A reproducible, clinically relevant, intensively managed, pig model of acute liver failure for testing of therapies aimed to prolong survival

A clinically relevant, translational large animal model of acute liver failure (ALF) is required for testing of novel therapies to prolong survival in acute liver failure, to permit spontaneous liver recovery or to act as a bridge to transplantation.


Abstract
Background: A clinically relevant, translational large animal model of acute liver failure (ALF) is required for testing of novel therapies to prolong survival in acute liver failure, to permit spontaneous liver recovery or to act as a bridge to transplantation. Aims: The aim was to establish a pig model of acetaminophen-induced ALF that mimics the human clinical syndrome, is managed as in a human intensive care unit and has a predictable survival time. Methods: Nine female pigs were anaesthetised and instrumented for continuous intensive care monitoring and management using: target-driven protocols for treatment of cardiovascular collapse, metabolic acidosis and electrolyte abnormalities; intermittent positive pressure ventilation; and continuous renal replacement therapy. Six animals were induced to ALF with acetaminophen (paracetamol). Three animals acted as controls. Results: Irreversible acute liver failure, defined as rise in prothrombin time >3 times normal, occurred 19.3 ± 1.8 h after the onset of acetaminophen administration. Death occurred predictably 12.6 ± 2.7 h thereafter, with acute hepatocellular necrosis in all animals. Clinical progression of liver failure mimicked the human condition including development of coagulopathy, intracranial hypertension, hyperammonaemia, cardiovascular collapse, elevation in creatinine, metabolic acidosis and hyperlactataemia. In addition, cardiovascular monitoring clearly demonstrated progressive cardiac dysfunction in ALF. Conclusions: A reproducible, clinically relevant, intensively managed, large animal model of acute liver failure, with death as a result of multi-organ failure, has been successfully validated for translational studies of disease progression and therapies designed to prolong survival in man.
Acute liver failure (ALF) is the sudden loss of hepatic function, in the absence of pre-existing liver disease that results from severe liver injury associated with hepatocellular necrosis. Orthotopic liver transplantation (OLT) is the only treatment proven to prolong survival in patients with ALF, but OLT remains a limited resource (1,2). Search for alternative therapies to prolong survival in ALF requires a clinically relevant, translational large animal model of ALF, which is clinically relevant, mimics the human clinical syndrome, is managed as in an intensive care setting, has a predictable time course to death, but also has the potential for reversal of liver failure (3).
An acetaminophen (Paracetamol, APAP)-induced pig model of ALF would be clinically relevant, as APAP toxicity is the leading cause of ALF in the UK and USA. Work in the 70s and 80s dismissed such a model because of: variable survival times; fatal methaemoglobi-naemia; acute anaemia; and poor correlation among APAP dose, survival time and plasma APAP concentration (4,5). In 2010, Newsome and colleagues (6) achieved stable blood APAP concentrations without fatal methaemoglobinaemia, but survival times remained highly variable. In 2011, Thiel and colleagues (7) used jejunal administration of APAP and reported a reduced survival time variation following the onset of ALF of 21 ± 5 h.
The purpose of this study was to improve on the previously published porcine models of APAP-induced ALF. The specific aims for the current model were that it should: (i) demonstrate a clinical course that mimics the human condition; (ii) be managed as in a human intensive care unit using standardised protocols to permit consistency and reproducibility; and (iii) demonstrate a predictable survival time following the onset of ALF.

Materials and methods
Nine 30-40 kg, female, Landrace cross Large White pigs were used: ALF was induced in six (APAP Group) and three were used as controls (Control Group). All animal procedures complied with the Animals (Scientific Procedures) Act 1986. Detailed study protocols are included in the supplementary data.

Model set-up
Pigs were maintained under general anaesthesia, with intermittent positive pressure ventilation (IPPV) via orotracheal intubation, throughout this study. Omeprazole (40 mg intravenous) was given as prophylaxis for gastric ulceration and cefuroxime (20 mg/kg intravenous) as a peri-operative prophylactic antibiotic.
A fibreoptic Camino 110-4L catheter tip intracranial pressure (ICP) catheter (Integra neurosciences, Hampshire, UK) was placed into the left frontal cortex. Vascular catheters were inserted for central venous pressure (CVP), direct arterial blood pressure, pulmonary artery pressure and cardiac output (CO) monitoring; fluid and drug administration; and blood sampling. Venous double lumen haemodialysis catheters were placed for continuous renal replacement therapy (CRRT). A cystostomy tube was placed for urine output. An oroduodenal feeding tube was placed for APAP dosing.

Induction of acute liver failure
Liver failure was induced with APAP tablets, suspended in 30-ml water and administered via the oro-duodenal feeding tube, flushed through with 30-ml water. A loading dose of 0.25 g/kg APAP was given at time zero, followed by a maintenance APAP dose every hour until point of 'irreversible ALF'. 'Irreversible ALF' was defined as the point at which PT exceeded 60s. Maintenance APAP dose was initially 3 g or 3.5 g for pigs <38 kg or ! 38 kg respectively. Maintenance APAP dose was adjusted between 0.5 and 3.5 g with aim to maintain arterial methaemoglobin (metHb) concentrations at 1-5% of total haemoglobin. Maintenance APAP dose was increased to 4 g after 10 h, if metHb remained at baseline levels or lower and there was no increase in PT. APAP was discontinued after point of irreversible ALF.

Supportive care
Crystalloid solutions were given intravenously at 5 ml/ kg/h throughout: compound sodium lactate at the beginning of this study; 0.9% sodium chloride from 12 h after time zero; and 0.18% sodium chloride with 4% glucose if risk of hypernatraemia. Potassium chloride was added if plasma potassium concentrations fell below 3.5 mmol/l. In addition, 6% hydroxyethyl starch ('Voluven 6%', Fresenius Kabi Ltd, Cheshire, UK) was given intravenously starting at 1 ml/kg/h. Twenty per cent human albumin (Bio Products Laboratory, Hertfordshire, UK) was given intravenously for the first 12 h following time zero at 8 ml/h and thereafter at 20 ml/h. Parenteral nutrition containing 10% glucose and 5.9% amino acids ('Vamin 9 Glucose', Fresenius Kabi Ltd, Cheshire, UK) was given intravenously at 1 ml/kg/h from time zero. Two units of porcine fresh frozen plasma (FFP) were given intravenously, as a bolus, at the point of irreversible ALF.
Care aimed to maintain mean arterial blood pressure (MAP) at >70 mmHg for the first 12 h, >60 mmHg from 12 h to onset of irreversible ALF and >50 mmHg following onset of irreversible ALF. If MAP fell below these targets, three possible actions were taken. Firstly, a 5 ml/kg bolus of Voluven was given over 15 min. This was repeated up to three times as long as CVP remained <9 mmHg, followed by increase in maintenance infusion of Voluven by 1 ml/kg/h. Secondly, noradrenaline was initiated at 0.1 lg/kg/min intravenously and increased by 0.1 lg/kg/min every 15 min. Thirdly, terlipressin was initiated at 2 mg/24 h intravenously when noradrenaline exceeded 0.5 lg/kg/min and further increased to 4 mg/24 h when noradrenaline exceeded 1.0 lg/kg/min. 8.4% Sodium bicarbonate was given intravenously to correct metabolic acidosis. Ten per cent calcium gluconate was given intravenously to correct hypocalcaemia (calcium <1.1 mmol/l).

Continuous renal replacement therapy
Continuous venovenous haemofiltration was started at the onset of irreversible ALF using the PRISMA system, HF1000 set (polyarylethersulphone membrane) and 'PrimaSol BGK 2/0', as replacement solution (Gambro Dialysatoren GmbH, Rostock, Germany). Blood flow rate was 120 ml/min and dialysate flow rate was at 35 ml/kg/h with zero net fluid removal.

Study end and controls
The study end for the APAP group was the point of death. For control animals, protocols for induction to ALF were used for 20 h, then protocols for onset of irreversible ALF for 20 h, followed by sacrifice with intravenous pentobarbital. The only exception was that APAP was excluded from the water administered into the duodenum.

Data analysis and statistics
Calculations for cardiovascular variables are given in Table 1. Survival analysis from point of irreversible ALF to death for the APAP group and point of onset of irreversible ALF protocols to termination for the Control group was carried out using log rank analysis.
For each animal, continuous variables collected by the patient anaesthetic monitor and ventilator and drug and fluid infusion rates were recorded every 15 min. For each animal, mean values for every hour were calculated and used for the calculation of group descriptive statistics for the APAP group and the Control group. All data points for arterial blood gas data, PT times and biochemical analyses were used for the calculation of APAP and Control group descriptive statistics. Data for each group were summarised as mean ± SD.
Change in physiological variables with time was analysed within and between Control and APAP groups for each of the following three phases of the experiment: time from onset of APAP or placebo dosing to onset of irreversible ALF (APAP group) or onset of irreversible ALF protocols (Control group); time from onset of irreversible ALF (APAP group) or onset of irreversible ALF protocols (Control group) to 2 h prior to death; and 2 h prior to death. Linear mixed effects models were used for all analyses, and first degree auto-regressive (co)variance structure was used to account for the correlation between repeated measures. Significance was set at the 5% level. All analyses were carried out using SAS version 9.1 for PC (Copyright, SAS Institute Inc. Cary, NC, USA).

Induction of acute liver failure
In the APAP group, mean total dose of APAP administered to induce irreversible ALF was 59.6 ± 10.5 g (1.6 ± 0.2 g/kg body weight; Figs 1 and 2). Peak serum APAP concentrations of 367 ± 30 mg/L occurred at 12 h (range from 4 to 12 h; Fig. 1). Percentage arterial metHb began to increase at 7.6 ± 2.7 h to a peak of 5.0 ± 4.1%. Increase in PT from a median of 22.5 s (range from 21 to 25 s) at time zero to >60 s at 19.3 ± 1.8 h marked the point of irreversible ALF. Survival following the point of irreversible ALF was 12.6 ± 2.7 h with coefficient of variation of 0.226. Methaemoglobinaemia, APAP and coagulopathy were not detected in Controls, all of which survived to study end.

Intracranial pressure
Mean ICP at time zero in the APAP and Control groups was 17.0 ± 3.3 mmHg and 15.3 ± 4.5 mmHg, respectively, reflective of the face-down, dorsal recumbency position of the animal (Fig. 2). There was a gradual significant increase in ICP with time in both groups (P = 0.006). However, increase in ICP in the APAP group following the onset of irreversible ALF was significantly greater than that in the Control group (P = 0.008) with a sudden marked increase in the hour prior to death. At study end, mean ICP in the APAP and Control groups was 41.2 ± 8.6 mmHg and 22.7 ± 2.5 mmHg respectively. In the APAP group, mean CPP at time zero, onset of irreversible ALF and 2 h prior to death was 49 ± 7 mmHg, 37 ± 8 mmHg and 20 ± 8 mmHg, respectively, with a further dramatic fall just prior to death. Whereas, mean CPP in the Control group was relatively stable throughout this study: mean CPP at time zero, onset of ALF protocols and at termination was 51 ± 7 mmHg, 65 ± 15 mmHg and 62 ± 10 mmHg respectively (P < 0.001).

Haemodynamics
In the APAP group from 12 h after time zero to the point of irreversible ALF, gradual onset hypotension (P = 0.001) and tachycardia (P < 0.001) were noted with fall in mean MAP from 79 ± 8 to 53 ± 7 mmHg and increase in mean HR from 120 ± 13 to 197 ± 34 bpm, despite adherence to cardiovascular support protocols (Fig. 3). Tachycardia maintained mean cardiac index (CI) between 4.5 and 6.0 L/min/m 2 , despite fall in mean stroke volume index (SVI) from 47 ± 3 to 21 ± 5 ml/beat/m 2 (P < 0.001). Mean systemic vascular resistance index (SVRI) also fell from 1139 ± 72 to 913 ± 491 dynes.s/cm 5 /m 2 . Progressive hypotension resulted in an increasing requirement for intravenous colloid therapy and inotropes (P < 0.001). 0.5 lg/kg/min noradrenaline and 2 mg/24 h terlipressin were required by 16.5 h after the onset of APAP dosing, i.e. 3.0 ± 1.5 h before ALF. Prior to the onset of irreversible ALF, mean CVP and PCWP were maintained between 4 to 6 mmHg and 5 to 8 mmHg respectively.
In the APAP group, at the onset of irreversible ALF, rapid infusion of two units of fresh frozen plasma resulted in temporary increase in MAP and SVRI. Thereafter, mean MAP and mean heart rate were maintained between 51 to 61 mmHg and 144 to 199 bpm respectively. SVRI remained low. On-going fluid therapy was associated with progressive peripheral oedema and a significant rise in CVP (P < 0.001) and PCWP (P < 0.001), but CVP values remained less than the target maximum of 9 mmHg until 10 h after the onset of irreversible ALF. From 10 h following the onset of irreversible ALF to point of death, there was a rapid rise in CVP and PCWP with mean values in the last hour of life of 17 ± 3 mmHg and 20 ± 4 mmHg respectively. These terminal values were associated with bradycardia and fall in MAP that was unresponsive to inotropes. In the APAP group, left ventricular stroke work index (LVSWI) and right ventricular stroke work index (RVSWI) decreased significantly from the onset of APAP dosing to the point of irreversible ALF: LVSWI decreased from 43 ± 7 to 22 ± 14 g/m/m 2 (P < 0.001) and RVSWI decreased from 9 ± 3 to 4 ± 1 g/m/m 2 (P < 0.001). Thereafter, LVSWI remained between 14 ± 5 and 19 ± 7 g/m/m 2 and RVSWI between 3 ± 1 and 6 ± 0.5 g/m/m 2 .
In the Control group, sustained hypotension and tachycardia were not observed, resulting in significantly lower requirement for Voluven (P < 0.050) and noradrenaline (P < 0.010) throughout this study compared with the APAP group. Prior to the onset of ALF protocols, mean CI increased from 4.5 to 6.8 L/min/m 2 (P = 0.017) likely because of intravenous colloids and albumin therapy in the absence of requirement. After the onset of ALF protocols, an increase in MAP, CVP and CI associated with FFP and albumin infusion was associated with a fall in SVRI. In addition, LVSWI and RVSWI increased gradually from 47 ± 1 to 71 ± 15 g/ m/m 2 and 9 ± 1 to 14 ± 4 g/m/m 2 respectively (P < 0.001).
In the APAP Group, progressive haemoconcentration was noted with the rise in haemoglobin from 8.2 ± 1.1 to 11.5 ± 1.4 g/L at irreversible ALF. Following the onset of irreversible ALF, a gradual decrease in haemoglobin was noted with a mean haemoglobin concentration 2 h prior to death of 3.9 ± 1.2 g/L, which was significantly lower than controls (P = 0.017). In the Control group, mean haemoglobin concentrations ranged from 9.8 ± 1.4 g/L at time zero to 8.3 ± 9.7 g/L at study end.

Respiration and ventilation
Respiratory support was similar between the APAP and Control Groups with requirement for increasing airway pressures to achieve adequate ventilation. In the APAP group, mean arterial partial pressure of oxygen (PaO 2 ) was maintained above 100 mmHg, mean arterial partial pressure of carbon dioxide (PaCO 2 ) between 38 ± 3 and 52 ± 8 mmHg and inspired tidal volumes (TVinsp) above 6 ml/kg. Similar ventilation results were seen in the Control group, but duration of anaesthesia was on average 8 h longer than in the APAP group. During these final 8 h, further failure of ventilation resulted in dips in mean PaO 2 to <75 mmHg for 3 h and increases in mean PaCO 2 of >55 mmHg for 4 h.

Electrolyte and acid-base status
In the APAP group, progression towards irreversible ALF was associated with metabolic acidosis, with decreases in mean pH from 7.4 ± 0.1 to 7.3 ± 0.1 (P = 0.002), mean bicarbonate from 25 ± 2.3 to 21 ± 2.9 mmol/L (P < 0.001) and mean base excess from 0.6 ± 2.6 to À4.6 ± 3.9 mEq/L (P < 0.001). These changes occurred despite bicarbonate therapy, initiated 13.0 ± 5.3 h after the onset of APAP dosing. CRRT improved acid-base control. In the Control group, metabolic acidosis was not noted. However, respiratory acidosis associated with increase in PaCO 2 was noted approximately 9 h after the onset of ALF protocols.
In the APAP group, there was a gradual increase in lactate to 4.7 ± 2.1 mmol/L at the onset of irreversible ALF, with a further increase thereafter to 7.6 ± 3.4 mmol/L 10 h after the onset of irreversible ALF. This was significantly different compared with Controls, which had mean lactate concentrations of <1 mmol/L in the 20 h prior to termination (P < 0.001).

Liver and kidney biochemistry profiles and urine output
In the APAP group, progressive severe hypoalbuminaemia and elevations in serum creatinine occurred in the APAP group, compared with the Control group (P < 0.001; Table 2). Significant hyperammonaemia was observed shortly before death in the APAP group, but not in Controls (P = 0.010). In the APAP group, urine output was 1.7 ± 0.5 ml/kg/h and 1.2 ± 0.4 ml/kg/h before and after the onset of irreversible ALF, which was significantly lower than the Controls (P < 0.001), in which urine output was 2.4 ± 1.1 ml/kg/h and 3.7 ± 0.1 ml/kg/ h before and after the onset of ALF protocols.

Histopathology
At time of death, liver specimens from all APAP group animals demonstrated moderate (n = 4) to severe (n = 2) acute centrilobular and midzonal hepatocyte degeneration and necrosis with mild haemorrhage and neutrophilia (Fig. 4). This was not seen in any of the Control animals. Pulmonary pathology was present in all Control and APAP post-mortem lung specimens including diffuse oedema of interlobular septa; alveolar over-distension and emphysema; and mild thickening of alveolar walls with histiocytosis.

Discussion
This article describes a reproducible porcine model of APAP-induced acute liver failure that results in acute hepatocellular necrosis in all animals. This model dem- Table 2. Kidney and liver biochemistry profiles in APAP and Control groups before and after the point of irreversible acute liver failure (ALF). Significant p values for between-group comparisons for each time point are given in brackets. Death of animals in the APAP group after 12 hours after ALF, precluded analyses after this time point (−). Ammonia analyses were not performed at all time points (x). Final ammonia concentrations were significantly different between groups (a, p=0.010).  onstrates a clinical course that mimics the human condition including development of coagulopathy, intracranial hypertension, hyperammonaemia, cardiovascular collapse, elevation in creatinine, metabolic acidosis and hyperlactataemia. Consistent management of all animals was achieved through: deep sedation; intensive care monitoring; target driven protocols for treatment of cardiovascular collapse, metabolic acidosis and electrolyte abnormalities; IPPV; and CRRT. This resulted in a predictable time course to death, with a mean survival time following the onset of irreversible ALF of 12.6 ± 2.7 h with a coefficient of variation of 0.226. In addition, data presented demonstrate cardiac dysfunction in ALF, which has not been systematically studied in ALF and support the use of a combination of noradrenaline and terlipressin to manage systemic hypotension in ALF. Significant methaemoglobinaemia was prevented by adjusting APAP dosing to percentage methaemoglobin, as our own pilot studies (data not shown) revealed that once methaemoglobin exceeded 1%, serum APAP concentrations were >300 mg/L and APAP dose could be reduced. Mechanism for APAP-induced methaemoglobinaemia in pigs has not been defined, but likely results from co-oxidation of oxyhaemoglobin and subsequent redox cycling by an APAP metabolite, which is produced in higher concentrations in pigs compared with humans (8).
In APAP-induced ALF, PT is used as a criterion for OLT, justifying its use as an indicator of irreversible ALF in this model (1). FFP was given at the onset of ALF to prevent uncontrolled bleeding prior to progression of ALF to multi-organ failure. Also, a proton pump inhibitor was given to decrease risk of gastric mucosal bleeding in the presence of coagulopathy, because feed withdrawal in pigs results in gastric ulcerations in 80-100% of otherwise healthy animals (9). Coagulopathy-related blood loss from catheter sites may have contributed to decline in haemoglobin prior to death, despite the placement of all catheters prior to the onset of coagulopathy.
Intracranial hypertension is a reliable marker of existing or impending cerebral oedema in ALF patients with hepatic encephalopathy and is thought to be caused, at least partly, by hyperammonemia, which was observed in this model (10). If intracranial hypertension is uncontrolled, it has been shown to be associated with progression to death in ALF patients (11). In the current model, terminal spike in ICP likely contributed to death, as such marked elevations in ICP are known to result in bradycardia, widening pulse pressure and irregular respiration via the Cushing reflex.
In the current model, elevation in bilirubin and liver enzymes was not observed despite hyperammonaemia and liver histopathological features characteristic of APAP overdose. We report serum aspartate aminotransferase (AST) concentrations, as in the pig, AST has a high sensitivity for hepatocyte injury, unlike alanine aminotransferase (12). Nevertheless, failure to detect increase in AST, likely reflects the hyperacute nature of the current model (13) and the latency between hepatocyte injury and transfer of hepatocyte cytosolic contents to the blood. This is supported by previous ALF porcine studies, in which significant elevations in AST were only seen 24 h after elevation in International Normalized Ratio >3 (7). The current model therefore highlights the importance of the liver injury that occurs in the early stages of progression to ALF prior to the onset of liver inflammation (14).
The current model demonstrates the circulatory dysfunction seen in human ALF patients including marked hypotension, peripheral vasodilation, elevated cardiac output and decreased systemic vascular resistance (15). Falling systemic vascular resistance, progressive hypoalbuminaemia and haemoconcentration is associated with inappropriate vasodilation and excessive leak of intravascular fluids into the interstitium. Moreover, the current data demonstrate cardiac dysfunction indicated by a lack of increase in cardiac index and progressive reduction in both LVSWI and RVSWI. Cardiac dysfunction is well recognised in cirrhotic patients, but has not been systematically studied in ALF (16). The underlying mechanism is not clear, but may be similar to those in sepsis (17).
In cirrhotic patients, albumin has been shown to be a highly effective plasma expander (18,19). There are no data to support its routine use in ALF. The albumindosing regimen in this study was based on pilot studies in which albumin levels dropped to less than 10 g/L, resulting in marked haemodynamic instability. The dose of 100 g/day was chosen as this has been shown to be safe in cirrhotic patients (19) and improved haemodynamic stability in pilot studies. This study suggests that an appropriately controlled clinical trial of albumin infusion in ALF patients may be justifiable.
The use of noradrenaline, an a adrenergic agonist, to normalise vascular tone and improve arterial blood pressure in the current model was based on recommendations for human ALF patients (15). Use of terlipressin, which stimulates vasoconstriction via V1 receptors and possibly by a local inhibitory effect on nitric oxide production, in ALF is controversial, owing to demonstration that terlipressin increases ICP in ALF patients (20). However, studies have shown that addition of terlipressin to pre-existing noradrenaline therapy increases cerebral perfusion pressure without adverse effects on ICP (21). In our study the concurrent use of noradrenaline and terlipressin appeared to act synergistically to improve systemic haemodynamics in ALF.
Continuous renal replacement therapy (CRRT) was initiated in all animals to manage metabolic acidosis and acute kidney injury (AKI). In our own pilot studies, severe metabolic acidosis resulting in pH approaching 7.25 was noted in the absence of CRRT. Elevation in creatinine in this study was suggestive of AKI, which is commonly seen in severe APAP-induced ALF patients (22). Whether AKI in the current model is caused by direct APAP toxicity or a consequence of liver injury will be clarified in future specifically designed studies.
In this study, evidence of acute lung injury was seen at post-mortem in both control and ALF animals, implicating prolonged mechanical ventilation as a causal factor. Acute lung injury may also have occurred secondary to ALF (23), but this cannot be confirmed by the data presented in this article.
In conclusion, a reproducible, clinically relevant, intensively managed, pig model of acute liver failure with death as a result of multi-organ failure has been successfully validated for testing of novel therapies and/ or devices designed to prolong survival.