1.1 CARBON DIOXIDE AND MECHANISM OF ACTION

In the normal brain, the constancy of CBF and cerebral blood volume relies upon the intrinsic ability of the cerebral resistance vessels to alter their diameter in response to variations in blood pressure (CBF autoregulation) and changes in metabolic demands. One of the major products of cerebral metabolism, CO 2 , can alter cerebrovascular resistance and ultimately affect CBF [15]. CO 2 is freely diffusible between arterial blood and brain tissue. Thus it can diffuse into the vascular smooth muscle cell from either brain tissue or the vessel lumen, while hydrogen ions in the vessel lumen are prevented from reaching the smooth muscle cell by the blood-brain barrier [16]. Even though the chemical reaction between CO 2 and tissue water is simple:

CO 2 + H 2 O ´ H 2 CO 2 ´ HCO -3 + H +

the physiologic mechanisms that underlie the responses of the cerebral vessels to changes in CO 2 have not been simple to ascertain. The main mechanism of this effect upon cerebral vessels appears to be perivascular pH [17, 18], but there may be additional, yet unknown, mechanisms. There is also evidence to support that factors such as prostanoids [19-21] and nitric oxide [22] are involved in the response, whereas the influence of peripheral/central [23] nerves is conflicting, at least during hypercapnia. The influence of these factors on the cerebral carbon dioxide reactivity has mostly been explored during hypercapnia , while only few studies have addressed their influences in hypocapnic responses.

1.1.1 Extracellular fluid [H + ]

In 1961, Gotoh et al suggested that the action of CO 2 was mediated by direct effect of [H + ] on cerebrovascular smooth muscles [24]. Indeed, the majority of experimental studies support the concept that CO 2 regulates the cerebral circulation primarily by changes in pH in the extracellular fluid surrounding the vessels [17, 18, 25-28]. Kontos et al, found that marked changes in PaCO 2 did not affect cerebral vessel diameter unless a change in extracellular fluid pH occurred [17, 18]. However, it is not certain whether the effects on cerebral vessels are due solely to changes in extracellular fluid pH, or whether and to what extent changes in intracellular pH also contribute [29]. There is experimental evidence to support that the extracellular pH influences the vasodilatory response to a greater extent than the intracellular pH, as Toda et al [30], found that hypercapnic vasodilation in dog cerebral artery strips was reversed by infusion of sodium-bicarbonate, i.e., pH was raised while PCO 2 remained stable. In other experimental studies, blockade of adenosine triphosphate (ATP)-sensitive potassium channels completely inhibits the hypercapnia induced vasodilation [31, 32] as well as the hypocapnia-induced vasoconstriction [33]. These findings suggest that opening and closure of these channels are of importance for the vascular responses to alterations in PaCO 2 . Recently, Nakahata et al [34], showed that blockade of ATP-sensitive potassium channels with glibenclamide completely abolished hypercapnia-induced vasodilation only when pH was decreased. Thus, CO 2 -induced alteration in pH appears to affect ATP-sensitive potassium channels either by opening (hypercapnia-induced acidosis) or closure (hypocapnia-induced alkalosis) of the potassium channels.

1.1.2 Prostaglandins

Prostaglandins appear to be mediators of the cerebral CO 2 response to hypercapnia [19]. Several studies have shown that hypercapnia elicits vasodilation that is accompanied by increased prostanoid synthesis, and that this vasodilation is inhibited by indomethacin, a cyclo-oxygenase inhibitor [19, 35, 36]. By contrast, other studies in other species did not find an effect of indomethacin on CBF [37, 38]. Also, indomethacin had no effect on dilator or constrictor prostanoids concentration in cerebrospinal fluid of newborn pigs in response to hypocapnia [39]. Nor did indomethacin affect pial arteriolar constriction in response to hypocapnia , or vasodilation when PaCO 2 was raised from hypocapnia to normocapnia [39].

1.1.3 Nitric oxide

Nitric oxide plays an important role in the tonic regulation of cerebrovascular tone and contributes to the changes in CBF produced by hypercapnia [40-42]. It has been suggested that the extracellular acidosis associated with hypercapnia might activate nitric oxide synthetase and increase nitric oxide production [22]. However, other studies suggest that nitric oxide is not solely responsible for the hypercapnia-induced vasodilation [41, 42]. It is assumed that the vasodilator response to nitric oxide in the cerebral circulation is secondary to stimulation of guanylyl cyclase [40] and involves activation of potassium channels [43].

1.1.4 Neural pathways

Definitive evidence to support the participation of either central and/or peripheral neural reflex pathways in mediating effects of CO 2 on the cerebral vessels are lacking [44]. However, in newborn piglets, Moore et al [45] showed that the hypocapnic vasoconstriction was attenuated after alpha blockade in the forebrain and brainstem and after cord transection, whereas the hypercapnic reactivity was unaltered by these procedures.

1.2 AIMS

The aims of the present clinical studies were to evaluate CBF and oxidative and amino acid metabolism in patients with FHF with special focus on the effect of short-term mechanical hyperventilation upon these measures. Also the influence of hyperventilation upon arterial content and cerebral fluxes of certain biomarkers and neuropeptides is investigated and discussed.

2.1 PATIENTS

From February 1996 until June 2000, patients with FHF referred to the department of Hepatology, Rigshospitalet, were enrolled consecutively. All patients were investigated within 24 hrs after the first appearance of stage III-IV hepatic encephalopathy, and after intubation and mechanical ventilation had been instituted. All patients had arterial lines, central venous catheters and a catheter placed retrogradely in the internal jugular bulb. They all received N-acetylcysteine and 20% glucose intravenously, and were all sedated with midazolam. No forms of extracorporeal renal or liver assist devices were used before or during the studies.

Patients included in paper I were not included in other studies, and no quantitative measurements of absolute CBF were performed. Data in paper II-VII were obtained from 16 consecutively enrolled patients with FHF. Except for the study of regional CBF (II) and the validation study of the transcranial doppler (III), metabolic data and CBF were obtained in all 16 patients. Data on some of the brain fluxes during hyperventilation were only available in 14 patients.

Healthy subjects in the study of regional CBF (II) only participated in that study, while the group of healthy subjects are the same in paper IV-VII. Likewise data on patients with cirrhosis of the liver without hepatic encephalopathy are obtained from the same 5 patients in paper IV-VII. Additionally arterial blood samples of biomarkers of brain cell damage (VI) were collected in all patients with FHF and patients with acute on chronic liver disease.

Consequently, many data were obtained in the same group of patients, and many comparisons were made between the groups. In order to adjust for these multiple comparisons, Bonferroni correction factor 5 was applied to P values between 2 unpaired/paired groups. When more groups were compared Kruskal-Wallis analysis of variance (ANOVA) using Dunn's method for multiple comparison was applied.

2.2 METHODS TO EVALUATE CBF
AND CEREBRAL CO 2 REACTIVITY

The aims of the present clinical studies were to determine the effect of hyperventilation upon global and regional CBF as well as the cerebral metabolism of different substances. For this purpose tracer-kinetic methods were used. These methods are based upon the cerebral uptake and wash-out of a freely diffusible inert tracer administered either intravenously or by inhalation. Two different tracer kinetic methods were applied, i.e., the Kety-Schmidt technique and Single Photon Emission Computed Tomography (SPECT) (II-VII). As tracer substances, 133 Xenon and Technetium-99m Hexamethyl propyleneamine oxide ( 99m Tc-HMPAO) were used. In the clinical setting of FHF where mechanical hyperventilation is necessary, these methods are, however, too time-consuming. Accordingly, Transcranial Doppler technique (TCD) and the arterio-venous oxygen difference (AVDO 2 ) technique were evaluated as bed-side methods to monitor CBF responses to hyperventilation in patients with FHF (III).

2.2.1 The Kety-Schmidt technique

The Kety-Schmidt technique is based on Fick's principle and is considered to be the "Gold Standard" for measurement of global CBF in man [46]. In our clinical studies (II-VII), we applied the modified Kety-Schmidt technique, i.e., the desaturation mode. In this setting the brain is saturated by infusion of 133Xenon at a constant rate for 30 minutes, and followed by a desaturation period of 10 minutes, where blood samples are obtained pairwise from the radial artery and internal jugular vein [47]. The Kety-Schmidt technique is widely acknowledged as an accurate technique for determination of global CBF. However, one critical assumption of the Kety-Schmidt technique is that the average 133Xenon tension of the brain is the same as the 133Xenon tension of the cerebral venous blood. In the experimental setting with a wash-out period of 10 minutes this assumption is not usually met, due to heterogeneously perfused cerebral tissues, leading to under representation of low flow areas such as cerebral white matter and hence, a systematic overestimation of global CBF [47, 48]. By computer simulation Madsen et al [47], demonstrated that the modified Kety-Schmidt technique over-estimates global CBF by 10 to 15% in healthy subjects, and proposed a simple conversion algorithm of measured CBF to "ideal" CBF. However, since the experiments by Madsen et al [47] were performed on healthy subjects, the assumptions by which the correction procedure is based may be invalid when applied to patients with FHF. Accordingly, this correction algorithm was not applied in the present clinical studies.

2.2.2 Single Photon Emission Computed Tomography

SPECT was applied to evaluate the regional CBF distribution pattern (II) as well as the global and regional cerebral CO 2 reactivity to hypocapnia in patients with FHF (III). Quantitative measurements of CBF during normoventilation and hyperventilation were obtained with 133Xenon as the tracer substance. The use of freely diffusible tracers for the tomographic CBF calculation method is more complex, but founded on the same physical assumptions as originally proposed and described by Kety and Smith [46]. 1) simultaneous arrival of the tracer to all parts of the brain following administration, 2) instantaneous equilibrium between brain and blood, and 3) cerebral blood flow remains constant during the period of measurement. CBF values derived from SPECT studies are prone to inaccuracies due to problems with attenuation correction, Compton scattered radiation, the partial volume effect, and estimation of the arterial input function from the lung curve input function [49, 50]. Apart from the methodological problems, the most critical comparison between the Kety-Schmidt technique and SPECT, is that SPECT studies derive their CBF-values from one representative whole brain slice, while CBF obtained by the Kety-Schmidt technique is from the entire brain volume. To assure that no major systematic error was applied in our studies (III-VII), we determined global CBF by both techniques in six patients with FHF, and found similar CBF values for the Kety-Schmidt technique and SPECT, i.e., 41 ± 4 vs. 43 ± 9 mL/(100 g × min) (NS) during normoventilation, and 34 ± 5 vs. 36 ± 7 mL/(100 g × min) (NS) during hyperventilation (unpublished data) ( Table 1 ).

Since 133Xenon only allows for evaluation of regional CBF with low-image resolution, 99m Tc-HMPAO was used as tracer to obtain high-resolution SPECT images of regional CBF distribution pattern during hyperventilation (II). This tracer only yields relative CBF values. After intravenous administration of 99m Tc-HMPAO approximately 70-80% of the primary complex reaching the brain crosses the blood brain barrier [51]. Inside the brain it is rapidly converted to its non-diffusible hydrophilic complex and retained for many hours [52]. Similar to the Kety-Schmidt technique, SPECT has a low time resolution and requires transportation of the patient to the tomograph. The advantage is that it allows for regional CBF determinations.

2.2.3 Transcranial Doppler Sonography.

The TCD method determines the mean flow velocity (V mean ) of flow in blood vessels. Measurements of V mean were obtained from the middle cerebral artery through the temporal window with a 2 mHz probe and an insonation depth of 45-55 mm. The middle cerebral artery is the most extensively evaluated intracerebral vessel studied by TCD. The average V mean in healthy adult subjects derived from previous studies is 62 (33-90) cm/s [53, 54], but may be affected by a variety of factors such as hematocrit, PaO2 and PaCO 2 in the same manner and direction as these factors affect CBF. The TCD technique has previously been evaluated as a reliable tool to evaluate relative changes in CBF during autoregulation studies where mean arterial pressure is raised by infusion of norepinephrine both in healthy man and in patients with FHF [55, 56]. Relative changes in V mean however only reflect actual changes in CBF when the diameter of the insonated vessel and the perfusion territories remain unchanged during the study. In patients with FHF, it was found that the TCD technique correctly reflected changes in CBF during hyperventilation, but the numerical value of the CO 2 reactivity was slightly underestimated as compared to CO 2 reactivity obtained by the Kety-Schmidt technique and SPECT ( Figure 1 ) (III). This slight discrepancy is probably caused by a decrease in the diameter of the middle cerebral artery produced during hyperventilation in patients with FHF, since the CBF/V mean ratio decreased slightly from rest to hyperventilation, i.e., from 0.014 (range, 0.011 to 0.020) to 0.012 (range, 0.009 to 0.021) cm 2 /g (p=0.047) (III). Although, the TCD method only provides relative CBF values, it has several advantages over both the Kety-Schmidt technique and SPECT: it is non-invasive, relatively inexpensive, less time consuming and has a high time resolution. Moreover, the technique can be applied bedside and provides on-line information on cerebral circulation, and alterations in this.

2.2.4 Arterio-venous oxygen difference method

This method is based on Fick's principle. As long as the cerebral metabolic rate of oxygen is constant, then changes in AVDO 2 will reflect variations in CBF, i.e., AVDO 2 varies inversely with CBF. However, it must be emphasised that this relationship only holds true as long as the normal relationship between CBF and metabolism is maintained [57]. The AVDO 2 method has most of the same advantages as the TCD technique, but requires placement of an internal jugular bulb catheter and an arterial line. The disadvantage is that blood samples are obtained from one internal jugular vein only, thus the possibility exists of overlooking severe ischemia in brain tissue drained by the opposite internal jugular vein, and although the catheter is correctly placed in the internal jugular bulb there will be an extracranial contamination of 2% to 3% [48]. In patients with FHF, we found that the AVDO 2 method yielded the same global CO 2 reactivity in response to hyperventilation as those obtained by the Kety-Schmidt technique (Figure 1) (III). Thus, the AVDO 2 method is a reliable and easy way to evaluate the effect of mechanical hyperventilation upon CBF in the clinical setting of FHF, as long as the normal relationship between CBF and metabolism is maintained.

3.1. CBF IN FHF

Wide variations in CBF have been reported in patients with FHF ( Table 2 ). In the present work (III-VII), a reduced CBF was found in patients with FHF using both TCD technique (I) as well as tracer methods to measure absolute CBF (III-VII). This finding is in accordance with a number of previous studies [12, 14, 58-61]. Only, Ede et al [62] and Jalan et al [63-65] have reported significantly higher CBF values than that reported in other previous studies (Table 1). In the study by Ede et al [62] they report a normal value of 120 mL/(100 g × min) using their technique. This is a much higher global CBF value than reported previously in healthy subjects [66], i.e., ~50 mL/(100 g × min). The studies by Jalan et al [63-65] were, contrary to other previously published studies, performed during ongoing intracranial hypertension. In one of the clinical papers by Jalan et al [65], two groups of patients were included, one with intracranial hypertension and another without intracranial hypertension, demonstrating that only the group of patients with intracranial hypertension had increased CBF, i.e., 85 (23-134) mL/(100 g × min), whereas the patients without intracranial hypertension had a reduced CBF, i.e., 45 (23-56) mL/(100 g × min), corresponding to other studies of patients with FHF without intracranial hypertension (Table 2). Thus, there seems to be evidence to support that CBF is low unless intracranial hypertension has evolved in patients with FHF. Whether this increase in CBF develops gradually or immediately before surges of intracranial hypertension in the clinical setting is not possible to unravel from the present studies in this thesis (I-VII) but studies of rats have shown that CBF gradually increases during the course of FHF [67].

The pathophysiological reason for the decreased CBF found in this thesis may rely on a number of different mechanisms:

3.1.1 Hepatic encephalopathy and sedation

Increased neural activity increases the energy expenditure for ion-pumping and transmitter synthesis, resulting in increased energy production due to increased oxidative glucose consumption, which is supplied by an increase in CBF. Conversely, CBF decreases during states with decreased neuronal activity, such as during coma [68, 69] and sleep [70]. Patients with FHF in the present study were in deep coma; hence the finding of a reduced CBF is most probable due to the decreased cerebral metabolism and neuronal activity (VII). Additionally, administration of midazolam may also have contributed to the reduced CBF in the present studies. In healthy volunteers intravenous administration of midazolam (0.15 mg/kg) given as a single bolus over 15 sec, resulted in a reduction in CBF of ~30% 6 minutes after the injection [71]. After a bolus injection of diazepam, Sari et al [72] showed that CBF was maximally reduced after 10 minutes, and reached normal values after 90-120 minutes. The studied patients in this thesis were administered midazolam continuously (~0.125 mg/kg/h). Thus, the percent reduction in CBF due to administration of bolus injection of midazolam is not directly comparable to the data in this thesis. During craniotomy of patients with cerebral tumors where midazolam was administered continuously (0.125 mg/kg/h vs. 0.250 mg/kg/h) Knudsen et al found reduced CBF values with no relationship to the dose administered [73]. However, CBF was not measured prior to administration of midazolam, thus it is not possible to determine to what extent CBF was reduced from normal values.

3.1.2 Hyperammonia and glutamine

To avoid the deleterious effects of ammonia, humans detoxify ammonia by incorporating it into urea. Urea is generated in the liver by the urea cycle, when the liver is failing blood ammonia levels increase. At high levels ammonia is neurotoxic, and leads to functional disturbances of the central nervous system [74], but it is conflicting whether ammonia per se affects CBF. Some studies have demonstrated that acute ammonia infusion dilate cerebral vessels and increases CBF [10], while others found that CBF decrease [75]. In the present thesis, there was no relationship between CBF and arterial ammonia levels in patients with FHF ( Figure 2 ) (IV). Contrary, Jalan et al [65] found a positive correlation between CBF and arterial ammonia. This discrepancy may be due to time differences, as the patients in this thesis were investigated well before the development of cerebral edema and intracranial hypertension, while the patients in the study by Jalan et al [65] were investigated later during the cerebral illness of FHF.

Since the brain lacks urea cycle enzymes, ammonia removal from the brain relies on the formation of different amino acids, mostly glutamine and alanine, which are the main nitrogen carriers out of the brain (IV). A recent study shows that accumulation of glutamine per se only plays a limited role as a cause of cerebral edema in FHF, as mild hypothermia prevented cerebral edema in an animal model of FHF despite glutamine accumulation [76]. However, these findings do not preclude other important contributions of glutamine to the cerebral complications in FHF as hypothermia may have several other effects on the brain that may have contributed to the protective effect. In accordance with the experimental study by Master et al [77] we found that patients who subsequently died of intracranial hypertension had significantly higher cerebral ammonia uptake and cerebral glutamine efflux as compared to patients who survived (IV), suggesting that ammonia and glutamine plays an important role for the subsequent surges of intracranial hypertension. Whether or not this effect is cytotoxic, or is a combination of a cytotoxic and vasogenic effect is not possible to settle from the present study (IV).

Although the exact mechanism of ammonia toxicity is unresolved, hyperammonemia and cerebral glutamine accumulation appears to have several other effects on brain and cerebral metabolism that may contribute to or aggravate cerebral edema formation and induce cerebral vasodilation, including effects on cerebral energy metabolism [78], lactate/pyruvate production [79], astrocytic glutamate transport [80], brain ATP depletion by activation of NMDA receptors [81], nitrosactive/oxidative stress and induction of the mitochondrial permeability transition in cultured astrocytes [82]. The present clinical studies do not allow for any conclusions on the cellular effects of hyperammonemia and cerebral glutamine accumulation, nor do they allow for conclusion on their effects upon CBF later during the disease course of FHF. However, in the early stages of FHF hyperammonemia and glutamine accumulation did not affect CBF and cerebral oxidative metabolism (VII).

3.1.3 Acetaminophen and CBF

Recent studies have revealed evidence that acetaminophen inhibits prostaglandin E 2 production in rat cerebral endothelial cells possibly by acting against cyclooxygenase-2 [83]. Accordingly, inhibition of prostaglandin E2 production could also to some extent have influenced CBF in patients with FHF, as prostaglandin E2 is a vasodilator and acetaminophen intoxication was the reason for FHF in most of the patients (I-VII). Notwithstanding, CBF in patients without acetaminophen intoxication was similar to patients with acetaminophen intoxication, i.e., 38 (28-55) vs. 40 (28-54) mL/(100 g × min) (NS). Patients with acetaminophen intoxication appears to have a better outcome in larger series of FHF, and it is possible that this inhibitory effect upon cyclooxygenase-2 may play a role in this setting by inhibiting the gradual increase in CBF that seems to evolve during the disease course.

In conclusion CBF is reduced within the first 24 h after development of stage III-IV hepatic encephalopathy. Increase in CBF seems to be a phenomenon that takes place later during the disease course, and only evolve in patients who subsequently develop intracranial hypertension. The low CBF values found in the studied patients in this thesis can be explained by the presence of hepatic encephalopathy and sedation by midazolam.

3.2 THE EFFECT OF HYPERVENTILATION ON CBF
AND METABOLISM IN FHF

Cerebral CO 2 reactivity is the change in CBF per unit change in PaCO 2 , defined as the % change in CBF divided by the ΔPaCO 2 (in mmHg). At first the relationship between PaCO 2 and CBF was thought to be linear, however, later studies have shown that it is sigmoid, with a CO 2 reactivity that increases at high PaCO 2 levels and decreases at low PaCO 2 levels.

3.2.1 Global cerebral CO 2 reactivity

In this thesis global CO 2 reactivity was found normal in patients with FHF compared to controls. All clinical studies of cerebral CO 2 reactivity to hypocapnia performed on patients with FHF are displayed in Table 3 (III) [12, 14, 58, 84, 85]. All these studies reported almost similar cerebral CO 2 reactivity, except for one study where the hypocapnic CO 2 reactivity appeared much higher [85]. One explanation for this apparent discrepancy could be that by Sari et al [85], contained pooled data of patients with hepatic encephalopathy and septic encephalopathy. Thus, that study was not completely comparable with the other studies, which only contained patients with FHF. As can be seen from Table 3, values of CO 2 reactivity varied widely among patients with FHF. In two of the studies [12, 58] a paradox increase in CBF to hypocapnia, i.e., a negative CO 2 reactivity, was found in one patient (Table 3). Neither mean arterial blood pressure nor intracranial pressure was measured in these studies. Alteration of these pressures during the study period may have accounted for the apparent increase in CBF to hypocapnia. That is, if intracranial pressure was high before institution of hypocapnia, and subsequently was reduced during hypocapnia, then the resultant cerebral perfusion pressure is increased, and thereby also CBF. Likewise, if mean arterial pressure drops significantly during hyperventilation, then the resultant cerebral perfusion pressure is reduced and thereby CBF. Methodological problems should also be considered as well as time difference, as it cannot be excluded that cerebral CO 2 reactivity is completely lost later during the course of FHF [84].

The cerebral CO 2 reactivity is influenced both by the oxygen status and by the mean arterial blood pressure, as both hypoxia and hypotension induce vasodilation [86]. Thus, vasodilation induced by either hypoxia or hypotension may blunt the cerebral CO 2 reactivity during hypercapnia. In 1996, Larsen et al [84] explored the cerebral CO 2 reactivity in a prospective study including both patients with FHF and rats with thioacetamide-induced liver failure. It was found that patients with FHF had a reduced cerebral CO 2 reactivity during hypercapnia as compared to healthy subjects, ~2.2 vs. ~4.6% mmHg -1 , while it was normal during hypocapnia (Table 3) [84]. This finding was in accordance with a retrospective study of patients with FHF published by Durham et al a year before [58]. Accordingly, Larsen et al suggested that the cerebral CO 2 reactivity curve is left-shifted in FHF, i.e., CO 2 reactivity decreases during hypercapnia, while it is relatively preserved during hypocapnia ( Figure 3 ) [84].

Animal studies have reported that cerebrovascular reactivity to hypercapnia is blunted following acute elevation of blood ammonia levels [87-89]. Thus, it could speculated that the increased blood ammonia levels could be responsible for the blunted CO 2 reactivity in patients with FHF, as arterial ammonia levels are significantly elevated (IV) [90].

In the brain, ammonia is detoxified by the formation of glutamine from glutamate. A recent study by Okada et al [91], have demonstrated that glutamine exerts a modulatory effect on the cerebrovascular reactivity to CO 2 . They showed that a threefold increase in plasma glutamine concentration induced by infusion of glutamine blunted the cerebrovascular CO 2 reactivity in rats. Furthermore, they showed that co-infusion with arginine completely counteracted the effect of glutamine upon cerebrovascular reactivity, i.e., restored CO 2 reactivity. From these studies it is suggested that glutamine inhibits the recycling of citruline to arginine. Consequently the availability of arginine for nitric oxide synthesis is reduced [91]. In patients with FHF arterial (IV) and brain glutamine content is increased [92], thus glutamine could be responsible for the blunted CO 2 reactivity in FHF during hypercapnia.

Hypothermia appears to restore the hypercapnic cerebral CO 2 reactivity in patients with uncontrolled intracranial hypertension [93]. This may be explained by the vasoconstriction induced by hypothermia, which shifts the autoregulation curve to the left, thereby restoring at least some of the vasodilation capacity of the cerebral vessels.

In patients with FHF, we detected high circulating levels of two potent vasodilators, calcitonin gene-related peptide and vasoactive intestinal peptide (V). These neuropeptides exerts their effects on vascular smooth muscle receptors localised on the abluminal side of the blood-brain barrier and they do not seem to pass the blood-brain barrier to a greater extent. Thus, the circulating neuropeptides probably do not influence cerebrovascular tone to a major extent, unless the blood-brain barrier is disrupted [94, 95]. As they are stored in perivascular nerves, it cannot be excluded that high brain levels of one or both these two vasodilators play a role for the blunted cerebral CO 2 reactivity in FHF. Since, the effect of calcitonin gene-related peptide on cerebral vessels is mediated through opening of potassium channels, especially the calcium dependent but also to a minor degree the ATP dependent potassium channels [96], it could be speculated that the blunted CO 2 reactivity to hypercapnia in FHF is due to already opened ATP dependent potassium channels produced partly by high brain levels of calcitonin gene-related peptide.

3.2.2 Regional CO 2 reactivity

Even though global CO 2 reactivity is normal in patients with FHF there may be alterations of regional CBF and CO 2 reactivity, which may turn imminent regional cerebral ischemia into manifest ischemia. Furthermore, alterations of PaCO 2 , may result in redistribution of blood flow from regions with a relatively high tissue pressure and low CO 2 reactivity to regions with a high CO 2 reactivity and relatively low tissue pressure, referred to as a Steal phenomenon. On the other hand hypocapnia may redistribute blood flow from regions with low tissue pressure and high CO 2 reactivity to regions with high tissue pressure and relatively low CO 2 reactivity, the so-called inverse steal phenomenon [97]. In a study addressing the regional CBF distribution pattern prior to and during hyperventilation in patients with FHF, we found neither steal nor inverse steal phenomena in patients with FHF (II). We also found preserved and similar CO 2 reactivity to hypocapnia in all brain regions ( Figure 4 ) (II), which is in accordance with a retrospective study by Durham et al [58]. Thus, global as well as regional cerebral CO 2 reactivity to hypocapnia is preserved in patients with FHF.

3.2.3 Cerebral autoregulation and hyperventilation.

Cerebral autoregulation is impaired in patients with FHF [56], and is re-established shortly after recovery of liver function [98]. The reason for the impaired cerebral autoregulation in FHF is not completely settled, but has been suggested to result from gradual cerebral vasodilation [11]. This vasodilation hypothesis was supported by the observation where moderate hypocapnia (PaCO 2 ~3.0 kPa) restored cerebral autoregulation in five of seven patients with FHF ( Figure 5 ) (I). On the other hand global CBF was found to be decreased compared to normal values in the subsequent studies (III-VII). This finding does not support the vasodilation theory, as responsible for impairment of autoregulation (III-VII). If vasodilation does not account for the impaired cerebral autoregulation in patients with FHF what could then be the explanation of the restoration of cerebral autoregulation following hyperventilation? Although the nature of the association between hypocapnia and cerebral autoregulation was not investigated in the present work, a number of possible explanations may be considered.

If patients prior to intubation had been hyperventilating spontaneously for some time, then the institution of mechanical normoventilation would render the patient relatively hypercapnic if mechanical ventilation was instituted at a level of PaCO 2 that was higher than the spontaneous PaCO 2 . Hypercapnia shortens the autoregulatory plateau, i.e., increases the lower limit and decreases the upper limit of autoregulation, whereas hypocapnia widens the autoregulatory plateau ( Figure 6 ). Hence, in a state of relative hypercapnia cerebral autoregulation will be abolished, while it would subsequently be restored by re-institution of hyperventilation. It cannot be excluded that relative hypercapnia was of importance for the impaired cerebral autoregulation, but in other conditions with impaired cerebral autoregulation where spontaneous hyperventilation is not a prominent feature, e.g. neurotrauma, autoregulation is also restored by mechanical hyperventilation. Another explanation could be that the cerebral autoregulation curve is right-shifted, i.e., lower limit reached at a higher cerebral perfusion pressure, corresponding to a functionally impaired cerebral autoregulation. Thus, cerebral autoregulation is impaired in the "physiological" range of mean arterial blood pressures, and since hypocapnia widens the plateau of the cerebral autoregulation curve it may be restored during hyperventilation. Indeed, this could be an explanation as rats subjected to portacaval anastomosis demonstrated a right-shift of the autoregulation curve that was not affected by ammonia infusion [99]. In the study by Dethloff et al [99], the lower limit of autoregulation was ~25% right-shifted. In healthy subjects lower limit of autoregulation is ~60 mmHg, corresponding to a lower limit of ~75 mmHg if the lower limit of autoregulation is right-shifted to the same extent as in rats with portacaval anastomosis. Mean arterial blood pressure was > 75 mmHg in 6 of 7 patients with FHF (I), thus a right-shift of the autoregulation curve does not appear to be the only explanation, although it cannot be excluded that the lower limit of autoregulation is present at even higher blood pressure levels in FHF.

To conclude , loss of cerebral autoregulation and recovery during hyperventilation have been described in other diseases affecting the brain, such as acute bacterial meningitis [100-102], and acute head injury [103]. It is likely that a common denominator, which is influenced by alkalosis is responsible. Whether this common denominator is cerebrovascular tone, metabolic changes, low-grade cerebral edema, increased sympathetic tone, a right-shifted autoregulation curve or other yet undetermined factors remains to be explored.

3.2.4 CBF and hyperventilation

The major concern of hyperventilation is that it may reduce CBF to critically low levels resulting in decreased oxygen delivery capacity. According to the data obtained in previous studies of FHF a change in PaCO 2 of ~1 kPa results in ~20-30% decrease in CBF (Table 3). Since the overall mean CBF reported in patients with FHF is 33-40 mL (100 g min) -1 (IV) [12, 14, 58, 59, 104], this moderate decrease in PaCO 2 will reduce CBF to values ~25-32 mL (100 g min) -1 , which is well above the ischemic threshold, i.e., CBF > 18-20 mL (100 g min) -1 reported in patients with head injury [105-107]. Furthermore, a recent study in patients with traumatic brain injury using positron emission tomography, demonstrated that although short-term hyperventilation produced large reductions in CBF, cerebral metabolism of oxygen (CMRO 2 ) remained unchanged due to the low metabolic demands and compensatory increases in oxygen extraction fraction [108]. Thus, the reduction in global CBF observed during hyperventilation in FHF is unlikely to have caused ischemia.

3.2.5 Regional CBF (rCBF) and hyperventilation

In a high resolution image study addressing regional CBF in patients with FHF (II), we found lower cerebral perfusion in frontal regions and basal ganglia as compared to the other brain regions and healthy subjects, respectively. This altered regional CBF distribution pattern is in accordance with another study of patients with FHF [58]. The regional CBF distribution pattern was not altered by moderate hyperventilation, i.e., PaCO 2 decreased from 4.8 (range, 3.8 to 5.5) kPa to 3.7 (range, 3.2 to 4.4) kPa (II). Thus, CBF decreased to the same extent in frontal brain regions and basal ganglia as compared to other brain regions (Figure 4), which may expose these regions to hypoperfusion during prolonged and pronounced hyperventilation. However, it is not possible from our study (II) to conclude that frontal hypoperfusion is present in FHF and that this is aggravated by moderate hyperventilation, as we did not have any measures of regional cerebral oxidative metabolism.

The presented data in this thesis have demonstrated that the frontal areas and the basal ganglia have relatively lower rCBF than the posterior regions. This distribution pattern has also been demonstrated in other conditions such as acute bacterial meningitis [109] suggesting that it is either the unconsciousness that results in this altered distribution pattern or the sedatives administered. Against administration of sedatives as the reason for the altered rCBF distribution pattern in patients with FHF speaks that the same rCBF distribution pattern has been found in patients with subclinical hepatic encephalopathy [110], where sedatives were not administered.

3.2.6 Cerebral oxidative metabolism and hyperventilation

In accordance with other studies on patients with FHF cerebral oxygen metabolism was reduced in patients with FHF investigated in this thesis ( Table 4 ) (VII) [12, 14, 61, 63]. By contrast, the results on cerebral glucose metabolism in patients with FHF have ranged from reduced [14, 61] to normal [14], and even increased cerebral glucose metabolism [63]. The study with increased cerebral glucose metabolism [63] was contrary to the other studies performed on patients with ongoing intracranial hypertension. It has been suggested that ammonia may interfere with cerebral energy metabolism, including stimulation of certain glycolytic enzymes, inhibition of certain enzymes in the TCA cycle, and induction of the mitochondrial permeability transition in astrocytes leading to energy failure [78, 111]. The results on cerebral glucose metabolism in patients with FHF investigated in this thesis (VII) do not corroborate this hypothesis, as all glucose taken up by the brain was aerobically metabolized, i.e., the oxygen to glucose index was normal. It cannot be excluded that these metabolic changes takes place later during the disease course. Alterations in cerebral metabolism later during the disease course may be supported by the study by Jalan et al [63] which was performed during ongoing intracranial hypertension in patients with FHF, i.e., the oxygen to glucose index calculated from the reported cerebral metabolic rates is ~ 0.5 (Table 4). Notwithstanding there may be either a miscalculation in their reported metabolic rates or a misprint of the reported arterial and venous oxygen contents in their articles, i.e., avDO 2 is 3.1 mL/dL, and CBF is 103 mL/(100 g min) corresponding to a cmrO 2 on ~ 3.2 mL/(100 g min) [63]. If this recalculation of cmrO 2 is correct the oxygen to glucose index is ~4.2 in the patients studied by Jalan et al [63] which is indeed lower than required for aerobic glycolysis.

Only one other study has evaluated the effect of shortterm hyperventilation on cerebral oxidative metabolism in FHF ( Table 5 ) (VII) [14]. Wendon et al [14], found a slight cerebral lactate efflux in patients with FHF and suggested that the reduction in CMRO 2 was inappropriate for cerebral metabolic requirements, and therefore a further reduction in CBF by hyperventilation was anticipated to be harmful for patients with FHF. In that study Wendon et al [14], reported a slight increase in cerebral lactate efflux during short-term mechanical hyperventilation in patients with FHF, i.e., the lactate-glucose index became more negative from -0.05 to - 0.5, indicating that the amount of cerebral glucose uptake, that was accounted for by lactate excretion, increased from 2.5% to 25%. In normal resting subjects, a slight cerebral efflux of lactate is normally present, and this cerebral lactate efflux increases during hyperventilation (Table 5) [102]. When indices of cerebral oxidative metabolism are calculated from the reported median values in the study by Wendon et al [14], it appears that anaerobic metabolism was present during normoventilation, i.e., the oxygen-glucose index was ª 1.6, while hyperventilation brought about aerobic metabolism, i.e., the oxygen-glucose index increased to ª 9.5 [14]. Furthermore, studies using brain microdialysis in patients with FHF [112], has demonstrated that the cerebral lactate concentration remains unchanged during moderate short-term hyperventilation. In our study (VII), cerebral oxidative metabolism was not compromised neither during normoventilation nor during hyperventilation as a striking net cerebral lactate uptake was demonstrated during normoventilation, which was only slightly decreased by hyperventilation, i.e., the lactate-oxygen index remained positive, and the cerebral oxygen to glucose index remained unchanged and normal (VII). Contrary to our findings, Wendon et al [14] found that hyperventilation produced a significant reduction both in CMRO 2 and cerebral glucose metabolism in patients with FHF. As can be seen from Table 5 both the CMRO 2 and cerebral glucose metabolism varied widely among patients with FHF in the study by Wendon et al. Since they also found negative values of cerebral glucose metabolism (Table 5) it is likely that not all of their patients were in a steady state condition at the time of their measurements that normally is required for accurate sampling of CBF and AV differences of various metabolites.

Measurements of markers of astroglial (S-100b) and neuronal (NSE) damage during hyperventilation in patients with FHF, may support that moderate hyperventilation is safe, as the arterial concentrations as well as net brain flux of these markers were unaltered (VI). A borderline ischemia that is turned into manifest ischemia will result in astroglial and neuronal cell death; thus, a cerebral efflux of these markers would have been expected to arise if hyperventilation induced severe ischemia. However, it cannot be excluded that efflux of biomarkers requires a longer time to arise. Thus, from the current studies performed in patients with FHF there is no convincing evidence that short-term hyperventilation, corresponding to a PaCO 2 of ~25-30 mmHg, is detrimental to global CBF and oxidative metabolism.

3.2.7 Cerebral nitrogen balance and hyperventilation

Critically ill patients are most often in a state of negative whole body nitrogen balance (IV) [113, 114] which is due in part to muscle protein wasting. In patients with FHF (IV) also the brain has a negative nitrogen balance that is primarily caused by a manifest cerebral glutamine efflux (IV). No other comparable human studies have been published on cerebral nitrogen balance in FHF or other conditions. Thus, it is not possible to conclude that this is a specific feature of FHF, but there is evidence to support that hyperammonemia causes cerebral protein breakdown, as experimental studies have shown that ammonia results in a reduced cerebral protein content [115], and loss of glial fibrillary acidic protein has been demonstrated in humans with hepatic encephalopathy [116], as well as in experimental liver failure [117].

Short-term mechanical hyperventilation significantly reduced the cerebral nitrogen release, i.e., it became less negative (from -14.81 ± 14.02 to - 4.69 ± 6.26 μ mol (100 g min) -1 ) ( Figure 7 ) (IV). The ameliorated cerebral nitrogen balance was primarily caused by a reduction of the cerebral glutamine efflux, i.e., from -6.11 ± 5.19 μ mol (100 g min) -1 during normoventilation to - 2.91 ± 3.22 μ mol (100 g min) -1 during hyperventilation ( Figure 8 ) (IV). However, a decreased cerebral proline, alanine and tyrosine efflux, and increased cerebral uptake of branched chain amino acids, also contributed slightly to normalisation of the cerebral nitrogen balance during hyperventilation (IV). In that study, (IV) pH increased from 7.46 ± 0.05 to 7.54 ± 0.05 during hyperventilation. Since the pKa for glutamine is ~2.17 for the carboxyl group and ~9.28 for the amino group, respectively, this slight increase in pH cannot have influenced the amount of glutamine available for transport across the blood-brain barrier to a major extent, as only a small amount of glutamine will become ionised at that pH. Increased flux into the brain and/or decreased flux out of the brain could also account for the apparent reduction in glutamine efflux during hyperventilation. Recent studies have shown that active transport from brain to blood through sodium-dependent transport carrier systems is pH dependent, i.e., "flux" decreases with decreasing pH [118, 119]. Some of these transporter systems are also located on astrocytes and neurons. The astrocytic form being pH sensitive [120], while the neuronal form is not [121]. If the effect of pH on these carrier transport systems were linear, an increased transport out of brain as well as an increased astrocytic glutamine uptake would be expected to arise during alkalosis, with a resulting decrease of extracellular glutamine. We cannot exclude that pH may have influenced our results to some extent. However, alkalosis have been shown to stimulate muscle protein synthesis in critically ill patients [114], and in the perfused working heart [122]. During normoventilation we found evidence of cerebral protein degradation, as there was a net brain efflux of tyrosine and threonine, two essential amino acids. During hyperventilation this flux became zero, indicating that alkalosis may also inhibit protein degradation. A firmer conclusion on whether alkalosis stimulates protein synthesis or inhibits protein breakdown or a combination of both is not possible to determine from the data presented in this thesis.