Early changes in physiological variables after stroke

Several aspects of physiology, notably blood pressure, body temperature, blood glucose, and blood oxygen saturation, may be altered after an ischemic stroke and intracerebral hemorrhage. Generally, blood pressure and temperature rise acutely after a stroke, before returning to normal. Blood glucose and oxygen levels may be abnormal in individuals, but they do not follow a set pattern. Several aspects of these physiological alterations remain unclear, including their principal determinants - whether they genuinely affect prognosis (as opposed to merely representing underlying processes such as inflammation or a stress response), whether these effects are adaptive or maladaptive, whether the effects are specific to certain subgroups (e.g. lacunar stroke) and whether modifying physiology also modifies its prognostic effect. Hypertension and hyperglycemia may be helpful or harmful, depending on the perfusion status after an ischemic stroke; the therapeutic response to their lowering may be correspondingly variable. Hypothermia may provide benefits, in addition to preventing harm through protection from hyperthermia. Hypoxia is harmful, but normobaric hyperoxia is unhelpful or even harmful in normoxic patients. Hyperbaric hyperoxia, however, may be beneficial, though this remains unproven. The above-mentioned uncertainties necessitate generally conservative measures for physiology management, although there are notably specific recommendations for thrombolysis-eligible patients. Stroke unit care is associated with better outcome, possibly through better management of poststroke physiology. Stroke units can also facilitate research to clarify the relationship between physiology and prognosis, and to subsequently clarify management guidelines.


Introduction
It is well established that stroke unit care improves outcome. [1] How this is achieved remains unclear, but close monitoring and maintenance of physiological homeostasis may contribute significantly to this beneÞ t. [2][3][4] However, our understanding of the changes in the main modifiable physiological parameters, namely blood pressure, body temperature, blood glucose and blood oxygen saturation, and the impact that these changes have on stroke outcome remains incomplete. Threshold levels for instituting treatment to modify these parameters, targets to be achieved with treatment and the eff ectiveness of such treatment remain uncertain. In this paper, we will review what is known about acute poststroke changes in these physiological variables, the causes of these changes, their prognostic significance, and the effects of treatment to modify these parameters.

Ischemic stroke
Changes in blood pressure (BP) aft er an ischemic stroke are important because there is impairment of cerebral autoregulation, [5] which, in normal circumstances, acts to maintain constant cerebral blood ß ow despite changes in systemic BP. With impairment of autoregulation, changes in systemic BP may aff ect cerebral perfusion, especially in the penumbral tissues, and may therefore aff ect the survival of ischemic tissue and the neurological outcome.
Both the systolic and diastolic BP are higher aft er stroke. It appears to rise acutely at the time of stroke. [6] Blood pressure then falls over the next 7-10 days, [7,8] with most of this fall occurring within the Þ rst 1-2 days. [9,10] The mechanisms driving these BP changes are uncertain, though there are several plausible explanations. A relationship between severe stroke and high poststroke BP has been documented, [11] although others have found that severe stroke is associated with a lower BP than stroke of mild to moderate severity. [10,12] How stroke severity might modulate BP is unknown. Acute psychological stress related to the process of hospital admission has been suggested to be a cause, [13] although a study in which high BP was documented aft er admission with stroke but not aft er medical admissions with other acute problems suggests that the BP elevation is stroke-speciÞ c. [14] However, since psychological stress cannot be quantiÞ ed, it is possible that the stroke group experienced more psychological stress than the control group. A neuroendocrine stress response may contribute and BP has been correlated with salivary [15] and serum [16] cortisol and urinary catecholamine [16] levels, although others have found no correlation between serum cortisol and BP. [17] Inß ammation may also play a role, one group Þ nding an association between BP and C-reactive protein (CRP) levels. [18] Inflammation and a neuroendocrine stress response may be linked, given the reported relationship between interleukin 6 (IL-6) and cortisol levels. [19] Other associations with high poststroke BP include a history of hypertension [10] and the requirement for antihypertensive medications, [10] although the latt er is likely to be an eff ect rather than a cause of higher BP.
It is hypothesized that BP elevations aft er ischemic stroke represent an adaptive response that helps to maintain the cerebral blood ß ow and perfusion of the ischemic penumbra, despite loss of cerebral autoregulation. [20] Conversely, it is of clinical concern that excessive rise in BP could lead to neurological deterioration from hemorrhagic transformation, especially in the presence of a damaged blood brain barrier.
The association between BP and the outcome of stroke is unclear, with poor outcome being associated with either absolute high [11] or low [21] BP, or having no association with BP. [22] Recent studies have identified a U-shaped relationship between BP and outcome, with poor outcome at either end of the BP spectrum. [12,23] Methodological issues, including cohort selection, the timing of BP measurements, and the timing and selection of endpoints, might contribute to the diff ering conclusions drawn by investigators.
The prognostic impact of BP levels appears to vary between ischemic stroke subtypes. One study found a U-shaped relationship between admission BP and mortality only in cardioembolic but not lacunar strokes [24] and another found that the relationship between admission BP and mortality in lacunar stroke was linear, with no harm resulting from low BP. [23] These Þ ndings might be explained by the lack of an ischemic penumbra in a lacunar stroke, [25] where low BP might not signiÞ cantly worsen hypoperfusion and therefore, tissue survival in the ischemic territory.
Lower admission BP has also been associated with poor outcome in thrombolysis-eligible patients. [26] Patients eligible for thrombolysis must present within three hours of stroke onset, when there is likely to be a large penumbra and, therefore, a proportionately greater vulnerability to hypoperfusion. However, thrombolysis protocols also exclude patients with high BP (>185/110 mm Hg); so any deleterious eff ect of high BP might not be evident in these cohorts.
One study also showed that cardioembolic stroke patients had lower BP in the Þ rst 24 h and poorer outcome, [21] as compared to those with non-cardioembolic strokes, although this may merely reß ect the impact that more severe stroke in cardioembolic patients has on both BP levels and outcome.
Although absolute systolic blood pressure (SBP), diastolic blood pressure (DBP) and mean arterial pressure (MAP) values are the most commonly studied indices of BP, other aspects of BP have been linked with stroke outcome. Elevated pulse pressure, [27] wide ß uctuations in both SBP and DBP within the Þ rst three hours of stroke, [28] greater variability of the DBP within the Þ rst 72 hours [26] and increased beat-to-beat SBP and DBP variability within 24-72 hours of stroke [29] have all been associated with poor prognosis.
The normal circadian variation in BP, where nocturnal BP falls relative to daytime BP (dipping), may be disrupted aft er stroke. [30,31] This is possibly more common after right-sided [32] or insular or thromboembolic [33] infarction, and may recover in the subacute and chronic phases. [16] There is some evidence that loss of dipping may be associated with poor outcome [34] and that its preservation is associated with bett er outcome. [35] There are no trials investigating the eff ects of maintaining a nocturnal BP dip, although there is some interest in the chronotherapeutic maintenance of circadian rhythms in other situations. [36] The prognostic signiÞ cance of dynamic ß uctuations in postadmission blood pressure varies. Poor outcome may be seen with spontaneous falls in systolic [37] and diastolic [38] BP; it can also be seen with spontaneous BP elevations. [11] Conversely, good outcome has been seen with spontaneous BP falls. [39] This has resulted in the investigation of both pharmacological BP elevation and reduction as potential strategies for the treatment of acute stroke.
Therapeutic BP elevation has been studied particularly in cases where relative hypoperfusion was suspected.
Such cases include patients with lower BP (SBP <140 mm Hg), [40] clinical deÞ cits that ß uctuate in parallel with BP changes [41] or perfusion-diff usion mismatch (a marker for the presence of penumbral tissue) on MRI. [42] In these small studies, BP elevation was shown to be safe and to partially reverse neurological deficits. [40][41][42] Pharmacological BP elevation may be beneÞ cial in carefully selected stroke subgroups, but this requires further study. [43] Pharmacological BP lowering after ischemic stroke remains a controversial topic. Long-term BP lowering is clearly beneÞ cial, signiÞ cantly reducing the incidence of recurrent stroke events. [44] However, what remains uncertain is the optimum timing of the initiation of therapy; speciÞ cally, whether it is safe to initiate BP therapy within the time window for the persistence of penumbral tissue, with the attendant theoretical risk of affecting penumbral salvage and overall neurological outcome. Whether to continue pre-existing antihypertensive therapy is also unclear and the subject of an ongoing study. [45] Caution is suggested by the results of research on the calcium channel blocker nimodipine, which, when studied as a neuroprotective agent, produced a worse outcome, possibly as a result of BP lowering in the treated group, [46] although this eff ect may have varied, depending on the stroke subtype. [47] Conversely, the ACCESS study was terminated prematurely when it was found that early treatment with candesartan was associated with a reduction in 12 month mortality and recurrent vascular events, despite there being no signiÞ cant diff erence in the BP between treatment arms; however, no eff ect on early functional outcome was seen. [48] Some antihypertensive agents may lower systemic BP without aff ecting cerebral blood ß ow, as seen in the studies of perindopril [49] and glyceryl trinitrate. [50] Such agents may be safe to use early, or even beneÞ cial; however, this remains unproven.
Some unrelated interventions and even routine activities can influence systemic BP after stroke. Insulin therapy, [51] therapeutic hypothermia [52] and experimental neuroprotective agents [46] may alter BP, potentially aff ecting the results of clinical trials; therefore, knowledge of the relationship between BP and prognosis is critical for the accurate interpretation of trial results. Even elevating the head of the bed to 15° or 30° can cause the MAP and cerebral perfusion pressure to fall signiÞ cantly. [53] In another study, aft er a meal, SBP fell >10 mm Hg in a quarter of the stroke patients. [54] Clarifying whether these alterations in systemic and cerebral pressure are clinically signiÞ cant is important, given that these activities are potentially modiÞ able in almost all stroke patients.

Intracerebral hemorrhage
Blood pressure also rises aft er intracerebral hemorrhage (ICH), perhaps even more than aft er an ischemic stroke, and falls over the next seven days. [55] It is also possible that the rise may sometimes precede and trigger the ICH. [56] The mechanisms behind BP changes aft er ICH remain uncertain. Because there is no ischemic penumbra around hematomas, [57] the BP rise is unlikely to be adaptive, especially as autoregulation seems to be preserved aft er ICH, and changes in systemic BP are not completely transmitt ed to the cerebral circulation. [58] A rise in BP may also be harmful, higher BP having been associated with hematoma enlargement [59] and poor prognoses. [60] Systemic BP, however, has not been universally shown to aff ect hematoma enlargement [61] or ICH prognosis [62] and lowering of BP has not yet been shown to improve prognosis, [63] although prospective trials are ongoing. [64] Temperature Ischemic stroke Body temperature rises by about 0.2°C over the Þ rst 24-36 h [65,66] aft er an ischemic stroke. The temperature rise is greater aft er moderate to severe stroke (0.35°C in a cohort with National Institutes of Health Stroke Scale (NIHSS) scores ≥ 6). [66] Boysen and Christensen, for example, observed a temperature rise only in those with severe stroke (Scandinavian Stroke Score, SSS ≤ 25), with no temperature change aft er mild to moderate stroke. [67] They also found that the admission temperature was more often low (<36°C) after severe stroke, possibly due to immobility and low environmental temperatures. [67] Higher temperatures have also been seen in cohorts where severe stroke was deÞ ned by large infarct volumes. [68] Following this rise, temperatures appear to fall over the Þ rst Þ ve days aft er stroke, [69] although the rate of normalization has not been speciÞ cally studied.
The normal circadian rhythm of temperature is also disrupted aft er severe stroke, possibly due to impaired consciousness or physical inactivity [70] and this may be associated with a poor prognosis, [71] although it is not clear that the relationship is independent of stroke severity.
Although the association between temperature elevation and stroke severity seems Þ rm, the mechanism through which this eff ect is generated remains unclear.
Infection is common aft er stroke, aff ecting 25-35% of the patients. [72,73] In some cases, the infection may precede the stroke and has been implicated in the causation. [74] Vila et al. found that IL-6 levels were associated with higher temperatures and larger infarctions in a cohort of patients without infection. [75] Audebert et al. found that the CRP, white cell count and temperature were high in the acute phase of stroke, but normalized within days of successful thrombolysis. [76] These results suggest that ischemic or infarcted brain tissue promotes inß ammation and hyperthermia. There may also be a role for a neuroendocrine stress response, with a correlation between cortisol levels and serial temperatures being documented. [17,77] Higher temperatures are more common in those requiring antipyretics, but this is more likely to have been a consequence of higher temperatures rather than a cause. [66,67] Low admission temperatures have been correlated with poor prognosis, [67] although this relationship was not shown to be independent of stroke severity. Most investigators, however, have found the opposite eff ect.
In one meta-analysis, high poststroke temperatures increased mortality with an odds ratio (OR) of 1.19 [78] and this relationship has been replicated subsequently. [79] High temperatures appear particularly harmful in those treated with thrombolysis, [80] although the mechanism for this remains unclear.
Although temperature appears to influence stroke prognosis, [78] stroke severity [67] and inflammation [81] may be responsible for some or all of its apparent eff ect. Stroke severity inß uences both stroke prognosis [67] and temperature [66] and the eff ect of temperature on outcome has been variably lost [82] or retained [83] aft er adjustment for stroke severity. Inß ammation similarly inß uences both temperature [76] and stroke prognosis. [81] Again, the relationship between temperature and outcome has been variably lost [68] or retained [75] aft er adjustment for inß ammation. It is also possible that inß ammation itself is an epiphenomenon of stroke severity, [84] although this theory is not universally supported. [85] Alternatively, each of these factors could independently aff ect prognosis, and one study found independent associations between all three factors and early neurological worsening. [75] These relationships require clariÞ cation to enable determining which of these factors might be suitable targets for treatment. It is, of course, hoped that simply lowering body temperature poststroke might improve outcome.
To this end, antipyretic agents administered orally, [86] rectally [87] or intravenously [87] have been used to modify temperature. The eff ect of antipyretics varies from zero [87] to around 0.2°C. [65,88] The eff ect of antipyretics, however, is more marked when larger temperature rises are expected; a 0.4°C treatment eff ect was seen in a cohort treated within 24 h of stroke onset. [69] There is no eff ect from these agents at 5 days poststroke. [69] These Þ ndings suggest that antipyretic agents prevent fever rather than actively lower temperature poststroke. ClariÞ cation of the eff ects of these agents would facilitate the assessment of protocols that combine treatments. [89] Active cooling techniques are highly eff ective at lowering body temperature and are generally employed with the aim of inducing hypothermia. The two main methods used are surface [90][91][92] and endovascular [93][94][95] cooling. Methods to selectively cool the brain, [96] for example with a cooling helmet, [97] have also been tried. Very low temperatures can be achieved with these techniques, but most protocols aim for a target temperature of 32-33°C. [94,98] This generally requires the patient to be intubated, with neuromuscular blockade used to manage shivering. Protocols avoiding the need for intubation would be more widely applicable, but the depth of hypothermia possible in awake patients is generally limited to around 35.5-36.4°C, [90,92] although a recent study achieved median core body temperatures of 33.4°C with endovascular cooling, in awake patients. [93] Several technical issues regarding induced hypothermia require clariÞ cation before this treatment can be evaluated in large randomized trials. These issues are related to the following -the most appropriate patients to be treated, the time window for commencing temperature lowering treatments, how low the temperature should be reduced to, how long the hypothermia should be maintained and how quickly to re-warm the patient. [99] Given the intensive resources required for monitoring patients undergoing therapeutic hypothermia and the need for large potential treatment eff ects to Þ nd beneÞ t with small numbers, most studies, to date, have focused on patients with more severe stroke, [90] for example large middle cerebral artery (MCA) territory infarcts. [91,94] There are no studies that have speciÞ cally addressed the issue of the time window for initiation of hypothermia. For practical reasons, subjects have had hypothermia initiated between 4 h [91] and 60 h [90] poststroke. It would seem likely that the maximum beneÞ t will result from earlier initiation of treatment.
Most investigators have maintained hypothermia for one to three days, [52,90,94,95] although durations from six hours [92] to three weeks [98] have been studied. The optimal duration of hypothermia is unclear, these reports suggesting no clear beneÞ ts from either shorter or longer durations of hypothermia.
Slower rewarming seems to be safer than more rapid rewarming, there being a lower incidence of rebound elevations in Intracranial Pressure (ICP) and a lower risk of transtentorial herniation. [100] Other side eff ects of these techniques include hypotension (patients oft en require BP support), for example, with inotropes [90,95] and pneumonia. [95] To date, studies involving these techniques have primarily been aimed at evaluating the feasibility and safety of treatment rather than at demonstrating its eff ectiveness. Overall, these techniques do appear to be feasible and relatively safe. [92,95,98] Clinical endpoints have not clearly been improved with treatment, but there are studies demonstrating improvement in surrogate outcome; for example, microdialysis studies have shown lower concentrations of glycerol (a neuronal membrane component released from infarcted neurons) and lactate (indicating anerobic metabolism) with body temperatures under 34°C, and less ICP elevation [91] and cerebral edema [101] with lower body temperatures. DeÞ nitive studies are awaited.

Intracerebral hemorrhage
Temperature rises aft er ICH, [67] probably more so than aft er an ischemic stroke. [102] The magnitude of the rise seen by Boysen and Christensen,[67] aft er severe (SSS < 25) ICH was about 1°C, beginning at about 4-6 h and being complete by about 12 h aft er the stroke. As with ischemic stroke, the temperature rise is more pronounced aft er severe ICH as assessed either clinically [67] or by hematoma volume. [103] As opposed to the uncertainty surrounding temperature changes aft er ischemic stroke, temperature elevations after ICH seem to be mostly an epiphenomenon of the severity of the hemorrhage. Unadjusted analyses have suggested a poor prognosis, with higher temperatures. [103,104] However, analyses incorporating adjustment for stroke severity have found no association with prognosis. [67,105] One study did show that 24-48 h or >48 h with a temperature above 37.5°C was associated with a poor Glasgow Outcome Scale (GOS) score at discharge, independent of Glasgow Coma Score (GCS) and hematoma volume, [106] suggesting that there may be a dose eff ect.
There are a few studies examining the eff ects of temperature lowering aft er ICH. Bett er 3-month outcome have been seen with nasopharyngeal cooling and indomethacin, as compared with no treatment. [107] Other groups have studied surface cooling in cohorts including both ischemic stroke and ICH. These results were discussed previously. [90] Overall, however, litt le is known about the beneÞ ts of temperature lowering aft er ICH.

Ischemic stroke
Glucose levels appear to rise aft er stroke. One study involving nondiabetic patients demonstrated a rise in median blood glucose level from 5.9 mmol/L at 2.5 h to 6.2 mmol/L at 6 h poststroke. [22] Indeed, poststroke hyperglycemia is a frequent phenomenon, with up to 50% of the patients having an initial blood glucose of above 6.0-7.0 mmol/L, [108,109] glucose levels that, in the fasting state, would be consistent with a pre-diabetic state. [110] In comparison, the prevalence of diabetes or impaired fasting glycemia is 34% in people aged over 60 years. [111] As one might expect, glucose levels poststroke are higher in patients with diabetes than in those without diabetes. [112][113][114] Some have suggested that the initial poststroke hyperglycemia resolves spontaneously in the acute phase. [112,115] Others have found no such change. [108] Dysregulated glucose metabolism has also been shown to extend for months poststroke. In patients without known diabetes, half of one cohort [116] and two-thirds of another [117] had diabetes or impaired glucose tolerance at three months poststroke. At three years aft er stroke, two thirds of another cohort had either diabetes or a pre-diabetic state, with half of these patients unaware of their status. [118] Hyperglycemia appears to be associated with more severe stroke, as assessed either with a clinical stroke scale [119] or by lesion volume. [120] A neuroendocrine stress response [121] and an inß ammatory response [122] may also play a role in generating hyperglycemia.
Stroke location may be important, with infarction of the insular cortex, a structure involved in the autonomic control of the neuroendocrine stress response associated with hyperglycemia; [123] however this association has been refuted. [124] Support for a neuroendocrine stress response comes from groups that have reported a correlation between glucose and cortisol levels aft er stroke, [17,125] although others have not found glucose to be associated with either cortisol [77] or catecholamine levels. [126] Although not a universal finding, [127] poststroke hyperglycemia has been associated with poor outcome [128] and seems to particularly aff ect outcome in patients without diabetes. In the meta-analysis by Capes et al., the relative risk of in-hospital/30-day mortality in patients with admission hyperglycemia (>6.1-7.0 mmol/L) was 3.28 (95% CI 2.32 to 4.64) in ischemic stroke patients without diabetes, but not significantly increased in patients with diabetes. [129] Similar Þ ndings have been published since this meta-analysis. [114,130] Outcome other than mortality have also been shown to be worse in hyperglycemic patients without diabetes than in other stroke patients. [113,131] Why hyperglycemia particularly affects stroke prognosis in patients without diabetes is unclear. Despite being apparently protective against ischemic damage in vitro, in vivo studies have consistently associated high glucose levels with harm. [132] Diabetes is associated with microcirculatory abnormalities in the brain, including arteriovenous shunting and a reduction in glucose transport across the blood-brain barrier. [133]. These processes would reduce the delivery of glucose from the blood to the brain of a patient with diabetes, thus possibly protecting them from high glucose levels aft er stroke.
Other factors could potentially explain the relationship between glucose and stroke prognosis. Hyperglycemia only increases growth of the infarct core in patients with surrounding hypoperfusion, [134] suggesting that hyperglycemic blood is only toxic to ischemic brain. Similarly, several studies have shown that non-ischemic brain surrounding lacunar infarcts [135] and, in turn, stroke prognosis, is unaffected [136] or perhaps even improved [137] by hyperglycemia. Conversely, Toni et al. found that a collateral blood supply could improve the prognosis if hyperglycemia coexists, [138] suggesting that glucose can protect ischemic brain. Several biochemical mechanisms, including excessive glutamate or lactate, vascular reactivity, or oedema formation [132] could possibly link glucose and stroke prognosis. However, these relationships remain too unclear to reconcile these apparently contradictory Þ ndings.
Hyperglycemia has a particularly potent adverse eff ect after thrombolysis. Hyperglycemic patients more commonly develop intracerebral hemorrhage after thrombolysis [139] and have overall poorer clinical [140] and radiological [141] outcome. Hyperglycemic patients are also less likely to recanalise with thrombolysis. [142] Even if recanalization occurs, hyperglycemic patients are more likely to deteriorate, [143] particularly if hyperglycemia occurs early aft er recanalization. [144] Hyperglycemia may be merely an epiphenomenon of other underlying processes. Given the association between stroke severity and hyperglycemia, [22,119] the repeated Þ nding that hyperglycemia has no association with prognosis aft er adjustment for stroke severity [82,145] suggests that in some cases hyperglycemia is an epiphenomenon of stroke severity. In other cases, however, hyperglycemia has aff ected prognosis independent of stroke severity. [146,147] Glucose has variably lost [148] or retained [75] an independent eff ect on stroke prognosis, aft er adjustment for IL-6 levels. Similarly, glucose has variably lost [125] or retained [149] its independent association with stroke prognosis aft er adjustment for cortisol levels. As glucose is associated with both inß ammation [75,122] and cortisol levels, [17,125] and as stroke prognosis is also associated with both inß ammation [81] and cortisol, [77] it is possible that these factors drive stroke prognosis and that hyperglycemia is merely an epiphenomenon of one or both of these factors.
Given the frequency of hyperglycemia and its eff ect on outcome, glucose-lowering therapy has potential as a widely applicable treatment aft er stroke. Insulin, speciÞ cally glucose-potassium-insulin (GKI) infusions, have been shown to be feasible and safe in acute stroke patients. [150,151] Insulin appears to have beneÞ cial eff ects, including anti-inflammatory, antioxidant and nitric oxide eff ects, which are independent of its ability to lower glucose levels, and which may be beneÞ cial in stroke. [152] Unfortunately, the large Glucose Insulin in Stroke Trial (GIST-UK) [51] was terminated prematurely due to slow recruitment, after 933 patients were randomized to either GKI or intravenous saline. Being correspondingly underpowered, the study failed to identify a treatment eff ect on mortality or other outcome. Further studies are clearly warranted.
There have been few studies with other agents used in the treatment of diabetes, although sulfonylureas were not shown to aff ect stroke prognosis in one trial. [153] Intracerebral hemorrhage Hyperglycemia aft er ICH is less well-characterized than aft er ischemic stroke. Apart from diabetes, [154] the most signiÞ cant determinant of hyperglycemia is the severity of the ICH, as assessed by the hematoma size [155] or other markers of severity. [154] In the meta-analysis by Capes et al., admission hyperglycemia was not associated with higher mortality in unadjusted analyses of either diabetic or nondiabetic ICH patients. [129] Since this meta-analysis, one group found that hyperglycemia had no association with outcome aft er adjustment for ICH volume and growth, two strong predictors of outcome aft er ICH. [62] Other studies, however, have shown an independent eff ect from glucose, even aft er adjusting for the volume [106,156] or other markers of mass eff ect [154] of the ICH. These subsequent results suggest that glucose does have an independent eff ect on ICH prognosis.
There are no treatment trials speciÞ cally addressing the eff ects of glucose lowering in ICH patients.

Oxygenation
Hypoxia is frequently reported aft er stroke, although the frequency depends on the deÞ nition used. Pulse oximetry identiÞ ed arterial oxygen saturations (SaO2) <90% for >10% of the recording time in 20% of one stroke cohort, [157] while 63% of another cohort had SaO2 < 96% for at least 5 min. [158] Hypoxia in stroke patients is commonly associated with co-morbidities such as respiratory tract infections and cardiac failure. [157] It also occurs overnight, with SaO2 falling below 90% for >30 min in 25% of patients in one study. [159] Hypoxia does not seem to be related to stroke severity. [157] Transient hypoxia has been observed during routine maneuvers such as MRI scanning (18% patients had SaO2 <90% for at least 1 min); [160] bed transfers (SaO2 fell ≥ 3% in 18% patients) [157] and nasogastric tube insertion, when diffi cult or prolonged (SaO2 fell below 90% in 21% patients). [161] Hypoxia is not related to overnight nasogastric tube feeding. [162] Hypoxia may [163] or may not [164] occur with oral feeding; this inconsistency may be due to the use of diff ering criteria for classifying patients as safe to feed orally.
Posture may aff ect oxygenation, with higher SaO2 levels when sitt ing upright or semi-recumbent, as compared with fully recumbent posture. [165] Positioning may be more likely to aff ect oxygenation in stroke patients who have coexisting respiratory problems. [166] It is generally assumed that hypoxia carries a poor prognosis and is rarely left untreated in clinical practice or research sett ings. In a mixed cohort of ischemic and hemorrhagic stroke, there was a univariate association between hypoxia (SaO2 < 90% for > 10% of the recording time) and death at three months poststroke, but this was nonsigniÞ cant aft er adjusting for stroke severity. [157] There is a corresponding lack of evidence that correction of hypoxia improves stroke prognosis.
Oxygen supplementation has been studied as a way to improve stroke outcome in normoxic patients. Understandably, most research has focused on ischemic stroke, where oxygen supplementation is seen as having a potential for neuroprotection. Potential benefits suggested by the results of animal studies include suppression of excitotoxicity, oxidative and nitrosative stress, inß ammation, and apoptosis. [167] Despite theoretical concerns about the production of harmful superoxide free radicals and hyperoxia-induced vasoconstriction, some human studies suggest an overall beneÞ cial eff ect. Indeed, oxygen-induced vasoconstriction in nonischemic areas has been shown to shunt blood towards ischemic areas, thus paradoxically improving perfusion. [168] Oxygen supplementation using both hyperbaric (HBO) and normobaric oxygen (NBO) to improve outcome in normoxic patients has been studied. Hyperbaric oxygen (HBO) has been more widely studied and the early studies showed occasional dramatic improvements temporally related to HBO therapy, with some of these improvements occurring even several weeks and months aft er stroke. [169,170] More recent studies have focused on acute stroke, but the results have either been neutral [171] or have suggested bett er outcome in the non-HBO group. [172] Various methodological explanations for these conß icting results have been offered. The optimum treatment pressure remains unclear, with pressures from 1.5 atmospheres absolute (ATA) to 2.5 ATA promoted as being the most appropriate. [167,173] The use of oxygen therapies by control groups, usually in an att empt to maintain blinding, may have off ered some treatment eff ect and, therefore, may also have aff ected the results in some studies. [171,172] The initiation of HBO therapy beyond the Þ rst few hours aft er stroke has also been criticized, because extrapolation from other stroke therapies [173] and animal studies [174] suggests that poor outcome from treatment this late aft er stroke are to be expected.
Normobaric oxygen (NBO) is capable of increasing the SaO2 in normoxic stroke patients, [175] but unselected patients appear to do poorly with NBO therapy. [176] Patients presenting with large MCA territory infarcts appeared to beneÞ t from oxygen supplementation (FiO2 40% by Venturi mask), when treated within 48 h of stroke, perhaps because these patients are more likely to have a penumbra. [177] Similarly, diff usion weighted imaging (DWI) volumes were smaller during treatment with highß ow (45 L/min) mask oxygen given within 12 h of stroke in patients with MRI evidence of perfusion-diff usion mismatch. [168] As with HBO, the dose and timing issues are unclear for NBO therapy also.
Oxygen therapy might protect the ischemic penumbra by suppressing harmful processes such as edema, [178] inflammation and apoptosis. [167] While NBO does not appear to provide permanent protection from recruitment of the ischemic penumbra into the infarct core, [179] it might be used to maintain the penumbra whilst awaiting reperfusion and thus prolonging the therapeutic time window for reperfusion therapies. [168] Applying HBO in the same way poses signiÞ cant logistical challenges,. Treatment would be restricted to a small subgroup of the already small percentage of thrombolysis-eligible patients. [180] Monitoring of physiology in the stroke unit While stroke-unit care reduces mortality and improves functional independence, [1] the mechanisms for this benefit remain unclear. One mode that stroke-unit care may help patients is through careful monitoring and optimization of physiological variables. [181]  (Please refer to full text of guidelines for details of specifi c recommendations) Stroke units might facilitate the maintenance of physiological homeostasis by bringing together medical, nursing and other staff experienced in stroke care; the necessary equipment for monitoring physiological parameters; and protocols to manage aberrant physiology. Monitoring in stroke units also facilitates data collection for subsequent reÞ nement of management protocols.
There is some evidence that intensive physiological monitoring in a stroke unit improves outcome. [2,3] In a randomized study of 54 ischemic stroke patients, [2] continuous monitoring of BP, electrocardiogram (ECG), body temperature, and pulse oximetry for at least 48 h improved 3-month mortality, as compared with 6thhourly manual measurements of BP, temperature, and heart rate and intermitt ent pulse oximetry (3.7% vs 25.9%; P=0.05). The same protocol for managing abnormal physiology was used in both groups, suggesting that early detection and management of abnormalities through continuous monitoring improved outcome. Improved outcome was also observed in a nonrandomized comparison between continuous monitoring of BP, ECG, temperature, oxygen saturation, respiratory frequency and electroencephalogram (EEG), and intermittent monitoring of BP, heart rate, and temperature. Aberrant physiology was more frequently detected in the continuously monitored group (64% vs 19%; P<0.0001), perhaps leading to more timely intervention and resulting in improved outcome. [3] One randomized study of 206 ischemic stroke patients showed that less DBP variability was associated with more frequent discharge to home at six weeks, although it is unclear whether this was due to more frequent monitoring or earlier mobilization in the stroke unit. [182] As can be seen from the above discussion, the optimum strategies for managing these physiological variables remain uncertain. Many stroke units have developed local management protocols guided by the available evidence, informed by experience, and modiÞ ed according to the resources available locally. Internationally, published evidence-based stroke-management guidelines provide some guidance. Table 1 summarizes the recommendations from published American, European, and Australian stroke treatment guidelines, regarding management of physiological parameters aft er acute stroke. There is general consistency across the guidelines, with some minor diff erences, generally in areas in which there is a lack of strong evidence to guide practice.

What do published guidelines say about management of physiological variables?
The guidelines generally recommend tolerating higher than usual BP aft er both ischemic stroke and intracerebral hemorrhage. Excessive elevations should be treated, but thresholds for initiating treatment and targets for treatment vary. All caution should be taken against aggressive and abrupt BP reductions. Treatment of fever with antipyretic agents is recommended, with a search for infection and antibiotics when appropriate. Treatment of hyperglycemia for both forms of stroke is recommended, although thresholds for initiating therapy vary. Treatment of hypoglycemia is advised. Hypoxic stroke patients should receive oxygen therapy, but evidence is insuffi cient for the guidelines to recommend oxygen therapy for normoxic patients.

Conclusions
Our understanding of the changes in BP, temperature, glucose, and oxygen levels after stroke remains incomplete. Further study should clarify the natural history of acute physiological changes, the mechanisms behind these changes, and the eff ect these changes have on outcome. If physiological abnormalities contribute significantly to poststroke outcome, they offer an obvious target for treatment. Intensive early monitoring and control of physiological parameters might then become the standard of care in the stroke units of the 21st century.