Ischemic stroke is among the most frequent neurological events in MCS [22,23]. Mechanisms include thromboembolism originating from the device or cannula, embolism related to underlying cardiac disease/left ventricle stasis, and systemic pro-thrombotic states [11]. In addition, reduced pulsatility in CF devices compromises CVAR, limiting adaptive cerebrum responses to blood pressure (BP) fluctuations [24]. These effects particularly impair perfusion in key regions, which reduces collateral flow during acute ischemic ABI and amplifies the severity and extent of the infarct [25].

HIBI may occur before, during, or after initiation of MCS, especially in patients supported with VA ECMO. Prolonged low-flow states during cardiac arrest, even with cardiopulmonary resuscitation, often provide the initial insult, while hypotension, gas exchange instability, or abrupt carbon dioxide (CO₂) shifts during cannulation or implantation can exacerbate cerebral vulnerability [26]. Even after hemodynamic stabilization, recurrent hypotension, absence of pulsatility, systemic desaturation, or differential oxygenation can propagate diffuse cortical and subcortical ABI [27]. Acting together, differential hypoxia, impaired microvascular perfusion, and pressure-determined cerebral perfusion in CF states diminish the compensatory capacity of the brain.

ICH represent a major cause of morbidity and mortality in MCS [11,28], most commonly manifesting as intraparenchymal or subarachnoid hemorrhage, or subdural hematoma. Systemic anticoagulation, acquired von Willebrand syndrome, and platelet dysfunction contribute to impaired hemostasis, while ischemic lesions may undergo reperfusion-associated bleeding or hemorrhagic transformation [23]. Impaired CVAR and microvascular fragility facilitate direct transmission of hemodynamic stress into cerebral vessels, further predisposing patients on MCS to hemorrhagic events [11].

Beyond overt episodes of stroke, many patients develop subclinical ABI. Advanced neuroimaging techniques frequently reveal silent infarcts and cerebral microbleeds [29,30], likely reflecting chronic microcirculatory dysfunction driven by endothelial activation, impaired nitric oxide (NO) signaling, inflammation, and prothrombotic changes [31]. These changes may contribute to cognitive decline, mood disorders, and reduced functional recovery [32], highlighting the importance of surveillance even in the absence of major clinical events.

A common pathophysiological theme emerges across these ABI phenotypes: cerebral microcirculation failure and impaired CVAR to buffer the brain against hemodynamic, metabolic, and gas exchange perturbations inherent to MCS. Understanding these shared mechanisms is essential to link device physiology with neurological injury risk, and to identify monitoring and intervention targets.

Microcirculation designates the vascular network responsible for oxygen and nutrient exchange. The capillary network represents the largest surface area of the vasculature, and it is critically responsible for delivering oxygen to tissues and removing CO₂ [33,34]. Arteriolar smooth muscle constriction or dilation adjusts downstream capillary hydrostatic pressure, thereby regulating blood distribution and ensuring adequate exchange (Figure 1A) [35,36]. Overall, a functional microcapillary network is essential to maintain tissue oxygenation and metabolite exchange, and even subtle drops in functional capillary density or flow can impair organ function [37].

Microcirculatory dysfunction is increasingly recognized as a central contributor to organ failure in patients using MCS. While systemic hemodynamic targets such as BP or cardiac output (CO) may normalize, the microcirculation frequently remains impaired [33,38,39]. This uncoupling of microcirculatory flow from microcirculatory targets, known as “loss of hemodynamic coherence,” results in persistent tissue hypoxia, elevation of lactate levels, and organ dysfunction [33]. Direct evidence of cerebral microvasculature injury during MCS is provided by experimental and clinical data. In a randomized study on an ovine VA ECMO model for differential hypoxemia, low ECMO flow resulted in a 58% reduction in brain tissue oxygen tension (PbtO₂) with accompanying histopathological evidence of neuronal shrinkage, vascular congestion, and perivascular edema, while high flow conditions preserved cerebral oxygenation. Complementing these findings, neuropathological examination of patients supported with CF LVADs demonstrated cerebral microvasculature injury in over 90% of cases, including cerebral microbleeds in approximately 80%, the majority of which were clinically silent, consistent with chronic microcirculatory and endothelial dysfunction [40,41]. Together, these findings suggest a mechanistic examination of how non-pulsatile flow, endothelial injury, and venous congestion disrupt microvascular regulation during MCS (Table 1).

Under normal physiological conditions, arterioles possess smooth vascular muscle that intrinsically responds to pressure and flow changes. For instance, an increase in transmural pressure activates mechanosensitive channels leading to vessel constriction, which stabilizes capillary flow despite BP fluctuations [35,42]. Endothelial function also plays a central role: endothelial cells sense pulsatile blood flow and respond by releasing vasodilators such as NO, prostacyclin, and endothelium-derived hyperpolarizing factor [43,44]. NO is continuously generated in response to laminar pulsatile flow, and is a key regulator of basal microvascular tone. In continuous flow states such as CF LVAD, reduced pulsatility blunts this signaling, decreasing NO production and contributing to lower functional capillary density [45].

Continuous flow states also injure the endothelial glycocalyx layer, impairing both vasoreactivity and barrier function, leading to tissue edema [38,46]. Glycocalyx shedding during cardiopulmonary bypass (CPB) correlates with impaired capillary perfusion and higher postoperative complication rates [46], while ECMO and CF LVAD support are associated with biochemical markers of endothelial activation, including elevated angiopoietin-2 and activation of coagulation pathways [47-51].

Consistent with these mechanisms, sublingual microscopy in patients using CF LVAD (n=42) shows a reduction of 44% in functional capillary density compared to healthy individuals (P<0.001), and of 23% compared to patients with chronic heart failure on standard pharmacotherapy (P=0.042) [52]. Similarly, studies of CPB have shown better microcirculation preservation with pulsatile versus continuous flow [53-56]. These findings demonstrate that perfusion pressure restoration alone cannot compensate for loss of pulsatility, as endothelial injury drives persistent microvascular dysfunction, impaired capillary recruitment, and local hypoxia despite the normalization of mean arterial pressure or the cardiac index [57-60].

Importantly, capillaries lack smooth muscle or innervation, and therefore perfusion is passively governed by the pressure gradient from upstream arterioles to downstream venules, which is known as microcirculatory driving pressure. Since most of this pressure is dissipated across arterioles, the capillary bed is a lowpressure compartment: capillary perfusion pressure is more influenced by venous instead of arterial pressure, especially when central venous pressure (CVP) is elevated [36].

In patients with cardiogenic shock using ECMO, venous congestion is common due to right ventricular dysfunction and increased preload, resulting in higher CVP [38]. Evidence demonstrates that this elevation is physiologically meaningful: in patients with sepsis, CVP >12 mm Hg reduces microvascular driving pressure and is associated with impaired sublingual microvascular flow despite normal mean arterial pressure (MAP) and CO [61]. In the brain, venous outflow pressure is also a determinant of cerebral perfusion pressure (CPP, which is calculated by subtracting intracranial pressure [ICP] to MAP), as elevated venous pressure may increase ICP or impair effective cerebral venous drainage. This provides a mechanism by which venous congestion could exacerbate cerebral hypoperfusion during MCS. However, direct evidence linking elevated CVP or venous congestion to cerebral ischemia or ABI in adults using MCS remains limited. Accordingly, this should be considered as a hypothesis, and highlights an important area for future investigation involving cerebral venous pressure, ICP, and CPP assessment in MCS cohorts.

Several techniques now permit direct or indirect microcirculation assessment at the bedside (Table 2). Incident dark field and side stream dark field video microscopy provides direct, realtime visualization of capillary density, flow velocity, and heterogeneity [62-65], and has been applied in cohorts experiencing shock or using ECMO or CF LVAD [16,18,66-68]. Biochemical and clinical markers such as lactate clearance, capillary refill time, and mottling score serve as simple indicators of impaired perfusion in MCS and VA ECMO, but are non-specific and often cannot reflect dynamic changes [69-75]. An elevated central venous-to-arterial PaCO₂ gap (>6 mm Hg) provides another useful clue, as it implies reduced washout of CO₂ from tissues, and has been linked to microcirculatory hypoperfusion in sepsis and to early mortality risk during VA ECMO [72,76].

Near-infrared spectroscopy (NIRS) offers an indirect measure of microcirculatory perfusion by reflecting tissue oxygen saturation (StO₂) in the microvasculature. Cerebral NIRS involving forehead sensors is widely used in cardiac surgery and intensive care unit settings because it provides a continuous, non-invasive indicator of regional oxyhemoglobin levels, has been validated for the tracking of cerebral perfusion changes, and, when correlated with variations, can assess autoregulation during CPB [77-79]. While subject to artifacts and limited spatial coverage, NIRS is the most practical bedside tool for MCS, and is recommended in patients using ECMO to detect cerebral hypoxemia and guide perfusion targets [80-84].

CF MCS devices, including CF LVAD and VA ECMO, markedly reduce arterial pulsatility, creating a cerebrovascular environment that differs fundamentally from normal physiology [31,90,96,97]. Pul-satility supports autoregulation through cyclic stretch and shear stress, which trigger NO release and allow vascular smooth muscle to reset tone with each cardiac cycle [31,96]. Prior studies in animals and humans have raised concerns that constant perfusion pressure may narrow the autoregulatory curve or blunt CO₂ reactivity [24,98,99]. Regarding VA ECMO, “pressure-passive” cerebral circulation has been reported in case series, particularly after cardiac arrest or severe shock, when cerebral perfusion becomes directly dependent on pump flow and MAP [100]. In a swine model for ECPR, loss of pulsatility and reduced pulse pressure were strongly associated with impaired autoregulation, reflected by elevated pressure reactivity index (PRx) values despite restoration of circulation. After return of spontaneous circulation (ROSC), the return of pulsatility improved autoregulatory function, although variability in PRx across subjects highlighted individual differences in recovery [21]. Likewise, low pulse pressure (<20 mm Hg) on VA ECMO is associated with ABI, with each decrease of 10 mm Hg in pulse pressure resulting in an increase of 30% in neu-rological risk [101-103]. Increasing pulsatility with an intra-aortic balloon pump (IABP) has been shown to improve cerebral oxygenation surrogates and enhance autoregulatory responsiveness [102,103]. Adjusting pump speed, using adjunct devices such as a microaxial flow pump (mAFP), or titrating inotropes can also restore pulsatility in VA ECMO. In summary, while universal pulse-pressure targets remain undefined and are an active area of investigation, severe reductions in pulsatility (particularly below 20 mm Hg) appear to determine a high-risk physiological state, and modulation of pulse pressure may represent a practical strategy for enhancing cerebral protection during VA ECMO.

Patients on VA ECMO (Figure 3) often have severe brain insults such as global brain injury, which can disrupt CVAR [100,104,105]. Observational studies using NIRS-based indices report periods of impaired CVAR during the first 24-48 hours of ECMO [97,100]. This pattern is consistent with post-cardiac arrest syndrome, which involves endothelial dysfunction, inflammation, and loss of vascular reactivity [106]. In such cases, clinicians sometimes pursue higher MAP targets to drive adequate cerebral perfusion, as it is crucial to avoid hypotension during pressure-dependent periods [107,108]. In a porcine model for ECPR, higher MAP targets transiently preserved CVAR before defibrillation and ROSC, but this benefit was no longer evident at later stages, with worsening hemodynamics and impaired CVAR [109]. Observational data similarly suggest that MAP targets during ECPR should be guided by the autoregulatory status, as both hypotension and excessive hypertension may exacerbate cerebral injury through pressuredependent perfusion and hyperperfusion [97].

PaCO₂ is a powerful modulator of cerebral perfusion, although the findings in the literature regarding its precise effects on CVAR are contradictory. Despite this heterogeneity, emerging evidence suggests that both the rate and amplitude of PaCO₂ correction contribute to CVAR impairment and ABI. The magnitude of these changes is typically highest during ECMO initiation and coincides with the first 24 hours—the period of highest CVAR vulnerability [97]. In a porcine model for ECPR, rapid PaCO₂ correction (from approximately 60 to approximately 30 mm Hg within 5 min) significantly impaired CVAR (PRx ≥0.5) and increased neuronal injury in vulnerable regions that included the putamen and hippocampus, while gradual correction better preserved CVAR and reduced ischemic damage [93]. Additionally, hypercarbia itself has been shown to elevate the lower limit of autoregulation (LLA), suggesting that CO₂ levels directly influence the CPP threshold needed to maintain intact autoregulation [14,99]. While the exact mechanisms remain to be fully elucidated, there is sufficient evidence that PaCO₂ alterations during ECMO transition may predispose patients to cerebral injury through pressure-passive CBF. This underscores the importance of careful sweep gas titration to avoid both rapid corrections and extreme CO₂ values during ECMO management.

Patients on CF LVAD present a unique physiological paradox: despite long-term exposure to nearly nonpulsatile circulation, many demonstrate apparently preserved CVAR and CO₂ reactivity (Figure 3) [101,110,111]. However, these observations likely reflect chronic vascular adaptation rather than true preservation of normal autoregulatory mechanisms. During acute transition to continuous flow, loss of pulsatile shear stress blunts NO-mediated vasomotion and narrows the autoregulatory range, render-ing CBF more pressure-passive. After months of CF LVAD support, compensatory remodeling occurs: vascular smooth muscle tone is down-regulated, and endothelial signaling partially returns to an steady-flow environment, resulting in a new “pseudo-steady” autoregulatory state [96]. This adaptation allows patients to maintain near-normal CO₂ reactivity and tolerate modest BP changes, even though their dynamic response capacity is reduced. In this remodeled state, the lower limit of CVAR may shift upward because baseline arteriolar tone remains dilated, leaving less reserve for further vasodilation during hypotension. Structural remodeling, including adventitial fibrosis and loss of elastic lamina integrity, further limits vascular compliance and may predispose to both ischemic and hemorrhagic complications [112]. These findings suggest that CVAR under chronic CF support is not physiologically “intact” but rather represents a redefined equilibrium with narrower safety margins and reduced flexibility to acute perturbations in pressure or gas tension (Table 3).

The combination of these functional and structural alterations may partly explain why patients on CF LVAD show a high incidence of ABI, with ischemic and hemorrhagic strokes accounting for nearly one-fifth of deaths in this population [11]. While many strokes are attributed to pump-related emboli or anticoagulation, there is speculation that altered autoregulation and cerebral perfusion may also contribute [22]. In one report, approximately one-third of stable outpatients on CF LVAD showed periods of pressure-dependent CBF on transcranial Doppler (TCD) monitoring, while others maintained good autoregulatory indices [111]. Interestingly, some patients with very low pulse pressure (<10 mm Hg) still had intact static autoregulation, suggesting that pulsatile flow is not absolutely required for CVAR [113]. However, this study evaluated only short-term experimental reductions in pulse pressure and therefore does not address the long-term cerebrovascular consequences of chronic CF, which remain incompletely understood. The introduction of partial pulsatility has been associated with improved middle cerebral artery flow patterns and could theoretically reduce the workload on autoregulatory mechanisms [102]. Ongoing research is evaluating whether actively introducing pulsatility into CF devices can improve cerebrovascular outcomes [114].

Several techniques are used for bedside monitoring of CVAR in the critical care and MCS setting. Each relies on detecting the relationship between CBF (or a proxy) and BP over time:

(1) PRx: PRx is a widely used index in neurocritical care that correlates slow waves of ICP with MAP. It is calculated as the moving Pearson correlation coefficient between ICP and MAP oscillations, typically over 30-second to 5-minute windows [115]. When autoregulation is intact, changes in MAP invoke compensatory changes in cerebral vessel di-ameter that buffer ICP fluctuations, yielding a negative or near-zero correlation. If autoregulation is impaired, ICP tends to vary directly with MAP, resulting in a positive value. PRx is considered the gold-standard measure for cerebrovascular reactivity, and has been applied in research on patients with ECMO and cardiac arrest by using an external ventricular drain or parenchymal ICP monitor to derive the index. Its strength lies in its clear physiological interpretation and its ability to provide an “optimal MAP” (the pressure at which PRx is minimized), indicating the best autoregulation for individual patients [116,117]. The main limitation is its invasiveness and the need for high-fidelity ICP data, which is not routinely available for most patients using MCS.

(2) Cerebral oximetry index (COx): COx is an analogous index that replaces ICP with cerebral oxygenation signals using NIRS [78]. Here, fluctuations in StO₂ are correlated with MAP changes. If CVAR is intact, StO₂ remains steady despite MAP variability, yielding a low correlation. Conversely, impaired autoregulation leads to StO₂ tracking MAP. COx can be monitored noninvasively via NIRS pads on the forehead, which offers a practical way for bedside monitoring of dynamic autoregulation and even for MAP titration guidance in real time [80].

(3) TCD monitoring: TCD ultrasound can measure blood flow velocity in basal cerebral arteries. Dynamic autoregulation can be assessed by assessing changes in flow velocity in response to spontaneous or induced BP fluctuations. One approach is to calculate the mean flow index (Mx), which correlates MAP and middle cerebral artery flow velocity. Mx is analogous to PRx but based on flow velocity instead of ICP [118,119]. A positive value suggests pressure-dependent flow. Another approach is to perform autoregulation tests with TCD, such as the cuff test [120]. TCD provides excellent temporal resolution and direct information on flow dynamics. It has been used to determine intact and impaired autoregulatory responses during CPB and in patients using CF LVAD. However, continuous TCD monitoring is technically challenging, requiring skilled operators, stable probe fixation, and sufficient acoustic bone windows. It also provides readings on the flow in large vessels and not in the microcirculation. Consequently, TCD is often used for discrete assessments rather than continuous monitoring [121].

(4) PbtO₂: in neurocritical care, a thin probe can be inserted into brain tissue (often frontal white matter) to directly measure PbtO₂, which reflects the balance between local oxygen supply and demand. From this measurement, an oxygen reactivity index can be derived by correlating PbtO₂ with MAP, as in the case of COx [122,123]. In the context of MCS, PbtO₂ monitoring is not typically used solely to assess autoregulation because of its invasiveness, but it may be employed as an adjunct in patients already undergoing ICP monitoring, offering complementary information about local oxygenation and potential ischemia. While PbtO₂ provides direct data on local tissue oxygenation, this may not capture global perfusion [122].

Emerging evidence supports an individualized approach to MAP targets in patients using MCS [124,125]. If autoregulation is intact, MAP should be kept within the autoregulatory range of the patient— the “optimal MAP”—identified as the pressure with indices such PRx or COx having the lowest value [78,126,127]. Deviation from this range is linked to higher incidence of acute neurological events in patients on ECMO [97,128]. Both hypo- and hypertension can be detrimental: in ECPR cohorts, MAP <70 mm Hg has been associated with worse neurological outcomes [129,130], while pediatric studies on ECMO show that sustained hypertension (above the autoregulatory upper limit) predisposes to ICH [131]. Thus, an intermediate range (65-75 mm Hg) is generally recommended, with real-time autoregulation monitoring offering patientspecific fine control.

Achieving optimal cerebral perfusion in continuous flow MCS often requires proactive adjustments of hemodynamic and respiratory parameters. Vasopressors are frequently used to maintain MAP above the LLA, thereby preventing pressure-dependent cerebral hypoperfusion when autoregulation is blunted [132]. However, care must be taken to avoid overshooting with vasopressors, since severe hypertension in the setting of impaired CVAR can precipitate breakthrough hyperemia or ICH [131].

Device-specific physiology further shapes perfusion strategies. In patients using CF LVAD, chronic hypertension raises the risk of ICH. Maintaining MAP in a moderate range not only mitigates this risk but also aligns with autoregulatory preservation, as sustained high pressures can shift autoregulatory curves rightward [133,134]. Pump speed adjustments add nuance: higher speeds increase MAP but decrease pulsatility, potentially blunting cerebral flow responsiveness [31,113], whereas running devices at slightly lower speeds to permit residual pulsatility may support a more physiological BP regulation [31]. In VA ECMO, adjunctive devices such as a IABP can restore diastolic augmentation and introduce a pulse pressure, potentially re-engaging the myogenic response [102,135]. Importantly, IABP effects depend on residual endogenous CO—increasing mean CBF when intrinsic pulsatility exists but potentially reducing it when this output is minimal [102]. Similarly, the more recent mAFPs (e.g., Impella) or tailored pump speed adjustments offer partial pulsatility that may better support cerebrovascular dynamics. While direct evidence remains limited, expert consensus suggests that maintaining some degree of arterial pulsatility and physiological BP levels is beneficial for cerebrovascular integrity during MCS [134,135].

Respiratory management is another key component in the optimization of cerebral perfusion in MCS. CO₂ is a powerful modulator of cerebrovascular tone, and maintaining normocapnia is particularly important when autoregulation is impaired. Excessive sweep gas flow on ECMO can decrease PaCO₂ excessively, provoking cerebral vasoconstriction and making CBF pressure-passive [97]. Rapid shifts in PaCO₂ at the time of ECMO initiation can also be detrimental, as abrupt correction of hypercapnia or hypocapnia can destabilize CVAR and predispose to cerebral injury [136,137]. Thus, clinicians titrate the sweep gas flow to maintain PaCO₂ within normal to slightly elevated values that support stable cerebral perfusion [97]. Oxygenation is managed with similar caution: adequate arterial oxygen content is ensured, but extremes of hyperoxia are avoided to limit oxidative stress [95]. Considered together, the optimization of blood gas targets to favor autoregulation complements support with vasopressor medication and pulsatility enhancement, resulting in a multifaceted strategy to preserve cerebral perfusion and mitigate neurological injury during ECMO.

Bedside assessments with frequent checks of pupil reactivity and standardized neurological examinations remain the foundation of ABI early detection [5]. In comatose or sedated patients, continuous electroencephalography is often used to detect seizures or ischemic activity [80]. Some centers also collect routine head computed tomography images within the first 24-48 hours of VA ECMO, recognizing the high incidence of clinically silent ischemic and hemorrhagic events in this setting. When imaging abnormalities are suspected, early neuroimaging should be pursued. Portable magnetic resonance imaging may facilitate timely detection of subclinical ischemic or hemorrhagic injury [15,100].

Physiological monitoring provides additional warning signs of evolving ABI. A sustained correlation between MAP variability and NIRS-derived cerebral oxygenation indicates loss of CVAR, prompting implementation of the autoregulation-guided MAP titration strategies described above. Similarly, detection of cerebral microembolic signals on TCD should prompt intensification of anticoagulation therapy and evaluation of the ECMO circuit for thrombus formation or flow disturbances [138,139].

Standardized responses are also important once complications are identified. Acute ischemic stroke or ICH is managed with urgent neuroimaging, early neurology or neurosurgery consultation, and careful titration of anticoagulation medication to balance neurological stability and device-related thrombotic risk. Broader neurocritical care measures, including seizure control, fever management, optimization of O₂/CO₂, and avoidance of large BP fluctuations, complement these targeted interventions. The first days under MCS represent a period of neurological vulnerability when interdisciplinary coordination among critical care, neurology, and surgical teams is essential to minimize secondary brain injury and optimize recovery.