Introduction

Multi-modal monitoring of cerebral physiology in traumatic brain injury (TBI) is increasingly becoming common place in advanced neurocritical care units worldwide [25]. However, not only do the raw physiology parameters returned by these devices carry potential importance in directing therapies and prognostication, but metrics derived through combination of various biomedical signal processing techniques can expand our understanding of cerebral physiology [10]. Continuous cerebrovascular reactivity monitoring in TBI is one such derived metric. Recently, data suggests that impaired cerebrovascular reactivity after TBI appears to remain unaffected by current intensive care unit (ICU)-based therapeutics [13, 54], with large portions of a patients ICU stay spent in an impaired state [54], and relatively unchanged rates of impairment over the past 25 years of Brain Trauma Foundation (BTF)-based changes in care [13]. Thus, impaired cerebrovascular reactivity could potentially account for the relatively unchanged mortality rates seen in moderate-to-severe TBI over the last few decades, despite overall advances in ICU and post-ICU care. Though, it should be acknowledged that such findings are from few studies, requiring further validation.

Over the previous two decades, literature has supported continuous cerebrovascular reactivity monitoring in adult TBI, with the most commonly quoted measures being those derived from intracranial pressure (ICP) [10, 45]. However, other metrics based on brain tissue oxygen (PbtO2) [1, 20, 22], near-infrared spectroscopy (NIRS) [28], thermal diffusion blood flow [29], and transcranial Doppler [8, 35] have been derived in TBI populations [45]. Such metrics are derived through evaluation of the correlation between slow-wave vasogenic fluctuations in a surrogate measure of pulsatile cerebral blood volume (CBV) or cerebral blood flow (CBF), and a driving pressure for flow, such as mean arterial pressure (MAP) or cerebral perfusion pressure (CPP). The ICP waveform is utilized as a surrogate for pulsatile CBV, for the most commonly described cerebrovascular reactivity metric, the pressure reactivity index (PRx) [10]. Specifically, PRx is derived through the moving Pearson correlation coefficient between slow-wave vasogenic fluctuations in ICP and MAP. Initially, raw full-waveform ICP and arterial blood pressure (ABP) data is decimated to 0.1 Hz (Hz) using a 10-s non-overlapping moving average filter, in order to focus on slow-wave frequency ranges associated with cerebral autoregulation. Then, Pearson correlation is conducted on thirty consecutive 10 s mean measures of ICP and MAP (i.e., 5 min worth of data). This is updated typically every minute, through sliding the calculation window. Classically, negative values are believed to represent “intact” cerebrovascular reactivity, and positive values represent “impaired” reactivity. This methodology focuses only on the time-domain relationship between signals. As an aside, there are alternatives that rely on frequency-domain assessments of the relationship between measured physiologies, which require more complex methodologies [15, 26]. Such measures are not widely employed in clinical monitoring for TBI at this time, and will not be discussed further here.

To date, PRx has displayed a strong association with global outcomes in adult TBI [10, 36, 51, 55], with an independent association with outcome beyond that accounted for by ICP [55]. PRx is one of the only measures receiving some validation in experimental animal models as a measure of the lower limit of autoregulation [6, 47, 53]. In addition, recent data has shown that impaired cerebrovascular reactivity metrics may predict lesion growth on follow-up computed tomography (CT) of the brain in TBI [30, 31]. Furthermore, there is a strong link between impaired cerebral metabolism and deranged PRx values [43]. Critical thresholds associated with 6-month outcomes have also been defined for PRx in adult TBI [36, 51]. Finally, PRx has been utilized in the exploration of individualized CPP and ICP targets in TBI care [2, 24, 38, 56, 57], sparking an ongoing Phase II study for individualized CPP targets [4].

However, despite the promise of PRx, it remains experimental still in the world of advanced monitoring in TBI, with alternative ICP-derived continuous cerebrovascular reactivity measures emerging in the literature [3, 11, 19, 49]. These alternative ICP-based metrics have been overshadowed by the large, and ever growing, body of literature on PRx [45]. With that said, these other ICP-based metrics do deserve some attention and discussion around when they should be considered for monitoring. The goal of this scoping review is to provide a comprehensive overview of the available literature on these alternative (i.e., non-PRx) ICP-based continuous metrics of cerebrovascular reactivity in adult TBI.

Methods

In order to be comprehensive in this scoping overview of alternative ICP-based continuous cerebrovascular reactivity metrics, a systematic search of the literature was conducted, using the methodology outlined in the Cochrane Handbook for Systematic Reviewers [18], with reporting in keeping with the Preferred Reporting In Systematic Reviews and Meta-Analysis (PRISMA) [32] and PRISMA Scoping Review (PRISMA-ScR) guidelines [44].

Search question, population, inclusion and exclusion criteria

The question posed for this systematic review was what literature is available on time-domain non-PRx ICP-based continuous cerebrovascular reactivity measures in adult TBI? The following time domain continuous pressure techniques were searched: pulse amplitude index (PAx), correlation between pulse amplitude of ICP and CPP (RAC), bandpass filtered PRx55-15, and low-resolution-based PRx alternatives (long-PRx (L-PRx) and LAx). Low resolution is defined as using minute-by-minute mean physiologic measures in the derivation of the index. We purposefully did not evaluate frequency-domain-based alternative measures, such as continuously updating wavelet phase-shift methodologies [26], given the current focus clinically is on time-domain-based metrics, such as PRx and its alternatives, given their current greater ease of derivation and clinical application.

Inclusion/exclusion criteria

Inclusion criteria were all full manuscript studies including human subjects with TBI (any severity), studies with 5 or more patients, adults only (age 18 or older), and the use of a continuous non-PRx ICP-based cerebrovascular reactivity metric (as listed above). Exclusion criteria were non-English studies, animal studies, exclusively pediatric patient populations, studies of less than 5 patients, and studies on non-PRx ICP-based continuous measures of cerebrovascular reactivity not based in the time domain. We chose a limit of 5 patients or more, so as to avoid small incidental case reports on the topic of ICP-derived cerebrovascular reactivity measures.

Search strategy

MEDLINE, BIOSIS, EMBASE, Global Health, SCOPUS, and Cochrane Library from inception to December 2019 were searched using individualized search strategies. The search strategy for MEDLINE can be seen in Appendix A of the supplementary materials. Finally, reference lists of any review articles on autoregulation/cerebrovascular reactivity techniques were reviewed for relevant studies on continuous techniques in adult TBI.

Study selection

A two-step review of all articles returned by our search strategies was performed. First, the reviewer independently screened titles and abstracts of the returned articles to decide if they met the inclusion criteria. Second, full text of the chosen articles was then assessed to confirm if they met the inclusion criteria. Any meeting abstracts identified within the database search results were crosschecked with MEDLINE in order to determine if any full manuscripts were subsequently published based on these abstract results.

Data collection

Data were extracted from the selected articles and stored in an electronic database. Data fields included patient demographics, type of study, article country of origin, number of patients, technique described, and study outcomes described. We split the techniques into high resolution (i.e., PAx, RAC, PRx55-15) and low resolution (i.e., L-PRx and LAx), with Table 1 displaying the high-resolution studies, and Table 2 displaying the low-resolution studies.

Table 1 High-resolution non-PRx ICP-based cerebrovascular reactivity studies in adult TBI—study characteristics
Full size table
Table 2 Low-resolution non-PRx ICP-based cerebrovascular reactivity studies in adult TBI—study characteristics
Full size table

Results

Search results

A total of 111 results were obtained from the search strategy highlighted in Appendix A of the Supplementary Materials. Appendix B provides the PRISMA flow diagram for the search results. Through the first filter and second filter, a total of 23 articles met the inclusion criteria to be included in the review. A total of 16 studies were pertaining to high-resolution techniques, while 7 pertained to low-resolution techniques. All studies, except one [41], focused on entirely European and United Kingdom data. Tables 1 and 2 provide the details of the high-resolution and low-resolution studies, respectively.

“The basics”

The reader will find the following sub-sections describing some more details regarding the specific ICP-derived continuous metrics. In general, the derivation of these time-domain-based measures follows similar principals, evaluating the relationship between pulsatile CBV and a driving pressure (MAP or CPP). First, both the ICP and MAP/CPP signals are processed to focus on a frequency range associated with cerebrovascular reactivity. This is typically in the range of 0.05 to 0.005 Hz, and involves the application of either non-overlapping smoothing average filters, low-frequency bandpass filters, or both, depending on the metric of interest. Second, with the signal data processed, the update frequency for the data is either every 10 s, or minute, depending on the measure. Third, a window of data for the processed ICP and MAP/CPP data is then utilized to derive a Pearson correlation coefficient. This data window varies depending on the measure, with most using a 5-min window of data for the high-resolution measures. Longer window lengths are used for low-resolution metrics such as LAx and L-PRx. Fourth, this calculation is then updated every minute. In general, positive index values are believe to represent “impaired” cerebrovascular reactivity, while negative values denote “intact” cerebrovascular reactivity.

High-resolution techniques

Table 1 provides an overview of the available literature body for high-resolution non-PRx ICP-derived continuous cerebrovascular reactivity metrics in adult TBI. The following sub-sections will highlight each of PAx, RAC, and PRx55-15, their derivation method, and potential benefit over PRx in TBI monitoring. For details regarding individual studies, we refer the reader to Table 1, and the referenced literature.

  1. 1.

    PAx

Pulse amplitude index (PAx) is derived using the relationship between slow-wave fluctuations in the fundamental amplitude of ICP (AMP) and MAP. AMP is derived using the fundamental Fourier amplitude of full-waveform ICP data over a 10-s window, updated every 10 s. A moving Pearson correlation coefficient is then derived between AMP and MAP, in keeping with the methodology described previously for PRx. The rationale behind its derivation focuses on the potential for AMP to provide a better surrogate metric for pulsatile CBV. Statistically, PAx appears to behave similarly to PRx, RAC, near-infrared-based indices, and transcranial Dopple-based systolic flow index [46, 48].

Clinical literature on the use of PAx in TBI patients is limited to date. Initial retrospective studies suggested a potential benefit of PAx over PRx in 6-month outcome prediction, but only for those with low mean ICP values (i.e., less than 15 mmHg) over their ICU stay. In this study, PAx demonstrated a superior ability to predict mortality over PRx, in those with mean ICP values below 15 mmHg (− 0.04 ± 0.16 vs. − 0.14 ± 0.16, χ2 = 6, p = 0.01) [3]. Similarly, PAx was shown in one study to correlate with transcranial Doppler-based mean flow index (R2 = 0.46, p < 0.0002) [33]. Furthermore, critical thresholds for PAx in association with 6-month outcomes have been defined as 0, for favorable/unfavorable outcome, and + 0.25 for mortality [51]. This was conducted in non-craniectomized patients, so as to avoid any potential influence of decompressive hemicraniectomy on cerebrovascular reactivity measures [42]. Finally, impaired cerebrovascular reactivity, as defined by PAx, appears to be associated with admission CT imaging markers of diffuse intracranial injury, and advanced age [50]. Recent neuroimaging work has demonstrated that worse PAx values (r = 0.64, p = 0.0006) may be associated with CT-based lesion progression, particularly peri-contusional edema, during the acute phase of ICU care [30].

However, recent larger studies, both retrospective and prospective, have demonstrate inferiority of PAx in outcome prediction over PRx and RAC [51, 57, 58]. Furthermore, optimal CPP (CPPopt) based on PAx appear to have limited association with 6-month outcomes, with most PAx-CPPopt measures failing to reach any statistically significant associations with mortality or favorable/unfavorable outcomes at 6 months in the multi-center Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI) high-resolution cohort [57]. Thus, the role of PAx and its derived physiology targets remains in question in adult TBI monitoring.

  1. 2.

    RAC

RAC, defined as the correlation (R) between AMP (A) and CPP (C), is derived from the moving Pearson correlation between AMP and CPP. A relatively newly described index of cerebrovascular reactivity in TBI, its accurate interpretation is complex and beyond the scope of the review. Detailed description of this index can be found in the referenced articles [49]. In general, using CPP as the driving metric for flow, and AMP as the surrogate measure for pulsatile CBV, RAC provides information regarding both cerebrovascular reactivity [3] and cerebral compensatory reserve [9, 52]. Continuously measured cerebral compensatory reserve had been described in various sources in both the hydrocephalus literature and adult TBI literature through the index RAP—the correlation (R) between vasogenic slow-waves in AMP (A) and ICP (P) [21, 52]. Thus, with RAC being derived from the correlation between AMP and CPP, it provides information both from PAx (described in the previous section) and RAC indices, or so it is believed [49]. Its main limitation currently is complexity of interpretation. RAC appears to behave similarly to PRx, PAx, near-infrared-based indices, and transcranial Doppler-based systolic flow index, statistically [46, 48].

RAC and its evaluation in adult TBI patients is still considered exploratory and experimental in nature, as highlighted by the initial study fully describing this index [49]. Compared to PRx, RAC appears to perform similarly in its association with both mortality and favorable/unfavorable outcome at 6 months, in both retrospective [51] and multi-center prospective data sets [58]. In fact, RAC trends to higher AUC for favorable/unfavorable outcome association, compared to PRx, suggesting it may prove superior in this area of outcome prognostication. Thresholds for 6-month mortality and favorable/unfavorable outcome have also been described as − 0.10 and − 0.05, respectively [51]. Furthermore, RAC appears to display a more uniform parabolic relationship with CPP, suggesting it may carry a role in CPPopt determination [49, 58]. Some preliminary work in RAC-based CPPopt determination has demonstrated comparable performance versus PRx [57], though this work is quite exploratory in nature at this time. Finally, recent neuroimaging work has demonstrated that worse RAC values may be associated with CT-based lesion progression, particularly peri-contusional edema, during the acute phase of ICU care [30].

  1. 3.

    PRx55-15

In an attempt to focus on purely the vasogenic frequency range associated with cerebral autoregulation (i.e., 0.02 to 0.07 Hz) [14, 19, 37, 59], PRx55-15 was developed to remove very-low-frequency bands (i.e., 0.01 to 0.005 Hz) of ICP and MAP outside of this range typically found in classic PRx calculations, and provide a “pure” metric [19]. PRx classically contains the slow-wave vasogenic frequency band of 0.05 to 0.005 Hz [7, 10, 14]. In order to derive PRx55-15, raw full-waveform ICP and MAP data require processing using a frequency bandpass filter for 0.018 to 0.067 Hz (i.e., oscillations with a period of 55 to 15 s in duration) [19]. Then, 5-min sliding data windows are used to derive the cerebrovascular reactivity metric, similar to classic PRx [10].

Given the relatively new nature of this index, there exists very limited literature on its application in TBI. Small retrospective studies have found that PRx55-15 is independently associated with global 6-month outcome, and displays reduced variability compared to PRx [19, 39]. Furthermore, there appears to be a positive correlation between microdialysis-based lactate:pyruvate ratio and impaired cerebrovascular reactivity, as measured through PRx55-15 [40]. However, much further validation of this metric is required prior to widespread adoption.

Low-resolution techniques

Table 2 provides an overview of the literature for low-resolution ICP-derived continuous cerebrovascular reactivity metrics in adult TBI. Low-frequency autoregulation index (LAx) and long-PRx (L-PRx) variants exist, with nomenclature depending on the center and study. In general, these measures are derived using minute-by-minute mean ICP and MAP data. Moving Pearson correlation coefficients are then calculated between these minute-by-minute measures, using various described window lengths of 5 min up to and including 120 min. The most commonly quoted window lengths for calculation are on the order of 20 to 30 min in length. These metrics were developed for situations where high-resolution full-waveform ICP and MAP data was not available. Many commercially available ICU monitors have data output frequencies limited to 1 Hz or less, leaving classic PRx, or other high-resolution techniques not feasible. Similarly, many centers do not have the biomedical engineering expertise to extract, store and process full-waveform ICP and MAP data, if the ability to export such data exists. As such, in an attempt to expand the accessibility of cerebrovascular reactivity monitoring outside of a limited set of expert academic centers, these metrics were developed. The main limitation is that they evaluate the ultra-low-frequency band (i.e., below 0.005 Hz) for ICP and MAP, and uncertainty regarding the appropriate window length for calculation. It remains unknown how much, if any, autoregulation information is contained within this frequency range [19, 34]. Though, recent data suggests that shorter window lengths demonstrate strong statistical associations with classic PRx [41].

Various clinical studies exist correlating L-PRx/LAx metrics to high-resolution PRx, or other non-ICU-based continuous measures of cerebrovascular reactivity. Using a 20-min calculation window, L-PRx has been shown to be moderately correlated with ICP (r = 0.467, p = 0.011), and elevated lactate:pyruvate ratio on microdialysis [34], similar to larger microdidalysis studies with high-resolution PRx [43]. Similarly, LAx, studied using different window lengths, demonstrated strong outcome association at 6 months [16, 17]. Further, impaired cerebrovascular reactivity, defined by LAx, was shown to reduce the CPP threshold for insults tolerated in this patient cohort [17]. Furthermore, it has been demonstrated that LAx may be utilized to determine CPPopt, with strong outcome associations [11, 12]. However, some studies have questioned the strength of outcome association for these low-resolution metrics, displaying superior performance of PRx over L-PRx [23]. Thus, given the current limited and conflicting literature body, these metrics should still remain exploratory in nature.

Discussion

We aimed to produce a comprehensive scoping overview of the literature regarding ICP-based non-PRx continuous cerebrovascular reactivity metrics in adult TBI. The literature remains scare to date regarding the applicability of these metrics. However, it must be acknowledged that many of these measures display promise above and beyond classic PRx monitoring. The main drawback with PRx monitoring and management strategies is its dependency on a few selected ICU monitoring solutions; thus, if other indices show potential, it could offer solutions for more widespread implementation. Some important aspects deserve highlighting.

First, the high-resolution ICP-based alternative metrics appear to display some promise. PAx may have a role in those patients with low mean ICP for cerebrovascular reactivity assessments [3, 33], and potentially even CPPopt determination [57]. RAC, though complex in interpretation, carries information regarding both cerebrovascular reactivity and compensatory reserve [49], with trends toward stronger favorable/unfavorable outcome association at 6 months compared to PRx [51, 58]. Further, RAC may have an emerging role in CPPopt determination, given the stronger parabolic relationship seen between RAC and CPP, compared to PRx and CPP, on preliminary analysis [49, 57]. Finally, PRx55-15 provides a cerebrovascular reactivity index with more focus on the true autoregulation slow-wave frequency range [19], with strong outcome association [19, 39]. However, overall the literature on these three metrics is limited, leaving their application currently as experimental. Furthermore, all three indices are much more complex to derive, requiring comfort with either Fourier analysis [3, 49], or bandpass filtering techniques [19]. This may limit this adoption to non-expert clinical centers. Much further investigation is required to validate their associations with outcome, assess their role in prognostication, and determine applicability in the derivation of individualized physiologic targets in TBI.

Second, the low-resolution metrics, L-PRx and LAx, have very limited literature to date [11, 34]. They show promise in that they can be derived using minute-by-minute data, opening the doors for widespread adoption into centers without biomedical engineering expertise, or the ability to obtain high-frequency full-waveform physiologic data. Preliminary data indicates that L-PRx and LAx have some moderate association with high-resolution PRx [23], and demonstrate strong associations with outcome [11, 16]. In addition, they may even be applicable to the derivation of individualized CPPopt [12], potentially expanding this type of personalized medicine to non-specialized centers. However, L-PRx and LAx should still be considered very experimental, given it is unclear what aspect of the autoregulation they are truly measuring [14, 19, 23], and the optimal window length for calculation remains unclear [41].

Third, the literature body surrounding ICP-derived PRx alternatives is complex. Over the years, various measures of cerebrovascular reactivity have been derived based on ICP signals. Yet, none have gained the traction of the original metric in this field, PRx. As new measures are derived, they tend to be compared to PRx. This yields a situation where newer metrics tend to take a back-seat to PRx, regardless of their comparative performance. Indices which perform poorly in comparison to PRx tend to be “shelved,” rarely emerging in future studies on advanced neuromonitoring. Similarly, those metrics which perform similar to PRx also tend to be overlooked, as PRx is easier to derive and has a larger captive audience based on it being the historical first measure. This produces a situation where it is difficult for newer measures to gain traction and acceptance in the literature and amongst end-users, despite their nuanced potential. As multi-center collaborative groups and data sets emerge, it is imperative that these ICP-derived alternatives be re-explored.

Fourth, the strength of relationships and conclusions in specific studies must be considered preliminary and exploratory. As mentioned above, many studies compare newer ICP-derived metrics to PRx. Thus, we are comparing ICP to ICP-derived measures. This brings up the issue of collinearity in comparison between measures of similar origin. As such, the strength of statistically significant results requires some degree of skepticism in their interpretation given most have not accounted for such collinearity. One way around this would be to assess the relationship between two, or more, such measures while evaluating the variance inflation factor analysis. Another would be to explore the higher-resolution behavior of such measures over time using time-series methodologies, which would also account for autocorrelation within a given metric over time [41]. To date, very little work in this area has been conducted, and will be the focus of future multi-center collaborative data sets [5].

In general, we are still a ways away from adoption of either the high-resolution or low-resolution ICP-based alternative measures. All should remain experimental at this time. Future work in the area will take coordination between multiple international centers of excellent in cerebrovascular reactivity monitoring, pooled data, and ongoing prospective data collection strategies, in order to validate current findings and truly understand the role of these measures in adult TBI care.

Limitations

Despite a systematically conducted scoping overview of the literature being conducted, there are some significant limitations of this review and the outlined literature body, which deserve mentioning. First, despite an all-inclusive search strategy, it appears that there remains limited literature for ICP-derived continuous cerebrovascular reactivity monitoring, outside of PRx. Thus, even though some interesting trends regarding specific alternative indices were mentioned in this review, these metrics remain exploratory at this time. They should not be adapted to widespread clinical monitoring of cerebrovascular reactivity without proper validation and exploration in multi-center datasets. Furthermore, most studies were generated from a select few centers in Europe with specialized expertise in biomedical signal processing and its application to advanced monitoring in TBI patients. As such, many studies have overlapping patient populations from these few centers of excellence, given such data is both cumbersome to collect, analyze, and interpret. Given many studies in the area of biomedical engineering techniques in TBI monitoring are exploratory, the same datasets are utilized to try to answer different questions/aspects of cerebral physiology post-injury. It is impossible to explore and answer all questions related to different measures of cerebral physiology or cerebrovascular reactivity in one or two papers, given the complexity of analysis and interpretation, as well as the myriad of very different questions being answered. The archived data sets from the few centers of excellence who lead this area are optimally positioned to be able to answer such preliminary questions, and have thus generated many papers from a relatively small data source. Of course, validation of such findings is another story, and highlights the need for future multi-center data collection strategies in TBI, which may be able to both validate and provide clarity regarding the utility of such measures.

Second, most of the studies included were retrospective in nature. The patient populations were heterogeneous in nature, with varied periods of high-frequency digital physiologic monitoring described. Many of these studies came from the same centers, with overlapping patient populations between articles. Thus, it is unknown how many unique patients were studied overall. Furthermore, these patients were adult TBI populations receiving active treatment for ICP and CPP during their ICU care. Meaning, their described behavior does not represent the true untreated behavior, or nature behavior, in adult TBI. As such, lack of association between these alternative metrics and classical PRx must be taken with some degree of skepticism. No strong conclusions regarding these alternative measures can be made at this time, as they all suffer from significant and varied biases.

Third, given the heterogenous nature of these studies, formal meta-analysis is not possible, nor would it have led to interpretable results. As such, we are left with a simple scoping overview of the literature. This emphasizes the preliminary and exploratory nature of the described studies, necessitating much further investigation in larger multi-center cohorts.

Fourth, at this time, we are unable to make definitive comments on the utility of longer window versions of PRx, or weighted window approaches. One of the included studies documented various window lengths in the determination of CPPopt based on LAx [11, 12]. It is possible that LAx or L-PRx provide different pieces of information regarding cerebral autoregulatory status, by variations in window length and/or weighting used in the derivation of such indices. Shorter versus longer windows may provide information on dynamic or static aspects of cerebral vasomotion. However, it must be acknowledged that increasing window lengths does run the risk of including information beyond the vasogenic frequency range associated with autoregulation, further muddying the picture. Thus, at this time, much further work is required to both validate such measures and investigate their ability to measure aspects of the Lassen curve in experimental models, and their application in TBI monitoring. Such work will be the ongoing efforts of the CENTER-TBI high-resolution sub-study [27] and the CAnadian High Resolution Traumatic Brain Injury (CAHR-TBI) Research Collaborative [5].

Fifth, given the small number of studies, from a few centers in Europe, with occasional overlapping data sets, there runs the risk that the identified studies in this area suffer from publication bias, with only positive findings being described. This is true of many emerging and niche areas in medicine, where the focus is only on positive findings. As such, this is another factor which limits the ability to extrapolate the findings of the studies included in this scoping review. However, one must acknowledge that recent multi-center data sets have begun to provide evidence in support of the previous single center retrospective findings [57, 58]. This again highlights the need for ongoing multi-center initiatives in advanced neuromonitoring [5, 27].

Finally, we limited our search to continuous metrics, based in the time domain. This was purposefully conducted, as current application of continuous cerebrovascular reactivity monitoring in adult TBI utilizes time-domain-based approaches. Such approaches are convenient and have a high ease of application for clinicians at the bedside. With that said, there are some frequency-domain ICP-based continuous measures that were not described in this review, such was wavelet-PRx [26]. These metrics may prove useful in the future, with improved background processes for derivation and better understanding of their interpretation. However, at this time, they have limited applicability in the clinical monitoring of the TBI patient.

Future directions

Based on this scoping review, it is clear that the literature on ICP-derived continuous cerebrovascular reactivity indices is limited. As alluded to above, there exists a role for future evaluation of such measures. Future work necessitates multi-center prospectively acquired high-frequency physiologic data sets in order to provide validation to these previous exploratory works. This work will focus not only on the derived cerebrovascular reactivity metrics themselves, but also on individualized physiology thresholds, such as CPPopt or individual ICP thresholds, in order to assess which measure may be optimally positioned to derive such personalized targets. This will require large patient numbers, in addition to quality control in the data collection/archival processes. Such initiatives are underway with the CENTER-TBI HR cohort [27] and the recently developed CAHR-TBI initiative in Canada [5]. Furthermore, evaluation of lower-resolution metrics, such as LAx and L-PRx, is a focus of our collaborative research group between Canada, Sweden, and Finland [41]. Such work will focus on evaluating the role of these alternative measures in the clinical monitoring of cerebrovascular reactivity, and will employ more complex statistical methodologies, accounting for various patient factors, injury patterns, and collinearity. This work will also employ time-series methodologies to account for autocorrelation, and hopefully better describe the higher-frequency relationship between such measures in vivo in TBI. Finally, we plan to explore machine learning methodologies in the assessment of co-variance and behavior of these various indices over time. Such initiatives will be multi-disciplinary endeavors, including computer science, engineering, neuroscience, physiology, and population health experts.

However, future work cannot only rely on clinical data sets, as every cerebrovascular reactivity measure has yet to be validated as a reliable measure of the Lassen curve. As such, future work necessitates experimental large animal studies to evaluate the ability of such metrics to detect both the lower limit and upper limit of autoregulation, during extremes of both MAP and ICP. Such large animal model work is expensive, and will require multi-center collaborative approaches in order to accomplish such goals.

Conclusions

Non-PRx continuous ICP-based cerebrovascular reactivity metrics can be split into high-resolution and low-resolution measures. High-resolution indices include PAx, RAC, and PRx55-15, while low-resolution indices include L-PRx and LAx. It remains unclear the true role for this metrics beyond classic PRx. Each displays situations where it may prove superior over PRx, given limitations with this currently widely accept measure. Much future investigation into each of these alternative metrics is required prior to adoption into the clinical monitoring armamentarium in adult TBI.