Acute liver failure (ALF) is a dramatic clinical syndrome that results from massive hepatic necrosis. The mortality of ALF remains high despite progress in the fields of intensive care management and liver transplantation [1]. ALF refers to the rapid development of severe acute liver injury with impaired synthetic function and encephalopathy in subjects with previously normal, or well-compensated liver disease [2]. Generation of a suitable animal model would allow for a greater understanding of the underlying pathophysiology of this condition and stimulate the development and refinement of treatments for ALF. There is a need for a suitable large animal model of hepatic injury to facilitate the development of bio-artificial liver support systems in humans and to assist pre-clinical assessment [3].

Since the recognition of acetaminophen-induced ALF in the 1960s, acetaminophen overdose has become the commonest cause of liver failure in the United Kingdom [4]. Although an effective antidote (N-acetylcysteine) has been available for many years, the late presentation of many cases ensures a significant number develop liver failure. Early large animal models of acetaminophen toxicity were unreliable, producing inconsistent toxicity from animal to animal. Gazzard et al [5] showed that dogs receiving oral doses greater than 1 g/kg died of nonhepatic causes, becoming cyanotic immediately before death with methylene blue-resistant methaemoglobinaemia. Human studies have shown that serum acetaminophen levels must be >140 mg/mL for greater than 20 hours for consistent hepatic necrosis to occur [6]. Using this, Francavilla et al [7] studied various doses and dosing schedules on 52 beagle dogs. These initially received 750 mg/kg acetaminophen subcutaneously followed by 200 mg/kg nine hours later, and then a further 200 mg/kg twenty-four hours later. Mortality rates in this group were: 0% at 24 hours, 10% at 48 hours, and 90% at 72 hours. Serum transaminase levels were greatly increased to 8206 +/- 3000 U/L at 48 hours and 21 253 +/- 3746 U/L at 72 hours. This correlated with liver biopsy samples that revealed severe centrilobular necrosis.

Several approaches have been adopted to control methaemoglobinaemia in animal models: Kelly et al [8] using dogs, suspected high plasma acetaminophen levels were responsible for the lethal methaemoglobinaemia that characterised their early experiments. They controlled acetaminophen infusion in an attempt to maintain its levels at 175 to 200 mg/mL for the first 20 hours. Methylene blue was administered at 10 mg/kg if blood samples appeared brown on visual inspection. Pilot data showed that levels of acetaminophen greater than 300 mg/mL were associated with severe methemoglobinemia.

Cellular glutathione (GSH) levels must be depleted to approximately 20% of their original levels before toxic acetaminophen damage can occur, [9, 10]. To reduce variation in the acetaminophen dose required and the time taken for the development of signs, Kelly et al [8] attempted to deplete GSH levels by administering 2 mmol/kg of buthionine sulfoximine (BSO) 2 hours before acetaminophen. BSO [11], an analogue for glutathione synthetase, binds and irreversibly inactivates this enzyme which is the final enzyme in GSH synthesis. The rapid turnover of cellular GSH, ensures that enzyme inactivation leads to a rapid reduction in the hepatocyte GSH concentrations. Kelly et al [8] showed that BSO administration to mice decreased hepatic GSH stores to 15% to 20% of normal within 2 hours. They showed greater liver damage in the BSO group in terms of alanine aminotransferase activity, prothrombin time, bilirubin level, and histological characteristics. Miller et al [12] administered phenobarbital to induce cytochrome P-450 enzymes and thus potentiate subsequent acetaminophen toxicity. This increased liver damage and decreased animal survival.

There has been limited success in the development of suitable large animal models of acetaminophen induced liver injury. In the study described here we used bed-side biochemical and invasive haemodynamic monitoring during acetaminophen-induced liver damage in a reproducible and controlled manner. Using this approach we markedly reduced previously reported problems, i.e., methaemoglobinaemia and cardiovascular collapse, and produced a large animal model which reliably reproduces the syndrome of ALF observed in clinical practice.

Animal experiments were performed in accordance with the Home Office regulations under the Animal (Scientific Procedures) Act 1986 as per Project licence 60/2389.

All animals received humane care and study protocols complied with our institution's guidelines.

13 Large White pigs (median body mass 35 kg) were maintained under. Anaesthesia was induced with Ketamine and Midazolam, and subsequently maintained with isoflurane and nitrous oxide according to tidal volume. Background hydration was maintained at a rate of 2 mls/kg/hr using a combination of 0.9% Normal Saline and 5% Dextrose according to electrolyte results from arterial blood gas sampling and urine output. Boluses of colloid (Gelofusine) were administered for episodes of hypotension. Heater pads were used to ensure that core (rectal) temperature was maintained between 37-39°C (the normal temperature range for pigs) [13]. Haemodynamic variables were monitored continuously using a Datex AS/3 monitoring system (Datex Ohmeda, Stirling, UK) and recorded at one minute intervals using the Datex Collect programme on a laptop computer. Measurements were made from arterial, central venous and pulmonary arterial cannulae. The latter allowed hourly left atrial pressure (from pulmonary capillary wedge pressure) and cardiac output (thermodilution) measurements: 10 mls isotonic saline aliquots at 4°C were injected through the injectate port of the pulmonary artery catheter using a closed injectate system (CO-Set plus, Baxter Edwards Critical Care, Irvine, CA, USA). Three measurements were made at random points during the respiratory cycle and an average value taken for each measurement. Intracranial pressure (ICP) was monitored with a subdural pressure transducer (Camino 110-4B, Integra Lifesciences Corp, New Jersey, USA). Correct positioning was confirmed by obtaining an appropriate waveform that fluctuated with ventilation. Urine output was monitored via a urethrostomy. Microbiological testing was performed routinely for all animals.

Cytochrome p450 levels measured in homogenised liver tissue from phenobarbital pre-treated animals (n = 3 pigs)were significantly higher than in animals which had not received phenobarbital (300 +/- 86 vs 100 +/- 31, p = 0.01) pmol/mg protein. All animals in this study subsequently received phenobarbital

Acetaminophen toxicity and methaemoglobinaemia

Acetaminophen was administered by intravenous infusion for 12 hours with the intention of maintaining blood levels between 200-300 mg/l. By accurate monitoring of acetaminophen and methaemoglobin levels this was largely accomplished (figure 1a) although in two (of nine) animals levels exceeded 400 mg/l. The consequence of high acetaminophen levels, namely an increasing methaemoglobin percentage, is demonstrated in figure 1b. In injury 7 we were unable to measure acetaminophen levels, and despite following a previously calculated infusion pattern the pig developed severe methaemoglobinaemia (peak of 13%). This reinforces the importance of close monitoring of acetaminophen levels. In general, methaemoglobinaemia (>4%) occurred approximately 2-3 hours after acetaminophen levels exceeded 350 mg/l. The finding of other groups, i.e., that brown blood discolouration presages the development of methaemoglobinaemia was not confirmed in the current study. Similarly the use of methylene blue (MB) which has previously been described as an effective treatment for methaemoglobinaemia, was associated with death within 5 minutes of injection in the single animal of the current study in which it was administered.

Animal survival

Control pigs (n = 4) survived the 28 hour anaesthetic without complications, whilst only two of the nine injured pigs survived to this time-point (see Figure 2a). Two of the animals which received acetaminophen were critically unwell, and were euthanased at 25 hours. Three pigs developed a syndrome of multi-organ failure (MOF) before death, which is in keeping with the clinical condition [4]. Failure of two or more organ systems was required for MOF to be confirmed: most commonly these were arterial hypotension (refractory to fluid resuscitation) and ventilatory failure. Sepsis was confirmed in two pigs by blood culture (Klebsiella pneumoniae and Pseudomonas paucimobilis). Refractory hypotension and/or difficulties in ensuring successful ventilation led to a joint decision to euthanase pigs.

Cardio respiratory evaluation

Figures 3a and 3b demonstrate that injured pigs (as opposed to uninjured pigs) developed the typical haemodynamic pattern of acute liver injury characterised by arterial hypotension (40.8 ± 5.9 vs 59 ± 2.0 mmHg, p = 0.16) (MAP, mean arterial pressure), increased cardiac output (CO) (7.26 ± 1.86 vs 3.30 ± 0.40 l/min, p = 0.01) and decreased systemic vascular resistance (SVR) (8.48 ± 2.75 vs 16.2 ± 1.76 mPa·s/m³, p = 0.03) that was unresponsive to fluid infused in an attempt to maintain normal pulmonary arterial occlusion pressures (PAOP) (figure 4). Positive pressure lung ventilation was associated with a reduction in total compliance as injury progressed and the pulmonary vascular resistance (PVR) (636 ± 95 vs 301 ± 26.9 mPa·s/m³, p < 0.01) increased in this group (figure 4). This may have indicated the onset of an adult respiratory distress syndrome (ARDS)-like picture.

Laboratory analysis

Biochemical analysis revealed a marked increase in serum AST (471.5 ± 210 vs 42 ± 8.14, p = 0.03) in conjunction with a marked reduction in serum albumin (11.5 ± 1.71 vs 25 ± 1 g/dL, p = 0.03) in injured animals (figure 5a). AST rather than ALT was used to quantify liver damage as the assay employed was unable to measure porcine ALT from homogenised porcine liver tissue. This hepatic dysfunction was confirmed by pH (7.23 ± 0.05 vs 7.45 ± 0.02, p = 0.02) and prothrombin time (36 ± 2 vs 8.87 ± 0.33 seconds, p = 0.002) (figure 5b). Arterial pH decreased significantly in the injured group in parallel with increasing prothrombin times (to a mean of 36 seconds at 24 hours). The liver components of prothrombin time, factors V and VII were (figure 6a) to decrease after injury to levels of 9.3% and 15.5% of normal values respectively. Factor VIII levels remained unchanged indicating that Disseminated intravascular coagulation (DIC) was not the main cause for the altered levels of other coagulation factors.

In addition to liver injury there was evidence of renal impairment as shown in figure 7 where serum creatinine rises above 250 μmol/l in injured animals, although serum potassium remains below 6 mmol/l. Mean levels of creatinine and potassium were 280.2 ± 36.5 vs 131.6 ± 9.33 μmol/l (p = 0.002) and 5.65 ± 0.53 vs 4.6 ± 0.12 mmol/L (p = 0.27) respectively.

Evidence of encephalopathy: Fischer's ratio

In the liver injury group there was a significant increase in tyrosine (0.03 ± 0.004 vs 0.40 ± 0.02) (p < 0.008) and phenylalanine concentrations (0.05 ± 0.004 vs 0.71 ± 0.08) (p < 0.001) by the end of the experiment. There were no statistically significant changes in leucine levels (0.17 ± 0.02 vs 0.175 ± 0.02) although isoleucine increased significantly (0.08 ± 0.01 vs 0.2 ± 0.01) (p < 0.05) while valine levels decreased (0.2 ± 0.02 vs 0.13 ± 0.01) (p < 0.01). Fischer's ratio (BCAA/AAA) was 5.6 ± 1.1 at the beginning and 0.45 ± 0.06 at the end of the study (see Figure 8). The ratio was unchanged in uninjured pigs.

Histology

Histological analysis of liver tissue revealed perivenular and mid-zone confluent hepatocytic necrosis in the worst-affected animals, with microvesicular change in preserved viable parenchyma. Other animals developed less injury as judged by the extent of necrosis, but qualitatively similar patterns (see Figure 9) were seen. Four animals demonstrated moderately severe liver injury, 2 developed severe coagulative necrosis and 3 developed mild (no zonal confluent necrosis) liver injury. Similarly, renal histopathology of seven pigs revealed severe acute tubular injury as judged by the prevalence of vacuolar change to the cortical tubular epithelium.

The major role for an ALF model is to provide a more controlled experimental environment than currently exists [18]. An ideal model should satisfy the following criteria: 1) induced hepatic failure should be reversible; 2) the features of liver damage should be reproducible; 3) selective damage should lead to fatal liver failure over a period similar to that encountered in the human condition; but 4) with sufficient time to allow the option for successful treatment; 5) the model should be large enough to accommodate anthropocentric therapies; and 6) the toxins used should not be hazardous to laboratory personnel. A pig model would be preferable as it has the potential to fulfil all of these criteria, and also importantly shares very similar liver metabolic function with human liver [19].

The porcine model described here develops most of the signs typifying acute liver injury/failure, namely: tachycardia, hypotension, increased cardiac output, decreased systemic vascular resistance, coagulopathy, renal failure and altered hepatic biochemistry. Consequently, it is the only animal model of acetaminophen-induced liver injury large enough to allow monitoring at a level equivalent to intensive care. Whilst the outcome of acetaminophen-injured pigs is relatively reproducible there still remain some inherent variations highlighting the difficulty of such studies. This reinforces the need to perform large studies when using such animals to ensure that individual variations can be accounted for.

Our model showed the typical increase in cardiac output and reciprocal reductions in SVR found in the clinical setting. Of note this is accelerated within the porcine model, starting within 16 hours of the onset of liver damage. This is likely a reflection of pre-treatment with phenobarbitone. The increased PVR in the context of normal PaO₂ values was in keeping with the increased requirement for positive end-expiration pressure (PEEP) during ventilation. This may have indicated the onset of an ARDS-like picture, although lung histology would be important to support this conclusion. That PAOP did not increase suggests that the ventilatory changes were not associated with pulmonary oedema secondary to left ventricular failure. A non-significant reduction in PVR Index after liver injury has been described [20]. The reason for this difference is unclear. Acute lung injury is common in patients with acetaminophen-induced fulminant hepatic failure and is associated with systemic circulatory failure, cerebral oedema and high mortality.

The marked reduction in Factor V and VII levels in the current model were consistent with severe liver injury. The Clichy criteria [21] deemed a reduction of Factor V levels to <20-30% as the trigger for liver transplantation; this was the case in most animals in this experiment. The absence of changes in Factor VIII levels indicate DIC was not the cause of altered levels of other coagulation factors.

An increase in AST levels indicated the development of liver necrosis, although these were not as high as are seen in patients. Similarly, a rise in bilirubin did not occur. This may reflect a species-specificity of the assay used, or insufficient time for bilirubin to rise. It is possible that there are subtle variations in this model compared to humans in that some systemic features prevail over a liver focused pattern of injury. Liver tissue analysis revealed severe coagulative necrosis in some animals with moderate and mild injury in others. This supports the probability that significant liver injury accounts for the clinical syndrome reported here. Although there is some variability in the liver injury seen in this model, there is still significant liver injury in the majority of animals. Indeed, this variation closely mimics what is seen clinically.

There are several explanations for the failure to observe changes in ICP. The time course may have been too brief. However, a raised ICP is not always encountered in the clinical setting, or only in association with other factors such as sepsis. It may also reflect differences in the pathophysiology of drug-induced ALF as opposed to ischaemic models of ALF. Measurements of arterial ammonia, cerebral blood flow or the use of cerebral microdialysis may help to clarify intracranial effects in future studies. Biochemical analysis suggested that animals may have been developing encephalopathy, although in the absence of ICP changes and with general anaesthesia it is difficult to be certain. Fischer's ratio [16], the ratio of the aromatic amino acid (AAA; tyrosine and phenylalanine) to branch chain amino acid (BCAA; leucine, isoleucine and valine) concentrations has a value of 3 to 4 in clinical liver failure but when this ratio is reversed and is in absolute figures <1.4 then most patients develop hepatic encephalopathy.

There was evidence of marked renal impairment, although fatal hyperkalaemia was not observed. Renal dysfunction is an important feature of acetaminophen induced liver failure [22], and the renal histology in treated animals was consistent with a toxic injury rather than the distinctive pattern of frank outer medullary tubular necrosis encountered with ischaemia, i.e., critical hypotension.