controlling for Admission IMPACT-Core Variables
A total of 127 of the 128 patients with identifiable individual ICP thresholds had complete IMPACT-Core admission variables. Controlling for these admission characteristics in multivariable logistic regression, it was found that comparing models with baseline characteristics and mean hourly dose of ICP above 20 or 22 mm Hg, to those with mean hourly dose above individual ICP threshold, that the models with mean hourly dose above individual threshold trended toward higher statistically significant AUC’s, for both dichotomized outcomes. This confirms that the mean hourly dose above individual ICP threshold maintains significance, when controlling for IMPACT-Core covariates. A table summarizing the findings for the multivariable logistic regression analysis is available in Appendix C (Supplementary Digital Content 3, http://links.lww.com/JNA/A164).
This validation study provides multicenter confirmation of the presence of individual epidemiologic ICP thresholds, and replicates the strong association between time spent above this threshold and global outcome in adult TBI. There are some important aspects which deserve highlighting.
First, we have been able to display that individual ICP thresholds in moderate/severe TBI can be detected in 65.3% of patients from this cohort. This is in keeping with prior retrospective single center results on the topic.4 This is an important finding because not only does it validate previous results, but it also suggests that future studies will need to take this into account in order to be powered appropriately. Failure to detect individual threshold may be attributed to low ICP (never disturbing autoregulation) or too high ICP, when autoregulation is continuously disturbed. The wide distribution of individualized thresholds (interquartile range) from 14.9 to 29.8 mm Hg underlines the importance of such approaches to examining the individually defined burden of intracranial hypertension, as opposed to accepting fixed thresholds that are identical across patients. The individual thresholds identified for ICP below the BTF guideline ICP thresholds of 20 or 22 mm Hg are at this point still unclear in significance. This methodology is still very much nascent, with the current work being only the second in the literature, and requires substantial validation and exploration in other TBI populations as well as controlled experimental models. Thus, individual ICP thresholds below 20 mm Hg require further investigation, and we in no way suggest that ICP targets should be changed to target such low values. There needs to be a substantial subpopulation analysis in those patients who display low individual ICP thresholds in order to explain why such values may exist. This will be the focus of future studies on the topic. As mentioned, the goal of this study was only to provide a multicenter validation of the previously published single-center retrospective results from Cambridge.4
Second, we have, for the first time, created a semiautomated algorithm for the detection of individual ICP thresholds, an improvement over prior completely manual determination from plots of PRx and ICP. Though a first attempt, the accuracy rate in this study was 83.2%. It must be acknowledged that the notion of using an abnormal ICP compliance curve does not require a computer to determine, and can in fact be identified by inspection of the plotted physiology at the bedside by the treating clinician. Thus, our semiautomated algorithmic process would benefit from refinement and optimization, which will be the focus of future analyses in this area. Further to this, there are other potential options for assessing individual patient ICP thresholds, employing cerebral compliance indices, such as RAP (correlation between pulse amplitude of ICP and ICP),18,19 or using ICP waveform analysis.20,21 Exploration into these techniques as means to derive individual ICP thresholds is required, but may prove fruitful.
Third, we have been able to confirm the strong association between mortality and dose of ICP above individual ICP threshold, which was shown in the original publication describing this relationship,6 and done so in a multicenter data set. These results validate the presence and detectability of individual ICP thresholds, and provide a conceptual framework for developing these as treatment targets in the future targets, as we move toward individualized medicine. Support for such an approach is justified in the stronger association between mean hourly dose of ICP above individual threshold and both dichotomized 6 to 12-month outcomes, using ULR and multivariable logistic regression controlling for standard IMPACT-Core admission characteristics. ICP dose (time×intensity) calculated above individual thresholds were much more strongly associated with outcome compared with the dose above BTF defined thresholds of 20 and 22 mm Hg.12 The current analysis focuses on confirmation of past findings, but subsequent work will examine the impact of individual ICP thresholds in more complex multivariable models which include covariates beyond those used in the IMPACT-Core prediction model. Thus, there is still limited data to support the adoption of individual ICP thresholds as a clinically utilized measure at this time.
Fourth, an important finding reiterated by the results of this work is that ICP and burden of ICP suffered after TBI is linked to outcome. Particularly the dose of ICP spent above BTF thresholds, as well as individual ICP threshold, was statistically significantly associated with outcome. This is important to emphasize as recent literature has led to questions regarding the utility of ICP monitoring in adult TBI,22,23 leading to confusion for some providers as to the need for such monitoring devices. However, the results within this work added to the existing large body of evidence, supporting the link between ICP and patient outcome in TBI.7,12,24
One shortcoming of the approach implemented in this study is that individualized thresholds were calculated based on all of the ICP values across the patient stay. This approach clearly does not lend itself to providing a management target early in the course of the patient’s management, which is when it is needed. However, we hypothesize that individual thresholds of ICP may be detectable on-line (on the basis of recent ICP monitoring data points), and provide decision support for individualized management across all tiers of ICP therapy—starting from hypertonic solutions and finishing with better targeted decompressive craniectomy. Such a concept is still experimental and would require the use of sliding windows of data over time, to calculate the intersect between the PRx versus ICP function and PRx of +0.20. We envision such methodology to be similar to current CPP optimum sliding window determinations employed in real-time.1–3 However, it should be acknowledged that the feasibility of this has not been tested, and the concept is only a theory requiring much further investigation. If proven feasible, this would allow for a continuously updating individual ICP threshold value which could then account for changes in individual thresholds over time, which the current described methodology is incapable of accomplishing.
Important limitations of this study deserve highlighting. First, as mentioned in previously published studies from this cohort,14 despite the data from the CENTER-TBI high-resolution cohort being collected in a prospective manner, the treatments and therapies received by patients remain heterogenous. Such heterogeneity may have impacted the individual ICP threshold determination, and it is currently unclear whether individual therapeutic measures directed at ICP differentially impact the derivation of individualized thresholds. Such analysis, including the impact of injury and patient heterogeneity, will require even larger prospectively collected high-resolution data sets.
Second, our methodology for identification of individual ICP thresholds relies on the use of PRx as previously described.4 This current study was conducted as a simple validation of this previous retrospective single-center work. However, given that the methodology of individual ICP thresholds is still new there is the potential that other methods for estimating such thresholds may prove equivalent or superior. There is the potential that thresholding ICP based on autoregulation may be too simplistic, and other measures, such as compensatory reserve metrics,18,19 may provide superior information for stratifying critical values of ICP. The concept of individual ICP thresholds using PRx is still experimental. This concept is based on individual ICP thresholds derived through impairment in cerebrovascular reactivity, through epidemiologically defined critical values from previous retrospective studies,7 not compensatory reserve. It remains unclear if using a pure compliance/compensatory reserve index, such as RAP,18,19 would provide different information for the determination of individual ICP thresholds. Cerebrovascular reactivity can be impaired in both settings of normal and elevated ICP in adult TBI, where compensatory reserve indices tend to remain normal until extreme ICP elevations. Hence, we decided to employ a method of individual ICP threshold determination using vascular reactivity. It is unclear if these calculated thresholds occurring at lower ICP values represent normal brain or simply dysautoregulation and pressure-passivity at low ICP. Further work is required to correlate these findings with other continuously derived cerebral physiological metrics (such as blood flow velocity, cerebral blood flow, brain tissue oxygen partial pressure, or near-infrared spectroscopy based measures) and neuroimaging biomarkers, in order to determine whether the brain is in a “normal” state when individual ICP thresholds are determined to be below 20 mm Hg. As such, the current methodology should be considered an experimental starting point for such analysis, and not employed in the treatment of patients. There are plans for much further analysis of other physiologic metrics for the derivation of individual ICP thresholds, and these will the focus of various other studies on both the Cambridge retrospective TBI database and the CENTER-TBI high resolution ICU cohort.
Third, the overall patient numbers with an identifiable individual ICP threshold was low, at 128 and only 127 with full IMPACT-Core admission variables and an identifiable individual ICP threshold. Although based on the initial population size with a documented outcome and presence of baseline characteristics (n=196), a yield of 65.3% for individual ICP threshold is in keeping with prior larger retrospective studies on the topic.4 This relatively small population effect may be exemplified by the low AUC values on univariate analysis, and during correction for baseline IMPACT-Core covariates, despite reaching statistical significance. As such, future investigations into individualized ICP thresholds will definitely require larger cohorts. At the moment, we are unable to make definitive comments on the characteristics related to not being able to derive an individual ICP threshold. It is possible that patient admission demographics and both extracranial and intracranial injury burden characteristics will be predictive of those patients in whom an individual ICP is not identifiable. Such analysis was not the focus of this study, and will form the basis for a much larger analysis conducted on an amalgamated cohort from the large retrospective Cambridge TBI database and the CENTER-TBI high resolution ICU cohort. The hope is with such larger patient cohorts, we will be able to shed some light on the topic.
Fourth, despite the automated portion of the algorithm for detection of individualized ICP thresholds demonstrating an acceptable accuracy rate of 83.2%, there remains substantial room for improvement. As this was the first attempt at producing an semiautomated approach for individual ICP threshold determination, we feel encouraged about being able to improve upon this, as previous methods required a completely manual inspection of plots.4 This will be the focus of future work.
Fifth, despite our results indicating that those patients with no discernable individual ICP threshold displayed higher mean PRx and ICP values, our understanding as to the characteristics of such patients is limited. Future analysis of individual ICP thresholds will not only need to focus on those with an identifiable threshold, but also on those without, so that we can better understand what contributes to a lack of a patient-specific threshold.
Finally, despite the finding that ICP doses derived from individualized ICP thresholds display potentially stronger associations with outcome compared with BTF defined thresholds, the concept of individualized threshold should still be considered experimental. Currently, individual ICP thresholds should not replace the BTF defined thresholds in monitoring and care of moderate and severe TBI patients. Much further evidence is required to validate these individualized targets as clinically valuable in TBI.
Individual epidemiologic ICP thresholds are present in two thirds of the adult TBI population. Mean hourly dose of ICP above a patient’s individual epidemiologic ICP threshold demonstrates a stronger association with mortality compared with the dose above BTF defined thresholds of 20 or 22 mm Hg, confirming prior single center findings. Further studies on individual patient-specific epidemiologic ICP thresholds are warranted.
1. Steiner LA, Czosnyka M, Piechnik SK, et al. Continuous monitoring of cerebrovascular pressure reactivity allows determination of optimal cerebral perfusion pressure in patients with traumatic brain injury. Crit Care Med. 2002;30:733–738.
2. Aries MJH, Czosnyka M, Budohoski KP, et al. Continuous determination of optimal cerebral perfusion pressure in traumatic brain injury. Crit Care Med. 2012;40:2456–2463.
3. Needham E, McFadyen C, Newcombe V, et al. Cerebral perfusion pressure targets individualized to pressure-reactivity index in moderate to severe traumatic brain injury: a systematic review. J Neurotrauma. 2017;34:963–970.
4. Lazaridis C, DeSantis SM, Smielewski P, et al. Patient-specific thresholds of intracranial pressure in severe traumatic brain injury. J Neurosurg. 2014;120:893–900.
5. Lazaridis C, Smielewski P, Menon DK, et al. Patient-specific thresholds and doses of intracranial hypertension in severe traumatic brain injury. Acta Neurochir Suppl. 2016;122:117–120.
6. Lazaridis C, Czosnyka M. Patient-specific intracranial pressure. Response. J Neurosurg. 2014;120:892.
7. Sorrentino E, Diedler J, Kasprowicz M, et al. Critical thresholds for cerebrovascular reactivity after traumatic brain injury. Neurocrit Care. 2012;16:258–266.
8. Brady KM, Lee JK, Kibler KK, et al. Continuous measurement of autoregulation by spontaneous fluctuations in cerebral perfusion pressure: comparison of 3 methods. Stroke. 2008;39:2531–2537.
9. Zeiler FA, Donnelly J, Calviello L, et al. Validation of pressure reactivity and pulse amplitude indices against the lower limit of autoregulation, Part I: experimental intra-cranial hypertension. J Neurotrauma. 2018;35:2803–2811.
10. Zeiler FA, Donnelly J, Smieleweski P, et al. Critical thresholds of ICP derived continuous cerebrovascular reactivity indices for outcome prediction in non-craniectomized TBI patients: PRx
, PAx and RAC. J Neurotrauma. 2018;35:1107–1115.
11. Maas AIR, Menon DK, Steyerberg EW, et al. Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI): a prospective longitudinal observational study. Neurosurgery. 2015;76:67–80.
12. Carney N, Totten AM, O’Reilly C, et al. Guidelines for the management of severe traumatic brain injury, fourth edition. Neurosurgery. 2017;80:6–15.
13. Doiron D, Marcon Y, Fortier I, et al. Software Application Profile: Opal and Mica: open-source software solutions for epidemiological data management, harmonization and dissemination. Int J Epidemiol. 2017;46:1372–1378.
14. Zeiler FA, Ercole A, Cabeleira M, et al. Comparison of performance of different optimal cerebral perfusion pressure parameters for outcome prediction in adult TBI: A CENTER-TBI Study. J Neurotrauma. 2019;36:1505–1517.
15. Zeiler FA, Ercole A, Cabeleira M, et al. Univariate comparison of performance of different cerebrovascular reactivity indices for outcome association in adult TBI: a CENTER-TBI study. Acta Neurochir (Wien). 2019;161:1217–1227.
16. Czosnyka M, Smielewski P, Kirkpatrick P, et al. Continuous assessment of the cerebral vasomotor reactivity in head injury. Neurosurgery. 1997;41:11–17; discussion 17–19.
17. Steyerberg EW, Mushkudiani N, Perel P, et al. Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics. PLoS Med. 2008;5:e165; discussion e165.
18. Calviello L, Donnelly J, Cardim D, et al. Compensatory-reserve-weighted intracranial pressure and its association with outcome after traumatic brain injury. Neurocrit Care. 2018;28:212–220.
19. Zeiler FA, Kim D-J, Cabeleira M, et al. Impaired cerebral compensatory reserve is associated with admission imaging characteristics of diffuse insult in traumatic brain injury. Acta Neurochir (Wien). 2018;160:2277–2287.
20. Nucci CG, De Bonis P, Mangiola A, et al. Intracranial pressure wave morphological classification: automated analysis and clinical validation. Acta Neurochir (Wien). 2016;158:581–588; discussion 588.
21. Fan J-Y, Kirkness C, Vicini P, et al. Intracranial pressure waveform morphology and intracranial adaptive capacity. Am J Crit Care. 2008;17:545–554.
22. Chesnut RM, Temkin N, Carney N, et al. A trial of intracranial-pressure monitoring in traumatic brain injury. N Engl J Med. 2012;367:2471–2481.
23. Ropper AE, Chi JH. Treatment of traumatic brain injury without direct intracranial pressure monitoring. Neurosurgery. 2013;72:N19–N20.
24. Adams H, Donnelly J, Czosnyka M, et al. Temporal profile of intracranial pressure and cerebrovascular reactivity in severe traumatic brain injury and association with fatal outcome: an observational study. PLoS Med. 2017;14:e1002353.