Perioperative stroke: A perspective on challenges and opportunities for experimental treatment and diagnostic strategies

Abstract Perioperative stroke is an ischemic or hemorrhagic cerebral event during or up to 30 days after surgery. It is a feared condition due to a relatively high incidence, difficulties in timely detection, and unfavorable outcome compared to spontaneously occurring stroke. Recent preclinical data suggest that specific pathophysiological mechanisms such as aggravated neuroinflammation contribute to the detrimental impact of perioperative stroke. Conventional treatment options are limited in the perioperative setting due to difficult diagnosis and medications affecting coagulation in may cases. On the contrary, the chance to anticipate cerebrovascular events at the time of surgery may pave the way for prevention strategies. This review provides an overview on perioperative stroke incidence, related problems, and underlying pathophysiological mechanisms. Based on this analysis, we assess experimental stroke treatments including neuroprotective approaches, cell therapies, and conditioning medicine strategies regarding their potential use in perioperative stroke. Interestingly, the specific aspects of perioperative stroke might enable a more effective application of experimental treatment strategies such as classical neuroprotection whereas others including cell therapies may be of limited use. We also discuss experimental diagnostic options for perioperative stroke augmenting classical clinical and imaging stroke diagnosis. While some experimental stroke treatments may have specific advantages in perioperative stroke, the paucity of established guidelines or multicenter clinical research initiatives currently limits their thorough investigation.

mechanisms such as aggravated neuroinflammation contribute to the detrimental impact of perioperative stroke. Conventional treatment options are limited in the perioperative setting due to difficult diagnosis and medications affecting coagulation in may cases. On the contrary, the chance to anticipate cerebrovascular events at the time of surgery may pave the way for prevention strategies. This review provides an overview on perioperative stroke incidence, related problems, and underlying pathophysiological mechanisms. Based on this analysis, we assess experimental stroke treatments including neuroprotective approaches, cell therapies, and conditioning medicine strategies regarding their potential use in perioperative stroke. Interestingly, the specific aspects of perioperative stroke might enable a more effective application of experimental treatment strategies such as classical neuroprotection whereas others including cell therapies may be of limited use. We also discuss experimental diagnostic options for perioperative stroke augmenting classical clinical and imaging stroke diagnosis. While some experimental stroke treatments may have specific advantages in perioperative stroke, the paucity of established guidelines or multicenter clinical research initiatives currently limits their thorough investigation.

K E Y W O R D S
experimental treatment, neuroprotection, perioperative stroke, stroke, translational research

| INTRODUC TI ON
According to the Society for Neuroscience in Anesthesiology and Critical Care (SNACC) definitions, perioperative stroke (PS) describes an ischemic or hemorrhagic cerebral event during or up to 30 days after surgery. 1 Non-modifiable risk factors for PS are higher age and female sex. 2 The reasons for the higher PS risk in female patients are not well understood, but it is supposed that it may be related to faster progression of atherosclerosis after menopause. 2 PS is most frequent after cardiac and neurosurgery, as well as after vascular thoracic and transplantation surgeries. 3 The main etiology of ischemic PS seems to be embolic, but perfusion-related events are also of relevance. For instance, severe anesthesia-induced hypotension is associated with PS or can contribute to cerebral mal-perfusion in intra-interventional PS. 4 Previously undiagnosed stenosis of a large brain-supplying artery may also play a role. The vast majority of PS, about 95%, is of ischemic nature. 5,6 Consequently, this review will focus on ischemic events.
Perioperative stroke is a feared condition for numerous reasons.
First, it is relatively frequent. In-depth analysis of more than 520,000 cases reported an overall incidence of 0.1% in patients undergoing non-cardiac surgery, which increased to 1.9% in high-risk populations. 7 A large-scale retrospective investigation based on the US National Inpatient Sample involving over 10.5 million cases revealed a PS rate of 0.52% in patients undergoing non-cardiac surgery in 2004. Interestingly, the rate increased to 0.77% in 2013. 8 Although these relative numbers may seem moderate, the large amounts of surgeries performed each year result in a considerable number of PS cases. Moreover, PS can be much more frequent in some populations of surgical patients, especially those who undergo cardiac surgery. In detail, PS rates of 8.8% were reported for mitral valve surgery, and up to 9.7% for double or triple valve surgery, respectively. 9 Second, PS can appear as a clinically silent event, sometimes referred to as covert PS, which is unrecognized on onset but detected later, for instance by means of brain imaging. Reasons for covert PS can be minor, subtle, or not properly classified symptoms. Moreover, the patient may be anesthetized or sedated during PS onset thus does not present obvious clinical symptoms. A recent multicenter clinical study revealed that covert PS appeared in 1 out of 14 (7.1%) non-cardiac surgical patients aged 65 or older. PS was also associated with a higher risk for cognitive decline 1 year after surgery. 10 Third, outcome after PS tends to be worse than in non-surgical (NS) stroke patients. In detail, mortality rates in PS patients can be up to eightfold higher than in comparable NS populations, with mortality rates of up to 26%. 11,12 This illustrates the additional burden that PS inflicts on patients, their relatives, caretakers, and healthcare systems.
Common clinical reasons for the unfavorable outcome after PS are delayed diagnosis and intervention in case of covert PS, and potential difficulties in applying recanalization treatments (thrombolysis and thrombectomy) due to the recent surgery, medications that affect the coagulation system or the general condition of the patient. 13 Recent SNACC guidelines recommend that early endovascular thrombectomy should be considered in PS patients to restore the cerebral blood flow. 14 However, this therapeutic strategy requires clinical detection of the ischemic event and cerebral imaging including visualization of cerebral arteries in a narrow time window. PS patients also have an increased risk of hemorrhage independent of the cerebral event, which limits the use of intravenous thrombolysis markedly. 15,16 The patients' general condition may further complicate early initiations of rehabilitative actions.
This complex clinical situation warrants assessment of experimental treatment options that are currently under development to augment established therapeutic strategies for stroke, regarding their potential applicability, limitations, and benefits in PS. This includes consideration of potential PS-specific pathophysiological mechanisms, related to the surgery or the underlying condition requiring it, which may contribute to the unfavorable PS outcome.

| DIFFEREN CE S IN PATHOPHYS IOLOG IC AL MECHANIS MS IN PERIOPER ATIVE VER SUS S P ONTANEOUS LY O CCURRING S TROK E
Clinical and logistical necessities affecting PS management and contributing to the unfavorable PS outcome are well known. However, basic pathophysiological reasons for the unfavorable outcome are incompletely understood and subject to current research. A general limitation of these studies is that a specific animal model for PS does not exist. Instead, models of surgical interventions are combined with commonly used approaches to induce focal cerebral ischemia in an experimental subject. Nevertheless, preclinical studies using these PS models provide preliminary evidence for important pathophysiological differences between PS and NS stroke that may substantially contribute to inferior PS outcome. 7,17 For instance, preclinical data suggest that the relatively high PS incidence in elderly and comorbid patients could be related to atherosclerotic plaque instability caused by perioperative stress. 18,19 Plaque volume, stability, and signs for plaque rupture were investigated in apolipoprotein-Edeficient mice. Mice were fed a high cholesterol diet, inducing a wellaccepted model to investigate plaque vulnerability. 20 Some animals underwent laparotomy plus a major (about 20%) blood withdrawal, determined as the double hit paradigm. 21 Plaque volumes were significantly increased in the double hit group, and more plaques were classified as bearing signs of rupture 3 days after inducing perioperative stress. Surgery or blood loss alone also increased plaque volumes, but inter-group differences were not statistically significant.
Treatment with 80 mg/kg atorvastatin starting 3 days prior to surgery reduced cholesterol as well as plaque volume and instability. 21 Bone fracture models simulate a condition requiring surgical intervention and the surgical procedure itself. A surgically induced tibia fracture 6 or 24 h prior to experimental stroke by permanent middle cerebral artery occlusion (pMCAO) in mice increased lesion volumes and impaired functional outcome. Injury was most severe when pMCAO was induced 6 h after bone fracture. 22 There were increased numbers of microglia and bone marrow-derived macrophages around the ischemic lesion in mice with tibia fracture, indicating increased inflammation. 23,24 Moreover, there was a pronounced blood-brain barrier (BBB) breakdown. 25 Detrimental effects were also seen when bone fracture was induced after pMCAO. 26 Poststroke bone fractures are further associated with larger brain edema volumes in subacute stroke stages, 27 potentially due to mentioned BBB breakdown. The detrimental effects of bone fracture on stroke outcome are partially reversed by α-7 nicotinic acetylcholine receptor agonists diminishing neuroinflammation and promoting proregenerative mechanisms. 28,29 These preclinical results strongly indicate that aggravated (neuro-)inflammatory reactions, promoted by pre-or post-stroke bone fracture and surgery, strongly contribute to unfavorable PS outcome. However, the exact mechanisms have not been fully clarified.
Importantly, comorbidities frequently seen in elderly or surgically stressed patients such as hypertension can further aggravate innate immune and inflammatory reactions, 30 suggesting a potential relevance in PS.

| CHALLENG E S AND OPP ORTUNITIE S FOR E XPERIMENTAL TRE ATMENT S TR ATEG IE S IN PS
Once proven safe and effective, experimental treatment strategies for stroke may help to counter PS. The idea is supported by the fact that many experimental stroke treatment strategies target neuroinflammation which seems to play a prominent pathophysiological role in PS. 31 In general, experimental treatments should be compatible to recanalization procedures, but may also be applied in PS patients being ineligible for recanalization. 15,16 However, feasibility, safety, and potential efficacy of experimental treatments must be reviewed considering the particular clinical circumstances in which they would be applied for PS. In this context, it is important to consider that about half of all PS occur within 24 h and 93% within 72 h after surgery. 11 Table 1 provides an overview on potential experimental treatment that may be feasible for PS.

| Pharmacological neuroprotection
Recent advances in recanalization therapies, especially by mechanical thrombectomy, provide novel opportunities for the restoration of the cerebral blood supply and thus for additional pharmacological neuroprotection. 32,33 Combining pharmacological neuroprotection with recanalization enables delivery of neuroprotectants exactly where and when needed. Next-generation neuroprotectants such as nerinetide, a postsynaptic density protein 95 inhibitor counteracting oxidative stress, are clinically tested in combination with recanalization therapies. 34  strategies such as mechanical thrombectomy are still available to the minority of patients with spontaneously occurring stroke, primarily due to a lack of salvageable brain tissue, contraindications, or lack of access to centers performing these techniques. 35 PS patients may experience additional therapeutic restrictions due to late diagnosis (covert PS), their general condition, or specific medications impeding the use of intravenous thrombolysis and partly mechanical thrombectomy. However, these patients would still be eligible for pharmacological neuroprotection as a stand-alone treatment.
For instance, targeting neuroinflammation pathways is a potential neuroprotective strategy for treating ischemic stroke.
Salmeron et al. unexpectedly found that the pro-inflammatory cytokine interleukin (IL)-1α exerts therapeutic effects after stroke by promoting proangiogenic and neurogenic mechanisms, potentially outweighing its pro-inflammatory effects. 36  On the contrary, there is a long history of neutral and negative efficacy trials of stand-alone neuroprotection in the clinic. 38 In addition to quality limitations in some preclinical studies, the previous failure of pharmacological neuroprotection emerged from design differences in preclinical versus clinical trials. 39 For instance, many pharmacological neuroprotectants are most effective when given shortly after stroke onset. This also accounts for nerinetide, which was most effective when given 1.5 h after stroke onset in primates. 40  in mice. Prophylactic, that is, 30 min prior to the hypoxic-ischemic injury, but not delayed edaravone application mitigated ischemic tissue damage and improved functional outcome. 42 Despite these encouraging findings, we still lack profound proof of concept for the perioperative neuroprotection paradigm, and future preclinical research is needed to explore potential therapeutic opportunities.
Another interesting option is metabolic support of brain tissue at risk for infarction. This paradigm would neither prevent PS nor longterm tissue damage. However, it may decelerate penumbra decline and thus increase the time available for recanalizing interventions.
Options comprise the pharmacological increase of collateral perfusion 44,45 or replenishment of energy-containing substrates such as adenosine triphosphate. 46 Although additional preclinical investigations are required to further assess and optimize the metabolic support paradigm particularly in an experimental PS setting, it was already applied successfully in aged animals and subjects expressing comorbidities similar to those of patients at risk for PS. 46 Although successful, therapeutic effects are moderate and best only if the metabolic support is started relatively early after stroke onset. The approach could be initiated prior to surgical interventions in patients characterized by a relevant risk for PS. It might particularly benefit those in whom PS is not caused by a major embolic event blocking blood flow to a certain brain area, but primarily on cerebral hypoperfusion during surgery with a residual blood flow to the affected area.
Notably, specific agents may also extend the time window for vascular recanalization and further neuroprotectant administration. In summary, pharmacological neuroprotection represents a promising strategy in the setting of PS based on the unique fact that initiating neuroprotection is possible prior to the ischemic event.

| Inhalative substances
Some frequently used volatile anesthetics and other gaseous compounds are reported to have neuroprotective capabilities, which might be particularly valuable in a perioperative setting.
Isoflurane and sevoflurane exposure prior to, during, and after cerebral ischemia was shown to reduce brain damage in rodents. 48,49 Therapeutic effects include a reduction in lesion volume, improvement of functional outcome, 50 attenuation of reactive astrogliosis, 51 and promotion of anti-inflammatory microglial and macrophage phenotypes. 52 Research revealed numerous molecular pathways that mediate the beneficial effects of volatile anesthetics.
There is strong evidence for isoflurane-induced neuroprotection. Ischemic damage is inversely related to duration and dose of isoflurane anesthesia in tMCAO. 53 Neuroprotective effects persisted at least 8 weeks after focal cerebral ischemia in rats. 48 These effects are so strong that they may even mask the therapeutic benefits of potential neuroprotective drugs in stroke models. 54 Moreover, volatile anesthetics including isoflurane show a dosedependent protective effect on the incidence and severity of early postoperative ischemic stroke in patients undergoing non-cardiac surgery. 55 On the contrary, both isoflurane and sevoflurane also exert detrimental effects on cognitive function in fetal and aged brains both in experimental models and patients [56][57][58] although their impact is still controversially discussed for healthy adults. 59 Both agents are also believed to induce inflammation-mediated neurotoxicity in the young adult and aged brain. 60,61 Isoflurane also opens the bloodbrain barrier. 62 Moreover, anesthesia is often related to hypotension what is critical in the context of ischemic stroke.
Interestingly, neurotoxic effects predominantly emerge in the absence of brain injury, while neuroprotective effects are observed in cases of ischemic or traumatic brain injury. 63  Normobaric and hyperbaric oxygenation have been discussed as potential neuroprotective approaches for decades. Currently, normobaric oxygenation is revisited. 66 Meta-analysis indicates a potential benefit of the approach, 67 but large-scale multicenter clinical trials are needed to assess this potential benefit. The situation is comparable for hyperbaric oxygenation. 68 Results of these ongoing assessments in spontaneous stroke should be awaited before discussing a potential application in PS.
Profound neuroprotective effects were also shown for other substances such as inhalational nitric oxide (iNO). iNO dilatates blood vessels in areas of low oxygen concentrations. This effect is clinically utilized since the 1990s, for instance in newborns with persistent hypertension due to pulmonary vasoconstriction. 69 iNO also increases cerebral blood flow in the penumbra during acute stroke. 70 However, long-term application of iNO comes with drawbacks such as methemoglobin formation and may cause mild to moderate hypotension. 71,72 Enhanced NO levels in the neuronal compartment are toxic, but the relatively short half-live of NO and its rapid binding to hemoglobin make iNO an unlikely cause of neurotoxicity, especially when applied at concentrations of 50 ppm or lower. These low con- The application of hydrogen has been reported in experimental stroke. The primary mode of action seems to be antioxidation that counters cytotoxic effects of reactive oxygen species. 73 Hydrogen also possesses strong anti-inflammatory effects that can be beneficial after stroke. For instance, hydrogen can attenuate post-stroke activation of microglia while shifting microglial polarization toward the anti-inflammatory M2 type. 74 Hydrogen also attenuates the increase of pro-inflammatory M1 macrophages. 75 Anti-apoptotic effects of hydrogen have been reported, 76 and it stabilizes the blood-brain barrier after stroke in hypertensive rats. 77 Of note, hydrogen can also be applied intravenously after being enriched in saline. The procedure is safe in patients after ischemic stroke including those treated with tissue plasminogen activator. 78 Hydrogen application seems to be safe even in higher concentrations although the expression of certain enzymes such as aspartate aminotransferase, alanine aminotransferase, and γ-glutathione transferase can decline after exposure. 79 Application of hydrogen to mitigate PS may be considered after defining the most effective way of administration (gaseous versus enriched in saline) and further confirming safety. An advantage may be its compatibility with recanalization approaches and effectiveness in comorbid individuals.
Noble gases such as xenon and argon can exert neuroprotective effects. Argon may be preferred over xenon as it is less sedative and shows robust beneficial effects in experimental stroke and global cerebral ischemia. Argon reduces ischemic lesion size and improves functional outcome after 2 h of tMCAO, 80 and prolonged exposure of 24 h is safe in rodents. 81 Argon application is also safe in hemorrhagic stroke. Preclinical data are missing on PS, and argon was not yet applied clinically for NS stroke or PS. However, it might be an interesting therapeutic candidate for both PS prevention and treatment even in patients being ineligible for recanalization therapies because it is even beneficial after pMCAO. 81 When applied therapeutically, the need for timely PS diagnosis remains.

| Intravenous anesthetics for neuroprotection
Propofol was found to exert neuroprotective effects after ischemic Ketamine and dexmedetomidine are other, frequently applied anesthetic drugs with potential neuroprotective effects. Ketamine has anti-apoptotic and anti-neuroinflammatory effects after brain damage and preserves cognitive function after major cardiac surgery in humans. [85][86][87] Most recently, Xiong et al. showed that MCAO-induced brain damage was significantly attenuated by administration of (R)-ketamine, but not (S)-ketamine. The effect was seen after treatment prior to or after MCAO. 88 Dexmedetomidine is a selective α2-adrenergic agonist that was tested as a neuroprotective agent in experimental stroke models. Dexmedetomidine at a dose of 3 μg/kg given subcutaneously 30 min before and 3, 12, 24, and 48 h after global ischemia in gerbils protected neurons in the hippocampal CA3 and dentate gyrus. 89 Intravenous infusion of dexmedetomidine (9 μg/kg) during tMCAO decreased infarct volume by 40%, being more effective than the NMDA receptor antagonist CGS-19755, although a minor increase in blood glucose and hypotension was observed. 90 Despite these early yet promising results, dexmedetomidine was never tested in clinical neuroprotection trials.

| Cell therapies
Immune and stem cell therapies (ICT and SCT, respectively) were extensively investigated in animal models of ischemic stroke and are considered a promising therapeutic option. 39,97,98 In the field of ICTs, regulatory T-cell therapy is an important approach.
CD4 + CD25 + Foxp3 + regulatory T (Treg) cells exert neuroprotection after ischemic stroke by suppressing neuroinflammation. 99,100 Adoptive transfer of Tregs was shown to ameliorate BBB damage in the acute phase of stroke 101 and accelerates neural stem cell propagation in the subsequent brain repair phase. 102 Inhibitory signaling molecules such as cytotoxic T-lymphocyte-associated protein 4, programmed cell death-ligand 1, and IL-10 contribute to the Tregmediated attenuation of neuroinflammation after stroke. [103][104][105] Tregs also promote neural stem cell proliferation by secreting IL-10 and by suppressing astrogliosis. 106 press astrogliosis and foster neurological recovery, and that these processes peak about 2 weeks after stroke. 108 However, the therapeutic window for Treg transplantation is not known. Age 109 and sex 110,111 can influence the potency of Tregs and also affect microglial function. 112 Any beneficial effects of ICTs in NS stroke or PS remain to be clinically investigated.
Another option of cell therapy for NS stroke is SCT. First clinical trials are underway to assess SCT safety and efficacy in NS stroke. [113][114][115][116] Adult cell populations such as mesenchymal stem cells (MSCs) from bone marrow 117,118

| Pre-and postconditioning strategies
Conditioning strategies are an emerging paradigm in experimental stroke research and already showed remarkable effects in other ischemic conditions such as myocardial infarction. 136 Conditioning could be applied before (ischemic preconditioning; iPreC) or after the ischemic event (ischemic postconditioning; iPoC). The principle of iPreC is to expose the target organ (or a peripheral target) to one or multiple brief ischemic episodes. This induces tolerance against subsequent and more severe ischemic injuries, including in the brain. 137 The effect is mediated by profound changes in cellular gene expression, for instance downregulating metabolism and immune responses. 138 Similar to neuroprotection prior to the ischemic injury, iPreC may not play a major role in NS stroke, but would be applicable in patients at high risk for PS. Although the brain itself can be conditioned effectively in experimental subjects, 139,140 the clinically preferred approach would be remote iPreC, for instance in a limb (peripheral conditioning).
The therapeutic mechanisms of iPoC differ from those of iPreC. They include the mitigation of reperfusion injury 138 and post-stoke edema, 141 BBB protection, 142 and reduction in apoptosis. 143 Some studies suggest that the therapeutic time window for remote iPoC may be surprisingly wide. Although a reduction in infarct volume was achieved only when applying iPoC within 24 h after 60 min tMCAO in mice, best functional recovery and even limited neuroregeneration was seen when iPoC started 5 days after tMCAO. 144 Moreover, iPoC may extend the therapeutic time window for recanalization. 145 Importantly, conditioning approaches are already tested clinically, 146 and there is preliminary evidence for a modest therapeutic benefit in non-human primates 147 and humans. 110 However, as in other fields, there are design differences between preclinical and clinical studies 148 so final appraisal of the therapeutic impact is not yet possible.
There are also non-ischemic conditioning strategies. For instance, a single pretreatment with isoflurane induces ischemic tolerance in the rat brain. 149

| Restorative approaches
Tissue restoration after stroke, regardless of occurring spontaneously or as PS, is currently thought to be challenging. The reason is a lack of anatomical, cellular, and molecular cues allowing the recapitulation of brain development as seen during ontogenesis. 120 Biomaterials and scaffolds may help to provide such cues in larger lesions, 152 whereas smaller lesions or those predominantly situated in the white matter may be attractive targets for restorative therapies. 25,153 However, a profound preclinical proof of concept of restorative approaches is not available so far, and any clinical application is unlikely to happen in the near future.
Another aspect of a potential brain repair is attracting increasing attention recently. Immune cell-mediated long-term neuroinflammation is an intriguing target for strategies trying to address or modulate various endogenous regenerative processes such as neurogenesis, white matter repair, and others. 154

CO N FLI C T O F I NTE R E S T
The authors declare that there are no competing interests.