Continuous Time-Domain Cerebrovascular Reactivity Metrics and Discriminate Capacity for the Upper and Lower Limits of Autoregulation: A Scoping Review of the Animal Literature

Over a wide range of systemic arterial pressures, cerebral blood flow (CBF) is regulated fairly constantly by the cerebral vessels in a process termed cerebral autoregulation (CA), which is depicted by the Lassen autoregulatory curve. After traumatic brain injury (TBI), CA can get impaired and these impairments manifest in changes of the Lassen autoregulatory curve. Continuous surrogate metrics of pressure-based CA, termed cerebrovascular reactivity (CVR) metrics, evaluate the relationship between slow vasogenic fluctuations in a driving pressure for cerebral blood flow, and the most commonly studied and utilized measures are based in the time domain and have been increasingly applied in bedside TBI care and have sparked the investigation of individualized cerebral perfusion pressure targets. However, not all CVR metrics have been validated as true measures of autoregulation in the pre-clinical setting. We reviewed all available pre-clinical animal literature that assessed the association between continuous time-domain metrics of CVR and some aspect of the Lassen autoregulatory curve. All 15 articles found associated the evaluated continuous metrics to the lower limit of autoregulation curve whereas none looked at the upper limit. Most of the evaluated metrics showed the ability to discriminate the lower limit of autoregulation with various methods of perturbation. Further work is required to evaluate the utility of such surrogate measures against the upper limit of autoregulation, while also providing validation to the existing literature supporting specific indices and their ability to discriminate the lower limit.


Introduction
The innate ability of the cerebral vessels to maintain cerebral blood flow (CBF), relatively constantly, over a wide range of systemic arterial pressures, is termed cerebral autoregulation (CA). 1,2 The Lassen autoregulatory curve depicts CA where the CBF is relatively constant between the lower and upper limits of autoregulation (LLA and ULA). The Lassen autoregulatory curve is plotted with cerebral perfusion pressure (CPP) or mean arterial pressure (MAP) along the x-axis, and CBF along the y-axis, where CBF has classically been invasively measured. 3 Impairment in CA has been documented in various neuropathological states, including post-stroke 4-7 and after traumatic brain injury (TBI), [8][9][10][11][12][13][14] across the spectrum of injury severity. Such impairment manifests in changes in the position of the LLA or ULA on the Lassen autoregulatory curve or, in the worst instance, absence of the autoregulatory curve altogether. 7,15 These alterations in the autoregulatory curve expose the brain to pressure-passive flow states, where low MAP or CPP leads to hypoperfusion, whereas high MAP or CPP leads to hyperperfusion. Recent literature suggests that the exposure burden of impaired CA is a significant driver of poor long-term outcomes in various neurological conditions. [10][11][12][16][17][18] Subsequently, to avoid subjecting the brain to hypo-or hyperperfusion conditions, it is imperative that we have the capacity to continuously and accurately monitor CA at the bedside.
Direct measurement of CA is not possible at the bedside, given that continuous and accurate measures of CBF are not readily available to the treating clinician. As such, we must rely on surrogate metrics of CA, termed cerebrovascular reactivity (CVR) metrics, involving changes in gas levels, pressure, or flow velocity to drive alterations in vascular caliber. 8,15,19 Continuous indices that have not been fully validated as measures of the Lassen autoregulatory curve have been given the term CVR metrics. The clinical realm has made a concerted effort to refer to these indices as CVR metrics to avoid misrepresenting them as a completely validated measure of autoregulation. We will maintain such consistent nomenclature throughout this article in keeping with this designation within the field. Such continuous CVR metrics evaluate the relationship between slow vasogenic fluctuations in a driving pressure for CBF (i.e., MAP or CPP) and a surrogate for pulsatile cerebral blood volume (CBV) or CBF. The table in Supplementary Appendix SB pro-vides a list of continuous CVR metrics derived from various cerebral monitoring techniques, which are of interest given that they can be adopted in the bedside clinical care of TBI patients. As such, CVR indices of interest are only those that have continuously updating data streams.
The most readily studied and utilized measures are based in the time domain and are derived as moving Pearson's correlation coefficients between raw physiological signals, with negative values denoting ''intact'' autoregulation and positive values denoting ''impaired'' autoregulation. Cerebral monitoring devices utilized to obtain surrogate measures for pulsatile CBV/CBF derived from raw continuous physiological signals include both invasive and non-invasive modalities, such as: intracranial pressure (ICP), near-infrared spectroscopy (NIRS), transcranial Doppler (TCD), brain tissue oxygen (PbtO 2 ), thermal diffusion flowmetry (TDF) CBF, and laser Doppler flowmetry (LDF) CBF. 19,20 With these devices, various CVR indices can be derived at the bedside in humans, with most literature to date focused on TBI patient cohorts. We mainly focused on time-domain metrics given that these measures have seen widespread adoption by clinicians at the bedside because they give greater simplicity in continuous derivation and interpretation over frequency-domain measures. However, not all indices have been explored in pre-clinical models as measures of the Lassen autoregulatory curve, and, as such, we conducted this study.
However, recent literature in moderate/severe TBI suggests that not all such continuously derived CVR metrics are the same, with varying levels of covariance between them. [21][22][23] Further, literature validating these metrics against the Lassen autoregulatory curve in pre-clinical models has been scarce and scattered across various subspecialty journals. 19,20,24,25 This ambiguity has left confusion among clinicians as to which indices are truly validated measures of autoregulation, despite support for their use from international consensus groups. 14,[26][27][28] Thus, it is imperative that we understand which measures truly evaluate aspects of the Lassen autoregulatory curve and which do not. Such knowledge will improve confidence in their use to manage TBI patients and highlight areas for future work. As such, the goal of this study was to perform a systematically conducted scoping review of the preclinical animal literature, assessing for any documented association between continuous time-domain metrics of CVR and some aspect of the Lassen autoregulatory curve.

Methods
A systematically conducted scoping review of the available literature was conducted using the methodology outlined in the Cochrane Handbook for Systematic Reviews. 29 The data were reported in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA). 30 Supplementary Appendix SA of the Supplementary Materials provides the PRISMA checklist. The search strategy and methodology are similar to other scoping reviews published by our group. [31][32][33] Review questions and search strategy were decided upon by the supervisor (F.A.Z.) and the primary author (A.S.S.).

Search questions, population, and inclusion and exclusion criteria
The question posed for this scoping systematic review was: What pre-clinical animal literature exists to validate continuous measures of CVR as true measures of the Lassen autoregulatory curve? All studies, either prospective or retrospective, of any size were included.
The primary outcome measure was the association between continuous CVR measures and the Lassen autoregulatory curve. 3 Continuous CVR measures were defined as those which are moving Pearson's correlation coefficients between slow-wave fluctuations in a driving pressure for CBF (i.e., either MAP or CPP) and a surrogate for pulsatile CBV/CBF, as defined in the existing literature body. 19,20 The following cerebral monitoring techniques were considered eligible for the derivation of continuous CVR metrics: ICP, NIRS, TCD, PbtO 2 , TDF CBF, or LDF CBF. Supplementary Appendix SB provides a table of the CVR metrics derived from these devices which were of interest in this review.
All studies, whether prospective or retrospective, of all sizes, including any animal model type that evaluated time-domain continuous metrics of pressurebased CVR and documented an association between the continuous metric and some aspect of the Lassen autoregulatory curve were eligible for inclusion in this review. Exclusion criteria were the following: being non-English-language studies, human studies, noncontinuous CVR assessments, and non-pressure-based CVR measures (i.e., chemoreactivity or CO 2 reactivity testing).
Search strategy MEDLINE, BIOSIS, EMBASE, Global Health, SCOPUS, and the Cochrane Library from inception to the end of January 2021 were searched using individualized search strategies for each database. The search strategies for all the databases can be seen in Supplementary Appendix SC of the Supplementary Materials. Finally, the reference lists of reviewed articles on CVR were examined to ensure that no references were left out.

Study selection
Using two reviewers (A.S.S. and L.F.), a two-step review of all articles returned by our search strategies was performed. First, the reviewers independently screened all titles and abstracts of the returned articles to decide whether they met the inclusion criteria. Second, the full text of the chosen articles was assessed to confirm whether they met the inclusion criteria, and that the primary outcome of CVR was documented. Finally, any discrepancies between the two reviewers were resolved by a third party (F.A.Z.).

Data collection
Data fields included the following: animal model details, the study's goal, aspects of Lassen autoregulatory curve assessed, primary/secondary outcomes, limitations, CVR indices measured, method of perturbation, and conclusions regarding continuous indices.

Bias assessment
Given that the goal of this review was to provide a comprehensive scoping overview of the available literature, a formal bias assessment was not conducted.

Statistical analysis
A meta-analysis was not performed in this study because of the heterogeneity of study designs and data.

Search results and study characteristics
Results of the search strategy across all databases and other sources are summarized in Figure 1. There were a total of 8513 articles identified from the databases searched. A total of 2408 articles were removed because of duplicated references, leaving 6105 articles to review. By applying the inclusion/exclusion criteria to the title and abstract of these articles, we identified 98 articles that fit these criteria. No articles were added from reference sections of pertinent review articles, which left 98 articles to review. On applying the inclusion/ exclusion criteria to the full-text documents, only 15 were found eligible for inclusion in the systematic review, all from the database search. Articles were excluded because they either did not report details around the association of time-domain continuous metrics and some aspect of the Lassen autoregulatory curve, were review articles, were non-animal literature, or were non-relevant given that they reported on nonpressure-based CVR measures.
Supplementary Appendix SD gives a general overview table of all 15 animal studies with primary and secondary outcomes of the study and its limitations. 19,20,25,[34][35][36][37][38][39][40][41][42][43][44][45] Tables 1, 2, and 3 describe three categories of perturbation methods: arterial hypotension/ hypertension, 19,25,[34][35][36][37][38][39][40][41]20,35,43,45 and other methods of inducing perturbation (such as cardiac arrest, hypothermia, etc.), 25,37,38,[40][41][42]44,45 respectively. Articles that used mixed methods will be found in more than one of Tables 1, 2, and 3 along with the study's conclusions regarding continuous indices. To measure CBF, 12 studies used LDF, 19,20,25,34,35,[38][39][40][41][42][43][44] two studies used laser Doppler index (LDx), 36,37 and one study used a diffusion coefficient measured with a custom-built diffusion coefficient spectroscopy. 45 All eligible studies reported the association between time-domain continuous metrics and the LLA aspect of the Lassen's autoregulatory curve. Only one study fully reconstructed the Lassen autoregulatory curve with both the LLA and ULA, but only related the continuous metrics to the curve's plateau (the region between LLA and ULA). 45 Figure 2 gives an illustration of variations in the autoregulatory curve noted in the studies along with indicating the LLA assessed in all the studies. Most of the studies measured multiple CVR indices and they include: six studies that measured cerebral-oximetry index (COx, correlation between regional cerebral oximetry [rSO 2 ] and CPP), 25,[34][35][36][37][38]40 one measured COx-a (correlation between rSO 2 and MAP), 36 one measured wavelet COx (wCOx, correlation between wavelet phase shift in rSO 2 and CPP), 25 seven studies measured hemoglobin volume index (HVx, correlation between relative total hemoglobin and MAP), 25,[38][39][40][41]43,44 one measured wavelet HVx (wHVx, wavelet correlation between relative total hemoglobin and MAP), 25 two measured LDx (correlation between laser-Doppler flux and MAP), 34,35 one measured mean flow index (correlation between mean flow velocity and CPP), 20 two measured pulse amplitude index (PAx, correlation between pulse amplitude of ICP), 19,20 10 studies measured pressurereactivity index (PRx, correlation between ICP and MAP), 19,20,25,35,37,39,[42][43][44][45] one measured induced PRx (iPRx, correlation between ICP and MAP with induced variations in MAP), 37 two measured wavelet PRx (wPRx, correlation between wavelet phase shift in ICP and CPP), 25,42 two measured the correlation between pulse amplitude of ICP and CPP (RAC), 19,20 one measured systolic flow index (correlation between systolic flow velocity and CPP), 20 and one measured MAP-ICP phase shift (Du AI ). 37 Eleven studies used neonatal swine 19,25,[34][35][36][37][38][39][40][41][42] as their animal model whereas two studies used juvenile domestic pigs, 43,44 one study used rhesus macaque, 45 and one study used New Zealand rabbits. 20 Hypotension/hypertension perturbation Most of the studies that induced hypotension did it by gradually inflating the balloon catheter in the inferior vena cava (Table 1), 19,25,[34][35][36][38][39][40][41] except for one where hypotension occurred with the gradual hemorrhage by syringe-pump withdrawal at a calculated rate of blood volume per hour. 37 From the 10 studies that induced hypotension, only three of them also  20,35,43 ; and curve C represents an autoregulatory curve during hypercarbia. 44 The LLA assessed is depicted in the figure by: ''I'' represents 14 studies 19,25,[34][35][36][37][38][39][40][41][42][43][44][45] ; ''II'' represents three studies 20,35,43 ; and ''III'' represents one study. 44 The asterisk (''*'') represents the ULA, which was not assessed in any study. CBF, cerebral blood flow; ICP, intracranial pressure; LLA, lower limit of autoregulation; MAP, mean arterial pressure; ULA, upper limit of autoregulation. induced hypertension by slowly inflating an aortic balloon catheter. 38,40,41 Overall, all the CVR indices that were measured in the hypotension model were able to distinguish whether CPP was above or below the LLA, and these indices included: COx, 25,[34][35][36]38,40,41 COx-a, 36 wCOx, 25 HVx, 25,38-41 wHVx, 25 LDx, 34,35 PAx, 19 PRx, 19,25,35,37,39 iPRx, 37 wPRx, 25 RAC, 19 and Du AI . 37 Results from studies show that during hypotension, PRx, COx, LDx, and HVx accurately detected CPP/ MAP above and below LLA. 25,34,35,[38][39][40][41] One study mentioned that LDx performed better than COx, and the agreement between these indices was limited but greatly improved with averaging values. 34 The increase in PRx to more positive values (denoting impaired CVR) had a blunted response at CPP values around LLA, when compared with autoregulation curves derived from COx and LDx. 35 In corollary, one study had shown that HVx is an excellent alternative to PRx. 39 In the same animal model, it has been shown that MAP-derived COx has an accuracy that is very much comparable to CPP-based COx, with COx-a having a slightly higher threshold for discriminating the LLA as compared to COx. 36 Variants of PRx displayed varying performance with regard to their availability measure aspects of the Lassen autoregulatory curve. During hypotension with mean pressures below LLA, the iPRx and Du AI were significantly lower as compared to normotension and hypotension above LLA where both iPRx and Du AI were not different. 37 Wavelet indices (wPRx, wCOx, and wHVx) did not increase as much as their corresponding correlation indices (PRx, COx, and HVx) did as MAP decreased below LLA. Also, all wavelet indices had lower variability than their correlation index counterparts given that the wavelet indices used a methodology to decrease signal noise. 25 Finally, non-PRx ICP-derived CVR indices were only sparingly mentioned. Strong conclusions about PAx and RAC cannot be made given that only one study evaluated them against PRx with a small number of animals, but the preliminary evidence shows that both of these indices validate against LLA within a model of hypotension whereas PRx seemed to be superior. 19 The results from the handful of studies that induced hypertension showed that the values of CVR indices measured (COx, HVx) were consistent with preserved autoregulation. 38,40,41 Of note, these studies did not successfully evaluate the ULA. Subsequently, we cannot comment on the ability of any of the described indices' ability to estimate aspects of the ULA at this time.

Intracranial pressure perturbation
There were only four studies on ICP perturbation, as seen in Table 2, and they induced ICP change using various methods, including: cerebrospinal fluid (CSF) infusion, 20,35 gradual inflation of a latex balloon catheter in the superior vena cava, 43 and influencing fluid to flow from a saline reservoir to lateral ventricle in the brain by lumbar catheter. 45 Under both naïve and elevated ICP conditions, similar values were obtained above LLA using PRx, COx, HVx, and LDx, 35,43 but two studies were able to relate high ICP to low CPP where CA becomes impaired 20,45 and both PRx and PAx correlated well with LLA. 20

Perturbation in pathological states
Other methods of inducing perturbation in the studies included cardiac arrest, 41 cardiac arrest with hypothermia, 25,38,40 hypercarbia, 44 and positive end-expiratory pressure (PEEP) oscillations as seen in Table 3. 37,42,45 Cardiac arrest was induced by hypoxic-asphyxic in all the related studies 25,38,40,41 whereas hypothermia was induced by cooling the temperature for a period of time or multiple short time periods. 25,38,40 PRx, wPRx, COx, wCOx, HVx, and wHVx indices were able to discriminate CPP/MAP above and below the LLA in cardiac arrest and hypothermia, 25,38,40 but it should be noted that the CPP/MAP change had to be driven to evaluate LLA taking the form of arterial hypotension or ICP increase in a brain-injured state attributable to cardiac arrest or hypothermia. One study reported that the accuracy of COx and HVx was slightly less in post-cardiac arrest animals. 40 There was only one study on hypercarbia, and CO 2 levels were elevated by bleeding CO 2 from a cylinder into the gas mixture inspired by the piglets. Despite the presence of hypercarbia, PRx and HVx were able to accurately detect the LLA, and the elevated CO 2 did not interfere with the NIRS readings. 44 Finally, PRx variants were evaluated during PEEP modulation. With PEEP modulation, oscillations in MAP were induced 37,42,45 and the precision of iPRx was improved compared to PRx. 37 The wPRx produced a more stable result than PRx while distinguishing CPP above and below the LLA. However, both indices increased significantly while CPP was decreased below LLA in the PEEP group. 42

Discussion
Through comprehensive evaluation of the pre-clinical literature surrounding continuous CVR measures and    Each piglet's laser Doppler flow-derived LLA was considered to be the gold standard to test whether autoregulation indices identified MAP above or below the LLA. All indices discriminated MAP above and below the LLA ( p < 0.001).
Distinguishing MAP above from below LLA did not differ between COx and wCOx or between PRx and wPRx whereas HVx had a significantly higher discriminatory value than wHVx in detecting dysfunctional autoregulation.
All the correlation indices became positive and approached +1 as blood pressure deviated farther below the LLA during induced hypotension. Wavelet indices increased as MAP decreased below LLA, but not as much as that observed with the correlation indices.
Mean values for all indices were each significantly lower when MAP was above LLA than below in paired comparisons ( p < 0.001 for each index).
All wavelet indices had lower variability than their corresponding correlation index. The wavelet methodology uses a coherence threshold to eliminate poorly coherent signals to decrease signal noise.

Hypercarbia
Nusbaum et al.   their ability to measure aspects of the autoregulation curve, some interesting findings deserve highlighting. First, the available literature lacked the inclusion of most of the CVR metrics of interest, provided in Supplementary Appendix SB. Thus, we can only comment on a select number of metrics based on ICP, NIRS, and TCD (to a limited extent). The handful of metrics evaluated by the eligible literature show that they can accurately detect CPP/MAP above the LLA and below the LLA. Thus, based on the available literature, end-users of such continuous measures can take some comfort in knowing that the listed metrics can estimate the LLA. However, all of the studies looked at the relationship between the continuous metrics and the LLA aspect of the Lassen autoregulatory curve, but none of the studies looked at relating the continuous metrics to the ULA aspect of the Lassen autoregulatory curve. Even though only one study could fully reconstruct the Lassen autoregulatory curve with LLA and ULA, it had only associated the continuous metrics to the intact autoregulation region of the curve. 45 Traditionally, the region between LLA and ULA has been described as a plateau, but emerging evidence suggests that the intact autoregulation region of the curve is not a plateau. 46,47 The study by Ruesch and colleagues 45 showed that fentanyl-anesthetized non-human primates had an averaged negative trend between the LLA and ULA, which had been generally regarded as the plateau. This highlights the need for future validation work for both the LLA and new studies to evaluate the ULA along with exploring the intact autoregulation region, particularly in different large animals given that responses may vary. Similarly, with the ever-increasing number of metrics emerging based on continuous multi-modal monitoring of cerebral physiology, new pre-clinical studies encompassing all multi-modal metrics concurrently while evaluating the LLA and ULA are required to get a better perspective on their ability to estimate the Lassen autoregulatory curve.
Second, regarding the NIRS-based potentially noninvasive metrics, some important aspects regarding their ability to discern the LLA were noted. A study comparing COx with LDx during arterial hypotension concluded that LDx performed better than COx 34 ; however, this should not be a surprise given that LDx is derived from LDF and the LLA was defined at the intersection of two regression lines on the LDF versus CPP scatterplot. The results from studies show that during hypotension, all the CVR metrics evaluated were able to detect CPP/MAP above and below LLA accurately. Among these studies, the commonly evaluated CVR indices were COx, PRx, and HVx, where HVx was mentioned to be an excellent alternative to PRx 39 and COx derived from either MAP or CPP had comparable accuracy. 36 Also, it has been shown that the wavelet counterparts of the three commonly evaluated indices had lower variability attributable to decreased signal noise. 25 Of the 10 studies categorized as arterial hypotension/hypertension, only three of them induced hypertension along with hypotension 38,40,41 whereas the rest only induced hypotension. 19,25,[34][35][36][37]39 The results from the hypertension studies showed that autoregulation was preserved, as indicated from COx and HVx values, 38,40,41 but, as mentioned above, evaluation of these CVR indices to estimate aspects of ULA cannot be commented on because ULA was not successfully evaluated.
Third, with respect to different ICP-derived measures, preliminary evidence shows that the newer ICP-derived CVR indices, PAx and RAC, can estimate the LLA. 19 However, PRx appears to remain superior in measuring the LLA in arterial hypotension models. 19 Similarly, PAx and PRx correlated well with LLA during intracranial hypertension, where RAC performed poorly. 20 Two other studies obtained similar values above LLA using PRx, COx, HVx, and LDx under both naïve and elevated ICP conditions. 35,43 Thus, despite some literature demonstrating the ability of PAx and RAC to estimate aspects of the LLA, their exact role beyond PRx in TBI monitoring remains unclear.
Fourth, some studies used other models of inducing perturbation where specific pathological states were studied; common models were cardiac arrest with hypothermia and PEEP oscillations to ICP and/or MAP. PRx, COx, and HVx, along with their wavelet counterparts, accurately discriminated CPP/MAP above and below LLA in the model of cardiac arrest and hypothermia. 25,38,40,41 With PEEP oscillations to modulate MAP, iPRx and wPRx gave improved and stable results of distinguishing CPP above and below LLA as compared to PRx. 37,42 There was only one study on hypercarbia where elevated CO 2 did not affect the ability of PRx and HVx to detect the LLA during gradual arterial hypotension induced by continuous hemorrhage. 44 Thus, it appears that the ability of PRx (and its variants), COx, and HVx to estimate the LLA is unaffected by the pathological state or co-manipulations of systemic physiology. Hence, it was important to review the pre-clinical literature given that such findings offer some confidence in their ability to perform as continuous measures at the bedside for TBI patients.
Finally, most of the articles reviewed looked at static aspects of CA. Static autoregulation measures refer to those metrics that are derived when both aspects of physiological measures (i.e., surrogate for pulsatile CBF/CBV and driving pressure for flow) have reached a steady state, where autoregulatory responses are assessed over a longer time frame of minutes to hours. 46 This is in contrast to dynamic autoregulation, where autoregulatory capacity is assessed in a much shorter temporal scale, on the order of seconds to minutes, in response to rapid changes in a driving pressure to flow. 46 However, despite most of the articles having evaluated static aspects of CA and some fundamental physiological differences between the techniques, the findings here carry potential importance for both static and dynamic autoregulation measurement techniques. There is evidence that points to the general agreement of static and dynamic aspects of CA in impaired autoregulation under challenges in ICP and blood pressure. 34,45,48 In normal anesthetized adults with intact or impaired autoregulation, 48 and in non-human primates under anesthesia-induced autoregulation impairment, 45 it has been demonstrated that there is a good correlation between both techniques to assess CA, although dynamic CA seems to be initially more reduced than static CA. Such an agreement between static and dynamic CA would be consistent given that the more momentary physiological factors assessed by dynamic CA would be sufficiently identical to the longer methods of assessment in static CA. However, in situations where paroxysmal physiological responses occur, static autoregulation may be unaffected and thus a disagreement between dynamic and static CA assessments. This enforces the idea that determination of CA must be validated through continuous methods.

Limitations
Despite the interesting findings outlined in this scoping review, there are some significant limitations that deserve highlighting. First, the uncovered literature was very heterogeneous in design, which means that there was limited ability to cross-sectionally evaluate the relationship between studies based on various combinations of CVR indices used in each method of perturbation. Further to this, the different animal models and species utilized limit the comparability between similar indices and perturbation results. Second, most studies used a small number of animals in their model design. This is likely secondary to the cost of such models and experiments. Such small numbers limit the definitiveness of the findings outlined and emphasize the need for further validation studies. Third, some models described used neonatal animals. This raises concerns of cerebrovascular immaturity and translatability of findings to adult animals or humans. Fourth, from the numerous multi-modal cerebral physiological monitoring devices noted in bedside monitoring to derive continuous CVR indices, 19,20 our review highlights that only a few have been studied pre-clinically. Thus, we cannot definitively indicate the utility of TCD, PbtO 2 , and HemedexÔ CBF-based techniques in their ability to measure the autoregulatory curve.
Fifth, we were unable to find literature documenting the ability of such continuous CVR metrics to estimate the ULA, which could be attributable to the occurrence of heart failure at higher CPP as mentioned by a couple of studies. As such, at this time, it is unknown whether such a metrics measure can discriminate more than just the LLA. Finally, and most important, the results from studies with animal models do not directly translate to human patients, though preliminary findings in humans support their ability to estimate the LLA. 49,50 Future directions Moving forward, there are some important areas for future research. First, validation of the above findings is required, given that most are based on singleinstitution, single-study findings in a limited number of animals per study. Such work will require specialized centers for large animal research, dedicated to precision medicine in neural injury. These large animal models would be best positioned to facilitate translatability to adult human neurocritical care. Second, studying the performance of such indices in both healthy-and injury-state models is key to be able to differentiate between intact and dysfunctional autoregulation using each of the indices. Given that different neuropathological states may cause different changes to the shape and nature of the autoregulatory curve, having control preclinical models in the healthy state are critical to our understanding of fundamental changes in cerebral autoregulation that may occur in disease.
Further, in providing validation for a continuous surrogate measure of CA, we must have the utmost confidence in their ability to estimate various aspects of autoregulation during both the healthy and diseased states before they are adopted into routine clinical-care provision. This requires, again, specialized center expertise in model development, multi-modal cerebral physiological monitoring, and biomedical engineering capability. Such models would benefit from having both focal and diffuse injury patterns, including those modeling stroke, TBI, and anoxic brain injury. Third, work here would be bolstered by evaluating differences between both neonatal and adult models. Existing literature supports the use of pig models to facilitate such work. 51 Fourth, integrating physiological monitoring with histopathological correlations could improve our fundamental understanding of tissue consequences of impaired CA, based on different monitoring modalities, across different model ages and across the spectrum of neuropathological states. Finally, such large animal-model platforms would provide the ideal setting to explore directed therapeutics aimed at prevention and treatment of impaired CA, facilitating the needed pre-clinical investigation of precision therapeutics before application in humans.

Conclusion
This literature demonstrates that most of the measured indices (COx, COx-a, wCOx, HVx, wHVx, LDx, PAx, PRx, iPRx, wPRx, RAC, and Du AI ) showed the ability to distinguish CPP/MAP from above and below the LLA with various methods of perturbation in animal models. However, most studies focused on a small number of animal populations, with only a few simulating ICP elevations observed in TBI, so the conclusions drawn from these studies should be taken with caution. Also, none of the indices evaluated the ULA aspect of the Lassen autoregulatory curve. Therefore, further research is required to fully assess the relationship between both the LLA and ULA aspects of the Lassen autoregulatory curve to CVR indices in a larger pre-clinical population size.