Monitoring of cerebral blood flow autoregulation: physiologic basis, measurement, and clinical implications

Summary

Cerebral blood flow (CBF) autoregulation is the physiologic process whereby blood supply to the brain is kept constant over a range of cerebral perfusion pressures ensuring a constant supply of metabolic substrate. Clinical methods for monitoring CBF autoregulation were first developed for neurocritically ill patients and have been extended to surgical patients. These methods are based on measuring the relationship between cerebral perfusion pressure and surrogates of CBF or cerebral blood volume (CBV) at low frequencies (<0.05 Hz) of autoregulation using time or frequency domain analyses. Initially intracranial pressure monitoring or transcranial Doppler assessment of CBF velocity was utilised relative to changes in cerebral perfusion pressure or mean arterial pressure. A more clinically practical approach utilising filtered signals from near infrared spectroscopy monitors as an estimate of CBF has been validated. In contrast to the traditional teaching that 50 mm Hg is the autoregulation threshold, these investigations have found wide interindividual variability of the lower limit of autoregulation ranging from 40 to 90 mm Hg in adults and 20–55 mm Hg in children. Observational data have linked impaired CBF autoregulation metrics to adverse outcomes in patients with traumatic brain injury, ischaemic stroke, subarachnoid haemorrhage, intracerebral haemorrhage, and in surgical patients. CBF autoregulation monitoring has been described in both cardiac and noncardiac surgery. Data from a single-centre randomised study in adults found that targeting arterial pressure during cardiopulmonary bypass to above the lower limit of autoregulation led to a reduction of postoperative delirium and improved memory 1 month after surgery compared with usual care. Together, the growing body of evidence suggests that monitoring CBF autoregulation provides prognostic information on eventual patient outcomes and offers potential for therapeutic intervention. For surgical patients, personalised blood pressure management based on CBF autoregulation data holds promise as a strategy to improve patient neurocognitive outcomes.

Editor’s key points.

• •Cerebral blood flow autoregulation maintains cerebral blood over a range of cerebral perfusion pressures, which is critical to preserving brain function at extremes of arterial blood pressure.
• •Various methods are available for monitoring autoregulation experimentally, but widespread clinical application awaits more practical approaches.
• •Studies of perioperative cerebral autoregulation demonstrate wide interindividual variability in the lower limit of autoregulation of 40–90 mm Hg in adults and 20–55 mm Hg in children.
• •Impaired CBF autoregulation is associated with adverse outcomes in traumatic brain injury, ischaemic stroke, subarachnoid haemorrhage, intracerebral haemorrhage, and surgical patients.
• •Personalised blood pressure management based on cerebral blood flow autoregulation is a promising strategy to improve perioperative neurocognitive outcomes.

While monitoring blood pressure in critically ill and surgical patients is standard of care, defining hypotension is without universal consensus.1, 2, 3 A mean arterial pressure (MAP) of at least 50 mm Hg has been widely viewed as a clinically acceptable blood pressure in many instances. This is based in part on doctrine established by Lassen⁴ in the 1950s who suggested that 50 mm Hg is the lower limit of cerebral blood flow (CBF) autoregulation (or LLA). Reporting of intraoperative MAP values <50 mm Hg has, in fact, been proposed as an intraoperative quality metric.⁵ In critically ill patients, recommendations for arterial pressure management for many conditions are based on observational data or randomised studies that have limitations because of their lack of consideration of the individual and dynamic nature of CBF autoregulation.6, 7, 8, 9 Defining the upper tolerable blood pressure in surgical or critical care settings has been even less well studied.

Monitoring of CBF autoregulation at the bedside is possible using methods that are scalable.^10,11 Consequently, these methods might be utilised clinically as a means for individualising arterial pressure management. In this review, we discuss the physiologic mechanisms of CBF autoregulation, methods for clinical autoregulation monitoring, and observational data linking impaired autoregulation and adverse patient outcomes. We further discuss interventional studies that support autoregulation monitoring as a strategy for optimising perioperative perfusion pressure.

Mechanistic regulation of cerebral blood flow

Cerebral blood flow-pressure autoregulation broadly refers to the mechanisms by which brain blood flow is regulated across changes in cerebral perfusion pressure (CPP).⁴ MAP is oftentimes used as a surrogate for CPP with the assumption that intracranial pressure (ICP) or central venous pressure (CVP) are normal or low. Below the LLA, CBF decreases and becomes pressure-dependent. Above the upper limit of autoregulation (ULA), CBF is also pressure-dependent increasing in response to elevated CPP. Between the LLA and ULA, CBF remains relatively constant despite changes in CPP. Since the original description of autoregulation by Lassen⁴ in 1959, more contemporary data suggest that the plateau region is narrower than previously thought, and there is a greater slope of the ΔCBF/ΔCPP relationship below the LLA compared with the slope of ΔCBF/ΔCPP above the ULA (Fig. 1).¹² Since Lassen's pivotal studies, data have emerged demonstrating wide interindividual variability in the LLA ranging from 30 to 90 mm Hg in adults and 20–55 mm Hg in children during cardiopulmonary bypass (CPB).^13,14 These data are consistent with earlier human and animal studies that reported that the LLA varies between 33 and 113 mm Hg.¹⁵

Fig 1.

Autoregulation depicting relationship between cerebral blood flow (CBF) and cerebral perfusion pressure or mean arterial blood pressure. The left is the representation of the autoregulation curve described by Lassen⁴ in 1959 based on between-subject analysis of patients during pharmacological interventions and pathologies. The right is a contemporary re-conceptualised version of the autoregulation curve with a smaller plateau region showing a %ΔCBF/%ΔMAP relationship below the lower limit of autoregulation (LLA) of 0.81 (0.77) compared with a %ΔCBF/%ΔMAP relationship above the ULA of 0.21 (0.47).¹² ULA, MAP at the upper limit of cerebral blood flow autoregulation. Figure adapted from Willie and colleagues¹⁴⁸ with permission.

The physiological processes that maintain CBF commensurate with metabolic demand are complex.¹⁶ When interpreted in the context of autoregulation, there are multiple mechanisms involved with variable contributions. Regulation of CBF is thought to be mediated by a number of factors including:

• •Myogenic mechanisms. These mechanisms involve the response of vascular smooth muscle cells which constrict or dilate in response to changes in transmural pressure. The myogenic response is initiated by vessel wall tension resulting in a negative feedback response leading to vasoconstriction. The mechanism involves membrane depolarisation, which activates voltage-gated calcium channels; the influx of calcium into the smooth muscle cell activates myosin light chain kinase which in turn phosphorylates myosin light chain and leads to vasoconstriction.¹⁷ Studies suggest that myogenic mechanisms of CBF autoregulation are not only complex, but also mediated by several types of potassium channels, chloride channels, stretch-activated cation channels, and transient receptor potential (TRP) channel homologs (TRPC6).^18,19
• •Neurogenic mechanisms. Neurogenic mechanisms for CBF regulation refer to the processes by which neurones and glia regulate smooth muscle function via adrenergic and cholinergic fibres releasing various neurotransmitters. Neurogenic control of the cerebral vasculature is suggested to be an important factor in autoregulation, particularly in large vessels. An important role of neurogenic autoregulation is to buffer acute increases in perfusion pressure via sympathetic nervous system activity (preventing hyperperfusion).20, 21, 22 In animal models, parasympathetic stimulation is associated with an increase in CBF, and sympathetic stimulation or parasympathetic denervation results in decreased CBF and a shift of the LLA towards higher blood pressures.23, 24, 25 These neurogenic mechanisms contribute to fine-tuning of blood flow and play a key role in regional variability of CBF. There are complex functional and anatomical characteristics between areas of cerebral circulation involving sympathetic/parasympathetic reflexes and receptors that are ongoing areas of study, and beyond the scope of this review.²⁶
• •Endothelial mechanisms. Local endothelial factors modulate vascular tone and thus autoregulation. The vasoconstrictors thromboxane A₂ and endothelin-1 and the vasodilator nitric oxide are important regulators of CBF. Endothelial factors can be released by physical stimuli (shear stress or haemorrhage), neurotransmitters, or cytokines.27, 28, 29
• •Metabolic mechanisms. Metabolic mechanisms of CBF regulation involve variations of metabolic activity that adjust blood flow. While the effects of Paco₂, Pao₂, and pH are well described in altering CBF, the overall effects on autoregulation are complex and inconsistent across experimental models of autoregulation. Studies have indicated that each mm Hg increase in Paco₂ above normal corresponds with a 3–6% increase in CBF, while each mm Hg decrease in Paco₂ corresponds with a 1–3% reduction in CBF.³⁰ Below a Pao₂ of 50 mm Hg, CBF increases as a result of a reduction of vascular smooth muscle tone via increased activation of membrane potassium channels and inhibition of transmembrane calcium flux.31, 32, 33 With hypercapnia, the CBF response to hypoxemia is accentuated.³⁴ Cerebral vascular smooth muscle cells constrict with increased pH and relax with decreased pH.³⁵ While cerebrovascular vessels generally dilate with increased Paco₂, acidosis, or decreased Pao₂, the neurovascular unit has complex relationships between neurones, perivascular astrocytes, endothelial cells, and the extracellular membrane.

There are numerous studies that have described the effects of volatile and intravenous anaesthetics on CBF.36, 37, 38, 39, 40, 41 Volatile anaesthetics can influence CBF autoregulation and have been shown to shorten the autoregulatory plateau at concentrations ≥1 minimum alveolar concentration (MAC).^11,42 The effects of temperature on autoregulation have been questioned and are particularly relevant during cardiac surgery when deep hypothermia is utilised. In a neonatal piglet model, the LLA is similar in normothermia and hypothermia (20°C).⁴³ Prior clinical studies on this topic have likely been confounded because of the association with hypotension in states of deep hypothermia.

Measurement and indices of autoregulation

There are a variety of methods to measure CBF autoregulation, but some disagreement about measurement methods exists.⁴⁴ A distinction between CBF autoregulation measurement (i.e. assessment over a finite time period) and clinical monitoring (i.e. continuous assessment) should be considered. The latter affords the opportunity for a modifiable target, such as clinical alterations in blood pressure or ICP. The two broad categories of CBF autoregulation assessment are static and dynamic measurements. Static refers to longer-term fluctuations in response to changes in MAP or ICP as represented by Lassen's model.^4,45 Static autoregulatory changes are the changes that occur over minutes to hours.⁴⁶ Dynamic refers to the transitory fluctuations in CBF in response to MAP changes. The dynamic autoregulatory response can be measured using models proposed by Aaslid and colleagues⁴⁷ and Tiecks and colleagues⁴⁸ where rapid deflation of a blood pressure cuff on the thigh is utilised to induce a response change in CBF velocity (measured by transcranial Doppler). In this approach, CBF autoregulation is quantified by determining the best fit response of changes in CBF velocity to predetermined modelled velocity responses (nine potential stepwise models). The resulting stepwise regression models termed ‘autoregulatory index’ range from zero (no autoregulation) to nine (intact autoregulation). In a comparison between dynamic and static methods by Tiecks and colleagues,⁴⁸ both dynamic and static measurements of autoregulation ultimately yield similar results. Another technique to measure CBF autoregulation is the pressure reactivity index (PRx) described by Czosnyka and colleagues.⁴⁹ In comparison to the models by Aaslid and colleagues⁴⁷ and Tiecks and colleagues,⁴⁸ PRx does not require dynamic blood pressure cuff deflation, nor does it require assumptions of critical closing pressure values. The PRx method has been considered as a hybrid pseudodynamic measurement of CBF autoregulation as it measures both static and dynamic changes.⁴⁶

The PRx CBF autoregulation monitoring method is based on analysis of naturally occurring ICP waves (i.e. Lundberg ‘B waves’) that occur at a frequency of 0.5–2 per min.⁵⁰ A variety of waveform analytics in both frequency and time domains have been applied to quantify the relationship between ICP and CPP waveforms. When autoregulation is intact, CPP is on the autoregulatory plateau whereby the transmission of CPP to ICP is phase shifted resulting in negative correlation. Conversely, when CPP is below the LLA or above the ULA (a state of vasomotor passivity) the transmission of CPP to ICP is in phase resulting in a positive correlation. Methods of autoregulation monitoring use either correlation or phase shift analytics to discriminate intact or impaired autoregulation with varying sensitivity and specificity. Transfer function analyses utilise spectral estimates in the frequency domain of CPP/MAP and CBF velocities to obtain estimates of the coherence function, gain, and phase frequency responses. To calculate autoregulation metrics with transfer function analysis, it has been recommended that spontaneous fluctuations of pressure measurements occur at a minimum of 5 min for sufficient frequency resolution with time windows ≥100 s.⁵¹ However, in a study comparing the various frequency and time domain methods in a piglet hypoxic-ischaemic brain injury model, correlation and phase-based indices outperformed frequency methods such as gain of transfer and coherence in delineating the LLA.⁵² In that study, autoregulation curves were generated by graphing CBF (y-axis) measured with laser Doppler flux vs CPP (x-axis). Piecewise linear regression was utilised to fit the data, and the LLA was defined as the CPP at which the two regression lines intersect. In this important study, the gain of transfer and coherence methods had receiver operator characteristic (ROC) area under the curve (AUC) values of 0.5–0.6 and 0.6–0.75, respectively. Phase shift and correlation methods demonstrated ROC AUC values of 0.8–0.9 and 0.8–0.95, respectively. The best discriminator of LLA was when correlation methods were utilised with relative total haemoglobin measurements (see below), resulting in ROC AUC values >0.9.

Correlation techniques are the most studied and effective in distinguishing intact vs impaired pressure autoregulation. In animal models, this method has delineated the limits of autoregulation with high sensitivity and specificity.53, 54, 55, 56, 57 These methods utilise correlation between CPP and surrogates of CBF or CBV at the frequency of slow waves (0.003–0.05 Hz). A positive correlation occurs when CPP is outside the limits of the autoregulatory plateau.⁵⁸ The optimal frequency response to measure autoregulation has been determined in experimental models to occur at waves of 0.017 Hz with a CPP-ICP phase shift rendering a cut-off frequency of 1.8 min⁻¹ when a Butterworth filter was applied.⁵⁸ Thus, CBV is an indicator of cerebral vasoreactivity that mediates CBF. Vasoconstriction in response to rising CPP leads to reduced CBV, and vasodilation in response to falling CPP leads to an increased CBV. As above, a moving correlation coefficient between ICP and CPP as an index of autoregulation is the PRx.⁵⁹ Similarly, mean velocity index (Mx) correlates CBF velocity measured by transcranial Doppler and CPP.¹⁰ However, continuous transcranial Doppler measurements can be challenging to obtain in clinical environments.

Use of cerebral near infrared spectroscopy was validated as a practical method of CBF autoregulation monitoring. When processed regional cerebral oxygenation (%rSO₂) signals are utilised as a surrogate of CBF, the autoregulation index is termed the cerebral oximetry index (COx) or tissue oxygenation index (TOx).^53,60 Subsequently, an index of autoregulation was developed using near infrared spectroscopy technology where the isosbestic wavelength of deoxy- and oxy-haemoglobin (805–810 nm) is used to estimate the relative total haemoglobin which serves as an indicator of arteriolar vasoreactivity and CBV. That is, total haemoglobin concentration varies with arteriolar size. The latter measure of autoregulation is termed the haemoglobin volume index (HVx).^53,61 In a piglet CPB model, the LLA measured with HVx corresponded to the LLA measured by laser Doppler and brain tissue oxygen tension. The thresholds for defining the LLA for Mx, COx, and HVx have been determined in a study of 59 patients undergoing cardiac surgery. Autoregulation curves were plotted between transcranial Doppler measured CBF velocity of the middle cerebral artery vs MAP. Using the reference LLA from this comparison, the LLA thresholds for Mx, COx, and HVx were determined to be 0.45, 0.35, and 0.30, respectively.⁶² Autoregulation monitoring/signal processing can be achieved through commercially available programs (ICM+, Cambridge University; Moberg CNS Monitor, Moberg Research Inc.; Sickbay, Medical Informatics Corp.; Matlab, Mathworks Inc.) to render autoregulation indices described in the literature.^10,53,60,63

Considerations for CBF autoregulation-based arterial pressure management

While Lassen⁴ referenced MAP in his seminal review on CBF autoregulation, it should be emphasised that CPP is the determinant of brain perfusion. High ICP not only impedes cerebral perfusion, but can also lead to a rightward shift of the LLA. Elevated CVP, especially relevant to congenital heart disease, is associated with a rightward shift of the LLA.63, 64, 65, 66

In conscious subjects, CBF correlates with cardiac output.⁶⁷ This may have relevance when cardiac output is elevated (e.g. sepsis, liver failure, inotropic drugs) or reduced (e.g. heart failure). The setting of CPB provides an experimental model where systemic flow (equivalent of cardiac output) can be manipulated without administration of vasoactive drugs that can potentially confound CBF autoregulation studies. During CPB, CBF autoregulation remains functional with varying systemic flows when alpha-stat pH management is used.⁴² Those findings suggest independence of CBF autoregulation to blood flow. Other physiologic factors thought to impair CBF autoregulation include Paco₂. In laboratory experiments, hypercarbia in the absence of acidosis or elevated ICP is associated with a rightward shift of the LLA that can lead to impaired CBF autoregulation. There was correlation between Paco₂ and the LLA (Spearman correlation, 0.8243, P<0.0001). The LLA was 75 mm Hg (95% confidence interval [CI], 73–77 mm Hg) when Paco₂ was 80 mm Hg compared with 45 mm Hg (95% CI, 43–47 mmHg) in the normocarbic group.⁶⁸

The concept of the brain serving as the ‘index organ’ to ensure perfusion to other organs particularly during cardiac surgery has been suggested because of a paucity of monitors to detect perfusion to most organs including the splanchnic circulation.⁶⁹ Use of CBF autoregulation metrics to guide to blood pressure management, however, must be interpreted in the context of adequate perfusion to other organs. For example, in a piglet model of haemorrhagic shock, renal and CBF autoregulation were both present at baseline. Renal blood flow, however, declined during haemorrhage even at renal perfusion pressures previously associated with autoregulation. In contrast, CBF was maintained as long as blood pressure was greater than the LLA. Renal blood flow declined to 75%, 50%, and 25% of baseline when MAP was 60, 45, or 40 mm Hg, respectively. Similar reductions in CBF did not occur until MAP was 30, 25, and 15 mm Hg, respectively.⁷⁰ These experiments support that the kidney depends on both blood flow (i.e. cardiac output or CPB flow) and blood pressure for effective perfusion.

Perfusion to other organs can be further compromised at perfusion pressures >LLA. A patient with critical coronary artery stenosis, for example, might require a higher blood pressure than that needed to ensure CBF. Evidence of myocardial ischaemia with lowered arterial pressure would provide clinical evidence that a higher MAP is needed. In congenital heart surgery, afterload reduction has been demonstrated to be a key factor in improving outcomes, reducing circulatory collapse, and increasing survival.71, 72, 73 There is a fine balance between optimising blood pressure targets above LLA while ensuring adequate afterload reduction in the neonatal/paediatric heart after CPB supporting the potential benefits of bedside CBF autoregulation monitoring.

Impaired CBF autoregulation and patient outcomes

Impaired cerebral autoregulation can result from the direct cerebral vascular consequences of central nervous system disease or from CPP falling outside the limits of autoregulation. The latter can result in central nervous system ischaemia, as CBF is CPP dependent. Inherent physiologic brain reserve, though, can protect the brain from ischaemic injury despite a CPP<LLA. Normal global CBF is 50–60 ml 100 g⁻¹ min⁻¹, but laboratory studies suggest the EEG does not become isoelectric until CBF is 15 ml 100 g⁻¹ min⁻¹, and neuronal damage does not occur until CBF is 6–10 ml 100 g⁻¹ min⁻¹.⁷⁴ In non-anaesthetised humans, symptoms of cerebral ischaemia are reported not to occur until CBF is reduced 35%–40% of baseline.⁷⁵

Hypertension affects nearly 30% of the world's population and >60% of adult patients undergoing cardiac surgery.76, 77, 78 Impairment of CBF autoregulation has been demonstrated in patients with hypertension which can result from vascular smooth muscle or endothelial dysfunction.^79,80 This CBF autoregulation impairment can normalise when blood pressure is restored to a normal range.⁸¹ Cerebral small vessel disease resulting from chronic hypertension can compromise the brain's physiologic reserve reducing the safety margin of CPP<LLA. Cerebral small vessel disease manifests on brain MRI as white matter hyperintensities. This finding is present on brain imaging in >50% of people >65 yr of age and nearly all people >80 yr.⁸² In non-surgical populations, cerebral small vessel disease is linked to 30% of ischaemic strokes, 80% of intracerebral haemorrhage,^83,84 45% of dementia cases, late-onset refractory depression, and significant functional impairment.^85,86 Cerebral small vessel disease is associated with reduced vascular reactivity to CO₂, an indicator of autoregulatory dysfunction.⁸⁷ Brain MRI white matter hyperintensities are associated with impaired CBF autoregulation during cardiac surgery and perioperative stroke.⁸⁸ A growing body of data link impairment of CBF autoregulation and adverse patient outcomes. Monitoring CBF autoregulation could potentially provide a means for assessing the severity of disease and effectiveness of treatments.

Acute neurological conditions

The association between impaired CBF autoregulation and adverse patient outcomes was demonstrated early for patients with traumatic brain injury (TBI). The use of ICP monitoring in affected patients provides the opportunity to assess CBF autoregulation by correlating changes in ICP with CPP (PRx). This work focused on identifying the optimal CPP (CPP_OPT) where autoregulation indices are the lowest (i.e. indicating the CPP for ‘best’ autoregulation).^10,63,66,89 Aries and colleagues⁹⁰ reported a significant association between the degree of CPP<CPP_OPT (indicating cerebral hypoperfusion) and mortality in a study of 327 patients with TBI. In contrast, the degree of CPP>CPP_OPT is associated with poor functional outcomes. In a systematic review of 20 publications and 924 patients, PRx, Mx, and autoregulation reactivity index predicted mortality and Glasgow Outcome Scale score for patients with TBI.⁹¹

Impaired autoregulation has been demonstrated after ischaemic stroke as a result of multiple mechanisms including endothelial dysfunction, blood–brain barrier disruption, and neuroinflammation.⁷⁹ Interestingly, CBF autoregulation was observed to be functional early after stroke but deteriorated 4–7 days later.^79,92 This early preservation of CBF autoregulation was associated with smaller cerebral infarct volume and better functional outcomes compared with patients having impaired autoregulation.^79,93 Persistent impaired CBF autoregulation after an ischaemic stroke was associated with cerebral oedema and risk for haemorrhagic transformation.^79,94 Appropriate administration of thrombolytics early after stroke is associated with preserved CBF autoregulation.^79,95

The effects of ischaemic stroke on CBF autoregulation might be impacted by stroke subtype. Patients with large-vessel stroke demonstrated impaired CBF autoregulation to the affected hemisphere that normalises over a period of several weeks.⁹⁶ In contrast, patients with lacunar strokes demonstrate impaired CBF autoregulation in both hemispheres, which is consistent with the presence of pre-existing small cerebral vessel disease.^79,97 Stroke as a result of intracranial arterial stenosis is associated with impaired CBF autoregulation in the vascular territory of the arterial stenosis but not in unaffected areas; in patients with hypoperfusion-related stroke, this impairment is more widespread.^79,98 Compromised CBF to the penumbral tissue resulting from impaired CBF autoregulation is of great concern. Brain tissue oxygen saturation measurement in relation to changes in CPP demonstrated that patients with functional CBF were more likely to have an uneventful clinical course after stroke than patients with impaired autoregulation.^79,97

Impaired CBF autoregulation can occur early after subarachnoid haemorrhage for many reasons including pre-existing hypertension, raised ICP, neuroinflammation, endothelial dysfunction, and autonomic nervous system imbalance.^79,99,100 Its presence is associated with vasospasm, delayed cerebral ischaemia and other adverse outcomes including worse Glasgow Outcome Scale score.^79,91,99, 100, 101, 102 The presence of impaired CBF autoregulation had a 92% positive predictive value and 77% negative predictive value for angiographic vasospasm.^79,102 In a study of 80 patients with acute subarachnoid haemorrhage, impaired CBF autoregulation averaged over 8 days after the event was predictive of poor 6-month neurological outcome.¹⁰³ For both TBI and subarachnoid haemorrhage, the predictive capacity of CBF autoregulation monitoring increases with duration of monitoring.⁹¹ In a study of 31 patients with subarachnoid haemorrhage, CBF autoregulation metrics based on ICP and near infrared spectroscopy was used to devise personalised haemodynamic goals.¹⁰⁴ Deviation of CPP outside of the optimal autoregulation goals was associated with modified Rankin Scale score ≥3 at both hospital discharge and 90 days later. Continuous CBF autoregulation monitoring after intracerebral haemorrhage demonstrated bilateral impairment that persisted for up to 12 days after the event recovering 1 month later.¹⁰⁵ Haematoma volume was shown to be predictive of ipsilateral impaired CBF autoregulation.¹⁰⁶ Studies of patients with intracerebral haemorrhage have linked impaired CBF autoregulation with poor 90-day neurological outcomes.^79,107

Intraoperative autoregulation monitoring

Results from observational studies of CBF autoregulation monitoring in patients undergoing cardiac and noncardiac surgery are listed in Table 1, Table 2. Autoregulation metrics are based on MAP during CPB because CVP is kept low by gravitational or suction-assisted venous drainage to the circuit reservoir. The link between MAP below the LLA and evidence of other organ hypoperfusion supports the need for individual blood pressure management. In patients undergoing cardiac surgery, there is a relationship between excursions of MAP below the LLA during CPB and acute kidney injury, risk for mortality, and major morbidity.^108,109 In addition, a relationship between impaired autoregulation during CPB and perioperative stroke has been demonstrated.¹¹⁰ Others have shown a link between abnormal CBF autoregulation metrics and postoperative delirium or neurocognitive dysfunction after surgery.111, 112, 113

Table 1.

Observational studies that have evaluated the relationship between cerebral blood flow autoregulation metrics and patient outcomes after cardiac surgery. The study size (N), method of cerebral autoregulation monitoring, patient exposure, and main outcomes are listed. CI, confidence interval; COx, Pearson correlation between slow waves of regional cerebral oximetry and mean arterial pressure; CPB, cardiopulmonary bypass; LLA, MAP at the lower limit of cerebral blood flow autoregulation; MMOM, composite outcome of major morbidity or mortality defined as mortality within 30 days of surgery, stroke, renal failure mechanical lung ventilation >48 h; inotrope use >24 h, intra-aortic balloon pump insertion; MAP, mean arterial pressure; Mx, Pearson correlation between slow waves of transcranial Doppler and mean arterial pressure; TCD, ; ULA, MAP at the upper limit of cerebral blood flow autoregulation.

Study	N	Exposure	Main outcome
Ono and colleagues110	234	Impaired autoregulation (mean velocity index ≥0.4 at all MAPs during CPB	Impaired autoregulation occurred in 20% of patients; stroke more common in patients with impaired autoregulation (P=0.011)
Ono and colleagues108	410	Duration and magnitude MAP<LLA determined with COx during CPB	Acute kidney injury after surgery predicted by excursions of MAP<LLA (relative risk, 1.02; 95% CI, 1.01–1.3, P<0.0001)
Ono and colleagues109	450	Duration and magnitude MAP<LLA determined with COx during CPB	Area under curve of MAP<LLA during CPB associated with MMOM (odds ratio, 1.36, 95% CI, 1.08–1.71, P=0.008)
Hori and colleagues116	491	Duration and magnitude MAP>ULA determined with COx during CPB	Area under curve of MAP>ULA during CPB associated with risk for clinical delirium (1.09, 95% CI, 1.03–1.15, P=0.001)
Hori and colleagues106	121	Duration and magnitude MAP less than that MAP associated with best autoregulation (i.e. optimal MAP) determined with COx in the intensive care unit after surgery	Duration and magnitude that MAP < optimal MAP after surgery associated with release of plasma brain injury biomarker glial fibrillary acid protein
Caldas and colleagues111	67	Autoregulation index (transfer function between TCD and MAP)	Lower autoregulation index measured over 5-min intervals before surgery and 24 h after surgery lower in patients with vs without postoperative delirium (P<0.001)
Kumpaitiene and colleagues112	59	TCD technological method for monitoring cerebral blood volume vs MAP	Duration of impaired cerebral blood flow autoregulation during CPB longer in those with vs those without delayed neurocognitive recovery (odds ratio for >5 min, 14.5, 95% CI, 3.9–51.8)

Table 2.

Retrospective studies that have evaluated the relationship between cerebral blood flow autoregulation metrics and patient outcomes after noncardiac surgery. The study size (N), method of cerebral autoregulation monitoring, patient exposure, and main outcomes are listed. CBF, cerebral blood flow; CI, confidence interval; COx or TOx, Pearson correlation between slow waves of regional cerebral oximetry and mean arterial pressure; MAP, mean arterial pressure; Mx, Pearson correlation between slow waves of transcranial Doppler and mean arterial pressure; NIRS, near infrared spectroscopy.

Study	N	Exposure	Main outcome
Laflam and colleagues117	218	COx monitoring during shoulder surgery	COx was higher (i.e. impaired autoregulation) in patients having surgery in beach chair vs lateral decubitus position adjusted for age and history of hypertension (P=0.035). No differences in neurocognitive outcomes were observed between positioning groups
Goettel and colleagues113	82	Patients ≥65 yr old undergoing major noncardiac surgery Mx and TOx monitoring	Higher intraoperative Mx (i.e. impaired autoregulation) increased the odds of delayed cognitive recovery 1 week after surgery in univariate but not multivariate logistic analysis (odds ratio, 1.39; 95% CI, 1.01–1.90, P=0.08)
Kahl and colleagues120	195	Surgery for visceral, urologic, or gynaecological cancer	Time-correlation method of CBF autoregulation monitoring with NIRS. Delayed cognitive recovery before hospital discharge associated with impaired CBF autoregulation (odds ratio, 1.042, 95% CI, 1.005–1.080, P=0.028)
Chuan and colleagues119	140	Major noncardiac surgery monitored with TOx	Impaired autoregulation associated with lower postoperative quality of recovery scale results on postoperative day 3 (P=0.02). TOx >0.1 identified 72.7% of patients with delayed cognitive recovery at day 3 after surgery (odds ratio, 3.3, 95% CI, 1.1–9.9, P=0.03) and major adverse events (odds ratio, 4.7, 95% CI, 1.3–17.2, P=0.02)

In addition to blood pressure below the LLA, CPP above the ULA can expose patients to brain injury from mechanisms similar to that seen with hypertensive emergency and associated encephalopathy. In that condition, cerebral hyperaemia results when autoregulatory mechanisms are exhausted leading to endothelial damage, breakdown of the blood–brain barrier, and vasogenic oedema.^114,115 The magnitude and duration that MAP is above the ULA during CPB is associated with clinically detected delirium after surgery.¹¹⁶ Consequently, simply increasing empirical MAP targets during CPB can have unintended consequences of brain injury from MAP above the ULA.

In a study of CBF autoregulation in patients undergoing shoulder surgery in either the beach chair (n=109) or lateral decubitus (n=109) position,¹¹⁷ continuous MAP was measured noninvasively using finger plethysmography to calculate COx. Patients in the beach chair position had higher COx (indicating impaired CBF autoregulation) and lower regional cerebral oxygen saturation than those in the lateral decubitus positions. There was no difference between the different surgical positioning groups in a composite neurocognitive score from baseline to 7–10 days or 4–6 weeks. A relationship between impaired CBF autoregulation and adverse neurocognitive outcomes after noncardiac surgery has been demonstrated although the data are inconsistent.^113,117, 118, 119

Postoperative autoregulation monitoring

Management of blood pressure after cardiac surgery in the ICU balances the goals of minimising mediastinal blood loss while ensuring organ perfusion. Given the wide interindividual variability in CPP at the LLA, basing postoperative blood pressure targets on empiric, historical practices might lead to organ injury. In a study of 121 adult patients after cardiac surgery, hypotension defined by CBF autoregulation monitoring vs standard definitions was associated with increased plasma levels of the brain-specific injury biomarker glial fibrillary acidic protein (GFAP).¹²⁰ GFAP is an astrocyte cytoskeleton protein that has been demonstrated to be elevated in patients with TBI, stroke, and cardiac arrest.121, 122, 123, 124 A link has been found between delayed neurocognitive recovery after noncardiac surgery and elevated blood GFAP levels.¹²⁵

To more closely characterise intraoperative and postoperative ICU CBF autoregulation metrics, we performed a prospective observational study of 134 patients undergoing cardiac surgery with CPB.¹²⁶ Based on COx, impaired CBF autoregulation was more frequent in the operating room than in the ICU (40% vs 13%, P<0.001). Excursions of MAP outside the limits of CBF autoregulation were as common in the ICU as in the OR (median 16.9 mm Hg × h; inter-quartile range [IQR] 10.1–38.8 vs 16.9 mm Hg × h; IQR 5.4–35.1 mm Hg, P=0.20). In exploratory analysis, impaired CBF autoregulation in the ICU but not the OR was associated with postoperative delirium detected with 3D-CAM. These data suggest that with usual care, MAP in the ICU is often outside the limits of CBF autoregulation, which might be associated with brain ischaemic injury and postoperative delirium.

Interventional trial of personalised blood pressure management

Data from observational studies linking MAP outside the limits of CBF autoregulation and adverse patient outcomes could be biased by unmeasured covariates. To more closely evaluate this question, 461 patients undergoing cardiac surgery were randomised to have MAP targets during CPB above the LLA vs usual care (Table 3). While there was no difference between MAP groups in the primary composite outcome (stroke or cognitive decline 1 month after surgery or new MRI brain injury), the frequency of clinically detected delirium was lower in the autoregulation arm than the usual care group (8.2% vs 14.9%, P=0.035). Further, the autoregulation-targeted MAP group had higher scores on memory testing 1 month after surgery compared with the usual care group (P=0.019). A subset of patients in that study underwent comprehensive assessments for delirium prospectively for 4 days after surgery. In that analysis, the frequency of postoperative delirium was reduced from 53% in patients in the usual care group compared with 38% in the autoregulation group (P=0.04).¹²⁷ The potential benefits of prospectively maintaining CPP above CPP_OPT have also been reported in pilot studies of patients with TBI.^128,129

Table 3.

Results of a randomised clinical trial of 461 patients where mean arterial pressure targets during cardiopulmonary bypass were chosen to be above the lower limit of cerebral blood flow autoregulation (transcranial Doppler measured mean velocity index) vs usual care. The primary outcomes of the parent study and its components are listed. The number of patients denotes the number of patients (percentage) with outcomes recorded for that event. Lower than projected number of patients with postoperative brain MRI examination impacted the number of patients with composite outcome data. In a sub-study, a subset of patients (n=199) underwent comprehensive delirium assessments for the first four postoperative days including a structured examination, chart review, and adjudication of events by an expert panel. Outcomes for that sub-study along with sensitivity analysis of different definitions of delirium are listed. 95% CI= 95% confidence interval; CAM, confusion assessment method or confusion assessment method ICU; OR, odds ratio; RR, risk ratio.

	MAP target > lower limit of autoregulation	Usual care MAP target	Risk reduction
Hogue and colleagues77 Parent trial (n=461)
Primary composite outcome (%)	70/93 (75.3)	79/101 (78.2)	RR: 0.96 (95% CI, 0.82–1.12, P=0.075)
Component outcomes clinical stroke (%)	13/232 (5.6)	9/228 (3.9)	RR: 1.42 (95% CI, 0.62–3.26, P=0.539)
Delayed neurocognitive recovery at 1 month (%)	27/172 (15.7)	35/164 (21.3)	RR: 0.74 (95% CI, 0.47–1.16, P=0.233)
Brain MRI diffusion weighted ischaemic lesion (%)	48/79 (60.8)	47/85 (55.3)	RR: 1.10 (95% CI, 0.85–1.43, P=0.582)
Clinically detected delirium (%)	19/232 (8.2)	34/228 (14.9)	RR: 0.55, 95% CI, 0.32–0.93, P=0.035)
Brown and colleagues127 Focused delirium assessment sub-study (n=199)
Postoperative delirium (%)	39/103 (37.9)	48/91 (52.7)	OR: 0.55 (95% CI, 0.31–0.97, P=0.04)
Sensitivity analysis Consensus panel (%)	41/105 (39.1)	51/94 (54.5)	OR: 0.54 (95% CI, 0.31–0.95, P=0.03)
Consensus panel adjudication with medical record review (%)	45/103 (43.7)	53/91 (58.2)	OR: 0.56 (95% CI, 0.32–0.98, P=0.04)
Research assistant CAM assessment	39/103 (37.9)	53/91 (58.2)	OR: 0.44 (95% CI, 0.25–0.78, P=0.005)

Paediatric cerebral autoregulation monitoring in cardiac surgery and TBI

In neonatal and paediatric congenital heart surgery, personalisation of MAP goals is challenging with poor definitions for the LLA in various age groups. Neurological injury occurs in 40–50% of newborns who undergo cardiac surgery with CPB.^130,131 Factors associated with its most common form, white matter injury, include low blood pressure, duration of CPB, duration of deep hypothermic circulatory arrest, and hypoxaemia.132, 133, 134, 135 Moreover, goal-directed blood pressure targets in this population are largely unknown and based on the experience of perioperative care teams. Near infrared spectroscopy-based autoregulation with COx has been described in an observational pilot study of 54 neonates and children undergoing surgery with CPB. In this study, the LLA was calculated in 77% of patients and found to be 42 (7) mm Hg.¹³⁶ However, there was a large degree of variability in this population with LLA ranging from 20 to 55 mm Hg, suggesting the need for personalised autoregulation monitoring.

Autoregulation monitoring has played a large role in paediatric TBI, which is a leading cause of death in children.¹³⁷ In children with TBI several groups have found associations between impaired cerebral autoregulation and poor 6-month outcomes, low Glasgow Coma Scores, and death.138, 139, 140, 141 Through development of computational and real-time methods for continuous assessment of autoregulation in head injury, Czosnyka and colleagues⁶³ described a method that allows for targeted management of CPP in TBI. Consequently, autoregulation measured by the PRx is a promising tool in the treatment of infant and paediatric TBI and has been suggested as a monitor to allow for early intervention. With such a tool, the time with severely impaired autoregulation (PRx >0.2) can be reduced with the goal to achieve better neurological outcomes.^142,143

Conclusions

Cerebral blood flow autoregulation has matured from the conceptual model introduced by Lassen⁴ in 1959 to computational methods measuring autoregulation indices. Animal studies have been instrumental in validation of the autoregulation measurement techniques and principles. A limitation of current technology is the response time to render a blood pressure target based on autoregulation monitoring. Prior studies suggest that vascular changes lead to a signal in the autoregulation response when slow wave changes last 30–60 s.⁵⁸ As a result, time-domain correlation methods average data over 300 s updated with 10-s overlapping windows of time. Graphical user interfaces have been designed to provide the clinical care team real-time autoregulation metrics for MAP targets to stay within the LLA and ULA (Fig. 2).¹⁴⁴

Fig 2.

Autoregulation monitoring during infant congenital heart surgery with cardiopulmonary bypass using a derived virtual patient monitor. Digital signals from patient monitor data were captured and analysed to yield an autoregulation metric that can be displayed on a variety of devices including tablet or smart phone. Top panel: the top graph displays trend of mean arterial blood pressure (mm Hg) over hours. Fluctuations in mean arterial blood pressure (ABPm) are expected during surgery because of factors including surgical stimulation, arterial line flushes, and vasoactive medications. The limits of autoregulation are calculated and updated in the green shaded region (lower limit of autoregulation of 46 mm Hg and upper limit of autoregulation of 90 mm Hg). The right bar denotes the current patient blood pressure (mean arterial blood pressure of 46 mm Hg) in relationship to the limits of autoregulation (shaded in green). The histogram inset to the left represents the duration of the recording at each blood pressure. Bottom panels: the lower autoregulation bar graphs are updated in real time representing cerebral oxygenation index (COx) for each 5 mm Hg bin of blood pressure for channel 1 (Ch 1)—right NIRS and channel 2 (Ch 2)—left NIRS. Channel 1 is utilised to determine limits of autoregulation and display in the top graph. The threshold of COx >0.3 is selected for limits of autoregulation. A clear indication of MAP associated with intact cerebral blood flow autoregulation is demonstrated by values of COx ≤0.3.

Based on clinical experience with real-time autoregulation monitors in the cardiac operating room, it can take 10–15 min to delineate useful measures of autoregulation indices. Novel methods involving co-trending have been proposed to examine the behaviour of MAP and near infrared spectroscopy cerebral saturation signals in 50-s windows (vs 300 s) to make a more rapid assessment of cerebral autoregulation.¹⁴⁵ The trending behaviour of the two signals can be designated as intact, impaired, or indeterminate. Other proposed methods of reducing the amount of time to render autoregulation indices involve inducing oscillations in the MAP signal by varying positive end-expiratory pressure (PEEP) at fixed intervals. This method of PEEP modulation is able to induce a low-frequency wave; with this method, there is a reduction in noise, increase in precision, and increase in accuracy for determination of the LLA.¹⁴⁶ PEEP oscillations have been applied to the clinical setting for adult TBI patients and similarly demonstrated promise as a feasible method for improved determination of the PRx with reduced variability.¹⁴⁷

As we continue to improve techniques for determining autoregulation indices, blood pressure might be a modifiable factor to mitigate adverse outcomes associated with hypoperfusion and hyperperfusion. Once feasibility, safety, and efficacy are established in a bedside autoregulation monitor, application of targeted and personalised blood pressure monitoring to optimise cerebral blood flow will become practical.

Authors’ contributions

Conception and intellectual content of this review: ELV, KMB, CWH

Drafting of the manuscript, and approval of the final version of the manuscript: all authors

Acknowledgements

The authors are grateful for the collaboration with Marek Czosnyka and Peter Smielewski (University of Cambridge, Cambridge, UK) and with Kathy Kibler (Texas Heart Institute, Houston, TX, USA). Their collaborations have been fundamental for the research that made this review possible.

Declarations of interest

CHB has a data share and research contract with Medtronic, with all funding through Johns Hopkins University (Baltimore, MD, USA). CWH serves on a data and monitoring committee for Merck (Kenilworth, NJ, USA), and serves as an advisor, provides lectures, and receives research funding from Medtronic, Inc. (Boulder, CO, USA) and Edwards Lifesciences (Irvine, CA, USA). The other authors declare that they have no conflicts of interest.

Funding

The research that forms the basis for this review was supported in part by a grant from the Additional Venture Single Ventricle Research Fund (www.additionalventures.org, to KMB); the National Institutes of Health (Bethesda, MD, USA) (R01HL092259 to CWH, K76AG057020 to CHB, RF1 AG072387 to CHB).

Handling Editor: Hugh C Hemmings Jr