Multiscale Study of the Hepatic Volume Evolution After Major Hepatectomie in a Porcine Model Mohamed Bekheit

Multiscale study of the hepatic volume evolution after major hepatectomie in a porcine model Mohamed Bekheit

To cite this version:

Mohamed Bekheit. Multiscale study of the hepatic volume evolution after major hepatectomie in a porcine model. Surgery. Université Paris Saclay (COmUE), 2018. English. NNT : 2018SACLS033. tel-01753156

HAL Id: tel-01753156 https://tel.archives-ouvertes.fr/tel-01753156 Submitted on 29 Mar 2018

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Etude multi-échelle de l`évolution du volume du foie après hépatectomie majeure chez un modèle porcine

Multiscale study of the hepatic volume evolution after 2018SACLS033

major hepatectomy in a porcine model NNT : Thèse de doctorat de l'Université Paris-Saclay préparée à l'Université Paris-Sud et l`INSERM U1193, CHB, Paul Brousse

École doctorale n°569 : innovation thérapeutique: du fondamental à l'appliqué (ITFA) et sigle Spécialité de doctorat: SC

Thèse présentée et soutenue à Villejuif, le 26-1-2018, par Mohamed Bekheit

Composition du Jury :

Iréne VIGNON-CLEMENTEL Dir Rech, Établissement : INRIA & UPMC, Paris Président

Stéphanie TRUANT Pr, Établissement : Université de Lille Rapporteur

Ewen HARRISON Dr, Établissement: Université d`Edinburgh Rapporteur

Emilie GREGOIRE Dr, Établissement APHM Université de Marseille Examinateur

Eric VIBERT Pr, Établissement Université Paris Saclay Examinateur

Titre : Etude multi-échelle de l`évolution du volume du foie après hépatectomie majeure chez un modèle porcine

Mots clés : Hepatectomie majeure, porc, modulation de flux, modelisation, architecture, regeneration.

Résumé : L’ablation partielle du foie est une Des mesures expétimentales sont nécessaires chirurgie qui intervient dans le traitement des pour la construction et la validation de ces lésions du foie ainsi que lors d’une modèles. Des ablations du foie de différentes transplantation partielle de foie (donneur tailles sont effectuées sur des porcs et pendant vivant). Grâce à la capacité de régénération du ces chirurgies plusieurs pressions et débits sont foie, quelques mois après la chirurgie il retrouve mesurés. De plus, un colorant fluorescent est sa masse initiale. Les complications de cette injecté avant ou après l’ablation partielle, et la chirurgie sont l’insufisance hépatique et après fluorescence de ce composé est mesurée. Dans une transplantation le syndrome du trop petit une première partie, la procédure chirurgicale, foie. Ces deux complications sont liées à une les conditions expérimentales ainsi que les fonction hépatique post-opératoire faible. Les mesures obtenues sont détaillées. Ensuite, les relations entre l’hémodynamique du foie, son changements hémodynamiques, conséquence volume et ses fonctions restent à élucider pour de l’ablation partielle du foie. Le modèle permet mieux comprendre les causes de ces de prendre en compte les changements de complications. Lors de la chirurgie, volume sanguins qui peuvent se produire l’hémodynamique du foie est alterée suite à (saignements) lors de la chirurgie. Par l’augmentation de la résistance au flux sanguin conséquent, ce modèle propose une explication de l’organe. Les conséquences de cette chirurgie de la variabilité des mesures acquises lors de ces sur l’hémodynamique sont difficiles à analyser chirurgies. du fait de la double perfusion sanguine du foie. Puis, le transport dans le sang d’un composé En effet, le foie reçoit du sang oxygéné via ainsi que son traitement par le foie sont l’artère hépatique et du sang riche en nutriment modélisés. La dynamique d’un composé depuis via la veine porte. De plus, la régénération du l’injection intraveineuse jusqu’au moment où il foie semble dépendante des changements de atteint les vaisseaux du foie est analysé avec débit et de pression dans la veine porte. Dans ce des modèles. contexte, le objectif de cette thèse est de mieux Le contrôle des changements de débit et de comprendre, grâce à des modèles pression de la veine porte après une mathématiques, l’influence de l’hépatectomie hepatectomie pourrait protèger le foie restant sur l’hémodynamique. L`objectif est l’analyse de (ou le greffon) et améliorer sa régénération la perfusion et de la fonction du foie. Un modèle post-opératoire. Les deux sujets abordés dans de transport dans le sang d’un composé ainsi cette thèse ont pour but d’améliorer l’efficacité que la modélisation du traitement de ce d’un dispositif médical (anneau ajustable MID- composé par le foie sont développés. AVRTM) permettant ce contrôle.

Université Paris-Saclay Espace Technologique / Immeuble Discovery Route de l’Orme aux Merisiers RD 128 / 91190 Saint-Aubin, France

Title : Multiscale Study of Hepatic Volume Evolution after Major Hepatectomy in a Porcine Model

Keywords : Liver, resection, regeneration, mathematical model, hemodynamics, flow modulation

Abstract : The liver is a unique organ with a The observations made during the surgical multitude of characteristics. One of these, is the procedures and information gathered through remarkable ability to regenerate. In the context other parameters were integrated into multiscale of secondary liver cancers and given the simulations to understand the interactions extensive bilobar disease that we often see in the between the change of the hemodynamics and clinical practice, most of the patients are deemed the change of the hepatic volume that result from inoperable. One of the major reasons is that this resection or regeneration. Subsequently, a category of patients, if subjected to curative modulation device was used to test the effect of resection, will leave them with insufficient liver hemodynamic adjustments on the regenration for survival. The mechanisms by which the and survival. This was done primarly in a 75% resultant syndrome; small for size; is not fully resection model then on a smaller scale in a 94% understood. resection model. In the former model, the effect The beforehand work addressed certain was more seen on the level of the function, some translational aspects regarding the histological parameters but not the volume nor pathophysiology of the syndrome, the prediction the survival. On the other hand, the experiments of its occurrence prior to surgery. The setting with the ring (modulation device) lead to a and design involved the major and ultra-major surviving animal as opposed to no survival of resection in a porcine model. Along with animals without the ring. At the moment, this continuous hemodynamic monitoring and ICG shows a promising path for researchers to take dynamic assessment, the present work lead to a forward to help reducing the significant few steps forward in the knowledge and mortality that could potentially follow such understanding of the interaction between the major resection. Not only that but this sould hepatic, systemic circulation, and the hepatic allow more patient s to have access to the regeneration. The experimentations involved curative treatment they would other not have it. CT scanning prior to the operation to gain a comprehensive knowledge on the anatomy and volume, which was repeated in a temporal fashion following a strict protocol to monitor the volume changes as well. This was accompanied by 2D and 3D histological analysis and quantification of various regeneration parameters.

Université Paris-Saclay Espace Technologique / Immeuble Discovery Route de l’Orme aux Merisiers RD 128 / 91190 Saint-Aubin, France Synthèse de la Thèse :

L’ablation partielle du foie est une chirurgie qui intervient dans le traitement des lésions du foie ainsi que lors d’une transplantation partielle de foie (donneur vivant). Grâce à la capacité de régénération du foie, quelques mois après la chirurgie il retrouve sa masse initiale. Les complications de cette chirurgie sont l’insufisance hépatique et après une transplantation le syndrome du trop petit foie. Ces deux complications sont liées à une fonction hépatique post-opératoire faible. Les relations entre l’hémodynamique du foie, son volume et ses fonctions restent à élucider pour mieux comprendre les causes de ces complications. Lors de la chirurgie, l’hémodynamique du foie est alterée suite à l’augmentation de la résistance au flux sanguin de l’organe. Les conséquences de cette chirurgie sur l’hémodynamique sont difficiles à analyser du fait de la double perfusion sanguine du foie. En effet, le foie reçoit du sang oxygéné via l’artère hépatique et du sang riche en nutriment via la veine porte. De plus, la régénération du foie semble dépendante des changements de débit et de pression dans la veine porte.

Dans ce contexte, le objectif de cette thèse est de mieux comprendre, grâce à des modèles mathématiques, l’influence de l’hépatectomie sur l’hémodynamique. L`objectif est l’analyse de la perfusion et de la fonction du foie. Un modèle de transport dans le sang d’un composé ainsi que la modélisation du traitement de ce composé par le foie sont développés.

Dans une première partie, la procédure chirurgicale, les conditions expérimentales ainsi que les mesures obtenues sont détaillées. Ensuite, les changements hémodynamiques, conséquence de l’ablation partielle du foie. Le modèle permet de prendre en compte les changements de volume sanguins qui peuvent se produire (saignements) lors de la chirurgie.

Par conséquent, ce modèle propose une explication de la variabilité des mesures acquises lors de ces chirurgies.

Puis, le transport dans le sang d’un composé ainsi que son traitement par le foie sont modélisés. La dynamique d’un composé depuis l’injection intraveineuse jusqu’au moment où il atteint les vaisseaux du foie (artère hépatique et veine porte) est analysé avec des modèles 1D et 0D. Les résultats des simulations numériques sont comparés aux mesures de fluorescence de vert d’indocyanine. Afin d’analyser la dynamique du traitement du vert d’indocyanine par le foie, un modèle pharmacocinétique est développé. De plus, grâce aux mesures, les paramètres du modèle sont estimés dans le but de proposer une nouvelle méthode pour estimer la fonction du foie (pendant la chirurgie).

En effet, pour contrôler l’hémodynamique de la veine porte avec l’anneau, il faut tout d’abord connaitre l’impact de l’hépatectomie sur l’hémodynamique. De plus, l’efficacité de l’anneau pourrait être vérifiée grâce à l’estimation de la fonction hépatique pendant l’opération, en utilisant la mesure defluorescence du vert d’indocyanine. Cette thèse propose des premières pistes de réflexion dans le but d’améliorer la chirurgie hépatique. I. Table of contents

I. Table of contents ...... i II. List of Tables: ...... iv III. List of Figures: ...... vii IV. List of Equations...... xv V. Preface ...... xv VI. Translational statement:...... xviii VII. List of publications ...... xxii I. Introduction ...... 1

i. Anatomy of the liver ...... 4

ii. Assessment of the hepatic functions ...... 6

a. Biochemical assessment ...... 6

b. Breath tests ...... 8

c. Radiological based assessment of the liver function ...... 8

d. Indocyanine green plasma disappearance rate (ICG-PDR) ...... 9

e. Evaluation of the hepatic architecture ...... 10

f. Imaging modalities: ...... 12

g. Evaluation of the hepatic hemodynamic ...... 13

h. Global live assessment of the hepatic perfusion, architecture, and excretory function ...... 13

iii. Surgical disorders of the liver ...... 14

iv. Role of chemotherapy in colorectal liver metastasis:...... 16

v. Effect of chemotherapy on the liver functions: ...... 18

vi. Expanding the criteria for liver resection in colorectal liver metastasis: ...... 20

a. Portal vein embolization and ligation...... 20

b. Tumor progression after portal vein embolization in patients with colorectal liver metastasis ...... 21

c. Two stage resection ...... 22

d. ALPPS as a variant of the two-stage hepatectomy and variation ...... 23

e. Single stage resection ...... 23

vii. Liver regeneration ...... 24

a. Phase I: The priming: ...... 25

b. Phase II: The proliferation...... 25

c. Phase III: The termination ...... 25

d. Role of architecture in the regeneration process: ...... 26

viii. The relation between hemodynamics and regeneration after resection ...... 26

a. Necessity for increase in portal venous pressure/flow per unit volume ...... 26

b. The role of portal vein arterialization and the theory of sinusoidal shear stress versus portal venous contents: ...... 27

c. Small for size/flow syndrome ...... 28

d. Limits of the safe increase in the portal venous pressure/flow ...... 29

e. The portal flow modulation: ...... 30

ix. Mathematical modeling and their role in liver research ...... 31 II. Methods ...... 33

i. Rationale of the layout ...... 33

iii. Ethical approval...... 34

iv. Funding...... 34

v. Study settings ...... 34

vi. Description of the INRA ...... 35

vii. Study design ...... 37

viii. Housing and preoperative accomodation ...... 39

ix. Anaesthetic protocol ...... 40

x. Radiological protocols ...... 42

a. CT scan and Volume analysis ...... 42

b. Magnetic resonance based studies...... 45

c. Fluorescence imaging:...... 52

xi. Biopsy sampling ...... 54

xii. Histology of the porcine liver...... 58

a. Histopathological Analysis Protocol ...... 58

xiii. Sacrifice ...... 59

xiv. Statistical methods...... 60 III. Results & Discussions: ...... 61

i. Anatomy of the porcine liver and technical implications ...... 61

a. Introduction to Hepatic Anatomy ...... 62

b. Anatomical features...... 62

c. CT scan depicted description ...... 65

d. Implications ...... 78

ii. Physiology...... 89

a. Normal Laboratory Values ...... 89

iii

b. Evolution of the hematological parameters after different resection percentages: ...... 97

c. Evolution of the blood gases parameters: ...... 111

d. Hemodynamics ...... 132

e. Kinetics of the hepatic volume evolution and the architectural changes following 75% resection ...... 152

f. Normal ICG handling detected by live imaging ...... 163

g. Changes following resection ...... 165

h. Mortality and survival ...... 168

iii. Clinical Application ...... 171

i. Radiology ...... 171

j. Software Applications ...... 195

k. Flow Modulation Device ...... 195 IV. Limitations and perspectives: ...... 226 V. References ...... 228 VI. Appendix ...... 252

iv. MRI Elastography protocol ...... 252

a. Context ...... 252

b. Method ...... 252

c. Results ...... 252

d. Questions ...... 261 VII. Abstrait ...... 264

II. List of Tables:

Table 1: ‘Indirect’ serological markers for the prediction of liver fibrosis ...... 12

Table 2: The different common anti-cancerous classes in clinical use. Adapted from (115) ...... 16

Table 3: Common chemotherapeutic regimens for colonic cancer...... 18

Table 4: Examples of the known hepato-specific adverse effects of the commonly used chemotherapeutic agents...... 19

Table 5: Total number of specimens collected from group 2-4 experiments ...... 57

Table 6: The summary of the segmented lobar volumes in the studied animals ...... 78

Table 7: Kidney and major electrolytes profile (samples collected from jugular vein) ...... 90

Table 8: Arterial blood gases, PH and lactate levels ...... 91

Table 9: Venous blood gases, PH and lactate levels ...... 92

Table 10: Portal vein blood gases, PH and lactate levels ...... 93

Table 11: Hepatic profile from samples collected from the internal jugular vein...... 94

Table 12: major hematogram parameters in samples collected from the internal jugular vein...... 95

Table 13: Hemoglobin and hematocrit levels measured in each sample type of gas analysis ...... 96

Table 14: Calcium levels as measured by gas analyser in the arterial, venous and portal venous samples and the hepatic oxygen consumption and the net lactate production...... 97

Table 15: Descriptive Statistics for the evolution of hematological parameters following 75% liver resection ...... 97

Table 16: Descriptive Statistics for the evolution of hematological parameters following 90% liver resection ...... 102

Table 17: Summary table of the available data on prothrombin time and activity...... 105

Table 18: Descriptive statistics of the calcium levels in the blood gas samples...... 111

Table 19: Summary of lactate level in both resection groups ...... 115

Table 20: Oxygen tension (mmHg) in blood gases and its evolution in groups submitted to two different resection volume .. 119

Table 21: Oxygen saturation in blood gases and its evolution in groups submitted to two different resection volume ...... 122

Table 22: Hepatic oxygen consumption levels in relation to the timing of liver resection...... 124

Table 23: Evolution of the hemoglobin and haematocrit level in blood gases ...... 126

Table 24: Summary of the normal pressure and flow parameters prior to clamping or resection...... 132

Table 25: Evolution due to mechanical ventilation and heartbeat average and standard deviation...... 133

Table 26: Hemodynamics measurements before and after resection...... 140

Table 27: The different mass assumptions description of the total liver, left lobe, right lobe and median lobe...... 143

Table 28: Summary of the evolution of the hepatic functions and flow parameters in the 75% resection group ...... 157

Table 29: Examples from different animals with different DWI values at different time points in relation to surgery ...... 174

Table 30: The liver volume at each experimental time point for all groups ...... 186

Table 31: The liver volume at each experimental time point for each resection %...... 187

Table 32: Pre-hepatectomy data from both groups ...... 211

Table 33: Different parameters measured after liver resection and on day-7 post-operative ...... 213

Table 34: List of parameters quantified in 3D for bile canalicular network...... 215

Table 35: The microarchitectural damage score for both groups...... 220

Table 36: The main liver functions of the animals in both groups ...... 223

III. List of Figures:

Figure 1: Illustration of typical hepatic segmental anatomy...... 5

Figure 2: Simplified scheme showing the components necessary for integral hepatic functions assessment ...... 7

Figure 3: Schematic representation of the microarchitecture in healthy (top) and Cirrhotic (bottom) livers...... 11

Figure 4: Simplified scheme demonstrating the clinically relevant hepatic function assessment tests...... 14

Figure 5: Tityus (c.1532), Michelangelo Buonarroti, Royal Collection, Windsor Castle, U.K...... 24

Figure 6: Illustration of the various experimental models of liver regeneration...... 32

Figure 7: One of the operating theatres at the INRA ...... 36

Figure 8: The equipment set up at the CIRE platform...... 36

vii

Figure 9: Flow chart demonstrating the workflow within the different iFLOW groups ...... 38

Figure 10: Photograph demonstrating one animal during the CT scan images acquisition...... 43

Figure 11: Screen shot showing the timing protocol for image acquisition during CT scan...... 43

Figure 12: Hepatic arterial and portal venous segmentations using SyngoVia, Siemens...... 44

Figure 13: 3D reconstructed images for planned a) 75% and b) 90% resection...... 44

Figure 14: PC-MRI flow measurements. ROI on celiac aorta (lumen)...... 50

Figure 15: Fluorescence element of reference to normalize the measured intensities...... 53

Figure 16: The four regions of interest are present, the liver tissue, the hepatic artery, the portal vein and the common bile duct

...... 54

Figure 17: Intra-operative photograph depicting the post-resectional incisional biopsy...... 56

Figure 18: photograph of an incisional biopsy immersed into 4% formaldehyde...... 56

Figure 19: core needle biopsy with the deep end marked with Fuscin green...... 57

Figure 20: A reconstructed image of the hepatic vascular anatomy...... 61

Figure 21: Branches of the celiac trunk and the hepatic artery...... 70

Figure 22: Volume rendering technique extracting the part of the arterial tree in relation to the portal vein...... 71

viii

Figure 23: Volume rendering technique extracting the part of the arterial tree in relation to the portal vein...... 72

Figure 24: Volume rendering technique extracting the part of the arterial tree in relation to the portal vein...... 72

Figure 25: Intraoperative photo showing the minute arterial branch to the caudate lobe crossing over the portal trunk to the caudate lobe...... 73

Figure 26: 3D ROI reconstruction for segmented portal and hepatic veins...... 74

Figure 27: 3D ROI reconstruction for segmented portal and hepatic veins...... 75

Figure 28: An intraoperative photo demonstrating portal venous branch crossing to the left lateral sector at the base of the fissure separating the left lateral from the right lateral sectors...... 75

Figure 29: Cranial view of a 3D reconstruction of the segmented hepatic sectors showing the common trunk of draining the veins from the right lateral and one of two right medial sectors...... 76

Figure 30: Caudal view of a 3D reconstruction of the five hepatic sectors showing their volume...... 77

Figure 31: Neck incision with cannulated jugular vein (J) and carotid artery (ca), probes installed for flow measurements around (p) portal vein, (H) hepatic artery ...... 82

Figure 32: technical aspects during liver resection ...... 85

Figure 33: Tightening of the clamping tourniquet around the 75% mass ...... 86

Figure 34: the passage of right angled clamp behind the first (a) and the second (b) portal pedicles ...... 87

Figure 35: for 90% to be completed; a) passage and b) tightening of the clamping tourniquet ...... 87

Figure 36: Box-plot with dots representing the alkaline phosphatase levels in both groups ...... 106

Figure 37: Box-plot with dots showing ALT levels over time in both groups ...... 107

Figure 38: Box-plot with dots showing ALT levels over time in both groups with the outlier excluded ...... 107

Figure 39: Box-plot with dots showing Amonia levels over time in both groups...... 108

Figure 40: Box-plot with dots showing AST levels over time in both groups...... 108

Figure 41: Box-plot with dots showing Direct Bilirubin levels over time in both groups...... 109

Figure 42: Box-plot with dots showing Total Bilirubin levels over time in both groups...... 109

Figure 43: Box-plot with dots showing GGT levels over time in both groups...... 110

Figure 44: Box-plot with dots showing the evolution of the prothrombin activity...... 110

Figure 45: Box-plot with dots representing the evolution of calcium in the arterial system ...... 113

Figure 46: Box-plot with dots representing the evolution of calcium level in the portal sample ...... 113

Figure 47: Box-plot with dots representing the evolution of calcium level in the supra-hepatic venous sample ...... 114

Figure 48: Box-plot with dots representing the evolution of lactate level in the arterial blood gas sample...... 116

Figure 49: Box-plot with dots representing the evolution of lactate level in the portal venous blood gas sample ...... 117

Figure 50: Box-plot with dots representing the evolution of lactate level in the supra-hepatic venous sample ...... 117

Figure 51: Box-plot with dots representing the evolution of the transhepatic lactate level in groups ...... 118

Figure 52: Error bar demonstrating the evolution of the ammonia level seen in the haematology samples...... 118

Figure 53: Box-plot with dots illustrating the evolution of the oxygen tension in the arterial blood ...... 120

Figure 54: Box-plot with dots illustrating the evolution of the oxygen tension in the portal venous blood ...... 120

Figure 55: Box-plot with dots illustrating the evolution of the oxygen tension in the hepatic venous blood ...... 121

Figure 56: Box-plot with dots illustrating the evolution of arterial oxygen saturation...... 123

Figure 57: Box-plot with dots illustrating the evolution of portal venous oxygen saturation...... 123

Figure 58: Box-plot with dots illustrating the evolution of hepatic venous oxygen saturation...... 124

Figure 59: Box-plot with dots representing the evolution of hepatic oxygen consumption...... 125

Figure 60: Pressures in mmHg over a few respiratory cycles before clamping or liver resection...... 135

Figure 61: Flows in Liter/minute over a few respiratory cycles at the beginning of surgery, before clamping...... 135

Figure 62: Pressure panel (a) showing the increase in the hepatic artery and portal venous pressures upon clamping, while the central venous and carotid artery pressures have not changed...... 137

Figure 63: Portal vein (a) and hepatic artery (b) flow per liver weight before and after 75% liver resection...... 137

Figure 64: Box-plot demonstrating the amount of change in the different pressure values following the resection in each type of resection...... 138

Figure 65: Box-plot demonstrating the global reduction in the flow values following resection in the different groups...... 139

Figure 66: Simplification of the dimensions of the mathematical fluid modelling...... 141

Figure 67: Schematic representation of the 0D closed-loop cardiovascular and liver blood circulations. RCR block and liver lobe parameters are shown...... 144

Figure 68: Pre-resection measurements vs simulation values in log/log scale, for each variable (unique color) ...... 146

Figure 69: Measurements (full) and simulations (dash) at different states of the surgery: pre-resection, post-resection...... 147

Figure 70: Screenshot of the observed hepatic arterial waveform before and after hepatic pedicle clamping ...... 152

Figure 71: CT scan estimated liver volume in the peri-operative period ...... 153

Figure 72: Portal vein (left) and hepatic artery (right) flow per liver volume ...... 154

Figure 73: Histopathological analysis under light microscopy of the porcine liver ...... 155

Figure 74: Boxplot representing the temporal change in the Ki67 and Bar diagram plotting the CD31 expression...... 156

Figure 75: Normal fluorescence signal intensity curve in the four target ROIs ...... 164

Figure 76: Fluorescence signal intensity curve after 75% resection in the four target ROIs ...... 165

Figure 77: The fluorescence signal pattern after the 90% resection...... 166

Figure 78: The fluorescence signal curve on the first day after 75% resection...... 167

xii

Figure 79: The fluorescence signal curve on the seventh postoperative day after 75% resection...... 168

Figure 80: Biexponentional model fitting example using 9 b-values for parameter estimation, day-3 after resection...... 173

Figure 81: Ki67 proliferative index in specimens taken at different time points...... 175

Figure 82: 3D reconstruction of confocal microscopy images ...... 176

Figure 83: Boxplot of the blood flow measurements in the celiac aorta (Qoc), portal vein (Qpv) and the hepatic artery (Qha) in

the MRI (MR) and the transit time (TT)...... 177

Figure 84: Least square regression for PC-MRI and TT flow measurements. PC-MRI flow measurements...... 179

Figure 85: Bland-Altman plot indicating the systematic difference between the flow reading in MRI and Transit time in the

aorta...... 180

Figure 86: Bland-Altman plot indicating the non-signifcant difference in estimation of the flow in the hepatic artery using either the MRI or the transit time method...... 181

Figure 87: Bland-Altman plot revealing the systematic underestimation of the MRI flow readings compared to the Transit time

flow readings...... 182

Figure 88 Box-plot with dots representing the hepatic volume evolution in both groups...... 188

Figure 89: Boxplot with dots representing the rate of hepatic mass recovery following resection in both resection %...... 189

Figure 90: Axial CT scan image depicting the planned resected volume (blue), and the planned residual volume (pink) for 75%

resection...... 190

xiii

Figure 91: Axial CT scan image depicting the recovered hepatic volume at day 3 following 75% resection (pink) and the spleen in blue...... 190

Figure 92: Axial CT scan image depicting the residual hepatic volume after a 90% resection (pink). The estimated volume is

114 cm3...... 191

Figure 93: 3D reconstruction image at day 3 following a 90% resection with portal ring placement...... 191

Figure 94: Coronal CT scan image at day 3 following a 90% resection with portal ring placement...... 192

Figure 95: Axial CT image at day 3 following a 90% resection with portal ring placement...... 192

Figure 96: The MID-AVR™ in its different shapes according to the degree of balloon inflation...... 201

Figure 97: Scatter diagram showing animal mortality in both groups stratified according to the change in portal flow per unit liver mass...... 205

Figure 98: Porto-caval pressure gradient was significantly higher in the control group compared to the MID-AVRTM ...... 206

Figure 99: Total bilirubin level was higher in the control group than in the MID-AVRTM group...... 207

Figure 100: Biopsies taken from the control (left panel) and ring (right panel) groups ...... 208

Figure 101: Analysis of hepatocyte proliferation in regenerating pig livers ...... 209

Figure 102: Analysis of hepatocyte proliferation index in regenerating pig liver...... 210

Figure 103: Evolution of the damage score after the hepatic resection in both groups...... 221

xiv

Figure 104: CT scan volumetry of the hepatic remnant and evolution in 90% resection ...... 222

Figure 105: The junction between the valve and the tube is showing small leak ...... 225

IV. List of Equations

Equation 1: Intravoxel incoherent motion ...... 45

Equation 2: The phase difference ...... 47

Equation 3: The hepatic O2 consumption ...... 51

Equation 4 : The hepatic net lactate production ...... 51

Equation 5: 1D Model equation ...... 151

Equation 6: The estimate of bias between both measures using Bland-Altman method ...... 180

Equation 7: The estimate of bias in the transit time measurement ...... 181

V. Preface

This work is the culmination of the scientific wealth of the Unité Inserm 1193 (Hôpital Paul Brousse, Villejuif – Directeur Pr.

D. Samuel) under the supervision of Prof. Eric VIBERT. This project is one of many projects that target the improvement of patients` care, centered on the hepatic remnant interaction with the surgically induced alterations in the hemodynamics.

A unique feature of this project is its multidisciplinary nature. It encompassed the collaboration of Surgeon, the Paul Brousse

Team represented by Eric Vibert, CHU Tours represented by Petru Bucur, and me, mathematicians, the INRIA of Paris teams

represented by the group REO of J-F. Gerbeau, I. Vignon-Clémentel, and C. Audebert and the team MAMBA represented by

D. Drasdo, and N. Boissier with their contribution and expertise in developing mathematical models to predict the interaction

at a multiscale level, which has the potential of being integrated into software applications, hapato-scientists from IFADO in

Dortmund, Germany represented by J. Hengstler, B. Begher-Tibbe, S. Hammad, and A. Othman whose contribution was unique in developing staining protocols and three dimensional imaging of the hepatic microarchitecture, the team from

Fluoptics France, represented by S. Guillermet, P. Rizo, A. Daures whose contribution targeted the development of novel algorithms to precisely estimate the live fluorescence signal and its interpretation, and the team from MID France, led by

Ludovic Cazenave with their novel device for portal flow modulation. The project was granted financial support from TecSan

de l’Agence Nationale de Recherche en 2013 (ANR-13-TECS-0006, iFLOW project). With this large scientific power, this collaboration was fruitful in achieving many of the proposed objectives.

Lastly, I am overwhelmed with the sincere mentorship I received during the entire working time from Professor Eric Vibert.

The story of me joining the project and the University to obtain this degree is something that I would remember for life.

Notably, the guidance and help I received from my honest and kind colleague; Petru Bucur, who transferred to me his considerable experience in the porcine hepatic surgery. I do not think I will be able to forget those few years I spent learning from him. I will not overlook the support I got from the INRA team and their collaboration to make the work flows smoothly and without much of a hassle.

There are too many people – to whom I am grateful – to mentioning by name. However, I must acknowledge the support I got from the mathematicians; Irene is on the top of the list. She was one of the reasons this work is walking to see the light. I am also grateful to Dr. Seddik Hammad and Noemie Boissier who guided me through the microarchitecture work.

Mohamed Bekheit

xvi

Aberdeen 2017

xvii

VI. Translational statement:

The liver has a remarkable capability to regenerate. However, patients undergoing liver surgery often suffer from liver diseases accompanied by a significant reduction in liver function and regenerative capacity. Postoperative liver failure is the main cause of short-term mortality (around 5% at 3 months after surgery) after hepatectomy (5000 / year in France) due to insufficient functional liver mass. Today, limits of liver surgery and partial liver transplantation are based on empiric minimal acceptable liver volume that is preoperatively defined on volumetric reconstruction using CT-scan. These limits are defined according to a priori ratios between liver volume and liver function that depend on the quality of liver parenchyma and the therapeutic situation. However, the evaluation based on these threshold values fails if it leads to important modification of the hepatic hemodynamics. Porcine model is one of the extensively used models for study of liver regeneration and factors implicated in the regeneration. The study of porcine liver regeneration – among other areas – is based on the similarities to the human anatomy and physiology. Therefore, the results of these studies are closely applicable to humans.

The primary objective of this project is to investigate the mechanistic interplay between the hemodynamic changes of the hepatic inflow and the post-resection failure and regeneration. The understanding phase is designed using various tools, which are commonly used in the clinical practice in order to obtain a valid translational value from this work. In addition to understanding the pathophysiology; we targeted the deployment of novel prediction tools that will enable the hepatic surgeons to anticipate with precision the clinical consequences of the required liver resection highlighting instead the feasible and safe resection. Moreover, the project aims at examining the role of a newly invented intervention device to enhance the hepatic microarchitecture integrity – a vital mediator for the proper liver function – subsequently improving the quality of the regenerating parenchyma after major liver resection.

The study protocol was designed in a staged fashion to enable goals achievements. In this regards, animals were allocated to the following groups:

xviii

· Group 1: composed of randomly allocated animals to undergo 75% liver resection with or without the application of the

adjustable vascular ring around the portal vein. This group consisted of 17 animals, of them 8 animals underwent liver resection

and ring placement and 9 animals underwent only liver resection. This allocation targeted the study of the safety and the efficacy

of the vascular ring in ameliorating the hepatic function and regeneration following this type of hepatectomy.

· Group 2: consecutive series of 16 animals that underwent 75% resection with 7th day sacrifice protocol. This allocation targeted

optimization of the invasive hemodynamics monitoring of the porcine model while conducting liver resection, which would

allow building up accurate predictive mathematical models at multi-scale level and investigating the role of ICG (indocyanine

green) in in-vivo evaluation of liver function during major liver resection.

· Group 3: composed of consecutively allocated animals to undergo 90% resection with and without the application of the

adjustable vascular ring around the portal vein. This group consisted of 6 animals, of them 3 animals underwent 90% liver

resection with the application of the vascular ring around the portal vein and 3 animals underwent only liver resection. This

allocation targeted the study of the efficacy of the vascular ring in ameliorating liver function and regeneration following this

potentially lethal liver resection.

· Group 4: consecutive series of 6 animals that underwent 75% liver resection with 3rd day sacrifice protocol. This allocated

targeted the investigation of early volume and histopathological changes.

Originality and goals

The specific goals of this project are

1) Study the volume evolution in association with the hemodynamic and biochemical changes induced by surgery.

2) To evaluate the potential of an implantable surgical device able to modulate the portal hemodynamics to improve their

surgical outcome.

xix

3) To develop a mathematical model relating blood liver perfusion, architecture and function to assess the fluorescence signal obtained in (1) with regard to potential liver failure and to predict the likely effect of treatment by an adjustable ring as in (2).

The model will address three levels, the whole organ level by a compartment model (macro scale), the tissue scale in which larger vessels are represented while flow between them are mimicked by a porous media approach (meso scale) and the lobule level in which the detailed architecture of the individual liver lobule, the smallest repetitive functional unit of liver, is represented (micro scale). The parameters of the meso scale will be calibrated with those on the micro scale, the parameters of the macro scale with those on the meso scale.

The main findings:

1. The technique we developed to resect the porcine liver is simple and efficient for both 75% and 90 % resection.

2. An accurate CT based anatomy as well liver volume estimation of the porcine liver is now available.

3. A critical description of the surgical anatomy of pigs to enable safe surgery and monitoring of the hemodynamic

parameters is now available.

4. A protocol to perform diffusion weighted imaging- MRI (DWI-MRI) on regenerating porcine liver is now optimized.

5. A comparison between the MRI estimated hepatic blood flow and the transit time based estimates revealed that the MRI

tends to underestimate the flow values compared to the transit time. However, the MRI values are also valid.

6. The 75% resection is unlikely to cause liver failure.

7. 90% resection is generally lethal.

8. The portal ring has promising potentials in modifying the lethal outcome of the 90% resection.

9. The porcine liver regains the majority of its resected mass during the first 3 days

10. Explanatory hemodynamic mathematical models have been constructed

Pending work and perspectives:

1. 3D histopathological quantification to integrate into the correlation with the DWI of regenerating livers.

2. Optimization of the fluorescence model to enable predictive interpretation prior to resection.

xxi

VII. List of publications

Journal Articles:

Published

1. Bekheit M, Bucur P, Vibert E, Andres C. The reference values for hepatic oxygen consumption and net lactate

production, blood gasses, hemogram, major electrolytes, and kidney and liver profiles in anesthetized large white swine

model. Transl Surg 2016;1:95-100

2. Bekheit M, Bucur P, Wartenberg M, Vibert E, Computerized tomography based anatomical description of the porcine

liver, Journal of Surgical Research (2016), doi: 10.1016/j.jss.2016.11.004.

3. Bucur P, Bekheit M, Audebert C, V-Clemente I, Vibert E. Simplified technique for 75 & 90 resection of porcine liver

with hemodynamic monitoring. Accepted Journal of Surgical research.

4. Bekheit M, Bucur P, Audebert C, Othman A, Hammad S, Cazenave L, Dirk Drasdo, Jan G. Hengstler, Irene Vignon-

Clementel and Eric Vibert. New Technique for Portal Flow Modulation in Porcine Major Liver Resection. Clin Surg.

2016; 1: 1219

5. Bucur P, Bekheit M, Chloe´ Audebert, Amnah Othman, Seddik Hammad, Mylene Sebagh, Marc-Antoine Allard,

Benoıˆt Decante, Adrian Friebel, Dirk Drasdo, Elodie Miquelestorena-Standley, Jan G. Hengstler, Irene Vignon-

Clementel, Eric Vibert. Modulating Portal Hemodynamics With Vascular Ring Allows Efficient Regeneration After

Partial Hepatectomy in a Porcine Model. Ann Surg 2017 Feb 1. doi: 10.1097/SLA.0000000000002146

xxii

6. Chloe Audebert, Bekheit M, Petru Bucur, Irene Vignon-Clemente, Eric Vibert. Partial hepatectomy hemodynamics

changes: experimental data explained by closed-loop lumped modeling. Journal of biomechanics.

http://dx.doi.org/10.1016/j.jbiomech.2016.11.037

7. Chloe Audebert, Petru Bucur, Bekheit M, Eric Vibert, Irene Vignon-Clementel, et al.. Kinetic scheme for arterial and

venous blood flow, and application to partial hepatectomy modeling.. Computer Methods in Applied Mechanics and

Engineering, Elsevier, 2016, <10.1016/j.cma.2016.07.009>. .

8. Bekheit M, Bucur P, Vibert E. The ideal porcine model for major liver resection, is there any yet?. J Surg Res. 2017

Mar 7. pii: S0022-4804(17)30132-4. doi: 10.1016/j.jss.2017.03.00

Submitted:

9. Bekheit M, Petru Bucur, Audebert C, Miquelestorena-Standley E, V-Clemente I, Vibert E. Hepatic volume evolution

after major resection in a porcine model: an insight on the correlation between the volume increase and the pathological

changes.

10. Bekheit M, Petru Bucur, Audebert C, Adriaensen H, Bled E, Wartenberg M, , V-Clemente I, Vibert E Validation of the

hepatic and aortic blood flow values measured using phase contrast MR imaging using transit time perivascular probes

in a porcine model.

11. Chloe Audebert, Petru Bucur, Bekheit M, Eric Vibert, Irene Vignon-Clementel. Impact of 75% partial hepatectomy

on hemodynamics in porcine model. Submitted journal of physiology.

xxiii

12. Noemie Boissier, Chloe Audebert, Bekheit M, Petru Bucur, Eric Vibert, Irene Vignon-Clementel.

Mathematical modeling for quantifying the impact of micro-architectural changes on hepatic and systemic

hemodynamics: insights after partial hepatectomy.

Book Chapter:

13. Bekheit M; Vibert E: Fluorescent guided liver surgery: Paul Brousse experiences and perspective. in Concepts and

Applications of Fluorescence Imaging for Surgeons (Dip and Ishizawa), Springer 2015, X pages 415

Conference proceedings:

14. C Audebert, P. Bucur, Bekheit M., E. Vibert, J-F. Gerbeau, I. E. Vignon-Clementel. Mathematical modeling of liver

hemodynamics during partial liver ablation. EUROMECH Colloquium 595: Biomechanics and computer assisted surgery

meets medical reality 29-31 August 2017, Centrale Lille, Villeneuve d’Ascq, France

15. Hans Adriansen, Bekheit M, Chloe Audebert, Petru Bucur, Eric Vibert. Elastographie par Résonance Magnétique sur

porcelet: méthode alternative et rapide pour vérifier la résistance hépatique, utilisée conjointement avec le calcul de

débit sanguin. Gen2bio 26/3/2015 La Baule, France.

16. Chloe Audebert, Petru Bucur, Bekheit M, Irene Vignon-Clement, Eric Vibert, Jean Federic. Cardiovascular closed-loop

model and partial hepatectomy simulation. The 8th Bio-fluid 12-14 Feb-2016 Caltech.

xxiv

I. Introduction

Major liver resection (partial hepatectomy) is being performed to treat liver lesions or for adult-to-adult living donor liver transplantation. Due to liver regeneration, during the postoperative period lasting a few months, the patient regains a normal liver mass. The major complications of these surgeries are postoperative liver failure (after partial hepatectomy) and small-for- size syndrome (after partial transplantation). Both complications are related to a poor liver function. The links between liver hemodynamics, liver volume, and liver function remain unclear and are necessary to understand these complications better.

The surgery increases the resistance to blood flow in the organ. Therefore it modifies liver hemodynamics. These changes are difficult to understand, partly because the liver receives arterial (through the hepatic artery) and venous blood (through the portal vein). Large modifications of the portal vein hemodynamics have been associated with poor liver regeneration.

Moreover, the liver receives 25% of the cardiac outflow. Therefore liver surgery may impact the whole blood circulation.

In this context, the first goal of this thesis is to investigate with mathematical models the impact of liver surgery on liver hemodynamics. The second goal is to study the liver perfusion and function with mathematical models of the transport of a compound and its processing by the liver. Data is required to build and verify these mathematical models. Therefore, liver resections were performed on pigs, during which various pressures and flows were recorded. Moreover, indocyanine green, a dye exclusively eliminated by the liver into the bile, is intravenously injected before or after liver resection and its fluorescence is measured.

The first part of this thesis describes the experimental conditions and reports the measurements recorded. Then, the second part focuses on the liver hemodynamics during partial hepatectomy. On the one hand, the hemodynamics during several surgeries is quantitatively reproduced and explained by a closed-loop model based on ordinary differential equations. The closed-loop model has enabled to take into account blood volume changes occurring during liver surgeries. On the other hand, the variation of waveforms observed after different levels of liver resection is reproduced with a model of the global circulation, including

0D and 1D equations. This may contribute to a better understanding of the change of liver architecture induced by

hepatectomy.

Next, the transport in the blood of a compound, as well as the indocyanine green processing by the liver, is studied. The 0D

and 1D equations model the dynamics of a compound from its intravenous injection to when it reaches the hepatic artery and

the portal vein. The results are compared to the indocyanine green fluorescence measurements. A new framework is

established to analyze indocyanine green fluorescence dynamics quantitatively. It consists in the development of a specific

pharmacokinetics model and its parameter identification. The aim is to provide a novel estimation of the liver function(s) peri-

operatively using indocyanine green fluorescence measurements.

The modulation of portal vein hemodynamics may protect the liver and improve the regeneration, as the portal hemodynamics

impacts the regeneration. The two topics of this thesis aim at improving the efficacy of a medical device controlling portal vein

hemodynamics (vascular ring MID-AVRTM). Knowing the hemodynamics evolution during liver surgery enable to know how

to modulate these hemodynamics with the ring. Moreover, the ring efficacy could be controlled by the indocyanine green

fluorescence measurements to quantify and thus evaluate the liver function intra-operatively. This thesis can be seen as the first

step to reach the final goals.

Understanding the organ development is crucial to comprehend the underlying mechanisms of liver regeneration following a

hepatic mass loss. The liver originates from the endoderm of the ventral bud of the foregut tube which originates from the

embryonic ventral streak. Embryonic stem cells migrate into this streak to form the definitive endoderm that further constitutes the foregut after rotation. The embryonic stem cells (Sox17+ cells) develop from this line giving the hepatobiliary pancreatic progenitor cell line that further gives pancreatic progenitor cells and hepatoblasts (1).

Their specification into hepatoblasts requires, among others, the endothelium (2). Endothelial cells were found significant in the development of the liver bud and their ablation, as in VEGF receptor deficient mice, intervenes with the development of hepatic bud.

Following their specification, these bipotential hepatoblasts proliferate and differentiate into hepatocytes and cholangiocytes and express hepatocyte (αFP, albumin) and cholangiocyte (CK19) markers. Maturation is known by the expression of albumin by the hepatocytes, while the commitment of the hepatoblasts to the hepatocytes cell line is represented by the expression of

αFP (3).

A subpopulation of stem cells exhibits different markers compared to the hepatoblasts (negative for albumin and αFP and adult liver cell marker [cytochrome P450s] but positive for CD34, CD133) with different characteristics where they are capable of

150 population doubling and generation of hepatoblasts (4).

The embryonic ductal plate goes into remodeling to become the canal of Hering in pediatric and adult liver (5). The intrahepatic biliary ducts develop from the periportal hepatoblasts (6). The elongation of the hilar hepatic duct forms the large intrahepatic biliary ducts and the peribiliary glands. The caudal part of the hepatic diverticulum connected to the foregut tube forms the extrahepatic biliary tract and the ventral pancreas (7).

Multiple stem cell niches persist in the adult life in specific locations. Biliary stem cells are located in the peribiliary glands around the large intrahepatic and extrahepatic biliary tree, hepatic stem cells located near or in the canal of Hering, and the pancreatic stem cells confined to the biliary tree (6). During hepatic regeneration following injury, the oval cells residing in the biliary radicles plays a major role in restitution of hepatic cell mass by giving rise to hepatocytes (8). In humans, the cholangiocytes develop from the hepatoblasts located near the portal vein, while those located at a distance from the portal vein give rise to the hepatocytes (9). Upon maturation, hepatocytes form a heterogeneous cell population which is characterized by their different zonal metabolic characteristics (10).

The hepatic stellate cells are derived from the embryonic septum mesothelium (11). Since they express germinal markers similar to those arising from the three germ layers, it is possible that these cells are derived from multiple sources or have a pluripotent potential. They are located in the space of Disse representing a major constituent of the hepatic non-parenchymal cells. These cells are responsible for deposition of the extracellular matrix (ECM) when activated into a-SMA-expressing

myofibroblasts. Their activation and quiescence are dependent on the Wnt pathway, and the knockout of B-catenin resulted in

the development of activated stellate cells (a-SMA-expressing) with dilated sinusoids (12). They are also involved in the

differentiation of hepatocytes as well as recruiting non-hepatocyte hematopoietic cells into fetal hepatic cells (13).

The kupffer cells that reside in the sinusoids and constitute around 15 % of the hepatic cells (14) are thought to have a driving influence on the bipotential progenitor cells to undergo hepatocyte development (15).

In Zebrafish model, cells of biliary origin participated in hepatocyte formation following extensive hepatocytes loss (16). The transdifferentiation of hepatocyte to biliary epithelium was seen in a rat model where hepatocytes contributed to the restoration of biliary epithelium after a massive loss (17).

i. Anatomy of the liver

The normal adult human liver weighs between 968- 1860 grams (18). It receives – uniquely – a dual blood supply with 75%

from the portal vein and 25% from the hepatic artery (19,20). Based on the further divisions of the portal vein, the liver is

divided into two hemilivers (first order branching), further divided into four sections (second order branches) and 9 segments

(third order branches) (21). The segmental division of the liver has been illustrated in Figure 1.

A further ramification of the portal vein and the hepatic artery take place to a microscale level interlacing the hepatic cords –

composed of hepatocytes intervened by bile canaliculi and spaces of Disse – to form the micro-unit of the liver (22). Two –

commonly used – but different descriptions of the smallest unit exit. The hepatic lobule is organized in a hexagonal fashion

with six portal vein ramifications representing its outline and a central vein located at the center of this hexagon. This follows

the macroscale anatomical description as an extension of the portal based subdivisions of the organ (23). Each lobule is subsequently divided into 6 sinusoidal segments or plates (24).

On the other hand, the acinus division seems to be more functional than anatomical (10). In this divisional concept, the unit is organized around a single portal vein and bounded by small hepatic veins (central veins of the lobule). This concept of

functional units is supported – from the functional perspective – by studies addressing the different gene and metabolic

functions expressions in the different zones (at various distances) from the portal branch (25).

The previous description was largely based on two-dimensional structural stains. Further, in depth analysis of the ultrastructure of the liver units suggests a more complex formation. In three dimensional analysis of the porcine hepatic unit structure, fibrous connective tissue was found to surround the lobule in two different patterns (simple and compound) (26).

A modular structure of the hepatic micro-organization has been proposed relatively recently by Teutsch et al. initially in rats

(24) then in humans to describe the variability in the dimensions and size of the liver units (27). The modular unit is composed

of a secondary lobule that contains around 14 lobules of different dimensions surrounded by a portal vascular tree connected

and draining into a sublobular vein (27). The main interest of this information is that it shows the limited likelihood of

obtaining information representative to the truth from only a two-dimensional scan. Therefore, complex systems of scanning

and analysis have to be developed for better understanding of the normal histology and pathologic anatomy.

Figure 1: Illustration of typical hepatic segmental anatomy where segments are indicated with Latin numbers. Portal vein in green and the vena-cava in blue receiving the main hepatic veins in blue as well.

ii. Assessment of the hepatic functions

a. Biochemical assessment

The liver is a unique organ from several perspectives. Apart from the complex anatomy that was only revealed in the 1950s, the complexity of its functions makes it difficult to test. The hepatic functions depend on an equilibrium among the blood supply, architecture, and intrinsic functions as simplified in Figure 2. The liver has various functions that are broadly classified

into synthetic, storage, detoxification, and excretory functions. The relevant matter to this manuscript is the lack of a single

summative laboratory test that could be useful in depicting the hepatic dysfunction (28).

Therefore various methods are concurrently used in the assessment of the hepatic functions. The traditional system proposed

by Dr. Child and Dr. Turcotte from University of Michigan (29) was modified by Dr. Pugh (30) and has remained in use until

today.

In essence, the Child-Turcotte-Pugh (CTP) scoring grades the different major hepatic functions (efficiency of portal flow via

the presence of ascites, detoxification via the presence of encephalopathy, the excretory function via the serum bilirubin level,

and the synthetic function with prothrombin and albumin levels) in one system. This system scores high as a prognostic index

for mortality following interventions for portal hypertension, albeit, superseded by a newer model of end stage liver disease

(MELD score) (31). Nonetheless, the MELD score did not show superior performance over the CTP in the setting of liver

transplantation (32). This demonstrates the complexity of the assessment process and confirms that there currently is no one

system that could be used solely for as a prognostic marker for the different procedures or disease conditions. Of note, the

traditionally known hepatic functions tests (AST and ALT) could be of diagnostic value in some conditions (33) but not of

prognostic value (34). More biomarkers come into practice with clinically relevant value in the prediction of outcomes (35).

Figure 2: Simplified scheme showing the components necessary for integral hepatic functions assessment and some examples of their common methods of assessment. Each of these methods could express more than one function.

A more advanced set of tests was deployed over the years aiming at a more precise assessment of the liver function in a

quantifiable manner suitable for the modern complexity of disease and treatment (36). These tests use at least one hepatic

specific substrate that is cleared from the circulation by hepatic uptake and excretion, metabolism, or both (37). These

substrates could be exogenous (e.g., indocyanine green), natural (e.g., caffeine), or endogenous (e.g., bile). This implies that

the measure of a hepatic function using these substrates will depend on the hepatic clearance capacity determined by the hepatic flow, the efficiency of extraction, and the hepatocytes metabolism capacity.(36).

This will naturally classify the hepatic clearance rates of substances into high and low. Thus, those with high extraction – the indocyanine green – could be sensitive to the changes in the hepatic perfusion (38,39) as opposed to those with a low clearance which could be more susceptible to the metabolism (40).

b. Breath tests

The fundamental principle of breath test based assessment of liver functions is that an ingested substrate is metabolized in the liver and exhaled via the respiratory system (41). 13C-phenylalanine (PheBT) and 13C-galactose (GBT) breath tests

noninvasively assess the hepatic function by measuring the activities of two enzymes that are localized to the hepatocellular cytosol (41).

A 13C-methacetin elimination cytochrome P45 breath test was used in a LiMAx test to assess the maximum hepatic function

(primary non-function) in a transplantation setting with a sensitivity and positive predictive value of 1 (42). The test ability to

predict the outcome of liver resection was also investigated and have shown promising with possible risk stratification ability

as well (43). The LiMAx test showed superior predictive value – of morbidity and mortality – to the Indocyanine green plasma

disappearance rate (ICG-PDR) in a setting of sepsis related hepatic dysfunction (44).

c. Radiological based assessment of the liver function

The role of imaging in the evaluation of hepatic function is growing. A combination of SUV mean from a PET/CT and the

AST/Platelet index (APRI) showed a high accuracy in predicting the hepatic insufficiency in patients who underwent liver

resection (45). The MRI scans showed potentials for not only the global assessment but also for segmental functional

assessment based on the signal intensity in hepatobiliary phase (46). Different techniques were utilized in the MRI based

imaging to obtain a better sensitivity as in the Volume-assisted MR relaxometry technique, which showed better performance

than MR relaxometry (47). It seems that the dynamic contrast studies could have potentials in the non-invasive quantitative

assessment of hepatic functions (48).

d. Indocyanine green plasma disappearance rate (ICG-PDR)

The role of indocyanine green in the evaluation of liver function was developed in the past century (49). The invasive serial

measurement of the plasma disappearance rate remains the gold standard for measurement despite the development of newer

tools for both invasive and non-invasive measurement (49).

Clinically, the ICG-PDR has an established role in the assessment of hepatic functions in various settings (49). It has a value in

the evaluation of the intrinsic hepatic uptake function in cirrhosis as the primary determinate of the reduction of the plasma

clearance rate (50). Besides, it has also an established role in the preoperative assessment and indication of surgery algorithms

in cirrhotic patients (51), as well as in non-cirrhotic patients (52).

The indocyanine green is an organic anion dye that is exclusively eliminated by the liver. In human livers, the sinusoidal

transport is mainly mediated by the organic anion transporting polypeptide (OATP), whereas canalicular efflux is mediated by

the multidrug resistance associated protein (MRP2) and the multidrug resistance P-glycoprotein (MDR3) (53).

The ICG-PDR provide a promising global assessment of the hepatic perfusion, uptake, and excretion. However, the

contribution of the hepatic uptake and excretion mechanisms leading to disappearance rate is poorly understood in the in-vivo

situation owing to the lack of information on the temporal profile of the bile appearance rate (54). Nonetheless, the usefulness

of the indocyanine green elimination rate in assessment of the hepatic functions – not only the perfusion and excretion

parameters – is related to competitive inhibition by the hyperbilirubinemia state resulting in the assessment of some of the energy based hepatocytes activity (55).

Mathematical models were deployed to discriminate the contribution of the different compartments – sinusoidal flow, hepatocellular uptake, and biliary excretion and their reflection on the overall hepatic functions (56). Despite the plausible understanding that these models provide, their primary role remains theoretical and perhaps in applied technology for software applications. Beyond that, the fluorescence characteristics of the ICG were found to be clinically useful in the assessment of the hepatic function (57). It was also used for the detection of tumors (39,58) and biliary surgery (59).

Currently, there is a need for an optimized in vivo assessment of the multi-parameter hepatic function, which could potentially be provided by an optimized use of the fluorescence characteristics of the indocyanine green (38).

e. Evaluation of the hepatic architecture

The most obvious example of the importance of the integrity of hepatic architecture to the function is seen in liver cirrhosis

(60). The bridging fibrosis will reduce the functional sinusoidal surface area and creates a variety of intra- and extra hepatic shunts as presented in Figure 3. There are several structural and functional changes observed in colorectal liver metastasis without chemotherapy, which ultimately lead to a degree of vulnerability of the parenchyma (61). It was noted that the development of sinusoidal obstruction syndrome – irrespective of the cause – significantly leads to hepatic function impairment and at least partial mechanism is attributed to the structural alteration (62,63) and makes it susceptible to posthepatectomy failure (64). This leads to change in the operative strategies and increased the minimum safe remnant liver volume (65).

Figure 3: Schematic representation of the microarchitecture in healthy (top) and Cirrhotic (bottom) livers. The functional

sinusoidal area is significantly reduced with the rearrangement of the lobule. H.A: hepatic arteriole, P.V: portal venule, H.V:

Hepatic venule, F.S: functional sinusoid, C.S: capillarized sinusoid, I.H.S: intrahepatic shunt, and E.H.S: extrahepatic shunt.

Liver biopsy remains the gold standard evaluation tool of the hepatic architecture (66). Controversies, however, exist on the

definition of the adequacy of samples and the representation of the scoring systems in use as these scores are mostly qualitative

(67). Moreover, it is not complication free (68). Collagen proportionate ratio is an example of quantitative assessment

modifications implemented in the evaluation of liver fibrosis. There is potentially significant inhomogeneity in the hepatic

architecture in the two-dimensional analysis (66) for which a three-dimensional analysis was utilized (69) and showed

reproducibility but also demonstrated the important regional variability of the microarchitecture (26).

Novel invasive tools have also been tried in an attempt for in-vivo live analysis of the hepatic architecture. The confocal needle endomicroscopy showed a reasonable discriminative ability to distinguish cirrhosis from normal architecture in animals with no significant role in a clinical setting till now (70). Less invasive biopsy (the transjugular) was used as an alternative to the percutaneous biopsy aiming to reduce the potential complications – particularly in ascites – with a comparable diagnostic yield, albeit with a similar rate of morbidity and mortality (71). Noninvasive tools have been introduced to assess the alteration

in microarchitecture as alternatives to the invasive tools (72). Several serological markers were used in an attempt to estimate

the degree of fibrosis (73). Some examples are listed in Table 1.

Table 1: ‘Indirect’ serological markers for the prediction of liver fibrosis

Name Parameters Sensitivity/Specificity

PGA index Prothrombin index, gGT, 91/81 (74) apolipoprotein A1

PGAA index Prothrombin index, gGT, 79/89 (75) apolipoprotein A1, a2-macroglobulin

Fibrotest a2-Macroglobulin, haptoglobin, g- 75/85 (76) globulin, apoliprotein A1, bilirubin

Forns Age, platelet count, gGT, cholesterol 94/51 fibrosis index (77)

APR index AST/Platelet ratio 89/75 (78)

GPR index GGT/Platelet ratio 79/58 (79)

f. Imaging modalities:

Fibroscan – which is an ultrasound based elastography – is perhaps one of the most widely studied imaging modality to assess

the hepatic fibrosis (80). The combined use of fibroscan and fibrotest is shown to estimate similar value to the liver biopsy in

assessing fibrosis (81). Interestingly, our team demonstrated the clinical utility of this modality in the prediction of post-

hepatectomy liver failure (82). Assessment of the microarchitecture using in-vivo fluorescence imaging showed promise in evaluating the reperfusion injury of the transplanted grafts but yet to be quantified (39). Ex-vivo fluorescence based photon

microscopy 3D images are also useful for the quantification of the architectural parameters (83).

Diffusion weighted MRI was used to assess the hepatic microarchitecture and liver fibrosis (84). Ultrasound spectrometry was

also used to characterize the tissue ultrastructure (85). Novel methods, utilizing the contrast enhanced ultrasound, were

developed to visualize the hepatic vascular architecture in-vivo with accuracy (86). However, none of these tools was yet taken

into the clinical setting of surgical resection. The relatively new field of the assessment of the microarchitecture is the

mathematical and computational models (87).

g. Evaluation of the hepatic hemodynamic

The clinical significance of the assessment of liver hemodynamics has been a subject of extensive research and ubiquitous

correlation (88). Various methods – invasive and non-invasive – have been proposed to assess the hepatic hemodynamics.

Interestingly, a substance like the D-sorbitol was found to have a higher hepatic clearance compared to indocyanine green with

an elimination rate that is closer to the sinusoidal plasma flow (89). These findings were reproducible in volunteers with

temporary, drug induced reduction in the hepatic circulation with high accuracy (90). As mentioned above, most of the

substrates used for assessment of perfusion are useful for the assessment of the hepatic perfusion with different degrees.

h. Global live assessment of the hepatic perfusion, architecture, and excretory function

The majority of the above-described assessment methods cover one or more of the essential parameters in assessing the hepatic

function. However, none of them is suitable for an across parameter evaluation in an instant live situation.

Due to the high complexity of the liver functions, it is currently not feasible for a single test to precisely assess the entire set of

functions. Substance clearance rates remain representative for the dynamic global assessment of hepatic function without

consideration of the individual processes involved in the clearance, and therefore a clinically meaningful margin of error

persists in the available methods (91). The main reason for which no predictive model has a universally good performance is that all there models do not have the capacity to assess the complex function. However, there is a general agreement that the assessment tool could be either dynamic or static as simplified in Figure 4 (55). Moreover, assessment of the segmental function of the liver was found to be helpful to avoid overestimation of the remnant potentials to recover (92). It was demonstrated that the dynamic function tests are more reliable in the prediction of the postoperative course compared to the static tests (93). Functional assessment of the Asialoglyprotien receptor was found to have an excellent prediction of the postoperative course using 99mTc-GSA SPECT/CT, with a sensitivity and specificity of 100 and 92%, respectively (94).

However, a combination of tests is the common and the standard of practice for the time being (95).

Figure 4: Simplified scheme demonstrating the clinically relevant hepatic function assessment tests. Adapted from De Gasperi

2016 (55).

iii. Surgical disorders of the liver

The surgical disorders of the liver can be classified into disorders treatable by resection of some form, replacement via transplantation, and vascular procedures. Oncologic liver resection is one of the common procedures on the liver (96,97), and it

is the one of concern to this manuscript. Hepatic tumors are diverse in their pathophysiology and the disease process.

Traditionally classified as benign and malignant, the latter is classified as primary – this includes tumors of parenchymal or

mesenchymal origin – and secondary tumors, which include metastasis from other sources as ovarian, colorectal, pancreatic

sources (98,99). An important, influential factor on the outcome is the background status of the liver (100). This is well known

as a predictor of outcome in a variety of diseases (101).

For this manuscript, it seems appropriate to classify the surgical intervention based on the background liver condition. This

classification seems technically relevant as the minimum safe amount of remnant liver is different in those with normal

background from those with the cirrhotic background (102). Indeed, there is a significant contrast to the role of the remnant volume in the development of post-hepatectomy liver failure in the cirrhotic population as opposed to the non-cirrhotic patients

(103), albeit the general agreement that the residual volume is larger in the pathological background liver tissue than normal livers (103). In patients with preoperative chemotherapy, the residual liver volume is required to be greater than in those without chemotherapy – due to the hepatotoxic effects of the chemotherapy on the liver – which necessitates the importance of preoperative remnant volume induction (104).

The reason for which the remnant liver volume plays a pivotal role in the decision of operation is that it is important to obtain a radical resection while preserving the remainder liver function to support the metabolic needs of the body. If this balance is disrupted, it is either that the remnant liver will not be able to support the host and death will be the result of the intervention

(105) or that the radical resection will not be achieved and hence the intervention would be potentially a futile one (106).

It is thought however that it is not only the absolute liver volume that matters but its function as well (101,107). That is one of the reasons for which the critical remnant volume is significantly larger in cirrhotic compared to liver received chemotherapy

(108) or non-cirrhotic livers (109). The presence of clinically significant portal hypertension secondary to cirrhosis was a poor prognostic factor after liver resection (110). Various methods – including the assessment of indocyanine green elimination rate at 15 minutes in relation to the standardized future liver remnant (109) – were used as integrated parameters to estimate the

quality of the hepatic parenchyma. This indicates that the architectural integrity is also an important parameter for the

determination of the post-resection outcome (95).

Most of the metastatic tumors of the liver take place on a normal liver background (111). It is also more common to be

multiple causing significant disease burden on the liver and was often non-resectable (112).

iv. Role of chemotherapy in colorectal liver metastasis:

The initially irresectable liver metastasis was considered a palliative disease until the advent of effective chemotherapeutic agents (113). There has been a substantial change in the criteria of palliative disease over the past few decades (114). Today, there is no doubt on the value of the chemotherapy in the treatment of colorectal liver metastasis. The common classes of the anti-cancerous agents are summarized in Table 2

Table 2: The different common anti-cancerous classes in clinical use. Adapted from (115)

Class Example Mechanism

Alkylating agents Platinum (cisplatin, oxaliplatin), Nitrogen DNA damage, not phase specific mustard derivatives (chlorambucil), Nitorsurea (carmustin)

Antimetabolite Pyrimidine inhibitors (5-FU, Gemcitabine, DNA/RNA replication in S phase of cell Capecitabine), methotrexate, 6- division mercaptopurine

Anthracycline (Antitumor-Antibiotics) Doxorubicin, Bleomycin Interfere with enzymes needed for DNA replication

Isomerase Inhibitors Irinotecan Interfere with enzymes required for DNA replication

Mitotic inhibitors Taxanes, vinca alkaloids (vincristine, Plant alkaloids that inhibit mitosis in tumor vinblastine) cells

Hormone therapy Tamoxifen Alter the action and production of hormones

Immune therapy rituximab, Thalidomide, BCG, Alpha Stimulates immune system to recognize and interferon, interleukin2 attack cancer cells

Differentiating agents Retinoids Act on cancer cells to make them differentiate into normal cells

Targeted therapy Imatinib-Gleevec, gefitinib Iressam More precise targeting of the cancerous erlotinib Tarceva, sunitinib Sutent and cells at various gene and receptor levels bortezomib

Steroids Prednisolone Decreases inflammation and enhance effectiveness of other drugs

It has not only improved the resectability (116) but also improved the survival (106). It is now foreseeable to consider a cure status for patients who had an R1 resection of liver metastasis from colorectal cancer attributed to the advent in chemotherapy; agents and regimens (117). The most common agents are summarized Table 3.

Table 3: Common chemotherapeutic regimens for colonic cancer.

o 5FU and folinic acid Ø The 5FU is an antimetabolite, and its action is potentiated with

(Leucovorin) leucovorin

o 5FU and oxaliplatin. Ø When combined with 5FU and folinic acid it is known as the

FOLFOX regimen.

o Capecitabine. Ø Is an oral prodrug of 5FU and is used in patients with metastatic

disease or patients with palliative intent therapy

o Irinotecan. Ø May be used in combination with 5FU and folinic acid, the FOLFIRI

regimen

o The monoclonal Ø may be useful for patients with wild type K RAS. It is either given

antibody cetuximab alone or together with irinotecan

v. Effect of chemotherapy on the liver functions:

The hepatotoxicity of chemotherapy has been well recognized for decades (118). Irinotecan-based regimens are associated

with the development of chemotherapy associated steatohepatitis (CASH) in patients with colorectal liver metastasis,

independent of their body mass index (119). One of the famous complications of the chemotherapeutic agents in this setting is the sinusoidal obstruction syndrome (SOS), which is a sinusoidal injury characterized by damage to sinusoidal endothelial cells (SECs) in association with hepatocellular atrophy, disruption of the sinusoidal membrane, sinusoidal dilatation,

collagenization of the space of Disse, centrilobular or perisinusoidal fibrosis (120). The common hepatic adverse effects of the chemotherapeutic agents are summarized in Table 4

Table 4: Examples of the known hepato-specific adverse effects of the commonly used chemotherapeutic agents. Adapted from (115).

Agents Pattern Features

Oxaliplatin, 6-MP. Dacarbazine, Sinusoidal obstruction Microvascular injury (blue liver) cyclophosphamide, and syndrome (SOS) vincristine

Gemcitabine, FOLFOX Pseudocirrhosis Radiological features of macronodular cirrhosis irinotecan, oxaliplatin, and Steatosis Fatty liver and increased cetuximab, tamoxifen transaminases

Cisplatin Vinblastine Rituximab Acute hepatitis Significant rise in transaminases and jaundice

Mithramycin, L asparaginase, Hepatic necrosis Liver failure methotrexate

There are several other potential morbidities related to the administration chemotherapy (121). Interestingly, mortality from chemotherapy is not common (122). There are however, several reports indicating that the outcome of surgery is not adversely affected by certain chemotherapeutic regimens (123). This highlights that the balance between toxicity and treatment is necessary to maintain.

Progression on chemotherapy is another aspect that needs consideration (124). Notably, the disappearing lesion following chemotherapy is another challenge encountered in this category of patients (125). The change of the radiological appearance of the hepatic parenchyma and the tumor response to chemotherapy – whether partial or complete – makes the detection of the some of the previously seen lesions difficult.

Intraoperative ultrasound and contrast-enhanced intraoperative ultrasound seem to have the best pick up rates for these lesions

(125,126). The treatment of such lesions is an ongoing subject of debate (127,128).

vi. Expanding the criteria for liver resection in colorectal liver metastasis:

Only 10-20% of patients who present with colorectal liver metastasis are resectable at presentation (129). Surgeons have been

pushing the frontiers of the treatment of the hepatic malignancy assisted by the cumulative knowledge of the pathophysiology

of the disease process and the advances in technology and oncotherapy. Previously irresectable colorectal liver metastasis has

become resectable based on the core concepts of stimulation of liver regeneration via portal flow diversion (108).

a. Portal vein embolization and ligation

It was more than a century when Rous and Larimore described the interesting phenomenon of ipsilateral liver atrophy and

contralateral hypertrophy of the liver on occlusion of a portal vein branch (130). Similarly, an observation by Bax in the fifties

documented a similar effect of ligation of the bile duct (131). The powerful clinical application of this concept observed

following the publication from Takayasu et al. in patients with hilar cholangiocarcinoma with portal vein stenosis (132). This

observation led to a clinical application on 14 patients with the same disorder, and the results were encouraging and reduced the morbidity of the major hepatic resection (133). Several years later the same group reported their experience with preoperative portal vein embolization in patients with HCC on cirrhosis (134), which also showed a favorable radiological – regarding volume gain – and clinical outcome. Nonetheless, the volume increase in this series was lower than the reported one

in patients with normal background liver (135). Currently, preoperative portal vein embolization is one of the main stay interventions for pre operative stimulation of the remnant liver to gain an acceptable volume (136).

The primary indication of portal vein embolization is to stimulate the remnant liver growth to a metabolically sufficient volume to sustain the host. Various techniques were proposed, and it seems that the majority of these techniques have similar clinical value (137). The concept of segmental blood flow deprivation and diversion to the future liver remnant was taken even further to a sequential occlusion of the portal and hepatic veins, where the synergistic effect was reflected on enhanced volume gain after the following procedure compared to the portal vein embolization alone (138).

Extensive animal models were developed to examine the different indications and mechanisms of the portal vein embolization, and the choice between large and small animal models depends mainly on the objective of the study and whether it is concerned with volume assessment or molecular investigation, respectively (139).

b. Tumor progression after portal vein embolization in patients with colorectal liver metastasis

The relative advantage of the portal vein embolization in the context of extensive tumor burden is related to enabling a radical tumor resection after stimulation of an otherwise small liver remnant to attain a sufficient volume. This could be jeopardized by the associated tumor growth rendering potentially resectable disease irresectable (140). In a recent analysis of the published literature, 30% of patients previously indicated for resection following portal vein embolization were rendered irresectable mostly due to tumor progression (136). As there is a minimum waiting time to radical resection with a risk of tumor progression, Chemotherapy as a bridge to surgery has been suggested (141). Ultimately, large series of portal vein embolization have demonstrated post-hepatectomy liver failure of 2.5 % in those who proceeded for liver resection (104).

Portal vein embolization increases the tumor growth rate by more than seven times (142), with a tumor doubling time significantly shorter than the natural history of these tumors (143). Despite that there were new lesions discovered in the remnant liver in at least one study (142), it is not clear if this is attributed to the waiting time between the portal vein

embolization and surgery or due to the stimulus that follows the embolization. It could be, however, that these tumors were residing in the remnant liver not detected by preoperative imaging and the portal vein embolization has stimulated their growth. This hypothesis is supported by the long residing time extrapolated from mathematical simulations (144), which shows that tumors reside for nearly a decade before discovery. Moreover, there was an earlier recurrence in patients who underwent resection who had tumor progression following portal vein embolization (145). Increase in tumor size is related to an increase in the tumor shedding and circulating tumor cells (146). Interestingly there was an increase in the circulating tumor cells following non-surgical procedures compared to surgical resection, which led to an immediate reduction of the circulating

tumor cells (147). Despite that, the latter study did not include portal vein embolization, yet the concept is still valid to apply.

The role of circulating tumor cells as a prognostic factor for early recurrence after surgery is unclear and was found by

Lalmahomed et al. (148). On the other hand, Brudvik et al. found that patients with circulating tumor cells have a significantly higher recurrence rate compared to those with negative circulating tumor cells (149). It seems that if the circulating tumor cells do not disappear after resection, the probability of recurrence is higher (149).

To overcome the problem of tumor growth in the remnant liver, clinicians established the role of different other modalities as a bridge to curative resection. Combined portal vein embolization of the resectable liver and radiofrequency ablation of the tumors in the remnant liver demonstrated promising approach (150).

c. Two stage resection

In the setting of bilobar metastasis, a preoperative portal vein embolization is inadequate due to the tumor progression mentioned above. A two stage approach was designed to clear the remnant liver from visible tumors in association with portal vein ligation of embolization was found useful (116).

d. ALPPS as a variant of the two-stage hepatectomy and variation

The associated liver partition and portal vein ligation (ALPPS) is a relatively new procedure which is developed to overcome

various problems associated with both the two stage liver resection and preoperative portal vein embolization in the context of

bilobar colorectal liver metastasis in the setting for small for size remnant (151). The technique has undergone many variations and different approaches were used to perform it, and it gained wide popularity at its introduction (152). The relatively simple technique, the extended potential for radically resecting the hepatic disease, and the significantly shorter waiting time between the first and the second stages, were all favorable factors for the ALPPS. However, this is hampered by the significantly higher inpatient mortality and morbidity of the procedure compared to the standard two stage procedure (153). The true potential and disadvantages of this operation are not yet fully explored, and it seems that the five year survival could be a good indicator of the usefulness of this technique, which is still awaited (154).

One of the major histologic differences in the regenerating remnant liver is the immaturity of the hepatocytes in the ALPPS technique compared to portal vein embolization (155).

e. Single stage resection

An appealing alternative but not widely adopted due to the high skill set required is the single stage liver resection for bilobar

colorectal liver metastasis. This approach is largely based on the careful use of intraoperative ultrasound (156). The idea is to

use the ultrasound for visualization of the tumors being resection and the structures being conserved in a way that preserves the maximum parenchyma while a radical resection of all tumors is achieved in a single stage (157). The outcome of this approach in qualified hands is superior to the two stage hepatectomy regarding morbidity with a similar oncologic outcome (158).

This study aimed to simulate the extensive resection of livers with the normal background liver scenario, which is close to the clinical situation of the colorectal liver metastasis, therefore, the normal liver background state will be our focus of discussion in the next sections.

vii. Liver regeneration

The hepatic surgery is driven by the remarkable ability of the liver to regenerate, which is thought be known from the ancient ages (159). A demonstration of the myth by Michelangelo in (Figure 4) highlights the ancient notion of the liver potentials to regenerate.

Figure 5: Tityus (c.1532), Michelangelo Buonarroti, Royal Collection, Windsor Castle, U.K. In contrast to the original myth according to which two vultures attack Titus, Michelangelo erroneously painted only one. (Copied with permission (159))

The stimulus for liver regeneration comes through various inciting factors such as toxicity, surgery, and blood flow diversion

in portal vein occlusion by either ligation, embolization, or even as a result of interruption of the blood flow due to tumor

invasion of portal branches.

The regeneration seen after significant hepato-toxicity is different from the perspective that the architecture is preserved.

However, there are generally three phases of liver regeneration:

a. Phase I: The priming:

The quiescent hepatocyte replication rate is quite low; 1/20,000 (160). A breakdown in the cellular bonds and extracellular

matrix digestion starts the priming secondary to several stimuli (161). This priming is initiated by cytokines signaling and it

leads to increase in a replication of cell cycle frequency. There is evidence for the role of lipopolysaccharides from the gut

microbiota in initiation of the regeneration process (162). The molecule activates the Kupffer cells to release interleukins and

tumor necrosis factors, which then and through a cascade of molecular activation, activates the hepatocytes transition from G0

to G1 in the cell cycle (163).

b. Phase II: The proliferation

The hepatocytes enter the S phase at around 12 hours, prior to the non-parenchyma cells at 48 hours (162). The DNA synthesis

was associated with this phase. Zonal proliferative differentiation was also noted in association with the proliferation (163).

These cells enter into replication and mitosis by not completely clarified mechanisms but the end results is around 1.6 cycles –

in average two consecutive cycles (162,163). The second wave of the DNA replication, which usually peaks at 48 hours is

usually in the centrilobular zone (164).

c. Phase III: The termination

The termination of the proliferation is not fully understood. It represents however the summit of the orchestrating process.

Interplay of the extracellular matrix, the pro-HGF, mTOR pathways, STAT3 pathways, and the down regulation of the cytokines pathways are proposed candidates of “steering the wheel” or “proliferation brake” (162,163,165).

Telomerase activity was found to be central in the process of liver regeneration and a five-fold increase in the enzyme activity

– compared to the quiescent liver – was observed in association with regeneration following resection (166). A large number of pathways and molecules contribute to the liver regeneration.

d. Role of architecture in the regeneration process:

There is an established role of components other than hepatocytes in the regenerating liver (167). Cholangiocytes lose their plasticity and differentiate into hepatocytes when required (168). The oval cells were found contributing to the regenerating liver (169). Hepatic stellate cells induced the differentiation of the hepatic stem cells into hepatocytes when co-cultured in the laboratory (13). The non-hepatocytes were also involved in priming the regeneration by acting as a reservoir to the hepatocyte growth factor and mediating the interaction on the extracellular matrix and the cytokines pathways (170). Kupffer cells have on the other hand both stimulatory and suppressive effects on the regeneration and their depletion led to augmentation of proliferative activity, in vitro (171).

viii. The relation between hemodynamics and regeneration after resection

Theories around the mechanism by which the liver regenerates following resection or portal vein occlusion are to an extent

converging and allude to the pivotal role of the shear stress as a stimulus to cellular proliferation. There is however counter

evidence that the role of portal shear stress is itself the sole factor, as highlighted later.

a. Necessity for increase in portal venous pressure/flow per unit volume

There is a cumulative knowledge on the role of portal hemodynamics and liver regeneration (172).

The immediate increase in portal pressure was found to correspond to be the stimulating factors of liver regeneration (173).

Increased portal pressure or flow per remnant liver volume is necessary to induce hepatic regeneration. In a study of molecular

and DNA changes following partial hepatectomy in rats, there was a lack of activation of hepatocyte growth factor in the group

with surgically porto-systemic shunt (that prevents the increase in the portal flow following resection) versus higher activation

in the group without shunt (174). Similarly, the apoptotic index was higher in the group with shunts (174). Interestingly, the

shunt group showed narrower sinusoidal spaces (174). This indicates that an increase in the portal pressure is required to

stimulate the liver regeneration. Several other studies have shown that there is an increase in the proliferating cellular nuclear

antigen (PCNA) in animals following the portal vein ligation (175).

b. The role of portal vein arterialization and the theory of sinusoidal shear stress versus portal venous contents:

Li et al. demonstrated that portal vein arterialization in a rat model of 68% resection resulted in significant improvement in the

regeneration rate as well as in the liver function represented in higher albumin levels with no difference in survival at 28 days

(176). Fan et al. have ascertained the findings of other studies in that the portal venous blood is required for initiation and

maintenance of liver regeneration and they also supported the role of the total arterialization of the portal vein in maintaining

liver regeneration in a rat model at the same resection percent (177). They also found that replacement of the portal inflow with

systemic venous flow leads to important suppression of the regeneration with failure to recover at the end of their experimental

time line (177).

Shimizu et al. have also identified the beneficial effect of portal vein arterialization on the hepatic regeneration rate and end

point hepatic volume with improvement in the hepatic metabolic profile owing to the increased oxygenation of the portal vein

(178). They have reported that the portal venous pressure after arterialization is higher than before arterialization but have not changed significantly after 10 days (178). Interestingly, their experimental setting was on a >80% resection (178).

In reality this could still be confusing giving that this resection would increase the portal venous pressure and flow per unit

volume and the superadded arterialization effect should in theory result in suppressive effect of the regeneration (179).

In a different setting and experimental design, Mortensen et al. constructed a controlled arterial shunt – contrary to the above studies where there was no control over the amount of flow in the arterialized vein – to the left portal vein to deliver four times the estimated normal portal flow to the target segments via an aortic graft without liver resection (180). They reported a preferential increase in the segments without shunts. Of more interest, they demonstrated that the arterialized lobules were smaller and condensed with different molecular and genetic expressions compared to the shame animals (180). The main differences in design between this study and the above ones is that the stimulatory effects of the resection was not elicited

(181,182). Of note, the arteriolar resistance sites are pre-sinusoidal while the resistance sites for the portal veins are likely to be postsinusoidal (183). Several studies have demonstrated that deprivation of the liver from the portal venous flow leads to atrophy of the deprived segments (184).

c. Small for size/flow syndrome

Definition:

The initial hypothesis of the syndrome of post-transplant liver failure as being attributed to a mismatch between the graft size and the recipient size has not been shown a plausible definition giving the weak predictability of the outcome (185). The plethora of evidence point towards the pivotal role of the hepatic hemodynamics alterations in the development of the posthepatectomy or post partial liver transplant failure (186,187). Recent interpretation of the literature has suggested defining the syndrome as clinical condition represented in hepatic insufficiency as a result of sinusoidal barotrauma (188).

Definitive diagnostic/prognostic criteria are lacking for this syndrome. The most famous “50-50” (189) criteria appears to be consistent and was validated in a prospective study by the same group (190) and it out performed the criteria proposed by the

ISGLS in a different study (91). Different grades were also proposed with different management strategies (191).

The portal hyperperfusion was recognized as a central mechanism in the development of hepatic failure (192). Three portal hemodynamic parameters were identified to interact with the development of small for size syndrome, a syndrome that could

be better understood based on the hemodynamic changes rather than mere size (186). These are the absolute portal venous pressure, the porto-caval pressure gradient, and the flow per unit of remnant volume.

Dearterialization is another contributing element to the pathophysiology of the small for size syndrome (193). The dearterialization is described as an exacerbated reduction of the arterial inflow as result of the excessive increase in the portal flow (193). As a compensatory strategy, Ozeki et al. reported improvement of the dearterialized liver after hepatectomy (194).

The relation between the cellular proliferation index and the development of liver failure was investigated in a small clinical series as well in experimental settings where findings are not converging smoothly. There was an increased in the Ki67 staining in patients who developed small for size liver grafts compared to those who survived (195). The syndrome in this series was associated with centrilobular cholestasis indicating that other functional or structural mechanisms contribute to the development of the hepatic failure (195). The regeneration – tested by the BrdU activity – was found to be suppressed in the presence of a small for size syndrome compared to non-small for size livers (196).

The liver-gut axis is now recognized by different interactions with the hepatic gene expression in various stages of liver

regeneration (197). The axis has various inflictions on the occurrence of liver injury (198). Alteration of the gut microbiota has

a direct influence on the liver regeneration (199). However, this area is still primitive in knowledge and clinical applications.

d. Limits of the safe increase in the portal venous pressure/flow

The previous two sections addressed the necessity of the portal venous flow to increase in order to stimulate the hepatic

regeneration following resection and also the hazardous uncontrolled increase in that parameter that would lead to hepatic

failure. In this section, we aim to shed the light on studies addressing the definition of the safe limit of the increase in the portal

flow per unit volume and/or the pressure parameters that safeguards against the development of liver failure.

Our group have demonstrated a critical value of the absolute portal venous pressure of 20 mmHg (200). The porto-caval pressure gradient was found more representative to the transhepatic shear stress as it takes into account the alteration in the

systemic circulation and was found to better correlate with the transplant outcomes (201). The hepatic venous pressure gradient

(HVPG), which is an equivalent to the portocaval pressure gradient but measured via an occlusive wedge pressure catheter

through the suprahepatic veins (202,203), was found to be an important parameter in the evaluation of the hemodynamic

adequacy and prediction of the graft outcome in a transplantation setting (204). Our group have also validated the role of the

portal blood flow per unit remnant mass (205).

e. The portal flow modulation:

· Surgical modulation:

The portal flow modulation is a term that describes a group of interventions that aims to adjust the portal inflow to the remnant

liver volume. All of the studies mentioned under the necessity of increase in portal pressure/flow are essentially an expression

of the effect of portal flow modulation on the liver regeneration, whether negative or positive. This section is concerned mainly

with exploration of the studies with controlled flow modulation.

Partial portal vein ligation – to reduce the portal flow by 40% of its normal – in a model of partial liver resection was

associated with a reduced regeneration and remnant volume compared to the control group (206).

The barotraumatic effect of the increased portal flow to the small graft and the deleterious excessive reduction of the portal

flow by means of uncontrolled shunts was reduced by the deployment of a controlled shunt that decompresses the portal bed to

limit the increase in the portal flow to around twice the baseline (207).

Splenectomy or splenic artery ligation would reduce the portal flow by reducing the splenic venous reduction through the

portal vein and also contribute to alleviate the dearterialization effect through redirecting the blood flow to the hepatic artery

(208). The role of splenic artery embolization has been confirmed in other studies (209).

The notion of decompression of the portal flow via shunts was experimented clinically with success (210). However, the main

adverse effect of this type of shunts is the non-reversibility that might lead to a drastic atrophy of the graft (211).

· Non-surgical modulation:

Octreotide and esmolol infusion in a potentially small for size transplanted liver grafts have improved the hepo-portal venous

gradient and reduced the portal flow and despite that this has not improved the ICG-PDR, it improved the portal blood

oxygenation (204). This was validated by different groups (212). In contrast, Olprinone – a phosphodiestrase III inhibitor that

acts via reduction of the portal venous pressure – led to improvement of not only in the histological and hemodynamic

parameters but also in the ICG-PDR (213).

Other molecules have shown promising in alleviating the clinical consequences of the syndrome such as the granulocyte

colony stimulating factor (G-CSF) (214). Nitric oxide infusion was found to improve the hepatic arterial flow reducing as such the dearterialization component of the syndrome (215).

ix. Mathematical modeling and their role in liver research

The advances in technology and the accumulating knowledge on liver physiology and anatomy in addition to the complexity of

live modeling, makes it a fertile land for quantitative mathematical research (216).

Various research experiments were designed to address particular translational settings as summarized in Figure 6. Our

experimental design – as detailed later – is concerned with the resection setting.

Figure 6: Illustration of the various experimental models of liver regeneration. (Copied from Rychtrmorc (163) with permission)

II. Methods

i. Rationale of the layout

The iFLOW is a huge project due to the multi-scale nature of the research and integrated questions cascade that was addressed within the duration of investigation. It was a difficult task to gather all the pieces together to highlight this enormous work in a concise fashion. I thought that the layout should be meaningful in itself and as much as I could, I tried to make the writing flow easily despite the complex content.

The manuscript flows from an introduction that targets the display of a common background knowledge and the extraction of the common methods from the individual research questions to prevent repetition and inflation of the text. However, it was deemed necessary to keep a certain minimum that is required for the read of the text in each section. Otherwise, it would have been very difficult for the reader to grasp the idea, while reading the highlights of each research question. Towards that end, the subsequent text was organized in a question based format, where each section is intended to address a particular problem.

This layout is based on the fact that each section represented a research question in itself.

By the time of disclosing this manuscript, most of the sections would have been already published and all of them should have been submitted. Noteworthy, that I was not able to explore the full potentials of this project within the restrictions of the degree. Therefore, the work will not come to an end by the submission of this manuscript. The output from this project will continue and I will make sure to take it forward to the maximum production possible. I tried as much possible to document most the methods and protocols used and the findings of these studies for those who would carry on this mission later on, trying as much as possible to avoid loose ends and ambiguous context.

iii. Ethical approval

The study was approved by the regional committee of ethics of animal research, and by the French Government authorities, complying with the European Union Directive N° 2010/63/EU. The application for ethical approval is provided in the appendix.

iv. Funding

This research had several funding sources. The main funding source was by the “Agence Nationale de Recherche” through the

ANR-13-TECS-0006 grant for the project (IFlow). There was also an important contribution of the “Agence de la

Biomedecine” through its program of Research (AOR 2009) and the Virtual Liver Network (German Bundesministerium für

Bildung und Forschung [BMBF]).

v. Study settings

Surgeries in the first set of experiments were performed at the experimental animal surgical unit at the Marie Lannelongue

Center, Le Plessis Robinson, France. Surgeries and radiological investigations in the second set of experiments were performed at the CIRE plateform, INRA (Institut national de la recherche agronomique, Val-de-loire, France), an authorised center for animal research by the EU under the number E37-175-2.

Histopathological analysis was performed at the Leibniz Research Centre for Working Environment and Human Factors

(IFADO), TU Dortmund University, Dortmund, Germany. This group was concerned mainly with tissue processing for immunohistochemical staining and 3D reconstruction and quantification. The H&E staining and interpretation were performed at the Pathology Department, AP-HP, Hôpital Kremlin-Bicètre, Kremlin-Bicètre, France.

The pipelines and the software used for 3D quantification of the stained microarchitecture were developed at the INRIA Paris-

Rocquencourt, France and the Interdisciplinary centre for Bioinformatics (IZBI), University of Leipzig, Leipzig, Germany.

Mathematical models were built at the INRIA Paris-Rocquencourt, France. The percutaneously adjustable vascular ring was designed and manufactured in collaboration between the Centre Hépato-Biliaire, AP-HP, Hôpital Paul Brousse and MID SAS,

Dardilly, France.

vi. Description of the INRA

The INRA center represents one of the largest and most advanced centers in the center of France for experiments on large

animals. The CIRE platform (center de Chirurgie et d’Imagerie pour la Recherche et l’Enseignement) is the center of

surgery and radiology for research and education is fully equipped with advanced MRI 3T (MRI Magentum Vario 3T), CT

scan (Somatom, Definition AS, Siemens, Forchheim, Germany), C-arm X-ray fluoroscopy (), Ultrasound machine (), and Fiber

optic microscopy (CellVizio Dual Band, Munakea, France). Figure 7 and Figure 8 are photos taken for the experimental setup

at the INRA.

Figure 7: One of the operating theatres at the INRA with an operating table in the middle and instrumentations tables around the central operating table. The anaesthesia machine is located at the head of the table.

Figure 8: The equipment set up at the CIRE platform. From left to right, MRI, C-arm fluoroscope, CT scan, Ultrasound and Fiber optic microscopy at the bottom right.

vii. Study design

In view of the objectives; animals were allocated to the following groups; also in Figure 9:

· Group 1: composed of randomly allocated animals to undergo 75% liver resection with or without the application of the adjustable vascular ring around the portal vein. This group consisted of 17 animals, of them 8 animals underwent liver resection and ring placement and 9 animals underwent only liver resection. This allocation targeted the study of the safety and the efficacy of the vascular ring in ameliorating the hepatic function and regeneration following this type of hepatectomy.

· Group 2: consecutive series of 16 animals that underwent 75% resection with 7th day sacrifice protocol. This allocation targeted optimization of the invasive hemodynamics monitoring of the porcine model while conducting liver resection, which would allow building up accurate predictive mathematical models at multi-scale level and investigating the role of ICG in in- vivo evaluation of liver function during major liver resection.

· Group 3: composed of consecutively allocated animals to undergo 90% resection with and without the application of the adjustable vascular ring around the portal vein. This group consisted of 6 animals, of them 3 animals underwent 90% liver resection with the application of the vascular ring around the portal vein and 3 animals underwent only liver resection. This allocation targeted the study of the efficacy of the vascular ring in ameliorating liver function and regeneration following this potentially lethal liver resection.

· Group 4: consecutive series of 6 animals that underwent 75% liver resection with 3rd day sacrifice protocol. This allocated targeted the investigation of early volume and histopathological changes.

Each of the above mentioned group had a timeline for interventions and sacrifice that is summarized in the following flow chart Figure 9.

Figure 9: Flow chart demonstrating the workflow within the different iFLOW groups and the major tasks assigned to each time point.

viii. Housing and preoperative accomodation

Experiments were conducted on Large White pigs. All of them were females except one. The average age= 3 months ± 10 days and the average weight=35.3 ± 5 kg. Animals were brought to the experimental centers one week before invasive procedures in order to attenuate the influence of changing housing on the bio-physiological data. Animals were guarded in individualized cages with wide spaces between the bars between cages to allow communication between animals. Temperature was maintained around 27° C and checked with an over the wall thermometer. Nearly half of the cages` floors were covered with straw to provide mattress for animals to sleep and when required heating lamps were available. Lighting was natural and was available through close to ceiling wide windows.

Free access to water through water source as well as feeding plates was available at the entry of each cage. At the night of surgery, animals were allowed free access to water only and food was restricted overnight. Nonetheless, this practice was modified over the course of experiments and animals were allowed free access to water and food till surgery.

Right after surgery, animals were transferred to a dedicated recovery room covered with aluminum blankets till fully awake then were transferred to their cages in a dedicated room for the operated animals. This room has a matching set up to the one used in the preoperative housing. Animals were allowed free access to water and food after surgery. However, where intake was inadequate or was not commenced, replacement fluid therapy was given through the vascular access inserted during the surgical intervention. Animals were kept in these cages till sacrifice.

ix. Anaesthetic protocol

All surgeries as well as procedures requiring still animals were performed under general anaesthesia. On the day of surgery,

animals were given in their individualized cages 30 mg/kg ketamine (Ketamin, Panpharma) and 0.03 mg/kg acepromazine

(Calmivet, Vetoquinol, France) for sedation. Then animals were transferred to anaesthesia and recovery room for completion

of anaesthesia and securing airway.

Each pig received 100 mg of xylazine 2% (Rompun, Bayer Healthcare) with750 mg ketamine for induction of anaesthesia followed by tracheal intubation (6-7 mm in size, Portex, France). Subsequently, animals were transferred to the operating

theatre and inhalational anaesthesia was started using a 60% FiO2 with 2% isoflurane (Isoflurane, Belamont, France) in

assisted ventilation. The ventilator was set up on volume control mode delivering 350-400 ml at a rate of 17-20 cycles/minute.

During surgery; pancuroniumbromide (Pavulon, Schering-Plough) for muscle relaxation, at a rate of 0.3mg/kg/h and fentanyl

(Fentanyl Janssen 100µg/2ml) for pain control, at a rate of 5μg/kg/h were continuously perfused intravenously. Crystalloid fluids were given at a rate of 2ml/kg/h-fasting in addition to 500-1000ml, which was increased as required.

During surgery animals were covered with heat blankets and gastric aspiration through an oro-gastric 38 Fr tube was attempted if gastric distension was observed. Surgery was performed under sterile conditions and the vital parameters were monitored.

Cefotaxime 1g (Cefotaxime, Mylan) and gentamicine 80mg (Gentalline, Schering-Plough) were given intramuscularly once a day for 5-days. In addition, animals received Pantoprazole 40mg/i.v (Inipomp, Nycomed) and enoxaparine 0.2 ml/S.C

(Lovenox, Sanofi Aventis) and 0.5mg/kg/b.i.d Nalbuphine (Nalbuphine Serb, 20mg/2ml).

At the end of surgery, the wound was infiltrated with ropivacaïne 150mg (Naropeine, AstraZeneca, 7.5mg/ml). Blood samples were collected before and after liver resection as well as on the 3rd, 5th, and the 7th postoperative days.

The same preoperative measures were repeated at the time of sacrifice on the 7th postoperative day. Few modifications were adopted on the aforementioned protocol over the course of experiments. Muscle relaxation and analgesics were given in shots instead of continuous infusion during the experiments performed at the INRA. At the end of surgery a fentanyl transdermal system 75 mcg/h (Duragesic®, Janssen -Cilag Pty Ltd, Australia) was applied to the right lower chest, after shaving, for 72 hours. Pain control was complemented with acetaminophen (perfalgan). Noradrenaline infusion was administered in 4 animals in the newer experimental setting (INRA), where it was stopped thereafter due to the important circulatory disturbances manifested during the administration.

Anaesthesia for imaging and minimal invasive procedures (eg: percutaneous ultrasound guided liver biopsy) did not require muscle relaxation. The anaesthesia induction and maintenance as well as the anti-bio-prophylaxis continued to be similar to the operative protocol.

x. Radiological protocols

a. CT scan and Volume analysis

An abdominal CT scan was conducted in every animal. On average, 80 ml (2 ml/kg) of iodinated contrast (Omnipaque TM,

GE healthcare, Carrigtohil, Irland) was injected through an intravenous catheter with a rate of 4 ml/s. CT scans were performed with a Somatom (Definition AS, Siemens, Forchheim, Germany) (Figure 10). The CT image acquisition was performed in 4 contrast phases, arterial at 15 second of injection, early portal at 35 seconds, portal venous at 55 second of injection and at 75 second after the injection, the hepatic venous phase was done (Figure 11). Where indicated, the contrast phase was omitted to avoid renal overwhelming with iodinated contrast toxicity. This was only adopted in a few animals in the immediate postoperative scan or the first day scan.

Syngo.via software (Siemens healthcare global) was used in a cardiac analysis module to subtract mainly the hepatic artery and its branches. Upon launching the module, cropping of the volume; limiting the display from the diaphragm to the celiac trunk was performed. Then, the image display parameters were adapted to the hepatic arterial and portal venous branches (Figure

11). Image cropping was further performed from ventral and dorsal aspects to remove the parietal vessels if necessary.

Volume analysis was performed using the Myrian® XP-Liver 1.14.1 software (Intrasense, Montpellier, France). Segmentation of the portal venous branches as well as the hepatic venous branches was performed at first. Subsequently, the liver parenchyma was segmented (Figure 13). In order to segment the different hepatic sectors, each sector was segmented based on the supplying portal vein, the draining hepatic vein and the fissure boundary with the adjacent sector. Images were segmented most successfully in the 75 second hepatic venous phase.

Verification of the segmentation was performed in two distinct situations. In 75% and 90% resection groups, CT scan was performed after resection of three and four hepatic sectors, respectively. This was performed to verify the segmentation of the right lateral and the caudate lobe in CT scan, respectively.

Figure 10: Photograph demonstrating one animal during the CT scan images acquisition.

Figure 11: Screen shot showing the timing protocol for image acquisition during CT scan.

Figure 12: Hepatic arterial and portal venous segmentations using SyngoVia, Siemens. The common hepatic artery trifurcates to give off the gastrodudenal artery, the right and left main hepatic arteries which further subdivide to the corresponding sectors.

Figure 13: 3D reconstructed images for planned a) 75% and b) 90% resection. The images demonstrate the planned resected lobes (S) and the residual lobes (R) with preserved portal inflow and hepatic venous outflow. PV=portal vein, IVC= inferior vena cava.

b. Magnetic resonance based studies

· Diffusion weighted images (DWI)

In this study, Diffusion Weighted Imaging (DWI) spin echo Magnetic Resonance Imaging (MRI) sequence was used for

assessment of pigs’ liver parenchyma during mechanical ventilation. 2D multi-slice Echo Planar Imaging (EPI) images were acquired in the axial plane in conjunction with respiratory gating so as to prevent respiratory artifacts. The MRI system was a

3T Verio (Siemens, Erlangen, Germany) with maximum gradient amplitude of 45 mT/m. A 32 Radio Frequency (RF) coil split into two parts (16 anterior channels and 16 posterior channels) was placed either side ventro-dorsal of the pig. Depending on

the accuracy of the respiratory gating synchronisation, the DWI sequence could be achieved over a period of time ranging from

15 until 45 minutes. 3 averages and 3 concatenations were set with 9 weighted b-values ranging from 0 until 1100 s/mm² such

as 0, 50, 100, 150, 200, 350, 650, 800, 1100 s/mm², parallel imaging technique imaging was used with a factor of 2, TR/TE

were 2800/78 ms, matrix size: 170×160, and a Field Of View (FOV): 330 mm², resulting in a voxel size of 2.1x1.9x5 mm3 as

the slice thickness was 5 mm. In order to gain time, the partial Fourier phase was 6/8. The bandwidth was 1730 Hz/Px. Typical

image quality is shown in Figure 14.

The DWI sequence was placed on the T2 anatomical gradient echo images produced in the three different sagittal, coronal and

axial planes with 20, 20 and 25 slices respectively of 6mm thick each. This T2 2D sequence was acquired within 2 minutes. 65

concatenations were used, TR/TE was 8.3/3.69 ms, flip angle: 20 degrees, FOV: 400 mm², matrix: 256x205, bandwidth: 320

Hz/Px and only one excitation was necessary.

Image analysis diffusion parameters were performed with the Intra-Voxel Incoherent Motion (IVIM) method. Within biologic tissues, IVIM includes microcirculation of blood in the capillary network, which is also called perfusion. The relation between signal variation and b factors with an IVIM-type sequence can be expressed by using the formula in Equation 1.

Equation 1: Intravoxel incoherent motion

S = S0[(1-f).exp(-b.D)+f.exp(-b.(D+D*))]

where S is the mean signal intensity, f is the fraction of the diffusion linked to microcirculation thus the blood volume

circulation, D is the diffusion parameter representing pure molecular diffusion (slow component of diffusion representing the

intra cellular and extra cellular partitions), and D* is the diffusion parameter representing incoherent microcirculation within

the voxel (perfusion-related diffusion, or fast component of diffusion); assuming that the term D+D* represents the fast

diffusion component and thus simplified to D* as D is << than D*. Its influence on signal decay can be neglected for b factors

greater than 200 sec/mm². Equation 1 can then be simplified, and the estimation of D can be obtained by using only b values

greater than 200 sec/mm2.

D, D* and f are calculated by using a nonlinear regression algorithm based on Equation 1. All regression algorithms were

implemented with TableCurve 2D. ROIs were positioned to avoid large vessels identified on IVIM DW images acquired with

a b factor of 0 sec/mm².

Three fixed diameter ROIs were placed on the right lobar parenchyma to estimate the signal intensity for a given z stack

composed of the 9 b-values defined above. The placement of the ROIs on the right lobe of the liver was chosen to enable

tracing the changes on subsequent imaging following surgery. The three ROIs were averaged out ant the mean value of each

parameter f, D, and D* were obtained.

· Phase Contrast MRI Flowmetry (PC-MRI)

u Image acquisition:

Phase contrast MRI flow measurements: spatially registered functional flow information simultaneously with the

morphological data within a single experiment in porcine model were acquired. Flow was measured in the Hepatic Artery

(HA), the Portal Vein (PV), and the Celiac Aorta (CA) using the phase contrast MRI and compared to the flow measurements

obtained from the intraoperative transit time values. The aortic flow was measured to control for the fallacy that might be

introduced by the small arterial diameter of the hepatic artery and the large diameter of the portal vein.

The intrinsic motion sensitivity of MRI is used to quantify blood flow. Based on the fact, that the local spin magnetization is a vector quantity, in addition to magnitude data, phase images are extracted from the measured MR signal.

The general physical principle is the following: using a linear magnetic field with spins in motion, the phase difference (∆Φ),

[Equation 2] (217,218) accumulated by the spins is proportional to its velocity. Consequently, velocities are calculated from the pixel intensity of the Phase Contrast (PC) images.

Equation 2: The phase difference

= ϒ. m. v where: ∆Φ ∆

∆Φ represents the phase difference,

ϒ is the gyromagnétique ratio of the proton,

∆m is the difference of the first moment of the gradient-temporal curve, and v the velocity.

In this study, a 3T scanner (Siemens Magnetom® Verio) available at the Chirurgie et Imagerie pour le Recherche et l’Enseignement (CIRE) Platform was used. The MRI scanner generates magnetic field gradients of 40 mT/m which are used in conjunction with a 32 phased array torso coil (16 anterior and 16 posterior). The anatomical sequences were composed of three

2D True Fast Imaging with Steady-State free Precession (FISP) acquired in the axial, coronal, and sagittal planes without fat saturation. The True FISP sequence is used for semi-anatomical recognition that is to say anatomy combined with localization of the vascular structures. These three sequences were gated based upon the respiratory motion. The gating box was placed on the diaphragm so as to have half of it on the lung (no MR signal) and half on the liver (with MR signal). This gating technique

uses the quiescent expiratory phase for data acquisition and reconstruction. The parameters for these anatomical images were

as follows:

3 - T2 True FISP coronal: 2x1.6x3 mm , acquisition time: 1.40 min, TR = 3.85 ms, TE = 1.67 ms, Field-of-View (FOV): 400

mm², matrix: 256x204;

3 - T2 True FISP axial: 1.6x1.3x3 mm , acquisition time: 3.04 min, TR = 3.95 ms, TE = 1.66 ms, FOV: 400 mm², matrix:

320x256;

3 - T2 True FISP sagittal: 1.2x0.9x3 mm , acquisition time: 1.18 min, TR = 3.95 ms, TE = 1.66 ms, FOV: 300 mm², matrix:

320x256; and with the following common parameters: flip angle = 60 degrees, bandwidth: 488 Hz/Px, slice thickness: 3 mm

and an integrated Parallel Acquisition Technique (iPAT) of 2.

Velocity images were acquired using gradient-echo fast phase-contrast pulse sequences. PC-MRI was performed during

mechanical ventilation. The data acquisition window is defined by the trigger point and the trigger window values. The trigger

point, defines the point at which data acquisition begins if a valid pulse trigger is detected. The trigger window defines the

period of the respiratory cycle in which data acquisition does not occur. This parameter, which depends on the animal’s heart

rate, was set so as to include 2–3 RR intervals in the available imaging time. Each series of reconstructed data consisted of

phase images associated with the corresponding magnitude images. Flow rates were calculated from 20 velocity images

spanning the cardiac cycle.

MR parameters included a minimum TR (ca. 75 ms), a minimum TE (ca. 4.5 ms), a FOV of 160x160 mm and matrix size of

192 x 134 resulting in an in-plane resolution of 1.2x0.8 mm². The slice thickness was set to 5 mm, a flip angle of 25 degrees was used, the bandwidth was 554 Hz/Px, 5 views per segment and an iPAT of 2 was set up. The acquisition time was heart rate dependent, therefore lasting more or less 2 min per flow sequence. Encoding velocities were set to 80, 140 and 20 cm/s for the

hepatic artery, the aortic artery and the portal vein respectively. The blood flow acquisition plane was selected strictly

perpendicular to these 3 different vessels.

u Image analysis:

The DICOM images exported into the Syngo.Via (Siemens healthcare) software. Analysis was performed using an automated

Region Of Interest (ROI) selector when possible (Figure 14). In case that the operator judged a non-spot-on ROI onto the

vessel of interest of the 20 different PC and magnitude images, manual ROIs selections were then performed.

This postprocessing tool allows automatic construction, based on the heart rate, of the temporal flow rate curves Q (t) during a

cardiac cycle and calculation of the main characteristic parameters of the flow curves such as the temporal maximum and mean

Spatial velocities, vessel cross-section, flow rate during a cardiac cycle were performed. This flow rate is defined as the

product of the mean pixel velocity value inside the lumen and the vessel cross-sectional area.

u Statistical methods:

Data summary was displayed in means and standard deviation. Bland-Altman method was used to examine the agreement of readings between the MRI and those measured directly using transit time method. Data were analyzed in XLSTAT ® Pearson edition, Addinsoft. Intraclass classification coefficient was used to examine the reliability of the MRI measurement. For this purpose; SPSS V 22 (IBM ®, SPSS ® Statistics for Windows. Armonk, NY: IBM Corp.) was used.

Figure 14: PC-MRI flow measurements. ROI on celiac aorta (lumen) with a) phase contrast image, b) magnitude image and c) temporal flow rate curve showing the 20 phase contrast points of a cardiac cycle.

· Elastography: see appendix · Blood sampling and analyzed parameters

Blood samples were collected from all animals before surgery. In groups 1, 2 and 3, blood samples were collected

approximately 1 hour after liver resection, on the third and seventh post-operative days. In group 4, blood samples were collected also 1 hour after liver resection as well as on the first post-operative and the third post-operative days.

Blood samples withdrawal:

The right internal jugular vein was canulated with an 8 Fr vascular Desivalve (Vygon) cannula. The vascular cannulation is

performed through a midline neck incision that is used for contemporaneous carotid artery cannulation for the hemodynamic

monitoring.

Thirty ml of blood is withdrawn from the jugular vein and sent to the central laboratory unit in universal tubes for analysis.

Blood samples were collected by venipuncture 20 minutes after the induction of anaesthesia after an overnight fast. Samples for hemtological analysis were collected in EDTA tube. For chemistry analysis; samples were collected in fluoride tubes and buffered sodium citrate. Serum was mixed with 0.8mg aprotinin, centrifuged at 2,000xg at 4 °C for 15 minutes within 30 min after puncture and incubated in an ice container till delivery. Delivery time to the laboratory was within 45 minutes following centrifugation. For the biochemistry analysis, the flowing parameters were analyzed: Na, K, Cl, Glucose, urea, creatinine, bilirubin direct and indirect, proteins, albumin, AST, ALT, GGT, PAL, ammonia and lactate.

For the hematological analysis: Hemoglobin and hematocrit levels and the PT, INR level as well were analyzed

After midline abdominal incision, blood samples for gas analysis were withdrawn as follows and were analyzed indoor with a dedicated analyzer:

1. 1 ml of arterial blood is withdrawn from the right carotid artery.

2. 1 ml of venous blood is withdrawn from the right suprahepatic vein by direct puncture using a 30 gauge needle.

3. 1 ml of portal venous blood is withdrawn from the portal vein by direct puncture of the portal vein using a 30 gauge

needle.

Hepatic oxygen consumption is calculated as described elsewhere (219). The hepatic oxygen consumption is calculated using the equation: Equation 3: The hepatic O2 consumption

= ((Hemoglobin g/dl* 1.34 * SaO2 + 0.003 * PaO2) * HAF) + ((Hb * 1.34 * SpO2 + 0.003 * PpO2) * PVF) – ((Hb* 1.34 * SvO2 + 0.003 * PvO2) * (HAF +PVF))

The flow used in the equation is adjusted to the liver volume by a simple transformation (flow*100/liver weight).

The net hepatic lactate production is calculated using the same principle: Equation 4 : The hepatic net lactate production

= (Arterial lactate * HAF) + (Portal lactate* PVF) – (Hepatic venous lactate* (HAF +PVF))

c. Fluorescence imaging:

The principle of fluorescence imaging is to illuminate the region of interest with a light, at the excitation wavelength of the

fluophores. Under excitement, the fluophores radiate a fluorescence signal (at emission wavelength). A fluorescence imaging

device is used to capture the signal and to obtain an image with different intensity zones. The fluorescence imaging device

used here is Fluobeam®, developed by the Fluoptics. The intensity of the signal received depends directly on the intensity of

the source light but also on the distance to the tissues and the light in the room. The experimental conditions were imitated

closely for all animals with minimal daylight leak. A reference object, with stable intensity in time, is added inside the field to

reduce the dependency of measurements on the experimental conditions (Figure 15). The reference object was placed at the

same distance to the camera than the regions of interest. The intensities measured were normalized by the measured intensity

of this reference object. Fluorescence was measured in four regions of interest: liver tissue, hepatic artery, portal vein and

common bile duct. The camera was placed such that all the regions of interest could be observed at the same time (Figure 16).

The camera was fixed around 20 cm from the tissue surface, and when recording, the surgery was temporarily stopped to avoid

major perturbations on the signal. Intensities were recorded for two minutes, with a frame approximately every 0.1 second, just

after ICG injection. Then, with a 2 minute frequency, the intensity was recorded for 30 seconds, for a total recording of 20 to

50 minutes. For each frame the average light intensity of the region of interest over time was assessed. The regions of interest

were moving due to the mechanical ventilation. An in-house algorithm was developed by Fluoptics to follow the same region of interest in all images. Curves of the intensity over time in each region of interest were thus obtained.

Figure 15: Fluorescence element of reference to normalize the measured intensities, in order to compare the measurements for different animals and different days.

Figure 16: The four regions of interest are present, the liver tissue, the hepatic artery, the portal vein and the common bile duct

xi. Biopsy sampling

Samples of liver tissue were retrieved following the previously illustrated schedule. Routine pre-resectional biopsy was taken

either at the time of the pre-operative CT scan, under guidance of ultrasound or on the day of surgery just after the abdominal

incision, before clamping of the resected sectors (Figure 17). These samples served as a control for quantification of the

subsequent changes. The subsequent specimen was taken from the remnant liver lobe 45 minutes after completion of hepatic

resection. Subsequent biopsies were performed on the third day (in groups 1-3) and on the first post-operative day (in groups

4) an ultrasound guided biopsy was performed. Another biopsy was performed at the day of sacrifice (day 7 in groups 1-3 and day 3 in group 4).

Core needle biopsy was the standard for the per-cutaneous biopsies at any time point. At the time of surgery (at any time point) needle biopsy was also performed in conjunction to the incisional biopsy. Initially the needle was of 18 gauge and subsequently a 16 gauge needle was also used in conjunction with the other one to allow for a larger specimen to be taken and reconstruction to be made efficiently. For any type of biopsy, two specimens were collected were possible.

The main specimens were immediately immersed into 4% formaldehyde for preservation till analysis is due (Figure 18). Were available, the second specimens were immersed into liquid nitrogen for one of the stemmed out projects.

In needle biopsy, the deep end of the specimen was marked with a Fuscin green (Figure 19). This was to investigate one of the hypotheses on liver regeneration which proposes that the regeneration of the hepatic parenchyma is predominantly deep away from the capsular surface.

Summary of the available specimens is presented in the tables below (Table 5).

Figure 17: Intra-operative photograph depicting the post-resectional incisional biopsy.

Figure 18: photograph of an incisional biopsy immersed into 4% formaldehyde.

Figure 19: core needle biopsy with the deep end marked with Fuscin green.

Table 5: Total number of specimens collected from group 2-4 experiments

D -7 D 0 before D 0 after D 1 D 3 D 7

Needle 14 25 25 2 9 6

Incision 0 25 25 2 3 6

xii. Histology of the porcine liver

a. Histopathological Analysis Protocol

· Cell proliferation

To assess the proliferation index in the liver, Ki-67 immunohistochemistry was performed according to Hammad et al. (220).

Briefly, 5-micron-liver sections formalin-fixed, paraffin-embedded tissues was deparaffinised and rehydrated in rotihistol

(Carl-Roth, Karlsruhe, Germany) and a descending ethanol gradient, respectively. To unmask the antigens, the liver sections were boiled twice in a microwave oven for 7min in 0.01M citrate buffer (Carl-Roth, Karlsruhe, Germany; pH 6.0).

Endogenous peroxidase was blocked using a solution of 7.5% H2O2 in methanol for 30 minutes at room temperature. All

further incubations were performed in a humidified chamber. Unspecific binding sites were blocked by 3% BSA, 0.3%

TweenR80 using 200μl per section for 2 hours. The liver sections were incubated with monoclonal mouse anti-Human Ki-67

antigen Clone MIB-1 (Dako, Glostrup, Denmark, M7240, 1:50) overnight at 4ºC. After three washing steps, the

VECTASTAIN ABC Kit (Mouse IgG, Vector Lab., Burlingame, USA, PK-4002) was used according to the manufacturer’s

protocol and three washing steps were performed. 3,3’-diaminobenzidine-Peroxidase (DAB-HRP) substrate Kit (Vector Lab.,

Burlingame, USA, SK-4100) was applied according to the manufacturer’sguidelinethen hematoxylin counterstaining was

done. Proliferation index was determined by the percentage of hepatocyte nuclei that display nuclear staining to the total

number of hepatocyte nuclei (five fields per slide, 40X).

· Image processing and analysis

In order to reconstruct and analyse the bile canalicular network in three-dimensional pattern, 100 µm liver slices were

immunostained according to the protocols (220,221). Subsequently, Z-stacks (n=6-9 per arm [in group 1] ; were captured by a

60 fold objective using a confocal laser scanning microscope (Olympus, Germany, FV1000) from the midzonal compartment.

Each dataset typically consisted of 75-150 spatially consecutive images and each revealing a depth of 0.54 μm per layer.

Moreover, the image preprocessing were carried out by Autoquant X3 Version X3.03 64 Bit Edition (Bitplane) and image

analysis as well as quantification of bile canaliculi were achieved by Ti-Quant (www.msysbio.com/tiquant) according to

(220,221). An arbitrary voxel size of 0.207 μm × 0.207 μm × 0.54 μm was used for visual depth to the stack of signals. To

reconstruct the bile canalicular network and the nuclei the segmented files were imported to Imaris (Version 7.7.1, Bitplane)

according to the Bitplane guidelines.

xiii. Sacrifice

Depending on the allocation to groups, animals underwent a planned sacrifice either on the 7th day for groups 1-3 or the 3rd day

post-hepatectomy for group 4. Animals were sacrificed following a predefined protocol that involved hemodynamic monitoring similar to that at the day of surgery. Portal pressure measurement took place through a direct puncture of the portal vein with a 14 G cannula. The pressure probe was inserted into the cannula that was then retrieved backwards and a fine 4/0 polypropylene suture was used to secure the probe in place. The other probes were placed exactly in the same manner as before hepatectomy. Sacrifice was executed through an i.v injection of a lethal dose of barbiturate (100mg/kg) after removal of the remnant liver. The remnant liver was weighted after euthanasia.

Occasionally, premature sacrifice was resorted to whenever an animal was in suffering due to post-operative uncorrectable illness. In these cases, a lethal dose barbiturate was injected and autopsy was performed to identify the cause of death and retrieve the liver for analysis. Animals that were found dead in their cages were also autopsied for inspection.

xiv. Statistical methods

Analysis was performed in multiple platforms according the investigated hypothesis. As a general role, data was presented according to the variable type (quantitative or qualitative). After normality testing, summary of continuous data were represented in either mean ± standard deviation or median and range. Summary of nominal data were presented in percentages.

Odds ratio was reported where required.

Mann-Whitney-U test and Chi-square tests were used to assess the difference between groups, where applicable, for non- parametric variables, while t-test was calculated for parametric variables. ANOVA with repeated measures and Friedman tests were used to compare the evolution of parameters for parametric and non-parametric data respectively. Significance threshold was set at a p-value of 0.05.

For the production of results multiple statistical software packages were used. MedCalc Statistical Software version 14.8.1

(MedCalc Software bvba, Ostend, Belgium), Graphpad Prism version 6 (GraphPad Software, Inc., La Jolla, USA), SPSS 22.0

(IBM SPSS Statistics for Windows, Armonk, NY: IBM Corp.) with IBM SPSS Statistics Version 22 – Essentials for R

Version 22.0.0 connected to R-version 2.15.3, and XLSTAT, Addinsoft TM, France. Each of these packages was used to produce particular group of analyses depending on the familiarity of the user and the functionality of the software.

Strategy for graphical presentation of the data:

Wherever possible, graphs were produced in a dotted boxes to represent the data in a detailed fashion. This is in order to see the outliers in particular.

III. Results & Discussion:

i. Anatomy of the porcine liver and technical implications

Figure 20: A reconstructed image of the hepatic vascular anatomy. 1 common hepatic artery, 2 gastrodudenal artery, 3 proper hepatic artery, 4 right hepatic artery, 5 left gastric artery originating from the left hepatic common trunk, 6 left hepatic artery proper, 7 artery correpsondng to segment II, 8 artery to the segment III, 9 artery to the segment IV. The portal venous branches: 10 main portal vein, 11 branches to caudate lobe, 12 right main portal branch, 13 left main portal branch, 14 branch to the left medial lobe. Hepatic veins and IVC: 15 IVC, 16 common branch draining the left lateral and left medial veins, 17 trunk draining segment II and III, 18 vein draining the left medial lobe, 19 vein draining the right lateral lobe.

a. Introduction to Hepatic Anatomy

Complications after liver surgery might be drastic (200). In order to reduce the risk of complications; particularly after major liver resection, swine has been in use as a surgical model to better study and understand the different surgical situations for a long time (222). This is particularly true since the animal is readily available at a reasonable price and that the anatomy is to an extent similar to human anatomy (223). Experiments on liver regeneration after different percentages of partial hepatectomy have been performed to simulate situations with small for size syndrome (224).

One of the key factors on which the success and the reproducibility of the experimental model are based is the precise knowledge of the normal anatomy along with its variations. In humans, an aberrant right hepatic artery originating from the superior mesenteric artery or a right bile duct originating from the left bile duct could change the outcome. Despite that the swine model has been in use for decades, little information on the detailed anatomical features of the porcine liver is available in the literature (223).

Few studies have concentrated on the ultrastructure of the hepatic micro-architecture (225), others based their studies on casting (223). Furthermore, detailed descriptive data from CT does not exist. This study aimed at describing the normal porcine liver anatomy based on CT scan.

b. Anatomical features

The porcine liver is divided into five lobes; the left lateral and medial lobes, the right medial and lateral lobes and the caudate lobe (226). Apparently the left lateral lobe represents around 25% of the total liver volume (139) and is consistently the largest of all lobes (223).

Hepatic hilum: The hepatic artery runs off the celiac artery on the posterior aspect to the postero medial aspect of the portal vein, where it gives off the gastro duodenal artery just above or behind the head of pancreas. Nearly at the same level in a more superficial plane; the gastrodudenal vein joins the portal vein. The bile duct is situated on the antero-medial side of the portal

vein. The portal vein and the hepatic artery are dissected carefully throughout their extra-hepatic course, and the lymph nodes on the posterolateral and medial sides of the portal vein are removed to facilitate the positioning of the probes.

Aorta

While dissecting the supra celiac part of the aorta, the plane is accessed through an opening in between the oesophageal and aortic crura. The pleura insert low into the coverings of the aorta on the anterior and the left aspects of the vessel, around one to two centimetres above the origin of the celiac trunk and the landmark to its insertion is where the dorsal sling converges with the ventral sling of the right crus. Trying to access the plane around the aorta below the insertion of the pleura might expose the suprarenal gland to manipulations to which the pig is highly sensitive. Manipulation around this area usually caused the pigs to manifest hemodynamic instability. Having that observed during the first few pigs; we therefore preferred to access the plane a little bit higher, taking the risk of opening the pleura. Opening of the pleura could be nevertheless, avoided if the plane was accessed between the two crural slings and below the white fibers of the diaphragm. At the end of the intervention, the pleura, if opened; is drained using a suction drain that is activated after the abdominal closure and left in place for around

30 minutes (ie; time for pig to recover from anaesthesia).

On the right side of the aorta, a large lymphatic vessel will be consistently found. This lymphatic channel is sometimes found creeping on the anterior aspect of the aorta and becomes difficult to avoid injuring.

At the origin of the renal arteries portion of the aorta, the large lymphatic vessels surround the aorta from many directions and were breached nearly constantly while positioning the supra-renal flow meter. The injury to the lymphatic channels at that level is responsible for loss of significant amounts of fluids which increases the risk of mortality after surgery.

RESECTION SURGERY

One of the advantages offered by the illustrated clamping technique is that it avoids pedicle dissection. In that way the time

required for resection is theoretically shorter. Moreover, the intraparenchymal individual pedicle ligation avoids injury to the

right lateral bile duct that commonly inserts into the left main duct (223).

We planned the resection based on the CT volumetry. Nonetheless, resections were performed based on the anatomic

segmentation leading to a discrepancy of 13.5 ±11.9 % between the planned and real resected volume. The error in estimating

the resected liver volume before surgery could be related to the difficulty in appreciation of the anatomic lobes in pigs, despite

of the prominent fissures between lobes that are not visible on the CT scan; since the estimated resected volume was lower

than the resected one and the estimated residual prior to surgery was higher than that estimated after surgery. However the

anatomic resection resulted in a resection that corresponds to the desired resection in both 75 and 90%.

An estimation error is systematically reported difference between the CT measured volume and the actual liver volume and is

attributed to the intrahepatic blood volume that is included in the CT measurement but not during the operative measurement

(227), which could be adjusted for using a factor of 0.85 in the calculation of volume based on CT (228).

The later hypothesis is supported by the presence of a similar difference between the estimated volume at 7 days postoperatively using CT and the directly obtained weights 1 hour later during the sacrifice, which was nearly equal to the difference at day 0 (12±11%). Moreover, the error in residual at day 0 was larger than that for the resected volume, considering that the initially planned resected volume was around 60%. Despite that error; the correlation between the resected and the estimated volumes was excellent (r=0.8, p=0.003).

The variation range in the real resection volume percent indicates that anatomic resections might not always result in the desired percent of resection; similar to what was found in humans (229). This is particularly evident in the estimation of 75% resection as opposed to the lower error in the estimation of 90%, which suggests that the contribution of the four lobes to the whole liver volume could be slightly different from animal to animal. Nevertheless, the relatively narrow range of standard deviation suggests its usefulness as a preoperative planning tool.

For partial hepatectomy pig model, resection usually starts from the left lateral lobe then further resections are performed in an

anti-clock wise manner. The caudate lobe has a peculiar position, together with the uncountable small veins that drains the

segment directly in the IVC which resides inside the parenchyma (230); make the resection of this segment is very difficult to

pursue.

We used to measure the hepatic hemodynamics in humans using a small needle connected to a water column barometer. This

technique although widely adapted yet it conveys some difficulties due to the sensitivity of the measurement set to calibration

and positioning. We find that the hereby described measurement setting is much more reproducible and robust. Despite that it

technically invasive, the electromagnet flow meters require the vessel of interest to be liberated all around, while the pressure

probe could be introduced through a small vessel that could be sacrificed, it could also be introduced directly into a vessel that

will be repaired after extraction of the barometer probe.

c. CT scan depicted description

Artery

The proper hepatic artery originates from the common hepatic artery after it gives off the gastroduodenal artery. It then runs for a varying length before it divides into two or more branches. The average length of the common hepatic artery, measured from its origin from the celiac trunk to the origin of the gastroduodenal artery, was 42.8 ±5.3 mm, while the proper hepatic artery length was 10 ±7.6 mm. The usual pattern of division of the proper hepatic artery is trifurcation (Figure 21). It gives off

left, middle and right hepatic trunks that further divide outside the liver parenchyma to supply the corresponding hepatic

sectors.

A right gastric branch supplying the lesser curvature of the stomach coming from the left hepatic branch or less commonly

from the proper hepatic artery was consistently seen (Figure 22).

The left lateral sector is supplied by a branch that leaves the left hepatic arterial trunk, which gives off two branches to the left

lateral and the left medial sectors in most of the cases (Figure 23) or from the proper hepatic artery.

A large middle artery that usually originates separately from the proper hepatic artery supplies the right medial sector but

sometimes it supplies the left medial sector (Figure 23). In the absence of the middle hepatic artery, the right medial sector is

supplied by a branch from the right hepatic artery that supplies the right lateral and the right medial sectors. This artery crosses

in front of the portal vein to supply the right lateral lobe. In about 80% of the examined animals, another hepatic artery is seen

crossing behind the portal venous branch to the right lateral sector to supply the posterior part of the parenchyma of the right

lateral lobe (Figure 24).

The caudate lobe receives its branch that crosses in front of the portal vein, from the artery to the right lateral sector or in cases

of bifurcated proper hepatic artery from the right hepatic artery trunk (Figure 25). Each one of the previously mentioned

arterial branches further divide, usually outside the hepatic parenchyma, to supply the corresponding segments.

Portal vein

The portal vein appears to course a long track towards the left side from which the different branches appear to stem off and

enter consecutively into the corresponding sectors. Usually there is more than a single branch that leaves the main portal vein

trunk to supply each sector.

The first branch to take off from the portal vein trunk is the vein to the caudate lobe. The caudate lobe is supplied by a single

relatively large portal vein that comes off from the posterior aspect of the portal vein before its divisions (Figure 26). This is

usually accompanied by multiple small veins that originate at different levels from the posterior aspect of the main portal

trunk.

After a short distance, a large branch enters the right lateral sector nearly at the level of its anterior border. In some cases, the

caudate lobe is supplied by a branch from the vein to the right lateral sector. In these cases, a single branch first leaves the

main trunk to divide into a branch that supplies the anterior portion of the right lateral sector (equivalent to segment VI) and

another one that further divides to supply the caudate lobe as well as the posterior portion of the right lateral parenchyma

(equivalent to segment VII) (Figure 27). In less frequent situations, it receives dual branches.

The next branch that leaves the portal vein is one that divides into two or directly two portal branches to the supply the right

medial sector (equivalent to segment V and VIII in human livers). Further forward, the continuation of the portal vein enters

the substance of the left medial sector. This branch supplies the upper portion of the left medial sector (segment IV a) with

another branch that comes off more distal to supply the lower portion of this sector (segment IV b).

Thereafter, the continuation of the main trunk gives two branches. One bridges the base of the fissure from the left medial to

the left lateral sector (Figure 28) to supply the lower portion of the left lateral sector (segment III). The second branch crosses

to the upper portion of the left lateral sector (segment II) more deep inside the parenchyma and before the branch supplying the

left medial sector.

The left medial sector receives consistently two small branches from the left side of the portal branch to the right medial sector.

The supra-hepatic veins:

Each sector is traversed by one large hepatic vein that receives smaller tributaries throughout its course. The right lateral vein

unites with the vein draining the right medial sector around 1 cm inside the parenchyma before its joining to the IVC. This

common trunk runs within the substance of the hepatic parenchyma beyond the base of the fissure that separates the right

lateral from the right medial sectors before it opens into the IVC separately (Figure 29). There is a consistent vein draining the

deep parenchyma of the right lateral sector that joins the vein that drains mainly the right medial sector. A second hepatic vein

is seen traversing the deep parenchyma of the right medial sector, which opens independently into the IVC. Less frequently, a

branch draining the right medial sector is seen joining the main draining vein of the left medial sector.

The left medial sector is traversed by one single vein that drains separately into the IVC in the majority of animals. Less frequently, this branch unites with the vein from the left lateral sector to form a large common opening into the IVC.

The left lateral sector drains through two well defined tributaries, one from segment II and another from segment III. Both branches unite into a common trunk before they drain into the IVC. Frequently, one or two smaller separate branches draining segment II join the IVC below the larger trunk.

The caudate lobe drains into the IVC directly usually with a single short vein that joins that IVC on its right posterior aspect.

However, there are innumerable veins that contribute to the drainage of this segment which is not evident in the CT.

Frequently, the portal venous branches interlace with the hepatic venous plexus of a neighboring sector, which frequently bridges over the plane where the fissure lies. Those fissures divide the liver approximately in the distal 3-4 fifth of the parenchyma and more towards the visceral rather than the diaphragmatic surface as seen in the fissure between the right and left medial sectors.

Liver segmentation and sectors:

The porcine liver is divided into five distinct sectors by four major fissures, the left lateral sector, left medial sector, right medial sector, right lateral sector and the caudate lobe. The hepatic sectors are not quite distinct in the CT scan as they are in- vivo. However, some prominent fissures are easily seen, particularly distally towards the liver border. The most prominent fissures depicted in the CT are the one separating the left lateral sector from the left medial one and the one between the right medial and the right lateral sectors. The fissure between the right and the left medial sectors is on the visceral surface more than the diaphragmatic one and it is seen in the CT scan only distally. The least evident one is the one separating the right lateral sector from the caudate lobe, which its visualization depends mainly on the volume of the caudate lobe.

The mean total liver volume was 915±159 ml. The largest sector in the liver was the right medial one representing around

28±5.7% of the total liver volume. Next in order is the right lateral sector, which constitutes around 24±5%. It`s volume is very

close to the volume of the left medial sector, which represents around 22±4.7% of the total liver volume. The caudate lobe

represents around 8±2% of the total liver volume (Figure 30). Controlling for the age, partial correlation technique showed a significant correlation between the weight of the animal and the total liver volume (r=0.75, p=0.001). There was no relation between the individual lobar volume and the weight. Furthermore, there was a significant negative correlation between the volume of the left lateral and the left medial sectors and between the right medial and the right lateral sectors (r=-7.9, p<0.001 and r=-6.5, p=0.01 respectively) (Table 6).

Figure 21: Branches of the celiac trunk and the hepatic artery as depicted from a 3D reconstruction of the segmented arterial tree from the celiac trunk. 1) common hepatic artery, 2) gastrodudenal artery, 3) the left hepatic artery, 4) the right hepatic artery, 5) the right gastric artery, 6) artery to the left lateral sector, 7) artery to the left medial sector, 8) middle hepatic artery to the right medial lobe, 9)

posterior artery to the right lateral sector, 10) branch from the right hepatic artery that divides to supply the right lateral and the right medial sectors, 11) left gastric artery from the celiac trunk, 12) splenic artery, 13) celiac trunk.

Figure 22: Volume rendering technique extracting the part of the arterial tree in relation to the portal vein. The right gastric artery coming from the left hepatic artery depicted in anterior view.

Figure 23: Volume rendering technique extracting the part of the arterial tree in relation to the portal vein. The middle hepatic artery supplying the left medial sector depicted in anterior view.

Figure 24: Volume rendering technique extracting the part of the arterial tree in relation to the portal vein. The posterior right artery to the right lateral lobe turning behind the corresponding portal venous branch depicted in anterior view.

Figure 25: Intraoperative photo showing the minute arterial branch to the caudate lobe crossing over the portal trunk to the caudate lobe.

Figure 26: 3D ROI reconstruction for segmented portal and hepatic veins. The green color represents the hepatic vein and the blue represents the portal vein. The portal venous branch is the first branch from the portal venous trunk itself to the caudate lobe depicted in caudal view.

Figure 27: 3D ROI reconstruction for segmented portal and hepatic veins. The green color represents the hepatic vein and the blue represents the portal vein. The first branch coming off is a common branch that further divides into two; one supplying the caudate lobe and inferior portion of the right lateral sector, while the other branch supplying the upper portion of the right lateral sector.

Figure 28: An intraoperative photo demonstrating portal venous branch crossing to the left lateral sector at the base of the fissure separating the left lateral from the right lateral sectors.

Figure 29: Cranial view of a 3D reconstruction of the segmented hepatic sectors showing the common trunk of draining the veins from the right lateral and one of two right medial sectors.

Figure 30: Caudal view of a 3D reconstruction of the five hepatic sectors showing their volume. 1) the left lateral sector, 2) the left medial sector, 3) the right medial sector, 4) the right lateral sector, and 5) the caudate lobe.

Table 6: The summary of the segmented lobar volumes in the studied animals

Left Right Right Left lateral Caudate medial medial lateral sector lobe sector sector sector

Mean ml (%) 168.6 (18) 201 (22) 255 (28) 216.2 (24) 74.6 (8)

SD ml (%) 70 (6) 57.2 (4.7) 58 (5.7) 63.8 (5) 23.6 (2)

Min ml (%) 55 (7.3) 135 (12.8) 160 (17) 121 (17) 36 (4)

Max ml (%) 335 (30.5) 339 (31) 368 (40.7) 375 (37) 109 (13)

d. Implications

Prior knowledge of the anatomical variation of the liver is fundamental in order to perform safe and efficient surgical

technique. This study is the first to comprehensively describe the CT anatomy of the porcine liver. Beside the prominent

fissures in pigs, there are several other differences between the macro-anatomical features between the human and porcine

liver. The supply of the caudate lobe in humans comes mainly from the left hepatic artery and main or the left portal vein while

in pigs it comes from the right side (231).

The portal venous supply to the caudate lobe was reported to come consistently from the right portal vein in the study by Court

(223), however, this was not the case in our study. We found that in the majority of animals a single portal branch from the main portal trunk is the main supplying branch to the caudate lobe and in a lower frequency from the right branch supplying the right lateral lobe. Furthermore, there was not obvious right and left divisions of the portal vein. We observed a long trunk from which consecutive branches come out.

One of the multiple similarities of the pattern of the blood supply is at the left lateral sector, where the portal vein runs mostly

into the fissure separating this sector from the left medial sector and gives out two branches to the left lateral sector and two to

the left medial one. This is similar to the pattern of the left portal vein divisions in humans (232).

Variations of the arterial branching were reported in a study by Gravante and coworkers (226). They did not report however the consistent branch that leave the proper hepatic artery to supply the stomach. This artery might be mistaken as a hepatic artery if not differentiated. In our study, we found that the anterior branch supplying the right lateral sector is the consistent

one and that the posterior branch was present in the majority of animals. This is In contrast to what was reported by Court

(230), who reported that the anterior artery is the accessory one.

Noteworthy, the discrepancy in the reported anatomy of swine originate from the differences in species. Zanchet et al (233)

reported only two common drainage sites for the supra-hepatic veins into the IVC of Landrace pigs. In our study, the usual pattern was three opening joining the IVC. The two drainage sites were also encountered. However, the venous trunk of the left medial sector frequently joins separately the IVC.

The hereby mentioned pattern of divisions and arrangement has some technical implications. In liver resection models, 70 and

90% models are most commonly used to study the hemodynamic changes induced by resection (234). These models generally aim to study the behavior of small for size syndrome. However, in order to successfully carry out the intended experiment, an accurate knowledge of the anatomy should be acquired. These percentages correspond to removal of the left three and four sectors respectively.

The pattern of vascular array and the deployment of the hepatic sectors around the IVC facilitate the resection of the more peripheral sectors. For that reason, liver resection in pigs is left oriented (235).

the right lateral sector. This division occurs close to the base of the fissure separating the right lateral from the caudate lobe

(236). In such case, resection should be performed not to compromise this supply, hence should be conducted above 1 cm

distance from the base of the fissure depending on the employed technique.

Moreover, the percentage of contribution of each sector to the whole liver volume varies from animal to another. The

importance of this becomes evident with the increment of the percentage of resection. For 90 % resection, which is performed

leaving only the caudate lobe in place. The caudate lobe might constitute to as low as 4% of the total liver volume, which

might impose inaccuracies and increased risk of undue mortality.

The porcine liver is reported to be divided into 5 different lobes by 4 fissures (223,226). In our opinion, sectors rather than

lobes is a more appropriate term to the anatomical characteristics of the porcine liver that retains, in general, similar inflow and

outflow pattern to the human liver (237).

· Installment of the Circulatory Monitors

u Incisions and hemodynamic set up

Longitudinal median neck incision was performed and canulation of the right carotid artery and the right internal jugular vein was performed with a 5 and 8 Fr vascular Desivalve (Vygon, Ecouen, France) cannula. A water column based barometer was connected to the side line of the arterial cannula in order to continuously measure the carotid pressure through the anaesthesia monitor. Millar pressure probes (Millar, Texas, USA) were then inserted into the internal jugular vein (5 Fr, SPR 350) and carotid artery (4 Fr, SPR 340). Subsequently a midline abdominal incision was performed. Table mounted costal retractors and self-retaining abdominal retractors were installed.

Carotid artery cannulation was performed after dissecting a 4-5 cm segment. The artery was surrounded by a vascular tape on which traction was applied in a cranial direction. The caudal end of the dissected portion was gently grasped with a vascular forceps till the puncture was performed and the guide wire was introduced. The artery was grasped again over the guide wire to reduce bleeding from the puncture, and then the vascular cannula was introduced. The jugular vein was cannulated in a similar

way except that the dissection around the vein was performed keeping a fine tissue pad around the vein to prevent its shearing.

Noteworthy, care should be taken to avoid false passages when an excess tissue pad is present around the vein. One more difference is that the caudal application of the forceps was not performed in order to keep the vein distended.

Hepatic artery cannulation was performed through the gastrodudenal artery, which was dissected above the head of pancreas to

its confluence with the hepatic artery. Two absorbable 4/0 threads were passed under the vessel. The distal one (close to the

head of pancreas) was ligated to direct all the flow that passes across the hepatic artery towards the liver. The vessel was

snipped with fine scissors, and then a 3Fr SPR-330 Millar pressure probe was introduced through this opening toward hepatic artery. Then the proximal ligature was tied over the probe.

The portal vein pressure probe (5 Fr, SPR-350 Millar) was inserted through the gastrodudenal vein in a similar way to the hepatic artery. Both probes were further secured in place with the proximal ties. In few occasions, the gastrodudenal vein was hanged over a proximally tied knot to allow field exposure for the arterial maneuver.

A 5 Fr and 4 Fr pressure probes were inserted into the right jugular vein (tip positioned in superior vena cava, verified by waveform) and carotid artery respectively (Figure 31a); while a 3 Fr pressure probe was inserted into the hepatic artery through the ligated gastro duodenal artery and a 5 Fr pressure probe was inserted into the portal vein through a small medial gastro-duodenal tributary. For flow measurements, we used TranSonic transit time technology (TranSonic System Inc, Ithaca,

USA). A 12 or 14 mm flow probe (MA12 PAU or MA14 PAU) surrounded the portal vein and a 4 to 6 Fr flow meter probe

(MA4PSS or MA 6PSS) was used around the hepatic artery, proximal to the gastroduodenal artery (Figure 31b). A 10 to 14 Fr flow meter probe (MA 10/12/14 PAU) was positioned around the supra-celiac aorta and a 8 or 10 Fr flow meter probe (MA

8PSS or MA 10PAU) surrounded the aorta between the mesenteric and the renal arteries Pressure probes were inserted to reside in the main stream of the designated vessel andwere always of the same caliber, the flow meters size could be changed depending on the vessel size depicted from the preoperative CT scan.

The perivascular flow meters are connected to Transonic® T403 (Transonic Systems, Inc., Ithaca, NY, USA), and the Millar

Mikro-Tip® Pressure Transducer Catheters are connected to pressure control unit PCU-2000 with two-channel connected

to four Millar catheters (Millar Inc, Houston, Texas, USA).

Measurements are continuously recorded during surgery with a 16-channel amplifier connected to a computer running IOX2

acquisition software (Emka TECHNOLOGIES, Paris, FRANCE) (Figure 31c). Recordings were tabulated and the mean values

of 20 consecutive seconds were exported for analytic tables.

Figure 31: a) Neck incision with cannulated jugular vein (J) and carotid artery (ca), b) probes installed for flow measurements around (p) portal vein, (H) hepatic artery and for pressure measurement inside the (G) gastrodudenal artery and the (V) gastrodudenal vein. C) Monitoring setup consisted of measuring apparatus and display monitor. Recordings were saved on a Laptop (not shown).

· Simplified technique of liver resection

u Resection of 75% of liver volume:

CT scans were not performed for group 1 and determination of resected and remnant volumes based on the assumption that the left three lobes constitute around 75% of the liver volume (230).

In subsequent experimental groups (ie: 2-4); CT scans were routinely performed before and after surgery. Based on CT volumetric study, a 75% hepatectomy was planned for through resection of the left lateral, left medial and the right medial hepatic lobes leaving in place the right lateral lobe and the caudate lobe surrounding the inferior vena cava (IVC).

The liver was freed from its diaphragmatic attachments and from the lesser omentum (Figure 24a). A capsular incision was performed all around the resected lobes at a 1 cm distance from the insertion of these lobes to the IVC on the dorsal aspect of the liver (Figure 24b) while this incision was performed around the portal pedicles on the ventral aspect. A Kelly clamp was used for parenchymal fractionation of a 2 cm segment at the right border of the right medial lobe (Figure 24c) to facilitate the placement of a clamping Rummel tourniquet (Figure 24d), which was then placed to squeeze the parenchyma, inflow and outflow of the lobes to be resected in mass (Figure 25a,b). Care must be taken to protect the right hepatic vein that passes from right lateral hepatic lobe to the IVC close to the fissure between the right medial and right lateral lobes.

Each individual lobe pedicle was then ligated separately with a single absorbable 0 tie. The tie was passed around the pedicle with the aid of a right angled clamp that crosses, through the parenchyma, posterior to the whole pedicle (Figure 26 a,b).

Subsequently, a common left hepatic trunk draining the left lateral and left medial lobes was ligated close to its confluence with the IVC. The hepatic vein draining the right medial lobe was then ligated and the remainder of the parenchyma was fractionated with a few Kelly crushings.

Gradual declamping takes place to complete the haemostasis. Haemostasis was completed using polypropylene 5/0 or bipolar electrocautery (Figure 26 c).

u Technique for 90% hepatectomy

A more extended resection was performed in 6 out of the 22 animals. The surgery was performed in the same manner till completion of the 75% resection; thereafter a small parenchymal dissection was performed on the lateral border of the right lateral lobe using a bipolar electrocautery device. Then, using a monopolar device; the capsule of this lobe was incised all

around its circumference close to the origin of its pedicle in order to facilitate the use of a Rummel tourniquet, which was then

tightened (Figure 27 a,b). Subsequently, the liver parenchyma was dissected using a combination of Kelly clamp crushings and

ties. Two large suprahepatic veins are generally encountered and tied with absorbable 0 sutures. Similarly, a single large portal

pedicle was tied before its further subdivisions. Once resection was completed, gradual declamping and hemostais was

performed in the same manner as in the 75% resection. After resection was completed, adapted pieces of gloves were left

around the vessels as well as between the liver and the stomach to prevent adhesion formation in order to facilitate dissection

at the time of subsequent surgery.

Figure 32: technical aspects during liver resection a) dissection of the medial attachment of the left lobe to the lesser omentum, b) capsular incision marking the line of resection, c) crushing of the parenchyma at the lateral border of the right medial lobe with Kelly clamp to facilicate the positioning of d) the tourniquet

Figure 33: a) tightening of the clamping tourniquet around the 75% mass represented in the left lateral, left medial and right medial lobes. b) the efficacy of clamping is demonstrated by the colour difference between the clamped and unclamped lobes.

Figure 34: the passage of right angled clamp behind the first (a) and the second (b) portal pedicles supplying the resected lobes, and c) an after 75% resection view.

Figure 35: for 90% to be completed; a) passage and b) tightening of the clamping tourniquet around the right lateral lobe leaving just below it the caudate lobe.

u Anatomy of the Abdominal Lymphatic

The distribution of the abdominal lymphatic in the Large White pigs is quite tricky and it seems that there is an extensive plexus of lymphatics surrounding the aorta. On the right side of the aorta, at its entry through the diaphragmatic crura, a large lymphatic vessel will be consistently found. This lymphatic channel is most of the times found creeping on the anterior and posterior aspects of the aorta and becomes difficult to dissect without injuring.

At the origin of the renal arteries portion of the aorta, the large lymphatic vessels surround the aorta from all directions and were breached nearly constantly while positioning the supra-renal flow meter. The injury to the lymphatic channels at that level is responsible for loss of significant amounts of fluids which increased the risk of mortality after surgery.

ii. Physiology

a. Normal Laboratory Values

In the gas analysis, the mean arterial PH was 7.393±0.09117, the venous PH was 7.378±0.168 and the portal PH was

7.335±0.08816. The mean arterial pCO2 was 45.89±12.54 mmHg, the venous pCO2 was 55.37±13.36 mmHg and the portal pCO2 was 56.82±13.73 mmHg. The mean arterial pO2 was 273.6±63.8, the venous pO2 was 37.32±9.108 mmHg and the portal pO2 was 45.65±10.12 mmHg. The mean arterial SO2 was 99.81±0.1311 %, the venous SO2 was 64.27±14.89 % and the portal SO2 was 75.35±8.802 %. The mean arterial lactate 1.462 ±0.4654 mmol/l, the venous lactate was 1.272±0.3431 mmol/l and the portal lactate was 1.451±0.4228 mmol/l. Table 7, Table 8, and Table 9 expand the various dispersion limits for these parameters.

The mean BUN level was 3.208±1.348 mmol/L and the mean creatinine level was 95.5±20.51 µmol/L. The mean plasma Na level was 138.3 ±2.188 mEq/L and K 4.875±0.9196 mEq/L. Table 10 summarizes the main electrolytes and kidney function parameters.

The mean total bilirubin level was 5.583±3.175 µmol/L and the mean total protein level was 50.67±6.88 g/L. The mean ammonia level was 49.1±45.67 the mean lactate level was 2.013±0.9827 mmol/L. Table 11 summarizes the main hepatic functions and enzyme levels. The mean venous hemoglobin level was 97.67±15.84 g/L and the mean hemoatocrite level was 29.41± 5.037 %. The mean venous WBCs was 13.51±3.355 103/mm3 and the mean platelets count was 279±104.6 103/mm3. Table 12 summarizes the data set on the basic hematogram.

The mean hepatic oxygen consumption was 17.32±9.671 ml/100 gr liver tissue/ minute and the mean net hepatic lactate production was 0.01682±0.02719 mmol/l. Table 13 summarizes this data with their dispersion.

The mean arterial Hg 8.508±1.738 g/dl and the mean arterial hematocrite % is 25.00±5.081 %. The mean portal hemoglobin level was 8.625±1.805 g/dl and the hematocrite was 25.18±5.564 %. These data are summarized in Table 14.

The mean arterial calcium level was 0.8961±0.2321 mmol/l, the venous calcium level was 0.7861±0.2276 mmol/l and the mean portal calcium was 0.77±0.2326 mmol/l. Table 15 summarizes the calcium level data in the different samples.

Table 7: Kidney and major electrolytes profile (samples collected from jugular vein)

BUN Creatinine Na K (mEq/L) Cl (mEq/L) (mmol/L) (µmol/L) (mEq/L)

Minimum 1.5 70 135 3.9 96

25% Percentile 2.275 75 137 4.3 98

Median 2.95 91 137 4.65 99

75% Percentile 4.25 113 140.8 5.275 101

Maximum 5.8 129 142 7.3 105

Mean 3.208 95.5 138.3 4.875 99.5

Std. Deviation 1.348 20.51 2.188 0.9196 2.355

Std. Error of Mean 0.3891 5.922 0.6317 0.2655 0.6798

Lower 95% CI of mean 2.352 82.47 136.9 4.291 98

Upper 95% CI of mean 4.065 108.5 139.7 5.459 101

Table 8: Arterial blood gases, PH and lactate levels

pCO2 mmHg pO2 mmHg SO2 % PH Lactate (mmol/l)

Minimum 28.7 166.5 99.6 7.22 0.92

25% Percentile 37.4 218.4 99.7 7.353 1.035

Median 44.25 276.6 99.85 7.38 1.365

75% Percentile 50.6 327.4 99.9 7.43 1.78

Maximum 72.3 375.7 100 7.55 2.4

Mean 45.89 273.6 99.81 7.393 1.462

Std. Deviation 12.54 63.8 0.1311 0.09117 0.4654

Std. Error of Mean 3.619 18.42 0.03786 0.02632 0.1097

Lower 95% CI of mean 37.93 233 99.73 7.335 1.23

Upper 95% CI of mean 53.86 314.1 99.89 7.45 1.693

Table 9: Venous blood gases, PH and lactate levels

pCO2 pO2 SO2 PH Lactate (mmol/l)

Minimum 36.6 19.8 23.9 7.18 0.83

25% Percentile 47.28 32.98 59.28 7.268 1.028

Median 53.45 36.05 65.85 7.34 1.145

75% Percentile 59.5 45.63 74.53 7.433 1.47

Maximum 84.1 53.6 80.9 7.83 2.14

Mean 55.37 37.32 64.27 7.378 1.272

Std. Deviation 13.36 9.108 14.89 0.168 0.3431

Std. Error of Mean 3.857 2.629 4.299 0.04849 0.08087

Lower 95% CI of mean 46.88 31.53 54.81 7.271 1.101

Upper 95% CI of mean 63.86 43.1 73.73 7.484 1.442

Table 10: Portal vein blood gases, PH and lactate levels

pCO2 pO2 SO2 PH Lactate (mmol/l)

Minimum 35.1 33.7 61.7 7.2 1.05

25% Percentile 46.3 38.3 67.85 7.265 1.105

Median 56.45 45.1 76.05 7.325 1.28

75% Percentile 64.95 49.53 83.73 7.373 1.743

Maximum 78.6 71.7 89 7.52 2.45

Mean 56.82 45.65 75.35 7.335 1.451

Std. Deviation 13.73 10.12 8.802 0.08816 0.4228

Std. Error of Mean 3.963 2.921 2.541 0.02545 0.09966

Lower 95% CI of mean 48.09 39.22 69.76 7.279 1.24

Upper 95% CI of mean 65.54 52.08 80.94 7.391 1.661

Table 11: Hepatic profile from samples collected from the internal jugular vein.

ASAT ALAT GGT ALP Bilirubin Bilirubin Proteins Albumin Ammonia Lactate (IU/L) (IU/L) (IU/L) (IU/L) Total Direct (g/L) (g/L) (mmol/L) (mmol/ (µmol/L) (µmol/L) L)

Minimum 31 34 19 231 2 0.7 41 24 10 0.95

25% Percentile 33.25 39 27.25 266.8 3 1 46 25 21.75 1.4

Median 41.5 43 34 305.5 4.5 1.65 50.5 30 31.5 1.65

75% Percentile 139.8 47.5 40.5 333.5 8.75 2 56.5 32 68.5 2.983

Maximum 258 53 46 344 11 3 64 35 164 3.72

Mean 83.75 43.67 33.58 296.5 5.583 1.6 50.67 29.18 49.1 2.013

Std. Deviation 73.97 5.867 8.628 39.67 3.175 0.7336 6.88 3.92 45.67 0.9827

Std. Error of Mean 21.35 1.694 2.491 11.45 0.9167 0.2118 1.986 1.182 14.44 0.3475

Lower 95% CI of mean 36.75 39.94 28.1 271.3 3.566 1.134 46.3 26.55 16.43 1.191

Upper 95% CI of mean 130.7 47.39 39.07 321.7 7.601 2.066 55.04 31.82 81.77 2.834

Table 12: major hematogram parameters in samples collected from the internal jugular vein.

Hemoglobine (g/L) Hematocrite % WBCs /mm3 Platelets / mm3

Minimum 71 19.5 9.4 141

25% Percentile 85 26.7 10.15 212.3

Median 97 30.3 13.8 236.5

75% Percentile 105 31.2 15.4 394.5

Maximum 133 40.1 20.4 442

Mean 97.67 29.41 13.51 279

Std. Deviation 15.84 5.037 3.355 104.6

Std. Error of Mean 4.775 1.519 1.061 33.07

Lower 95% CI of mean 87.03 26.03 11.11 204.2

Upper 95% CI of mean 108.3 32.79 15.91 353.8

Table 13: Hemoglobin and hematocrit levels measured in each sample type of gas analysis

Arterial Arterial Venous Venous Portal Portal

Hemoglubin Hematocrite Hemoglubin Hematocrite Hemoglubin Hematocrite (%) (g/dl) (%) (g/dl) (%) (g/dl)

Minimum 5.700 17.00 5.300 16.00 5.800 17.00

25% Percentile 7.200 21.50 6.950 20.25 7.800 23.00

Median 8.400 24.50 8.850 26.00 8.450 23.00

75% Percentile 9.400 27.75 10.98 32.25 10.25 31.00

Maximum 12.00 35.00 16.90 49.00 11.70 34.00

Mean 8.508 25.00 9.425 27.67 8.625 25.18

Std. Deviation 1.738 5.081 3.484 10.08 1.805 5.564

Std. Error of Mean 0.5017 1.467 1.006 2.911 0.5209 1.678

Lower 95% CI of mean 7.404 21.77 7.212 21.26 7.478 21.44

Upper 95% CI of mean 9.613 28.23 11.64 34.07 9.772 28.92

Table 14: Calcium levels as measured by gas analyser in the arterial, venous and portal venous samples and the hepatic oxygen consumption and the net lactate production.

Hepatic Net hepatic Ca Ca Ca oxygen lactate (mmol/l) (mmol/l) (mmol/l) consumption production Arterial Venous Portal ml/100 gr (mmol/l) liver tissue Minimum 0.42 0.42 0.4 4.293 -0.0297 25% 0.76 0.6575 0.5675 9.357 0.005325 Percentile Median 0.935 0.78 0.83 15.25 0.0109 75% 1.06 0.845 0.95 24.36 0.02693 Percentile Maximum 1.29 1.41 1.26 37.6 0.0861 Mean 0.8961 0.7861 0.77 17.32 0.01682 Std. 0.2321 0.2276 0.2326 9.671 0.02719 Deviation Std. Error 0.0547 0.05365 0.05481 2.792 0.00785 of Mean Lower 95% CI of 0.7807 0.6729 0.6544 11.18 -0.0004615 mean Upper 95% CI of 1.012 0.8993 0.8856 23.47 0.03409 mean

b. Evolution of the hematological parameters after different resection percentages:

Summary of these finding are presented in tables Table 15 and Table 17

Table 15: Descriptive Statistics for the evolution of hematological parameters following 75% liver resection

Min Max Mean Std. Dev Min Max Mean Std. Dev

Na Before Bilirubin 132 142 137.6 2.7 2 11 4.9 3.2 (n=19) Total Before

Na After Bilirubin 127 144 134.7 4.0 0.1 10 4.8 2.6 (n=18) Total After

Na Day3 Bilirubin 126 143 137.0 4.8 3 18 6.4 5.3 (n=9) Total Day-3

Na Day7 Bilirubin 131 139 136.8 2.9 3 13 6.8 3.4 (n=6) Total Day-7

Bilirubin K Before 3.9 7.3 4.7 0.8 0.7 3 1.5 0.7 Direct Before

Bilirubin K After 4.2 8.0 6.1 1.1 0.0 6 2.7 1.5 Direct After

Bilirubin K Day3 3.3 5.9 4.4 0.9 1.0 16 3.4 4.8 Direct Day-3

Bilirubin K Day7 3.1 5.7 4.6 1 1.0 4.3 2.4 1.1 Direct Day-7

Bilirubin Cl Before 11 105 95.1 20.5 Indirect 1.0 8.0 3.4 2.5 Before

Bilirubin Cl After 94 116 106.6 5.7 0.0 5.7 2.1 1.5 Indirect After

Bilirubin Cl Day3 90 110 103 5.8 Indirect Day- 1.4 3.0 2.3 0.6 3

Bilirubin Cl Day7 88 105 99 6.5 Indirect Day- 2.0 8.7 4.5 2.3 7

Creatinine 70 129 91.6 17.6 GGT Before 19.0 52.0 33.4 8.9 Before

Creatinine 76 248 140.8 42.4 GGT After 13.0 32.0 21.9 5.9 After

Creatinine 60 365 112.4 95.6 GGT Day3 18.0 47.0 33.2 8.9 Day3

Creatinine 67 203 133 43.4 GGT Day7 13.0 39.0 26.8 9.9 Day7

Urea 1.5 5.8 3.1 1.2 ALP Before 149.0 344.0 273.7 52.5 Before

Urea After 1.6 6.0 3.3 1.2 ALP After 140.0 436.0 315.1 80.9

Urea Day3 1.8 7.0 3.2 1.6 ALP Day-3 170.0 700.0 343.9 161.4

Urea Day7 2.4 7.7 4.5 2.2 ALP Day-7 74.0 202.0 137.3 50.8

Glucose Hemoglobin 2.9 15.2 7.1 3.6 8.7 133.0 85.3 34.9 Before Before

Glucose Hemoglobin 2.1 21.9 10.9 5.7 8.5 130.6 78.1 38.3 After After

Glucose Hemoglobin 4.8 18.8 7.5 4.6 8.1 108.0 70.2 38.9 Day3 Day3

Glucose Hemoglobin 4.7 7.7 6.1 1.3 75.0 114.0 91.4 16.1 Day7 Day7

Lactates Hematocrit 0.0 3.8 1.8 1.0 0.0 40.1 27.6 8.8 Before Before

Lactates Hematocrit 1.9 9.0 3.6 1.9 9.8 41.1 27.7 7.3 After After

Lactates Hematocrit 1.2 3.9 2.2 1.1 24.0 33.2 27.9 3.2 Day3 Day3

Lactates Hematocrit 1.7 6.1 4.0 2.1 21.7 33.8 28.2 4.5 Day7 Day7

Protein WBCs 25 64 49 8.7 9.4 28.1 14.5 5.0 Before Before

100

Protein 16 38 30.3 5.6 WBCs After 4.4 21.2 10.7 4.3 After

Protein 26 44 35.7 7.2 WBCs Day3 3.5 21.2 13.9 6.6 Day3

Protein 34 50 41.7 5.4 WBCs Day7 5.5 15.3 9.4 3.9 Day7

Albumin Platelets 15 35 27.9 4.8 141.0 561.0 311.4 120.6 Before Before

Albumin Platelets 8 35 17.9 5.6 112.0 455.0 243.3 121.4 After After

Albumin Platelets 14 24 20.3 3.3 12.0 426.0 227.3 124.4 Day-3 Day3

Albumin Platelets 16 24 19.8 2.9 155.0 405.0 272.2 111.4 Day-7 Day7

AST 31 258 69.1 61.6 ALT Before 25 74 42.5 10.8 Before

AST After 63 353 160 88.5 ALT After 22 37 29.3 4.6

AST Day-3 70 1012 335 317.5 ALT Day-3 35 1031 188.1 317.9

AST Day-7 29 231 74 78 ALT Day-7 31 58 46.5 10.5

101

Table 16: Descriptive Statistics for the evolution of hematological parameters following 90% liver resection

Std. Std. Min Max Mean Min Max Mean Dev Dev

Bilirubin Na (mmol/l) 136.0 140.0 138.3 1.5 (µmol/l) Total 2.0 10.0 5.8 3.1 Before (n=6) Before

Na After Bilirubin Total 131.0 140.0 136.7 3.1 4.0 9.0 6.8 1.9 (n=6) After

Na Day3 Bilirubin Total 145.0 145.0 145.0 . 38.0 38.0 38.0 . (n=1) Day-3

K (mmol/l) Bilirubin Direct 3.3 4.7 4.2 0.5 1.0 2.0 1.5 0.5 Before Before

Bilirubin Direct K After 5.6 7.4 6.3 0.8 2.0 4.0 3.7 0.8 After

Bilirubin Direct K Day3 3.2 3.2 3.2 . 14.0 14.0 14.0 . Day-3

Cl (mmol/l) Bilirubin 96.0 102.0 99.0 2.3 1.0 8.0 4.3 2.6 Before Indirect Before

Bilirubin Cl After 102.0 112.0 106.7 3.6 2.0 5.1 3.2 1.3 Indirect After

Bilirubin Cl Day3 104.0 104.0 104.0 . 24.0 24.0 24.0 . Indirect Day-3

102

Creatinine GGT (IU/l) (µmol/l) 79.0 117.0 97.3 17.5 4.0 48.0 32.3 16.1 Before Before

Creatinine 121.0 198.0 166.0 29.5 GGT After 6.0 33.0 24.2 9.7 After

Creatinine 87.0 87.0 87.0 . GGT Day3 45.0 45.0 45.0 . Day3

Urea ALP (IU/l) (mmol/l) 3.0 4.9 3.8 0.7 234.0 355.0 299.0 44.2 Before Before

Urea After 3.1 4.7 3.6 0.6 ALP After 296.0 662.0 408.0 131.3

Urea Day3 4.0 4.0 4.0 . ALP Day-3 613.0 613.0 613.0 .

Glucose Hemoglobin (mmol/l) 2.6 8.7 5.4 2.5 82.0 110.0 99.4 12.4 (g/dl) Before Before

Hemoglobin Glucose After 4.8 10.5 6.3 2.2 94.0 101.0 96.8 3.4 After

Glucose Hemoglobin 14.7 14.7 14.7 . 71.0 71.0 71.0 . Day3 Day3

Lactates Hematocrit (%) (mmol/l) 1.2 2.9 2.0 0.6 25.3 33.3 29.9 3.4 Before Before

Lactates 4.3 7.7 5.9 1.4 Hematocrit After 28.6 31.6 30.2 1.3 After

Lactates 4.2 4.2 4.2 . Hematocrit Day3 22.9 22.9 22.9 . Day3

103

Protein (g/l) WBCs (*1000) 44.0 60.0 51.7 5.2 11.0 18.3 14.4 3.1 Before Before

Protein After 27.0 37.0 31.3 3.7 WBCs After 9.0 14.6 12.0 2.4

Protein Day3 42.0 42.0 42.0 . WBCs Day3 11.3 11.3 11.3 .

Albumin (g/l) Platelets (*1000) 27.0 35.0 30.2 2.8 209.0 385.0 321.0 69.8 Before Before

Albumin 14.0 21.0 17.7 2.7 Platelets After 29.0 368.0 264.3 157.9 After

Albumin 24.0 24.0 24.0 . Platelets Day3 230.0 230.0 230.0 . Day-3

AST (IU/l) ALT (IU/l) 29.0 70.0 48.8 18.3 36.0 60.0 46.8 9.7 Before Before

AST After 184.0 244.0 212.3 28.2 ALT After 27.0 35.0 31.0 3.3

AST Day-3 562.0 562.0 562.0 . ALT Day-3 77.0 77.0 77.0 .

It is noteworthy to state that the prothrombin activity and the coagulation profile measurements in this study were inaccurate.

The majority of results came back from laboratory stating that the preoperative value of the prothrombin activity is >100%.

This was due to the limitation of the measuring kit that was not specific to pig blood. There was however measures that either

at the top maximum value of the reading kit or below this maximum value as summarized in the table (Table 17).

104

Table 17: Summary table of the available data on prothrombin time and activity.

75 Resection % 90 Resection %

Prothrombin activity (%) N Mean Minimum Maximum N Mean Minimum Maximum

Before 12 104.5 81 120 4 105 100 120

After 13 81 41 100 4 79.3 29 100

Day1 2 52 51 53 0

Day3 7 87.6 51 120 1 66 66 66

Day7 4 106.3 92 120 0

Prothrombin time (seconds) 12 12 10.5 13.8 5 11.7 10.4 12.5 Before

After 13 13.9 11.4 19.5 4 17.9 13.3 23.5

Day1 2 17.1 16.9 17.3 0

Day3 8 12.8 10.6 16.9 1 15.3 15.3 15.3

Day7 4 12.5 12.1 13.1 0

Graphical summary of the evolution of some hematological parameters in both groups is presented in Figure 36 to Figure 44 :

105

Figure 36: Box-plot with dots representing the alkaline phosphatase levels in both groups

106

Figure 37: Box-plot with dots showing ALT levels over time in both groups with an evidence of an outlier in the 75% resection group.

Figure 38: Box-plot with dots showing ALT levels over time in both groups with the outlier excluded

107

The ammonia levels appeared to be generally higher in the 90% resection compared to the 75% resection. However, this difference was insignificant (p=0.06) due to significant inequity of variances (p=0.012) (Figure 39).

Figure 39: Box-plot with dots showing Amonia levels over time in both groups.

Figure 40: Box-plot with dots showing AST levels over time in both groups.

108

Figure 41: Box-plot with dots showing Direct Bilirubin levels over time in both groups.

Figure 42: Box-plot with dots showing Total Bilirubin levels over time in both groups. The single case that survived till day 3 in the 90% resection group was from the portal ring subgroup. An overt liver failure is manifested in this animal.

109

Figure 43: Box-plot with dots showing GGT levels over time in both groups.

Figure 44: Box-plot with dots showing the evolution of the prothrombin activity over the experiments time points in both group.

110

c. Evolution of the blood gases parameters:

The evolution of the biological parameters measured using the gas analysis machine is described in the following table and

figures

u Calcium in gases

Table 18: Descriptive statistics of the calcium levels in the blood gas samples

Resection 75% 90%

Calcium in N Mean SD Minimum Maximum N Mean SD Minimum Maximum Gases (mmmol/l)

Before-A* 17 0.88 0.24 0.42 1.3 4 0.99 0.06 0.9 1

After-A 16 0.82 0.23 0.52 1.5 3 1.17 0.19 0.98 1.4

D1-A 2 0.95 0.3 0.74 1.17 0

D7-A 7 0.93 0.32 0.61 1.42 0

Before-P** 17 0.79 0.25 0.4 1.26 4 0.8 0.16 0.58 0.96

After-P 17 0.77 0.29 0.21 1.31 4 1 0.22 0.75 1.28

D1-P 2 0.84 0.01 0.83 0.85 0

D7-P 7 0.78 0.3 0.44 1.2 0

111

Before-V*** 17 0.77 0.22 0.42 1.2 4 0.97 0.29 0.77 1.4

After-V 17 0.76 0.22 0.4 1.3 3 1 0.14 0.9 1.2

D1-V 2 1 0.14 0.9 1.13 0

D7-V 6 0.86 0.29 0.46 1.19 0

*Arterial, **Portal, ***Venous

112

Figure 45: Box-plot with dots representing the evolution of calcium in the arterial system over the different time points of the experiment

Figure 46: Box-plot with dots representing the evolution of calcium level in the portal venous blood gas sample

113

Figure 47: Box-plot with dots representing the evolution of calcium level in the supra-hepatic venous blood gas sample

114

lactate in gas: Table 19: Summary of lactate level in both resection groups

Resection 75% 90%

Lactate N Mean SD Minimum Maximum N Mean SD Minimum Maximum

Before-A 17 1.4 0.5 0.7 2.4 4 1.2 0.3 1 1.7

After-A 17 3 2.3 1 7.8 3 5.6 2.6 3.7 8.5

D1-A 2 2.6 1.1 1.8 3.4 0

D7-A 7 2.9 2 1.4 6 0

Before-P 17 1.4 0.4 0.8 2.5 4 1.3 0.4 1 1.8

After-P 17 2.6 1.8 0.9 6.7 4 5 1.3 4 6.9

D1-P 2 2.4 0.6 1.9 2.8 0

D7-P 6 3.7 1.7 1.7 5.9 0

Before-V 17 1.2 0.4 0.5 2.1 4 1 0.3 0.8 1.6

After-V 17 2.7 1.9 0.9 7.1 3 5 1.9 3 7

D1-V 2 2.2 0.7 1.7 2.7 0

D7-V 6 2.9 1.8 1.4 5.3 0

115

The lactate level after liver resection in blood gases was generally higher in the groups who had 90% resection compared to the

75 resection. This was non-significant in the arterial samples (p=0.09) and in hepatic venous system (p=0.06) but significant in the portal samples (p=0.02). We were not able to reliably indicate a significant difference in the transhepatic lactate levels in groups with 90% resection compared to groups with 75% resection.

Figure 48: Box-plot with dots representing the evolution of lactate level in the arterial blood gas sample

116

Figure 49: Box-plot with dots representing the evolution of lactate level in the portal venous blood gas sample

Figure 50: Box-plot with dots representing the evolution of lactate level in the supra-hepatic venous blood gas sample

117

Figure 51: Box-plot with dots representing the evolution of the transhepatic lactate level in groups with different resection percentages.

u Ammonia

The changes of the ammonia level was not statistically significant as shown in Figure 52

Figure 52: Error bar demonstrating the evolution of the ammonia level seen in the haematology samples.

118

u Oxygen content, saturation and consumption

Table 20: Oxygen tension (mmHg) in blood gases and its evolution in groups submitted to two different resection volume

Resection 75% 90% 75% 90%

Arterial N Mean SD N Mean SD Venous N Mean SD N Mean SD

Before 17 286 61.2 4 355.7 11.8 Before 17 37.4 8 4 40.3 8.7

After 17 301 95.5 3 319 76.6 After 17 37 6 3 40.3 3.6

Day1 2 329.7 20.9 0 Day1 2 42.6 10.6 0

Day7 7 294 93.4 0 Day7 6 37.6 15.7 0

Portal Before 15 45.6 9.3 0

After 15 41 6.1 0

Day1 2 48.6 2.1 0

Day7 5 42.4 12.7 0

119

Figure 53: Box-plot with dots illustrating the evolution of the oxygen tension in the arterial blood

Figure 54: Box-plot with dots illustrating the evolution of the oxygen tension in the portal venous blood

120

Figure 55: Box-plot with dots illustrating the evolution of the oxygen tension in the hepatic venous blood

121

Table 21: Oxygen saturation in blood gases and its evolution in groups submitted to two different resection volume

Resection 75% 90% 75% 90%

Arterial N Mean SD N Mean SD Venous N Mean SD N Mean SD

Before 17 99.9 0.1 4 99.9 0.05 Before 17 64.6 12.8 4 69 9.9

After 17 98.5 5.8 3 99.9 0.06 After 17 62.4 9.8 3 65.9 6.8

Day1 2 99.9 0.07 0 Day1 2 77.9 3.3 0

Day7 7 102.8 51.2 0 Day7 6 70.9 23.8 0

Portal 17 76.3 8.4 4 80.3 4.2 Before

After 17 69.7 8.8 4 63.9 8.1

Day1 2 88.3 2.4 0

Day7 7 85.7 13.4 0

122

Figure 56: Box-plot with dots illustrating the evolution of arterial oxygen saturation.

Figure 57: Box-plot with dots illustrating the evolution of portal venous oxygen saturation.

123

Figure 58: Box-plot with dots illustrating the evolution of hepatic venous oxygen saturation.

Table 22: Hepatic oxygen consumption levels in relation to the timing of liver resection.

Resection % 75% 90%*

Hepatic oxygen consumption N Mean SD

(ml/min/100 ml liver volume)

Before-Clamp** 12 17.3 9.7

After-Clamp 12 23.7 18.7

After-Declamp 12 31.6 34.2

124

End** 12 31.6 34.2

Day7 5 18.9 19.7

*The calculation was neglected in 90% resection groups due to missing data that hindered the calculation.

**The timing of the 2 sets of gas samples used to calculate oxygen consumption. The difference between the calculation after-clamp and before-clamp is the flow value that is being considered at that time. These two sets were used in two clusters (prior to resection, including after clamping) and (after resection, including after-declamp)

There was an increase in the hepatic oxygen consumption following clamping and hepatic resection. This change was not

statistically significant (p=0.26) Figure 59. This data represent however, the hepatic oxygen consumption in the 75% resection

groups only. Data was insufficient to calculate that rate for the 90% resection groups.

Figure 59: Box-plot with dots representing the evolution of hepatic oxygen consumption in relation to the liver resection. This was pooled analysis since missing gas data from the 90% resection groups prevented clustered graph generation.

125

u Hemoglobin and hematocrit in gas

Table 23: Evolution of the hemoglobin and haematocrit level in blood gases

Hemoglobin 75% 90% Hematocrit 75% 90% Arterial N Mean SD N Mean SD Arterial N Mean SD N Mean SD Before-A 17 8.5 1.6 4 8.8 0.9 Before 17 25 4.6 4 26 2.9 After-A 17 8.9 2.6 3 7.8 1.8 After 17 26 7.9 3 23 5.3 D1-A 2 6.8 0.6 0 Day1 2 20 1.4 0 D7-A 7 8.4 1.8 0 Day7 7 24.6 5 0 Portal Portal 17 8.7 1.6 4 8.7 1.2 16 25 4.9 4 25.5 3.3 Before-P Before After-P 17 8.4 2.1 4 9.6 1.2 After 17 24.7 6 4 28.3 3.4 D1-P 2 8.9 3.2 0 Day1 2 26 9.9 0 D7-P 7 8.9 2.5 0 Day7 7 25.8 7.3 0 Venous Venous 17 9.3 3 4 11 0.9 17 27 8.7 4 33 2.8 Before-V Before After-V 17 8.9 2 3 9 1 After 17 26 6.5 3 26.7 3.1 D1-V 2 8.3 1 0 Day1 2 24.5 3.5 0 D7-V 6 7.5 1.7 0 Day7 6 22 5 0

There was not significant change in the hemoglobin levels in the blood gas samples over time. Nonetheless, the time dependent

variation in the 90% resection groups were more than those in the 75% groups.

u Evolution of the parameters: An insight

As for the evolution of the hematological values over time, it seems that the 75% resection did not lead to liver failure in the

surviving animals. In other words, this data support the findings that 75% resection is well tolerated in this model as was

reported elsewhere (212,234). This is however, is contradictory to what was reported by Arkadopoulos (238) for a different breed of pigs, who reported that 70% hepatectomy clearly induced hepatic failure in his model.

It appears also that the blood samples collected 1 hour after hepatectomy did not show much difference from that prior to resection. It is true, however, that there was only one animal who manifested liver failure at day 3 in the 90% resection group.

126

The rest of animals in this group died early after surgery. In the 75% resection group, there was no observable difference between values from samples taken 1 hours after surgery compared to those taken prior to surgery.

There was transient reduction in the prothrombin activity in the survived animals in the groups that had 75% hepatic resection.

The groups who had a 90% did show an early reduction in the activity as well. However, the only animal that survived continued to show a reduced activity at day 3, but in conjunction with high bilirubin levels indicating a hepatic failure. In a comprehensive review of literature (191), the increase in prothrombin time was always associated with a high bilirubin levels in the context of liver failure. It is important to note that reduction in the prothrombin time in its own is not a mark for liver failure following resection (239). It was reported that prothrombin time could slightly change at the end of major abdominal surgery, which tends to recover by the 3-7th post-operative day (240). This might be dilution due to the effect of intra-operative fluid infusion (241). A similar effect was found after non-reconstituted stress, which led to hemo-concentration (242). The prothrombin half-life does not explain the reduction in the activity (243).

Evolution of gases

Lactate level is one of the indicator of hepatic function following liver resection (244). It was demonstrated that there is a high correlation between the lactate level at the end of surgery and the operative outcome from surgery (35). The lactate level was one of the few parameters that showed interesting change at the end of liver resection. Interestingly, this was even more prominent in the 90% resection compared to the 75% resection in which the levels were minimally raised compared to the preoperative levels.

In our study, ANOVA test was not possible to conduct due to the missing data in the late time points in the 90% resection.

Therefore, the reported significance is based on the independent t- or Welch test based on the equality of variances.

Statistically this was not significant in all but in the portal venous samples.

In cases where liver resection is major or where there is an important impact of the surgery on the hepatocellular function, liver becomes a net producer to lactate and the level of hyperlactatemia has been shown to correlate with the level of

127

ischemia/reperfusion (245). This is in agreement with our observation, where the lactate levels were higher after 90% resection

compared to 75% resection. This finding, in conjunction with the values of ammonia level, indicates that there is a minimal

impact of the 75% on the early liver function. The evolution of these parameters indicates that in early stages ammonia and the

lactate levels might be indicators for hepatic insults following resection.

Oxygen consumption and hemoglobin level

There was a slight increase in the hepatic oxygen consumption following the inflow clamping of the lobes to be resected in the

75% resection. Since the calculation is standardized to a 100 ml of liver tissue, redirecting the inflow towards the future

remnant liver by clamping, which will lead to increase in the relative oxygen content and concentration in the blood received

by the remnant liver, remains a potentially neutral factor. Therefore, the increment in the hepatic oxygen consumption

following hepatic clamping and resection is representative to the hepatocellular stress and to the augmented oxygen

requirements.

The hepatic oxygen consumption returned back to levels close to the pre-resection levels, together with the lactate levels,

indicate that the hepatic oxygen requirement are normalized by day-7 post-operative. Despite that it was not possible to calculate the oxygen consumption in the 90% resection, the higher lactate levels in these groups might indicate the higher oxygen requirement.

The value of the regional oxygen content and the lactate flux as a representative to the metabolic demands has been a subject

of interest in limited number of studies. In fact, their absolute values are difficult to interpret out of the context of measurement

(246). It seems that liver continues to consume lactate at low Fio2, where the rest of the splanchnic region is producing lactate

(246). Hemoglobin levels might vary after resection due to possible blood loss. Hemodilution from the infused fluid or third

space fluid mobilization might also contribute to the reduction in hemoglobin levels (247).

We observed that the hemoglobin level was higher in the portal vein compared to the venous or arterial hemoglobin

concentration in the group under 90% resection. This observation might explain the higher portal venous oxygen concentration

128

observed elsewhere (248). But it contradicts our observation about the portal venous oxygen saturation, where it tended to be

lower than in other blood samples. This was more pronounced in the 90% resection groups compared to the 75% resection.

Perhaps that the partial clamping and resection led to increased splanchnic oxygen extraction. This increment in the extraction

might be partially explained by the reduced splanchnic inflow, as evidenced from the reduction in the flow in the artic artery

above the celiac trunk. It was documented that the reduction of arterial inflow into the splanchnic circulation leads to increase

in its fraction of oxygen extraction (249).

We hereby report probably the first series of comprehensive routine baseline values in large white pigs Samples were collected

under anesthesia were we used the protocol of xylazine-ketamine (250) for induction. This protocol was reported not to

influence the blood gas analysis (250). The lowest calcium concentration was in samples collected from the portal vein.

Kallner (251) found that postprandial calcium is lower than before meals. In our animals, the PH of the portal venous blood was lower than in other samples types. This makes the alkaline tide effect, where an increase in the PH would result is

reduction of the ionized calcium level (252), unlikely to be the explanation for this difference. Particularly that the lactate

levels in the portal blood were the highest.

The difference between the arterial and venous concentration of calcium is known in the literature (253). Perhaps this

difference in related to the increased protein contents (254) in the portal blood which reduces the ionized calcium levels, which

we did not test in our experimental setting. Calcium levels in our experiments were generally higher than values in humans

(254).

Perhaps the elevated serum calcium is a contributing factor to the higher coagulability of the animals’ blood compared to

humans (255). In this study, we found that the normal coagulation profile of large white pigs is exceeding the human profile.

Since we used kits dedicated for humans, we were not able to identify accurately the activity except that it exceeded 100%. It

was previously reported that prothrombin activity is higher in pigs than in humans (256). Analysis of the coagulation profile was not precise in our animals since their values were beyond the reference values in our laboratory. Platelets count, on the

129

other hand, was comparable to the human level (257). Furthermore, the variability in the level with the resection percent can be an indicator of the remnant liver stress given that it has a strong association with mitochondrial function.

The difference in hemoglobin concentration was small between the different samples, however, both hemoglobin and hematocrite levels were higher in venous blood than in arterial and portal blood samples. These values are lower than in humans (258,259). Leucocytes count was slightly higher than in healthy human subjects (260). BUN and serum creatinine levels are within ranges that correspond to the normal values of healthy human subjects (261). Similarly, serum sodium (262) and chloride levels are equivalent to those in healthy human subjects. However, potassium levels were slightly higher than in humans (263).

The pCO2 levels were significantly lower in arterial samples compared to venous and portal venous samples with a smaller difference than what was found in other studies (264). The arterial values of pCO2 in our animals were slightly higher than the corresponding values in humans (265). However, this increase in the arterial pCO2 values might be attributed to the effect of anesthesia (266) or as a result of ketamine induced respiratory depression (250). The gap between the arterial and venous pCO2 might reflect some degree of tissue hypoxia in anaesthetized animals (267) with a resultant increase in pCO2 production from anaerobic metabolism (268). However, this was not associated with significant reduction in pH, which might suggest that different physio-metabolic pathways play a major role in the difference in the observed values.

Lactate levels were lower in the venous samples, taken from the suprahepatic vein, than in the arterial or the portal vein. This implies that the liver actively metabolizes lactate (269). However, a net lactate production was seen in these animals, which is contrary to the findings observed in transplanted individuals (270). Nevertheless, it should be noted that the amount of splanchnic lactate is influenced by the fraction of inspired oxygen (246), which in our protocol is relatively high.

Lactate values in the portal blood were higher than in venous samples. This might be attributed to the activity of the different intestinal flora in production of lactate (271). This might contribute to the relatively high unconjugated bilirubin fraction compared to the human individual (272).

130

Oxygen tension and saturation were both higher in portal samples than in venous samples. The difference in oxygen saturation between the arterial and venous samples was reported to be an indicator of tissue oxygen delivery (273). However, this

interpretation is not feasible in our observation due to the artificial increase in oxygenation of animals under anesthesia. Post

hepatectomy hepatic oxygen consumption was identified as a mediator for liver regeneration (274). Similarly, serum lactate

has been described as a prognostic marker after live resection (244). Hepatic transaminases levels were higher in these animals

than reported for healthy human subjects (33). Both direct and total bilirubin levels were within a similar range to that

described for healthy humans.

Conclusions:

To ensure accurate interpretation of the experimental results, it is important to be aware of the reported porcine normal hematological and biological parameters. Some differences between these parameters and the corresponding human parameters do exist and it should be considered prior to extrapolation of the human parameters values onto the porcine values. This study

confirms agreement between some human and porcine parameters and reports some differences as well. It also emphasizes the

importance of special calibration when assessing the coagulation profile.

131

d. Hemodynamics

· Normal Hemodynamics The baseline parameters for the different arterial and venous pressures and flows are summarized in Table 24.

Table 24: Summary of the normal pressure and flow parameters prior to clamping or resection. P: pressure, Q: flow, pv: portal vein, ha: hepatic artery, and aoC: aorta above celiac trunk.

Mean Std. Deviation

Pjug mmHg 1.95 2.17

Pcar mmHg 46.70 9.06

Ppv mmHg 9.99 9.22

Pha mmHg 37.33 7.65

Qpv L/min .63 .17

Qha L/min .20 .08

QaoC L/min 1.90 .76

· Changes in Hemodynamics

u Changes with respiration and effect of ventilation

Measurements were driven from a continuous 20 seconds recording of a stable signal and were used to compute

averaged pressure and flow values. As expected the impact of theFigure cardiac 60 cyclesFigure on the 61 pressure. Table 25 parameters was larger

than the impact of the mechanical ventilation as demonstrated in and presents the average flow.and standard deviation of the estimated variation due to heartbeat and mechanical ventilation for each pressure and

132

Table 25: Evolution due to mechanical ventilation and heartbeat average and standard deviation. The variation around the average (in time) during 20 seconds was computed to estimate the impact of the ventilation for signals measured in veins. For arterial signals, the variation around the average in peak systole was computed. The variation around the average (in time) during two cardiac cycles was computed to evaluate the impact of a heartbeat.

Mechanical ventilation

PJug Pcar QaoC QaoI Pha Qha Qpv Ppv

(min-avg)/avg (%) -85 (53) -4 (2) -5 (2) -7 (1) -5 (1) -7 (3) -47 (30) -11 (3)

(max-avg)/avg (%) 39 (10) 5 (1) 7 (2) 6 (2) 7 (1) 9 (1) 35 (18) 17 (3)

Heartbeat

PJug Pcar QaoC QaoI Pha Qha Qpv Ppv

(min-avg)/avg (%) -84 (91) -23 (8) -74 (19) -94 (27) -19 (6) -38 (15) -12 (6) -3 (1)

(max-avg)/avg (%) 79 (79) 29 (11) 133 (54) 197 (103) 19 (11) 65 (39) 12 (8) 3 (1)

PJug: Jugular pressure, Pcar: Carotid pressure, QaoC: flow celiac aorta, Pha: Hepatic artery pressure, Qpv: portal

venous flow, Ppv: portal venous pressure.

133

ventilation

Theth impact of the cycle-3 on the flow measurements in the hepatic artery, the aorta- above the celiac and below

e renal arteries was around 1 % for all the measurements, while it was around 1 2% for the pressure variability attributed to the mechanical ventilation.

6Conversely,-11%. the variation, due to a cardiac cycle was larger for the carotid artery pressure and for hepatic- artery pressure

The variation of the flow rates attributed- to a cardiac cycle for the aorta above the celiac,: 74% (19%) / +133%

(39%).(54%), for the aorta beyond mesenteric artery: 94% (27%) / +197% (103%) and for hepatic artery: 38% (15%) / 65%

were

Contrary to arterial signals, the venous pressures and flows more sensitive to mechanical ventilation thanU the cardiac cycle except central vein pressure, which was sensitive to the ventilation, as well as to the heartbeat. nder mechanical ventilation, the central venous pressure was maximal at the end of inspiration and minimal at the end of expiration.- The same observations were made for portal vein pressure.- The variations attributed to the ventilation cycle was 85% (53%) / +39% (10%) for central venous pressure and 11% (3%) / +17% (3%) for portal vein pressure. T

The cardiac cycle impact on the portal- vein pressure was smaller than on central venous pressure.- he estimated variation due to a cardiac cycle was 84% (91%) / +79% (79%) for central venous pressure and 3% (1%) / 3% (1%) for portal vein pressure. influenced

The mechanical ventilation the portal vein flow (PV). Opposite to pressures, the portal flow was maximal at the end of the expiration cycle, and minimal at the end of the inspiration (for a mechanical ventilation). In comparison to the averaged value, the portal flow increased by 35% -(18%) and decreases by 47% (30%) during the ventilatorFigure cycles.61

The variation of portal vein flow due to heartbeat was 12% (6%) / +12%(8%). For the presented animal in the maximal variation of portal flow was around 0.6 L/min.

134

Figure 60: Pressures in mmHg over a few respiratory cycles before clamping or liver resection.

Figure 61: Flows in Liter/minute over a few respiratory cycles at the beginning of surgery, before clamping.

135

u Effect of hepatic pedicle clamping on the hemodynamics

Figure 62

The portal vein pressure seemed to slowly increase ( ). The increase of pressure occurred directly after the clamping. The amplitude of the hepatic Figureartery 62pressure increased afterTable the 26 75% clamping. The averaged hepatic artery flow decreased for the 75% clamping ( ), as illustrated in . Besides the mean value decrease, after the clamping two characteristic changes could be observed: the first peak was sharper, meaning the second peak was lower than before liver resection, and diastolic flow was at low values for longer time. These changes were typical and were

observed, Figurein all animals,62 at 75% clamping instant. At clamping instant, the portal flow seemed to decrease slowly as

shown in . The- central venous pressure was not affected by the clamping. The carotid artery pressure was not

changes from its pre resection value and waveform.

In the celiacFigure aorta 62 neither the flow waveform nor its mean value were impacted in the 10 seconds following the 75% claclampingmping. ( ). The decrease observed of celiac aorta mean flow was therefore not instantaneous following the

TheFigure change 63 in. the portal flow per unit liver mass was significant after resection yet the arterial flow was not as illustrated in

136

Figure 62: Pressure panel (a) showing the increase in the hepatic artery and portal venous pressures upon clamping, while the central venous and carotid artery pressures have not changed. Flow panel (b) showing the hepatic artery (HA), portal vein (PV) flow reduction, while the celiac aorta (AoC) flow was not affected. Black lines indicate the clamping time.

Figure 63: Portal vein (a) and hepatic artery (b) flow per liver weight before and after 75% liver resection.

137

u Effect of 75% resection

The evolution of the- hemodynamic parameters in relation- to- the resection was computed for each variable, with the following formula (z z0)/z0, where z and z0 are post and pre resection average measurements (either flow or pressure); respectively. by - . inThe central venous pressure tended to decrease after liver resection 50% (min: 169% / max: +977%) The decrease surgery the venousby pressure- is linked to the blood losses occurring(Table during 26 the surgery. The carotid pressure decreasedpressure durings

9% (min: 56% / max: +9%) was observed ). infusionThe reductionFigure of 64 arterial and venous are

linked to the blood loss and it was partially corrected with the fluid . . is showing the different variation

in pressure values after liver resection in each type of resection The variation in the venous pressure were more

pronounced compared to the changes in the arterial pressures.

Figure 64: Box-plot demonstrating the amount of change in the different pressure values following the resection in each type of resection.

138

- (Figure 62) T

The aorta flow above the celiac trunk decreased by 14% (min: 45% / max: 32%) . he aortic flow between mesenteric artery and kidney arteries increased after liver resection. Depending on the animals, the evolution of this aortic flow was different. After liver resection, the arterial tree resistance of the liver has increased leading to a lowerTable

26hepatic artery flow. Moreover the celiac aorta flow and the portal vein flow also decreased after liver resection (

). The aorta flow split is therefore reorganized leading different evolutions of the aortic flows. decreased - Table 26

Locally, the hepatic artery pressure by 11% (min: 59% / max: 46%) after the 75% liver resection ( ).

The decrease of pressure was mainly due to the blood losses occurring during surgery, similarly to the pressure measured in the carotid artery. The -

hepatic- artery flow rate(Figure decreased 65) T after liver resection. After 75% liver resection, a resectiondecrease of 74% (min: 90% /

max: 6%) was measured- . he portal vein flow tended to decreaseTable 26 after liver . The portal vein flow

dropped by 9% (min: 50% / max: 63%) after the 75% liver resection ( ).

Figure 65: Box-plot demonstrating the global reduction in the flow values following resection in the different groups. The reduction appears more in the higher resection percentage. B: before, E: end of surgery.

139

T -

he portal pressure increased 3% (min: 90% / max: 186%) at the end of( Figurethe surgery 64) and the central venous- pressure

decreased, leading to the drastic increase of portocaval pressure- gradient . In contrastthe to pre Tableresection 26). value, the venous pressure difference increased by 53% (min: 69% / max: 331%) at the end of resection ( Table 26: Hemodynamics measurements before and after resection average and standard deviation (two first columns), and median (min/max) post-resection evolution of the pressures and flows compared with pre-resection values.

D0 before (X0) D0 after (X) (X - X0)/X0

-50% (-169% / Pcv mmHg 2.1 (2.7) 0.89 (3.0) 977%)

Pcar mmHg 54.5 (8.2) 46.4 (10.1) -9% (-56% / 6%)

QaoC L/min 2.06 (0.83) 1.79 (0.78) -14% (-45% / 32%)

QaoI L/min 0.76 (0.41) 0.78 (0.36) -16% (-38% / 53%)

Pha mmHg 48.2 (12.4) 42.4 (11.0) -11% (-59% / 46%)

Qha L/min 0.18 (0.07) 0.05 (0.02) -74% (-90% / -6%)

Qpv L/min 0.74 (0.19) 0.66 (0.2) -9% (-50% / 63%)

Ppv mmHg 6.7 (3.2) 9.8 (12.0) 3% (-90% / 186%)

Ppv – Pvc mmHg 4.5 (2.5) 6.4 (2.3) 57% (-69% / 331%)

140

u The remnant volume and its impact on hemodynamics

To assess the relation between the residual volume and the alteration in the hemodynamics, we have ran a bivariate correlation

between the residual volume and the different hemodynamic parameters at approximately one hour after the resection (flow

and pressure). There was a significant negative correlation between the jugular venous pressure and the resection percentage

(r=-0.56, p=0.036). Subsequently, an independent t-test showed a mean difference in the hepatic artery pressure of 27 mmHg

(p=0.016).

We interpret that to a combination of larger amount of fluid loss associated with the higher resection percentage, which partially contributed to reduction of the jugular venous and the hepatic arterial pressure. In addition to the increased resistance across the remnant liver, which led to reduction of the venous return – hence lower central pressure – and a more pronounced hepatic arterial buffer response associated with the higher resection percentage. The purpose of these models was to simulate the observations and attempt to explain the interactions that occur among the different systems (concerned here with cardiovascular system mainly) in the setting of major liver resection. A simplification of the model concept is given is Figure

66. We started with a 0D model and escalated from there as required. The reason for this is that integration of the 3D equations is quite complex.

Figure 66: Simplification of the dimensions of the mathematical fluid modelling.

141

· Closed loop (0D) model

Different groups have worked on the modeling of the liver hemodynamics, at different scales and for various applications.

Liver lobule porous models were proposed to model glucose transport and metabolism (275), and to study the influence of a septum and tissue permeability (276), including in cirrhosis (277).

At the organ scale, lumped models for multiple vascular generations have been developed to study hypothermic machine perfusion (278). Electric analogue model of the liver is proposed to improve the setting of the machine perfusion. A lumped model of the splanchnic and liver circulation has been proposed to illustrate the link between hepatic venous pressure increase, vessel contractility and liver interstitial fluid (279). Different models have been developed on transport and diffusion of a compound in the liver, including whole-body pharmacokinetics models (280) or to study tumor detection with MRI (281)

Convection is based on resistive models of the different generations of arterial and venous trees. The flow in the hepatic arterial and venous trees is modeled for the first generations with the Bernouilli equation, while porous media models the flow in the smallest vessels. The trees geometry is based on CT-scans. Hepatic artery flow 3D computational fluid dynamics (CFD) simulations for rigid and flexible walls have been performed in (282) to study direct drug-targeting.

Liver models have also been developed to study the impact of liver surgery. Flow behavior for different H-Graft diameters has been studied with a resistive model and compared to clinical observations in (282). A 3D CFD simulation has been performed in the portal vein before and after right lobe hepatectomy in (283). The surgery was simulated by changing the geometry.

Similarly, for a two-lobe liver lumped model, driving conditions were kept unchanged before and after hepatectomy.

Various resection sizes and two different surgical techniques have been simulated using a resistance model, based on cast reconstruction, of rat liver vasculature (284). Most of these works thus do not consider the dynamics induced by the surgery or the interaction with the rest of the circulation. The present work aims at dynamically modelling partial liver resection, with a closed-loop 0D model of the cardiovascular system and the liver. In (285) we have proposed a numerical scheme for 1D hemodynamics models, and explored in a generic 1D-0D pig model the hepatic artery waveforms, to understand the

142

experimental changes observed during hepatectomy. Here, the impact of the surgery on the liver and on the whole body hemodynamics were identified. Moreover, the consequences of blood loss and fluid infusion were studied. The simulations, done on twelve pigs, were quantitatively compared to experimental measurements from 75% hepatectomy. Prediction of hemodynamics changes relies partly on liver lobe masses. Thus, several options are tested. We hereby present the available experimental measurements, the cardiovascular and liver models and the parameter estimation procedure. Then, we show the simulation results from the partial hepatectomy and the comparison with the measurements.

Experimental measurements: Hepatectomies were performed as described in our previous study (286), following the anatomical guidance previously published (287). During surgery, the various hemodynamic measurements were continuously recorded as described earlier and in our previous publication (286). The cardiac output (CO) was estimated assuming celiac aorta receives around 60% of CO (288) (assuming humans and pigs have similar distributions). Heart rate is computed from the CA pressure probe. Before and after the surgery a CT-scan is performed as referred to earlier. After resection, the removed liver is weighted and the left and median lobe masses are then assessed. To estimate the liver masses, four different assumptions are made as described in Table 27, with varying predictive capabilities.

Table 27: The different mass assumptions description of the total liver, left lobe, right lobe and median lobe. Their degree of certainty increases, and conversely their degree of predictability decreases from A1 to A4: preop calculation, peri-op calculation possible, post-op calculation.

143

0D closed-loop model: A 0D hemodynamics model (289,290) of the liver and the entire cardio-vascular system around is

developed. The model aims to represent hepatectomy, i.e. the resection but also other related phenomena. Hence, only the

involved organs are included, resulting in five blocks (Figure 67).

Figure 67: Schematic representation of the 0D closed-loop cardiovascular and liver blood circulations. RCR block and liver lobe parameters are shown. Qi;b is the infused or removed ow to account for blood volume changes. Rpv: resistance portal vein, Rha: resistance hepatic artery,

The detailed methods for developing this model were further explained in a previous publication (291).

144

The main results of this model that hepatic surgeons are particularly interested in specific pressure and flow values hypothesized to be linked to liver failure. Thus, here the variables of interest are: arterial pressure, PV pressure, the venous pressure drop Ppv-Pv across the liver, PV flow and HA flow. Simulated and measured values are compared for these variables.

Pre-resection stage

The tuning procedure described above, based on pre-resection measurements, gives good agreement between pre-resection simulated and measured values. Figure 68 displays the simulated pressures and flows of interest against the measured ones in logarithmic scale. The dots in Figure 68 are nicely aligned along the curve y = x illustrating the good match between the results and measurements, for all pigs. Parameters are tuned for each animal, thus the inter-animal variability is well captured.

Standard deviations for measured and simulated variables are: 10.4 mmHg for arterial pressure, 2.4 mmHg for PV pressure,

3.1 mmHg for the pressure drop, 0.06 L/min and 0.18 L/min for HA and PV flow rates respectively.

145

Figure 68: Pre-resection measurements vs simulation values in log/log scale, for each variable (unique color) and for each animal (one dot). Pressures are in mmHg and flow rates are in L/min.

Liver partial resection simulation

Impact of liver lobe mass assumptions. The post-resection simulated variables are impacted by the estimation of the liver lobe masses. Thus, the simulations are run, with the four different mass estimations, for twelve different pigs and compared with measurements of the relative error for the different variables and mass assumptions. The arterial pressure and PV flow are almost not impacted by the lobe mass differences (at most 20% difference). The lobe mass estimations have a significant impact on PV pressure (at most 68% difference), the pressure drop (at most 90% difference), and HA flow (140% maximum difference). The last mass estimation (A4) gives the smallest relative errors (in average), thus this mass assumption is kept for the rest of the simulations. It is indeed the least predictive assumption but gives as expected the best simulation results.

Hepatectomy simulation. The simulation result averages for twelve pigs are compared to the measurements (Figure 69) before and after resection. After liver resection, on average, a 45% increase PV pressure and a 98% increase in pressure drop are measured. A small decrease in arterial pressure of 12% is observed. Moreover, a large decrease in HA flow, 74%, and a

146

smaller decrease in PV flow, 30%, are measured. In the model, on average, a 66% increase in PV pressure and a 110% increase for pressure drop are simulated. The fact that these pressures increase is coherent with the measurements, but these increases are overestimated. The arterial pressure decrease is only 3%. The model underestimates the decrease of PV flow

(5%), but captures well the HA flow decrease of 75%.

Figure 69: Measurements (full) and simulations (dash) at different states of the surgery: pre-resection, post-resection. Simulations with (dashed green) and without (dashed red) blood volume changes are represented for the A4 mass assumption.

147

Taking into account changes in blood volume. During surgery, the total volume of blood in the circulation varies. Estimating its loss or gain is complex. However, pressures are strongly linked to the circulating blood volume. Thus, the changes observed in arterial pressure measurements are used to estimate the change in blood volume. The volume added or removed is chosen such that the simulated post-resection arterial pressure corresponds to the measurement. Therefore a decrease of 12% in arterial pressure is obtained with the model.

Figure 69 displays the simulated variables, taking into account the change in blood volume, averaged over twelve pigs.

Pressures are largely impacted by changes in blood volume. The model agrees well with the measurements for PV pressure and pressure drop, with increases of 43% and 82% respectively. HA flow is still correctly simulated, with a 79% decrease, and PV flow is improved, with 23% decrease, however is still overestimated.

Similarity of measured and simulated populations. The model ability to reproduce the animal population variability is verified with a two one-sided t -test. The A4 mass assumption is used to perform the simulations. The test is performed with and without blood volume changes. For each variable of interest and each simulation the relative error is computed. The test is performed on the obtained relative error vectors. The simulated variables and the measurements, before resection, are similar.

The post-resection simulation with blood volume changes and measurements are similar only if we incorporate the blood volume changes, while if we don`t, the simulation result in measurements different from what is observed.

Discussion and conclusion

Measurements explained by modeling. The behavior of the measured pressures and flows during 75% hepatectomy are analyzed using the model. The observed HA flow decrease corresponds exactly to the increase of the HA tree resistance due to the 75% liver resection, without the hepatic arterial buffer response (HABR) being needed. At leading order, the liver arterial system behaves as _P = RQ, with _P the arterial pressure drop, which remains almost constant, R the liver resistance and Q is the HA flow. The 75% liver resection induces the HA tree resistance increase of 75%, thus explaining the decrease of HA

148

flow. The arterial pressure is not impacted by the liver resection because the HA and liver resistances are small compared to the rest of the systemic circulation.

However in average it decreases by 12% in the measurements. This decrease is a consequence of the blood loss, as suggested by the model. PV flow measurements, in average, decrease by 30%. The main decrease is due to blood (volume) loss.

However the simulations show that liver resection tends to decrease portal flow by around 5%. This is reinforced by the fact that animals with larger blood losses have a more important impact on the hepatic artery flow measurements. It might be of interest to note that HABR is initially described based on the adenosine wash out theory in a non-resection situation, where the portal flow is reduced with different degrees and the arterial flow changed were evaluated. Those experiments were carried out in cats and the authors of this hypothesis have stated honestly that around 50% of their experimental animals did not respond to the initial test of adenosine infusion. Furthermore, they noticed that if the animals do not respond to that initial stimulation, under the basal conditions, then there was no observation of the buffer response (292). This plausible theory has been carried out to explain the observations seen in the resection setting where there is increase in portal flow per unit mass (not a pre- haptic reduction of the flow as in the experiments with adenosine). Our experiments did not address specifically the adenosine washout theory in response to the increase in portal flow per unit mass of liver after resection but the model integrated parameters did not need the addition of HABR to the model to explain what happen to the arterial flow following resection

(291). Interesting, it was observed that increased portal flow through mechanical pumping did not affect the hepatic arterial flow (293).

PV flow decrease. The measured PV pressure by 45% and venous pressure drop by 98% after the 75% resection. The portal pressure increase is expected given the increase of PV tree and liver resistances due to resection. However it is compensated by four mechanisms: HA flow decrease, interaction with the rest of the circulation which causes PV flow to decrease, the sinusoidal dilatation that lead to compensatory reduction in PV tree resistance after its initial increase and the general pressure decrease due to blood loss. Indeed, the simulations without blood volume change predict a 66% increase of portal pressure and adding the blood losses the increase is 43%. If the venous pressure remains constant during surgery, the pressure drop would increase by 110% as simulated with the model. However, the measured CV pressure decreases by 33%. Thus, the blood losses

149

lead to all venous pressures decrease, and an increase in the pressure drop estimated as 82% in the simulations. Sensitivity and

variability. The liver lobe mass estimations impact the post-resection simulated variables due to the fact that the liver lobe

resistances and capacitances depend on lobe masses. The mostly impacted variables are PV pressure, the pressure drop and HA

flow. This is expected since arterial pressure and portal flow, are strongly dependent on heart and digestive organs parameters

respectively. A better estimation of the liver lobe masses may improve post-resection simulated variables.

Moreover, several events happen during the surgery due to surgical acts, anesthetists interventions etc. Here, the model

demonstrates that taking into account the change in blood volume improves the simulated post-resection prediction, knowing

e.g. the change in arterial pressure measurements. In terms of variability, the simulated results are in good agreement with the

measurements, both for pre-resection and post-resection with blood volume changes, according to the two one sided tests

(TOST) tests. The tests also show that taking into account blood loss significantly improved the model outputs agreement with the measurements. Pressure and flow changes due to hepatectomy without any volume change are also simulated with the model. These results may represent the state after the surgery, once the blood volume is back to the pre-resection volume.

Under this assumption, the portal pressure and pressure drop, important for liver failure (Allard et al., 2013), may be

underestimated with intraoperative measurements. Thus, for a 75% liver ablation, the model predicts an increase, in the

following post-operative days of 110% instead of 82% for the pressure drop and 66% instead of 43% for portal pressure.

Among animals, various post-resection behaviors are obtained with the simulations. For example, without change in blood volume, for animal iF03, PV flow is almost unchanged (decrease by 1.7 %) compared to iF12 for which it decreases by 13%.

The simulated blood volume change is adapted on an animal basis to the arterial pressure change. For animal iF02 a loss of

200 ml is simulated compared to animal iF08 for which a volume infusion of 200 ml is modeled. Therefore, the model is able

to simulate the different hemodynamics states that occur post-resection.

· Kinetic scheme (1D) Model

Understanding the interaction between the hepatic blood flow, the hepatic resection, and regeneration is challenging on clinical

grounds. This is, in part, due to the anatomical complexity of the organ receiving dual blood supply with complex

microvascular network. Mathematical modelling seems to be a promising approach to highlight these interactions. This model

150

took into account the diameter and length of the medium and large size vessels of the animals extracted from the CT scans that were obtained as part of the protocol. The simple (0D) closed loop model described in the previous section did not simulate the waveform that results from the arterial pulsation. Therefore, a more complex model (1D) is used, which results in good simulation as depicted in Figure 70. The equation used in this model is given in

Equation 5

151

Figure 70: Screenshot of the observed hepatic arterial waveform before and after hepatic pedicle clamping indicated by black line in the middle of the graphs (top panel) and the resultant simulation from the 1D model.

e. Kinetics of the hepatic volume evolution and the architectural changes following 75% resection

The mean operative time was 4.6 ±0.9 (95% CI=4-5) hours. The mean time for hepatectomy was 1.23±0.76 (95% CI= .78-1.5)

hours. The mean initial liver volume was 971.9ml and 95%CI=880-1062ml. Following resection, the residual liver volume was

236.8ml and 95%CI=198–265ml. At day-1, day-3, and, day-7; the liver volume was 407ml 95%CI=173–626ml, 587ml

152

95%CI=504-669ml, and 631.5ml 95%CI=521–741ml; respectively. The mean HU at day-1 was 40±1.9 U, day-2= 32±0.7 U,

day-3= 48±1.4, and day-7=53±0.5 U (p<0.001) (Figure 71).

Figure 71: CT scan estimated liver volume in the peri-operative period a) 3D reconstruction of the total liver volume and the estimated residual hepatic volume, b) Axial CT in a venous phase showing the actual residual volume following resection, c) axial (non-contract enhanced) CT scan depicting the hepatic volume in the first postoperative day, and d) further increase in the hepatic volume at day 3 postoperative is shown in a contrast-enhanced axial CT scan. e) Boxplot with dots representing the hepatic volume evolution. Liver-V is the whole liver volume, V is volume and D is the day after surgery. f) Boxplot representing the change in the ROI density – measured by the Hounsfield units.

After liver resection, the portal vein flow decreased (p=0.02) but with a significant increase of portal vein flow per liver volume (p<0.001). Then, the latter gradually decreased. On postoperative day-7 the portal flow per liver volume was 54% larger than

153

baseline (p=0.02). The hepatic artery flow decreased after liver resection (p<0.001) and remained low. However, the hepatic artery flow per liver volume remained similar to baseline - with a trend to be lower at postoperative day-3 by 41% and at day-7 by 36% - (Figure 72).

Figure 72: Portal vein (left) and hepatic artery (right) flow per liver volume (in ml/min/LiverVolume(ml)), before and after resection (D0- B, D0-A) and during liver regeneration at day-1 (D1), day-3 (D3) and day-7 (D7).

There was an initial significant increase (ie: worsening) in the histological scoring for the micro-architectural changes induced by resection towards the 2nd post-operative day extending over the 3rd postoperative day. These changes faded towards the 7th post-operative day. This difference was significant (p<0.001) except between samples taken on the seventh day compared to those taken prior to surgery (p=0.39) (Figure 73).

154

Figure 73: Histopathological analysis under light microscopy of the porcine liver using H&E x200: a) Liver sample prior to resection showing the absence of inflammation or steatosis. b) Day 3 postoperative showing microvesicular steatosis (à). Ki67 x200: c) Liver sample prior to resection showing scanty Ki67 positive cells. d) At day 3 postoperative there is an increase in the

155

number of Ki67 positive cells. CD31 x100 : e) Scanty staining of the sinusoidal cells in samples prior to resection, f) increase in positive staining of the sinusoids at day 3 post-operative, g) Boxplot demonstrating the change in the pathological scores over time.

There was a significant increase in the Ki67 expression (p<0.0001) in the days following resection to peak on the second postoperative period and nearly normalized at day 7. The mean % of Ki67 positive cells in preoperative samples was 4.7±4.6 %, 5.7±5.3% in after-resection samples, 15.3± 23.2% in day-1, 52.9±17.9% in day-2, 42.4±23.7% in the 3rd postoperative days, and 7.8±7.2 % in the 7th postoperative day. The expression of CD31 in the analyzed samples increased significantly on the second and third days compared to the one prior to surgery samples (p=0.03) (Figure 74).

Figure 74: a) Boxplot representing the temporal change in the Ki67, b) Bar diagram plotting the CD31 expression. Significance was estimated with a Two-tailed test statistic (Wilcoxon signed rank) at an α-error of 0.05 using SPSS V.22.

A summary of the biochemical parameters and their values measured before and after the resection is presented in Table 28. Correlation coefficient and two-tail significance between the hepatic volume increase at day-3 and the hepatic oxygen consumption before resection and the net lactate production at the end of the procedure were r=-0.82 (p=0.01) and r=-0.70 (p=0.03) respectively. No other variable – among the ones studied – had significant correlation.

156

Table 28: Summary of the evolution of the hepatic functions and flow parameters in the 75% resection group

Lactate Before (mmol/L) resection After resection Day- 1 Day- 3 Day- 7

Mean 1.77 3.61 2.37 2.16 4.01

Std. Deviation 1.04 1.91 .57 1.11 2.08

ASAT (Ul/L)

Mean 69.1 160 452.33 335.22 74

Std. Deviation 61.6 88.47 159.63 317.47 78.03

ALAT (Ul/L)

Mean 42.47 29.33 71.66 188.11 46.5

Std. Deviation 10.8 4.62 22.03 317.87 10.46

Total Bilirubin (µmol/L)

Mean 4.94 4.83 7.33 6.44 6.83

Std. Deviation 3.2 2.5 2 5.3 3.4

157

Direct Bilirubin (µmol/L)

Mean 1.53 2.72 4.66 3.36 2.36

Std. Deviation .73 1.54 2.08 4.80 1.13

GGT (Ul/L)

Mean 33.36 21.94 30 33.22 26.83

Std. Deviation 8.92 5.91 6.08 8.89 9.92

ALP (Ul/L)

Mean 273.68 315.05 431.66 343.88 137.33

Std. Deviation 52.5 80.9 92.26 161.3 50.83

Platelets (/mm3)

Mean 311.3 243.28 171.5 227.3 272.2

Std. Deviation 120.6 121.3 2.1 124.4 111.4

158

Prothrombin activity (%)

Mean 104.5 81 52 87.57 106.25

Std. Deviation 11.6 16.3 1.4 25.9 12.6

Ammonia (µmol/L)

Mean 41.4 81.6 119.0 71.0 39.2

Std. Deviation 38.5 38.3 41.7 61.7 29.6

Hepatic artery flow HAF (L/min)

Mean 0.18 0.05 0.06 0.07 0.07

Std. Deviation 0.07 0.02 0.02 0.04 0.03

HAF per Liver Volume

( ml/min/LV(ml))

Mean 0.19 0.23 0.15 0.11 0.12

159

Std. Deviation 0.08 0.10 0.09 0.06 0.07

Portal vein flow

PVF (L/min)

Mean 0.77 0.68 0.93 0.98 0.78

Std. Deviation 0.19 0.19 0.25 0.40 0.19

PVF per Liver Volume

( ml/min/LV(ml))

Mean 0.83 3.11 2.41 1.70 1.27

Std. Deviation 0.27 0.19 0.93 0.60 0.37

u Interaction between remnant volume and regeneration

There was no statistically significant correlation between the residual volume and the volume gain at the day 3 (p=0.8).

u Volume evolution: An insight

This study documents the speed of liver regeneration in this porcine model following 75% hepatic resection. The estimated volume gains in the early postoperative days were almost double on the first postoperative day trending to a plateau between the 3rd and the 7th postoperative days. The study also highlights the architectural structure in the early phases of porcine hepatic regeneration, which is a

160

fundamental knowledge that would enable novel surgical applications – devised for the temporary portal hemodynamics modulation – to be precisely calibrated (205).

Hepatic artery flow decreased due to increased arterial hepatic resistance to flow after liver resection (291). Hepatic artery flow per liver volume remained in baseline ranges, suggesting that regeneration is mostly triggered by the portal flow per volume change. Indeed, despite that portal flow decreased after resection - mainly due to blood loss and the removal of the blood contained in the resected liver (291) -, portal flow per liver volume increased by 275%. On postoperative day-1 the latter was still larger than the baseline but decreased again on postoperative day three and seven.

Redistribution of the portal flow through the remnant liver is responsible for a shear force that subsequently contributes to liver regeneration (173). In our study, cell proliferation estimated with Ki67 expression peaked at day-2 and day-3, following an increase of portal vein flow per liver volume observed after resection and at day-1. The amount of portal flow diversion to the remnant was found to positively correlate with the degree of hypertrophy and the mitotic index following portal vein ligation in a small animal model (294).

It was reported that the cellular regeneration precedes vascular regeneration during liver regeneration following partial hepatectomy (295). In our study, the Ki67 peaked on the second day after the resection. The pattern of the CD31 scores indicated that the increase in the stained sinusoids in the studied fields over days could be related to the relative reduction in the edema and steatosis rather than true endothelial cell proliferation. The sinusoidal endothelial cell repopulation was found to occur over the first 48 hours following rat liver transplantation (296). A situation that could be different in its kinetics from the resection setting. The sinusoidal cell proliferation started at 48 hours and peaked at 72 hours in a resectional rat model (297).

A deeper analysis of these studies shows that the density of the sinusoidal spaces decreases in the first days along with a remarkable increase in the intervening parenchymal spaces prior to evident sinusoidal formation after the first 72 hours in rats. This is particularly visible in figure 2 provided by Sato (297). The principles of sequential cellular kinetics in the regenerating liver following partial resection were shown in a study by Navarrate (87). Ready hepatocytes with polyploidy divide rapidly from bi-nucleated to mononucleated cells and the mononucleated cells increment their ploidy early with a parallel increase in the cell volume. This also consolidates the hypothesis that was shown by Ding (295).

161

Capacitance of the liver could contribute with up to 20% of the volume change in normal circumstances. This mechanism is responsible for accommodating the increased portal flow per liver volume after major resection. However, this is the case till a limit, with regional variation within the parenchyma (183). Oedema and periportal exanguination are seen after partial hepatectomy in a small for size remnants but were also seen where the liver functions were preserved (298).

In non-small for size remnant model, the development of the full blown pathologic picture described by Demetris (193) is not expected. In our model, we observed transient changes similar to those changes but less severe. These changes are associated with disruption of the sinusoidal basement membrane and hemorrhagic tracking in the peri-sinusoidal spaces. It probably results from the excess in sinusoidal pressure and failure of the capacitance mechanism leading to stagnation and edema that could hypothetically contribute initially to the volume increase. The relation between the pathological scores and the hepatic functions has been linked to the microarchitecture changes in one of our previous studies (205).

There is evidence that the hepatic regeneration process is associated with transient steatosis (299), which is mainly derived from peripheral adiposity (300). Looking more closely at the corresponding H&E histological subscore, we observed an increase in the steatotic index following resection in this model. Furthermore, this is supported by the pattern of the HU change depicted in the CT scans. This suggests that the porcine liver regeneration follows more or less a pattern of regeneration that is similar to that of the rats indicating that cell replication is not the only factor responsible for such rapid volume increase.

Steatosis has been also demonstrated to contribute to the volume increase in patients receiving chemotherapy for colorectal cancer (301). The reduction of the Hounsfield Units – ie: density – corresponds to the transient development of hepatic steatosis. The mean values of HU measured in our model at the first postoperative day indicate that fat deposition occurs very early in the regeneration cycle and disappears over the course of the first few days. This further reinforces the evidence of fast regeneration in porcine liver following resection.

The hepatic fat accumulation was found to start prior to the first wave of hepatocyte proliferation and lasts during this phase and then fades with the subsequent three waves of hepatocytes proliferation (302). This is demonstrated as well in this study by the HU measures on the first day, which precedes the peak of the proliferative Ki67 index – ie: on the second day.

162

Hepatic steatosis has been linked to the oxidative mitochondrial activity and the hepatic oxygen consumption (303). We found a significant inverse correlation between the hepatic volume increase at the third postoperative day and the hepatic oxygen consumption after resection. The negative correlation between the hepatic volume increase and the net hepatic lactate production by the hepatic remnant liver at the end of the procedure is well explained in that context, as previously studied by our group (35). The increased oxygen requirement by the remnant liver – as reflected by the increased lactate production – is an expression of stress. Interestingly, Ozdogan and collegues (304) demonstrated that the supplementation of hyperaric oxygen enhanced the volume and the quality of the regenerating liver. Perhaps hepatic protection via administration of N-acetylcysteine (NAC) could be an option to improve the metabolic performance of the liver – as was found in the study from Torzilli group (305) – and improve hepatic volume gain as well in critical cases. However, this area is still under preclinical trials and has not been validated clinically to the best of our knowledge.

The main limitation of this study is the lack of numeric quantification of the non-hepatocytes stained structures. However, the inferences given from the stratified subjective quantification are consistent with the evidence drawn from the existing body of literature. Furthermore, although most measurements were performed on all animals for all days, only a relatively small number of animals were studied for postoperative days measurements. However, statistical significance was still found as reported.

Conclusion: The porcine hepatic volume evolution following major resection is a fast process. The majority of the volume gain is within the first three days. Cell proliferation was found maximum on second postoperative day, following an increase of portal vein flow per liver volume after resection and on postoperative day one.

f. Normal ICG handling detected by live imaging

The signal first appears in the hepatic artery with a few milliseconds until its appearance in the portal vein. The peak of the signal in the artery precedes the signal peak in the portal vein with a few milliseconds. This is followed by the parenchyma peak which we observe going into plateau as the signal in the bile duct (Figure 72). Noteworthy, the peak pattern is different between the vessels. The arterial peak tend to be sharp compared to the blunt portal peak. The peak pattern represents the large first pass uptake as represented by the arterial peak, while the blunt portal peak represent the slower redistribution from the gut vessels.

163

Before 7.6

6.9

6.2

5.5

4.8

4.1

3.4

2.7

1.3

0.6

-0.1 1 307 613 919 1225 1531 1837 2143 2449 2755 3061 3367 3673 3979 4285 4591 4897 5203 5509 5815 6121 6427 6733 7039 7345 7651 7957 8263 8569 8875 9181 9487 9793 10099 10405 10711 11017 11323 11629 11935 12241 12547 12853 13159 13465

HA-ICG PV-ICG Liv-ICG BileDuct-ICG

Figure 75: Normal fluorescence signal intensity curve in the four target ROIs (HA: hepatic artery, PV: portal vein, Liv: liver parenchyma, and bile duct).

164

g. Changes following resection

There was no obvious changes in the fluorescence signal following 75% liver resection. The signal curves followed similar pattern to the before resection one and a similar timeline, except that there is a slight lag of the portal venous signal peak (Figure 73).

Figure 76: Fluorescence signal intensity curve after 75% resection in the four target ROIs (HA: hepatic artery, PV: portal vein, Liv: liver parenchyma, and bile duct).

On the contrary, the signal pattern change after the 90% resection seemed to be different from the curve of the 75% resection in that the arterial and portal signal peaks were stunted with a larger delay as well as seen in Figure 74.

165

Figure 77: The fluorescence signal pattern after the 90% resection. Note the stunted arterial and portal venous peaks compared to the 75% as well as the larger delay between the two peaks.

The evolution of the signal intensity curves for the subsequent days is also interesting. The delay between the arterial and venous peaks is shortened and the peaks regained their amplitude. The pattern is illustrated in Figure 75

166

Figure 78: The fluorescence signal curve on the first day after 75% resection.

On the 7th postoperative day, the curves of animals underwent 75% resection has regained a pattern similar to that preoperative except that the parenchyma peak plateau showed a delay as shown in Figure 76.

167

Figure 79: The fluorescence signal curve on the seventh postoperative day after 75% resection.

The observed changes are – to an extent – in keeping with the hemodynamic changes observed after resection. The postoperative reduction in the absolute portal flow could be linked to the stunned signal peak observed for fluorescence. This is reinforced by the greater peak flattening after the 90% resection compared to the 75% resection and furthermore by the return of the pattern to preoperative pattern, which corresponds to the pattern of changes in the portal flow.

h. Mortality and survival

168

In general, the 75% resection is well tolerated in pigs (230). In this study, we had several mortalities in animals subjected to this resection. This might be in part explained by the invasiveness of the protocol, which none of the previous studies attempted. The purpose of this invasive setup was to accurately monitor and study the relationship between the changes in hemodynamics and the post-operative liver regeneration.

In group 1, there was still lower mortality among animals when compared to the mortality of the subsequent 75% resection group. In this group, animals who had a portal flow between 2.2-4 times the original portal flow per unit liver mass had a better survival regardless their subgroup allocation. This indicate the importance of the portal flow changes into survival and mortality of animals after liver resection. This assumption was based on previous reports, where the increase in portal pressure was essential for liver regeneration (174) and prevention of the pressure from exceeding certain limits is also essential for recovery (207). Despite that is most likely to be true, we still have some doubts about the applicability of this information in our study, since most of the animals died in the early hours after surgery.

The main difference in the adopted protocol in the subsequent group was related to invasive probe pressure monitoring and the dissection of the aorta above the renal artery. This dissection was inevitably responsible for lymphatic leak due to major ductal injury around the aorta. On re-exploration, at autopsy or sacrifice, animals who had aortic dissection at the renal artery level showed important fluid collection, which could be responsible for fatalities among the deceased animals.

Moreover, in the first 4 animals in group 2, where we transferred to a new center, we modified the anesthetic protocol to include a low dose noradrenaline infusion, based on an advice from the local veterinary. We observed marked instability of the circulatory parameters of these animals and three of them died prematurely.

169

On the other hand, groups which had a 90% resection showed no survivors except for one animal in which a portal ring has been placed. This animal survived till day3 with interesting increase in liver volume, however, this animal manifested significant liver failure and had to be sacrificed at day3. Interestingly, we abandoned the suprarenal dissection in the 90% resection groups to reduce the morbidity and kept the rest of hemodynamic installations the same. Nonetheless, there was a 100% mortality in the group with

90% resection, which did not have portal ring for flow modulation.

The high mortality in this group indicates that the magnitude of liver resection was intolerable to these animals. This is in fact contrary to what was previously reported (230). Still there could be a contribution to the invasive nature of this protocol, which required a prolonged operative time as well.

The main points to learn out of this discussion is that pigs could not tolerate excessive lymphatic loss after partial hepatectomy that otherwise is well tolerated. Furthermore, the 90% resection in this model was lethal. However, there is a strong potential for the portal flow modulation using the ring to improve survival in this animal model and perhaps the future experiments would show that.

170

iii. Clinical Application

i. Radiology

u Magnetic resonance based studies

DWMRI: preliminary work.

Quantifiable non-invasive estimation of the proliferative components during liver regeneration after partial liver resection in a porcine model. Applications of diffusion weighted images (DWI) have been explored extensively over the past decade (306). The majority of research has been dedicated, until recently to neuroimaging (307). Recent work in the liver revealed its potentials to detect and

characterize various kinds of tumors, as well as to follow up the response to treatment (308). Furthermore, it has been used to assess

the status of the hepatic parenchyma and to estimate liver fibrosis (309). The pattern of proliferation and the organization of the

hepatic micro architecture are responsible for the manifestation of hepatic failure (310).

Aim:

This study aims to assess the potential of the DWI to quantitatively estimate the regenerative hepatic components after partial liver

resection.

Methods:

171

Seven large white female pigs were subjected to a standard 75% hepatic resection. MRI was performed before surgery and at

predetermined post-operative days. A 3-T SIEMENS MAGNETOM Verio-syngo MR B19 Germany was used for imaging. The TR

was adjusted to 280 ms, and the TE to 81 ms.

A diffusion-weighted multisection spin- echo type of echo-planar imaging sequence that combined the motion-probing gradients

(MPGs) before and after the 180° pulse with an echo-planar imaging read- out was used for the intravoxel incoherent motion (IVIM)

echo-planar MR imaging. The sequence was repeated for nine values (0, 50, 100, 150, 200, 350, 650, 800, and 1100) of the MPG,

yielding images with nine diffusion weightings for each section. The perfusion (1-f, Dfast) and diffusion (f,Dslow) effects on the

attenuation of signal intensity (S) in each voxel are parameters estimated with the IVIM bi-exponential model Equation 1 (311):

Sb=S0*((1-f)*EXP(-b*Dfast)+f*EXP(-b*Dslow))

Where S is signal and b is the strength of the magnetic field. Exp is the exponential and D is the diffusion value in the - fast -

compartment which is the vascular and slow is the cellular compartment and f is the fraction volume.

Liver tissue specimens were taken the same days as DWI was performed, and analyzed for proliferative Ki67 index and 3D micro

architecture reconstruction and constituent quantification.

Preliminary results:

DWI parameter estimation: excellent model fitting was obtained following our imaging protocol. Estimated parametersare within the

biological bounds (312).The parameters on the Figure 80 (a, b, c and d) correspond to S0, Dfast, f and Dslow respectively.

172

Figure 80: Biexponentional model fitting example using 9 b-values for parameter estimation, day-3 after resection.

The estimated intravoxel cellular volume fraction was lowest before surgery and progressively increased over consecutive

examinations with the peak at the 3rd day post-hepatectomy. Paradoxical reduction of the intravoxel perfusion fraction f was estimated as in Table 29.

173

Table 29: Examples from different animals with different DWI values at different time points in relation to surgery

Sample animals Ffast Dfast Fslow Dslow

Animal-1 Before Surgery 0.355663 0.019074 0.644337 0.000227

Day-1 after surgery 0.262236 0.028559 0.737764 0.001067

Day-3 after surgery 0.141117 0.027917 0.858883 0.000901

Animal-2 Before Surgery 0.208566 0.02396 0.791434 0.000859

Day-1 after surgery 0.398516 0.025738 0.601484 0.000927

Day-3 after surgery 0.235218 0.019393 0.764782 0.000773

Animal-B6 Before Surgery 0.176956 0.023136 0.823044 0.00081

Day-3 after surgery 0.230817 0.052183 0.769183 0.000669

Histological analysis: In a parallel series analysis, Ki67 indices were highest at day-3 compared to before resection and decreased later on (Figure 81).

174

Figure 81: Ki67 proliferative index in specimens taken at different time points, showing a peak activity at day-3 with a trend towards normalization over the end of the first week after surgery.

Inferences from 3D volume quantification indicate that relative to the preoperative specimen, cellular components had higher density in the analyzed tissue volume, which is consistent with the estimated intravoxel cellular volume (Figure 82).

175

Figure 82: 3D reconstruction of confocal microscopy images showing the higher cellular density (blue stained nuclei using DAPI immunohistochemistry fluorescence) days after resection (right) compared to before resection (left). Red and green structures are sinudoidal and biliary networks respectively.

More quantifications are required to provide precise correlations with the estimated volume components, which are ongoing.

Conclusions:

DWI could be a useful non-invasive tool to estimate hepatic cellular proliferation and perfusion changes which might correlate with

the clinical outcome of surgery.

176

Phase contrast MRI (PC-MRI) for flow measurement compared to the intra-operative measures

Eighteen animals were included in this study. The mean age was 3 months ± 9 days and their mean weight was 37 ± 3 kg. The mean

blood flow in the portal vein measured in the PC-MRI was 0.55 ± 0.12 l/min, while in the transit time measurement it was 0.74 ± 0.17 l/min. The mean flow in the celiac aorta was 1.4 ± 0.47 l/min in the PC-MRI and 2 ± 0.6 l/min in the transit time readings. The flow in the hepatic artery was 0.17 ± 0.1 l/min in the PC-MRI, while it was 0.13 ± 0.06 l/min in the transit time measurements (Figure 83).

Figure 83: Boxplot of the blood flow measurements in the celiac aorta (Qoc), portal vein (Qpv) and the hepatic artery (Qha) in the MRI (MR) and the transit time (TT).

177

Agreement between the PC-MRI measurements was good for all vessel types. The intraclass correlation coefficients were 0.92, 0.93,

0.94 for the portal vein, the hepatic artery, and the celiac aorta, respectively). Similarly, the agreement between the measurements in transit time was good for all vessel type. The intraclass correlation coefficients were 0.99, 0.98, 0.99 for the portal vein, the hepatic artery, and the celiac aorta. These coefficients were for the average measures and very close results (>0.9) were obtained for single measurements as well. All p-values were <0.001 for these coefficients.

A linear correlation between the two methods was found taking into account all types of vessel, R2 = 0.81 (Figure 84). The flow ranges are varying between the different vessels, therefore the agreement between the two methods for each vessel type was then studied with the Bland-Altman method. There was a significant difference between the blood flow measured in the celiac aorta with

PC-MRI and with transit time (t=3.3, p=0.0023). Bland-Altman revealed an estimate of bias of 32% with a 95% CI= -49%-15%

(Figure 85).

178

Figure 84: Least square regression for PC-MRI and TT flow measurements. PC-MRI flow measurements vs. transit time flow measurements in the celiac aorta (CA), portal vein (PV) and hepatic artery (HA). The regression line equation is y = 1.29x +0.048.

179

Bland and Altman plot

40%

20%

0% 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3

-20%

-40% Difference (%)

-60%

-80%

-100%

-120% Average (QaoC1_J0_direct + Aort_CE_MO_MRI)/2

Bias CI Bias (95%) CI (95%)

Figure 85: Bland-Altman plot indicating the systematic difference between the flow reading in MRI and Transit time in the aorta.

From Bland-Altman bias estimation, a linear relation between the two measurements is introduced. Let k, XMR, XTT denote respectively the estimate bias from Bland-Altman, the PC-MRI flow measurement and the transit time flow measurement. They relate by definition as Equation 6, Equation 7: Equation 6: The estimate of bias between both measures using Bland-Altman method

XMR – XTT = 0.5 k (XMR + XTT)

180

which implies the Equation 7: The estimate of bias in the transit time measurement

XTT = XMR (2 k)/(2 + k)

∗ − The perioperative transit time measurement in the celiac aorta was then estimated, with the Bland-Altman bias, multiplying the pre-

operative PC-MRI flow measurement by 1.38. The average absolute difference between the true transit time measurement and the

estimation was 0.45 l/min. There was no significant difference between the mean blood flow in the hepatic artery measured with PC-

MRI or with the transit time probes (t=1.4, p=0.18). Bland-Altman identified a 17% bias with a 95%CI= -15 to 51% (Figure 86).

Bland and Altman plot

200%

150%

100%

50%

Difference (%) 0 0.05 0.1 0.15 0.2 0.25 0.3

-50%

-100%

-150% Average (Qha1_J0_direct + Hep_art_MO_MRI)/2

Bias CI Bias (95%) CI (95%)

Figure 86: Bland-Altman plot indicating the non-signifcant difference in estimation of the flow in the hepatic artery using either the

MRI or the transit time method.

181

The perioperative transit time flow can be estimated by multiplying the PC-MRI flow measurement by 0.84, with an average absolute error of 0.08 l/min.

There was a significant difference between the portal blood flow measured in PC-MRI and in the transit time technology (t=-4, p=0.0007). Bland-Altman detected a 40% bias in the estimated portal vein flow values in PC-MRI compared to the transit time method (Figure 87). For portal flow, the bias is almost constant for all measured flows, with an average absolute error of 0.21 l/min, the perioperative portal flow measurement can be evaluated by multipying the PC-MRI flow by 1.5. To further re-enforce the validity of these findings, there was no correlation between the average of both methods and their difference for each vessel (p>0.05).

Bland and Altman plot

100%

50%

0% 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8

-50% Difference (%)

-100%

-150% Average (Qpv_J0_direct + Port_MRI)/2

Bias CI Bias (95%) CI (95%)

Figure 87: Bland-Altman plot revealing the systematic underestimation of the MRI flow readings compared to the Transit time flow readings.

182

Discussion of the comparison between the PCI-MRI and the Transit-Time based measurement of the hepatic blood flow

To our knowledge, this is the first study comparing the hepatic blood flow measured using transit time ultrasound to PC-MRI. Herein,

there was an agreement between both methods.

There are many limitations with flow measurements from the transit time method being invasive, might requires additional dissection,

which increases the surgical burden. Furthermore, contact temperature, the type of the acoustic coupling medium, and the vessel probe

fit can influence the reading significantly as reported in the instructions for use (313). Noteworthy, the probe diameter could influence

the flow readings by up to one third (314). Likewise, the ratio of the outer to the inner diameter of the vessel such as in the hepatic artery could impact the flow readings significantly (315). This might render it non-ideal for a real blood flow evaluation. Up to now, a

reference method for comparison of the performance of PC-MRI in assessing hepatic flow is not readily available for clinical use

(316).

Noteworthy, some of these factors are shared with the PC-MRI. For instance, identification of the diameter is an important

determinant factor for the accuracy of the Doppler reading (317), which is similar to the importance of pixel identification in

determining the flow value in PC-MRI. Over evaluation of the area leads to over estimation of the flow. One factor to which this

might be attributed is the sensitivity to the pixel measurement (318). We have observed such influence when selecting manually

versus an automated contour selection, which might influence the average of flow calculation.

The advantages of PC-MRI for flow measurements are that it is a non-invasive procedure, compared to the transit time. It also has

minimal interobserver variability (319), hence better reproducibility. Our findings support this statement. Therefore, PC-MRI was

considered a reliable method for the flow measurement (320). The PC-MRI flow measurement would enable estimating the

parameters of the predictive model of hemodynamics pre-operatively and therefore perform the simulation before the surgery (285).

183

Interestingly, the variability of readings was wide but did not jeopardize the agreement depicted by the Bland-Altman method (316).

The results from the hepatic artery flow rates measured intraoperatively or with PC-MRI can be used with that perspective as absolute

values. This is similar to the finding driven from another study (316). However, results from transit time and PC-MRI are different in

absolute values for the portal vein and the celiac aorta and a correction coefficients was required to estimate the perioperative flow

with the PC-MRI measurements.

One of the possible explanations of the lower values of the PC-MRI is the effect of time lapse between both types of measurements, which was accompanied by an increase in the animals` age. The laparotomy, the fluid infusion and the dissection of the vessels may also impact the transit time measurements and could be responsible for at least some of the observed differences. There is also an important reduction in the hepatic flow parameters following closure of the abdomen (321). This might indicate that the observed difference could be attributed to the fact that the abdomen was opened during imaging.

In our study, animals had the same mechanical ventilation setting in surgery and imaging. Moreover, the compared flow data were taken at the end of expiration in both methods. Therefore, less likely that the readings were affected by the breathing cycle as was observed – up to 40% - in other studies (322). In the study by Yzet (320), all readings obtained by Doppler were higher than those estimated by PC-MRI. This is similar to a certain extent to the finding of our study.

Our data suggests that both the transit time measurements and the PC-MRI are not interchangeable in terms of absolute values but a correction coefficient could be used for that. However, caution should be exercised when adopting the PC-MRI as a tool for blood flow measurement.

Conclusions: The absolute hepatic artery blood flow values driven from the PC-MRI could be reliably used as an alternate to the

transit time method. However, a correction factor is required for the portal vein.

184

Elastography

We have developed a protocol for the assessment of the hepatic stiffness using the MRI with collaboration with the Mayo Clinic team.

We wish to acknowledge Richard Ehman, Robert Grimm, and Meng Yin from the Mayo Clinic for their guidance during the development of the protocol. A summary of the protocol is provided in the appendix.

185

u CT

Volume evolution and density

A table (Table 30) summarizing the hepatic volume evolution is presented below

Table 30: The liver volume at each experimental time point for all groups

Volume N Mean 95% CI SD Minimum Maximum

Before 28 992.4 917 – 1067 185.7 684 1306

Residual 28 199.5 165- 234 85.8 48 390

Volume-D1 6 400 174 – 626 91 295 456

Volume-D3 11 567.8 482- 653 127 375 740

Volume-D7 6 631.5 521- 742 105 495 795

186

Table 31: The liver volume at each experimental time point for each resection %.

Resection % 75% 90%

Liver volume in CT scan N Mean 95% CI N Mean 95% CI

Liver-V 22 971.9 880 - 1062 6 1056.5 852- 1261

Residual-V 22 236.8 198– 265 6 92.9 50.5- 135

Volume-D1 6 407 173 – 626 0

Volume-D3 10 587 504- 669 1 375

Volume-D7 6 631.5 521 – 741 0

187

Figure 88 Box-plot with dots representing the hepatic volume evolution in both groups. Liver-V is the whole liver volume, V is volume and D is day after surgery.

188

Figure 89: Boxplot with dots representing the rate of hepatic mass recovery following resection in both resection %.

189

Figure 90: Axial CT scan image depicting the planned resected volume (blue), and the planned residual volume (pink) for 75% resection.

Figure 91: Axial CT scan image depicting the recovered hepatic volume at day 3 following 75% resection (pink) and the spleen in blue.

190

Figure 92: Axial CT scan image depicting the residual hepatic volume after a 90% resection (pink). The estimated volume is 114 cm3.

Figure 93: 3D reconstruction image at day 3 following a 90% resection with portal ring placement. The image shows the portal ring induced constriction of the portal vein and the considerable volume regain.

191

Figure 94: Coronal CT scan image at day 3 following a 90% resection with portal ring placement. The image shows the portal ring with the induced constriction of the portal vein and the considerable volume regain. The estimated volume is 387 cm3.

Figure 95: Axial CT image at day 3 following a 90% resection with portal ring placement. The image shows the considerable volume regain. The estimated volume is 387 cm3, which is more than 3 times the residual size.

192

Surgical anatomy and implications

Prior knowledge of the anatomical variation of the liver is fundamental in order to perform safe and efficient surgical technique. This

study is the first to comprehensively describe the CT anatomy of the porcine liver. Beside the prominent fissures in pigs, there are

several other differences between the macro-anatomical features between the human and porcine liver. The supply of the caudate lobe in humans comes mainly from the left hepatic artery and main or the left portal vein while in pigs it comes from the right side (231).

The portal venous supply to the caudate lobe was reported to come consistently from the right portal vein in the study by Court (323),

however, this was not the case in our study. We found that in the majority of animals a single portal branch from the main portal trunk

is the main supplying branch to the caudate lobe and in a lower frequency from the right branch supplying the right lateral lobe.

Furthermore, there was not obvious right and left divisions of the portal vein. We observed a long trunk from which consecutive

branches come out.

One of the multiple similarities of the pattern of the blood supply is at the left lateral sector, where the portal vein runs mostly into the

fissure separating this sector from the left medial sector and gives out two branches to the left lateral sector and two to the left medial

one. This is similar to the pattern of the left portal vein divisions in humans (232).

Variations of the arterial branching were reported in a study by Gravante and coworkers (226). They did not report however the consistent branch that leaves the proper hepatic artery to supply the stomach. This artery might be mistaken as a hepatic artery if not

differentiated. In our study, we found that the anterior branch supplying the right lateral sector is the consistent one and that the

posterior branch was present in the majority of animals. This is In contrast to what was reported by court (223), who reported that the anterior artery is the accessory one.

193

Noteworthy, the discrepancy in the reported anatomy of swine originate from the differences in species. Zanchet et al (233) reported

only two common drainage sites for the supra-hepatic veins into the IVC of Landrace pigs. In our study, the usual pattern was three

opening joining the IVC. The two drainage sites were also encountered. However, the venous trunk of the left medial sector frequently

joins separately the IVC.

The hereby mentioned pattern of divisions and arrangement has some technical implications. In liver resection model, 70 and 90% models are most commonly used to study the hemodynamic changes induced by resection (234). These models generally aim to study the behavior of small for size syndrome. However, in order to successfully carry out the intended experiment, an accurate knowledge of the anatomy should be acquired. These percentages correspond to removal of the left three and four sectors respectively. The pattern of vascular array and the deployment of the hepatic sectors around the IVC facilitate the resection of the more peripheral sectors. For that reason, liver resection in pigs is left oriented (235).

In a model of 90% resection, the remnant liver parenchyma is the caudate lobe. The portal venous supply of this part comes directly from the main portal trunk in most of the cases, however, not unfrequently, it originates from the portal vein supplying the right lateral sector. This division occurs close to the base of the fissure separating the right lateral from the caudate lobe (236). In such case, resection should be performed not to compromise this supply, hence should be conducted above 1 cm distance from the base of the fissure depending on the employed technique.

Moreover, the percentage of contribution of each sector to the whole liver volume varies from animal to another. The importance of this becomes evident with the increment of the percentage of resection. For 90 % resection, which is performed leaving only the caudate lobe in place. The caudate lobe might constitute to as low as 4% of the total liver volume, which might impose inaccuracies and increased risk of undue mortality.

194

The porcine liver is reported to be divided into 5 different lobes by 4 fissures (223,226). In our opinion, sectors rather than lobes is a

more appropriate term to the anatomical characteristics of the porcine liver that retains, in general, similar inflow and outflow pattern

to the human liver (237).

u Fluoptics (confidential at the time of writing)

At the moment, the results related to this work remains confidential.

j. Software Applications

· Methematical Models and Potential Computer Applications (Tiquant pipelines)

This freeware is designed to quantify the changes in the individual components of the hepatic microarchitecture from images with fluorescent staining captured from photon microscopy. Algorithms based on the data extracted from this project were used extensively to calibrate the model algorithm in order to obtain robust quantification of the architecture of the porcine liver.

k. Flow Modulation Device

u Modulating portal hemodynamics with vascular ring allows efficient regeneration after partial hepatectomy in a porcine model

ABSTRACT

Objective: to investigate safety and efficacy of temporary portal hemodynamics modulation with a novel percutaneously adjustable vascular ring (MID-AVRTM) onto a porcine model of 75% hepatectomy.

Background data:

195

Post-operative liver failure is a leading cause of mortality after major hepatectomy. Portal flow modulation is an increasingly accepted concept to prevent postoperative liver failure. Nonetheless, the currently strategies have shortcomings.

Methods

Resection was performed under hemodynamic monitoring in 17 large-white pigs allocated into two groups. 8-pigs had ring around the

portal vein for 3 days with the aim of reducing changes in hemodynamics due to hepatectomy. Analysis of hemodynamics, laboratory,

and histopathological parameters was performed.

Results

Percutaneous inflation, deflation and removal of the MID-AVRTM were safe. Two (25%) pigs in the MID-AVRTM group and 4 (45%) controls died before day-3 (p=NS). A moderate increase of portal flow rate per liver mass after resection was associated with better survival (p=0.017). The portocaval pressure gradient was lower after hepatectomy in the MID-AVRTM group (p=0.001). Postoperative

serum bilirubin levels were lower in the MID-AVRTM group (p=0.007 at day-5). In the MID-AVRTM group, the Ki67 index was

significantly higher at day-3 (p=0.043) and the architectural derangement was lower (p<0.05). Morphometric quantification of the bile

canaliculi revealed a significantly lower number of intersection branches (p<0.05) and intersection nodes (p<0.001) at day-7 compared

to the preoperative specimen, in the control group. These differences were not found in the ring group.

Conclusion

MID-AVRTM is safe for portal hemodynamics modulation improving liver regeneration, protecting the microarchitecture.

196

Background

The liver function is dependent on the integrity of the micro-architecture that permits optimum exchange of metabolites between blood and hepatocytes (324). After resection, the portal flow through the sinusoidal network increases leading to stimulation of sinusoidal endothelial cells (295) and initiation of liver regeneration (173). Major hepatectomy or transplantation of small liver are, on the other hand, associated with disequilibrium between the portal flow rate, which is excessively increased at the sinusoidal level, and the liver volume, leading to “barotrauma” (188,325).

The volume and the quality of the future liver are important determinants for this disequilibrium (326,327). Thresholds for portal flow rate per 100 g of liver (328) and portal pressure (200) were identified above which the risk of post-operative liver failure is high (88).

This phenomenon is partially interplayed by the important reduction in the arterial flow “de-arterialization” of the remnant liver as a consequence of the excess in portal flow rate through the sinusoidal network (181).

Healthy remnant liver volume superior to 20% of the theoretical total liver volume (329) and/or superior to 0.5% of the body weight is considered mandatory after major hepatectomy (330,331) to keep a balance between volume and flow to avoid post-operative liver failure. If more resection is anticipated, which is the case in many liver malignancies, preoperative portal vein occlusion, by embolization or ligation, might be necessary to induce regeneration of the future liver (332). However, this approach may increase the risk of cancer progression (142,333).

Portal flow modulation was initially applied in living donor liver transplantation (334). Partial portal flow diversion (335), splenic artery ligation (192) or splenectomy were proposed to reduce the incidence of post-operative liver failure (174,336). These techniques do not allow precise control of the portal flow rate and might have adverse effects. For instance, excessive diversion of flow might have an equally deleterious effect on liver regeneration (174), to which it is essential (173). Therefore, it would be helpful to use a

197

modulation technique with flexible and reversible control over the portal hemodynamics to tailor it to the planned remnant volume. In

this field, a theoretical paper had postulated that portal vein banding device could prevent post-operative failure (337).

Towards that end, we developed an adjustable vascular ring “MID-AVRTM” to protect the hepatic microarchitecture from the initial

harmful barotrauma. The efficacy of the ring was assessed in terms of survival, liver function tests, liver regeneration, and changes in the micro-architecture.

Methods

Ethical approval: The study was approved by the regional committee of ethics of animal research, and by the French Government authorities, complying with the European Union Directive N° 2010/63/EU.

Animals: All animals received humane care according to the criteria outlined in the “Guide for the Care and Use of Laboratory

Animals” prepared by the National Academy of Sciences and published by the National Institutes of Health (338). Seventeen large white female pigs, which underwent 75% liver resection, were randomised in blocks into two groups. The control (no MID-AVRTM)

group included 9-animals and the ring group (MID-AVRTM) included 8-animals in which the vascular ring was positioned around the

portal vein. The average age of the included animals was 3months ±9days and their mean weight was 32.9±5.3 kg. Study setting:

Surgeries were performed at the experimental animal surgical unit at the Marie Lannelongue Center, Le Plessis Robinson, France.

Preoperative preparation: Animals were left fasting the night before surgery. On the day of surgery, animals were given in their

individualized cages 30 mg/kg ketamine (Ketamin, Panpharma) and 0.03 mg/kg acepromazine (Calmivet, Vetoquinol, France).

Anaesthesia: All surgeries were performed under general anaesthesia. Each pig received 100 mg of xylazine 2% (Rompun, Bayer

Healthcare) with750 mg ketamine for anaesthesia induction followed by tracheal intubation (6-7 mm in size, Portex, France).

198

Subsequently, inhalational anaesthesia was started using a 60% FiO2 with 2% isoflurane (Isoflurane, Belamont, France) in assisted ventilation.

Pancuroniumbromide (Pavulon, Schering-Plough), at a rate of 0.3mg/kg/h and fentanyl (Fentanyl Janssen 100µg/2ml), at a rate of

5μg/kg/h were continuously perfused intravenously. Crystalloid fluids were given at a rate of 2ml/kg/h-fasting in addition to 500-

1000ml, which was increased as required. At the end of surgery, the wound was infiltrated with ropivacaïne 150mg (Naropeine,

AstraZeneca, 7.5mg/ml).

During surgery animals were covered with heat blankets and gastric aspiration through an oro-gastric tube was attempted if gastric distension was observed.

Cefotaxime 1g (Cefotaxime, Mylan) and gentamicine 80mg (Gentalline, Schering-Plough) were given intramuscularly once a day for

5-days. In addition, animals received Pantoprazole 40mg/i.v (Inipomp, Nycomed) and enoxaparine 0.2 ml/S.C (Lovenox, Sanofi

Aventis) and 0.5mg/kg/b.i.d Nalbuphine (Nalbuphine Serb, 20mg/2ml).

The same preoperative measures were repeated at the time of sacrifice on the 7th postoperative day. Blood samples were collected before and after liver resection as well as on the 3rd, 5th, and the 7th postoperative days.

Surgical procedure:

Hemodynamic measurements: Median cervical incision and cannulation of the right internal jugular vein with an 8 Fr and the right carotid artery with a 5 Fr Desivalve (Vygon, Ecouen, France) vascular cannula were performed.

199

Sternotomy was performed to place the TranSonic 20 mm (20 PAX, TranSonic, Ithaca, NY, USA) transit time echo probe around the origin of the ascending aorta for measuring the cardiac output. Subsequently, a midline abdominal incision was performed. Upon

dissecting the hepatic hilum, two other flow meter probes, 14 mm and 4 mm, were positioned around the portal vein and the hepatic

artery; respectively. Portal and vena-cava pressures were measured by direct puncture with a 24 G needle connected to a built-in

electronic transducer in the anaesthetic monitor.

Flow per unit mass was calculated as the recorded flow rate in the main portal vein divided by the liver weight multiplied by 100

(ml/minute/100 gram of liver tissue), and the whole liver weight was estimated based on the fact that the left lateral, the left medial

and the right medial lobes of the pig liver constitute around 75% of the whole liver weight (230). The whole liver weight is calculated

as the resected liver weight *100/75.

MID-AVR™ positioning: The ring is silicone made (Figure 96 a,b); connected to a regulating valve via long tube. The two lips of the ring opening were fixed together with a fine non absorbable (polypropylene 8/0) suture that is readily broken upon over inflation of

200

the balloon (Figure 96c). The ring was placed around the portal vein (Figure 96d) in the designated group and calibrated before starting liver resection.

Figure 96: The MID-AVR™ in its different shapes according to the degree of balloon inflation. (A) the ring is closed while the balloon is completely deflated, (B) the ring is closed while the balloon is inflated with small amount of saline, (C) the ring tends to open while the balloon is inflated with large amount of saline, and (D) The MID-AVR™ is placed around the portal vein and the balloon on the inner surface is inflated with a small amount of saline, the portal vein shows moderate constriction.

Once the ring was placed, the balloon was progressively inflated with sterile saline solution with 0.1 ml steps. At each step the flow

rate in the portal vein as well as the pressure below and above the ring were measured during 4-5 minutes to ensure the stability of the

effect. The target portal flow rate in the MID-AVR™ group was 50% of its initial value, to limit the increase in the portal flow per the

remnant liver mass after a 75% liver resection to around twofold (207).

201

At the end of surgery, the valve was fixed subcutaneously on the xiphoid process with non-absorbable sutures for percutaneous control

and extraction.

Liver resection: Resection of the left lateral, left medial and the right medial hepatic lobes was done leaving in place the right lateral

and the caudate lobes. At the end of the procedure, a central venous catheter was placed in the internal jugular vein for postoperative

fluid administration and blood samples withdrawal. Biopsies from the remnant liver lobe were taken before and one hour after liver

resection. Pleuro-mediastinal suction drain was placed at the end of the procedure and removed on the first postoperative day.

Fifteen-minute indocyanine green retention% (ICG-R15): After resection, ICG-R15 was measured in serum from arterial samples taken before injection of 0.5 mg/kg ICG (Infracyanine, SERB, France) and 2, 4, 8 and 15 minutes after the injection.

Day-3 postoperative: An ultrasound guided liver biopsy using an 18 Fr needle was taken. In the MID-AVR™ group, removal of the ring was performed by reopening the midline incision in the first three animals, in order to ensure that the ring was completely open and that the anchoring sutures ruptured. Subsequently, it was removed through a small percutaneous incision over the valve.

Sacrifice: On the 7th postoperative day, animals were sacrificed following a similar protocol to that at the day of surgery. The remnant

liver was weighted after euthanasia.

Histological analysis, proliferation index, and 3D morphometric quantification of the bile canaliculi: The protocol for this analysis was detailed the histopathological section earlier (chapter xi).

202

Results

The experimental design of this group was set to study the effect of MID-AVRTM application; compared to no-application; on liver regeneration after 75% resection in two groups of large white pigs. Hemodynamic, laboratory, and histophalogical parameters were analysed at different time points as indicated in the methods. Both groups had similar baseline parameeters (Table 32).

Ring safety: Application of the ring was easy in all animals and its removal was safe in all animals. In the first three animals, removal under visual control demonstrated the efficiency of the over inflation to open the ring. Percutaneous removal in the subsequent animals was similarly safe and successful. Patency of the portal vein was confirmed in the surviving animals by ultrasound at day-3 and by direct visualization at sacrifice.

Survival: Six pigs (75%) in the MID-AVR™ group and 5 pigs (55.6%) in the control group survived till day-7 (p=0.62). However, when portal flow per unit liver mass at the end of surgery was within the range of 2.2-4 folds its baseline, only one animal died prematurely (10%). On the other hand, 5 animals (71%) died when these values were outside the range (p=0.017) (odds ratio=22, p=0.02) (Figure 97). Autopsy was performed for all premature deaths, and it was negative for macroscopic explanation for the mortality. None of the pigs showed portal vein thrombosis.

Hemodynamic measurements: The hemodynamic measurements at the end of hepatectomy are presented in (Table 33). In the ring group, the porto-caval pressure gradient was significantly lower than in the control group (p<0.01) (Figure 98a). In a stepwise multiple regression analysis, only the presence or absence of the ring was a significant predictor for the change of porto-caval pressure gradient

(F=16, p=0.001) whereas the systemic hemodynamic changes were not significant predictors. There was no difference in the pressure gradient at the post-operative day-7 between the two groups (i.e. 4 days after ring removal).

203

Changes in liver weight: The estimated residual liver weight was not significantly different between the control and the ring groups

(155 ± 32 gram and 152 ± 25 gram, respectively, p=1). The residual liver weight increased significantly at day-7, p<0.0001 (520 ± 90

gram and 478 ± 37 gram in the ring and the control groups respectively) (Figure 98b). There was no significant difference in the liver

weight between both groups at the day of sacrifice, but the gain was estimated to be slightly higher in the ring group compared to the

control group (mean difference= 353±68 ml versus 323±36 ml, respectively).

Laboratory results: The MID-AVRTM was a significant influential factor on the bilirubin level in repeated measures ANOVA

(p=0.033). Postoperatively, serum bilirubin level was lower in the MID-AVRTM group, particularly at day-5, than in the control group

(3.8 vs 6.6 µmol/L, p=0.007) (Figure 99a). Prothrombin activity (Figure 99b), was slightly higher after hepatectomy in the ring group

(p=ns). For the rest of the parameters, no differences were observed. Table 33 summarizes the main findings after resection and at sacrifice.

Indocyanine green retention%: [ICG-R15] was not significantly different between groups. However, the average retention% was slightly lower at day-7 in the ring group compared to the control group (Figure 99c).

Histopathological results: The mean pathological scores were significantly better in the ring group at the end of hepatectomy and at day-3 and day-7 (Figure 100a-f) (p<0.05, p<0.01, p<0.05, respectively). The Ki67 positive cells increased by 14.3% (95% CI= 10.8-

17.9%) at post-operative day-3 compared to before surgery and decreased again at day-7 with a difference in the activity of 1.5%

(95% CI=-2.8 to -0.2%) compared to baseline. Between the two groups the Ki67 positive cells was higher at day-3 (p=0.043) in the ring group and slightly higher in the control group at day-7 (p=NS) (Figure 101 a,b and Figure 102).

In 3D morphometric quantification of the biliary structures demonstrated that the number of intersection branches and the number of intersections nodes were significantly lower in the control group (but not in the ring group) at day-7 compared to the values before

204

resection (mean difference= 162, 95% CI=44-279 for intersection branches and mean difference=130, 95% CI=13-248 for intersection nodes, [Table 34]). The cumulative length of bile canaliculi was significantly lower at day-7 than in preoperative specimen in both groups (p=0.02) (Figure 101c,d). There was a non-significant higher density of biliary structures in the ring-group compared to the control group at day-7 (supplemental movies 2-4 (http://1drv.ms/1nIX0rW)).

Figure 97: Scatter diagram showing animal mortality in both groups stratified according to the change in portal flow per unit liver mass.

205

Figure 98: (A) porto-caval pressure gradient was significantly higher in the control group compared to the MID-AVRTM group only after hepatectomy (**p < 0.01) (B) the liver weight before and after resection and the regain at day-7 were not significantly different between both groups.

206

Figure 99: (A) total bilirubin level was higher in the control group than in the MID-AVRTM group, particularly at postoperative day-5 (*=p < 0.05), (B) prothrombin activity is not significantly different between both groups, (C) the indocyanine green retention test (ICG) at 15 minutes was, after resection or at day-7, similar in the control group and in the MID-AVRTM group.

207

Figure 100: Biopsies taken from the control (left panel) and ring (right panel) groups (A) H&E, x200, day-7 after surgery, bridging necrosis, thick cords forming pseudonodules, (B) H&E, x400, day-7 after surgery, cholangioductal proliferation, (C) Trichrome, x100, day-7, dilatation and hemorrhagic destruction of the sinusoids with surrounding thick parenchymal cords, (D) H&E, x200, day-7, Normal architecture, (E) H&E, x400,

208

day-7, normal aspect portal pedicle with mild arterial dilatation, and (F) Trichrome, x100, day-7, conserved architecture with normal thickness cords and portal spaces.

Figure 101: Analysis of hepatocyte proliferation in regenerating pig livers and reconstruction and quantification of bile canalicular network in regenerating pig livers. (A) There are increased numbers of Ki-67 labelled hepatocyte nuclei at day-3 compared to day-0, which decreased again at day-7 (Scale bars are 50 µm). Overview images are included in Figure 76. (B) Quantification of Ki-67 positive hepatocyte nuclei in five fields per specimen. Five to six pigs were used per time point, namely, 0, 3 and 7 days after partial hepatectomy. (C) Examples of reconstructed pig livers. Blue (left column): nuclei, green (middle column): bile canalicular network and merge (right column). The corresponding reconstructions of control, 7 days after partial hepatectomy with MID-AVRTM and 7 days after partial hepatectomy without MID-AVRTM are shown in

209

Supplemental movies 2-4 (http://1drv.ms/1nIX0rW), respectively. Scale bars are 50 µm and (D) Length of bile canalicular network in a given volume is shown. The bile canaliculi length is not influenced by application of MID-AVRTM ring. *p < 0.05 and **p < 0.01.

Figure 102: Analysis of hepatocyte proliferation index in regenerating pig liver. Ki-67 proliferation index from day-0, day-3 and day-7 with and without application of the adjustable vascular ring (MID-AVRTM). Scale bars are 100 µm.

210

Table 32: Pre-hepatectomy data from both groups

MID-AVR™ Group Control Group (n=9) p (n=8)

Weight (kg) 30.9 ± 5.2 32.8 ± 5.6 ns

Resected liver weight (g) 454 ± 76 465 ± 96 ns

Remnant liver weight (g)* 151 ± 25 155 ± 32 ns

Portal flow (ml/min) 852 ± 318 793 ± 312 ns

Portal flow per unit liver mass 142 ± 57 130 ± 52 ns (ml/min/g*100g)

Arterial flow (ml/min) 222 ± 76 185 ± 81 ns

Cardiac output (l/min) 2.5 ± 0.7 2.6 ± 0.7 ns

Mean arterial pressure (mmHg) 60 ± 4 60 ± 9 ns

Central venous pressure 5.2 ± 1.4 6.4 ± 1.5 ns (mmHg)

Portal pressure (mmHg) 7 ± 0.8 8.4 ± 1.6 ns

Porto-caval gradient (mmHg) 1.75 ± 1.3 2 ± 1.2 ns

Total bilirubin (µmol/l) 4.6 ± 3.4 4 ± 2.5 ns

211

Direct bilirubin (µmol/l) 0.75 ± 1.4 1.25 ± 0.7 ns

Prothrombin actvity (%) 107 ± 4 107 ± 7 ns

ASAT (UI /ml) 34 ± 29 25 ± 3 ns

ALAT (UI/ml) 52 ± 9 46 ± 6 ns

Platelets (X 103/µl) 272 ± 84 287 ± 129 ns

% 15 min ICG*** retention 23 ± 11 19 ± 8 ns

212

Table 33: Different parameters measured after liver resection and on day-7 post-operative

MID-AVR™ Group (n=8) Control Group p (n=9)

Portal flow (ml/min) 565 ± 164 544 ± 122 ns

Portal flow for remnant 372 ± 97 359 ± 95 ns liver mass (ml/min/g*100g)a

Arterial flow (ml/min) 92 ± 49 64 ± 60 ns

Cardiac output (l/min) 2.6 ± 0.8 2.3 ± 0.9 ns

Central venous pressure 5.9 ± 2.0 4.9 ± 1.6 ns (mmHg)

Portal pressure (mmHg) 7.5 ± 2.4 9.3 ± 2.3 ns

Porto-caval gradient 1.63 ± 1.3 4.4± 1.5 0.001 (mmHg)

ASAT (UI /ml) 169 ±60 197 ±110 ns

ALAT (UI/ml) 37.3±8.1 42 ± 6 ns

Platelets (X 103/µl) 348±101 312±31 ns

Analysed parameters at animal sacrifice

213

Animal weight (kg) 30 ± 3.8 30 ± 6.0 ns

Liver weight (g)b 507 ± 77 478 ± 37 ns

Portal flow (ml/min) 710 ± 224 698 ± 190 ns

Portal flow per unit liver 149 ± 54 93 ± 79 ns mass (ml/min/g*100g)

Hepatic arterial flow 151 ± 61 93 ± 79 ns (ml/min)

Cardiac output (l/min) 2.7 ± 1.2 2.8 ± 1.3 ns

Central venous pressure 4.7 ± 1.7 4.8 ± 1.1 ns (mmHg)

Portal pressure (mmHg) 8.2 ± 1.3 9.4 ± 0.9 ns

Porto-caval gradient 3.5 ± 1 4.6 ± 1.7 ns (mmHg)

% 15 min ICG retention c 28 ± 18 38 ± 7 ns

Platelets (X 103/µl) 375±30 435±133 ns

a The weight of the remnant liver was estimated based on the fact that resection of the left and median lobes is nearly equal to 75%. b Whole liver weight is the sum of the resected liver weight and the calculated liver weight. c ICG= Indocyanine green

214

Table 34: List of parameters quantified in 3D for bile canalicular network. Data are in means ± standard deviation of 6-9 stacks per group.

Bile canalicular (BC) morphometrya Before resection MID-AVR™ Control

0d 7d 7d

1 Number of dead end branches /mm3 427.95±111.79 395.74±118.67 391.21±70.00

2 Number of intersection branches 798.27±233.43 693.14±203.32 635.92±141.74** /mm3

3 Number of intersection nodes/mm3 714.01±186.63 620.98±175.69 583.45±115.77 *

4 Dead end branch length [µm] 7.54±1.14 7.38±1.12 7.50±0.71

5 First order branch length [µm] 7.04±0.88 6.44±1.48 7.20±1.03

6 Second order branch length [µm] 10.61±1.65 9.65±2.65 11.58±2.27

7 Number of branches per 3.01±0.009 3.01±0.005 3.01±0.004 intersection node

8 Number of BC nodes (x106)/mm3 19.21±1.99 15.72±3.08 16.28±2.19

9 BC volume in % of mm3 liver 1.48±0.52 1.23±0.31 1.11±0.26 tissue

*P<0.05 and **P<0.01

a The sketch below defines the number of dead end branches/mm3 (1), number of intersection branches/mm3 (2), the number of intersection nodes/mm3 (3), dead end branch length (4), first order branch length (5), and second order branch length (6) (7-9 are not shown on the schema)

215

Discussion

This present study presents the effects of a novel, adjustable, and less invasive technique for portal hemodynamics modulation on liver

regeneration after 75% hepatectomy in a porcine model. The application of the ring was safe and associated with better hepatic

function represented in the lower bilirubin levels during the 5th post-operative day as well as in the less histological derangement and

changes in the temporal proliferation pattern indicated by a higher Ki67 index at day 3 while the liver mass at day 7 in both groups did

not differ.

The proliferative activity was higher in the ring group at day-3 and returned to normal at day-7. Together with the higher pathological scores these results indicate that the control group had a temporally different cellular proliferation and a deranged micro-architecture that could be attributed to the effect of barotrauma. Sinusoidal barotrauma is thought to be a leading mechanism of post-operative liver

failure (192,325). In our study, each one of the animals in the ring group with their portal flow within this range of 2.2-4 folds

survived.

In this study, the majority of animals at the end of surgery in both groups had an estimated portal flow per unit mass within the target

range. However, the ring group showed lower portocaval pressure gradient, which indicates that the intra-sinusoidal pressure was

lower in ring group. Indeed, the flow in ring group after ring positioning does not reflect the intra-parecnhymal resistance or pressure

since it reflects the resistance imposed by the ring positioning, unlike the case in the control group. The parenchyma in the ring group

continues to receive a lower flow compared to the control group, which leads to lower intra-sinusoidal pressure, smaller portocaval

pressure gradient and lesser architectural damage. Therefore, we conclude that the ring application helped in protecting the hepatic

microarchitecture in the initial phase after resection.

The higher porto-caval pressure gradient, which is a manifestation of increased microvascular resistance (339), explains the more

evident changes in the microachitecture in the control group despite that there was no difference in the calculated portal flow per unit

mass. Furthermore, the higher bilirubin and ICG-R15 level, which are typical manifestations of portal hyperperfusion (213), could be

216

attributed to the higher intra-sinusoidal pressure in this group. Alone; the high bilirubin level was reported to be a sufficient laboratory feature for the diagnosis of post-hepatectomy liver failure (191). The aforementioned changes imply that the better hepatic function in the ring group is attributed to the maintained microarchitectural integrity in this group.

The relation between the portal pressure and the portal flow rate is not strictly linear owing to the interplay of the hepatic sinusoidal capacitance, which means that it would usually require more than doubling of the portal flow through the sinusoidal network for the portal pressure to slightly increase (183). Given the capacitance of the liver as a reservoir for blood (340), the absence of increased porto-caval pressure gradient in the ring group reflects the lower intra-sinusoidal pressure in spite of the relatively higher portal flow calculated from flow rate proximal to the ring in this group.

An increase in portal flow per liver mass is thought to be necessary for the stimulation of liver regeneration (174) and is associated with increased cellular proliferation through proliferative gene expression and apoptotic gene down regulation (341). On the contrary if there is an excessive increase in portal flow, liver functions are deranged (206) and suppression of liver regeneration paradoxically dominates, leading to liver failure (179). The liver weight gain was lower in the control group than in the ring group. Furthermore, in the 3D quantification of biliary structures, all morphometric parameters had lower values in both groups compared to specimens taken before resection, indicating a higher hepatocyte density, as supported by the interpretation of the pattern of Ki67 index. . These parameters were higher in the ring group than in the control group implying that the regeneration in the ring group was more balanced and organised, which might explain the better hepatic function in this group.

Hepatic inflow modulation is becoming an increasingly accepted strategy for the reduction of the initial damage caused by the small for flow syndrome (342,343). It targets the early protection of the architecture from the barotrauma inflicted by high portal flow during the inductive angiogenesis phase (295). Several techniques have been proposed for modulation of the portal inflow. Diversion of the portal flow through a partial porto-caval shunt would require a second intervention for the closure of the shunt (344).

Splenectomy offered more benefit compared to the portal flow diversion in terms of liver regeneration (343). This could be attributed to the negative effect of the excessive diversion on regeneration (345). Splenic artery embolization, as an alternative, exposes the

217

spleen to the risk of infarction (346). Those alternatives do not have the dynamic and minimal invasive advantages offered by the

reversible modulation with an adjustable vascular ring, which could be tailored to the extent of the planned liver resection.

In this study, the application of the ring prevented the increase in porto-caval pressure gradient and allowed a moderate increase in portal flow per unit liver. This effect resulted in protection of the remnant parenchyma without compromising liver regeneration.

However, comparing the two groups, only a few parameters reached the statistical significance.

The full potentials of this novel technique could be more evident in a situation where the increase in portal flow per unit mass is higher. Here, the goal was to first test the ring safety in a non-extreme situation to have a better chance to see its effect at day-3 and day-7. In spite of that, the present study demonstrates the high safety profile and the potential efficacy of the MID-AVRTM. Therefore,

we started a human clinical trial (phase I/II) registered at (clinicaltrials.gov) under the number (NCT02390713).

Conclusion

The adjustable vascular ring “MID-AVRTM” applied around the portal vein is a safe, precise, reversible and efficient mean to protect

the hepatic micro-architecture during the initial phase of liver regeneration. It promotes liver regeneration in a preserved

microarchitecture environment which might result in better hepatic function over the course of regeneration.

u The potential of preventing lethal consequences of near total hepatectomy using a novel portal flow modulation device

Summary of the experimental group and highlights of the main findings:

Extensive liver resection for radical surgery could be lethal. The remnant liver might fail to accommodate the excessive barotrauma

induced by the shift of portal flow (188). The condition is associated with disruption of the microarchitectural harmony (193), leading

to suppression of hepatic regeneration (196), which would normally occur following resection due to a modest increase in the portal

flow to the remnant liver (174).

218

Tailoring the increase in the portal flow was effective in improving the regeneration in both clinical (343) and experimental settings.

We previously demonstrated the potentials of the “MID-AVRTM” to ameliorating the architectural layout with the improvement of the clinically relevant endpoints in a 70% resection model (205). However, the 70% resection model was not challenging enough – given the little impact of this resection on the hemodynamic in this model – to examine the potentials of this device.

We, therefore, have designed a smaller scale exploratory study to test this device in a near total resection model. After ethical approval, six large white pigs were allocated to subtotal hepatectomy – leaving only the caudate lobe surrounding the IVC – following the anatomical guidance described elsewhere (287). A CT scan volumetry was performed to all animal before surgery and

immediately after, then at day-3 post-operative. Three animals had the MID-AVRTM – around the portal vein before the hepatic

resection – adjusted to reduce the portal flow to the remnant by approximately 50% of its initial value. The target is to deliver

approximately an increase in the portal flow to the remnant liver mass of around 5 times the baseline (347). The other three animals

did not have the ring, while both groups had the resection as described previously (286).

Biopsies were taken before and 1 hour after the resection to examine the impact of the surgery on the microarchitecture in standard

Hematoxylin-Eosin and Masson trichrome stains based on criteria adapted from Demetris et al. (193). Five criteria were formulated:

A= presence of inflammatory infiltrate and necrosis [sub-items: neutrophil infiltrate, lymphocytes infiltrate, and hepatocellular

necrosis], B=sinusoidal dilatation, C= ductular proliferation, D= steatosis [sub-items: macrovesicular/ballooning, microvesicular], and

E=regeneration [sub-items: pseudo-nodular, mitotic activity, and acinar formation]. Each item was scaled from 0=no change, 1=mild,

2=moderate, to 3=severe, in samples taken from deep liver tissue for a total score between 0 and 30 at each time point.

Hemodynamic parameters were recorded and analyzed for the pre- and post-resection phases. Installations were described in a

previous study (286). Blood samples were withdrawn and handled according to our previous report (348). One animal in the ring

group survived, while the other two died prematurely during the first postoperative day. There was a malfunction of the balloon

mechanism in those two animals. The first one was not seen intraoperative and was discovered during the autopsy, while the second

219

one was seen intraoperative but was thought to be managed by ligation, which seemed to be ineffective during the autopsy. All

animals in the no-ring group died on the first postoperative night.

The reduction in the central venous pressure was significantly greater in the group with the ring compared to the no-ring group

(median= 238% versus 82%). On the other hand, there was a non-significant larger increase in the portal venous pressure in the no-

ring compared to the ring group (median= 149% versus 139%). The portocaval pressure gradient increased significantly more in the

no-ring group compared to the ring group (median= 411% versus 133%). The portal vein flow was reduced in both groups (median=

47% in no-ring versus 44% in the ring group), while the hepatic artery flow reduction was larger in the no-ring group (median= 82%

versus 66%). The hepatic artery pressure decreased more in the no-ring group (median= 13% versus 8%). The damage scoring

increased significantly in both groups (p=0.017) (Figure 103). However, no statistical difference between groups was observed albeit

the tendency for a larger increase in the scores in the no-ring group (Table 35).

Table 35: The microarchitectural damage score for both groups.

No-Ring Ring

Damage score Median Percentile 25 Percentile 75 Median Percentile 25 Percentile 75

Before 2.00 2.00 3.00 1.00 1.00 2.00

After 4.00 4.00 5.00 2.00 2.00 4.00

Day 1 . . . 5.00 4.00 6.00

Day 3 . . . 5.00 5.00 5.00

220

Figure 103: Evolution of the damage score after the hepatic resection in both groups. The change was significant as estimated by Wilcoxon test, but the percentage of change between the groups was not significant.

The estimated median resected volume was 87% with a 95% CI of the median= 80 to 94%. The surviving animal showed a volume increase of

167% by the third post-operative day (Figure 104).

221

Figure 104: CT scan volumetry of the hepatic remnant. a) immediately post resection, showing the ring around the portal vein (indicated by the arrow: the white ring is the band, and the space between it and the portal vein is the balloon inflated), b) an operative photo of the ring positioned prior to resection with the relevant probes for monitoring of the inflow (Fa is the probe for the hepatic artery flow, Fp is the one for portal vein flow, and Pp is the pressure probe for the portal venous pressure), c) is the volume gain at day 3 in an axial CT scan image; enhanced with

222

volumetric tools, and d) is the 3D reconstruction of the liver remnant at day 3 with the constriction effect of the band on the portal vein is indicated by the arrow.

The blood analysis one hour after resection did not show any significant change in the measured parameters. However, the parameters from the

surviving animals in the ring group showed significant derangement of the liver functions consistent with the clinical picture of the unsteady gate of hepatic encephalopathy that this animal had before the euthanasia at the third postoperative day (Table 36).

Table 36: The main liver functions of the animals in both groups

Bilirubin Bilirubin Bilirubin AST ALT AST AST ALT ALT Total Total Total Day- Day- Before After Before After Before After Day-3 3 3 (U/I) (U/I) (U/I) (U/I) (µmol/l) (µmol/l) (µmol/l) (U/I) (U/I)

n. 6 6 1 6 6 1 6 6 1

No- 3.1 1.9 . 18.3 28.2 . 9.7 3.3 . Ring

Ring 5.8 6.8 38 48.8 212.3 562 46.8 31 77

This study demonstrates the potential of the innovative vascular ring to prevent fatalities in near total hepatectomy in a porcine model.

We have previously demonstrated the potentials its potentials in ameliorating the hepatic microarchitecture and function following a

major (75%) resection in a porcine model (205).

One animal in the ring group died during the first postoperative night, while a second animal was euthanized 30 hours following the

resection as judged by animal welfare onsite representative to be suffering. There was a malfunction of the balloon mechanism – leak

at the valve junction – in those two animals. The first one was not seen intraoperative, while the second one was seen intraoperative

but was thought to be managed by ligation distal to the leaking point. Both rings were found deflated at the autopsy. The third animal

was euthanized as per protocol on the third postoperative day. All animals in the no-ring group died on the first postoperative night.

The ring was effective in alleviating the triad of portal hyperperfusion syndrome (increased portal pressure, reduced portal flow and

hepatic artery flow) to the remnant liver. The portal pressure increased more in the no-ring group, while the portal flow decreased

223

more in the no-ring group indicating a relative higher intra-parenchymal resistance, which was also reflected by the higher reduction in the arterial flow in the no-ring group indicating a more significant de-arterialization in this group. The larger reduction of the central venous pressure in the ring group could be partially attributed to the reduced venous return as a result of reduced the hepatic inflow, which could have influenced the remnant blood volume and its capacitance (183), in addition to more blood loss encountered in this group. Unfortunately, the manufacturing error of the ring in two out of the three animals limited the interpretation of these results, and further studies are required to verify these findings.

The alternative techniques for portal flow modulation; such as the portal vein plication, or other methods such partial portal flow diversion or splenectomy to reduce the incidence of postoperative liver failure were reviewed in our previous report (205). But all have the disadvantage of imprecise tailoring or the irreversibility, which this ring is capable of avoiding. Ring Leak

224

Figure 105: The junction between the valve and the tube is showing small leak, but unfortunately affecting the balloon pressure and subsequently the modulation effect on the portal flow leading to loss of effect and animal mortality.

Future work. In summary, this work presents the first dynamics model of liver ablation. Its validation based on pig data is promising; the model is able to capture and explain the main features of hemodynamics changes due to the surgery as well as its variability among pigs, and may give insights about other states

(e.g. day post-surgery) when measurements are difficult to take. HABR does not seem needed to explain the data. Future work will include to adapt the liver model to human liver anatomy. Moreover, in this work, several pressure and flow measurements were available at different stages of the surgery. The pre-resection measurements were used for parameter tuning. In patient surgeries less measurements are available. Furthermore, the model and measurements in this work are for a healthy (pig) liver. However the liver of the patients treated with partial hepatectomy is generally not healthy. For example, collateral circulations can appear (349), and the liver resistance and capacitance parameters can change.

225

IV. Limitations and perspectives:

There is no human work that could be without flaws. This work is not an exception. It is of value to shed the light on some that we can see in order for others to learn and improve.

Protocol: The protocol was ambition. I would not say over ambition though. The main possible criticism of this protocol is that is tried to address several questions in the same. In reality, and despite that this could be true, it is important to understand that the protocol was designed to address the multiscale nature of events following liver resection. This should be considered a strong point rather than a weakness. Nonetheless, we have not been able to achieve all the endpoints within the allocated time.

Was it too invasive? One of the main difficulties that we have encountered during surgeries is the complex surgical setup. In order to accurately measure the gut inflow and its contribution to the portal venous flow, the setup included a flow probe to surround the arota above the renal arteries. This area was full of lymphatics surrounding the aorta that prevented clearance of the aorta without injury to these lymphatics. This injury led to significant fluid loss in the postoperative period that we believe contributed to the high mortality of the animals, notable in the 75%. There should be some tools made available to facilitate the measurements with minimal invasivity as alternative to the invasive probing. This should reduce the negative influence of the surgical manipulations and instalments.

Model: Was it the best choice? The young pigs have an important physiological changes over short period of time which might confound the interpretation of some of the data from this study.

Laboratory: It appears that the 1 hour blood testing did not show major alterations except for the coagulation profile. Given the fact that the laboratory assessment of the normal values of the coagulation parameters, the interpretation of the change was hampered.

Were they adequate to test the aim? The results of the laboratory analysis from the 90% resection which survived showed that the animals would manifest hepatic insufficiency on the standard laboratory results from the first post-operative day and not in the one- hour samples. Therefore, we think that the immediate post resection samples could be ignored in that context.

226

Histology: At the moment, the protocol for 3D analysis of the porcine liver microarchitecture is still to be standardized.

MRI: Where have we been in elastography? We could only achieve a standardized protocol to assess the hepatic stiffness with the

MRI elastography. One of the reasons was that the experimental protocol was heavy for the animals to well tolerate. So it was deemed

appropriate to omit the elastography on occasions. The result of that will be an available protocol at the research unit to future studies

addressing this issue.

227

V. References

1. Gordillo M, Evans T, Gouon-Evans V. Orchestrating liver development. Development. 2015;142(12):2094–108.

2. Matsumoto K, Yoshitomi H, Rossant J, Zaret KS. Liver organogenesis promoted by endothelial cells prior to vascular function. Science (80- ). 2001;294(5542):559–63.

3. Goldman O, Han S, Sourisseau M, Sourrisseau M, Dziedzic N, Hamou W, et al. KDR identifies a conserved human and murine hepatic progenitor and instructs early liver development. Cell Stem Cell. 2013;12(6):748–60.

4. Schmelzer E, Zhang L, Bruce A, Wauthier E, Ludlow J, Yao H, et al. Human hepatic stem cells from fetal and postnatal donors. J Exp Med. 2007;204(8):1973–87.

5. Carpentier R, Suñer RE, van Hul N, Kopp JL, Beaudry J-B, Cordi S, et al. Embryonic ductal plate cells give rise to cholangiocytes, periportal hepatocytes, and adult liver progenitor cells. Gastroenterology. 2011;141(4):1432–8, 1438.e1–4.

6. Lanzoni G, Cardinale V, Carpino G. The hepatic, biliary and pancreatic network of stem/progenitor cells niches in humans: A new reference frame for disease and regeneration. Hepatology. 2015;

7. Spence JR, Lange AW, Lin S-CJ, Kaestner KH, Lowy AM, Kim I, et al. Sox17 regulates organ lineage segregation of ventral foregut progenitor cells. Dev Cell. 2009;17(1):62–74.

8. Dorrell C, Grompe M. Liver repair by intra- and extrahepatic progenitors. Stem Cell Rev. 2005;1(1):61–4.

9. Ruebner BH, Blankenberg TA, Burrows DA, SooHoo W, Lund JK. Development and transformation of the ductal plate in the developing human liver. Pediatr Pathol. 1990;10(1-2):55–68.

10. Jungermann K, Kietzmann T. Zonation of parenchymal and nonparenchymal metabolism in liver. Annu Rev Nutr. 1996;16:179–203.

11. Yin C, Evason KJ, Asahina K, Stainier DYR. Hepatic stellate cells in liver development, regeneration, and cancer. J Clin Invest. 2013;123(5):1902–10.

12. Berg T, DeLanghe S, Al Alam D, Utley S, Estrada J, Wang KS. β-catenin regulates mesenchymal progenitor cell differentiation during hepatogenesis. J Surg Res. 2010;164(2):276–85.

13. Nagai H, Terada K, Watanabe G, Ueno Y, Aiba N, Shibuya T, et al. Differentiation of liver epithelial (stem-like) cells into hepatocytes induced by coculture with hepatic stellate cells. Biochem Biophys Res Commun. 2002;293(5):1420–5.

14. Naito M, Hasegawa G, Ebe Y, Yamamoto T. Differentiation and function of Kupffer cells. Med Electron Microsc. 2004;37(1):16–28.

228

15. Boulter L, Govaere O, Bird TG, Radulescu S, Ramachandran P, Pellicoro A, et al. Macrophage-derived Wnt opposes Notch signaling to specify hepatic progenitor cell fate in chronic liver disease. Nat Med. 2012;18(4):572–9.

16. Choi T-Y, Ninov N, Stainier DYR, Shin D. Extensive conversion of hepatic biliary epithelial cells to hepatocytes after near total loss of hepatocytes in zebrafish. Gastroenterology. 2014;146(3):776–88.

17. Michalopoulos GK, Barua L, Bowen WC. Transdifferentiation of rat hepatocytes into biliary cells after bile duct ligation and toxic biliary injury. Hepatology. 2005;41(3):535–44.

18. Molina DK, DiMaio VJM. Normal organ weights in men: part II-the brain, lungs, liver, spleen, and kidneys. Am J Forensic Med Pathol. 2012;33(4):368–72.

19. Sibulesky L. Normal liver anatomy. Clinical Liver Disease. Wiley Online Library; 2013;2(S1):S1–S3.

20. Rypins EB, Milne N, Sarfeh IJ, Lyons KP. Quantitation and fractionation of nutrient hepatic blood flow in normal persons, in persons with portal hypertensive cirrhosis, and after small-diameter portacaval H grafts. Surgery. 1988;104(2):335–42.

21. Couinaud C. Liver anatomy: portal (and suprahepatic) or biliary segmentation. Digestive surgery. Karger Publishers; 1999;16(6):459–67.

22. Krishna M. Microscopic anatomy of the liver. Clinical Liver Disease. Wiley Online Library; 2013;2(S1).

23. Bhunchet E, Wake K. The portal lobule in rat liver fibrosis: a re-evaluation of the liver unit. Hepatology. 1998;27(2):481–7.

24. Teutsch HF, Schuerfeld D, Groezinger E. Three-dimensional reconstruction of parenchymal units in the liver of the rat. Hepatology. 1999;29(2):494–505.

25. Lamers WH, Hilberts A, Furt E, Smith J, Jonges GN, van Noorden CJ, et al. Hepatic enzymic zonation: a reevaluation of the concept of the liver acinus. Hepatology. 1989;10(1):72–6.

26. Ekataksin W, Wake K. Liver units in three dimensions: I. Organization of argyrophilic connective tissue skeleton in porcine liver with particular reference to the “compound hepatic lobule.”American journal of anatomy. Wiley Online Library; 1991;191(2):113–53.

27. Teutsch HF. The modular microarchitecture of human liver. Hepatology. 2005;42(2):317–25.

28. Hoekstra LT, de Graaf W, Nibourg GA, Heger M, Bennink RJ, Stieger B, et al. Physiological and biochemical basis of clinical liver function tests: a review. Annals of surgery. LWW; 2013;257(1):27–36.

29. Child CG. Surgery and portal hypertension. The liver and portal hypertension. WB Saunders; 1964;

30. Pugh R, Murray-Lyon I, Dawson J, Pietroni M, Williams R. Transection of the oesophagus for bleeding oesophageal varices. British Journal of Surgery. Wiley Online Library; 1973;60(8):646–9.

31. Gaba RC, Couture PM, Bui JT, Knuttinen MG, Walzer NM, Kallwitz ER, et al. Prognostic capability of different liver disease scoring systems for prediction of early mortality after transjugular intrahepatic portosystemic shunt creation. J Vasc Interv

229

Radiol. 2013;24(3):411–20, 420.e1–4; quiz 421.

32. Cholongitas E, Marelli L, Shusang V, Senzolo M, Rolles K, Patch D, et al. A systematic review of the performance of the model for end-stage liver disease (MELD) in the setting of liver transplantation. Liver Transpl. 2006;12(7):1049–61.

33. Uemura H, Katsuura-Kamano S, Yamaguchi M, Sawachika F, Arisawa K. Serum hepatic enzyme activity and alcohol drinking status in relation to the prevalence of metabolic syndrome in the general Japanese population. PLoS One. 2014;9(4):e95981.

34. Boleslawski E, Vibert E, Pruvot F-R, Le Treut Y-P, Scatton O, Laurent C, et al. Relevance of postoperative peak transaminase after elective hepatectomy. Ann Surg. 2014;260(5):815–20; discussion 820–1.

35. Vibert E, Boleslawski E, Cosse C, Adam R, Castaing D, Cherqui D, et al. Arterial Lactate Concentration at the End of an Elective Hepatectomy Is an Early Predictor of the Postoperative Course and a Potential Surrogate of Intraoperative Events. Ann Surg. 2015;262(5):787–93.

36. Helmke S, Colmenero J, Everson GT. Non-Invasive Assessment of Liver Function. Curr Opin Gastroenterol. NIH Public Access; 2015;31(3):199.

37. Shrestha R, McKinley C, Showalter R, Wilner K, Marsano L, Vivian B, et al. Quantitative liver function tests define the functional severity of liver disease in early-stage cirrhosis. Liver Transpl Surg. 1997;3(2):166–73.

38. Narasaki H, Noji T, Wada H, Ebihara Y, Tsuchikawa T, Okamura K, et al. Intraoperative Real-Time Assessment of Liver Function with Near-Infrared Fluorescence Imaging. Eur Surg Res. 2017;58(5-6):235–45.

39. Bekheit M, Vibert E. Fluorescent-Guided Liver Surgery: Paul Brousse Experiences and Perspective. In: Fluorescence Imaging for Surgeons. Springer; 2015. p. 117–26.

40. Vesell ES. Noninvasive assessment in vivo of hepatic drug metabolism in health and disease. Ann N Y Acad Sci. 1984;428:293–307.

41. Ilan Y. Review article: the assessment of liver function using breath tests. Aliment Pharmacol Ther. 2007;26(10):1293–302.

42. Lock JF, Schwabauer E, Martus P, Videv N, Pratschke J, Malinowski M, et al. Early diagnosis of primary nonfunction and indication for reoperation after liver transplantation. Liver Transpl. 2010;16(2):172–80.

43. Stockmann M, Lock JF, Malinowski M, Niehues SM, Seehofer D, Neuhaus P. The LiMAx test: a new liver function test for predicting postoperative outcome in liver surgery. HPB (Oxford). 2010;12(2):139–46.

44. Kaffarnik MF, Lock JF, Vetter H, Ahmadi N, Lojewski C, Malinowski M, et al. Early diagnosis of sepsis-related hepatic dysfunction and its prognostic impact on survival: a prospective study with the LiMAx test. Crit Care. 2013;17(5):R259.

45. Cho A, Chung YE, Choi JS, Kim KS, Choi GH, Park YN, et al. Feasibility of Preoperative FDG PET/CT Total Hepatic Glycolysis in the Remnant Liver for the Prediction of Postoperative Liver Function. AJR Am J Roentgenol. 2017;208(3):624– 31.

230

46. Ünal E, Akata D, Karcaaltincaba M. Liver Function Assessment by Magnetic Resonance Imaging. Semin Ultrasound CT MR. 2016;37(6):549–60.

47. Haimerl M, Schlabeck M, Verloh N, Zeman F, Fellner C, Nickel D, et al. Volume-assisted estimation of liver function based on Gd-EOB-DTPA-enhanced MR relaxometry. Eur Radiol. 2016;26(4):1125–33.

48. Georgiou L, Penny J, Nicholls G, Woodhouse N, Blé F-X, Hubbard Cristinacce PL, et al. Quantitative Assessment of Liver Function Using Gadoxetate-Enhanced Magnetic Resonance Imaging: Monitoring Transporter-Mediated Processes in Healthy Volunteers. Invest Radiol. 2017;52(2):111–9.

49. Levesque E, Martin E, Dudau D, Lim C, Dhonneur G, Azoulay D. Current use and perspective of indocyanine green clearance in liver diseases. Anaesth Crit Care Pain Med. 2016;35(1):49–57.

50. Kawasaki S, Sugiyama Y, Iga T, Hanano M, Sanjo K, Beppu T, et al. Pharmacokinetic study on the hepatic uptake of indocyanine green in cirrhotic patients. Am J Gastroenterol. 1985;80(10):801–6.

51. Alander JT, Kaartinen I, Laakso A, Pätilä T, Spillmann T, Tuchin VV, et al. A review of indocyanine green fluorescent imaging in surgery. Int J Biomed Imaging. 2012;2012:940585.

52. Yokoyama Y, Ebata T, Igami T, Sugawara G, Mizuno T, Yamaguchi J, et al. The Predictive Value of Indocyanine Green Clearance in Future Liver Remnant for Posthepatectomy Liver Failure Following Hepatectomy with Extrahepatic Bile Duct Resection. World J Surg. 2016;40(6):1440–7.

53. Kusuhara H, Sugiyama Y. Pharmacokinetic modeling of the hepatobiliary transport mediated by cooperation of uptake and efflux transporters. Drug Metab Rev. 2010;42(3):539–50.

54. KETTERER SG, WIEGAND BD, RAPAPORT E. Hepatic uptake and biliary excretion of indocyanine green and its use in estimation of hepatic blood flow in dogs. Am J Physiol. 1960;199:481–4.

55. De Gasperi A, Mazza E, Prosperi M. Indocyanine green kinetics to assess liver function: Ready for a clinical dynamic assessment in major liver surgery? World J Hepatol. 2016;8(7):355–67.

56. Weiss M, Krejcie TC, Avram MJ. A physiologically based model of hepatic ICG clearance: Interplay between sinusoidal uptake and biliary excretion. European Journal of Pharmaceutical Sciences. Elsevier; 2011;44(3):359–65.

57. Recknagel P, Claus RA, Neugebauer U, Bauer M, Gonnert FA. In vivo imaging of hepatic excretory function in the rat by fluorescence microscopy. J Biophotonics. Wiley Online Library; 2012;5(7):571–81.

58. Ishizawa T, Masuda K, Urano Y, Kawaguchi Y, Satou S, Kaneko J, et al. Mechanistic background and clinical applications of indocyanine green fluorescence imaging of hepatocellular carcinoma. Ann Surg Oncol. 2014;21(2):440–8.

59. Hirayama Y, Iinuma Y, Yokoyama N, Otani T, Masui D, Komatsuzaki N, et al. Near-infrared fluorescence cholangiography with indocyanine green for biliary atresia. Real-time imaging during the Kasai procedure: a pilot study. Pediatr Surg Int. 2015;31(12):1177–82.

60. Muir AJ. Understanding the Complexities of Cirrhosis. Clin Ther. 2015;37(8):1822–36.

231

61. Vekemans K, Braet F. Structural and functional aspects of the liver and liver sinusoidal cells in relation to colon carcinoma metastasis. World J Gastroenterol. 2005;11(33):5095–102.

62. Wang X, Qi X, Guo X. Tusanqi-Related Sinusoidal Obstruction Syndrome in China: A Systematic Review of the Literatures. Medicine (Baltimore). 2015;94(23):e942.

63. Robinson SM, Mann J, Vasilaki A, Mathers J, Burt AD, Oakley F, et al. Pathogenesis of FOLFOX induced sinusoidal obstruction syndrome in a murine chemotherapy model. J Hepatol. 2013;59(2):318–26.

64. Narita M, Oussoultzoglou E, Bachellier P, Jaeck D, Uemoto S. Post-hepatectomy liver failure in patients with colorectal liver metastases. Surg Today. 2015;45(10):1218–26.

65. Shindoh J, Tzeng C-WD, Aloia TA, Curley SA, Zimmitti G, Wei SH, et al. Optimal future liver remnant in patients treated with extensive preoperative chemotherapy for colorectal liver metastases. Ann Surg Oncol. 2013;20(8):2493–500.

66. Crawford AR, Lin XZ, Crawford JM. The normal adult human liver biopsy: a quantitative reference standard. Hepatology. 1998;28(2):323–31.

67. Standish RA, Cholongitas E, Dhillon A, Burroughs AK, Dhillon AP. An appraisal of the histopathological assessment of liver fibrosis. Gut. 2006;55(4):569–78.

68. Machado NO. Complications of liver biopsy-Risk factors, management and recommendations. In: Liver Biopsy. InTech; 2011.

69. Cogger VC, McNerney GP, Nyunt T, DeLeve LD, McCourt P, Smedsrød B, et al. Three-dimensional structured illumination microscopy of liver sinusoidal endothelial cell fenestrations. J Struct Biol. 2010;171(3):382–8.

70. Mennone A, Nathanson MH. Needle-based confocal laser endomicroscopy to assess liver histology in vivo. Gastrointest Endosc. 2011;73(2):338–44.

71. Kalambokis G, Manousou P, Vibhakorn S, Marelli L, Cholongitas E, Senzolo M, et al. Transjugular liver biopsy-indications, adequacy, quality of specimens, and complications-a systematic review. J Hepatol. 2007;47(2):284–94.

72. Lombardi R, Buzzetti E, Roccarina D, Tsochatzis EA. Non-invasive assessment of liver fibrosis in patients with alcoholic liver disease. World J Gastroenterol. 2015;21(39):11044–52.

73. Pinzani M, Rombouts K, Colagrande S. Fibrosis in chronic liver diseases: diagnosis and management. J Hepatol. 2005;42 Suppl(1):S22–36.

74. Poynard T, Aubert A, Bedossa P, Abella A, Naveau S, Paraf F, et al. A simple biological index for detection of alcoholic liver disease in drinkers. Gastroenterology. 1991;100(5 Pt 1):1397–402.

75. Naveau S, Poynard T, Benattar C, Bedossa P, Chaput J-C. Alpha-2-macroglobulin and hepatic fibrosis. Digestive diseases and sciences. Springer; 1994;39(11):2426–32.

76. Imbert-Bismut F, Ratziu V, Pieroni L, Charlotte F, Benhamou Y, Poynard T. Biochemical markers of liver fibrosis in patients with hepatitis C virus infection: a prospective study. Lancet. 2001;357(9262):1069–75.

232

77. Forns X, Ampurdanès S, Llovet JM, Aponte J, Quintó L, Martínez-Bauer E, et al. Identification of chronic hepatitis C patients without hepatic fibrosis by a simple predictive model. Hepatology. 2002;36(4 Pt 1):986–92.

78. Wai C-T, Greenson JK, Fontana RJ, Kalbfleisch JD, Marrero JA, Conjeevaram HS, et al. A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C. Hepatology. 2003;38(2):518–26.

79. Zeng X, Xu C, He D, Li M, Zhang H, Wu Q, et al. Performance of several simple, noninvasive models for assessing significant liver fibrosis in patients with chronic hepatitis B. Croat Med J. 2015;56(3):272–9.

80. Wilder J, Patel K. The clinical utility of FibroScan\textregistered as a noninvasive diagnostic test for liver disease. Medical Devices (Auckland, NZ). Dove Press; 2014;7:107.

81. Castéra L, Vergniol J, Foucher J, Le Bail B, Chanteloup E, Haaser M, et al. Prospective comparison of transient elastography, Fibrotest, APRI, and liver biopsy for the assessment of fibrosis in chronic hepatitis C. Gastroenterology. 2005;128(2):343–50.

82. Rajakannu M, Cherqui D, Ciacio O, Golse N, Pittau G, Allard MA, et al. Liver stiffness measurement by transient elastography predicts late posthepatectomy outcomes in patients undergoing resection for hepatocellular carcinoma. Surgery. 2017;

83. Friebel A, Neitsch J, Johann T, Hammad S, Hengstler JG, Drasdo D, et al. TiQuant: Software for tissue analysis, quantification and surface reconstruction. arXiv preprint arXiv:14104598. 2015;31(19):1–5.

84. Li H, Chen T-W, Chen X-L, Zhang X-M, Li Z-L, Zeng N-L, et al. Magnetic resonance-based total liver volume and magnetic resonance-diffusion weighted imaging for staging liver fibrosis in mini-pigs. World J Gastroenterol. 2012;18(48):7225–33.

85. Muleki-Seya P, Guillermin R, Guglielmi J, Chen J, Pourcher T, Konofagou E, et al. High-Frequency Quantitative Ultrasound Spectroscopy of Excised Canine Livers and Mouse Tumors Using the Structure Factor Model. IEEE transactions on ultrasonics, ferroelectrics, and frequency control. IEEE; 2016;63(9):1335–50.

86. Kuroda H, Abe T, Kakisaka K, Fujiwara Y, Yoshida Y, Miyasaka A, et al. Visualizing the hepatic vascular architecture using superb microvascular imaging in patients with hepatitis C virus: A novel technique. World J Gastroenterol. 2016;22(26):6057– 64.

87. Morales-Navarrete H, Segovia-Miranda F, Klukowski P, Meyer K, Nonaka H, Marsico G, et al. A versatile pipeline for the multi-scale digital reconstruction and quantitative analysis of 3D tissue architecture. Elife. 2015;4.

88. Golse N, Bucur PO, Adam R, Castaing D, Sa Cunha A, Vibert E. New paradigms in post-hepatectomy liver failure. J Gastrointest Surg. 2013;17(3):593–605.

89. Molino G, Avagnina P, Belforte G, Bircher J. Assessment of the hepatic circulation in humans: new concepts based on evidence derived from a D-sorbitol clearance method. J Lab Clin Med. 1998;131(5):393–405.

90. Burggraaf J, Schoemaker RC, Lentjes EG, Cohen AF. Sorbitol as a marker for drug-induced decreases of variable duration in liver blood flow in healthy volunteers. Eur J Pharm Sci. 2000;12(2):133–9.

91. Skrzypczyk C, Truant S, Duhamel A, Langlois C, Boleslawski E, Koriche D, et al. Relevance of the ISGLS definition of posthepatectomy liver failure in early prediction of poor outcome after liver resection: study on 680 hepatectomies. Ann Surg.

233

2014;260(5):865–70; discussion 870.

92. Nilsson H, Karlgren S, Blomqvist L, Jonas E. The inhomogeneous distribution of liver function: possible impact on the prediction of post-operative remnant liver function. HPB (Oxford). 2015;17(3):272–7.

93. Greco E, Nanji S, Bromberg IL, Shah S, Wei AC, Moulton C-A, et al. Predictors of peri-opertative morbidity and liver dysfunction after hepatic resection in patients with chronic liver disease. HPB (Oxford). 2011;13(8):559–65.

94. Mao Y, Du S, Ba J, Li F, Yang H, Lu X, et al. Using Dynamic 99mT c-GSA SPECT/CT fusion images for hepatectomy planning and postoperative liver failure prediction. Ann Surg Oncol. 2015;22(4):1301–7.

95. Dong J, Zhang X-F, Zhu Y, Ma F, Liu C, Wang W-L, et al. The value of the combination of fibrosis index based on the four factors and future liver remnant volume ratios as a predictor on posthepatectomy outcomes. J Gastrointest Surg. 2015;19(4):682–91.

96. Farges O, Goutte N, Bendersky N, Falissard B. Incidence and risks of liver resection: an all-inclusive French nationwide study. Ann Surg. 2012;256(5):697–704; discussion 704–5.

97. Dimick JB, Wainess RM, Cowan JA, Upchurch GR, Knol JA, Colletti LM. National trends in the use and outcomes of hepatic resection. J Am Coll Surg. 2004;199(1):31–8.

98. Anderson BB, Ukah F, Tette A, Villaflor S-G, Koh D, Seton P. Primary tumors of the liver. Journal of the National Medical Association. National Medical Association; 1992;84(2):129.

99. Hamperl H. The classification of liver tumors. In: Tumors of the Liver. Springer; 1970. p. 4–14.

100. Pagadala MR, McCullough AJ. The relevance of liver histology to predicting clinically meaningful outcomes in nonalcoholic steatohepatitis. Clin Liver Dis. 2012;16(3):487–504.

101. Llovet JM, Brú C, Bruix J. Prognosis of hepatocellular carcinoma: the BCLC staging classification. In: Semin Liver Dis. 1999. p. 329–38.

102. Cieslak KP, Runge JH, Heger M, Stoker J, Bennink RJ, van Gulik TM. New perspectives in the assessment of future remnant liver. Dig Surg. 2014;31(4-5):255–68.

103. Guglielmi A, Ruzzenente A, Conci S, Valdegamberi A, Iacono C. How much remnant is enough in liver resection? Dig Surg. 2012;29(1):6–17.

104. Abdalla EK. Portal vein embolization (prior to major hepatectomy) effects on regeneration, resectability, and outcome. J Surg Oncol. 2010;102(8):960–7.

105. Mullen JT, Ribero D, Reddy SK, Donadon M, Zorzi D, Gautam S, et al. Hepatic insufficiency and mortality in 1,059 noncirrhotic patients undergoing major hepatectomy. J Am Coll Surg. 2007;204(5):854–62; discussion 862–4.

106. Viganò L, Capussotti L, Lapointe R, Barroso E, Hubert C, Giuliante F, et al. Early recurrence after liver resection for colorectal metastases: risk factors, prognosis, and treatment. A LiverMetSurvey-based study of 6,025 patients. Ann Surg Oncol.

234

2014;21(4):1276–86.

107. EASL-EORTC clinical practice guidelines: management of hepatocellular carcinoma. J Hepatol. 2012;56(4):908–43.

108. Azoulay D, Castaing D, Smail A, Adam R, Cailliez V, Laurent A, et al. Resection of nonresectable liver metastases from colorectal cancer after percutaneous portal vein embolization. Ann Surg. 2000;231(4):480–6.

109. Kim HJ, Kim CY, Park EK, Hur YH, Koh YS, Kim HJ, et al. Volumetric analysis and indocyanine green retention rate at 15 min as predictors of post-hepatectomy liver failure. HPB (Oxford). 2015;17(2):159–67.

110. Portal hypertension and the outcome of surgery for hepatocellular carcinoma in compensated cirrhosis: A systematic review and meta-analysis. Hepatology. 2016;63(1):349.

111. Pereira-Lima JE, Lichtenfels E, Barbosa F, Zettler CG, Kulczynski J. Prevalence study of metastases in cirrhotic livers. Hepatogastroenterology. 2003;50(53):1490–5.

112. Scheele J, Stang R, Altendorf-Hofmann A, Paul M. Resection of colorectal liver metastases. World journal of surgery. Springer; 1995;19(1):59–71.

113. Giacchetti S, Itzhaki M, Gruia G, Adam R, Zidani R, Kunstlinger F, et al. Long-term survival of patients with unresectable colorectal cancer liver metastases following infusional chemotherapy with 5-fluorouracil, leucovorin, oxaliplatin and surgery. Ann Oncol. 1999;10(6):663–9.

114. Costi R, Leonardi F, Zanoni D, Violi V, Roncoroni L. Palliative care and end-stage colorectal cancer management: the surgeon meets the oncologist. World J Gastroenterol. Baishideng Publishing Group Inc; 2014;20(24):7602.

115. Sharma A, Houshyar R, Bhosale P, Choi J-I, Gulati R, Lall C. Chemotherapy induced liver abnormalities: an imaging perspective. Clin Mol Hepatol. 2014;20(3):317–26.

116. Chun YS, Vauthey J-N, Ribero D, Donadon M, Mullen JT, Eng C, et al. Systemic chemotherapy and two-stage hepatectomy for extensive bilateral colorectal liver metastases: perioperative safety and survival. J Gastrointest Surg. 2007;11(11):1498–504; discussion 1504–5.

117. Hosokawa I, Allard M-A, Gelli M, Ciacio O, Vibert E, Cherqui D, et al. Long-Term Survival Benefit and Potential for Cure after R1 Resection for Colorectal Liver Metastases. Ann Surg Oncol. 2016;

118. Chang AE, Schneider PD, Sugarbaker PH, Simpson C, Culnane M, Steinberg SM. A prospective randomized trial of regional versus systemic continuous 5-fluorodeoxyuridine chemotherapy in the treatment of colorectal liver metastases. Ann Surg. 1987;206(6):685–93.

119. Rabinowich L, Shibolet O. Drug Induced Steatohepatitis: An Uncommon Culprit of a Common Disease. Biomed Res Int. 2015;2015:168905.

120. Deleve LD. Sinusoidal obstruction syndrome. Gastroenterol Hepatol (N Y). 2008;4(2):101–3.

121. Shimizu N, Ban N, Watanabe Y, Rikitake A, Watanabe R, Tanaka S, et al. The Elevation of Cardio-Ankle Vascular Index in a Patient With Malignant Lymphoma Treated With a Combination Therapy of Rituximab and Cyclophosphamide, Doxorubicin,

235

Vincristine, and Prednisolone. J Clin Med Res. 2017;9(8):729–32.

122. Nordlinger B, Sorbye H, Glimelius B, Poston GJ, Schlag PM, Rougier P, et al. Perioperative FOLFOX4 chemotherapy and surgery versus surgery alone for resectable liver metastases from colorectal cancer (EORTC 40983): long-term results of a randomised, controlled, phase 3 trial. Lancet Oncol. 2013;14(12):1208–15.

123. Bergeat D, Rayar M, Mouchel Y, Merdrignac A, Meunier B, Lièvre A, et al. Preoperative bevacizumab and surgery for colorectal liver metastases: a propensity score analysis. Langenbecks Arch Surg. 2017;402(1):57–67.

124. Adam R, Pascal G, Castaing D, Azoulay D, Delvart V, Paule B, et al. Tumor progression while on chemotherapy: a contraindication to liver resection for multiple colorectal metastases? Ann Surg. 2004;240(6):1052–61; discussion 1061–4.

125. Zendel A, Lahat E, Dreznik Y, Zakai BB, Eshkenazy R, Ariche A. “Vanishing liver metastases”-A real challenge for liver surgeons. Hepatobiliary Surg Nutr. 2014;3(5):295–302.

126. Sturesson C, Nilsson J, Lindell G, Andersson RG, Keussen I. Disappearing liver metastases from colorectal cancer: impact of modern imaging modalities. HPB (Oxford). 2015;17(11):983–7.

127. Spolverato G, Vitale A, Ejaz A, Cosgrove D, Cowzer D, Cillo U, et al. Hepatic Resection for Disappearing Liver Metastasis: a Cost-Utility Analysis. J Gastrointest Surg. 2015;19(9):1668–75.

128. Bischof D, Clary B, Maithel S, Pawlik TM. Surgical management of disappearing colorectal liver metastases. British Journal of Surgery. Wiley Online Library; 2013;100(11):1414–20.

129. Rocha FG, Helton WS. Resectability of colorectal liver metastases: an evolving definition. HPB (Oxford). 2012;14(5):283–4.

130. Rous P, Larimore LD. Relation of the portal blood to liver maintenance. Journal of Experimental Medicine. Rockefeller University Press; 1920;31(5):609–32.

131. Bax H, Mansens B, Schalm L. Atrophy of the liver after occlusion of the bile ducts or portal vein and compensatory hypertrophy of the unoccluded portion and its clinical importance. Gastroenterology. 1956;31(2):131–55.

132. Takayasu K, Muramatsu Y, Shima Y, Moriyama N, Yamada T, Makuuchi M. Hepatic lobar atrophy following obstruction of the ipsilateral portal vein from hilar cholangiocarcinoma. Radiology. 1986;160(2):389–93.

133. Makuuchi M, Thai BL, Takayasu K, Takayama T, Kosuge T, Gunvén P, et al. Preoperative portal embolization to increase safety of major hepatectomy for hilar bile duct carcinoma: a preliminary report. Surgery. 1990;107(5):521–7.

134. Sugawara Y, Yamamoto J, Higashi H, Yamasaki S, Shimada K, Kosuge T, et al. Preoperative portal embolization in patients with hepatocellular carcinoma. World journal of surgery. Springer; 2002;26(1):105–10.

135. De Baere T, Roche A, Elias D, Lasser P, Lagrange C, Bousson V. Preoperative portal vein embolization for extension of hepatectomy indications. Hepatology. 1996;24(6):1386–91.

136. Ironside N, Bell R, Bartlett A, McCall J, Powell J, Pandanaboyana S. Systematic review of perioperative and survival outcomes of liver resections with and without preoperative portal vein embolization for colorectal metastases. HPB (Oxford). 2017;

236

137. Pandanaboyana S, Bell R, Hidalgo E, Toogood G, Prasad KR, Bartlett A, et al. A systematic review and meta-analysis of portal vein ligation versus portal vein embolization for elective liver resection. Surgery. 2015;157(4):690–8.

138. Hwang S, Ha T-Y, Ko G-Y, Kwon D-I, Song G-W, Jung D-H, et al. Preoperative Sequential Portal and Hepatic Vein Embolization in Patients with Hepatobiliary Malignancy. World J Surg. 2015;39(12):2990–8.

139. Huisman F, van Lienden KP, Damude S, Hoekstra LT, van Gulik TM. A review of animal models for portal vein embolization. J Surg Res. 2014;191(1):179–88.

140. Pamecha V, Glantzounis G, Davies N, Fusai G, Sharma D, Davidson B. Long-term survival and disease recurrence following portal vein embolisation prior to major hepatectomy for colorectal metastases. Ann Surg Oncol. 2009;16(5):1202–7.

141. Van Gulik TM, van den Esschert JW, de Graaf W, van Lienden KP, Busch ORC, Heger M, et al. Controversies in the use of portal vein embolization. Dig Surg. 2008;25(6):436–44.

142. Hoekstra LT, van Lienden KP, Doets A, Busch ORC, Gouma DJ, van Gulik TM. Tumor progression after preoperative portal vein embolization. Ann Surg. 2012;256(5):812–7; discussion 817–8.

143. Finlay IG, Meek D, Brunton F, McArdle CS. Growth rate of hepatic metastases in colorectal carcinoma. Br J Surg. 1988;75(7):641–4.

144. Hori SS, Gambhir SS. Mathematical model identifies blood biomarker-based early cancer detection strategies and limitations. Sci Transl Med. 2011;3(109):109ra116.

145. Simoneau E, Hassanain M, Shaheen M, Aljiffry M, Molla N, Chaudhury P, et al. Portal vein embolization and its effect on tumour progression for colorectal cancer liver metastases. Br J Surg. 2015;102(10):1240–9.

146. Katoh M, Neumaier M, Nezam R, Izbicki JR, Schumacher U. Correlation of circulating tumor cells with tumor size and metastatic load in a spontaneous lung metastasis model. Anticancer Res. 24(3a):1421–5.

147. Jiao LR, Apostolopoulos C, Jacob J, Szydlo R, Johnson N, Tsim N, et al. Unique localization of circulating tumor cells in patients with hepatic metastases. J Clin Oncol. 2009;27(36):6160–5.

148. Lalmahomed ZS, Mostert B, Onstenk W, Kraan J, Ayez N, Gratama JW, et al. Prognostic value of circulating tumour cells for early recurrence after resection of colorectal liver metastases. Br J Cancer. 2015;112(3):556–61.

149. Brudvik KW, Seeberg LT, Hugenschmidt H, Renolen A, Schirmer CB, Brunborg C, et al. Detection of Circulating Tumor Cells at Surgery and at Follow-Up Assessment to Predict Survival After Two-Stage Liver Resection of Colorectal Liver Metastases. Ann Surg Oncol. 2015;22(12):4029–37.

150. Elias D, Santoro R, Ouellet J-F, Osmak L, de Baere T, Roche A. Simultaneous percutaneous right portal vein embolization and left liver tumor radiofrequency ablation prior to a major right hepatic resection for bilateral colorectal metastases. Hepatogastroenterology. 51(60):1788–91.

151. Schnitzbauer AA, Lang SA, Goessmann H, Nadalin S, Baumgart J, Farkas SA, et al. Right portal vein ligation combined with in situ splitting induces rapid left lateral liver lobe hypertrophy enabling 2-staged extended right hepatic resection in small-for-

237

size settings. Ann Surg. 2012;255(3):405–14.

152. Edmondson MJ, Sodergren MH, Pucher PH, Darzi A, Li J, Petrowsky H, et al. Variations and adaptations of associated liver partition and portal vein ligation for staged hepatectomy (ALPPS): Many routes to the summit. Surgery. 2015;

153. Qadan M, D’Angelica MI. Complex Surgical Strategies to Improve Resectability in Borderline-Resectable Disease. Curr Colorectal Cancer Rep. 2015;11(6):369–77.

154. Cai Y-L, Song P-P, Tang W, Cheng N-S. An updated systematic review of the evolution of ALPPS and evaluation of its advantages and disadvantages in accordance with current evidence. Medicine (Baltimore). 2016;95(24):e3941.

155. Matsuo K, Murakami T, Kawaguchi D, Hiroshima Y, Koda K, Yamazaki K, et al. Histologic features after surgery associating liver partition and portal vein ligation for staged hepatectomy versus those after hepatectomy with portal vein embolization. Surgery. 2016;159(5):1289–98.

156. Torzilli G, Leoni P, Gendarini A, Calliada F, Olivari N, Makuuchi M. Ultrasound-guided liver resections for hepatocellular carcinoma. Hepatogastroenterology. 2002;49(43):21–7.

157. Torzilli G, Viganò L, Gatti A, Costa G, Cimino M, Procopio F, et al. Twelve-year experience of “radical but conservative” liver surgery for colorectal metastases: impact on surgical practice and oncologic efficacy. HPB (Oxford). 2017;

158. Memeo R, de Blasi V, Adam R, Goéré D, Azoulay D, Ayav A, et al. Parenchymal-sparing hepatectomies (PSH) for bilobar colorectal liver metastases are associated with a lower morbidity and similar oncological results: a propensity score matching analysis. HPB (Oxford). 2016;18(9):781–90.

159. Tiniakos DG, Kandilis A, Geller SA. Tityus: a forgotten myth of liver regeneration. J Hepatol. 2010;53(2):357–61.

160. Steer CJ. Liver regeneration. FASEB J. 1995;9(14):1396–400.

161. Liu ML, Mars WM, Zarnegar R, Michalopoulos GK. Collagenase pretreatment and the mitogenic effects of hepatocyte growth factor and transforming growth factor-alpha in adult rat liver. Hepatology. 1994;19(6):1521–7.

162. Taub R. Liver regeneration: from myth to mechanism. Nat Rev Mol Cell Biol. 2004;5(10):836–47.

163. Rychtrmoc D, Libra A, Buncek M, Garnol T, Cervinková Z. Studying liver regeneration by means of molecular biology: how far we are in interpreting the findings? Acta Medica (Hradec Kralove). 2009;52(3):91–9.

164. Michalopoulos GK. Liver regeneration: molecular mechanisms of growth control. The FASEB Journal. FASEB; 1990;4(2):176–87.

165. Jin J, Hong I-H, Lewis K, Iakova P, Breaux M, Jiang Y, et al. Cooperation of C/EBP family proteins and chromatin remodeling proteins is essential for termination of liver regeneration. Hepatology. 2015;61(1):315–25.

166. Wege H, Müller A, Müller L, Petri S, Petersen J, Hillert C. Regeneration in pig livers by compensatory hyperplasia induces high levels of telomerase activity. Comp Hepatol. 2007;6:6.

238

167. Zhang W, Chen X-P, Zhang W-G, Zhang F, Xiang S, Dong H-H, et al. Hepatic non-parenchymal cells and extracellular matrix participate in oval cell-mediated liver regeneration. World J Gastroenterol. 2009;15(5):552–60.

168. Tanimizu N, Nakamura Y, Ichinohe N, Mizuguchi T, Hirata K, Mitaka T. Hepatic biliary epithelial cells acquire epithelial integrity but lose plasticity to differentiate into hepatocytes in vitro during development. J Cell Sci. 2013;126(Pt 22):5239–46.

169. Fausto N, Campbell JS. The role of hepatocytes and oval cells in liver regeneration and repopulation. Mechanisms of development. Elsevier; 2003;120(1):117–30.

170. Malik R, Selden C, Hodgson H. The role of non-parenchymal cells in liver growth. Semin Cell Dev Biol. 2002;13(6):425–31.

171. Boulton RA, Alison MR, Golding M, Selden C, Hodgson HJ. Augmentation of the early phase of liver regeneration after 70% partial hepatectomy in rats following selective Kupffer cell depletion. J Hepatol. 1998;29(2):271–80.

172. Rozga J, Demetriou A. Animal models of liver regeneration. In: Souba, W.Wiley and Wilmore, W.Douglas, editor. Surgical research. CA, USA: Academic Press; 2001. p. 703–8.

173. Niiya T, Murakami M, Aoki T, Murai N, Shimizu Y, Kusano M. Immediate increase of portal pressure, reflecting sinusoidal shear stress, induced liver regeneration after partial hepatectomy. J Hepatobiliary Pancreat Surg. 1999;6(3):275–80.

174. Marubashi S, Sakon M, Nagano H, Gotoh K, Hashimoto K, Kubota M, et al. Effect of portal hemodynamics on liver regeneration studied in a novel portohepatic shunt rat model. Surgery. 2004;136(5):1028–37.

175. Nagano Y, Nagahori K, Kamiyama M, Fujii Y, Kubota T, Endo I, et al. Improved functional reserve of hypertrophied contra lateral liver after portal vein ligation in rats. J Hepatol. 2002;37(1):72–7.

176. Li J, Cai C, Guo H, Guan X, Yang L, Li Y, et al. Portal vein arterialization promotes liver regeneration after extended partial hepatectomy in a rat model. J Biomed Res. 2015;29(1):69–75.

177. Fan Y-D, Praet M, Van Huysse J, Lelie B, De Hemptinne B. Effects of portal vein arterialization on liver regeneration after partial hepatectomy in the rat. Liver Transpl. 2002;8(2):146–52.

178. Shimizu Y, Miyazaki M, Shimizu H, Ito H, Nakagawa K, Ambiru S, et al. Beneficial effects of arterialization of the portal vein on extended hepatectomy. British Journal of Surgery. Wiley Online Library; 2000;87(6):784–9.

179. Pan N, Lv X, Liang R, Wang L, Liu Q. Suppression of graft regeneration, not ischemia/reperfusion injury, is the primary cause of small-for-size syndrome after partial liver transplantation in mice. PLoS One. 2014;9(4):e93636.

180. Mortensen KE, Conley LN, Nygaard I, Sorenesen P, Mortensen E, Bendixen C, et al. Increased sinusoidal flow is not the primary stimulus to liver regeneration. Comp Hepatol. 2010;9(2).

181. Michalopoulos GK. Liver regeneration after partial hepatectomy: critical analysis of mechanistic dilemmas. Am J Pathol. 2010;176(1):2–13.

182. Tashiro S. Mechanism of liver regeneration after liver resection and portal vein embolization (ligation) is different? J Hepatobiliary Pancreat Surg. 2009;16(3):292–9.

239

183. Lautt WW. Regulatory processes interacting to maintain hepatic blood flow constancy: Vascular compliance, hepatic arterial buffer response, hepatorenal reflex, liver regeneration, escape from vasoconstriction. Hepatol Res. 2007;37(11):891–903.

184. Kim RD, Kim JS, Watanabe G, Mohuczy D, Behrns KE. Liver regeneration and the atrophy-hypertrophy complex. Semin Intervent Radiol. 2008;25(2):92–103.

185. Hill MJ, Hughes M, Jie T, Cohen M, Lake J, Payne WD, et al. Graft weight/recipient weight ratio: how well does it predict outcome after partial liver transplants? Liver Transpl. 2009;15(9):1056–62.

186. Golriz M, Majlesara A, El Sakka S, Ashrafi M, Arwin J, Fard N, et al. Small for Size and Flow (SFSF) syndrome: An alternative description for posthepatectomy liver failure. Clin Res Hepatol Gastroenterol. 2016;40(3):267–75.

187. Yagi S, Uemoto S. Small-for-size syndrome in living donor liver transplantation. HBPD INT. 2012;11(6):570–6.

188. Glanemann M, Eipel C, Nussler AK, Vollmar B, Neuhaus P. Hyperperfusion syndrome in small-for-size livers. Eur Surg Res. 2005;37(6):335–41.

189. Balzan S, Belghiti J, Farges O, Ogata S, Sauvanet A, Delefosse D, et al. The “50-50 criteria” on postoperative day 5: an accurate predictor of liver failure and death after hepatectomy. Annals of surgery. Lippincott, Williams, and Wilkins; 2005;242(6):824.

190. Paugam-Burtz C, Janny S, Delefosse D, Dahmani S, Dondero F, Mantz J, et al. Prospective validation of the “fifty-fifty” criteria as an early and accurate predictor of death after liver resection in intensive care unit patients. Ann Surg. 2009;249(1):124–8.

191. Rahbari NN, Garden OJ, Padbury R, Maddern G, Koch M, Hugh TJ, et al. Post-hepatectomy haemorrhage: a definition and grading by the International Study Group of Liver Surgery (ISGLS). HPB (Oxford). 2011;13(8):528–35.

192. Lo C-M, Liu C-L, Fan S-T. Portal hyperperfusion injury as the cause of primary nonfunction in a small-for-size liver graft- successful treatment with splenic artery ligation. Liver Transpl. 2003;9(6):626–8.

193. Demetris AJ, Kelly DM, Eghtesad B, Fontes P, Wallis Marsh J, Tom K, et al. Pathophysiologic observations and histopathologic recognition of the portal hyperperfusion or small-for-size syndrome. Am J Surg Pathol. 2006;30(8):986–93.

194. Ozeki Y, Umemoto T, Tateyama K, Katagiri Y, Katayama M, Sugiyama A. Partial portal arterialization for dearterialized liver after hepatectomy. Br J Surg. 1997;84(7):1011.

195. Gruttadauria S, Pagano D, Liotta R, Tropea A, Tuzzolino F, Marrone G, et al. Liver Volume Restoration and Hepatic Microarchitecture in Small-for-Size Syndrome. Ann Transplant. 2015;20:381–9.

196. Zhong Z, Schwabe RF, Kai Y, He L, Yang L, Bunzendahl H, et al. Liver regeneration is suppressed in small-for-size liver grafts after transplantation: involvement of c-Jun N-terminal kinase, cyclin D1, and defective energy supply. Transplantation. 2006;82(2):241–50.

197. Liu H-X, Rocha CS, Dandekar S, Wan Y-JY. Functional analysis of the relationship between intestinal microbiota and the expression of hepatic genes and pathways during the course of liver regeneration. J Hepatol. 2016;64(3):641–50.

240

198. Moschen AR, Kaser S, Tilg H. Non-alcoholic steatohepatitis: a microbiota-driven disease. Trends Endocrinol Metab. 2013;24(11):537–45.

199. Wu X, Sun R, Chen Y, Zheng X, Bai L, Lian Z, et al. Oral ampicillin inhibits liver regeneration by breaking hepatic innate immune tolerance normally maintained by gut commensal bacteria. Hepatology. 2015;62(1):253–64.

200. Allard M-A, Adam R, Bucur P-O, Termos S, Cunha AS, Bismuth H, et al. Posthepatectomy portal vein pressure predicts liver failure and mortality after major liver resection on noncirrhotic liver. Ann Surg. 2013;258(5):822–9; discussion 829–30.

201. Sainz-Barriga M, Scudeller L, Costa MG, de Hemptinne B, Troisi RI. Lack of a correlation between portal vein flow and pressure: toward a shared interpretation of hemodynamic stress governing inflow modulation in liver transplantation. Liver Transplantation. Wiley Online Library; 2011;17(7):836–48.

202. Cucchetti A, Cescon M, Golfieri R, Piscaglia F, Renzulli M, Neri F, et al. Hepatic venous pressure gradient in the preoperative assessment of patients with resectable hepatocellular carcinoma. J Hepatol. 2016;64(1):79–86.

203. Lin HC, Tsai YT, Lee FY, Chang TT, Wang SS, Lay CS, et al. Comparison between portal vein pressure and wedged hepatic vein pressure in hepatitis B-related cirrhosis. J Hepatol. 1989;9(3):326–30.

204. Busani S, Marconi G, Schiavon L, Rinaldi L, Del Buono M, Masetti M, et al. Living donor liver transplantation and management of portal venous pressure. In: Transplantation proceedings. 2006. p. 1074–5.

205. Bucur PO, Bekheit M, Audebert C, Othman A, Hammad S, Sebagh M, et al. Modulating Portal Hemodynamics With Vascular Ring Allows Efficient Regeneration After Partial Hepatectomy in a Porcine Model. Ann Surg. 2017;

206. Nobuoka T, Mizuguchi T, Oshima H, Shibata T, Kimura Y, Mitaka T, et al. Portal blood flow regulates volume recovery of the rat liver after partial hepatectomy: molecular evaluation. Eur Surg Res. 2006;38(6):522–32.

207. Hessheimer AJ, Fondevila C, Taurá P, Muñoz J, Sánchez O, Fuster J, et al. Decompression of the portal bed and twice-baseline portal inflow are necessary for the functional recovery of a “small-for-size” graft. Ann Surg. 2011;253(6):1201–10.

208. Chan SC, Lo CM, Chok KS, Sharr WW, Cheung TT, Tsang SH, et al. Modulation of graft vascular inflow guided by flowmetry and manometry in liver transplantation. Hepatobiliary \& Pancreatic Diseases International. Elsevier; 2011;10(6):649–56.

209. Gruttadauria S, Mandala L, Miraglia R, Caruso S, Minervini MI, Biondo D, et al. Successful treatment of small-for-size syndrome in adult-to-adult living-related liver transplantation: single center series. Clinical transplantation. Wiley Online Library; 2007;21(6):761–6.

210. Yamada T, Tanaka K, Uryuhara K, Ito K, Takada Y, Uemoto S. Selective hemi-portocaval shunt based on portal vein pressure for small-for-size graft in adult living donor liver transplantation. Am J Transplant. 2008;8(4):847–53.

211. Oura T, Taniguchi M, Shimamura T, Suzuki T, Yamashita K, Uno M, et al. Does the permanent portacaval shunt for a small- for-size graft in a living donor liver transplantation do more harm than good? Am J Transplant. 2008;8(1):250–2.

212. Mohkam K, Darnis B, Schmitt Z, Duperret S, Ducerf C, Mabrut J-Y. Successful modulation of portal inflow by somatostatin in a porcine model of small-for-size syndrome. Am J Surg. 2016;212(2):321–6.

241

213. Iguchi K, Hatano E, Yamanaka K, Sato M, Yamamoto G, Kasai Y, et al. Hepatoprotective effect by pretreatment with olprinone in a swine partial hepatectomy model. Liver Transpl. 2014;20(7):838–49.

214. Ji Y, Dahmen U, Madrahimov N, Madrahimova F, Xing W, Dirsch O. G-CSF administration in a small-for-size liver model. J Invest Surg. 22(3):167–77.

215. Cantré D, Schuett H, Hildebrandt A, Dold S, Menger MD, Vollmar B, et al. Nitric oxide reduces organ injury and enhances regeneration of reduced-size livers by increasing hepatic arterial flow. Br J Surg. 2008;95(6):785–92.

216. Lin C, Khetani SR. Advances in Engineered Liver Models for Investigating Drug-Induced Liver Injury. Biomed Res Int. 2016;2016:1829148.

217. Lotz J, Meier C, Leppert A, Galanski M. Cardiovascular flow measurement with phase-contrast MR imaging: basic facts and implementation. Radiographics. 22(3):651–71.

218. Markl M. Velocity encoding and flow imaging. University Hospital Freiburg, Dept of Diagnostic Radiology, Germany. 2005;

219. Shiba H, Zhu X, Arakawa Y, Irefin S, Wang B, Trenti L, et al. Oxygen consumption predicts outcome in porcine partial liver grafts. J Surg Res. 2014;189(2):335–9.

220. Hammad S, Hoehme S, Friebel A, von Recklinghausen I, Othman A, Begher-Tibbe B, et al. Protocols for staining of bile canalicular and sinusoidal networks of human, mouse and pig livers, three-dimensional reconstruction and quantification of tissue microarchitecture by image processing and analysis. Arch Toxicol. 2014;88(5):1161–83.

221. Friebel A, Neitsch J, Johann T, Hammad S, Hengstler JG, Drasdo D, et al. TiQuant: software for tissue analysis, quantification and surface reconstruction. Bioinformatics. 2015;31(19):3234–6.

222. Norman JC, Franco FO, Brown ME, Saravis CA, Ackroyd FW, McDermott Jr WV. Techniques of obtaining and preparing the porcine liver for experimental and clinical temporary ex vivo perfusion. Journal of Surgical Research. Elsevier; 1966;6(3):117– 20.

223. Court FG, Wemyss-Holden SA, Morrison CP, Teague BD, Laws PE, Kew J, et al. Segmental nature of the porcine liver and its potential as a model for experimental partial hepatectomy. Br J Surg. 2003;90(4):440–4.

224. Iida T, Yagi S, Taniguchi K, Hori T, Uemoto S. Improvement of morphological changes after 70% hepatectomy with portocaval shunt: preclinical study in porcine model. J Surg Res. 2007;143(2):238–46.

225. Debbaut C, Vierendeels J, Casteleyn C, Cornillie P, Van Loo D, Simoens P, et al. Perfusion characteristics of the human hepatic microcirculation based on three-dimensional reconstructions and computational fluid dynamic analysis. J Biomech Eng. 2012;134(1):011003.

226. Gravante G, Ong SL, Metcalfe MS, Lloyd DM, Dennison AR. The porcine hepatic arterial supply, its variations and their influence on the extracorporeal perfusion of the liver. J Surg Res. 2011;168(1):56–61.

227. Niehues SM, Unger JK, Malinowski M, Neymeyer J, Hamm B, Stockmann M. Liver volume measurement: reason of the difference between in vivo CT-volumetry and intraoperative ex vivo determination and how to cope it. Eur J Med Res.

242

2010;15(8):345–50.

228. Karlo C, Reiner CS, Stolzmann P, Breitenstein S, Marincek B, Weishaupt D, et al. CT- and MRI-based volumetry of resected liver specimen: comparison to intraoperative volume and weight measurements and calculation of conversion factors. Eur J Radiol. 2010;75(1):e107–11.

229. Abdalla EK, Denys A, Chevalier P, Nemr RA, Vauthey J-N. Total and segmental liver volume variations: implications for liver surgery. Surgery. 2004;135(4):404–10.

230. Court FG, Laws PE, Morrison CP, Teague BD, Metcalfe MS, Wemyss-Holden SA, et al. Subtotal hepatectomy: a porcine model for the study of liver regeneration. J Surg Res. 2004;116(1):181–6.

231. Heloury Y, Leborgne J, Rogez J, Robert R, Barbin J, Hureau J. The caudate lobe of the liver. Surgical and Radiologic Anatomy. Springer; 1988;10(1):83–91.

232. Cho A, Asano T, Yamamoto H, Nagata M, Takiguchi N, Kainuma O, et al. Relationship between right portal and biliary systems based on reclassification of the liver. The American journal of surgery. Elsevier; 2007;193(1):1–4.

233. Zanchet DJ, Montero EF de S. Pig liver sectorization and segmentation and virtual reality depiction. Acta Cirurgica Brasileira. SciELO Brasil; 2002;17(6):381–7.

234. Darnis B, Mohkam K, Schmitt Z, Ledochowski S, Vial J-P, Duperret S, et al. Subtotal hepatectomy in swine for studying small-for-size syndrome and portal inflow modulation: is it reliable? HPB (Oxford). 2015;

235. Kahn D, Hickman R, Terblanche J, Von Sommoggy S. Partial hepatectomy and liver regeneration in pigs—the response to different resection sizes. Journal of Surgical Research. Elsevier; 1988;45(2):176–80.

236. Martins AC de A, Machado MAC, Ferraz ÁAB. Porcine liver: experimental model for the intra-hepatic glissonian approach. Acta Cirurgica Brasileira. SciELO Brasil; 2008;23(2):204–7.

237. Belghiti J, Clavien P, Gadzijev E. The Brisbane 2000 Terminology of Liver Anatomy and Resections. HPB. 2000;2(3):333–9.

238. Arkadopoulos N, Defterevos G, Nastos C, Papalois A, Kalimeris K, Papoutsidakis N, et al. Development of a porcine model of post-hepatectomy liver failure. J Surg Res. 2011;170(2):e233–42.

239. Shirabe K, Shimada M, Gion T, Hasegawa H, Takenaka K, Utsunomiya T, et al. Postoperative liver failure after major hepatic resection for hepatocellular carcinoma in the modern era with special reference to remnant liver volume. J Am Coll Surg. 1999;188(3):304–9.

240. Mahla E, Lang T, Vicenzi MN, Werkgartner G, Maier R, Probst C, et al. Thromboelastography for monitoring prolonged hypercoagulability after major abdominal surgery. Anesth Analg. 2001;92(3):572–7.

241. Darlington DN, Delgado AV, Kheirabadi BS, Fedyk CG, Scherer MR, Pusateri AE, et al. Effect of hemodilution on coagulation and recombinant factor VIIa efficacy in human blood in vitro. J Trauma. 2011;71(5):1152–63.

243

242. Austin AW, Wirtz PH, Patterson SM, Stutz M, von Känel R. Stress-induced alterations in coagulation: assessment of a new hemoconcentration correction technique. Psychosom Med. 2012;74(3):288–95.

243. Shapiro SS, Martinez J. Human prothrombin metabolism in normal man and in hypocoagulable subjects. J Clin Invest. 1969;48(7):1292–8.

244. Wrighton LJ, O’Bosky KR, Namm JP, Senthil M. Postoperative management after hepatic resection. J Gastrointest Oncol. 2012;3(1):41–7.

245. Theodoraki K, Arkadopoulos N, Fragulidis G, Voros D, Karapanos K, Markatou M, et al. Transhepatic lactate gradient in relation to liver ischemia/reperfusion injury during major hepatectomies. Liver Transpl. 2006;12(12):1825–31.

246. Pastor CM. Hepatic and splanchnic oxygen consumption during acute hypoxemic hypoxia in anesthetized pigs. Crit Care Med. 2000;28(3):765–73.

247. Lundvall J, Länne T. Large capacity in man for effective plasma volume control in hypovolaemia via fluid transfer from tissue to blood. Acta Physiol Scand. 1989;137(4):513–20.

248. Hughes SJ, Yang W, Juszczak M, Jones GL, Powis SH, Seifalian AM, et al. Effect of inspired oxygen on portal and hepatic oxygenation: effective arterialization of portal blood by hyperoxia. Cell Transplant. 2004;13(7-8):801–8.

249. Takala J. Determinants of splanchnic blood flow. Br J Anaesth. 1996;77(1):50–8.

250. Jacobson C. A novel anaesthetic regimen for surgical procedures in guineapigs. Lab Anim. 2001;35(3):271–6.

251. Kallner A. Preanalytical procedures in the measurement of ionized calcium in serum and plasma. Eur J Clin Chem Clin Biochem. 1996;34(1):53–8.

252. Seamonds B, Towfighi J, Arvan DA. Determination of ionized calcium in serum by use of an lon-selective electrode. I. Determination of normal values under physiologic conditions, with comments on the effects of food ingestion and hyperventilation. Clin Chem. 1972;18(2):155–60.

253. Bilkovski R, Cannon C, Adhikari S, Nasr I. Arterial and venous ionized calcium measurements: Is there a difference? Annals of Emergency Medicine. Elsevier; 2004;44(4):S56.

254. Goldstein D. Serum Calcium. In: Walker, H Kenneth and Hall, W Dallas and Hurst, J Willis, editor. Clinical Methods: The History, Physical, and Laboratory Examinations. Butterworths; 1990. p. 677–9.

255. Siller-Matula JM, Plasenzotti R, Spiel A, Quehenberger P, Jilma B. Interspecies differences in coagulation profile. Thromb Haemost. 2008;100(3):397–404.

256. Miura N, Kawaguchi H, Nagasato T, Yamada T, Ito T, Izumi H, et al. Coagulation activity and white thrombus formation in the microminipig. In Vivo (Brooklyn). 27(3):357–61.

257. BRECHER G, CRONKITE EP. Morphology and enumeration of human blood platelets. J Appl Physiol. 1950;3(6):365–77.

244

258. Burman D. Haemoglobin levels in normal infants aged 3 to 24 months, and the effect of iron. Arch Dis Child. 1972;47(252):261–71.

259. Janus J, Moerschel SK. Evaluation of anemia in children. Am Fam Physician. 2010;81(12):1462–71.

260. Tembe N, Joaquim O, Alfai E, Sitoe N, Viegas E, Macovela E, et al. Reference values for clinical laboratory parameters in young adults in Maputo, Mozambique. PLoS One. 2014;9(5):e97391.

261. Hosten AO. BUN and Creatinine. In: Walker, H Kenneth and Hall, W Dallas and Hurst, J Willis, editor. Clinical Methods: The History, Physical, and Laboratory Examinations. Butterworths, Boston; 1990. p. 874–8.

262. Ackerman GL. Serum Sodium. In: Walker HK, Hall WD, Hurst JW, editor. Clinical Methods: The History, Physical, and Laboratory Examinations. Boston: Butterworths; 1990. p. 879–83.

263. Rastegar A. Serum potassium. In: Walker HK, Hall WD, Hurst JW, editor. Clinical Methods: The History, Physical, and Laboratory Examinations. Boston: Butterworths; 1990. p. 884–7.

264. Martin GB, Carden DL, Nowak RM, Tomlanovich MC. Comparison of central venous and arterial pH and PCO2 during open- chest CPR in the canine model. Ann Emerg Med. 1985;14(6):529–33.

265. TRULOCK III E. Arterial Blood Gases. In: Clinical Methods: The History, Physical, and Laboratory Examinations. Butterworth-Heinemann; 1990. p. 254.

266. Andersen JB, Wang T. Effects of anaesthesia on blood gases, acid-base status and ions in the toad Bufo marinus. Comp Biochem Physiol, Part A Mol Integr Physiol. 2002;131(3):639–46.

267. Van der Linden P, Rausin I, Deltell A, Bekrar Y, Gilbart E, Bakker J, et al. Detection of tissue hypoxia by arteriovenous gradient for PCO2 and pH in anesthetized dogs during progressive hemorrhage. Anesth Analg. 1995;80(2):269–75.

268. Hannon JP, Wade CE, Bossone CA, Hunt MM, Loveday JA. Oxygen delivery and demand in conscious pigs subjected to fixed- volume hemorrhage and resuscitated with 7.5% NaCl in 6% Dextran. Circ Shock. 1989;29(3):205–17.

269. Adeva-Andany M, López-Ojén M, Funcasta-Calderón R, Ameneiros-Rodríguez E, Donapetry-García C, Vila-Altesor M, et al. Comprehensive review on lactate metabolism in human health. Mitochondrion. 2014;17:76–100.

270. Murphy ND, Kodakat SK, Wendon JA, Jooste CA, Muiesan P, Rela M, et al. Liver and intestinal lactate metabolism in patients with acute hepatic failure undergoing liver transplantation. Crit Care Med. 2001;29(11):2111–8.

271. Bongaerts GP, Tolboom JJ, Naber AH, Sperl WJ, Severijnen RS, Bakkeren JA, et al. Role of bacteria in the pathogenesis of short bowel syndrome-associated D-lactic acidemia. Microb Pathog. 1997;22(5):285–93.

272. Chambers AC, Bismohun SL, Davies H, White P, Patil AV. Predictive value of abnormally raised serum bilirubin in acute appendicitis: a cohort study. Int J Surg. 2015;13:207–10.

273. Shiao S-YPK. Accurate measurements of oxygen saturation in neonates: paired arterial and venous blood analyses. Newborn and Infant Nursing Reviews. Elsevier; 2005;5(4):170–8.

245

274. Ren Y-S, Qian N-S, Tang Y, Liao Y-H, Liu W-H, Raut V, et al. Beneficial effects of splenectomy on liver regeneration in a rat model of massive hepatectomy. HBPD INT. 2012;11(1):60–5.

275. Ricken T, Werner D, Holzhütter HG, König M, Dahmen U, Dirsch O. Modeling function-perfusion behavior in liver lobules including tissue, blood, glucose, lactate and glycogen by use of a coupled two-scale PDE-ODE approach. Biomech Model Mechanobiol. 2015;14(3):515–36.

276. Debbaut C, Vierendeels J, Siggers JH, Repetto R, Monbaliu D, Segers P. A 3D porous media liver lobule model: the importance of vascular septa and anisotropic permeability for homogeneous perfusion. Comput Methods Biomech Biomed Engin. 2014;17(12):1295–310.

277. Peeters G, Debbaut C, Cornillie P, De Schryver T, Monbaliu D, Laleman W, et al. A multilevel modeling framework to study hepatic perfusion characteristics in case of liver cirrhosis. J Biomech Eng. 2015;137(5):051007.

278. Van Der Plaats A, t Hart N, Verkerke G, Leuvenink H, Verdonck P, Ploeg R, et al. Numerical simulation of the hepatic circulation. Int J Artif Organs. 2004;27(3):222–30.

279. Chu T-M, Reddy NP. A lumped parameter mathematical model of the splanchnic circulation. Journal of biomechanical engineering. American Society of Mechanical Engineers; 1992;114(2):222–6.

280. Schwen LO, Schenk A, Kreutz C, Timmer J, Bartolomé Rodríguez MM, Kuepfer L, et al. Representative Sinusoids for Hepatic Four-Scale Pharmacokinetics Simulations. PLoS One. 2015;10(7):e0133653.

281. Bézy-Wendling J, Kretowski M, Mescam M, Jurczuk K, Eliat P-A. Simulation of hepatocellular carcinoma in MRI by combined macrovascular and pharmacokinetic models. In: Biomedical Imaging: From Nano to Macro, 2007 ISBI 2007 4th IEEE International Symposium on. 2007. p. 1272–5.

282. Childress EM, Kleinstreuer C. Impact of fluid-structure interaction on direct tumor-targeting in a representative hepatic artery system. Ann Biomed Eng. 2014;42(3):461–74.

283. Ho H, Sorrell K, Bartlett A, Hunter P. Modeling the hepatic arterial buffer response in the liver. Medical engineering \& physics. Elsevier; 2013;35(8):1053–8.

284. Debbaut C, De Wilde D, Casteleyn C, Cornillie P, Van Loo D, Van Hoorebeke L, et al. Modeling the impact of partial hepatectomy on the hepatic hemodynamics using a rat model. IEEE Trans Biomed Eng. 2012;59(12):3293–303.

285. Audebert C, Bucur P, Bekheit M, Vibert E, Vignon-Clementel IE, Gerbeau J-F. Kinetic scheme for arterial and venous blood flow, and application to partial hepatectomy modeling. Comput Methods Appl Mech Eng. Elsevier; 2016;

286. Bucur P, Bekheit M, Audebert C, Vignon-Clementel I, Vibert E. Simplified technique for 75% and 90% hepatic resection with hemodynamic monitoring in a large white swine model. J Surg Res. 2016;209:122–30.

287. Bekheit M, Bucur PO, Wartenberg M, Vibert E. Computerized tomography-based anatomic description of the porcine liver. J Surg Res. Elsevier; 2016;

246

288. Lantz BM, Foerster JM, Link DP, Holcroft JW. Regional distribution of cardiac output: normal values in man determined by video dilution technique. AJR Am J Roentgenol. 1981;137(5):903–7.

289. Liang F, Liu H. A closed-loop lumped parameter computational model for human cardiovascular system. JSME International Journal Series C Mechanical Systems, Machine Elements and Manufacturing. The Japan Society of Mechanical Engineers; 2005;48(4):484–93.

290. Segers P, Stergiopulos N, Westerhof N, Wouters P, Kolh P, Verdonck P. Systemic and pulmonary hemodynamics assessed with a lumped-parameter heart-arterial interaction model. Journal of Engineering Mathematics. Springer; 2003;47(3):185–99.

291. Audebert C, Bekheit M, Bucur P, Vibert E, Vignon-Clementel IE. Partial hepatectomy hemodynamics changes: Experimental data explained by closed-loop lumped modeling. J Biomech. 2017;50:202–8.

292. Ezzat WR, Lautt WW. Hepatic arterial pressure-flow autoregulation is adenosine mediated. Am J Physiol. 1987;252(4 Pt 2):H836–45.

293. Bernadich C, Fernández M, Bandi JC, Bosch J, Rodés J. Mechanical pumping of portal blood to the liver: hemodynamic effects of a new experimental treatment for portal hypertension. J Hepatol. 1996;25(1):98–105.

294. Lauber DT, Tihanyi DK, Czigány Z, Kovács T, Budai A, Drozgyik D, et al. Liver regeneration after different degrees of portal vein ligation. J Surg Res. Elsevier; 2016;203(2):451–8.

295. Ding B-S, Nolan DJ, Butler JM, James D, Babazadeh AO, Rosenwaks Z, et al. Inductive angiocrine signals from sinusoidal endothelium are required for liver regeneration. Nature. 2010;468(7321):310–5.

296. Stolz DB, Ross MA, Ikeda A, Tomiyama K, Kaizu T, Geller DA, et al. Sinusoidal endothelial cell repopulation following ischemia/reperfusion injury in rat liver transplantation. Hepatology. 2007;46(5):1464–75.

297. Sato T, El-Assal ON, Ono T, Yamanoi A, Dhar DK, Nagasue N. Sinusoidal endothelial cell proliferation and expression of angiopoietin/Tie family in regenerating rat liver. J Hepatol. 2001;34(5):690–8.

298. Asencio JM, García-Sabrido JL, López-Baena JA, Olmedilla L, Peligros I, Lozano P, et al. Preconditioning by portal vein embolization modulates hepatic hemodynamics and improves liver function in pigs with extended hepatectomy. Surgery. 2017;

299. Brasaemle DL. Cell biology. A metabolic push to proliferate. Science (80- ). 2006;313(5793):1581–2.

300. Thevananther S. Adipose to the rescue: peripheral fat fuels liver regeneration. Hepatology. 2010;52(6):1875–6.

301. Tani K, Shindoh J, Takamoto T, Shibahara J, Nishioka Y, Hashimoto T, et al. Kinetic Changes in Liver Parenchyma After Preoperative Chemotherapy for Patients with Colorectal Liver Metastases. J Gastrointest Surg. 2017;

302. Zou Y, Bao Q, Kumar S, Hu M, Wang G-Y, Dai G. Four waves of hepatocyte proliferation linked with three waves of hepatic fat accumulation during partial hepatectomy-induced liver regeneration. PLoS One. 2012;7(2):e30675.

303. Ulicná O, Istvánová B, Valachová A, Brixová E. [Oxidative phosphorylation in liver mitochondria after injury with carbon tetrachloride and during regeneration]. Bratisl Lek Listy. 1994;95(9):402–7.

247

304. Ozdogan M, Ersoy E, Dundar K, Albayrak L, Devay S, Gundogdu H. Beneficial effect of hyperbaric oxygenation on liver regeneration in cirrhosis. J Surg Res. 2005;129(2):260–4.

305. Donadon M, Molinari AF, Corazzi F, Rocchi L, Zito P, Cimino M, et al. Pharmacological Modulation of Ischemic-Reperfusion Injury during Pringle Maneuver in Hepatic Surgery. A Prospective Randomized Pilot Study. World J Surg. 2016;40(9):2202– 12.

306. Le Bihan D. Diffusion, confusion and functional MRI. Neuroimage. 2012;62(2):1131–6.

307. Le Bihan D. Apparent diffusion coefficient and beyond: what diffusion MR imaging can tell us about tissue structure. Radiology. Radiological Society of North America, Inc.; 2013;268(2):318–22.

308. Chandarana H, Taouli B. Diffusion and perfusion imaging of the liver. Eur J Radiol. 2010;76(3):348–58.

309. Tosun M, Inan N, Sarisoy HT, Akansel G, Gumustas S, Gürbüz Y, et al. Diagnostic performance of conventional diffusion weighted imaging and diffusion tensor imaging for the liver fibrosis and inflammation. European journal of radiology. Elsevier; 2013;82(2):203–7.

310. Li J, Liang L, Ma T, Yu X, Chen W, Xu G, et al. Sinusoidal microcirculatory changes after small-for-size liver transplantation in rats. Transplant International. Wiley Online Library; 2010;23(9):924–33.

311. Le Bihan D, Breton E, Lallemand D, Aubin M, Vignaud J, Laval-Jeantet M. Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology. 1988;168(2):497–505.

312. Yamada I, Aung W, Himeno Y, Nakagawa T, Shibuya H. Diffusion coefficients in abdominal organs and hepatic lesions: evaluation with intravoxel incoherent motion echo-planar MR imaging. Radiology. Radiological Society of North America; 1999;210(3):617–23.

313. Transonic. Keys to Accurate Flow Measurements with Transit-Time Ultrasound [Internet]. Business Park Stein 205 6181 MB Elsloo The Netherlands; Available from: http://www.transonic.com/resources/research/keys-to-accurate-perivascular-flow- measurements-with-transit-time-ultrasound/

314. Chow PK, Yu WK, Ng TH, Ong HS, Ooi PJ, Chan ST, et al. Influence of respiration and portal pressure on transabdominal duplex Doppler ultrasound measurement of portal blood flow: a porcine model for experimental studies. J Surg Res. 2000;89(1):66–73.

315. Edgerton RH. The effect of arterial wall thickness and conductivity on electromagnetic flowmeter readings. Medical and biological engineering. Springer; 1968;6(6):627–36.

316. Vermeulen MAR, Ligthart-Melis GC, Buijsman R, Siroen MPC, van de Poll MCG, Boelens PG, et al. Accurate perioperative flow measurement of the portal vein and hepatic and renal artery: a role for preoperative MRI? Eur J Radiol. 2012;81(9):2042– 8.

317. Carlisle KM, Halliwell M, Read AE, Wells PN. Estimation of total hepatic blood flow by duplex ultrasound. Gut. 1992;33(1):92–7.

248

318. Polzin JA, Alley MT, Korosec FR, Grist TM, Wang Y, Mistretta CA. A complex-difference phase-contrast technique for measurement of volume flow rates. J Magn Reson Imaging. 5(2):129–37.

319. Andersson C, Kihlberg J, Ebbers T, Lindström L, Carlhäll C-J, Engvall JE. Phase-contrast MRI volume flow - a comparison of breath held and navigator based acquisitions. BMC Med Imaging. 2016;16(1):26.

320. Yzet T, Bouzerar R, Allart J-D, Demuynck F, Legallais C, Robert B, et al. Hepatic vascular flow measurements by phase contrast MRI and doppler echography: a comparative and reproducibility study. J Magn Reson Imaging. 2010;31(3):579–88.

321. Payen DM, Fratacci MD, Dupuy P, Gatecel C, Vigouroux C, Ozier Y, et al. Portal and hepatic arterial blood flow measurements of human transplanted liver by implanted Doppler probes: interest for early complications and nutrition. Surgery. 1990;107(4):417–27.

322. Kurol M, Forsberg L. Ultrasonographic investigation of respiratory influence on diameters of portal vessels in normal subjects. Acta Radiol Diagn (Stockh). 27(6):675–80.

323. MISSING:gravante2011porcine. MISSING:gravante2011porcine. 2018;

324. Reichen J. The Role of the Sinusoidal Endothelium in Liver Function. News Physiol Sci. 1999;14:117–21.

325. Asencio JM, Vaquero J, Olmedilla L, García Sabrido JL. “Small-for-flow” syndrome: shifting the “size” paradigm. Med Hypotheses. 2013;80(5):573–7.

326. Eshkenazy R, Dreznik Y, Lahat E, Zakai BB, Zendel A, Ariche A. Small for size liver remnant following resection: prevention and management. Hepatobiliary Surg Nutr. 2014;3(5):303–12.

327. Adams RB, Aloia TA, Loyer E, Pawlik TM, Taouli B, Vauthey J-N. Selection for hepatic resection of colorectal liver metastases: expert consensus statement. HPB (Oxford). 2013;15(2):91–103.

328. Vasavada BB, Chen CL, Zakaria M. Portal flow is the main predictor of early graft dysfunction regardless of the GRWR status in living donor liver transplantation - a retrospective analysis of 134 patients. Int J Surg. 2014;12(2):177–80.

329. Kishi Y, Abdalla EK, Chun YS, Zorzi D, Madoff DC, Wallace MJ, et al. Three hundred and one consecutive extended right hepatectomies: evaluation of outcome based on systematic liver volumetry. Ann Surg. 2009;250(4):540–8.

330. Truant S, Oberlin O, Sergent G, Lebuffe G, Gambiez L, Ernst O, et al. Remnant liver volume to body weight ratio > or =0.5%: A new cut-off to estimate postoperative risks after extended resection in noncirrhotic liver. J Am Coll Surg. 2007;204(1):22–33.

331. Mullin EJ, Metcalfe MS, Maddern GJ. How much liver resection is too much? Am J Surg. 2005;190(1):87–97.

332. Van Lienden KP, van den Esschert JW, de Graaf W, Bipat S, Lameris JS, van Gulik TM, et al. Portal vein embolization before liver resection: a systematic review. Cardiovasc Intervent Radiol. 2013;36(1):25–34.

333. Al-Sharif E, Simoneau E, Hassanain M. Portal vein embolization effect on colorectal cancer liver metastasis progression: Lessons learned. World J Clin Oncol. 2015;6(5):142–6.

249

334. Troisi R, de Hemptinne B. Clinical relevance of adapting portal vein flow in living donor liver transplantation in adult patients. Liver Transpl. 2003;9(9):S36–41.

335. Hou P, Chen C, Tu Y-L, Zhu Z-M, Tan J-W. Extracorporeal continuous portal diversion plus temporal plasmapheresis for “small-for-size” syndrome. World J Gastroenterol. 2013;19(33):5464–72.

336. Kim J, Kim C-J, Ko I-G, Joo SH, Ahn HJ. Splenectomy affects the balance between hepatic growth factor and transforming growth factor-β and its effect on liver regeneration is dependent on the amount of liver resection in rats. J Korean Surg Soc. 2012;82(4):238–45.

337. Reyal J, Uemoto S. Percutaneously adjustable portal vein banding device could prevent post-operative liver failure-artificial control of portal venous flow is the key to a new therapeutic world. Med Hypotheses. 2009;73(5):640–50.

338. Council. NR. Guide for the care and use of laboratory animals. Eighth. Washington: Natl Academy Pr; 2007.

339. Bosch J, Groszmann RJ, Shah VH. Evolution in the understanding of the pathophysiological basis of portal hypertension: How changes in paradigm are leading to successful new treatments. J Hepatol. 2015;62(1 Suppl):S121–30.

340. Kjekshus H, Risoe C, Scholz T, Smiseth OA. Methods for assessing hepatic distending pressure and changes in hepatic capacitance in pigs. Am J Physiol Heart Circ Physiol. 2000;279(4):H1796–803.

341. Mueller L, Broering DC, Meyer J, Vashist Y, Goettsche J, Wilms C, et al. The induction of the immediate-early-genes Egr-1, PAI-1 and PRL-1 during liver regeneration in surgical models is related to increased portal flow. J Hepatol. 2002;37(5):606–12.

342. Vasavada B, Chen CL, Zakaria M. Using low graft/recipient’s body weight ratio graft with portal flow modulation an effective way to prevent small-for-size syndrome in living-donor liver transplant: a retrospective analysis. Exp Clin Transplant. 2014;12(5):437–42.

343. Umeda Y, Yagi T, Sadamori H, Matsukawa H, Matsuda H, Shinoura S, et al. Effects of prophylactic splenic artery modulation on portal overperfusion and liver regeneration in small-for-size graft. Transplantation. 2008;86(5):673–80.

344. Taniguchi M, Shimamura T, Suzuki T, Yamashita K, Oura T, Watanabe M, et al. Transient portacaval shunt for a small-for-size graft in living donor liver transplantation. Liver Transpl. 2007;13(6):932–4.

345. Wang X-Q, Xu Y-F, Tan J-W, Lv W-P, Liu Z, Zeng J-P, et al. Portal inflow preservation during portal diversion in small-for- size syndrome. World J Gastroenterol. 2014;20(4):1021–9.

346. Troisi R, Hesse UJ, Decruyenaere J, Morelli MC, Palazzo U, Pattyn P, et al. Functional, life-threatening disorders and splenectomy following liver transplantation. Clin Transplant. 1999;13(5):380–8.

347. Xiang L, Huang L, Wang X, Zhao Y, Liu Y, Tan J. How Much Portal Vein Flow Is Too Much for Liver Remnant in a Stable Porcine Model? Transplant Proc. 48(1):234–41.

348. Bekheit M, Bucur P, Vibert E, Andres C, others. The reference values for hepatic oxygen consumption and net lactate production, blood gasses, hemogram, major electrolytes, and kidney and liver profiles in anesthetized large white swine model. Transl Surg. Medknow Publications; 2016;1(4):95.

250

349. Pinzani M, Vizzutti F. Anatomy and vascular biology of the cells in the portal circulation. Portal hypertension. Springer; 2005;15–35.

251

VI. Appendix

iv. MRI Elastography protocol

This communication is concerned with demonstrating the reproducibility of the last set of results and targets the answers of three (or

perhaps more) questions:

Are these images interpretable and reliable?

Does the ROI given on the figures (under title ROIs) correspond directly to the kPa or do we need a transformation formula?

The position of the acquisition slice, is it a must to be perpendicular (see at the end)

a. Context

MRE experiment trial on piglet(LargeWhite) liver.

The piglet was 3 months old and weighed 35 kg. b. Method

The pig was laying on his back surrounded with 2 antennas one flex large channel laying on his abdomen and the spine placed directly onto the MRI table.

For the acquisition (approx. 30 s), the pig was put under apnea. Our primitive method to induce apnea is to use an ambu bag with controlled squeezing. c. Results

252

450

400

350

300

250 J-7 : 200 J3 150

100

0 surgery

Figure Appendix: Preoperative Kpa values of elastography measures compared to the measures at day 3.

The trials were done with 40% amplitude. Several trials are included below to demonstrate the reproducibility of the technique of imaging. Old Trial which you commented upon. We kept it since you had a query to which an answer was given

Phase image Magnitude image

Wave Stiffness map with 95%

image

SAGITTAL: Essai 1

253

Phase image Magnitude image

Wave image Stiffness map with 95%

CORONAL: Essai 2

AXIAL: Essai 3

Phase image Magnitude image

Wave Stiffness map with 95%

image

254

This axial slice is perfect for MRE imaging. Have you tried this slice for MRE imaging with through-plane motion sensitizing direction? It should be feasible with the commercial sequence because it is the same as the clinical patient protocol. It is said this slice was located at F51.8 in this figure, while it is F14.2 in the previous figure. They are 37.6mm apart. What is the liver size of a piglet? I am worried that the first image is too far away from the optimum slice prescription.

The average liver size in our animals is around 1000 ml

New experiments (which you did not see yet)

Monday 8th of June Elastography

Cochon 6B J-7

Addition of saturation bands on the three different acquisition planes.

SAGITTAL: Essai 1

255

Is this the approximate position recommended for the saturation bands?, below (the last trial) we added it much closer to the acquisition band. Which position is better close or far from the acquisition?

They are determined by the flow velocity and pattern. I usually define saturation bands 40 mm away from the slice acquisition.

CORONAL: Essai 2

256

AXIAL: Essai 3

257

On this axial acquisition the two saturation bands where placed either side of about 4 cm from the acquisition plane.

Are these two saturation bands intended to minimize flow artefacts?

Is this the correct placement?

Yes, these two saturation bands were used to minimize flow artifacts. I think the placement is appropriate.

It seems you may have a big hot spot artifact in the axial imaging plane at the right anterior corner of the liver. Waves might change its propagating direction when the boundary condition was not straight like a line.

ROIs

258

This hot spot is quite consistent in the axial imaging plane at the right anterior corner of the liver.

I think axial imaging plane may not perfect for this pig at this situation.

Considering the cylindrical symmetric property of the driver, coronal imaging planes across the driver also satisfy the rule of

perpendicular imaging planes. As shown in your upper right coronal imaging plane, the liver stiffness measurement was quite

homogeneous and the liver boundary changed very smoothly at an obtuse angle (not as sharp change at an angle of 90 degree as that

of axial imaging plane in the lower right image)

Tuesday17th of June Elastography

Cochon 3A J-12

259

Use of saturation bands close to the acquisition bands

SAGITTAL: Essai 1

Same as the comment in the previous page.

CORONAL: Essai 2

260

AXIAL: Essai 4

d. Questions

261

1- Are these images interpretable and reliable?

Yes, all images are interpretable. However, not all pixel values are reliable considering limitations of 2D-MRE. Areas at

the corners and edges may have hot spots artifacts.

2- Does the ROI given on the figures (last two) correspond directly to the kPa or do we need a transformation formula?

For Siemens scanner, you need divided by 100 to get mean liver stiffness measurement in kPa. You can find a clue in

the color bar contained within the elastogram.

3- We tried the most we can to applicate the principles you recommended : perpendicular, through plane, including major

meat of the liver, however, sometimes we perform inclination of the acquisition slice to avoid the passage through the

stomach and this usually does not affect the stiffness map but the wave map is affected a little where the waves are

complete in consecutive images rather than the same image (following the inclination of the acquisition slice). Do you

think this is sound or we stick to the perpendicularity ? (images not shown here since it was higher than required)

If liver stiffness measurement is consistent, it would be totally fine to choose coronal or oblique imaging planes to avoid

bowels or stomach. We had experience in some volunteers that “optimum imaging plane” may not work well and

consistently give hot spots. Usually we found that they have special contours (local bulges) of the liver or relationship

with other organs (bowel interposition).

In your cases of piglets MRE experiments, I felt that you have obtained

1) consistent and reliable results when using coronal imaging planes;

2) consistent and reliable results when using sagittal imaging planes with small artificial hot spots at the inferior edges

of the liver

262

3) consistent results when using axial imaging planes with a big artificial hot spot at the right anterior corner of the liver.

I think you can certainly keep collecting coronal and sagittal imaging planes in this piglet to get a reliable liver stiffness measurement. You can also keep collecting axial imaging plane to see whether the artificial hot spots are consistent in all pigs or only this one.

Finally, 3D-MRE should be the most appropriate solution to obtain reliable estimates without considering selection of imaging planes.

263

VII. Abstrait

u Résumé de la thèse :

Des mesures expétimentales sont nécessaires pour la construction et la validation de ces modèles. Des ablations du foie de différentes tailles sont effectuées sur des porcs et pendant ces chirurgies plusieurs pressions et débits sont mesurés. De plus, un colorant fluorescent est injecté avant ou après l’ablation partielle, et la fluorescence de ce composé est mesurée. Dans une première partie, la procédure chirurgicale, les conditions expérimentales ainsi que les mesures obtenues sont détaillées. Ensuite, les changements hémodynamiques, conséquence de l’ablation partielle du foie. Le modèle permet de prendre en compte les changements de volume sanguins qui peuvent se produire (saignements) lors de la chirurgie. Par conséquent, ce modèle propose une explication de la variabilité des mesures acquises lors de ces chirurgies.

264

Le contrôle des changements de débit et de pression de la veine porte après une hepatectomie pourrait protèger le foie restant (ou le greffon) et améliorer sa régénération post-opératoire. Les deux sujets abordés dans cette thèse ont pour but d’améliorer l’efficacité d’un dispositif médical (anneau ajustable MID-AVRTM) permettant ce contrôle. En effet, pour contrôler l’hémodynamique de la veine porte avec l’anneau, il faut tout d’abord connaitre l’impact de l’hépatectomie sur l’hémodynamique. De plus, l’efficacité de l’anneau pourrait

être vérifiée grâce à l’estimation de la fonction hépatique pendant l’opération, en utilisant la mesure defluorescence du vert d’indocyanine. Cette thèse propose des premières pistes de réflexion dans le but d’améliorer la chirurgie hépatique.

u Mots clés : Hepatectomie majeure, porc, modulation de flux, modelisation, architecture

u Laboratoire d’accueil : Laboratoire de Physiopathogénèse et Traitement des Maladies du Foie, CHB, Université

Paris-Sud 11, Faculté de médecine/pharmacie, 12 avenue Paul Vaillant Cuturier, 94200, Villejuif, France ; Directeur :

Professeur Dédier Samuel

u PÔLE : SCIENCES CHIRURGICALES UNIVERSITÉ PARIS-SUD 11

UFR «FACULTÉ DE PHARMACIE DE CHATENAY-MALABRY »

5, rue Jean Baptiste Clément

92296 CHÂTENAY-MALABRY Cedex

265

journalofsurgicalresearch april 2017 (210) 223e230

Available online at www.sciencedirect.com ScienceDirect

journal homepage: www.JournalofSurgicalResearch.com

Computerized tomographyebased anatomic description of the porcine liver

Mohamed Bekheit, MBChb, Ms.Chir, Ms.MiniChir, MRCS, MRCPS, PhD,a,b,c Petru O. Bucur, MD, PhD,a,b Mylene Wartenberg, MSc,d and Eric Vibert, MD, PhDa,b,* a Department of Liver Surgery and Transplantation, AP-HP, Hoˆpital Paul Brousse, Centre He´pato-Biliaire, Villejuif, France b Inserm Unite´ 1193, Villejuif, France c Institute of Biomedical Sciences, Division of Applied Medicine, University of Aberdeen, Aberdeen, UK d INRIA, Paris, France article info abstract

Article history: Background: The knowledge of the anatomic features is imperative for successful modeling Received 11 June 2016 of the different surgical situations. This study aims to describe the anatomic features of the Received in revised form porcine using computerized tomography (CT) scan. 28 October 2016 Methods: Thirty large, white, female pigs were included in this study. The CT image Accepted 2 November 2016 acquisition was performed in four-phase contrast study. Subsequently, analysis of the Available online 9 November 2016 images was performed using syngo.via software (Siemens) to subtract mainly the hepatic artery and its branches. Analysis of the portal and hepatic veins division pattern was Keywords: performed using the Myrian XP-Liver 1.14.1 software (Intrasense). CT Results: The mean total liver volume was 915 159 mL. The largest sector in the liver was the Anatomy right medial one representing around 28 5.7% of the total liver volume. Next in order is the Liver right lateral sector constituting around 24 5%. Its volume is very close to the volume of the Pig left medial sector, which represents around 22 4.7% of the total liver volume. The caudate lobe represents around 8 2% of the total liver volume.The portal vein did not show distinct right and left divisions rather than consecutive branches that come off the main trunk. The hepatic artery frequently trifurcates into left trunk that gives off the right gastric artery and the artery to the left lateral sector, the middle hepatic artery that supplies both the right and the left medial sectors and the right hepatic artery trunk that divides to give anterior branch to the right lateral lobe, branch to the right medial lobe, and at least a branch to the caudate lobe. Frequently, there is a posterior branch that crosses behind the portal vein to the right lateral lobe. The suprahepatic veins join the inferior vena cava in three distinct openings. There are communications between the suprahepatic veins that drain the adjacent sectors. The vein from the right lateral and the right medial sectors drains into a common trunk. The vein from the left lateral and from the left medial sectors drains into a common trunk. A separate opening is usually encountered draining the right medial sector. The caudate lobe drains separately into inferior vena cava caudal to the other veins.

* Corresponding author. AP-HP, Hoˆpital Paul Brousse, Centre He´pato-Biliaire, 12 Avenue Paul Vaillant Couturier, 94804 Villejuif Cedex, France. Tel.: 0 33 1 45 59 30 00; fax: 0 33 1 45 59 38 57. E-mail address: [email protected] (E. Vibert). 0022-4804/$ e see front matter ª 2016 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jss.2016.11.004 224 journalofsurgicalresearch april 2017 (210) 223e230

Conclusions: Knowledge of the anatomic features of the porcine liver is crucial to the performance of a successful surgical procedure. We herein describe the CT-depicted anatomic features of the porcine liver. ª 2016 Elsevier Inc. All rights reserved.

Background were kept under strict protocol. There was a period of conditioning before surgery varying from 4 to 6 d. The pigs were Complications after liver surgery might be drastic.1 To reduce housed in individual pens with temperature regulated at 23 the risk of complications, particularly after major liver resec- 1C at ambient humidity. Lighting was natural through close tion, swine has been in use as a surgical model to better study to ceiling wide windows. and understand the different surgical situations for a long time.2 This is particularly true because the animal is readily Anesthesia available at a reasonable price and that the anatomy is to an extent similar to human anatomy.3 Experiments on liver All animals were subjected to overnight fast before anesthesia regeneration after different percentages of partial hepatec- and before imaging that was performed under general anes- tomy have been performed to simulate situations with small- thesia as well. Each pig received 100 mg of xylazine 2% for-size syndrome.4 (Rompun; Bayer Healthcare, Mississauga, Ontario, Canada) One of the key factors on which the success and the with 750 mg ketamine for anesthesia induction followed by reproducibility of the experimental model are based is the tracheal intubation (6-7 mm in size, Portex; Smiths Medical, precise knowledge of the normal anatomy along with its France). Subsequently, inhalational anesthesia was started variations. In humans, an aberrant right hepatic artery origi- using a 60% FiO2 with 2% isoflurane (Isoflurane, Belmont, nating from the superior mesenteric artery or a right bile duct France) in assisted ventilation. The ventilator was set up on originating from the left bile duct could change the outcome. volume control mode delivering 350-400 mL at a rate of 17-20 Despite that the swine model has been in use for decades, cycles per minute. Crystalloid fluids were given at a rate of little information on the detailed anatomic features of the 2 mL/kg/h fasting. During imaging, animals were covered with porcine liver is available in the literature.3 blankets. Few studies have concentrated on the ultrastructure of the hepatic microarchitecture,5 others based their studies on Abdominal CT scan and volumetric study casting.3 Furthermore, detailed descriptive data from computerized tomography (CT) scan do not exist. This study Abdominal CT scan was conducted in every animal. On aimed at describing the normal porcine liver anatomy based average, an 80 mL (2 mL/kg) of iodinated contrast (Omnipaque; on CT scan. GE healthcare, Carrigtohill, Ireland) was injected through an intravenous catheter with a rate of 4 mL/s. CT scans were performed with a Somatom (Definition AS; Siemens, For- Methods chheim, Germany). The CT image acquisition was performed in four contrast phases, arterial at 15 s of injection, the early Ethical approval portal at 35 s, portal venous at 55 s of injection, and at 75 s after the injection, the hepatic venous phase was done. The study was approved by the regional committee of ethics Syngo.via software (Siemens healthcare global) was used in animal research and by the ministry of higher education in a cardiac analysis module to subtract mainly the hepatic and scientific research and ministry of agriculture and fishing artery and its branches. On launching the module, cropping of according to the European Union directives. the volume, limiting the display from the diaphragm to the celiac trunk was performed. Then, the image display param- Study setting eters were adapted to the hepatic artery. Image cropping was further performed from ventral and dorsal aspects to remove Surgeries were performed at the CIRE platform, INRA Centre the parietal vessels if necessary. Val de Loire, Nouzilly, France. Volume analysis was performed using the Myrian XP-Liver 1.14.1 software (Intrasense, Montpellier, France). Segmenta- Animals tion of the portal venous branches as well as the hepatic venous branches was performed initially. Subsequently, the Thirty large, white, female pigs were included in this study. liver parenchyma was segmented. To segment the different Females were chosen to neutralize the potential influence of hepatic sectors, each sector was segmented based on the gender on regeneration because these animals were subse- supplying portal vein, the draining hepatic vein, and the fissure quently used for that purpose. Besides, laparotomy in female boundary with the adjacent sector. Images were segmented pigs was relatively easier because urine leak from the male most successfully in the 75-s hepatic venous phase. penis located within the umbilical stalk has an extremely Verification of the segmentation was performed in CT scan unpleasant odor. The average age of the included animals was performed in two distinct postresection situations. The first 3mo 10 d and their mean weight was 35.3 5 kg. Animals was after resection of the three left hepatic sectors, and the bekheitetal ct anatomy of the porcine liver 225

Fig. 2 e Volume rendering technique extracting the part of the arterial tree in relation to the portal vein. The right gastric artery coming from the left hepatic artery depicted in anterior view. (Color version of ﬁgure is available online.)

further divide outside the liver parenchyma to supply the Fig. 1 e Branches of the celiac trunk and the hepatic artery corresponding hepatic sectors. as depicted from a three-dimensional reconstruction of the A right gastric branch supplying the lesser curvature of the segmented arterial tree from the celiac trunk. (1) common stomach coming from the left hepatic branch or less hepatic artery, (2) gastroduodenal artery, (3) the left hepatic commonly from the proper hepatic artery was consistently artery, (4) the right hepatic artery, (5) the right gastric seen (Fig. 2). The left lateral sector is supplied by a branch that artery, (6) artery to the left lateral sector, (7) artery to the left leaves the left hepatic arterial trunk, which gives off two medial sector, (8) middle hepatic artery to the right medial branches to the left lateral and the left medial sectors in most lobe, (9) posterior artery to the right lateral sector, (10) of the cases (Fig. 3) or from the proper hepatic artery. branch from the right hepatic artery that divides to supply A large middle artery that usually originates separately the right lateral and the right medial sectors, (11) left from the proper hepatic artery supplies the right medial sector gastric artery from the celiac trunk, (12) splenic artery, and but sometimes it supplies the left medial sector (Fig. 3). In the (13) celiac trunk. (Color version of ﬁgure is available online.) absence of the middle hepatic artery, the right medial sector is

second was after resection of the four sectors leaving behind only the caudate lobe in place. This was performed to verify the segmentation of the right lateral sector and the caudate lobe. All animals have been submitted to different percentages of hepatic resection.

Results

Artery

The proper hepatic artery originates from the common hepatic artery after it gives off the gastroduodenal artery. It then runs for a varying length before it divides into two or more branches. The average length of the common hepatic artery, measured from its origin from the celiac trunk to the origin of the gastroduodenal artery, was 42.8 5.3 mm, whereas the Fig. 3 e Volume rendering technique extracting the part of proper hepatic artery length was 10 7.6 mm. The usual the arterial tree in relation to the portal vein. The middle pattern of division of the proper hepatic artery is trifurcation hepatic artery supplying the left medial sector depicted in (Fig. 1). It gives off left, middle, and right hepatic trunks that anterior view. (Color version of ﬁgure is available online.) 226 journalofsurgicalresearch april 2017 (210) 223e230

Fig. 4 e Volume rendering technique extracting the part of the arterial tree in relation to the portal vein. The posterior right artery to the right lateral lobe turning behind the corresponding portal venous branch depicted in anterior view. (Color version of ﬁgure is available online.)

e supplied by a branch from the right hepatic artery that sup- Fig. 6 Three-dimensional region-of-interest plies the right lateral and the right medial sectors. This artery reconstruction for segmented portal and hepatic veins. The crosses in front of the portal vein to supply the right lateral green color represents the hepatic vein and the blue sector. In about 80% of the examined animals, another hepatic represents the portal vein. The portal venous branch is the artery is seen crossing behind the portal venous branch to the ﬁrst branch from the portal venous trunk itself to the caudate lobe depicted in caudal view. (Color version of ﬁgure is available online.)

right lateral sector to supply the posterior part of the parenchyma of the right lateral sector (Fig. 4). The caudate lobe receives its branch that crosses in front of the portal vein, from the artery to the right lateral sector or in cases of the bifurcated proper hepatic artery from the right hepatic artery trunk (Fig. 5). Each one of the previously mentioned arterial branches further divides, usually outside the hepatic parenchyma.

Portal vein

The portal vein appears to course a long track toward the left side from which the different branches appear to stem off and enter consecutively into the corresponding sectors. Usually, there is more than a single branch that leaves the main portal vein trunk to supply each sector. The first branch to take off from the portal vein trunk is the vein to the caudate lobe. The caudate lobe is supplied by a single relatively large portal vein that comes off from the posterior aspect of the portal vein before its divisions (Fig. 6). This is usually accompanied by multiple small veins that originate at different levels from the posterior aspect of the main portal trunk. After a short distance, a large branch enters the right Fig. 5 e An intraoperative photo showing the minute lateral sector nearly at the level of its anterior border. In arterial branch to the caudate lobe crossing over the portal some cases, the caudate lobe is supplied by a branch from trunk to the caudate lobe. (Color version of figure is the vein to the right lateral sector. In these cases, a single available online.) branch first leaves the main trunk to divide into a branch bekheitetal ct anatomy of the porcine liver 227

Fig. 9 e Cranial view of a three-dimensional reconstruction Fig. 7 e Three-dimensional region-of-interest of the segmented hepatic sectors showing the common reconstruction for segmented portal and hepatic veins. The trunk of draining the veins from the right lateral and one of green color represents the hepatic vein and the blue two right medial sectors. (Color version of figure is represents the portal vein. The first branch coming off is a available online.) common branch that further divides into two; one supplying the caudate lobe and inferior portion of the right lateral sector, whereas the other branch supplying the enters the substance of the left medial sector. This branch upper portion of the right lateral sector. (Color version of supplies the upper portion of the left medial sector (segment figure is available online.) IVa) with another branch that comes off more distal to supply the lower portion of this sector (segment IVb). Thereafter, the continuation of the main trunk gives two that supplies the anterior portion of the right lateral sector branches. One bridges the base of the fissure from the left (equivalent to segment VI) and another one that further medial to the left lateral sector (Fig. 8) to supply the lower divides to supply the caudate lobe as well as the posterior portion of the left lateral sector (segment III). The second portion of the right lateral parenchyma (equivalent to branch crosses to the upper portion of the left lateral sector segment VII; Fig. 7). In less frequent situations, it receives (segment II) more deep inside the parenchyma and before the dual branches. branch supplying the left medial sector. The left medial sector The next branch that leaves the portal vein is one that di- receives consistently two small branches from the left side of vides into two or directly to portal branches to supply the right the portal branch to the right medial sector. medial sector (equivalent to segments V and VIII in human livers). Further distal, the continuation of the portal vein The suprahepatic veins

Each sector is traversed by one large hepatic vein that receives smaller tributaries throughout its course. The right lateral vein unites with the vein draining the right medial sector around 1 cm inside the parenchyma before its joining to the inferior vena cava (IVC). This common trunk runs within the substance of the hepatic parenchyma beyond the base of the fissure that separates the right lateral from the right medial sectors before it opens into the IVC separately (Fig. 9). There is a consistent vein draining the deep parenchyma of the right lateral sector that joins the vein that drains mainly the right medial sector. A second hepatic vein is seen traversing the deep parenchyma of the right medial sector, which opens independently into the IVC. Less frequently, a branch draining the right medial sector is seen joining the main draining vein of the left medial sector. Fig. 8 e An intraoperative photo demonstrating portal The left medial sector is traversed by one single vein that venous branch crossing to the left lateral sector at the base drains separately into the IVC in most animals. Less of the fissure separating the left lateral from the right frequently, this branch unites with the vein from the left lateral sectors. (Color version of figure is available online.) lateral sector to form a large common opening into the IVC. 228 journalofsurgicalresearch april 2017 (210) 223e230

medial sector, right medial sector, right lateral sector, and the caudate lobe. The hepatic sectors are not quite distinct in the CT scan as they are in vivo. However, some prominent fissures are easily seen, particularly distally toward the liver border. The most prominent fissures depicted in the CT scan are the one separating the left lateral sector from the left medial one and the one between the right medial and the right lateral sectors. The fissure between the right and the left medial sectors is on the visceral surface more than the diaphragmatic one and it is seen in the CT scan only distally. The least evident one is the one separating the right lateral sector from the caudate lobe, which its visualization depends mainly on the volume of the caudate lobe. The mean total liver volume was 915 159 mL. The largest sector in the liver was the right medial one representing around 28 5.7% of the total liver volume. Next in order is the right lateral sector constituting around 24 5%. Its volume is very close to the volume of the left medial sector, which rep- Fig. 10 e Caudal view of a three-dimensional resents around 22 4.7% of the total liver volume. The reconstruction of the five hepatic sectors showing their caudate lobe represents around 8 2% of the total liver vol- volume. (1) The left lateral sector, (2) the left medial sector, ume (Fig. 10). (3) the right medial sector, (4) the right lateral sector, and (5) Controlling for the age, partial correlation technique the caudate lobe. (Color version of figure is available showed a significant correlation between the weight of the ¼ ¼ online.) animal and the total liver volume (r 0.75, P 0.001). There was no relation between the individual lobar volume and the weight. Furthermore, there was a significant negative corre- The left lateral sector drains through two well-defined lation between the volume of the left lateral and the left tributaries, one from segment II and another from segment medial sectors and between the right medial and the right III. Both branches unit into a common trunk before they lateral sectors (r ¼7.9, P < 0.001 and r ¼6.5, P ¼ 0.01 drain into the IVC. Frequently, one or two smaller separate respectively; Table). branches draining segment II join the IVC below the larger trunk. The caudate lobe drains into the IVC directly usually with a single short vein that joins that IVC on its right pos- Discussion terior aspect. However, there are innumerable veins that contribute to the drainage of this segment which is not Prior knowledge of the anatomic variation of the liver is evident in the CT scan. fundamental to perform safe and efficient surgical technique. Frequently, the portal venous branches interlace with the This study is the first to comprehensively describe the CT hepatic venous plexus of a neighboring sector, which anatomy of the porcine liver. This information is substantial frequently bridges over the plane where the fissure lies. Those for accurate surgical planning in order not to jeopardize the fissures divide the liver approximately in the distal three- to inflow or the outflow of the remnant liver after partial resec- four-fifths of the parenchyma and more toward the visceral tion. Resection without this accurate information might lead rather than the diaphragmatic surface as seen in the fissure to congestive segments or segments deprived from their between the right and left medial sectors. inflow which in turn might influence the study question. Besides the prominent fissures in pigs, there are several Liver segmentation and sectors other differences between the macro-anatomic features between the human and porcine livers. Some of these differ- The porcine liver is divided into the following five distinct ences are seen in Figures 10 and 11. The supply of the caudate sectors by four major fissures: the left lateral sector, left lobe in humans comes mainly from the left hepatic artery and

Table e The summary of the segmented lobar volumes in the studied animals. Hepatic sector/segment Mean, mL (%) SD, mL (%) Minimum, mL (%) Maximum, mL (%)

Left lateral sector 168.6 (18) 70 (6) 55 (7.3) 335 (30.5) Left medial sector 201 (22) 57.2 (4.7) 135 (12.8) 339 (31) Right medial sector 255 (28) 58 (5.7) 160 (17) 368 (40.7) Right lateral sector 216.2 (24) 63.8 (5) 121 (17) 375 (37) Caudate lobe 74.6 (8) 23.6 (2) 36 (4) 109 (13)

SD ¼ standard deviation. bekheitetal ct anatomy of the porcine liver 229

Fig. 11 e Three-dimensional reconstruction of human liver in (A) an oblique caudal view showing the main portal venous trunk (P on arrow) and its dense divisions along with the relative extension of the caudate lobe compared to the porcine liver as shown in Figure 10, (B) top view to show the hepatic veins and their junction with the IVC. (Color version of ﬁgure is available online.)

main or the left portal vein, whereas in pigs, it comes from the syndrome. However, to successfully carry out the intended right side.6 experiment, an accurate knowledge of anatomy should be The portal venous supply to the caudate lobe was reported acquired. These percentages correspond to removal of the left to come consistently from the right portal vein in the study three and four sectors respectively. by Court,3 however, this was not the case in our study. We The pattern of the vascular array and the deployment of found that in most animals, a single portal branch from the the hepatic sectors around the IVC facilitate the resection of main portal trunk is the main supplying branch to the the most peripheral sectors. For that reason, liver resection caudate lobe and in a lower frequency from the right branch in pigs is left oriented.11 In a model of 90% resection, the supplying the right lateral lobe. Furthermore, there were not remnant liver parenchyma is the caudate lobe. The portal obvious right and left divisions of the portal vein. We venous supply of this part comes directly from the main observed a long trunk from which consecutive branches portal trunk in most of the cases, however, not infrequently, come out. it originates from the portal vein supplying the right lateral One of the multiple similarities of the pattern of the blood sector. This division occurs close to the base of the fissure supply is at the left lateral sector, where the portal vein runs separating the right lateral from the caudate lobe.12 In such mostly into the fissure separating this sector from the left case, resection should be performed not to compromise this medial sector and gives out two branches to the left lateral supply, hence should be conducted above 1 cm distance sector and two to the left medial one. This is similar to the from the base of the fissure depending on the used pattern of the left portal vein divisions in humans.7 Variations technique. of the arterial branching were reported in a study by Gravante Moreover, the percentage of contribution of each sector to et al.8 They did not report however the consistent branch that the whole liver volume varies from animal to another. The leaves the proper hepatic artery to supply the stomach. This importance of this becomes evident with the increment of the artery might be mistaken as a hepatic artery if not percentage of resection. For 90% resection, which is per- differentiated. formed leaving only the caudate lobe in place. The caudate In our study, we found that the anterior branch supplying lobe might constitute to as low as 4% of the total liver volume, the right lateral sector is the consistent one and that the which might impose inaccuracies and increased risk of undue posterior branch was present in most animals. This is in mortality. contrast to what was reported by Court,3 who reported that The porcine liver is reported to be divided into five different the anterior artery is the accessory one. Noteworthy, the lobes by four fissures.3,8 In our opinion, sectors rather than discrepancy in the reported anatomy of swine originates from lobes is a more appropriate term to the anatomic character- the differences in species. Zanchet et al.9 reported only two istics of the porcine liver that retains, in general, similar common drainage sites for the suprahepatic veins into the IVC inflow and outflow patterns to the human liver.13 of Landrace pigs. In our study, the usual pattern was three opening joining the IVC. The two drainage sites were also encountered. However, the venous trunk of the left medial sector frequently joins separately the IVC. Conclusions The hereby mentioned pattern of divisions and arrangement has some technical implications. In liver resection Knowledge of the anatomic features of the porcine liver is model, 70% and 90% models are most commonly used to study crucial to the performance of a successful surgical procedure. the hemodynamic changes induced by resection.10 These We hereby describe the CT-depicted anatomic features of the models generally aim to study the behavior of small-for-size porcine liver. 230 journalofsurgicalresearch april 2017 (210) 223e230

and clinical temporary ex vivo perfusion. J Surg Res. Acknowledgment 1966;6:117e120. 3. Court FG, Wemyss-Holden SA, Morrison CP, et al. Segmental The authors would like to acknowledge of Myle`ne Wartenberg nature of the porcine liver and its potential as a model for e at the INRIA, Paris, France and Hans Adriansen, Francois Le experimental partial hepatectomy. Br J Surg. 2003;90:440 444. 4. Iida T, Yagi S, Taniguchi K, et al. Improvement of Compte at the INRA, Tours, France, for their role in data morphological changes after 70% hepatectomy with acquisition. portocaval shunt: preclinical study in porcine model. J Surg The source of funding: This study was funded mainly by Res. 2007;143:238e246. the “Agence de la Biomedicine” through its program of 5. Debbaut C, Vierendeels J, Casteleyn C, et al. Perfusion Research (AOR 2009). Eric Vibert, Petru O. Bucur, Mohamed characteristics of the human hepatic microcirculation Bekheit acknowledge funding by project ANR-13-TECS-0006 based on three-dimensional reconstructions and (IFlow). computational fluid dynamic analysis. J Biomech Eng. 2012;134:011003. Authors’ contributions: M.B. and P.B. have equally 6. Heloury Y, Leborgne J, Rogez J, et al. The caudate lobe of the contributed to this study and they are the first authors. M.B. liver. Surg Radiol Anat. 1988;10:83e91. contributed to study concept and design. M.B., P.B., and M.W. 7. Cho A, Asano T, Yamamoto H, et al. Relationship between contributed to acquisition of data. M.B., P.B., and M.W. right portal and biliary systems based on reclassification of contributed to analysis and interpretation of data. M.B. draf- the liver. Am J Surg. 2007;193:1e4. ted the article. M.B., P.B., M.W., and E.V. approved the final 8. Gravante G, Ong SL, Metcalfe MS, et al. The porcine hepatic arterial supply, its variations and their influence on the version of the article. extracorporeal perfusion of the liver. J Surg Res. 2011;168:56e61. Disclosure 9. Zanchet DJ, Montero EFS. Pig liver sectorization and segmentation and virtual reality depiction. Acta Cir Bras. 2002;17:381e387. The authors reported no proprietary or commercial interest in 10. Darnis B, Mohkam K, Schmitt Z, et al. Subtotal hepatectomy any product mentioned or concept discussed in the article. in swine for studying small-for-size syndrome and portal inflow modulation: is it reliable? HPB (Oxford). e references 2015;17:881 888. 11. Kahn D, Hickman R, Terblanche J, et al. Partial hepatectomy and liver regeneration in pigsdthe response to different resection sizes. J Surg Res. 1988;45:176e180. 1. Allard MA, Adam R, Bucur PO, et al. Posthepatectomy portal 12. Martins ACA, Machado MAC, Ferraz A´ AB. Porcine liver: vein pressure predicts liver failure and mortality after major experimental model for the intra-hepatic glissonian liver resection on noncirrhotic liver. Ann Surg. approach. Acta Cir Bras. 2008;23:204e207. 2013;258:822e829. discussion 829-830. 13. Belghiti J, Clavien P, Gadzijev E. The Brisbane 2000 2. Norman JC, Franco FO, Brown ME, et al. Techniques of terminology of liver anatomy and resections. HPB 2000; 2:333- obtaining and preparing the porcine liver for experimental 339. HPB (Oxford). 2002;4:99e100.

[Downloaded free from http://www.translsurg.com on Tuesday, January 03, 2017, IP: 86.132.192.189]

Original Article

The Reference Values for Hepatic Oxygen Consumption and Net Lactate Production, Blood Gasses, Hemogram, Major Electrolytes, and Kidney and Liver Profiles in Anesthetized Large White Swine Model

1,2 1,2 3, 1,2 Mohamed Bekheit , Petru Bucur , Christian Andres Eric Vibert 1Departments of Liver Surgery and Transplantation, Inserm, Unité 1193, Villejuif, France; 2AP‑HP – Hôpital Paul Brousse, Centre Hépato‑Biliaire, Villejuif, France; 3Service de Biochimie, CHRU Tours, Tours, France

Address for correspondence: Dr. Eric Vibert, AP‑HP – Hôpital Paul Brousse, 12 Avenue Paul Vaillant‑Couturier, 94804 Villejuif Cedex, France. E‑mail: [email protected] Received: 06-09-2016, Accepted: 09-12-2016

Abstract

Aim: Pigs are extensively used as experimental models to study the human physiology and pathophysiological conditions. Knowledge of the normal values of the commonly used parameters is indispensable to the correct interpretation of the test results. This study reports on the normal hemogram, blood gas, major electrolytes, kidney and liver profiles, hepatic oxygen consumption, and net lactate production in a large white pig model. Methods: Twenty-five female large white pigs were included in this study. Blood gas samples were collected from the portal and hepatic veins as well as the carotid artery. Results: The mean hemoglobin level was 97.7 ± 15.8 g/L. white blood cells were 13.5 ± 3.3 103/mm3, and platelet count was 279 ± 104.6 103/mm3. The mean aspartate aminotransferase, alanine aminotransferase, gamma amyltransferase,‑g lu t alkaline phosphatase, total bilirubin, and direct bilirubin were 83.8 ± 73.9 IU/L, 43.7 ± 5.9 IU/L, 33.6 ± 8.6 IU/L, 296.5 ± 39.7 IU/L, 5.6 ± 3.2 µmol/L, and 1.6 ± 0.73 µmol/L. The mean albumin level was 29 ± 3.9 g/L. The mean ammonia and arterial lactate levels were 49.1 ± 45.67 mmol/L and 1.5 ± 0.46 mmol/L. Kidney profile parameters were comparable to human values. Hepatic oxygen consumption was 17.3 ± 9.7 mL/100 g liver tissue/min and net hepatic lactate production was 0.017 ± 0.03 mmol/L. Conclusion: Knowledge of the normal parameters is mandatory for accurate interpretation of the experimental results that involves large white animals.

Key words: Blood gases, hepatic, lactate, liver, normal, oxygen, reference, swine

Introduction withdrawal presents challenges. For these reasons, blood samples are usually collected under anesthesia. However, Animals are extensively used in research to simulate various anesthesia has no significant effect on several measurements.8 clinical situations, and large white pigs are among those Knowledge of normal biochemistry and hematology values widely used as a large surgical model.1,2 In many situations, under practical experimental conditions is, therefore, blood samples are required at different stages of the various 3,4 fundamental. experiments conducted on these animals. The aim of this study was to describe the normal values of the Pigs, as well as other animals, are frequently infl by blood gas, major electrolytes within the portal and systemic the change in the environment in a manner that might alter blood, and the hepatic and renal function parameters. the results of the laboratory tests.5 Particularly, the blood 6 7 gas samples are sensitive to stress, similar to humans. In addition, restraining large animals for blood sample This is an open access article distributed under the terms of the Creative Commons Attributi onCommercial eAlike‑Shar 3.0 License, which allows others to remix, tweak, and build upon the work non ommercially,‑c as long as the author is credited and the new creati Access this article online are licensed under the identi al terms. Quick Response Code: For reprints contact: [email protected] Website: www.translsurg.com How to cite this article: Bekheit M, Bucur P, Vibert E, Andres C. The reference values for hepatic oxygen consumption and net lactate production, blood gasses, hemogram, major electrolytes, and kidney and liver profiles in DOI: anesthetized large white swine model. Transl Surg 2016;1:95-100. 10.4103/2468-5585.197495

[Downloaded free from http://www.translsurg.com on Tuesday, January 03, 2017, IP: 86.132.192.189]

Bekheit, et al.: Normal values for hematological and biochemical parameters in pigs

Methods • One milliliter of venous blood was withdrawn from the right suprahepatic vein by direct puncture using a Ethical approval 30-gauge needle Approval of the committee of ethics of animal research, • One milliliter of portal venous blood was withdrawn from Ministry of Higher Education and Scientific Research, and the portal vein by direct puncture of the portal vein using Ministry of Agriculture and Fishing was obtained, complying a 30-gauge needle. with the European Union Directive N° 2010/63/EU.

Study setting Hepatic oxygen consumption is calculated as described Surgeries were performed at the CIRE platform, INRA Centre elsewhere.9 The hepatic oxygen consumption is calculated Val de Loire, Nouzilly, France. Samples were taken to the central using the equation: laboratory at the University Hospital of Tours for analysis. ([hemoglobin (Hb) g/dL × 1.34 × SaO + 0.003 × PaO ] × hepatic 2 2

Animals arterial flow [HAF]) + ([Hb × 1.34 × SpO2 + 0.003 × PpO 2] ×

Twenty-five large white female pigs were included in the study. portal vein flow [PVF]) − ([Hb × 1.34 × SvO 2 + 0.003 × PvO 2] The average age was 3 months ± 15 days. The mean weight × [HAF + PVF]). [Equation 1] was 35.9 ± 7.5 kg. The flow used in the equation is adjusted to the liver volume Preoperative preparation by a simple transformation (flow × 100/liver volume). Animals fasted the night before the surgery. A preanesthetic The net hepatic lactate production is calculated using the same preparation was given to the animals in their individual cages principle: on the day of surgery as 30 mg/kg ketamine (KETAMINE PANPHARMA 250 mg/5 mL) and 0.03 mg/kg acepromazine (Arterial lactate × HAF) + (Portal lactate × PVF) − (Hepatic (Calmivet, Vetoquinol, France). venous lactate × [HAF + PVF]). [Equation 2] Anesthesia Data presentation and analysis Inside their cages, each pig received 2 mL xylazine 2%, with The data are presented in the summative form. Mean or 15 mL ketamine (KETAMINE PANPHARMA 250 mg/5 mL) median was used as central tendency measures, and multiple dispersion measures were given for precision and to facilitate for anesthesia induction. A tracheal tube, size 6–7 mm exploitation. The data summary was created using SPSS, (Portex, France), was then placed and secured. Subsequently, version 21 (IBM®SPSS®, Chicago, IL, USA). inhalational anesthesia was started using 60% FiO2 inhalational oxygen mixed with 2% isofl (Isofl Belamont, Centre des Spécialités Pharmaceutiques, France) at a rate of 2–3 mL Results mixed with 1.5–2 L/min oxygen in 1.5 L of air. In the gas analysis, the mean arterial pH was 7.393 ± 0.09117, Blood sample withdrawal the venous pH was 7.378 ± 0.168, and the portal pH was 7.335±0.08816.ThemeanarterialpCO was45.89±12.54mmHg, The right internal jugular vein was cannulated with an 8 Fr 2 the venous pCO was 55.37 ± 13.36 mmHg, and the portal vascular Desivalve (Vygon, Ecouen, France) cannula. The 2 pCO was 56.82 ± 13.73 mmHg. The mean arterial pO was vascular cannulation is performed through a midline neck 2 2 273.6 ± 63.8 mmHg, the venous pO was 37.32 ± 9.108 mmHg, incision that is used for contemporaneous carotid artery 2 cannulation for the hemodynamic monitoring. and the portal pO 2 was 45.65 ± 10.12 mmHg. The mean

arterial SO2 was 99.81 ± 0.1311%, the venous SO 2 was Thirty milliliters of blood was withdrawn from the jugular 64.27 ± 14.89%, and the portal SO 2 was 75.35 ± 8.802%. The vein and sent to the central laboratory unit in universal tubes mean arterial lactate was 1.462 ± 0.4654 mmol/L, the venous for analysis. lactate was 1.272 ± 0.3431 mmol/L, and the portal lactate was After the overnight fast, blood samples were collected 1.451 ± 0.4228 mmol/L [Tables 1-3]. by venipuncture 20 min after the induction of anesthesia. The mean blood urea nitrogen (BUN) level was Samples for hematological analysis were collected in 3.208 ± 1.348 mmol/L and the mean creatinine level ethylenediaminetetraacetic acid tube. For chemistry analysis; was 95.5 ± 20.51 µmol/L. The mean plasma sodium level was samples were collected in fluoride tubes and buffered sodium 138.3 ± 2.188 mEq/L and potassium 4.875 ± 0.9196 mEq/L citrate. Serum was mixed with 0.8 mg aprotinin, centrifuged at [Table 4]. 2000 ×g at 4°C for 15 min within 30 min after puncture, and The mean total bilirubin level was 5.583 ± 3.175 µmol/L and incubated in an ice container until delivery. Delivery time to the mean total protein level was 50.67 ± 6.88 g/L. The mean the laboratory occurred within 45 min following centrifugation. ammonia level was 49.1 ± 45.67 mmol/L and the mean lactate Blood samples for gas analysis after midline abdominal level was 2.013 ± 0.9827 mmol/L. Table 5 summarizes the incision: main hepatic functions and enzyme levels. The mean venous • One milliliter of arterial blood was withdrawn from the Hb level was 97.67 ± 15.84 g/L and the mean hematocrit level right carotid artery was 29.41 ± 5.037%. The mean venous white blood cell was

96 Translational Surgery ¦ Oct ‑2016D ec ¦ Volume 1 ¦ Issue 4

[Downloaded free from http://www.translsurg.com on Tuesday, January 03, 2017, IP: 86.132.192.189]

Bekheit, et al.: Normal values for hematological and biochemical parameters in pigs

Table 1: Arterial blood gasses, pH, and lactate levels

pCO (mmHg) pO (mmHg) SO (%) 2 2 2 pH Lactate (mmol/L) Minimum 28.7 166.5 99.6 7.22 0.92 25% percentile 37.4 218.4 99.7 7.353 1.035 Median 44.25 276.6 99.85 7.38 1.365 75% percentile 50.6 327.4 99.9 7.43 1.78 Maximum 72.3 375.7 100 7.55 2.4 Mean 45.89 273.6 99.81 7.393 1.462 SD 12.54 63.8 0.1311 0.09117 0.4654 SEM 3.619 18.42 0.03786 0.02632 0.1097 Lower 95% CI of mean 37.93 233 99.73 7.335 1.23 Upper 95% CI of mean 53.86 314.1 99.89 7.45 1.693 SD: Standard deviation, SEM: Standard error of mean, CI: Confidence interval

Table 2: Venous blood gasses, pH, and lactate levels The mean arterial Hg was 8.508 ± 1.738 g/dL and the mean arterial hematocrit is 25.00 ± 5.081%. The mean portal Hb level

pCO 2 pO 2 SO 2 pH Lactate was 8.625 ± 1.805 g/dL and the hematocrit was 25.18 ± 5.564% (mmol/L) [Table 8]. Minimum 36.6 19.8 23.9 7.18 0.83 25% percentile 47.28 32.98 59.28 7.268 1.028 The mean arterial calcium level was 0.8961 ± 0.2321 mmol/L, Median 53.45 36.05 65.85 7.34 1.145 the venous calcium level was 0.7861 ± 0.2276 mmol/L, 75% percentile 59.5 45.63 74.53 7.433 1.47 and the mean portal calcium was 0.77 ± 0.2326 mmol/L Maximum 84.1 53.6 80.9 7.83 2.14 [Table 9]. Mean 55.37 37.32 64.27 7.378 1.272 SD 13.36 9.108 14.89 0.168 0.3431 Discussion SEM 3.857 2.629 4.299 0.04849 0.08087 Lower 95% CI 46.88 31.53 54.81 7.271 1.101 Reported here is what is expected to be the first of a of mean series of comprehensive routine baseline values in large Upper 95% CI 63.86 43.1 73.73 7.484 1.442 white pigs. Samples were collected under anesthesia, and of mean xylazine‑ke tamine10 was used for induction. This protocol was 10 SD: Standard deviation, SEM: Standard error of mean, CI: Confi interval reported not to influence the blood gas analysis.

The lowest calcium concentration was found in samples Table 3: Portal vein blood gasses, pH, and lactate levels collected from the portal vein. Kallner11 found that

postprandial calcium is lower than before meals. In our pCO pO SO Lactate 2 2 2 pH animals, the pH of the portal venous blood was lower than (mmol/L) Minimum 35.1 33.7 61.7 7.2 1.05 in other sample types. This makes the alkaline tide effect, in 25% percentile 46.3 38.3 67.85 7.265 1.105 which an increase in the pH would result in a reduction of 12 Median 56.45 45.1 76.05 7.325 1.28 the ionized calcium level, unlikely to be the explanation 75% percentile 64.95 49.53 83.73 7.373 1.743 for this difference given that the lactate levels in the portal Maximum 78.6 71.7 89 7.52 2.45 blood were the highest. Mean 56.82 45.65 75.35 7.335 1.451 The difference between the arterial and venous concentration SD 13.73 10.12 8.802 0.08816 0.4228 of calcium is known in the literature.13 Perhaps, this difference SEM 3.963 2.921 2.541 0.02545 0.09966 is related to the increased protein contents14 in the portal blood Lower 95% CI 48.09 39.22 69.76 7.279 1.24 of mean which reduces the ionized calcium levels, which we did not test Upper 95% CI 65.54 52.08 80.94 7.391 1.661 in our experimental setting. Calcium levels in our experiments 14 of mean were generally higher than values in humans. SD: Standard deviation, SEM: Standard error of mean, CI: Confi interval Perhaps, the elevated serum calcium is a contributing factor 13.51 ± 3.355 103/mm3 and the mean platelets count was to the higher coagulability of blood of animals as compared 15 279 ± 104.6 103/mm3 [Table 6]. to that of humans. In this study, we found that the normal coagulation profile of large white pigs exceeds the human T he m e an he pa t ic oxyge n c onsumpti on was profile. Since kits dedicated for humans were used, the ability 17.32 ± 9.671 mL/100 g liver tissue/min and the mean net to accurately identify the activity except that it exceeded 100%. hepatic lactate production was 0.01682 ± 0.02719 mmol/L It was previously reported that prothrombin activity is higher in [Table 7]. pigs than in humans.16 Analysis of the coagulation profile was

Translational Surgery ¦ Oct ‑2016D ec ¦ Volume 1 ¦ Issue 4 97

[Downloaded free from http://www.translsurg.com on Tuesday, January 03, 2017, IP: 86.132.192.189]

Bekheit, et al.: Normal values for hematological and biochemical parameters in pigs

Table 4: BUN, creatinine, and major electrolytes levels in anesthetized animals

BUN (mmol/L) Creatinine (µmol/L) Na (mEq/L) K (mEq/L) Cl (mEq/L) Minimum 1.5 70 135 3.9 96 25% percentile 2.275 75 137 4.3 98 Median 2.95 91 137 4.65 99 75% percentile 4.25 113 140.8 5.275 101 Maximum 5.8 129 142 7.3 105 Mean 3.208 95.5 138.3 4.875 99.5 SD 1.348 20.51 2.188 0.9196 2.355 SEM 0.3891 5.922 0.6317 0.2655 0.6798 Lower 95% CI of mean 2.352 82.47 136.9 4.291 98 Upper 95% CI of mean 4.065 108.5 139.7 5.459 101 SD: Standard deviation, SEM: Standard error of mean, CI: Confidence interval, BUN: Blood urea nitrogen

Table 5: Hepatic profile from samples collected from the internal jugular vein

AST ALT GGT ALP Bilirubin total Bilirubin Proteins (g/L) Albumin (g/L) Ammonia Lactate (IU/L) (IU/L) (IU/L) (IU/L) (µmol/L) direct (mmol/L) (mmol/L) (µmol/L) Minimum 31 34 19 231 2 0.7 41 24 10 0.9510 25% percentile 33.25 39 27.25 266.8 3 1 46 25 21.75 1.421.75 Median 41.5 43 34 305.5 4.5 1.65 50.5 30 31.5 1.6531.5 75% percentile 139.8 47.5 40.5 333.5 8.75 2 56.5 32 68.5 2.98368.5 Maximum 258 53 46 344 11 3 64 35 164 3.72164 Mean 83.75 43.67 33.58 296.5 5.583 1.6 50.67 29.18 49.1 2.01349.1 SD 73.97 5.867 8.628 39.67 3.175 0.7336 6.88 3.92 45.67 0.982745.67 SEM 21.35 1.694 2.491 11.45 0.9167 0.2118 1.986 1.182 14.44 0.347514.44 Lower 95% CI 36.75 39.94 28.1 271.3 3.566 1.134 46.3 26.55 16.43 1.19116.43 of mean Upper 95% CI 130.7 47.39 39.07 321.7 7.601 2.066 55.04 31.82 81.77 2.83481.77 of mean SD: Standard deviation, SEM: Standard error of mean, CI: Confidence interval, AST: Aspartate aminotransferase, ALP: Alkaline phosphatase, ALT: Alanine aminotransferase, ALP: Alkaline phosphatase, GGT: Gamma-glutamyltransferase

Table 6: Major hemogram parameters in samples collected from the internal jugular vein

Hemoglobin (g/L) Hematocrit (%) WBCs (103/mm3) Platelets (103/mm3) Minimum 71 19.5 9.4 141 25% percentile 85 26.7 10.15 212.3 Median 97 30.3 13.8 236.5 75% percentile 105 31.2 15.4 394.5 Maximum 133 40.1 20.4 442 Mean 97.67 29.41 13.51 279 SD 15.84 5.037 3.355 104.6 SEM 4.775 1.519 1.061 33.07 Lower 95% CI of mean 87.03 26.03 11.11 204.2 Upper 95% CI of mean 108.3 32.79 15.91 353.8 SD: Standard deviation, SEM: Standard error of mean, CI: Confidence interval, WBC: White blood cell

not precise in our animals since their values were beyond the BUN and serum creatinine levels are within ranges that reference values in our laboratory. Platelet count, on the other correspond to the normal values of healthy humans.21 Similarly, 17 22 hand, was comparable to that of the human level. serum sodium and chloride levels are equivalent to those in healthy humans. Potassium levels, however, were slightly The difference in Hb concentration was small between the 23 higher than in humans. different samples; however, both Hb and hematocrit levels

were higher in venous blood than in arterial and portal blood The pCO2 levels were significantly lower in arterial samples samples. These values are lower than in humans.18,19 In addition, compared to venous and portal venous samples with a smaller 20 24 leukocyte count was slightly higher than in healthy humans. difference than what was found in other studies. The arterial

98 Translational Surgery ¦ Oct ‑2016D ec ¦ Volume 1 ¦ Issue 4

[Downloaded free from http://www.translsurg.com on Tuesday, January 03, 2017, IP: 86.132.192.189]

Bekheit, et al.: Normal values for hematological and biochemical parameters in pigs

values of pCO 2 in our animals were slightly higher than the might contribute to the relatively higher unconjugated bilirubin 25 32 corresponding values in humans. However, this increase fraction compared to the human individual.

in the arterial pCO 2 values might be attributed to the effect Oxygen tension and saturation were both higher in portal of anesthesia8 or as a result of ketamine ‑respiratoryinduced 10 samples than in venous samples. The difference in oxygen depression. The gap between the arterial and venous pCO 2 might reflect some degree of tissue hypoxia in anesthetized saturation between the arterial and venous samples was 33 animals26 with a resultant increase in pCO production from reported to be an indicator of tissue oxygen delivery. 2 anaerobic metabolism.27 This was not, however, associated However, upon observation, this interpretation is not feasible with a significant reduction in pH, which might suggest that due to the artificial increase in oxygenation of animals under different physio-metabolic pathways play a major role in the anesthesia. Posthepatectomy, hepatic oxygen consumption 34 difference in the observed values. was identified as a mediator for liver regeneration. Similarly, serum lactate has been described as a prognostic marker after Lactate levels were lower in the venous samples, taken from the 35 live resection. suprahepatic vein, than in the arterial or the portal vein. This implies that the liver actively metabolizes lactate.28 However, Hepatic transaminase levels were higher in these animals than a net lactate production was seen in these animals, which is reported for healthy humans.36 Both direct and total bilirubin contrary to the findings observed in transplanted individuals.29 levels were within a similar range to that described for healthy Nevertheless, it should be noted that the amount of splanchnic humans. lactate is influenced by the fraction of inspired oxygen,30 which in the current study’s protocol is relatively high. Conclusion Lactate values in the portal blood were higher than in venous To ensure accurate interpretation of the experimental results, samples. This might be attributed to the activity of the different it is important to be aware of the reported porcine normal intestinal flora in the production of lactate.31 In addition, this hematological and biological parameters. Some differences between these parameters and the corresponding human Table 7: Calculated hepatic net oxygen consumption and parameters do exist, and it should be considered before lactate production extrapolation of the human parameter values onto the porcine values. This study confirms agreement between some human Hepatic oxygen Net hepatic and porcine parameters and reports some differences as well. consumption lactate production (mL/100 g liver (mmol/L) Acknowledgment tissue/min) The authors would like to acknowledge of Hans Adriansen, Minimum 4.293 −0.0297 Francois Le Compte at the INRA, Tours, France, for their role 25% percentile 9.357 0.005325 in data acquisition. Median 15.25 0.0109 75% percentile 24.36 0.02693 Financial support and sponsorship Maximum 37.60 0.0861 This study was funded mainly by the “Agence de la Mean 17.32 0.01682 Biomedecine” through its program of Research (AOR 2009). SD 9.671 0.02719 Eric Vibert, Petru O. Bucur, and Mohamed Bekheit acknowledge SEM 2.792 0.007850 funding by project ANR-13-TECS-0006 (IFlow). Lower 95% CI of mean 11.18 −0.0004615 Upper 95% CI of mean 23.47 0.03409 Conflicts of interest SD: Standard deviation, SEM: Standard error of mean, CI: Confi interval There are no conflicts of interest.

Table 8: Hemoglobin and hematocrit levels measured in each sample type of gas analysis

Arterial Venous Portal Hemoglobin (g/dL) Hematocrit (%) Hemoglobin (g/dL) Hematocrit (%) Hemoglobin (g/dL) Hematocrit (%) Minimum 5.700 17.00 5.300 16.00 5.800 17.00 25% percentile 7.200 21.50 6.950 20.25 7.800 23.00 Median 8.400 24.50 8.850 26.00 8.450 23.00 75% percentile 9.400 27.75 10.98 32.25 10.25 31.00 Maximum 12.00 35.00 16.90 49.00 11.70 34.00 Mean 8.508 25.00 9.425 27.67 8.625 25.18 SD 1.738 5.081 3.484 10.08 1.805 5.564 SEM 0.5017 1.467 1.006 2.911 0.5209 1.678 Lower 95% CI of mean 7.404 21.77 7.212 21.26 7.478 21.44 Upper 95% CI of mean 9.613 28.23 11.64 34.07 9.772 28.92 SD: Standard deviation, SEM: Standard error of mean, CI: Confidence interval

Translational Surgery ¦ Oct ‑2016D ec ¦ Volume 1 ¦ Issue 4 99

[Downloaded free from http://www.translsurg.com on Tuesday, January 03, 2017, IP: 86.132.192.189]

Bekheit, et al.: Normal values for hematological and biochemical parameters in pigs

16. Miura N, Kawaguchi H, Nagasato T, Yamada T, Ito T, Izumi H, Table 9: Calcium levels as measured by gas analyzer in Shameshima H, Miyoshi N, Tanimoto A, Maruyama I. Coagulation the arterial, venous, and portal venous samples activity and white thrombus formation in the microminipig. In Vivo 2013;27(3):357-61. Ca (mmol/L) Ca (mmol/L) Ca (mmol/L) 17. Brecher G, Cronkite EP. Morphology and enumeration of human blood arterial venous portal platelets. J Appl Physiol 1950;3(6):365-77. Minimum 0.42 0.42 0.4 18. Burman D. Haemoglobin levels in normal infants aged 3 to 24 months, 25% percentile 0.76 0.6575 0.5675 and the effect of iron. Arch Dis Child 1972;47(252):261-71. 19. Janus J, Moerschel SK. Evaluation of anemia in children. Am Fam Median 0.935 0.78 0.83 Physician 2010;81(12):1462-71. 75% percentile 1.06 0.845 0.95 20. Tembe N, Joaquim O, Alfai E, Sitoe N, Viegas E, Macovela E, Maximum 1.29 1.41 1.26 Gonçalves E, Osman N, Andersson S, Jani I, Nilsson C. Reference Mean 0.8961 0.7861 0.77 values for clinical laboratory parameters in young adults in Maputo, SD 0.2321 0.2276 0.2326 Mozambique. PLoS One 2014;9(5):e97391. 21. Hosten AO. BUN and creatinine. In: Kenneth WH, Dallas HW, SEM 0.0547 0.05365 0.05481 Willis HJ, editors. Clinical Methods: The History, Physical, and Lower 95% CI of mean 0.7807 0.6729 0.6544 Laboratory Examinations. Boston: Butterworths; 1990. p. 874-8. Upper 95% CI of mean 1.012 0.8993 0.8856 22. Ackerman GL. Serum sodium. In: Walker HK, Hall WD, Hurst JW, SD: Standard deviation, SEM: Standard error of mean, CI: Confi interval editors. Clinical Methods: The History, Physical, and Laboratory Examinations. Boston: Butterworths; 1990. p. 879-83. 23. Rastegar A. Serum potassium. In: Walker HK, Hall WD, Hurst JW, References editors. Clinical Methods: The History, Physical, and Laboratory 1. Kahn D, Hickman R, Terblanche J. A porcine model for the study of Examinations. Boston: Butterworths; 1990. p. 884-7. liver regeneration. Invest Surg 1988;1(2):139-42. 24. Martin GB, Carden DL, Nowak RM, Tomlanovich MC. Comparison of 2. Pouyet M, Paquet C, Mechet I, Le Derf Y, Bernard P, Figueiredo P, central venous and arterial pH and PCO2 during open-chest CPR in the Berger F, Boillot O. Hemodynamic tolerance and rapid hypertrophy of canine model. Ann Emerg Med 1985;14(6):529-33. a hepatic graft corresponding to less than 30% of the ideal mass in pigs. 25. Trulock EP III. Arterial Blood Gases. In: Walker HK, Hall WD, Gastroenterol Clin Biol 2001;25(10):869-74. Hurst JW, editors. Source Clinical Methods: The History, Physical, and rd 3. Swindle MM, Smith AC. Perioperative care of swine. J Invest Surg Laboratory Examinations. 3 ed., Ch. 49. Boston: Butterworths; 1990. 2004;26(2):63-71. 26. Van der Linden P, Rausin I, Deltell A, Bekrar Y, Gilbart E, Bakker J, 4. Taniguchi H, Takada Y, Fukunaga K, Yuzawa K, Otsuka M, Todoroki K, Vincent JL. Detection of tissue hypoxia by arteriovenous gradient for Fukao K. Establishment of a swine model for auxiliary partial orthotopic PCO2 and pH in anesthetized dogs during progressive hemorrhage. liver transplantation. Transplant Proc 1998;30(7):3232-6. Anesth Analg 1995;80(2):269-75. 5. Giral M, Armengol C, Sánchez-Gómez S, Gavaldà A. Effects of changing 27. Hannon JP, Wade CE, Bossone CA, Hunt MM, Loveday JA. Oxygen to individually ventilated caging on guinea pigs (Cavia porcellus). J Am delivery and demand in conscious pigs subjected to fixed-volume Assoc Lab Anim Sci 2015;54(3):267-72. hemorrhage and resuscitated with 7.5% NaCl in 6% Dextran. Circ 6. Gelsomino S, Lorusso R, Livi U, Romagnoli S, Romano SM, Shock 1989;29(3):205-17. Carella R, Lucà F, Billè G, Matteucci F, Renzulli A, Bolotin G, 28. Adeva-Andany M, López-Ojén M, Funcasta-Calderón R, De Cicco G, Stefàno P, Maessen J, Gensini GF. Assessment of a Ameneiros-Rodríguez E, Donapetry-García C, Vila-Altesor M, continuous blood gas monitoring system in animals during circulatory Rodríguez-Seijas J. Comprehensive review on lactate metabolism stress. BMC Anesthesiology 2011;11:1. doi:10.1186/1471-2253-11-1. in human health. Mitochondrion 2014;17:76-100. doi: 10.1016/j. 7. Wallbridge PD, Hannan LM, Joosten SA, Irving LB, Steinfort DP. mito.2014.05.007. Review. PubMed PMID: 24929216. Clinical utility of sequential venous blood gas measurement in the 29. Murphy ND, Kodakat SK, Wendon JA, Jooste CA, Muiesan P, Rela M, assessment of ventilatory status during physiological stress. Intern Med Heaton ND. Liver and intestinal lactate metabolism in patients with J 2013;43(10):1075-80. acute hepatic failure undergoing liver transplantation. Crit Care Med 8. Andersen JB, Wang T. Effects of anaesthesia on blood gases, acid-base 2001;29(11):2111-8. status and ions in the toad Bufo marinus. Comp Biochem Physiol Part A 30. Pastor CM. Hepatic and splanchnic oxygen consumption during Mol Integr Physiol 2002;131(3):639-46. acute hypoxemic hypoxia in anesthetized pigs. Crit Care Med 9. Shiba H, Zhu X, Arakawa Y, Irefin S, Wang B, Trenti L, Fung JJ, 2000;28(3):765-73. Kelly DM. Oxygen consumption predicts outcome in porcine partial 31. Bongaerts GP, Tolboom JJ, Naber AH, Sperl WJ, Severijnen RS, liver grafts. J Surg Res 2014;189(2):335-9. Bakkeren JA, Willems JL. Role of bacteria in the pathogenesis of 10. Jacobson C. A novel anaesthetic regimen for surgical procedures in short bowel syndrome-associated D-lactic acidemia. Microb Pathog guineapigs. Lab Anim 2001;35(3):271-6. 1997;22(5):285-93. 11. Kallner A. Preanalytical procedures in the measurement of ionized 32. Chambers AC, Bismohun SL, Davies H, White P, Patil AV. Predictive calcium in serum and plasma. Eur J Clin Chem Clin Biochem value of abnormally raised serum bilirubin in acute appendicitis: a cohort 1996;34(1):53-8. study. Int J Surg 2015;13:207-10. doi: 10.1016/j.ijsu.2014.11.041. 12. Seamonds B, Towfighi J, Arvan DA. Determination of ionized calcium PubMed PMID: 25498496. in serum by use of an lon-selective electrode. I. Determination of 33. Shiao SY. Accurate measurements of oxygen saturation in neonates: normal values under physiologic conditions, with comments on the effects Paired arterial and venous blood analyses. Newborn Infant Nurs Rev of food ingestion and hyperventilation. Clin Chem 1972;18(2):155-60. 2005;5(4):170-8. 13. Bilkovski R, Cannon C, Adhikari S, Nasr I. Arterial and venous 34. Ren YS, Qian NS, Tang Y, Liao YH, Liu WH, Raut V, Tao KS, Dou KF. ionized calcium measurements: Is there a difference? Ann Emerg Med Beneficial effects of splenectomy on liver regeneration in a rat model of 2004;44(4):S56. massive hepatectomy. Hepatobiliary Pancreat Dis Int 2012;11(1):60-5. 14. Walker HK, Hall WD, Hurst JW, editors. Clinical Methods: The History, 35. Wrighton LJ, O’Bosky KR, Namm JP, Senthil M. Postoperative Physical, and Laboratory Examinations. 3rd ed., Ch. 143. Boston: management after hepatic resection. J Gastrointest Oncol 2012;3(1):41-7. Butterworths; 1990. 36. Uemura H, Katsuura-Kamano S, Yamaguchi M, Sawachika F, 15. Siller-Matula JM, Plasenzotti R, Spiel A, Quehenberger P, Jilma B. Arisawa K. Serum hepatic enzyme activity and alcohol drinking status Interspecies differences in coagulation profile. Thromb Haemost in relation to the prevalence of metabolic syndrome in the general 2008;100(3):397-404. Japanese population. PLoS One 2014;9(4):e95981.

100 Translational Surgery ¦ Oct ‑2016D ec ¦ Volume 1 ¦ Issue 4 Investigative Radiology

Measuring the hepatic blood flow: An applied comparison between the phase contrast MR imaging and the transit time perivascular probes in a porcine model. --Manuscript Draft--

Manuscript Number: Full Title: Measuring the hepatic blood flow: An applied comparison between the phase contrast MR imaging and the transit time perivascular probes in a porcine model. Article Type: Original Article Keywords: blood flow, phase contrast MRI, transit time ultrasound, porcine model, liver surgery. Corresponding Author: Mohamed Bekheit, MBChB, MSc Chir, MSc MiniChir, MRCS, MSc stat, PhD University of Aberdeen Institute of Medical Sciences Aberdeen, Aberdeen UNITED KINGDOM First Author: Mohamed Bekheit, MBChB, MSc Chir, MSc MiniChir, MRCS, MSc stat, PhD First Author Secondary Information: Order of Authors: Mohamed Bekheit, MBChB, MSc Chir, MSc MiniChir, MRCS, MSc stat, PhD chloe Audebert, PhD Petru Bucur, MD, PhD Hans Adriaensen, PhD Mylene Wartenberg, PhD student Irene vignon-clementel, PhD Eric Vibert, Md, PhD Manuscript Region of Origin: FRANCE Abstract: Background: The hepatic hemodynamics are essential parameters in surgical planning as well as in various disease processes. The transit time technology is widely used in clinical practice to evaluate the hepatic inflow, yet invasive. Being non-invasive, the phase- contrast-MRI (PC-MRI) is potentially attractive in assessing the hepatic blood flow. A comparison between the hepatic inflow rates measured using the PC-MRI to the transit time perivascular flow meters is the aim of this study.

Methods: Eighteen large white pigs anesthetized for PC-MRI and Surgery under the unified protocol. The flow was measured in the Hepatic Artery (Qha), the Portal Vein (Qpv), and the Aorta above the Celiac Trunk (Qca) using PC-MRI and compared to the transit time values. Two trained independent observers observed measurements. The Bland- Altman method was conducted.

Results: The mean Qpv measured in PC-MRI was 0.55 ± 0.12 l/min, while in the transit time was 0.74 ± 0.17 l/min. The average flow in the Qca was 1.4 ± 0.47 l/min in the PC-MRI and 2 ± 0.6 l/min in the transit time. The Qha was 0.17 ± 0.1 l/min in the PC-MRI and was 0.13 ± 0.06 l/min in the transit time.

The Bland-Altman method revealed an estimate of bias of 32% with a 95% CI= -49%- 15% for the Qca. A 17% bias with a 95% CI= -15 to 51% was observed for the Qha. There was a 40% bias in the estimated Qpv values in the PC-MRI, 95% CI=-62-18%. The TT had higher weight in predicting adverse outcomes after 75% resection compared to the PC-MRI (B=0.35-0.43 versus B=0.22-0.07, for architecture changes and premature death, respectively).

Conclusion: There is a tendency of the PC-MRI to underestimate the flow measured by the TT

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation probes. A linear relation is found between the two methods, and a correction term could be implemented. Suggested Reviewers: Opposed Reviewers:

Dear Editor,

We are delighted to submit our study entitled “Measuring the hepatic blood flow: An applied

comparison between the phase contrast MR imaging and the transit time perivascular probes in

a porcine model” for consideration of publication in your prestigious journal “investigative

radiology”, which perfectly fits the manuscript source and targets vast majority of its audience.

The study beforehand is focused in assessing the reliability of the phase contrast MRI in

replacing the transit time technology in estimating the hepatic blood flow. Therefore, we

conducted a series of animal experiments that was carried out in our laboratory in the central

region of France. The experiments included measuring the portal vein and hepatic artery as well

as the aorta above the celiac trunk using both the PC-MRI and transit time technology. Once

data were registered, the post processing was performed by two independent observers for

each method ie: each two were trained in one method. Then an intraclass correlation was

conducted to assess the inter-observer reliability. Further to that, a Bland-Altman analysis was

performed to assess the interchangeability of both methods. We thereafter constructed a PLS

model to assess the weight of each measure against the prediction of postoperative adverse

events.

The results of this study are interesting in many ways, particularly since it is the first of its kind

to assess the hepatic blood flow. Moreover, it addresses the potential over-estimation of the

role of this tool, which is being more and more emphasized in the literature.

We are looking forward to hearing from you

On behalf of the team

Mohamed Bekheit

MBChB, MSc Chir, MSc MiniChir, MSc Biostatistics, MRCS, PhD Surgical Sciences Best regards Title Page

Measuring the hepatic blood flow: An applied comparison between the phase contrast

MR imaging and the transit time perivascular probes in a porcine model.

Mohamed Bekheit 1,2

Chloe Audebert 3, 4

Petru Bucur 1, 5

Hans Adriaensen 6

Emilie Bled, 6

Mylène Wartenberg, 3

Irene Vignon-Clementel 3, 4

Eric Vibert 1,2

Authors` affiliations

1. Inserm Unité 1193, 112 Boulevard Paul Valliant Cuturier, Villejuif, France

2. Institute of medical sciences, University of Aberdeen, Aberdeen, UK

3. Centre de recherche Inria de Paris, Paris, France

4. Sorbonne Universités, UPMC Univ. of Paris 6, Laboratoire Jacques-Louis Lions,

Paris, France

5. AP-HP, Hôpital Paul Brousse, Centre Hépato-Biliaire, Villejuif, France

6. CHRU, hôpitaux de Tours, Chirurgie Hépato-biliaire et Pancréatique, Transplantation Hépatique, Tours, France

7. Plateforme Chirurgie et Imagerie pour le Recherche et l’Enseignement (CIRE),

Imagerie, UMR-PRC, 37380 Nouzilly, Centre INRA Val de Loire, France

Corresponding author:

Eric Vibert, MD, PhD 12 Avenue Paul Vaillant Couturier

94804 Villejuif Cedex

Tel: 00 33 1 45 59 30 00

Fax: 00 33 1 45 59 38 57

E-Mail: [email protected]

Manuscript word count: 3049

Authors` contribution:

Mohamed Bekheit and Chloe Audebert equally contributed in this manuscript and are both first authors.

Study concept and design: Petru Bucur, Mohamed Bekheit, Eric Vibert, Irene Vignon-

Clementel

Acquisition of data: Hans Adriaensen, Emilie Blend, Mylène Wartenberg, Chloe

Audebert, Mohamed Bekheit, Petru Bucur

Analysis and interpretation of data: Mohamed Bekheit, Petru Bucur, Chloe Audebert,

Mylène Wartenberg

Drafting of the manuscript: Mohamed Bekheit, Chloe Audebert.

Critical revision of the manuscript: Petru Bucur, Eric Vibert, Irene Vignon-Clementel

Final approval: Irene V-C, Eric Vibert

Disclosure and Conflict of interest statement:

MB, CA, PB, HA, EB, MW, IVC, EV declare no conflict of interest

Acknowledgment

Acknowledgment:

The authors are grateful to the INRA Plateforme CIRE (Nouzilly, France) staff for their

technical assistance in surgeries and imaging.

Source of funding:

This study was funded mainly by the “Agence de la Biomedecine” through its program of

Research (AOR 2009). Eric Vibert, Irene Vignon-Clementel, Petru O. Bucur, Mohamed

Bekheit, Mylène Wartenberg and Chloe Audebert acknowledge funding by project ANR-

13-TECS-0006 by the Agence Nationale de la Recherche.

Original manuscript (BLINDED)

Measuring the hepatic blood flow: An applied comparison between the phase contrast

MR imaging and the transit time perivascular probes in a porcine model.

Abstract

Background:

The hepatic hemodynamics are essential parameters in surgical planning as well as in

various disease processes. The transit time technology is widely used in clinical practice

to evaluate the hepatic inflow, yet invasive. Being non-invasive, the phase-contrast-MRI

(PC-MRI) is potentially attractive in assessing the hepatic blood flow. A comparison

between the hepatic inflow rates measured using the PC-MRI to the transit time

perivascular flow meters is the aim of this study.

Methods:

Eighteen large white pigs anesthetized for PC-MRI and Surgery under the unified

protocol. The flow was measured in the Hepatic Artery (Qha), the Portal Vein (Qpv), and

the Aorta above the Celiac Trunk (Qca) using PC-MRI and compared to the transit time

values. Two trained independent observers observed measurements. The Bland-Altman

method was conducted.

Results:

The mean Qpv measured in PC-MRI was 0.55 ± 0.12 l/min, while in the transit time was

0.74 ± 0.17 l/min. The average flow in the Qca was 1.4 ± 0.47 l/min in the PC-MRI and 2

± 0.6 l/min in the transit time. The Qha was 0.17 ± 0.1 l/min in the PC-MRI and was 0.13

± 0.06 l/min in the transit time.

The Bland-Altman method revealed an estimate of bias of 32% with a 95% CI= -49%-

15% for the Qca. A 17% bias with a 95% CI= -15 to 51% was observed for the Qha.

There was a 40% bias in the estimated Qpv values in the PC-MRI, 95% CI=-62-18%.

The TT had higher weight in predicting adverse outcomes after 75% resection compared to the PC-MRI (B=0.35-0.43 versus B=0.22-0.07, for architecture changes and premature death, respectively).

Conclusion:

There is a tendency of the PC-MRI to underestimate the flow measured by the TT probes.

A linear relation is found between the two methods, and a correction term could be implemented.

Keywords: hepatic blood flow, phase contrast MRI, transit time ultrasound, porcine model, liver surgery.

Running title: Comparing the PC-MRI hepatic flow measures to transit time

Introduction:

Alteration of the hepatic hemodynamics is considered a key contributing factor to the development of failure or success following hepatic surgery (1). Hepatic blood flow is an indicator of the condition of the liver (2) and may determine its functional reserve and subsequently, serves as decision-making tool (3). The need for an accurate estimate of hepatic flow parameters in the perioperative period has been demonstrated in a previous study (4).

The perivascular Transit Time technology is an established but invasive ultrasound based method for measuring the blood flow in a variety of clinical settings (5–7).

The Doppler Ultrasound is an alternative to the transit time method and relatively non- invasive (8). Nonetheless, the Doppler flow measurement is subject to high interobserver variability; up to 24% (9). On the other hand, the variability of the phase-contrast-MRI

(PC-MRI) flow readings is low; indicating its usefulness in determining absolute blood flow values (10).

The performance of PC-MRI in hepatic flow measurement was discriminative in the classification of the cirrhosis stage, unlike the measurements obtained using the Doppler ultrasound indicating a clinical relevance (11). It is also a non-invasive tool in the hands of clinical radiologists that would mitigate invasive measurements. In modern surgical practice; there is a strong trend towards non-invasive predictive tools to measure these parameters. Notably, this is important in the assessment and surveillance of diseases such as portal hypertension, where frequent measurements are required and the capability of such measures to predict important perioperative events – such as premature death or

4 architectural damage as indicator of hyperperfusion syndrome – is of paramount importance

Therefore, this study aimed at implementing a practical comparison between the performances of the phase contrast MRI in measuring the hepatic blood inflow and a commonly used clinical tool; the transit time ultrasound technology. The basic assumption is that the two methods can be used interchangeably.

Methods

Ethical approval:

The study was approved by the regional committee of ethics in animal research, and by the ministry of higher education and scientific research and ministry of agriculture and fishing, according to European Union directives.

Study setting:

The surgeries were performed at the Chirurgie et Imagerie pour le Recherche et l’Enseignement (CIRE) platform, INRA Centre Val de Loire, Nouzilly, France.

Animals:

Eighteen large white female pigs were included in this study. The average age was 3 months ± 9 days and weight was 37 ± 3 kg. Animals were kept under strict protocol.

There was a period of conditioning before experiments varying from 4 to 7 days. The pigs were housed in individual pens with temperature regulated at 23 ± 1°C at ambient humidity and natural lighting.

Pre-imaging preparation followed the same preoperative preparation protocol.

Preparation for imaging and surgery:

The PC-MRI was performed 5 days before surgical measurements. Animals were fasted by night and on the day of experiments were given a pre-anaesthetic preparation of 30 mg/kg ketamine (Ketamin, Panpharma) and 0.03 mg/kg acepromazine (Calmivet,

Vetoquinol, France).

Anaesthesia:

Each pig received 100 mg of xylazine 2% (Rompun, Bayer Healthcare) with 750 mg ketamine (Ketamin Panpharma) for induction followed by intubation (6-7 mm in size,

Portex, France). Subsequently, inhalational anesthesia was started using a 60% fraction of inspired oxygen (FiO2) inhalational oxygen mixed with 2% Isoflurane (Isoflurane,

Belmont, France) at a rate of 2-3 ml mixed with 1.5-2 l/min oxygen in 1.5 liters of air.

Animals were then transferred immediately to the Imaging or the Surgical Unit as per protocol to minimize the time lag between induction and measurements.

Afterward, and for surgery only, 1 gram of amoxicillin (Amoxycillin, Mylan) was given i.v. Fentanyl (1 ml) was given subcutaneously after intubation and at the end of surgery and a Fentanyl patch was placed on the shaved skin of the right side of the thoracic cage.

20 mg of atracurium (Tracrium, Glaxo Smith Klyne) muscle relaxant was given intramuscularly. Crystalloids were infused at a rate of 2 ml/kg/h fasting in addition to

500-1000 ml.

The rationale for the choice of the vessels:

The primary study focus was to examine the possibility of estimating the hepatic blood flow using the PC-MRI and how dependable could it be in replacing the transit time technology. There is a potentially considerable difference in the caliber and the flow rate of the hepatic artery –small in caliber and flow - and the portal vein, which is high in caliber and flow. Therefore, we opted to use the aortic artery above the celiac trunk to provide a referee with a high flow and large caliber characteristics.

Image acquisition:

Spatially registered functional flow information was acquired simultaneously with the morphological data. The flow was measured in the Hepatic Artery (Qha), the Portal Vein

(Qpv), and the Celiac Aorta (Qca) using the phase contrast MRI and compared to the flow measurements obtained from the intraoperative transit time values.

The intrinsic motion sensitivity of PC-MRI was used to quantify blood flow (12,13).

Consequently, velocities were calculated from the pixel intensity of the Phase Contrast

(PC) images. In this study, a 3T scanner (Siemens Magnetom® Verio) available at the

Chirurgie et Imagerie pour le Recherche et l’Enseignement (CIRE) Platform was used.

The gating technique uses the quiescent expiratory phase for data acquisition and reconstruction. The parameters for these anatomical images were as follows:

3 - T2 True FISP coronal: 2x1.6x3 mm , acquisition time: 1.40 min, TR = 3.85 ms, TE =

1.67 ms, Field-of-View (FOV): 400 mm², matrix: 256x204;

3 - T2 True FISP axial: 1.6x1.3x3 mm , acquisition time: 3.04 min, TR = 3.95 ms, TE =

1.66 ms, FOV: 400 mm², matrix: 320x256;

3 - T2 True FISP sagittal: 1.2x0.9x3 mm , acquisition time: 1.18 min, TR = 3.95 ms, TE =

1.66 ms, FOV: 300 mm², matrix: 320x256; moreover, with the following common parameters: flip angle = 60 degrees, bandwidth:

488 Hz/Px, slice thickness: 3 mm and an integrated Parallel Acquisition Technique

(iPAT) of 2.

Velocity images were acquired using gradient-echo fast phase-contrast pulse sequences.

PC-MRI was performed during mechanical ventilation with retrospective cardiac gating and 20 different phases were calculated to enable extrapolation if necessary. The data acquisition window was defined by the trigger point and the trigger window values. The trigger point defined the point at which data acquisition began if a valid pulse trigger was detected. The trigger window defined the period of the respiratory cycle in which data acquisition did not occur. This parameter, which depends on the animal’s heart rate, was set so as to include 2–3 respiratory rate intervals in the available imaging time. Each series of reconstructed data consisted of phase images associated with the corresponding magnitude images. Flow rates were calculated from 20 velocity images spanning the cardiac cycle.

MR parameters included a minimum TR (ca. 75 ms), a minimum TE (ca. 4.5 ms), an

FOV of 160x160 mm and a matrix size of 192 x 134 resulting in an in-plane resolution of

1.2x0.8 mm². The slice thickness was set to 5 mm, a flip angle of 25 degrees was used, the bandwidth was 554 Hz/Px, 5 views per segment and an iPAT of 2 was set up. The acquisition time was heart rate dependent, therefore lasting more or less 2 min per flow

8 sequence. Encoding velocities were set to 80, 140 and 20 cm/s for the hepatic artery, the celiac aorta, and the portal vein respectively. The blood flow acquisition plane was selected strictly perpendicular to these 3 different vessels. These parameters – including the encoding velocities – were identified with experiments on 3 pigs, prior to the real experiment and found to be the best to avoid aliasing, which could have been otherwise an issue.

The DICOM images exported into the Syngo.Via (Siemens Healthcare) software. The analysis was performed using an automated Region Of Interest (ROI) selector when possible (figure 1 b-d). In case that the operator judged a non-accurate detection of the

ROI onto the vessel of interest of the 20 different PC and magnitude images, manual

ROIs selections were then performed. All post-processing was conducted by two independent observers who were not involved in processing the data obtained from the transit time.

This postprocessing tool allowed automatic reconstruction- of the temporal flow curves Q

(t) during a cardiac cycle, and calculation of the principal parameters of the flow curves.

This flow rate was defined as the product of the mean pixel velocity value inside the lumen and the vessel cross-sectional area.

Surgery:

During surgery, animals were covered with heat blankets, and gastric aspiration through an orogastric tube was performed if gastric distension was observed at surgery. Surgery was under sterile conditions. Animals were positioned in a dorsal decubitus, and surgical

9 site disinfection and draping were routine. Oxygen saturation, temperature, pulse rate, arterial pressure, and central venous pressure were monitored during surgery.

For flow measurements, we used TranSonic transit time flowmeters (TranSonic System

Inc, Ithaca, USA). A 12 or 16 mm flow probe (MA12 PAU or MA16 PAU) surrounded the portal vein, and a 4 or 6 mm flow probe (MA4PSS or MA 6PSS) was used around the hepatic artery, proximal to the gastroduodenal artery. A 10 or 12 mm flow probe (MA

10/12 PAU) was positioned around the aortic artery. The probes size was chosen fitting to the vessel – without compression – to avoid sliding, which is also assisted by the probe design. The perivascular flow probes were connected to Transonic® T403

(TransonicSystems, Inc., Ithaca, NY, USA).

Measurements were continuously recorded – while no manipulation is conducted to avoid interference – with a 16-channel amplifier connected to a computer running IOX2 acquisition software (Emka TECHNOLOGIES, Paris, FRANCE).

Values were obtained from the mean of 20 seconds of continuous recording (figure 1), which is comparable to the 20 cardiac cycles gated images recordings in the phase contrast MRI. After that, data corresponding to the flow measures at the end of expiration was extracted from the consecutive cycles to simulate the measurement situation during

PC-MRI. Furthermore, data was obtained at a heart-rate that was as close as possible to the heart-rate during acquisition of the PC-MRI measurements. This process was repeated independently by two observers who were blinded from the PC-MRI results.

Three left lateral liver lobes were resected in all animals following a stable signal acquisition according to the technique described by Bucur et al. (14). This type of resection results in an average of 32% residual liver volume as detailed described (15).

Liver biopsy and scoring of the microarchitectural changes: Tissue biopsy was taken prior to and one hour after the resection and scores were given to specimens in standard hematoxylin-eosin and Masson trichrome stains based on criteria adapted from Demetris et al. (16). Five criteria were formulated: A= presence of inflammatory infiltrate and necrosis [sub-items: neutrophil infiltrate, lymphocytes infiltrate, and hepatocellular necrosis], B=sinusoidal dilatation, C= ductular proliferation, D= steatosis [sub-items: macrovesicular/ballooning, microvesicular], and E=regeneration [sub-items: pseudo-nodular, mitotic activity, and acinar formation]. Each item is graded from 0=no change, 1=mild, 2=moderate, to 3=severe, in samples taken from deep liver tissue for a total score between 0 and 30 at each time point. Experienced, but blinded on the temporal factor of the specimens, pathologist performed all pathological analyses.

Statistical methods:

We used means and standard deviations to present data summary. Intraclass classification coefficient was used to examine the interobserver reliability of the PC-MRI and transit- time measurements separately. For this purpose, we used SPSS V 22 (IBM ®, SPSS ®

Statistics for Windows. Armonk, NY: IBM Corp.). The Bland-Altman analysis was used to examine the agreement of readings between the PC-MRI and those measured directly using the transit time method. The percentage presented the estimate of bias between the two methods. From Bland-Altman bias estimation, a linear relationship between the two measurements was introduced. Let k, XMR, XTT denote the estimated bias from Bland-

Altman, the PC-MRI flow measurement, and the transit time flow measurement, respectively. They relate by definition as:

XMR – XTT = 0.5 k (XMR + XTT),

11 which implies that

XTT = XMR *(2-k)/(2+k).

Data was analyzed in XLSTAT ® Pearson edition, Addinsoft.

A simplified partial least square model was constructed based on the clinical knowledge, incorporating the two methods as predictive factors comparing their path coefficients.

Latent variables were constructed with a clinically relevant outcome (premature death or parenchymal damage score following resection) and each group of measurements were used in a composite of a separate latent variable. The model was built using WarpPLS version 6.0 available from (http://www.scriptwarp.com/warppls).

Results

The mean blood flow in the portal vein measured in the PC-MRI was 0.55 ± 0.12 l/min, while in the transit time measurement it was 0.74 ± 0.17 l/min. The mean flow in the celiac aorta was 1.4 ± 0.47 l/min in the PC-MRI and 2 ± 0.6 l/min in the transit time readings. The hepatic artery flow was 0.17 ± 0.1 l/min in the PC-MRI, while it was 0.13

± 0.06 l/min in the transit time measurements (figure 2a).

The intraclass correlation coefficients for the PC-MRI measurements were 0.92, 0.93,

0.94 for the portal vein, the hepatic artery, and the celiac aorta, respectively). Similarly, for the transit time measurements, were 0.99, 0.98, 0.99 for the portal vein, the hepatic artery, and the celiac aorta. These coefficients were for the average measures, and very

12 close results (>0.9) were obtained for single measurements as well. All p-values were

<0.001 for these coefficients.

There was a significant difference between the blood flow measured in the celiac aorta with PC-MRI and with transit time (t=3.3, p=0.0023). Bland-Altman revealed an estimate of bias of 32% with a 95% CI= -49%-15% (figure 2b).

The perioperative transit time measurement in the celiac aorta was then estimated, with the Bland-Altman bias, multiplying the pre-operative PC-MRI flow measurement by

1.38. The average absolute difference between the actual transit time measurement and the estimation was 0.45 l/min.

There was no significant difference between the mean blood flow in the hepatic artery measured with PC-MRI or with the transit time probes (t=1.4, p=0.18). Bland-Altman identified a 17% bias with a 95%CI= -15 to 51%. (figure 2c). The perioperative transit time flow can be estimated by multiplying the PC-MRI flow measurement by 0.84, with an average absolute error of 0.08 l/min.

(figure 2d). For portal flow, the bias is almost constant for all measurements, with an average absolute error of 0.21 l/min, the perioperative portal flow measurement can be evaluated by multiplying the PC-MRI flow by 1.5. To further re-enforce the validity of these findings, there was no correlation between the average of both methods and their difference for each vessel (p>0.05).

There was an initial significant increase (i.e., worsening) in the histological scoring for the micro-architectural changes induced by resection (figure 3).

The PLS model:

For the architectural changes, the average path coefficient (APC)=0.287, P=0.014, the average R-squared (ARS)=0.143, P=0.090, the average adjusted R-squared,

(AARS)=0.093, P=0.140, and the average block VIF (AVIF)=1.039. The transit time measures had a higher weight in estimating the incidence of premature than the PC-MRI measures (p<0.01, p=0.07, respectively) (figure 4a).

For premature death, the average path coefficient (APC)=0.251, P=0.024, the average R- squared (ARS)=0.218, P=0.038, the average adjusted R-squared (AARS)=0.172,

P=0.067, the average block VIF (AVIF)=1.286. The transit time measures had a higher weight in estimating the incidence of premature than the PC-MRI measures (p=0.002, p=0.3, respectively) (figure 4b).

Discussion

To our knowledge, this is the first study comparing the hepatic blood flow measured using transit time ultrasound to PC-MRI. Herein, there was an agreement between both methods. The purpose of this comparison was to identify how much would the two methods be interchangeable in assessment of the hepatic flow.

There is no ideal method for flow measurement that is readily available for a clinical setting and up to now, a reference method for comparison of the performance of PC-MRI in assessing hepatic flow does not exist (17). There are many limitations with flow measurements using the transit time method being invasive, requiring additional

14 dissection, which increases the surgical burden. Furthermore, contact temperature, the type of the acoustic coupling medium, and the vessel probe fit can influence the reading significantly as reported in the instructions for use (18). Noteworthy, the probe diameter might affect the flow readings by up to one-third(19). Likewise, the ratio of the outer to the inner diameter of the vessel such as in the hepatic artery could impact the flow readings significantly (20). Moreover, the surgical manipulation could have induced arterial vasospasm (21).

Some of these factors are shared with the PC-MRI. For instance, identification of the diameter is an important determinant factor for the accuracy of the Doppler reading (22), which is similar to the importance of pixel identification in determining the flow value in

PC-MRI. Over-evaluation of the area leads to overestimation of the flow. One factor to which this might be attributed is the sensitivity of the pixel measurement (23). We have observed such influence when selecting manually versus an automatic contour selection, which might affect the average of flow calculation.

The advantage of PC-MRI for flow measurements is that it is a non-invasive procedure, compared to the transit time. It also has minimal interobserver variability (24), hence better reproducibility. Our findings support this statement. Therefore, PC-MRI was considered a reliable method for the flow measurement (25). The PC-MRI flow measurement would enable estimating the parameters of the predictive model of hemodynamics pre-operatively and therefore perform the simulation before the surgery

(26).

The results from the hepatic artery flow rates measured intraoperatively or with PC-MRI can be used with that perspective as absolute values. This is similar to the finding driven by another study (17). However, results from transit time and PC-MRI are different in absolute values of the portal vein, and the celiac aorta and a correction coefficients were required to estimate the perioperative flow with the PC-MRI measurements.

The microarchitectural changes are indicative for the quality of regeneration and mortality is related to an extent to the hemodynamic alterations induced by liver resection

(27). The results from the PLS model suggests that a higher weight is given to the TT measures in predicting the post-operative mortality and the damage to the microarchitecture. This might indicate – given the important bias estimated with Bland-

Altman – that the two methods are not interchangeable.

One of the possible explanations of the observed bias is the effect of the time lapse between both types of measurements, which was accompanied by an increase in the animals` age. The laparotomy, the fluid infusion, and the dissection of the vessels may also impact the transit time measurements and could be responsible for at least some of the observed differences. There is also an important increase in the hepatic flow parameters following the laparotomy incision of the abdomen (6). This might indicate that the observed difference could be attributed to the fact that the abdomen was opened during imaging.

In our study, animals had the same mechanical ventilation setting in surgery and imaging.

Moreover, the compared flow data were taken at the end of expiration in both methods.

Therefore, less likely that the readings were affected by the breathing cycle as was observed – up to 40% - in other studies (28).

In the study by Yzet (25), all readings obtained by Doppler were higher than those estimated by PC-MRI. This is similar to a certain extent to the finding of our study. This study demonstrates that the absolute values of the transit time measurements are not interchangeable with the PC-MRI. However, a correction coefficient could be used for that, but caution should be exercised when adopting the PC-MRI as a tool for blood flow measurement.

Conclusions:

The hepatic artery blood flow values driven from the PC-MRI could be used as an alternative to the transit time method. However, a correction factor is required for the portal vein and the aorta above the celiac trunk to compensate for the systematic underestimation.

Figures` legends:

Figure 1: a) Transit time flow measurement over time. Transit time flow measurements in the portal vein (PV), hepatic artery (HA) and celiac aorta (aorta above celiac trunk) (CA) over time.b-d) PC-MRI flow measurements. ROI on celiac aorta

(lumen) with b) phase contrast image, c) magnitude image and d) temporal flow rate curve showing the 20 phase contrast points of a cardiac cycle.

Figure 2: a) Error bar for blood flow measurements. Error Bar representation of the blood flow measurements in the celiac aorta (Qca), portal vein (Qpv) and the hepatic artery (Qha) for PC-MRI (MR) and transit time ultrasound (TT). P value is significant below 0.05. b-d) Bland-Altman plot for celiac aorta flow measurements. b) Bland-

Altman plot is indicating the systematic difference between the flow reading in PC-MRI and transit time in the celiac aorta. c)Bland-Altman plot is indicating the no-significant difference in estimation of the flow in the hepatic artery using either PC-MRI or the transit time method. d) Bland-Altman plot for portal vein flow measurement. Bland-

Altman plot is revealing the systematic underestimation of the PC-MRI flow readings compared to the transit time flow readings in the portal vein.

Figure 3: Box-plot demonstrating the temporal changes in the architecture following resection.

Figure 4: PLS model path with coeeficients. The TT (transit time) measures and the PC-

MRI (phase contrast MRI) measures were used to build the model predicting the architectural changes (left) and the premature death (right).

References:

1. Hessheimer AJ, Fondevila C, Taurá P, Muñoz J, Sánchez O, Fuster J, et al. Decompression of the portal bed and twice-baseline portal inflow are necessary for the functional recovery of a “small-for-size” graft. Ann Surg. 2011;253(6):1201–10.

2. Tziafalia C, Vlychou M, Tepetes K, Kelekis N, Fezoulidis IV. Echo-Doppler measurements of portal vein and hepatic artery in asymptomatic patients with

hepatitis B virus and healthy adults. J Gastrointestin Liver Dis. 2006;15(4):343–6.

3. Nanashima A, Shibasaki S, Sakamoto I, Sueyoshi E, Sumida Y, Abo T, et al. Clinical evaluation of magnetic resonance imaging flowmetry of portal and hepatic veins in patients following hepatectomy. Liver Int. 2006;26(5):587–94.

4. Gontarczyk GW, Łagiewska B, Pacholczyk M, Trzebicki J, Jureczko L, Kołacz M, et al. Intraoperative blood flow measurements and liver allograft function: preliminary results. Transplant Proc. 38(1):234–6.

5. Huet P, Villenevue J, Marleau D, Viallet A. Hepatic circulation: Applicable human methodology. Hepatic circulation in health and disease Raven Press, New York. 1981;57.

6. Payen DM, Fratacci MD, Dupuy P, Gatecel C, Vigouroux C, Ozier Y, et al. Portal and hepatic arterial blood flow measurements of human transplanted liver by implanted Doppler probes: interest for early complications and nutrition. Surgery. 1990;107(4):417–27.

7. Ohnishi K, Saito M, Sato S, Terabayashi H, Iida S, Nomura F, et al. Portal hemodynamics in idiopathic portal hypertension (Banti’s syndrome). Comparison with chronic persistent hepatitis and normal subjects. Gastroenterology. 1987;92(3):751–8.

8. Chow PKH, Yu W-K, Soo K-C, Chan STF. The measurement of liver blood flow: a review of experimental and clinical methods. J Surg Res. 2003;112(1):1–11.

9. De Vries PJ, van Hattum J, Hoekstra JB, de Hooge P. Duplex Doppler measurements of portal venous flow in normal subjects. Inter- and intra-observer variability. J Hepatol. 1991;13(3):358–63.

10. Lycklama à Nijeholt GJ, Burggraaf K, Wasser MN, Schultze Kool LJ, Schoemaker RC, Cohen AF, et al. Variability of splanchnic blood flow measurements using MR velocity mapping under fasting and post-prandial conditions-comparison with echo- Doppler. J Hepatol. 1997;26(2):298–304.

11. Annet L, Materne R, Danse E, Jamart J, Horsmans Y, Van Beers BE. Hepatic flow parameters measured with MR imaging and Doppler US: correlations with degree of cirrhosis and portal hypertension. Radiology. 2003;229(2):409–14.

12. Lotz J, Meier C, Leppert A, Galanski M. Cardiovascular flow measurement with phase-contrast MR imaging: basic facts and implementation. Radiographics. 22(3):651–71.

13. Markl M. Velocity encoding and flow imaging. University Hospital Freiburg, Dept of Diagnostic Radiology, Germany. 2005;

14. Bucur P, Bekheit M, Audebert C, Vignon-Clementel I, Vibert E. Simplified technique for 75% and 90% hepatic resection with hemodynamic monitoring in a large white swine model. J Surg Res. 2016;209:122–30.

15. Bekheit M, Bucur PO, Wartenberg M, Vibert E. Computerized tomography-based anatomic description of the porcine liver. J Surg Res. Elsevier; 2016;

16. Demetris AJ, Kelly DM, Eghtesad B, Fontes P, Wallis Marsh J, Tom K, et al. Pathophysiologic observations and histopathologic recognition of the portal hyperperfusion or small-for-size syndrome. Am J Surg Pathol. 2006;30(8):986–93.

17. Vermeulen MAR, Ligthart-Melis GC, Buijsman R, Siroen MPC, van de Poll MCG, Boelens PG, et al. Accurate perioperative flow measurement of the portal vein and hepatic and renal artery: a role for preoperative MRI? Eur J Radiol. 2012;81(9):2042–8.

18. Transonic. Keys to Accurate Flow Measurements with Transit-Time Ultrasound [Internet]. Business Park Stein 205 6181 MB Elsloo The Netherlands; Available from: http://www.transonic.com/resources/research/keys-to-accurate-perivascular- flow-measurements-with-transit-time-ultrasound/

19. Chow PK, Yu WK, Ng TH, Ong HS, Ooi PJ, Chan ST, et al. Influence of respiration and portal pressure on transabdominal duplex Doppler ultrasound measurement of portal blood flow: a porcine model for experimental studies. J Surg Res. 2000;89(1):66–73.

20. Edgerton RH. The effect of arterial wall thickness and conductivity on electromagnetic flowmeter readings. Medical and biological engineering. Springer; 1968;6(6):627–36.

21. Kelly DM, Zhu X, Shiba H, Irefin S, Trenti L, Cocieru A, et al. Adenosine restores the hepatic artery buffer response and improves survival in a porcine model of small-for-size syndrome. Liver Transpl. 2009;15(11):1448–57.

22. Carlisle KM, Halliwell M, Read AE, Wells PN. Estimation of total hepatic blood flow by duplex ultrasound. Gut. 1992;33(1):92–7.

23. Polzin JA, Alley MT, Korosec FR, Grist TM, Wang Y, Mistretta CA. A complex- difference phase-contrast technique for measurement of volume flow rates. J Magn Reson Imaging. 5(2):129–37.

24. Andersson C, Kihlberg J, Ebbers T, Lindström L, Carlhäll C-J, Engvall JE. Phase- contrast MRI volume flow - a comparison of breath held and navigator based

acquisitions. BMC Med Imaging. 2016;16(1):26.

25. Yzet T, Bouzerar R, Allart J-D, Demuynck F, Legallais C, Robert B, et al. Hepatic vascular flow measurements by phase contrast MRI and doppler echography: a comparative and reproducibility study. J Magn Reson Imaging. 2010;31(3):579–88.

26. Audebert C, Bucur P, Bekheit M, Vibert E, Vignon-Clementel IE, Gerbeau J-F. Kinetic scheme for arterial and venous blood flow, and application to partial hepatectomy modeling. Comput Methods Appl Mech Eng. Elsevier; 2016;

27. Bucur PO, Bekheit M, Audebert C, Othman A, Hammad S, Sebagh M, et al. Modulating Portal Hemodynamics With Vascular Ring Allows Efficient Regeneration After Partial Hepatectomy in a Porcine Model. Ann Surg. 2017;

28. Kurol M, Forsberg L. Ultrasonographic investigation of respiratory influence on diameters of portal vessels in normal subjects. Acta Radiol Diagn (Stockh). 27(6):675–80.

Figure Click here to download Figure 001.tif Figure Click here to download Figure 002 Bland Altman and difference.tif Figure Click here to download Figure 003 Damage scores before and after.tif Figure Click here to download Figure 004 PLS model.tif

Click here to access/download Copyright Transfer and Disclosure Form Bekheit coi_disclosure.pdf

Click here to access/download Copyright Transfer and Disclosure Form Chloe coi_disclosure.pdf

Click here to access/download Copyright Transfer and Disclosure Form Emelie coi_disclosure.pdf

Click here to access/download Copyright Transfer and Disclosure Form Eric coi_disclosure.pdf

Click here to access/download Copyright Transfer and Disclosure Form Hans coi_disclosure.pdf

Click here to access/download Copyright Transfer and Disclosure Form Irene coi_disclosure.pdf

Click here to access/download Copyright Transfer and Disclosure Form Mylene coi_disclosure.pdf

Click here to access/download Copyright Transfer and Disclosure Form Petru coi_disclosure.pdf journalofsurgicalresearch march 2017 (209) 122e130

Available online at www.sciencedirect.com ScienceDirect

journal homepage: www.JournalofSurgicalResearch.com

Simpliﬁed technique for 75% and 90% hepatic resection with hemodynamic monitoring in a large white swine model

Petru Bucur, PhD,a,b Mohamed Bekheit, MBChB, MSc (Chir), MSc (Minivasive), MRCS, MRCPS, PhD,a,b Chloe Audebert, PhD,c,d Irene Vignon-Clementel, PhD,c,d and Eric Vibert, MDa,b,* a Centre He´pato-Biliaire, Hoˆpital Paul Brousse, AP-HP, Villejuif, France b Inserm Unite´ 1193, Villejuif, France c Inria, Paris, France d Laboratoire Jacques-Louis Lions, UPMC Univ. of Paris 6, Sorbonne University, Paris, France article info abstract

Article history: Background: Accurate measuring of the hepatic hemodynamic parameters in humans is Received 12 June 2016 inconvenient. Swine has been a favorite surgical model for the study of liver conditions due Received in revised form to many similarities with human livers. However, pigs cannot tolerate pedicle clamping 22 August 2016 and to reduce bleeding during resection a simplified technique is required. The aim of this Accepted 12 September 2016 study is to present a simplified technique for different percentages of hepatic resection in a Available online 30 September 2016 porcine model. Methods: Twenty-two consecutive large white pigs were operated with 75% and 90% liver Keywords: resection. Computarized tomography liver volumetry is performed before and after sur- Liver resection gery. In both types of surgery, hemodynamic monitoring was performed using a special- Hemodynamic monitoring ized apparatus. Technique Results: Resections were performed in both groups successfully. The residual volume in the Major planned 75% was 235 77 mL and 118 119 mL in the planned 90% resection. For 75% Pigs resection, the portal flow was reduced after resection by 8.13 28%, which might be part of systemic circulatory depression. However, the portal pressure increased by 20.1 51%. The hepatic artery flow decreased by 63.86 26.3% as well as the pressure by 5 28%. The central venous pressure at the start of surgery was 3.34 1.9 mm Hg and 2.8 2.2 mm Hg at the end of surgery. The portacaval pressure gradient was 4.4 2.9 mm Hg at the beginning of surgery and was 5.9 2.8 mm Hg at the end of surgery. For 90% resection, the portal flow decreased by 33.6 12.6% and the pressure increased by 104 58%. The hepatic artery flow decreased by 88 7%, and the pressure decreased by 5 14.8%. The central venous pressure was 3.5 1.7 mm Hg before resection and 3 2.5 mm Hg after resection. The portacaval pressure gradient was 3.8 1.1 mm Hg before resection and 8 3.7 mm Hg after resection. The mean anesthesia time was 6.6 1.05 h and 6.9 0.5 h for 75% and 90% resection, respectively. The mean operative time was 4.6 0.9 h and 5 0.7 h for 75% and 90% resections, respectively. The mean time for hepatectomy was 1.23 0.76 h and 2.4 0.1 h for 75% and 90% resection, respectively. The mean time consumed in the measurements was 2.28 1.4 h and 1.1 0.3 h for 75% and 90% resections, respectively.

* Corresponding author. 12 Avenue Paul VaillantCouturier, 94804 Villejuif Cedex, France. Tel.: þ33 1 45 59 30 00; fax: þ33 1 45 59 38 57. E-mail address: [email protected] (E. Vibert). 0022-4804/$ e see front matter ª 2016 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jss.2016.09.018 bucur et al technique for major porcine liver resection 123

The mean volume of aspirated fluid and blood in the 75% resection was 1062 512 mL, while it was 1050 354 mL in 90% resections. Conclusions: The hereby described technique is simple and easily applicable for major liver resection in a porcine model. Portal flow decreases after 90% resection more than in 75% due to the relative reduction of remnant hepatic mass. There was a larger increase in portal pressure following 90% compared to 75% resection. The hepatic artery flow decreases more in 90% than in 75% resections. ª 2016 Elsevier Inc. All rights reserved.

Background regulated at 23 1C at ambient humidity. Lighting was natural through close to ceiling wide windows. The liver is one of the most heavily studied organs in modern medicine. Surgical parenchymal resection is a strong stimu- Volumetric study lating factor for regeneration.1 Among the various factors, the change in portal flow and pressure after resection was re- Abdominal computarized tomography (CT) scanebased volu- 2 ported to be influential on regeneration. metric analysis was conducted to every pig 7 d before surgery Increased portal pressure after partial hepatic resection and immediately after the hepatic resection. 3 acts as a stimulus for liver regeneration. Nonetheless, a On average, an 80 mL (2 mL/kg) of iodinated contrast limiting threshold has been identified after which further in- (Omnipaque, GE healthcare, Carrigtwohill, Ireland) was 4 crease in portal pressure would lead to liver failure. Liver injected through an intravenous catheter with a rate of 3- 5 failure after extensive hepatectomy is a lethal complication. 4 mL/s. CT scans were performed with a Somatom (Definition Pressure and flow measurement of the splanchnic area in AS, Siemens, Forchheim, Germany). Volume analysis was clinical settings is not feasible. For that, animal models were performed using the Myrian XP-Liver 1.14.1 software (Intra- proposed as an alternative way to elaborate precise relations sense, Montpellier, France). The resection is designed between the hemodynamic parameters and liver regenera- following anatomic landmarks to keep the residual segment/s 6 tion. The development of an animal model to study the with intact outflow and inflow (Fig. 1A and B). relationship between the liver regeneration and changes in hepatic hemodynamics after major liver resection is neces- Preoperative preparation sary for better understanding and predicting the pathophysi- 7 ological mechanisms related to this issue. We hereby Animals were left fasting the night before surgery. On the day describe our technique for 75% and 90% hepatectomy with of surgery, animals were given in their individualized cages as mesenteric and hepatic hemodynamic monitoring and to a preanesthetic preparation of 30 mg/kg ketamine (Ketamine, discuss some critical anatomic features and technique based Panpharma) and 0.03 mg/kg acepromazine (Calmivet, Veto- on our experience. quinol, France).

Anesthesia and postoperative care Methods Each pig received 100 mg of xylazine 2% (Rompun, Bayer Ethical approval Healthcare) with 750 mg ketamine (Ketamine, Panpharma) for anesthesia induction followed by tracheal intubation (6-7 mm The study was approved by the regional committee of ethics in size, Portex, France). Subsequently, inhalational anesthesia in animal research and by the ministry of higher education is started using a 60% FiO2 inhalational oxygen mixed with 2% and scientific research and ministry of agriculture and fishing, isoflurane (Isoflurane, Belamont, France) at a rate of 2-3 mL according to European Union directives. mixed with 1.5-2 L/min oxygen in 1.5 L of air. Afterward, 1 g of amoxicillin (Amoxicillin, Mylan) was given i.v. Fentanyl (1 mL) was given subcutaneously after intubation and at the end of Study setting surgery, and a fentanyl patch was placed on the shaved skin of the right side of the thoracic cage after surgery. Twenty mil- Surgeries were performed at the CIRE platform, INRA Centre ligrams of atracurium (Tracrium, Glaxo Smith Klyne) muscle Val de Loire, Nouzilly, France. relaxant was given intramuscular. Crystalloids were infused at a rate of 2 mL/kg/h fasting in addition to 500-1000 mL. This Animals was increased according to demands. During surgery animals were covered with heat blankets, Twenty-two large white female pigs were included in this and gastric aspiration through an orogastric tube was per- study. The average age was 3 mo 10 d. The mean weight was formed if gastric distension was observed at surgery. Surgery 35.9 7.5 kg. The pigs were kept under strict protocol. There was conducted under sterile conditions. Animals were posi- was a period of conditioning before surgery varying from 4 to tioned in a dorsal decubitus, and surgical site disinfection and 6 d. The pigs were housed in individual pens with temperature draping were performed. Oxygen saturation, temperature, 124 journalofsurgicalresearch march 2017 (209) 122e130

Fig. 1 e Three-dimensional reconstructed images for planned (A) 75% and (B) 90% resection. The images demonstrate the planned resected lobes (S) and the residual lobes (R) with preserved portal inflow and hepatic venous outflow. PV [ portal vein. (Color version of figure is available online.)

pulse rate, arterial pressure, and central venous pressure were around the vein to prevent its shearing. One more difference is monitored during surgery. that the caudal application of the forceps was not performed Amoxicillin 1 g (Amoxicillin, Mylan) b.i.d was administered in order to keep the vein distended. intravenously for 5 d. In addition, animals received daily Hepatic artery cannulation was performed through the pantoprazole 40 mg i.v (Inipomp, Nycomed) and enoxaparin gastroduodenal artery, which was dissected above the head of 0.2 mL s.c (Lovenox, Sanofi Aventis). Animals were given free pancreas to its confluence with the hepatic artery. Two access to water and food, and if intake is not satisfactory, fluid absorbable 4/0 threads were passed under the vessel. The replacement was tailored. distal one (close to the head of pancreas) was ligated to direct all the flow that passes across the hepatic artery to the liver. Surgery The vessel was snipped with fine scissors, and then, a 3Fr SPR- 330 Millar pressure probe was introduced through this open- Longitudinal median neck incision was performed, and can- ing toward hepatic artery. Then, the proximal ligature was nulation of the right carotid artery and the right internal ju- tied over the probe. gular vein was performed with a 5 and 8 Fr vascular Desivalve The portal vein pressure probe (5 Fr, SPR-350 Millar) was (Vygon) cannula. A water column-based barometer was con- inserted through the gastroduodenal vein in a similar way to nected to the side line of the arterial cannula in order to the hepatic artery. Both probes were further secured in place continuously measure the carotid pressure through the with the proximal absorbable ties. anesthesia monitor. Millar pressure probes (Millar, TX) were then inserted into the internal jugular vein (5 Fr, SPR 350) and Probes positioning carotid artery (4 Fr, SPR 340). A 5 Fr and 4 Fr pressure probes were inserted into the right Subsequently, a midline abdominal incision was per- jugular vein (tip positioned in superior vena cava, verified by formed. Table-mounted costal retractors and self-retaining waveform) and carotid artery, respectively (Fig. 2A), while a 3 abdominal retractors were installed. Fr pressure probe was inserted into the hepatic artery through the ligated gastroduodenal artery and a 5 Fr pressure probe Vascular cannulation was inserted into the portal vein through a small medial Carotid artery cannulation was performed after dissecting a 4- gastroduodenal tributary (Fig. 2B). For flow measurements, we 5 cm segment. The artery was surrounded by a vascular tape used TranSonic transit time technology (TranSonic System through which traction was applied in a cranial direction. The Inc, Ithaca). A 12 or 14 Fr flow probe (MA12 PAU or MA14 PAU) caudal end of the dissected portion was gently grasped with a surrounded the portal vein and a 4-6 Fr flow meter probe vascular forceps till the puncture was performed, and the (MA4PSS or MA 6PSS) was used around the hepatic artery, guide wire was introduced. The artery was grasped again over proximal to the gastroduodenal artery. A 10-14 Fr flow meter the guide wire to reduce bleeding from the puncture, and probe (MA 10/12/14 PAU) was positioned around the supra- then, the vascular cannula was introduced. The jugular vein celiac aorta, and a 8 or 10 Fr flow meter probe (MA 8PSS or MA was cannulated in a similar way except that the dissection 10PAU) surrounded the aorta between the mesenteric and the around the vein was performed keeping a fine tissue pad renal arteries. Pressure probes were inserted to reside in the bucur et al technique for major porcine liver resection 125

Fig. 2 e (A) Neck incision with cannulated jugular vein (J) and carotid artery (Ca), (B) probes installed for ﬂow measurements around (P) portal vein, (H) hepatic artery and for pressure measurement inside (G) gastroduodenal artery, (V) gastroduodenal vein. (C) Monitoring setup. L [ liver. (Color version of ﬁgure is available online.)

main stream of the designated vessel and were always of the which was then placed to squeeze the parenchyma, inflow, same caliber; the flow meters size could be changed depend- and outflow of the lobes to be resected in mass (Fig. 4A and B). ing on the vessel size depicted from the preoperative CT scan. Care must be taken to protect the right hepatic vein that The perivascular flow meters are connected to Transonic passes from right lateral hepatic lobe to the IVC close to the T403 (Transonic Systems, Inc, Ithaca, NY), and the Millar fissure between the right medial and right lateral lobes. Each Mikro-Tip Pressure Transducer Catheters are connected to individual lobe pedicle is then ligated separately with a single pressure control unit PCU-2000 with two-channel connected absorbable 0 tie. The tie was passed around the pedicle with to four Millar catheters (Millar Inc, Houston, TX). the aid of a right-angled clamp that crosses posterior to the Measurements are continuously recorded during surgery whole pedicle (Fig. 5A and B). Subsequently, a common left with a 16-channel amplifier connected to a computer running hepatic trunk draining the left lateral and left medial lobes IOX2 acquisition software (Emka Technologies, Paris, France) was ligated close to its confluence with the IVC. The hepatic (Fig. 2C). vein draining the right medial lobe is then ligated, and the remainder of the parenchyma is fractionated with a few Kelly Liver resection crushings. Resection of 75% of liver volume. Based on the CT volumetric Gradual declamping takes place to complete the hemo- study, a 75% hepatectomy was planned for through resection stasis. Hemostasis is completed using polypropylene 5/0 or of the left lateral, left medial, and the right medial hepatic bipolar electrocautery (Fig. 5C). lobes leaving in place the right lateral lobe and the caudate lobe surrounding the inferior vena cava (IVC). Technique for 90% hepatectomy The liver was freed from its diaphragmatic attachments A more extended resection was performed in 6 out of the 22 and from the lesser omentum (Fig. 3A). A capsular incision animals. The surgery was performed in the same manner was performed all around the resected lobes at a 1 cm distance until completion of the 75% resection; thereafter, a small from the insertion of these lobes to the IVC on the dorsal parenchymal dissection was performed on the lateral border aspect of the liver (Fig. 3B), while this incision is performed of the right lateral lobe using a bipolar electrocautery device. around the portal pedicles on the ventral aspect. A Kelly clamp Then, using a monopolar device, the capsule of this lobe was was used for parenchymal fractionation of a 2-cm segment at incised all around its circumference close to the origin of its the right border of the right medial lobe (Fig. 3C) to facilitate pedicle in order to facilitate the use of a Rummel tourniquet, the placement of a clamping Rummel tourniquet (Fig. 3D), which is then tightened (Fig. 6A and B). Subsequently, the liver 126 journalofsurgicalresearch march 2017 (209) 122e130

Fig. 3 e Technical aspects during liver resection (A) dissection of the medial attachment of the left lobe to the lesser omentum, (B) capsular incision marking the line of resection, (C) crushing of the parenchyma at the lateral border of the right medial lobe with Kelly clamp to facilitate the positioning of (D) the tourniquet. (Color version of figure is available online.) parenchyma was dissected using a combination of Kelly residual volume was calculated from a CT scan performed clamp crushings and ties. Two large suprahepatic veins are immediately after the resection and it was 235 77 mL, while generally encountered and tied with absorbable 0 sutures. the planned residual volume was 365.6 113.5 mL. A Similarly, a single large portal pedicle was tied before its discrepancy of 13.5 11.9% between the planned and real further subdivisions. Once resection was completed, gradual resected volume was found. Likewise, a 31 23% difference declamping and hemostasis were performed in the same was found between the planned residual and the actual re- manner as in the 75% resection. sidual volume. After resection was completed, adapted pieces of gloves are left around the vessels as well as between the liver and the Ninety percent resection stomach to prevent adhesion formation in order to facilitate The animals’ weight was 35 4 kg; their whole liver volume was dissection at the time of sacrifice surgery. 994 288 mL. We planned to resect 876 170 mL (89 8%) of the whole liver volume, leaving a residual volume 118 119 mL. The actual resected volume was 952 267 mL, leaving a residual Results volume of 63 21 mL. The percentage of resection based on the both preoperative and postoperative volumetry was 94 0.4% Resection volume estimates with an error in estimation of 7.3 7.4%. The liver to body weight ratio was 2.8 0.5 before surgery Seventy-five percent resection and 0.11 0.04 after surgery. The mean animals’ weight was 35.6 4.5 kg. Their mean whole liver volume based on CT estimation was 926 184 mL. Operative time The planned resection volume for the 75% hepatectomy was 558 84.5 mL with an estimated planned residual For 70% resection, the mean anesthesia time was 6.6 1.05 365.6 113.5 mL. The planned resection volume based on CT (95% confidence interval [CI] ¼ 6-7.2) h. The mean operative volumetry was 60.9 5.2% of the whole liver. time was 4.6 0.9 (95% CI ¼ 4-5) h. The mean time for hepa- The anatomic resection resulted in a real resected liver tectomy was 1.23 0.76 (95% CI ¼ 0.78-1.5) h. The mean time volume of 711 175 mL (range, 506-982 mL). The percentage of consumed in measurements was 2.28 1.4 (95% CI ¼ 1.5-3) h. actual resection was 74.8 7.5%, based on the whole liver For 90% resection, the mean anesthesia time was volume obtained from CT before and after resection. The 6.9 0.5 h. The mean operative time was 5 0.7 h. The mean bucur et al technique for major porcine liver resection 127

The aortic flow measured above the celiac trunk was 2.3 0.78 L/m, which insignificantly decreased to 2.1 0.59 L/ m(t ¼ 0.93, P ¼ 0.36) at the end of surgery. The aortic flow above the origin of the renal arteries was 0.83 0.39 L/m at the start of surgery and was 0.74 0.4 L/m at the end of surgery (t ¼ 0.54, P ¼ 0.59) at the end of surgery. The carotid artery pressure was 52.5 7.6 mm Hg at the beginning of surgery and was 46.8 13.6 at the end of surgery (t ¼ 1.3, P ¼ 0.2). The central venous pressure at the start of surgery was 3.34 1.9 mm Hg and 2.8 2.2 mm Hg at the end of surgery (t ¼ 0.6, P ¼ 0.54). The portacaval pressure gradient was 4.4 2.9 mm Hg at the beginning of surgery and was 5.9 2.8 mm Hg at the end of surgery (t ¼1.3, P ¼ 0.2).

Ninety percent resection The mean portal flow at the start of surgery was 0.8 0.08 L/ min and 0.48 0.025 L/min the end of surgery (P ¼ 0.09). The percent in reduction was 33.6 12.6%. The portal pressure at the start of surgery was 6.4 2.3 mm Hg and 11.1 1.2 mm Hg at the end of surgery with a 104 58% increase after resection (P ¼ 0.6). The hepatic artery flow decreased by 88 7% from 0.16 0.06 L/min to 0.04 0.02 L/min (P ¼ 0.007), and the hepatic artery pressure decreased from 34.8 10.1 mm Hg to 22 14 mm Hg (P ¼ 0.8). The aortic flow measured above the celiac trunk was 2 1.2 L/min before resection and 1.2 0.76 L/min after Fig. 4 e (A) Tightening of the clamping tourniquet around resection. The carotid artery pressure was 43 11 mm Hg the 75% mass represented in the left lateral, left medial, before resection and 37 mm Hg after resection. and right medial lobes. (B) The efficacy of clamping is The central venous pressure was 3.5 1.7 mm Hg before ¼ demonstrated by the color difference between the clamped resection and 3 2.5 mm Hg after resection (P 0.06). The and unclamped lobes. (Color version of figure is available portacaval pressure gradient was 3.8 1.1 mm Hg before ¼ online.) resection and 8 3.7 mm Hg after resection (P 0.02).

Discussion hepatectomy time was 2.4 0.1 h. The mean time consumed for measurement was 1.1 0.3 h. The hereby described technique to perform 75% and 90% liver resection in a swine model as well as our technique for inva- Blood loss sive continuous monitoring of hemodynamics during liver resection is simple. There are certain precautions that should The blood loss could not be accurately estimated due to its mix be kept in mind while proceeding with this particular tech- with aspirated fluid, which is mainly lymphatic. The mean nique in a swine model. volume of aspirated fluid and blood in the 75% resection was 1062 512 mL, while it was 1050 354 mL in 90% resections. Anatomic consideration Hemodynamic assessment The porcine liver is divided into the left lateral and medial Seventy-five percent resection lobes, the right medial and lateral lobes, and the caudate At the start of surgery, the mean portal flow was lobe.8 The left lateral lobe represents around 25% of the total 0.775 0.198 L/m, while at the end of surgery, there was a liver volume9 and is consistently the largest of all lobes.10 reduction in the portal flow to 0.689 0.19 L/m. The percent in reduction was 8.13 28% (t ¼ 1.1, P ¼ 0.27). The portal pressure Hepatic hilum at the start of surgery was 7.8 2 mm Hg and 8.79 2mmHg The hepatic artery runs off the celiac artery on the posterior at the end of surgery. The increase in the portal pressure was aspect to the posteromedial aspect of the portal vein, where it equivalent to 20.1 51% (t ¼1.2, P ¼ 0.24). gives off the gastroduodenal artery just above or behind the The hepatic artery flow decreased by 63.86 26.3% from head of pancreas. Nearly at the same level in a more superfi- 0.18 0.071 L/m to 0.052 0.024 L/m (t ¼ 6.2, P < 0.001), and the cial plane, the gastroduodenal vein joins the portal vein. The hepatic artery pressure decreased by 5 28% from bile duct is situated on the anteromedial side of the portal 42.9 8.5 mm Hg to 39.5 11.3 mm Hg (t ¼ 0.87, P ¼ 0.39). vein. 128 journalofsurgicalresearch march 2017 (209) 122e130

Fig. 5 e The passage of right-angled clamp behind the ﬁrst (A) and the second (B) portal pedicles supplying the resected lobes and (C) an after 75% resection view. (Color version of ﬁgure is available online.)

The portal vein and the hepatic artery are dissected care- landmark to its insertion is where the dorsal sling converges fully throughout their extrahepatic course, and the lymph with the ventral sling of the right crus. Trying to access the nodes on the posterolateral and medial sides of the portal vein plane around the aorta below the insertion of the pleura might are removed to facilitate the positioning of the probes. expose the suprarenal gland to manipulations to which the pig is highly sensitive. Manipulation around the suprarenal Aorta area usually caused the pigs to manifest hemodynamic While dissecting the supra celiac part of the aorta, the plane is instability. Having that observed during the ﬁrst few pigs, we accessed through an opening in between the esophageal and therefore preferred to access the plane a little bit higher, aortic crura. The pleura insert low into the coverings of the taking the risk of opening the pleura. Opening of the pleura aorta on the anterior and the left aspects of the vessel, around could be, nevertheless, avoided if the plane was accessed be- 1 to 2 cm above the origin of the celiac trunk, and the tween the two crural slings and below the white ﬁbers of the

Fig. 6 e For 90% to be completed, (A) passage and (B) tightening of the clamping tourniquet around the right lateral lobe leaving just below it the caudate lobe. (Color version of figure is available online.) bucur et al technique for major porcine liver resection 129 diaphragm. At the end of the intervention, the pleura, if 90%, which suggests that the contribution of the four lobes to opened, is drained using a suction drain that is activated after the whole liver volume could be slightly different from animal the abdominal closure and left in place for around 30 min (i.e., to animal. Nevertheless, the relatively narrow range of stan- time for pig to recover from anesthesia). dard deviation suggests its usefulness as a preoperative On the right side of the aorta, a large lymphatic vessel will planning tool. be consistently found. This lymphatic channel is sometimes For partial hepatectomy pig model, resection usually starts found creeping on the anterior aspect of the aorta and be- from the left lateral lobe then further resections are per- comes difficult to avoid injuring. formed in an anticlock wise manner. The caudate lobe has a At the origin of the renal arteries portion of the aorta, the peculiar position, together with the uncountable small veins large lymphatic vessels surround the aorta from many di- that drain the segment directly in the IVC which resides inside rections and were breached nearly constantly while posi- the parenchyma15; making the resection of this segment is tioning the suprarenal flow meter. The injury to the lymphatic very difficult to pursue. channels at that level is responsible for loss of significant We used to measure the hepatic hemodynamics in amounts of fluids which increases the risk of mortality after humans using a small needle connected to a water column surgery. barometer. This technique although widely adapted yet it conveys some difficulties given the sensitivity of the measurement set to calibration and positioning. We find that the hereby described measurement setting is much more repro- Resection surgery ducible and robust. Despite that it is technically invasive, the electromagnet flow meters require the vessel of interest to be One of the advantages offered by the illustrated clamping librated all around and the pressure probe to be introduced technique is that it avoids pedicle dissection. In that way, the through a small vessel that could be sacrificed or introduced time required for resection is theoretically shorter. Moreover, directly into a vessel that will be repaired after extraction of the intraparenchymal individual pedicle ligation avoids injury the barometer probe. to the right lateral bile duct that commonly inserts into the left main duct.10 We planned the resection based on the CT volumetry. Conclusions Nonetheless, resections were performed based on the anatomic segmentation leading to a discrepancy of The hereby described technique is simple and reproducible. It 13.5 11.9% between the planned and real resected volume. avoids the pedicle clamping, which is intolerable in large The error in estimating the resected liver volume before sur- white pigs. The anatomic considerations described are gery could be related to the difficulty in appreciation of the important to consider to perform a successful surgery. Portal anatomic lobes in pigs despite the prominent fissures between flow decreases after 90% resection more than in 75% due to lobes that might be difficult to visualize in the CT scan. the reduction of remnant hepatic mass, which lead to in- However, the anatomic resection resulted in a resection that crease in tissue resistance. There was a larger increase in corresponds to the planned resection in the 90% group, since portal pressure following 90% compared to 75% resection. the delineation of the caudate lobe was precise in the preop- The hepatic artery flow decreases more in 90% than in 75% erative CT. resections. An estimation error is systematically reported different between the CT measured volume and the actual liver volume. This difference is attributed to the intrahepatic blood volume that is included in the CT measurement but not during Acknowledgment the operative measurement,11 which could be adjusted for using a factor of 0.85 in the calculation of volume based on The authors would like to acknowledge the contribution of CT.12 Hans Adriansen, Fracois Le Compte at the INRA, Tours, The later hypothesis is supported by the presence of an France, in data acquisition. error in residual at day 0, measured in CT scan, which was Funding: this study was funded mainly by the “Agence de larger than that for the resected volume, considering that the la Biomedecine” through its program of Research (AOR 2009). initially planned resected volume was around 60%. Despite Eric Vibert, Petru O. Bucur, and Mohamed Bekheit acknowl- that error, the correlation between the resected and the esti- edge funding by project ANR-13-TECS-0006 (IFlow). mated volumes was excellent (r ¼ 0.8, P ¼ 0.003). Attributing Authors’ contribution: P.B. and M.B. equally contributed in the intrahepatic blood volume to the difference between the this manuscript and are first author. Study concept and CT measured volume and the ex-vivo volume was reported in design were preformed by P.B. and M.B. Surgical technique one study.13 development was done by P.B. M.B., P.B., and C.A. contributed The variation range in the real resection volume percent to the acquisition of data. M.B., P.B., C.A., E.V., and I.V.-C. indicates that anatomic resections might not always result in analyzed and interpreted the data. M.B., P.B., and C.A. draf- the desired percent of resection, similar to what was found in ted the manuscript. E.V. and I.V.-C. contributed to the critical humans.14 This is particularly evident in the estimation of 75% revision of the manuscript. Final approval was done by E.V. resection as opposed to the lower error in the estimation of and M.B. 130 journalofsurgicalresearch march 2017 (209) 122e130

7. Arkadopoulos N, Defterevos G, Nastos C, et al. Development Disclosure of a porcine model of post-hepatectomy liver failure. J Surg Res. 2011;170:e233ee242. 8. Gravante G, Ong SL, Metcalfe MS, Lloyd DM, Dennison AR. The authors reported no proprietary or commercial interest in The porcine hepatic arterial supply, its variations and their any product mentioned or concept discussed in the article. influence on the extracorporeal perfusion of the liver. J Surg Res. 2011;168:56e61. references 9. Huisman F, van Lienden KP, Damude S, Hoekstra LT, van Gulik TM. A review of animal models for portal vein embolization. J Surg Res. 2014;191:179e188. 10. Court F, Wemyss-Holden S, Morrison C, et al. Segmental 1. Chen MF, Hwang TL, Hung CF. Human liver regeneration after nature of the porcine liver and its potential as a model for major hepatectomy. A study of liver volume by computed experimental partial hepatectomy. Br J Surg. tomography. Ann Surg. 1991;213:227e229. 2003;90:440e444. 2. Niiya T, Murakami M, Aoki T, Murai N, Shimizu Y, Kusano M. 11. Niehues S, Unger J, Malinowski M, Neymeyer J, Hamm B, Immediate increase of portal pressure, reflecting sinusoidal Stockmann M. Liver volume measurement: reason of the shear stress, induced liver regeneration after partial difference between in vivo CT-volumetry and intraoperative hepatectomy. J Hepatobiliary Pancreat Surg. 1999;6:275e280. ex vivo determination and how to cope it. Eur J Med Res. 3. Park M, Lee Y, Rha S, Oh S, Byun J, Kim D. Correlation of portal 2010;15:345e350. venous velocity and portal venous flow with short-term graft 12. Karlo C, Reiner C, Stolzmann P, et al. CT-and MRI-based regeneration in recipients of living donor liver transplants. volumetry of resected liver specimen: comparison to Transplant Proc. 2008;40:1488e1491. intraoperative volume and weight measurements and 4. Allard MA, Adam R, Bucur PO, et al. Posthepatectomy portal calculation of conversion factors. Eur J Radiol. vein pressure predicts liver failure and mortality after major 2010;75:e107ee111. liver resection on noncirrhotic liver. Ann Surg. 13. Mu¨ ller SA, Pianka F, Scho¨ binger M, et al. Computer-based 2013;258:822e830. liver volumetry in the liver perfusion simulator. J Surg Res. 5. Lin XJ, Yang J, Chen XB, Zhang M, Xu MQ. The critical value of 2011;171:87e93. remnant liver volume-to-body weight ratio to estimate 14. Abdalla EK, Denys A, Chevalier P, Nemr RA, Vauthey JN. Total posthepatectomy liver failure in cirrhotic patients. J Surg Res. and segmental liver volume variations: implications for liver 2014;188:489e495. surgery. Surgery. 2004;135:404e410. 6. Pouyet M, Mećhet I, Paquet C, Scoazec JY. Liver regeneration 15. Court FG, Laws PE, Morrison CP, et al. Subtotal hepatectomy: a and hemodynamics in pigs with mesocaval shunt. J Surg Res. porcine model for the study of liver regeneration. J Surg Res. 2007;138:128e134. 2004;116:181e186. The Journal of Physiology

https://jp.msubmit.net

JP-RP-2017-275012

Title: Impact of 75% partial hepatectomy on hemodynamics in porcine model

Authors: Chloe Audebert Petru Bucur Mohamed Bekheit Eric Vibert Irene Vignon-Clementel

Author Conflict: No competing interests declared

Author Contribution: Chloe Audebert: Acquisition or analysis or interpretation of data for the work; Drafting the work or revising it critically for important intellectual content; Final approval of the version to be published; Agreement to be accountable for all aspects of the work Petru Bucur: Acquisition or analysis or interpretation of data for the work; Drafting the work or revising it critically for important intellectual content; Final approval of the version to be published; Agreement to be accountable for all aspects of the work Mohamed Bekheit: Acquisition or analysis or interpretation of data for the work; Drafting the work or revising it critically for important intellectual content; Final approval of the version to be published; Agreement to be accountable for all aspects of the work Eric Vibert: Conception or design of the work; Drafting the work or revising it critically for important intellectual content; Final approval of the version to be published; Agreement to be accountable for all aspects of the work Irene Vignon-Clementel: Conception or design of the work; Acquisition or analysis or interpretation of data for the work; Drafting the work or revising it critically for important intellectual content; Final approval of the version to be published; Agreement to be accountable for all Disclaimer: This is a confidential document. aspects of the work

Running Title: Impact of 75% partial hepatectomy on hemodynamics in porcine model

Dual Publication: No

Funding: Agence Nationale de la Recherche (L' Agence Nationale de la Recherche): Chloe Audebert, Petru Bucur, Mohamed Bekheit, Eric Vibert, Irene E Vignon-Clementel, ANR-13-TECS-0006 iFLOW

Disclaimer: This is a confidential document. 1 Impact of 75% partial hepatectomy on hemodynamics in porcine model 2 Chloe Audebert1,2#*, Petru Bucur3#, Mohamed Bekheit4,5, Eric Vibert4 and Irene E. Vignon- 3 Clementel1,2 4 5 1 Inria Centre de recherche de Paris, Paris, France. 6 2 Sorbonne universités, laboratoire Jacques-Louis Lions, UPMC Paris VI, Paris, France. 7 3 Service de Chirurgie Digestive, CHU Trousseau, Tours, France. 8 4 Unité INSERM 1193, Centre Hépato-Biliaire, Hopital Paul Brousse, Villejuif, France. 9 5 Institute of Biomedical Sciences, University Of Aberdeen, Aberdeen, UK. 10 #The authors contributed equally to this article. 11 12 *Corresponding author: 13 Chloe Audebert 14 Inria centre de recherche de Paris 15 2 rue Simone Iff, 75012 Paris, France 16 [email protected] 17

18 Key points summary

19 • Impact of 75% partial hepatectomy in porcine model on liver and systemic hemodynamics is 20 analyzed. 21 • Decision trees to explain survival are built based on hemodynamics evolution during 75% 22 liver resection. 23 • Continuous recordings to investigate mechanical ventilation and cardiac cycle variation of 24 arterial and venous hepatic and systemic pressures and flows are presented. 25 • The mechanical ventilation and the blood volume changes occurring during surgery largely 26 impact venous pressures. Therefore, measurements of pressure difference (in particular of the 27 ‘hepatic venous pressure gradient’) are sensitive to experimental conditions.

28 Abstract (250w)

29 Post-operative liver failure is partly due to altered post-resection hemodynamics, especially portal vein 30 pressure and flow. The goal of this article is to investigate the impact of 75% partial hepatectomy on 31 the liver and systemic hemodynamics. A better knowledge on hemodynamics alteration after liver 32 resection might enable a control of hemodynamics changes to improve surgery outcomes. 33 Hemodynamics measurements are presented that were continuously recorded during 22 liver partial 34 ablations in a porcine model. 1

35 The evolution of time-averaged measurements revealed a small decrease of portal vein flow, probably 36 related to blood volume changes, and on average a small increase of portal pressure. To explain 37 survival, decision trees based on evolution of hemodynamics were built. The outcome suggested that 38 the surgery risk was reduced when the portal vein increase and arterial pressure reduction were limited 39 (respectively a less than 20% increase and a less than 30% reduction). In addition, another indicator 40 was proposed: a ratio of venous pressure drop and liver flows. This predictor includes the variation of 41 pressure with respect to flow occurring during liver surgery. The decision tree suggested that when 42 this ratio decrease and increase were limited (between 0.7 and 4.6 times baseline) the surgical risk was 43 reduced.

44 The continuous recording enabled us to study the impact of mechanical ventilation and cardiac-cycle 45 variations, and make measurements recommendations. Moreover, the dynamical changes occurring at 46 the clamping instant were analyzed. A typical change of cardiac-cycle waveform in the hepatic artery 47 flow was observed, presumably linked to the hepatic artery tree architecture.

49 Introduction 50 Post-operative liver failure is a condition encountered in liver transplantation and in major partial 51 hepatectomy (liver resection). This condition is partly due to a liver volume below a certain threshold. 52 It also depends on the liver hemodynamics, especially on the portal vein pressure and flow [1]–[5]. A 53 modulation of portal vein hemodynamics is assumed to protect the liver and improve the regeneration 54 process [6], [7]. To correctly modulate the portal vein hemodynamics during liver surgery, the impact 55 of this surgery must be known. Our goal was to investigate the impact of liver surgery on the liver and 56 systemic hemodynamics. Therefore, various pressures and flows were recorded – in a 75% liver 57 resection swine model – which is believed to resemble the human anatomy and physiology [8], [9]. 58 Moreover, its liver to body weight ratio is close to human’s [10]. 59 Several groups studied liver resection and the associated hemodynamics on a porcine model. Xiang et 60 al. [5] presented various resection volumes in pigs. This work has concluded that the liver fails to 61 regenerate if the post-resection portal venous flow per liver mass is increased more than a threshold 62 value. 63 Liver regeneration can be disrupted by changes in hemodynamics. Indeed, intraoperative ischemia, 64 during 70% liver resection on pigs, has been shown to damage the sinusoidal endothelium and to 65 cause post-operative liver failure [11]. In [12], the pre-treatment with Oprinone has shown 66 hepatoprotective effects, by modifying the portal vein pressure, for 70% liver resection in swine. The 67 liver hemodynamics after liver resection may be controlled by a shunt, linking the mesenteric vein to 68 the vena cava (using a vena cava graft from a pig donor). This technique has shown that liver 69 regeneration is preserved, despite a large reduction of liver portal inflow [13]. 2

70 This article presents the various pressures and flows recorded during liver surgeries - 75% liver 71 resection performed on pigs [14]. To our knowledge, no study has reported continuous recording of 72 pressures and flows during liver resection in porcine or other species. The purpose of measurement 73 analyses was to better understand the liver hemodynamics during liver resection, and which factors 74 may impact them to determine how to best take representative measurements when they cannot be 75 continuously recorded. 76 The paper is structured as follows. The surgical procedure and the measurements protocol are briefly 77 described. Then, the global changes due to liver surgery are studied with the evolution of time- 78 averaged pressures and flows. Decision trees based on evolution of hemodynamics are built to explain 79 survival. The continuous recordings provide additional information. First the impact of mechanical 80 ventilation and the variation during cardiac cycles are described. Then the changes occurring at the 81 clamping instant are also analyzed. The article ends with a discussion and some clinically relevant 82 conclusions. 83 84 Methods 85 Ethical approval 86 This study was approved by the regional committee of ethics in animal research, by the ministry of 87 higher education and scientific research and by the ministry of agriculture and fishing, according to 88 European Union directives. 89 Liver resection surgery 90 Protocol and studied animals 91 75% liver resection was performed on twenty-two pigs – as described in [6], [14] – with sacrifice 92 seven or three days after liver resection. CT-scans were performed before and immediately after the 93 surgery, with a Siemens Somatom AS definition 128 machine. Image acquisitions were performed 94 before, 15, 35, 55 and 75 seconds after injection of 75 ml of iohexol 350mg/ml (Omnipaque, GE 95 Healthcare) with a rate of 5 ml per second. The pre- and post-resection liver volumes were estimated 96 from the CT-scans (Myrian® XP-Liver 1.14.1 software (Intrasense, Montpellier, France)) 97 (Supplementary material Table 1). 98 During liver resection surgery, pressures and flows were measured (as described in paragraph 99 Hemodynamics measurements). Before and after liver surgery, the animals were kept in individual 100 boxes with daily care. Following the surgery, nine animals died before the third post-operative day. 101 Liver resection surgery 102 The surgical technique is detailed in [14]. The first stage of the surgery was a neck incision performed 103 to reach the right internal jugular vein and the carotid artery, in order to measure central venous and 104 systemic arterial pressures respectively. The pressure catheters were introduced in the jugular vein and 105 in the carotid artery using respectively 8 Fr and 5 Fr vascular Desivalve (Vygon) cannula. Then a 106 laparotomy was performed. First, the aorta above the celiac trunk (AoC) (Figure 1) was dissected to

107 place a flow probe. Second, the dissection of the aorta between the mesenteric artery and the kidney 108 arteries (AoI)(Figure 1) was performed to position another flow probe. 109 When the two aortic flow probes were in place, the gastroduodenal artery was ligated (Figure 1). The 110 hepatic artery pressure catheter was introduced through the gastroduodenal artery in order to set the 111 pressure catheter in the middle of the hepatic artery and obtain a precise pressure measurement. The 112 hepatic artery and the portal vein were then dissected to place the flow probes. Finally, the portal vein 113 pressure catheter was inserted into the vein. Figure 2 shows the hepatic artery and the portal vein with 114 the pressure catheters and the flow probes set. To perform the liver resection, the removed liver lobe 115 vessels were circled with a tape (a thin fabric strip) [14]. Next, the tape was pulled (clamping instant) 116 to block the inflow of the corresponding liver tissues. Finally, the liver lobes were resected and 117 weighted. Blood loss during the resection was reduced with this procedure. Finally, all the flow probes 118 and pressure catheters were removed and the animal incisions were closed. 119 Hemodynamics measurements 120 Millar Instruments pressure catheters were inserted to measure the pressures. 5Fr catheters were 121 chosen for the central venous and the portal vein pressure measurements, while 4Fr and 3.5Fr were 122 used for the carotid artery and the hepatic artery pressure measurements respectively. The four Millar 123 Instruments catheters were connected to two Millar Instruments pressure control units PCU-2000 with 124 two channels. A Transonic flowmeter (three channel perivascular flowmeter T403-PPP) was used for 125 flow measurements. The probe diameters were: 10mm or 12mm for portal vein, 4mm or 6mm for 126 hepatic artery, 10mm or 12mm for AoC and 8mm for AoI. The probe size was chosen according to the 127 blood vessel diameter for each animal. Flows and pressures were recorded continuously during the 128 surgery with Emka TECHNOLOGIES itf16USB usbAMP amplifier connected to a computer running 129 the iox2 acquisition software. In order to obtain reliable pressure and flow measurements, a calibration 130 in iox2 software was required. The calibration consists in the comparison of the values measured by 131 the device under test with a calibration standard. The calibration was performed at the beginning of 132 each surgery, before the incision time. The flow calibrations were done with two points using 133 electronic settings on the Transonic flowmeter T403-PPP. The pressure catheters were maintained in a 134 saline solution for 5 minutes before calibration. The two-point calibration procedure was done using 135 the electronic settings of the PCU-200 or with a manual inflation system controlled with a high 136 precision digital manometer (Lex 1 manometer from KELLER AG). 137 Hemodynamics measurements analysis 138 Data analysis 139 First the hemodynamics global changes occurring during liver surgery were studied. Therefore, for 140 each measured pressure or flow two key moments of the surgery were analyzed: the beginning of the 141 surgery before liver resection (right after position of all the probes) and the end of the surgery after 142 liver resection (just before incision closure, around 1 hour after clamping). During these moments, 20 143 seconds of stable signal (without surgeons intervention) were used to compute averaged pressure and

144 flow values (supplementary material Table 1). Moreover, the evolution with respect to the pre-

145 resection value was computed for each variable, with the following formula (X-X0)/X0, where X0 and 146 X are pre- and post-resection averaged measurements respectively. 147 Then, the dynamics of the pressure and flow signals was analyzed thanks to the continuous recordings. 148 The variations due to mechanical ventilation and during the cardiac cycle were estimated for 4 149 different animals. The variation at peak systole was computed to estimate the impact of mechanical 150 respiration on arterial pressures and flows during 20 seconds. For the venous signals, the impact of 151 mechanical ventilation was estimated by computing the variation around the mean during 20 seconds. 152 For both arterial and venous signals, the variation over the cardiac cycle was estimated with the 153 variations around the average during two cardiac cycles taken at the end of expiration. The variation 154 around the average was computed with the following formulas, 155 (average-minimum)/average ; (maximum-average)/average. 156 Finally, at the clamping instant, the signals waveform and the instantaneous response were also 157 described. 158 Decision tree analysis 159 Decision trees were built based on hemodynamics data, to explain animal deaths that occurred before 160 the third post-operative day. The goal of a decision tree is to “predict” the class variable (here: “alive 161 after day 3” or “died before day 3“) from the predictor variables (here: hemodynamics data). 162 The MATLAB function fittctree (MATLAB 2017a, The MathWorks, Inc., Natick, Massachusetts, 163 United States), based on the CART algorithm [15], was used to perform this analysis. In this 164 algorithm, the decision tree is built by recursively partitioning the data set, one predictor variable at a 165 time. In order to obtain a tree with a limited number of “decisions” the maximum number of split was 166 set to 4. The Gini's Diversity Index was used as splitting criterion, and the other parameters of the 167 function were set to the default ones. 168 Only the animals for which all the measurements were available were studied. The considered

169 predictors were the evolution with respect to pre-resection value of: portal vein flow (!!"), hepatic

170 artery flow (!!! ), carotid artery pressure (!! ), portal vein pressure (!!" ), pressure ‘gradient’

171 (!!" − !!") and pressure-flow ratio (!). The pressure-flow ratio was computed by the following 172 formula,

0.2(!!" − !!") ! = (!!! + !!")

173 where 0.2(!!" − !!") is an estimation of the drop of pressure across the liver tissue and hepatic veins, 174 assuming the pressure between liver tissue and hepatic vein is 20% of the total pressure drop [16]. 175 Note that the method (and thus the results) is insensitive to the precise % of pressure drop: this choice 176 was motivated by the consistency between the tissue pressure drop and the flow into it. The evolution

177 of the flows and pressures was defined by (X-X0)/X0 and for the pressure-flow ratio by X/X0, where X0 178 and X are pre- and post-resection values respectively. 5

179 The efficiency of the decision tree was assessed with the training error, meaning the misclassification 180 rate on the training set. The performance of the predictive modeling procedure was estimated with the 181 rate of misclassification in a leave-one-out procedure (MATLAB 2017a, The MathWorks, Inc., 182 Natick, Massachusetts, United States). 183 Statistical analysis 184 Due to the small sample size, a non-parametric statistical test was chosen to compare pre- and post- 185 resection averaged measurements and the significance threshold was set at a p-value of 0.05. A paired 186 two-sample Wilcoxon test in R (R Foundation for Statistical Computing, Vienna, Austria. ISBN 3- 187 900051-07-0, URL http://www.R-project.org) was employed. 188 189 Results 190 Average evolution 191 Table presents the average and relative standard deviation of the different measurements before and 192 after liver resection, the two-sample Wilcoxon paired tests p-value comparing pre- and post- resection 193 variables, as well as the median (minimum / maximum) of the evolution after liver resection with 194 respect to baseline. 195 The central venous pressure decreased during liver resection surgeries in average (p=0.01). An 196 averaged decrease of 1.2 (2.1) mmHg was observed at the end of the liver resection surgery. The 197 carotid pressure decreased during the surgery (p<0.001), a decrease of 7.9 (10.8) mmHg was measured 198 compared with pre-resection values. The two flow rates measured in the aorta were also impacted by 199 the surgery. After liver resection, the AoC flow decreased by 14% (min: -45% / max: 32%). In 200 average, the AoI flow decreased after liver resection (p=NS), the measured reduction median was of 201 16% (min: -38% / max: 53%). 202 Locally, the hepatic hemodynamics was modified. The hepatic artery pressure was reduced by 5.5 203 (12.7) mmHg after liver resection (p=NS). The hepatic artery flow rate decreased drastically 204 (p<0.001), a decrease of 74% (min: -90% / max: -6%) was measured. For the venous side of the 205 hepatic circulation, the portal vein flow decreased after liver resection (p=0.02), dropping by 9% (min: 206 -50% / max: 63%). 207 The changes of portal vein and hepatic artery flow per liver mass, during the liver resection, are 208 summarized in Figure 3. A large increase of portal vein flow per liver mass was observed (p < 0.001), 209 whereas the hepatic arterial flow per liver mass remained similar to baseline (p = NS). 210 Finally, on average a small increase of portal pressure of 0.6 (2.1) mmHg was measured at the end of 211 the surgery (p=NS). Due to liver resection, the portal pressure increased and the central venous 212 pressure decreased, leading to the increase of venous pressure difference across the liver (portocaval 213 gradient) (p=0.003). With respect to the pre-resection value, the venous pressure difference increased 214 by 1.9 (2.4) mmHg at the end of 75% liver resection (Table ). 215

216 Decision trees: explaining survival with hemodynamics

217 First a tree was built including all predictors (evolution of !!", !!!, !!, !!", !!" − !!" and !) with

218 13 animals. The resulting tree used only !! and !!" as predictors. Then, removing these two 219 predictors, the procedure was repeated and the second tree emerged with R only. Therefore, the

220 building of decision tree was performed once including only !!"# and !!" and once including only R. 221 Figure 4 shows a representation of the first decision tree. The animals with an increase of portal 222 pressure greater than 20% were classified in the ``Died before post-operative day 3’’ group. Moreover, 223 according to the decision tree, if the portal pressure increase was limited (less than 20%), but a large 224 decrease of the mean arterial pressure occurred (more than 30%), the animal was classified in the 225 ``Died before post-operative day 3’’ group. For this tree, the error rate on the training set was 1/13 ~ 226 7.7%. One animal was misclassified in the group ``Died before day 3’’. The leave-one-out error was 227 23% (number of misclassification in the procedure over size of the data set). 228 Figure 5 displays the second tree, built with the pressure-flow ratio. The results show that if R was 229 increased by more than 4.6 times or reduced by more than 0.7 times baseline, the animal was 230 classified in the ``Died before post-operative day 3’’ group. The error rate on the training set was 2/13 231 ~ 15% and the leave-one-out error was 31%. The misclassified animals, for this decision tree, were all 232 misclassified in the ``Alive after day 3’’ group. 233 234 Dynamical signals 235 Influence of respiration and variation during the cardiac cycle 236 Figure 6 presents the pressures and flows over time before the liver resection for three different 237 animals (data file for one animal is provided in supplementary material). The mechanical ventilation 238 had a small impact on arterial pressures and flows, unlike the heartbeat. The variations during a 239 cardiac cycle and due to mechanical ventilation were estimated as detailed in Data analysis section. 240 Table presents the average and standard deviation of the estimated variation during the cardiac cycle 241 and due to mechanical ventilation for each pressure and flow. 242 The variations around the average of arterial pressures and flows, at peak systole, due to mechanical 243 respiration were for hepatic artery (HA) flow: -5% (3%) / +9% (1%), for AoC flow: -5% (2%) / +7% 244 (2%), and for AoI flow: -7% (1%) / +6% (2%). Similarly, the sensitivity of the arterial pressures to 245 ventilation was small. For the carotid artery pressure: -4% (2%) / +5% (1%), and for the hepatic artery 246 pressure: -5% (1%) / +7% (1%). On the other hand, the variation during the cardiac cycle was larger 247 for the carotid artery pressure: -23% (8%) / +29% (11%) and for the hepatic artery pressure: -19% 248 (6%) / +19% (11%). The variation of the flow rates around the average during a cardiac cycle was for 249 the AoC: -74% (19%) / +133% (54%), for the AoI: -94% (27%) / +197% (103%) and for the hepatic 250 artery: 38% (15%) / 65% (39%).

251 Contrary to arterial signals, the venous pressures and flows were more sensitive to mechanical 252 ventilation than to cardiac cycle variation (except for the central vein pressure). Under mechanical 253 ventilation, the central venous pressure was maximal at the end of inspiration and minimal at the end 254 of expiration (Figure 6). The same observations were made for portal vein pressure. The variations due 255 to respiration were -85% (53%) / +122% (53%) for the central venous pressure and -11% (3%) / +17% 256 (3%) for the portal vein pressure. The variation during the cardiac cycle of the portal vein pressure was 257 smaller than for the central venous pressure, probably due to the distance to the heart. Indeed, the 258 estimated variation was -84% (91%) / +79% (79%) for the central venous pressure and -3% (1%) / 259 3% (1%) for the portal vein pressure. 260 Opposite to pressures, the portal flow (PV) was maximal at the end of the expiration cycle, and 261 minimal at the end of the inspiration (for mechanical ventilation) (Figure 6). The portal flow increased 262 by 35% (18%) and decreased by 47% (30%) during the respiration cycles. The variation during the 263 cardiac cycles was -12% (6%) / +12%(8%). 264 Hemodynamics changes due to clamping 265 Figure 7 shows the impact of the clamping on pressure and flow dynamics for 75% liver resection for 266 two different animals (a similar behavior was observed for all animals). 267 Central venous pressure was not affected by clamping (Figure 7). The carotid artery pressure was 268 varying at the clamping instant for the first animal, however it was because of the surgeon 269 intervention. Once the surgeons’ intervention stopped, but the vessel clamping remained effective, the 270 carotid artery pressure returned to its pre-resection value and waveform. For the second animal the 271 pressure in the carotid artery remained unchanged at clamping instant. The portal vein pressure 272 seemed to slowly increase (Figure 7) directly after the clamping. The amplitude of the hepatic artery 273 pressure increased after the 75% clamping (Figure 7). The averaged hepatic artery flow decreased at 274 clamping instant (Figure 7). Besides the mean value decrease, two characteristic changes could be 275 observed: the first peak was sharper, meaning the second peak was lower than before liver resection, 276 and diastolic flow was at low values for longer time. These changes were typical and were observed, 277 in all animals, at the clamping instant. The portal flow seemed to decrease slowly in one animal and 278 remained constant for the second one as shown in Figure 7. In the AoC neither the flow waveform nor 279 its mean value were impacted in the 10 seconds following the clamping (Figure 7). 280 281 Discussion 282 Mechanisms driving hemodynamics changes 283 This study enabled a precise description of the various changes of hemodynamics occurring in liver 284 surgery. With the continuous recordings, the impact of mechanical ventilation and variation during the 285 cardiac cycle were assessed. As expected, the venous flow rates and pressures were sensitive to 286 mechanical respiration. Whereas arterial pressures and flow rates were mainly sensitive to the heart 287 dynamics.

288 The decrease of the central venous pressure during hepatectomy was likely linked to the blood 289 losses and evaporation due to open-abdomen surgery. The blood losses were a consequence of 290 bleeding, but also of the removal of liver tissues (filled with blood). The anesthesiologists 291 compensated the blood volume changes with infusion of normal saline and 5% glucose solutions 292 (around 1.5L infused). The infusion tended to increase pressures, so the measured reduction of central 293 vein pressure resulted from the combination of blood losses and fluid infusion. The reduction of

294 arterial pressures (Pa , Pha) was also probably a consequence of blood volume changes. An increase of 295 3.1 mmHg between pre- and post-resection averaged portal pressure was observed similarly to the 296 increase reported in [11], [12] after 70% liver resection. 297 The impact of the ventilation on venous pressure was large, larger for central venous pressure than for 298 portal vein pressure. Thus, to estimate a venous pressure difference (across the liver for example), it is 299 essential to take the venous pressures at the same instant in the respiration cycle or to use pressures 300 averaged over several respiratory cycles. We observed that venous pressures were also largely 301 impacted by fluid infusion. However, all pressures are almost similarly impacted. The measurement of 302 the difference of two pressures (taken at the same instant in the respiration cycle) is then less sensitive 303 to fluid volume variations. 304 The portal flow was also impacted by the mechanical respiration. Since the portal vein flow 305 and the venous pressure difference were impacted by the respiratory cycle, what is sometimes called 306 the “liver resistance” (venous pressure difference divided by portal vein flow) was also impacted. Its 307 variations around the mean were very large (-99% (51%) / + 124% (75%)). 308 During 75% liver resection, the portal vein flow decreased similarly to AoC flow, suggesting that the 309 decrease of portal vein flow was linked to a decrease of cardiac output. This decrease may mainly be 310 due to blood volume changes. Therefore, when blood volume is restored (on post-operative days) 311 portal flow may increase. With the mathematical model proposed in [17], the numerical simulations 312 have shown that 75% liver resection without blood volume changes leads to only a 5% decrease of 313 portal flow. 314 Even if the portal vein flow decreased, the flow per liver mass increased drastically after liver 315 resection; it was multiplied by around 3.8. The portal flow per liver mass has been reported to be 316 multiplied by around 3, 4, 5.6 and 7.5 respectively, after 70%, 80%, 85% and 90% liver resection in 317 pigs [5], [11]. Moreover, [5] has concluded that liver fails to regenerate if post-operative portal flow 318 per liver mass is more than 5.6 times baseline. In [3] an increase of portal flow per liver mass greater 319 than 4 times the baseline has been associated with a risk of liver failure (after liver transplantation). 320 The hepatic artery flow decrease (74%) seemed to correspond to the liver mass reduction 321 (75%). The reduction of mass led to an increased resistance to arterial flow and therefore a decrease of 322 hepatic artery flow. The hepatic arterial buffer response (HABR) is a physiological mechanism in 323 which the hepatic artery dilates or constricts to compensate for the changes in the portal vein flow 324 [18]. The arterial buffer seemed unnecessary to explain the decrease of hepatic artery flow after 75%

325 liver resection. Indeed, in [17] the decrease was explained with a lumped mathematical model, that did 326 not include HABR. 327 In addition to the decrease of average flow, a typical change of waveform was observed in the hepatic 328 artery pressure and flow. The change of hepatic artery tree architecture after liver resection was 329 assumed responsible for the typical change of waveform (3 main liver lobes before resection and only 330 one main liver lobe after resection). In [19], a mathematical model has been proposed to study and 331 explain these particular waveform changes. 332 333 Survival explanation with decision trees 334 The decision tree suggested that to have a ``safe’’ partial hepatectomy, the portal vein pressure 335 increase and the arterial pressure decrease have to be limited. A large increase of portal pressure 336 (greater than 20%) seemed correlated to the animal early death. A similar result has been reported in 337 patients [1]. When the increase of portal pressure was limited, the early death seemed related to the 338 large decrease of arterial pressure (larger than 30%). A large drop of arterial pressure may be due to 339 blood losses. 340 The second decision tree was built with the pressure-flow ratio, combining information on pressure 341 ‘gradient’ (drop) across the liver as well as liver inflow. A ratio multiplied by a factor smaller than 0.7 342 or larger than 4.6 seemed to lead to an early death of the animal (before third post-operative day). The 343 large increase of pressure-flow ratio can be related to a large increase of pressure drop, a decrease of 344 liver inflow or both. A low portal vein flow may lead to a poor liver regeneration as reported in [20] 345 after partial liver transplant with portocaval shunt. A large decrease of pressure-flow ratio can be due 346 to a decrease of venous pressure drop across the liver, an increase of liver inflow or both. 347 The control of portal hemodynamics with a medical device, such as the one proposed by our group [6], 348 to limit the increase of this pressure-flow ratio (limiting the increase of portal vein pressure) may 349 result in reducing the risk of post-operative liver failure. 350 The first tree misclassified one animal (of the training set) in the group “died before post-operative 351 day 3”. Whereas for the second tree, misclassified animals were all misclassified in the ``Alive after 352 day 3’’ group. The second type of misclassification is more problematic than the first one. The small 353 number of animals explains the trees errors, and hopefully with a larger training set the tree threshold 354 would be more precise and the errors decrease. 355 356 Variability is challenging 357 The flows and arterial pressures had the smallest relative standard deviation, whereas the venous 358 pressures had the highest ones (Table 1). The calibration of the pressure catheters was more difficult 359 than for flow probes. Therefore, the large standard deviation of the pressure may be due to the 360 measurement technique. Furthermore, during the surgery we have observed that the venous pressures 361 were the most sensitive to fluid infusion. Following a fluid infusion all the pressures rise; since the

362 venous pressures are smaller than arterial ones, the relative increase was larger for venous pressure. 363 This may explained the large variability of venous pressure evolution. The computed increase of portal 364 vein pressure was low in average, however a large inter-animal variability was observed. 365 Different evolutions of AoI flow were also observed after liver resection, caused by the reorganization 366 of aortic flow partition. The AoC flow and the portal vein flow decreased after liver resection. In 367 addition, the arterial tree resistance of the liver increased leading to a lower hepatic artery flow. 368 Therefore, depending on the magnitude of these different changes, the resulting AoI flow sometimes 369 increased and other times decreased. 370 371 Conclusions 372 To conclude, the presented continuous recordings of pressure and flow enabled us to understand the 373 sensitivity of hemodynamics to mechanical ventilation. In addition, the dynamical changes due to liver 374 resection, and more specifically at the time of liver vessel clamping were also assessed. 375 Based on these measurements, the main findings that are clinically relevant are the following: 376 1. The portal vein flow decreased after liver resection. This decrease seemed mainly due to a 377 reduction of the total blood volume (caused by the blood losses and by the removal of the 378 blood contained in the resected liver lobes but also compensated by fluid infusion). 379 2. The portal and central venous pressures were largely impacted by the mechanical 380 ventilation and the blood volume changes occurring during surgery. Measurements of 381 pressure difference across the liver (often called in clinical papers ``pressure gradient’’) are 382 thus sensitive to experimental conditions. In order to obtain a good estimation of this 383 pressure difference, one-time measurements in both vessels should be taken at the same 384 instant in the respiration cycle, e.g at the end of expiration. If continuous recording is 385 possible, then average over several respiratory cycles may be used. 386 3. For 75% liver resection, the average decrease of hepatic artery flow was proportional to the 387 liver mass resection. After the resection, the liver resistance to arterial flow was increased, 388 leading to the reduction of arterial flow. The hepatic arterial buffer response did not seem to 389 play a role in this decrease of liver arterial flow after 75% liver resection [17]. 390 4. A typical change of cardiac-cycle waveform in the hepatic artery flow and pressure was 391 observed during 75% liver resection. This change is presumably linked to the hepatic artery 392 tree architecture [19]. 393 5. The decision tree suggested that the surgery risk is reduced when the increase of portal vein 394 pressure and the reduction of arterial pressure are limited (portal vein increase by less than 395 20% and arterial pressure decrease by less than 30%). In addition, when the indicator !.!(! !! ) 396 ! = !" !" was largely reduced or increased (multiplied by less than 0.7 or more than (!!!!!!") 397 4.6) compared to baseline, early death occurred. The previous works usually focused on the

398 increase of portal vein pressure [1], [7], [12], or flow [4], [5]. The relation between the two 399 seems also very important. The proposed predictor includes the variation of pressure with 400 respect to flow occurring during liver surgery. 401 402 403 Competing interests 404 The authors declare no conflicts of interest. 405 Author contributions. 406 Eric Vibert and Irene Vignon-Clementel conceived the experiments. Chloe Audebert, Mohamed 407 Bekheit, and Petru Bucur performed the experiments and measurements. Chloe Audebert, Petru Bucur 408 and Irene Vignon-Clementel analyzed and interpreted the data. All authors contributed to writing or 409 critical revision of the paper. 410 Funding 411 This material is based upon work supported by the French National Agency for Research ANR-13- 412 TECS-0006 iFLOW. 413 Acknowledgment 414 The authors are grateful to the INRA Plateforme CIRE (Nouzilly, France) staff for their technical 415 assistance in surgeries and imaging, and to Mylène Wartenberg for assistance in taking measurements. 416

417 Tables Variable units D0 before (X0) D0 after (X) p value (X - X0)/X0 (animal number)

Pcv mmHg (n=22) 2.1 (1.30) 0.89 (3.4) 0.01 -50% (-169% / 977%)

Pa mmHg (n=22) 54.5 (0.15) 46.4 (0.22) <0.001 -9% (-56% / 6%)

QAoC L/min (n=17) 2.06 (0.40) 1.79 (0.44) 0.004 -14% (-45% / 32%)

QAoI L/min (n=14) 0.87 (0.35) 0.78 (0.46) NS -16% (-38% / 53%)

Pha mmHg (n=21) 48.2 (0.26) 42.4 (0.26) NS -11% (-59% / 46%)

Qha L/min (n=21) 0.18 (0.37) 0.05 (0.45) <0.001 -74% (-90% / -6%)

Qpv L/min (n=21) 0.74 (0.26) 0.66 (0.30) 0.02 -9% (-50% / 63%)

Ppv mmHg (n=21) 6.7 (0.48) 7.3 (0.44) NS 3% (-90% / 186%)

Ppv – Pvc mmHg (n=21) 4.5 (0.55) 6.4 (0.36) 0.003 57% (-69% / 331%)

418 Table 1: Hemodynamics measurements before and after resection, average and in parenthesis relative 419 standard deviation (std/avg) (two first columns), two-sample Wilcoxon paired tests to compare pre- 420 and post-resection variables (third column) and median (min/max) post-resection evolution of the 421 pressures and flows compared with pre-resection values (fourth column). 422

Mechanical ventilation

Animals (n=4) Pvc Pa QaoC QaoI Pha Qha Qpv Ppv

(min-avg)/avg (%) -85 (53) -4 (2) -5 (2) -7 (1) -5 (1) -7 (3) -47 (30) -11 (3)

(max-avg)/avg (%) 122 (53) 5 (1) 7 (2) 6 (2) 7 (1) 9 (1) 35 (18) 17 (3)

Heartbeat

Animals (n=4) Pvc Pa QaoC QaoI Pha Qha Qpv Ppv

(min-avg)/avg (%) -84 (91) -23 (8) -74 (19) -94 (27) -19 (6) -38 (15) -12 (6) -3 (1)

(max-avg)/avg (%) 79 (79) 29 (11) 133 (54) 197 (103) 19 (11) 65 (39) 12 (8) 3 (1) 423 Table 2: Evolution due to mechanical ventilation and heartbeat, average and standard deviation. The 424 variation around the average (in time) during 20 seconds was computed to estimate the impact of the 425 ventilation for signals measured in veins. For arterial signals, the variation around the average in peak

426 systole was computed. The variation around the average (in time) during two cardiac cycles was 427 computed to evaluate the variations during cardiac cycles. 428 429 Figures

Hepatic artery Aorta above celiac Gastroduodenal artery trunk

Common hepatic artery

Celiac trunk

Mesenteric artery

Aorta between mesenteric artery and kidney arteries Right kidney artery Left kidney artery

430 431 Figure 1: Reconstruction from arterial time CT-scan. The zoom on celiac and splanchnic circulations 432 displays the vessels where flow and/or pressure are measured (framed in red).

433

434 435 Figure 2: Hepatic artery and portal vein pressure catheters and flow probes.

p<0.001 500 35

450 30 400

350 25

300 20 250 15 200

150 10

100 PV ﬂow per liver weight (ml/min/100gLW) HA ﬂow per liver weight (ml/min/100gLW) 5 50

0 0 pre-resec.on post-resec.on pre-resec.on post-resec.on 436 437 Figure 3: Portal vein (left) and hepatic artery (right) flow per liver weight before and after 75% liver 438 resection. Two-sample Wilcoxon paired tests were performed to compare pre- and post-resection 439 variables.

440 441

Portal vein pressure increased > 20%

FALSE TRUE

Carotid artery pressure decreased > 30% Died before D3

FALSE TRUE

Alive after D3 442 Died before D3 443 Figure 4: Decision tree built with 13 animals (among which 5 died before post-operative day 3) based 444 on evolution of carotid and portal vein pressure (with fitctree MATLAB function).

Pressure-flow ratio multiplied by > 4.6 or < 0.7

FALSE TRUE

Died before D3 445 Alive after D3 446 Figure 5: Decision tree built with 13 animals (among which 5 died before post-operative day 3) with 447 pressure-flow ratio (with fitctree MATLAB function).

448

449

450

451

452

453

454

455 Figure 6: Pressures (mmHg) (4 first plots: Pcv, Pa, Ppv, Pha) and flows (L/min) (4 last plots: Qpv, Qha,

456 QAoC, QAoI) over a few respiration cycles (left) and during two cardiac cycles (right) at the start of the 457 surgery, before the liver resection, for three different animals. Pressure and flow for one vessel are 458 represented by a single color (hepatic artery in red, portal vein in blue, carotid artery pressure and AoC 459 flow in pink). A data file for one animal is provided in supplementary material.

460

461

462 Figure 7: Central venous (CV), carotid artery (Pa), hepatic artery (HA) and portal vein (PV) pressures, 463 and hepatic artery, portal vein and celiac aorta (AoC) flows measured in two animals during the 464 clamping for 75% liver resection over time. Black lines indicate the clamping time. Pressure and flow 465 for one vessel are represented by a single color (hepatic artery in red, portal vein in blue, carotid artery 466 pressure and AoC flow in pink).

467 468

469 REFERENCES 470 [1] M.-A. Allard et al., “Posthepatectomy portal vein pressure predicts liver failure and mortality after 471 major liver resection on noncirrhotic liver,” Ann. Surg., vol. 258, no. 5, pp. 822–830, 2013. 472 [2] J. Asencio, J. Vaquero, L. Olmedilla, and J. G. Sabrido, “‘Small-for-flow’ syndrome: shifting the 473 ‘size’ paradigm,” Med. Hypotheses, vol. 80, no. 5, pp. 573–577, 2013. 474 [3] M. Sainz-Barriga, L. Scudeller, M. G. Costa, B. de Hemptinne, and R. I. Troisi, “Lack of a 475 correlation between portal vein flow and pressure: toward a shared interpretation of 476 hemodynamic stress governing inflow modulation in liver transplantation,” Liver Transpl., vol. 477 17, no. 7, pp. 836–848, 2011. 478 [4] B. B. Vasavada, C. L. Chen, and M. Zakaria, “Portal flow is the main predictor of early graft 479 dysfunction regardless of the GRWR,” Int. J. Surg., no. 2, pp. 177–180. 480 [5] L. Xiang, L. Huang, X. Wang, Y. Zhao, Y. Liu, and J. Tan, “How Much Portal Vein Flow Is Too 481 Much for Liver Remnant in a Stable Porcine Model?,” in Transplantation proceedings, 2016, 482 vol. 48, pp. 234–241. 483 [6] P. O. Bucur et al., “Modulating Portal Hemodynamics With Vascular Ring Allows Efficient 484 Regeneration After Partial Hepatectomy in a Porcine Model.,” Ann. Surg., 2017. 485 [7] Y. Ogura et al., “Portal pressure < 15 mm Hg is a key for successful adult living donor liver 486 transplantation utilizing smaller grafts than before,” Liver Transpl., vol. 16, no. 6, pp. 718–728, 487 2010. 488 [8] E. Bendixen, M. Danielsen, K. Larsen, and C. Bendixen, “Advances in porcine genomics and 489 proteomics—a toolbox for developing the pig as a model organism for molecular biomedical 490 research,” Brief. Funct. Genomics, vol. 9, no. 3, pp. 208–219, 2010. 491 [9] F. Court et al., “Segmental nature of the porcine liver and its potential as a model for experimental 492 partial hepatectomy,” Br. J. Surg., vol. 90, no. 4, pp. 440–444, 2003. 493 [10] H. Boxenbaum, “Interspecies variation in liver weight, hepatic blood flow, and antipyrine intrinsic 494 clearance: extrapolation of data to benzodiazepines and phenytoin,” J. Pharmacokinet. 495 Biopharm., vol. 8, no. 2, pp. 165–176, 1980. 496 [11] J. Leng, H. Xing, J. Tan, K. Chen, and J. Dong, “The safe minimally ischemic liver remnant for 497 small-for-size syndrome in porcine hepatectomy,” in Transplantation proceedings, 2013, vol. 498 45, pp. 2419–2424. 499 [12] K. Iguchi et al., “Hepatoprotective effect by pretreatment with olprinone in a swine partial 500 hepatectomy model,” Liver Transpl., vol. 20, no. 7, pp. 838–849, 2014. 501 [13] M. Pouyet, I. Méchet, C. Paquet, and J.-Y. Scoazec, “Liver regeneration and hemodynamics in 502 pigs with mesocaval shunt,” J. Surg. Res., vol. 138, no. 1, pp. 128–134, 2007. 503 [14] P. Bucur, M. Bekheit, C. Audebert, I. Vignon-Clementel, and E. Vibert, “Simplified technique for 504 75% and 90% hepatic resection with hemodynamic monitoring in a large white swine model,” J. 505 Surg. Res., 2016.

506 [15] L. Breiman, Ed., Classification and regression trees, Repr. Boca Raton: Chapman & Hall [u.a.], 507 1998. 508 [16] C. Debbaut et al., “Modeling the impact of partial hepatectomy on the hepatic hemodynamics 509 using a rat model,” IEEE Trans. Biomed. Eng., vol. 59, no. 12, pp. 3293–3303, Dec. 2012. 510 [17] C. Audebert, M. Bekheit, P. Bucur, E. Vibert, and I. E. Vignon-Clementel, “Partial hepatectomy 511 hemodynamics changes: Experimental data explained by closed-loop lumped modeling,” J. 512 Biomech., vol. 50, pp. 202–208, Jan. 2017. 513 [18] W. W. Lautt, “Regulatory processes interacting to maintain hepatic blood flow constancy: 514 Vascular compliance, hepatic arterial buffer response, hepatorenal reflex, liver regeneration, 515 escape from vasoconstriction,” Hepatol. Res., vol. 37, no. 11, pp. 891–903, Nov. 2007. 516 [19] C. Audebert, P. Bucur, M. Bekheit, E. Vibert, I. E. Vignon-Clementel, and J.-F. Gerbeau, “Kinetic 517 scheme for arterial and venous blood flow, and application to partial hepatectomy modeling,” 518 Comput. Methods Appl. Mech. Eng., vol. 314, pp. 102–125, 2017. 519 [20] A. J. Hessheimer et al., “Decompression of the Portal Bed and Twice-Baseline Portal Inflow Are 520 Necessary for the Functional Recovery of a ‘Small-for-Size’ Graft:,” Ann. Surg., vol. 253, no. 6, 521 pp. 1201–1210, Jun. 2011. 522 523

524 Supplementary material Animals 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Died before D3 0 1 0 1 1 1 0 0 1 0 0 1 1 0 0 0 1 0 0 0 1 Pre-resection Pcv mmHg 4.5 4.7 4.1 4.6 5.5 6.3 4.8 0.1 2.3 1.5 1.8 0.6 1.9 2.0 2.0 1.8 0.4 -0.3 -2.8 -5.0 1.1 5.0 Pa mmHg 64.7 55.4 48.8 53.4 50.1 41.6 35.9 59.8 52.9 57.4 58.3 53.0 51.4 77.0 56.0 54.5 52.7 49.4 47.5 57.5 63.9 57.0 QAoC L/min 2.23 2.38 1.75 2.50 2.57 1.71 2.71 1.75 1.55 3.43 4.17 1.03 1.62 1.57 1.33 1.24 1.42

QAoI L/min 0.45 0.85 0.60 1.33 1.25 0.55 0.70 0.91 0.84 1.50 0.77 0.87 0.81 0.71

Pha mmHg 48.3 60.0 32.6 43.5 42.9 43.6 32.1 36.9 34.8 36.3 41.8 37.8 53.0 75.0 55.6 44.2 54.6 37.1 62.7 63.1 70.2 54.0 Qha L/min 0.12 0.23 0.03 0.27 0.20 0.05 0.20 0.21 0.21 0.17 0.21 0.17 0.23 0.11 0.11 0.12 0.21 0.26 0.20 0.24 0.14

Qpv L/min 1.00 0.99 0.60 0.92 1.13 0.72 0.73 0.67 0.86 0.62 0.68 0.52 0.68 0.98 0.82 0.64 0.75 0.72 0.36 0.53 0.92 0.46 Ppv mmHg 6.6 7.3 7.0 7.5 8.6 10.5 6.0 8.0 9.0 10.5 10.9 7.3 4.1 5.0 8.1 3.8 7.2 -1.0 -1.1 7.3 9.0

Ppv-Pv mmHg 2.1 2.7 3.0 2.9 3.1 4.2 1.3 7.8 6.7 9.0 9.1 6.7 2.2 3.0 6.1 2.1 7.5 1.8 3.9 6.2 4.0

Liver Volume mL 963 1020 1085 1049 718 722 1254 684 908 805 1156 735 1093 1136 869 1209 887 905 997 845

Post-resection Pcv mmHg 2.4 1.3 1.1 6.2 7.2 0.7 2.2 1.4 1.7 0.1 4.9 0.2 2.1 0.7 -0.3 -1.2 0.3 -0.7 -3.8 -7.2 0.2 0.0 Pa mmHg 68.7 24.3 48.9 31.1 45.5 31.6 36.0 56.0 50.7 52.2 53.2 47.9 46.1 36.9 53.8 45.2 40.1 51.7 46.5 59.0 52.9 47.0 QAoC L/min 2.14 2.04 1.63 1.87 2.4 1.28 2.76 1.63 1.34 2.7 3.54 0.62 2.14 0.87 1.14 0.81 1.48

QAoI L/min 0.55 0.63 0.55 0.82 1.91 0.84 0.48 0.92 0.65 1.06 0.53 0.71 0.71 0.62

Pha mmHg 52.3 24.9 47.6 28.0 40.7 31.9 38.5 32.6 35.3 39.9 32.5 36.0 47.1 38.4 46.8 42.6 43.4 51.3 71.9 60.9 47.0

Qha L/min 0.1 0.06 0.03 0.05 0.04 0.02 0.08 0.02 0.05 0.04 0.07 0.08 0.04 0.03 0.06 0.03 0.06 0.05 0.03 0.08 0.025

Qpv L/min 0.77 0.6 0.81 0.87 0.63 1.17 0.59 0.54 0.8 0.54 0.63 0.49 0.62 0.49 0.54 0.76 0.81 0.33 0.44 0.93 0.41

Ppv mmHg 8.5 8.4 7.7 7.1 10.4 7.9 6.2 6.1 11.2 10.2 12.6 7.2 11.6 5.4 8.2 4.5 7.9 -0.1 -0.2 6.0 7.0

Ppv-Pv mmHg 6.1 7.1 6.6 0.9 3.2 7.2 4.0 4.7 9.5 10.1 7.7 7.0 9.5 4.7 8.5 5.7 8.6 3.7 7.0 5.9 7.0

Liver Volume mL 390 170 183 363 212 155 272 206 215 235 218 180 137 292 256 332 208 201 224 327

525 Table 1: Time-averaged hemodynamics measurements for 22 pigs before and after 75% liver resection. Pre-resection measurements were recorded at the 526 beginning of the surgery before liver resection (right after position of all the probes) and post-resection measurements were recorded at the end of the surgery 527 after liver resection (just before incision closure, around 1 hour after clamping).

528 529

Journal of Biomechanics 50 (2017) 202–208

Contents lists available at ScienceDirect

Journal of Biomechanics

journal homepage: www.elsevier.com/locate/jbiomech www.JBiomech.com

Partial hepatectomy hemodynamics changes: Experimental data explained by closed-loop lumped modeling

Chloe Audebert a,b, Mohamed Bekheit c, Petru Bucur c,d, Eric Vibert c,e, Irene E. Vignon-Clementel a,b,n a Inria, Paris, France b Sorbonne Universités UPMC Univ Paris 6, Laboratoire Jacques-Louis Lions, France c Inserm Unité 1193, Villejuif, France d CHRU, hôpitaux de Tours,Chirurgie Hépato-biliaire et Pancréatique, Transplantation Hépatique, Tours, France e AP-HP, Hôpital Paul Brousse, Centre Hépato-Biliaire, Villejuif, France article info abstract

Article history: The liver function may be degraded after partial liver ablation surgery. Adverse liver hemodynamics have Accepted 2 November 2016 been shown to be associated to liver failure. The link between these hemodynamics changes and ablation size is however poorly understood. This article proposes to explain with a closed-loop lumped model the Keywords: hemodynamics changes observed during twelve surgeries in pigs. The portal venous tree is modeled with Closed-loop lumped model a pressure-dependent variable resistor. The variables measured, before liver ablation, are used to tune the Hemodynamics model parameters. Then, the liver partial ablation is simulated with the model and the simulated Surgery simulation pressures and flows are compared with post-operative measurements. Fluid infusion and blood losses Hepatectomy occur during the surgery. The closed-loop model presented accounts for these blood volume changes. Validation Moreover, the impact of blood volume changes and the liver lobe mass estimations on the simulated Swine variables is studied. The typical increase of portal pressure, increase of liver pressure loss, slight decrease of portal flow and major decrease in arterial flow are quantitatively captured by the model for a 75% hepatectomy. It appears that the 75% decrease in hepatic arterial flow can be explained by the resistance increase induced by the surgery, and that no hepatic arterial buffer response (HABR) mechanism is needed to account for this change. The different post-operative states, observed in experiments, are reproduced with the proposed model. Thus, an explanation for inter-subjects post-operative variability is proposed. The presented framework can easily be adapted to other species circulations and to different pathologies for clinical hepatic applications. & 2016 Elsevier Ltd. All rights reserved.

1. Introduction hepatic venous pressure gradient (Sainz-Barriga et al., 2011) are associated with post-surgery liver failure, the link between Major liver resection (partial hepatectomy) is being performed resected volume and hemodynamics changes remains unclear. to treat liver lesions or for adult-to-adult living donor liver Since the liver receives around 25% of the cardiac output, hepa- transplantation. Due to liver regeneration, during the post- tectomy may impact the whole blood circulation. Thus the present operative period of a few months, the patient re-gains a normal work aims to develop a mathematical model to explain the various liver mass. However, sometimes liver function is poorly recovered hemodynamics changes observed in experimental surgeries of and post-operative liver failure may occur. twelve pigs. Pig is considered a good animal model for liver. Liver hemodynamics is modified by the surgery, which The proposed model is constructed to satisfy the following increases the resistance to flow of the organ. To understand it is requirements. First, the equations must be numerically fast to not easy, partly because the liver is perfused by both arterial and solve, to explore a diversity of hypotheses with all the pigs data. venous blood. Although high portal pressure (Allard et al., 2013), Second, the number of parameters must remain small enough so high portal flow (Iida et al., 2007; Vasavada et al., 2014), and high that calibration is tractable. Finally, the whole blood circulation must be taken into account, and hepatectomy dynamically modeled. Consequently, a closed-loop lumped model (also called n Corresponding author at: REO - A321, INRIA Paris, 2 rue Simone Iff, CS 42112, 75589 Paris Cedex 12, France. 0D model), taking into account the liver and groups of organs, is E-mail address: [email protected] (I.E. Vignon-Clementel). presented. http://dx.doi.org/10.1016/j.jbiomech.2016.11.037 0021-9290/& 2016 Elsevier Ltd. All rights reserved. C. Audebert et al. / Journal of Biomechanics 50 (2017) 202–208 203

Different groups have worked on liver hemodynamics mod- partly on liver lobe masses. Thus, several options are tested. eling, at different liver scales and for various applications. Liver The paper is organized in the following manner. Section 2 lobule porous models have been proposed, to model glucose presents the available experimental measurements, the cardi- transport and metabolism (Ricken et al., 2015), to study the ovascular and liver models and their parametrization. Section 3 influence of a septum and tissue permeability (Debbaut et al., shows partial hepatectomy simulation results and comparison 2014b), including in cirrhosis (Peeters et al., 2015)ortosimulate with measurements. Section 4 discusses model capabilities. the impact of deformation on pressure–flow relation (Bonfiglio et al., 2010). At the organ scale, liver π lumped models for multiple vascular generations have been used to study the hypo- 2. Methods thermic machine perfusion (Van Der Plaats et al., 2004; Debbaut – et al., 2014a). A lumped model of the splanchnic and liver cir- 2.1. Liver surgery experimental measurements culation has been proposed to illustrate the link between hepatic Hepatectomies are performed on several pigs to study the hemodynamics venous pressure increase, vessel contractility and liver interstitial impacts. Approval of the committee of ethics of animal research, ministry of higher fluid (Chu and Reddy, 1992). Models have been developed on education and scientific research and ministry of agriculture and fishing was transport and diffusion of a compound in the liver, including obtained. The pig liver is composed of five lobes, usually considered as three main whole-body pharmacokinetics models (Schwen et al., 2015) lobes (Court et al., 2003): left lobe, median lobe (subdivided in left medial and right medial lobes) and right lobe (subdivided in right lateral and caudate lobes). The or to study tumor detection with Magnetic Resonance Images median and left lobes are resected. Since the median lobe is around twice the size (Bezy-Wendling et al., 2007). Convection is based on resistive of left and right lobes, around 75% hepatectomy is performed. models of the different generations of arterial and venous trees. During surgery, several measurements are continuously recorded. Three pres- In Lukeš et al. (2014),theflow in liver arterial and venous trees is sures and three flows are the basis of parameter tuning and model validation. These modeled for the first generations with Bernoulli equation, while a measurements are averaged over 20 s during a stable state of the surgery. Pre- fl resection and post-resection (immediately after surgical clamping) states are con- porous media models the ow in the smallest vessels. The trees sidered. The carotid artery (CA), portal vein (PV) and central venous (CV) pressures geometry is based on CT-scans. Hepatic artery flow 3D CFD are measured. The latter is a surrogate for the hepatic vein (v) pressure. The flows simulations for rigid and flexible walls have been performed in are recorded in the aorta above the celiac trunk (celiac aorta), the hepatic artery Childress and Kleinstreuer (2014) to study direct drug-targeting. (HA) and the portal vein. Cardiac output (CO) is estimated assuming celiac aorta flow is around 60% of CO (Lantz et al., 1981) (assuming humans and pigs flow Liver models have also been developed to study the impact of distributions are similar (Swindle et al., 2012)). Heart rate is computed from the CA liver surgery. Flow behavior for different H-Graft diameters has pressure measurement. been studied with a resistive model and compared to clinical Before and after the surgery a CT-scan is performed with a Siemens Somatom observations in Rypins et al. (1987).A3DCFDsimulationhas AS definition 128 machine. Image acquisitions are done before, 15, 35, 55 and 75 s after injection of 75 ml of iohexol 350 mg/ml (Omnipaque, GE Healthcare) with a been performed in the portal vein before and after right lobe rate of 5 ml per second. From the CT-scans liver volumes are estimated. After hepatectomy in Ho et al. (2012). The surgery was simulated by ablation, the removed liver is weighted; left and median lobe masses are then changing the geometry. Similarly, for a two-lobe liver lumped assessed. To estimate the liver masses, four different assumptions are made as model, driving conditions were kept unchanged before and after described in Table 1, with varying predictive capabilities. hepatectomy. Various resection sizes and two different surgical techniques have been simulated using a resistance model, based 2.2. 0D closed-loop model on cast reconstruction, of rat liver vasculature (Debbaut et al., 2012). Most of these works thus do not consider the dynamics A 0D hemodynamics model of the entire cardiovascular system (Liang and Liu, 2005; Segers et al., 2003) is coupled to a new model of liver that is structured by induced by the surgery or the interaction with the rest of the lobes. The model aims to represent hepatectomy, i.e. the resection but also other circulation. related phenomena. Hence, only the involved organs are included, resulting in five The present work proposes to model liver partial ablation blocks (Fig. 1). dynamically, with a closed-loop 0D model of the cardiovascular Lungs (i¼L), digestive organs (i¼DO) and other organs (i¼OO) are represented by three-element Windkessel models: system and the liver. In Audebert et al. (2016),wehavepro- 8 posed a numerical scheme for 1D hemodynamics models, and > i > dPp > Ci ¼ Q i Q i explored in a generic 1D-0D pig model the hepatic artery < dt a v i i ¼ i i ð1Þ waveforms, to understand the experimental changes observed > RpQ a Pa Pp > :> i i ¼ i i during hepatectomy. Here, the impact of the surgery on the RdQ v Pp Pv liver and on the whole body hemodynamics is identified. i i fl i i i Moreover the consequences of blood loss and infusion are where for block iQa and Qv are arterial and venous ows, Pa, Pp and Pv are arterial, proximal and venous pressures, Ri , Ri and Ci are proximal and distal resistances studied. The simulations, done for twelve pigs, are quantita- p d and capacitance (Fig. 1). tively compared to experimental measurements from 75% pig Heart model: The heart model is based on Suga and Sagawa (1974), Pennati hepatectomy. Prediction of hemodynamics changes relies et al. (1997), Liang et al. (2009), and Blanco and Feijóo (2013). To obtain smooth, yet

Table 1 The different mass assumptions description of the total liver, left lobe, right lobe and median lobe. Their degree of certainty increases, and conversely their degree of predictability decreases from A1 to A4: preop calculation, peri-op calculation possible, post-op calculation.

Mass assumption A1 A2 A3 A4

Total liver mass estimate with estimate with sum of lobe sum of lobe pre-op CT-scan pre-op CT-scan masses masses Left lobe mass 1/3 planned resected weight after weight after weight after mass (pre-op CT-scan) resection resection resection Right lobe mass planned remaining total mass minus equal to left estimate with mass (pre-op CT-scan) left and median lobe masses lobe mass post-op CT-scan Median lobe mass 2/3 planned resected weight after weight after weight after mass (pre-op CT-scan) resection resection resection 204 C. Audebert et al. / Journal of Biomechanics 50 (2017) 202–208

Fig. 1. Schematic representation of the 0D closed-loop cardiovascular and liver blood circulations. RCR block and liver lobe i parameters are shown. Q i;b is the infused or removed ﬂow to account for blood volume changes.

sharp, transitions between open and closed valves, logistic functions are used for T75, the resection time: valves (Audebert et al., 2016). ( 1ift oT Liver model: The liver tissue is perfused with venous blood through PV and ð Þ¼ 75 ð Þ r t ð ð ÞÞ 4 arterial blood through HA and drained by the hepatic vein. The liver main lobes are exp 5 t T75 otherwise represented by three blocks in parallel (Fig. 1), related to the left heart (arterial Infusion or bleeding modeling: During surgery, blood and lymph losses and input), digestive organs (venous input) and right heart (venous output) compart- evaporation due to open-abdomen occur. Anesthetists thus infuse fluid. Conse- ments. Within each lobe, the HA tree is represented by a single resistance. The PV quently blood volume is not constant. In the model, systemic venous flow is thus tree is modeled by a non-linear resistance to represent pressure (P) dependent changed dynamically by adding or removing flow to represent bleeding or infusion resistance (R), the subscript labelling before resection: 0 (Fig. 1): ! : r r þ 5 0 75 V i=b 1 Ti=b t Ti=b Di=b R0 R0 Q ; ðtÞ¼ with Δt ¼ ð5Þ P P0 K ¼ 0; ð2Þ i b ð þ ð : ðΔ : ÞÞÞ R R Di=b 1 exp 0 1 t 0 01 0 otherwise

where Ti=b and Di=b are the time and duration of the infusion or bleeding and V i=b is This equation is derived from a vein tube law relating pressure and area (Shapiro, the added or removed volume. 10 1:5 1977; Pedley et al., 1996): P P0 ¼ KA=A0 A=A0 . Then, the area and the Numerical resolution: The model leads to a system of nonlinear differential- resistance are related assuming Poiseuille flow. Finally, liver tissue and hepatic vein algebraic equations, solved with the IDA package from SUNDIALS (Serban et al., trees are modeled with a capacitance and a resistance (Bennett and Rothe, 1981). 2015). A Backward Differentiation Formula is used for time integration, and a The resistances and capacitance in a lobe are assumed, as a first approximation, Newton method for the resulting nonlinear system. proportional to lobe mass inverse and lobe mass respectively. Thus, for each lobe: Statistic tests: A test of equivalence compares the model outputs and the measurements. The equivalence test null hypothesis is the dissimilarity of the 8 > M two populations. Thus, the test assumes the populations are different and uses > Pa P ; ¼ R Q ; > t i haM a i the data to prove otherwise (Robinson et al., 2005). A two one-sided t-test > i > (TOST) is used, with the R function TOST in the “equivalence” package > ¼ ð Þ M < Ppv Pt;i Rpv Ppv Q pv;i (R Development Core Team, 2008; Robinson, 2016). For the equivalence test, a Mi ð Þ > 3 region of indifference has to be defined; here a 10% relative error is chosen. If the > ¼ M > Pt;i Pv Rl Q v;i fi > M region of indifference is contained in the con dence interval, then the two > i > populations are deemed significantly similar. If not, the null hypothesis is not > dPt : C M ¼ Q ; þQ ; Q ; l i dt a i pv i v i rejected (Robinson and Froese, 2004).

where Pa, Ppv, Pt;i and Pv are respectively the pressure for artery, portal vein, ith lobe 2.3. Parameter tuning procedure fl tissue and venous pressure. Q a;i, Q pv;i and Q v;i are HA, PV and hepatic venous ows in lobe i respectively. M and Mi are the liver and lobe i mass estimations. Rha, Rpv, Rl Based on the available cardiac-cycle-averaged measurements and literature are resistances for HA tree, PV tree, and liver tissue and hepatic venous tree, data, model parameters are tuned to obtain similar pressures and flows to the respectively. Cl is the liver tissue capacitance per liver mass, the only liver non pig- measured ones before resection. specific parameter. Systemic circulation: CA and CV pressure measurements are used as target Partial hepatectomy simulation: To model median and left lobes resection, the arterial (Pa) and venous (Pv) pressures in the model. Combined with CO, the sys- corresponding HA and PV resistances are multiplied by a function dependent on temic equivalent resistance is computed: Req ¼ðPa PvÞ=CO. Similarly, the RCR total C. Audebert et al. / Journal of Biomechanics 50 (2017) 202–208 205

resistance for DO is computed with measured Pa, Ppv and Qpv. The liver tissue Table 3 pressure (P ) is estimated assuming that the pressure drop between PV and liver t Mean and standard deviation (in parenthesis) of relative error Erel for all the 12 tissue is 80% of the pressure drop between PV and hepatic vein (Ppv Pv)(Debbaut E ¼ jj =jj simulations, rel Xsimu Xmes Xmes with Xsimu the simulated and Xmes the mea- et al., 2012). Combined with the Pa and Qha measurements, the HA tree resistance is sured post-resection variable. computed: Mass assumption A1 A2 A3 A4 ¼ Pa Pt Rha Q ha Pa 0.2 (0.18) 0.2 (0.18) 0.21 (0.18) 0.2 (0.18)

Similarly, the liver tissue resistance and the initial portal resistance (R0 in Eq. (2)) Ppv 0.28 (0.19) 0.46 (0.57) 0.47 (0.46) 0.37 (0.26) are computed: Ppv Pv 0.39 (0.26) 0.74 (0.89) 0.56 (0.49) 0.46 (0.38) Qha 0.63 (0.59) 0.65 (0.75) 0.38 (0.27) 0.27 (0.27) ¼ Pt Pv ; ¼ Ppv Pt Rl Rpv Qpv 0.48 (0.62) 0.42 (0.59) 0.4 (0.6) 0.44 (0.6) Q ha þQ pv Q pv

The OO total resistance is then computed: ROO ¼ 1=ð1=Req 1=Rha 1=RDOÞ. In the RCR models, the proximal resistance is assumed to carry 10% and 5% of the total 3.1. Pre-resection stage resistance, for the DO and OO blocks respectively, within ranges in Raines et al. (1974), Vignon-Clementel et al. (2010), and LaDisa et al. (2011). Capacitances are fixed (same for all pigs). Other organs capacitance: COO ¼ 7:36 The tuning procedure described above (2.3), based on pre- 10 4 cm5=dyn (within ranges in Segers et al., 2000); digestive organs capacitance resection measurements, gives good agreement between pre- (manually tuned): CDO ¼ 4 10 4 cm5=dyn and liver tissue capacitance per mass: resection simulated and measured values. Fig. 2 displays the ¼ : 5 5= = Cl 1 5 10 cm dyn g(Bennett and Rothe, 1981). simulated pressures and flows of interest against the measured Heart and lung parameters: The amplitude and baseline of heart contraction functions are tuned, for each pig, to obtain arterial and venous average pressures ones in logarithmic scale. The dots in Fig. 2 are nicely aligned along measured before resection. The heart contraction times and cardiac cycle are based the curve y¼x illustrating the good match between the results and on the measured heart rate (Table 2). Lung parameters are based on Blanco and measurements, for all pigs. Parameters are tuned for each animal, Feijóo (2013): RL ¼ 53:33 dyn s/cm5; RL ¼ 53:33 dyn s=cm5 and CL ¼0.05 cm5/dyn. p d thus the inter-animal variability is well captured. Standard deviations for measured and simulated variables are: 10.4 mmHg for arterial pressure, 2.4 mmHg for PV pressure, 3.1 mmHg for the 3. Results pressure drop, 0.06 L/min and 0.18 L/min for HA and PV flow rates respectively. Hepatic surgeons are particularly interested in specific pressure and flow values hypothesized to be linked to liver failure. The variables of interest are: arterial pressure, PV pressure, the venous 3.2. Liver partial resection simulation pressure drop Ppv Pv across the liver, PV flow and HA flow. Simulated and measured values are compared for these variables. Impact of liver lobe mass assumptions: The post-resection simulated variables are impacted by the estimation of the liver

Table 2 lobe masses. Thus, the simulations are run, with the four different Durations and times for contraction and relaxation of the different cardiac cham- mass estimations (Table 1), for twelve different pigs and compared bers. These parameters are the same for the left and right hearts. with measurements. Table 3 shows mean and standard deviation of the relative error for the different variables and mass assump- Notation Detail Value in second tions. The arterial pressure and PV ﬂow are almost not impacted

Tcc Cardiac cycle 60 s/heart rate (bpm) by the lobe mass differences (at most 20% difference). The lobe : Tvc Ventricular contraction duration 0 34Tcc mass estimations have a signiﬁcant impact on PV pressure (at : Tvr Ventricular relaxation duration 0 15Tcc most 68% difference), the pressure drop (at most 90% difference), T Atrium contraction duration 0:17T ac cc and HA ﬂow (140% maximum difference). The last mass estimation Tar Atrium relaxation duration 0:17Tcc

tac Time atrium begins to contract 0:8Tcc gives the smallest relative errors (in average), thus this mass : tar Time atrium begins to relax 0 97Tcc assumption is kept for the rest of the simulations. It is indeed the least predictive assumption but gives as expected the best simulation results. Hepatectomy simulation: The simulation results averages for twelve pigs are compared to the measurements (Fig. 3) before and after resection. After liver resection, on average, 45% increase of PV pressure and 98% increase in pressure drop are measured. A small decrease in arterial pressure of 12% is observed. Moreover, a large decrease in HA flow, 74%, and a smaller decrease in PV flow, 30%, are measured. In the model, in average, 66% increase for PV pressure and 110% increase for pressure drop are simulated. The fact that these pressures increase is coherent with the measurements, but these increases are overestimated. The arterial pressure decrease is only 3%. The model underestimates the decrease of PV flow (5%), but captures well the HA flow decrease of 75%. Taking into account changes in blood volume: During surgery, the total volume of blood in the circulation varies. Estimating its loss or gain is complex. However, pressures are strongly linked to the circulating blood volume. Thus, the changes observed in arterial pressure measurements are used to estimate the change in blood volume. The volume added or removed is chosen such that the Fig. 2. Pre-resection measurements vs simulation values in log/log scale, for each simulated post-resection arterial pressure corresponds to the variable (unique color) and for each animal (one dot). Pressures are in mmHg and flow rates are in L/min. (For interpretation of the references to color in this figure measurement. Therefore a decrease of 12% in arterial pressure is caption, the reader is referred to the web version of this paper.) obtained with the model. 206 C. Audebert et al. / Journal of Biomechanics 50 (2017) 202–208

Table 4 p-values for TOST test between the 12 simulated post-resection variables and measurements. The simulations include or not the changes in blood volume.

TOST Before resection After resection test Blood volume constant Blood volume changes

p-value 3.3 1023 0.94 0.027 Result Reject Not reject Reject

the obtained relative error vectors. The results and p-values are given in Table 4. According to the test p-values the null hypothesis is rejected for the pre-resection simulation, with 10% relative error for indifference region. More precisely, the simulated variables and the measurements, before resection, are significantly similar. The null hypothesis is not rejected for post-resection simulation without blood loss changes, with 10% relative error for indifference region. Meaning, there is insufficient evidence to reject the null hypothesis. This result may occur, because the model output and the measurements really differ or because the sample size is too small to conclude. Finally, with a 10% relative error indifference region, post-resection simulation with blood volume changes and measurements are significantly similar. Taking into account blood volume changes knowing the change in arterial pressure, improves model outputs.

4. Discussion and conclusion

Measurements explained by modeling: The behavior of the measured pressures and flows during 75% hepatectomy are analyzed using the model. The observed HA flow decrease corresponds exactly to the increase of the HA tree resistance due to the 75% liver resection, without HABR being needed. At leading order, the liver arterial system behaves as ΔP ¼ RQ, with ΔP the arterial pressure drop, which remains almost constant, R the liver resistance and Q is the HA flow. The 75% liver resection induces the HA tree resistance increase of 75%, thus explaining the decrease of HA flow. The arterial pressure is not impacted by the liver resection because the HA and liver resistances are small compared to the rest of the systemic circulation. However in average it decreases by 12% in the measurements. This decrease is a consequence of the blood loss, as proven with the model. PV flow measurements, in average, decrease by 30%. The main decrease is due to blood (volume) loss. This is reinforced by the fact that animals with larger blood losses have a more important PV flow decrease. Fig. 3. Measurements (full) and simulations (dash) at different states of the sur- However the simulations, without blood loss, show that 75% liver gery: pre-resection, post-resection. Simulations with (dashed green) and without resection accounts for a decrease of portal flow of around 5%. The (dashed red) blood volume changes are represented for the A4 mass assumption (a) measured PV pressure and venous pressure drop increase by 45% pressures and (b) flows. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.) and 98% respectively. The portal pressure increase is expected given the increase of PV tree and liver resistances due to resection. However it is compensated by three mechanisms: interaction with Fig. 3 displays the simulated variables, taking into account the the rest of the circulation which causes PV flow to decrease, the change in blood volume, averaged over twelve pigs. Pressures are lessen increase in PV tree resistance due to dilation and the gen- largely impacted by changes in blood volume. The model agrees eral pressure decrease due to blood loss. Indeed, the simulations well with the measurements for PV pressure and pressure drop, without blood volume change predict a 66% increase of portal fl with increases of 43% and 82% respectively. HA ow is still cor- pressure and adding the blood losses the increase is 43%. If the fl rectly simulated, with a 79% decrease, and PV ow is improved, venous pressure remains constant during surgery, the pressure with 23% decrease, however is still overestimated. drop would increase by 110% as simulated with the model. How- Similarity of measured and simulated populations: The model ever, the measured CV pressure decreases by 33%. Thus, the blood fi ability to reproduce the animal population variability is veri ed losses lead to all venous pressures decrease, and an increase in the with a two one-sided t-test. The A4 mass assumption is used to pressure drop estimated as 82% in the simulations. perform the simulations. The test is performed with and without Sensitivity and variability: The liver lobe mass estimations blood volume changes. For each variable of interest and each impact the post-resection simulated variables due to the fact that simulation the relative error is computed. The test is performed on the liver lobe resistances and capacitances depend on lobe masses. C. Audebert et al. / Journal of Biomechanics 50 (2017) 202–208 207

The mostly impacted variables are PV pressure, the pressure drop Acknowledgments and HA flow. This is expected since arterial pressure and portal flow, are strongly dependent on heart and digestive organs para- This material is based upon work supported by the French meters respectively. A better estimation of the liver lobe masses National Agency for Research ANR-13-TECS-0006 iFLOW. The may improve post-resection simulated variables. authors are grateful to the INRA Plateforme CIRE (Nouzilly, France) Moreover, several events happen during the surgery due to staff for their technical assistance in surgeries and imaging, and to surgical acts, anesthetists interventions etc. Here, the model Mylène Wartenberg for assistance in taking measurements. demonstrates that taking into account the change in blood volume improves the simulated post-resection prediction, knowing e.g. the change in arterial pressure measurements. In terms of varia- Appendix A. Supplementary data bility, the simulated results are in good agreement with the measurements, both for pre-resection and post-resection with Supplementary data associated with this article can be found in blood volume changes, according to the TOST tests. The tests also the online version at http://dx.doi.org/S0021-9290(16)31214-3. show that taking into account blood loss significantly improved the model outputs agreement with the measurements. Pressure and flow changes due to hepatectomy without any volume change are also simulated with the model. Theses results References may represent the state after the surgery, once the blood volume is back to the pre-resection volume. Under this assumption, the portal Allard, M.A., Adam, R., Bucur, P.O., Termos, S., Cunha, A.S., Bismuth, H., Castaing, D., Vibert, E., 2013. Posthepatectomy portal vein pressure predicts liver failure and pressure and pressure drop, important for liver failure (Allard et al., mortality after major liver resection on noncirrhotic liver. Ann. Surg. 258, 2013), may be underestimated with intraoperative measurements. 822–830. Thus, for a 75% liver ablation, the model predicts an increase, in the Audebert, C., Bucur, P., Bekheit, M., Vibert, E., Vignon-Clementel, I.E., Gerbeau, J.F., 2016. Kinetic scheme for arterial and venous blood flow, and application to following post-operative days of 110% instead of 82% for the pres- partial hepatectomy modeling. Comput. Methods Appl. Mech. Eng. http://dx. sure drop and 66% instead of 43% for portal pressure. doi.org/10.1016/j.cma.2016.07.009. fl The simulation results for each animal are presented in Bennett, T.D., Rothe, C.F., 1981. Hepatic capacitance responses to changes in ow and hepatic venous pressure in dogs. Am. J. Physiol.-Heart Circul. Physiol. 240, Table 4. Among animals, various post-resection behaviors are H18–H28. obtained with the simulations. For example, without change in Bezy-Wendling, J., Kretowski, M., Mescam, M., Jurczuk, K., Eliat, P.A., 2007. Simu- fl lation of hepatocellular carcinoma in mri by combined macrovascular and blood volume, for animal iF03 PV ow is almost unchanged pharmacokinetic models. In: 4th IEEE International Symposium on Biomedical (decrease by 1.7 %) compared to iF12 for which it decreases by Imaging: From Nano to Macro, April 12-15, 2007 Crystal Gateway Marriott 13% (Table 4). The simulated blood volume change is adapted on Arlington, Virginia, USA, pp. 1272–1275. Blanco, P.J., Feijóo, R.A., 2013. A dimensionally-heterogeneous closed-loop model an animal basis to the arterial pressure change. For animal iF02 a for the cardiovascular system and its applications. Med Eng Phys 35, 652–667. loss of 200 ml is simulated compared to animal iF08 for which a http://dx.doi.org/10.1016/j.medengphy.2012.07.011. fi volume infusion of 200 ml is modeled (Table 4). Therefore, the Bon glio, A., Leungchavaphongse, K., Repetto, R., Siggers, J.H., 2010. Mathematical modeling of the circulation in the liver lobule. J. Biomech. Eng. 132, 111011. model is able to simulate the different hemodynamics states that Childress, E.M., Kleinstreuer, C., 2014. Impact of fluid–structure interaction on occur post-resection. direct tumor-targeting in a representative hepatic artery system. Ann. Biomed. – Future work: In summary, this work presents the first dynamics Eng. 42, 461 474. Chu, T.M., Reddy, N.P., 1992. A lumped parameter mathematical model of the model of liver ablation. Its validation based on pig data is promis- splanchnic circulation. J. Biomech. Eng. 114, 222–226. ing; the model is able to capture and explain the main features of Court, F., Wemyss-Holden, S., Morrison, C., Teague, B., Laws, P., Kew, J., Dennison, A., Maddern, G., 2003. Segmental nature of the porcine liver and its potential as a hemodynamics changes due to the surgery as well as its variability model for experimental partial hepatectomy. Br. J. Surg. 90, 440–444. among pigs, and may give insights about other states (e.g. day post- Debbaut, C., De Wilde, D., Casteleyn, C., Cornillie, P., Van Loo, D., Van Hoorebeke, L., surgery) when measurements are difficult to take. HABR does not Monbaliu, D., Fan, Y.D., Segers, P., 2012. Modeling the impact of partial hepatectomy on the hepatic hemodynamics using a rat model. IEEE Trans. Biomed. seem needed to explain the data. Future work will include the Eng. 59, 3293–3303. adaptation of the liver model to human liver anatomy. Moreover, in Debbaut, C., Monbaliu, D., Segers, P., 2014a. Validation and calibration of an elec- this work, several pressure and flow measurements were available trical analog model of human liver perfusion based on hypothermic machine perfusion experiments. Int. J. Artif. Organs 37, 486–498. at different stages of the surgery. The pre-resection measurements Debbaut, C., Vierendeels, J., Siggers, J.H., Repetto, R., Monbaliu, D., Segers, P., 2014b. were used for parameter tuning. In patient surgeries less mea- A 3d porous media liver lobule model: the importance of vascular septa and anisotropic permeability for homogeneous perfusion. Comput. Methods Bio- surements are available. Furthermore, the model and measure- mech. Biomed. Eng. 17, 1295–1310. ments in this work are for a healthy (pig) liver. However the liver of Ho, H., Sorrell, K., Bartlett, A., Hunter, P., 2012. Blood flow simulation for the liver the patients treated with partial hepatectomy is generally not after a virtual right lobe hepatectomy. In: MICCAI 2012, the 15th International Conference on Medical Image Computing and Computer Assisted Intervention, healthy. For example, collateral circulations can appear (Pinzani and 1st to 5th October 2012 in Nice, France, pp. 525–532. Vizzutti, 2005), and the liver resistance and capacitance parameters Iida, T., Yagi, S., Taniguchi, K., Hori, T., Uemoto, S., 2007. Improvement of morpho- can change. Thus, future work is needed to integrate these con- logical changes after 70% hepatectomy with portocaval shunt: preclinical study in porcine model. J. Surg. Res. 143, 238–246. siderations. Despite this, the framework in place can easily be LaDisa, J.F., Alberto Figueroa, C., Vignon-Clementel, I.E., Jin Kim, H., Xiao, N., Ellwein, adapted to include various measurements and different pathologies L.M., Chan, F.P., Feinstein, J.A., Taylor, C.A., 2011. Computational simulations for aortic coarctation: representative results from a sampling of patients. J. Bio- for clinical applications. Finally, the model can be a basis to compute mech. Eng. 133, 091008. http://dx.doi.org/10.1115/1.4004996. how much to control portal flow in order to avoid liver-failure Lantz, B., Foerster, J., Link, D., Holcroft, J., 1981. Regional distribution of cardiac related hemodynamics ranges (Iida et al., 2007; Vasavada et al., output: normal values in man determined by video dilution technique. Am. J. Roentgenol. 137, 903–907. 2014; Sainz-Barriga et al., 2011; Allard et al., 2013). Liang, F., Liu, H., 2005. A closed-loop lumped parameter computational model for human cardiovascular system. JSME Int. J. 48, 484–493. Liang, F., Takagi, S., Himeno, R., Liu, H., 2009. Multi-scale modeling of the human cardiovascular system with applications to aortic valvular and arterial stenoses. Conflict of interest statement Med. Biol. Eng. Comput. 47, 743–755. Lukeš, V., Jiřík, M., Jonášová, A., Rohan, E., Bublík, O., Cimrman, R., 2014. Numerical Simulation of Liver Perfusion: From CT Scans to FE Model. arXiv preprint None declared. arXiv:1412.6412. 208 C. Audebert et al. / Journal of Biomechanics 50 (2017) 202–208

Pedley, T.J., Brook, B.S., Seymour, R.S., 1996. Blood pressure and flow rate in the interpretation of hemodynamic stress governing inflow modulation in liver giraffe jugular vein. Philos Trans R Soc Lond B Biol Sci 351, 855–866. http://dx. transplantation. Liver Transplant. 17, 836–848. doi.org/10.1098/rstb.1996.0080. Schwen, L.O., Schenk, A., Kreutz, C., Timmer, J., Rodríguez, M.M.B., Kuepfer, L., Peeters, G., Debbaut, C., Cornillie, P., De Schryver, T., Monbaliu, D., Laleman, W., Preusser, T., 2015. Representative sinusoids for hepatic four-scale pharmaco- Segers, P., 2015. A multilevel modeling framework to study hepatic perfusion kinetics simulations. PloS One 10, e0133653. characteristics in case of liver cirrhosis. J. Biomech. Eng. 137, 051007. Segers, P., Stergiopulos, N., Schreuder, J.J., Westerhof, B.E., Westerhof, N., 2000. Left Pennati, G., Migliavacca, F., Dubini, G., Pietrabissa, R., de Leval, M.R., 1997. A ventricular wall stress normalization in chronic pressure-overloaded heart: a mathematical model of circulation in the presence of the bidirectional cavo- mathematical model study. Am. J. Physiol. – Heart Circul. Physiol. 279, pulmonary anastomosis in children with a univentricular heart. Med. Eng. Phys. H1120–H1127. 19, 223–234. Segers, P., Stergiopulos, N., Westerhof, N., Wouters, P., Kolh, P., Verdonck, P., 2003. Pinzani, M., Vizzutti, F., 2005. Anatomy and vascular biology of the cells in the Systemic and pulmonary hemodynamics assessed with a lumped-parameter portal circulation. In: Portal Hypertension. Springer, pp. 15–35 (chapter 2). heart-arterial interaction model. J. Eng. Math. 47, 185–199. R Development Core Team, 2008. R: A Language and Environment for Statistical Serban, R., Petra, C., Hindmarsh, A., 2015. User Documentation of ida v2.7.0. URL Computing. R Foundation for Statistical Computing. Vienna, Austria. 〈http:// 〈https://computation.llnl.gov/casc/sundials/documentation/documentation. www.R-project.org〉 ISBN 3-900051-07-0. html〉. Raines, J.K., Jaffrin, M.Y., Shapiro, A.H., 1974. A computer simulation of arterial Shapiro, A.H., 1977. Steady flow in collapsible tubes. J. Biomech. Eng. 99, 126–147. dynamics in the human leg. J. Biomech. 7, 77–91. Suga, H., Sagawa, K., 1974. Instantaneous pressure–volume relationships and their Ricken, T., Werner, D., Holzhutter, H., König, M., Dahmen, U., Dirsch, O., 2015. ratio in the excised, supported canine left ventricle. Circul. Res. 35, 117–126. Modeling function–perfusion behavior in liver lobules including tissue, blood, Swindle, M., Makin, A., Herron, A., Clubb, F., Frazier, K., 2012. Swine as models in glucose, lactate and glycogen by use of a coupled two-scale pde-ode approach. biomedical research and toxicology testing. Vet. Pathol. Online 49, 344–356. Biomech. Model. Mechanobiol. 14, 515–536. Van Der Plaats, A., t Hart, N., Verkerke, G., Leuvenink, H., Verdonck, P., Ploeg, R., Robinson, A.P., Froese, R.E., 2004. Model validation using equivalence tests. Ecol. Rakhorst, G., 2004. Numerical simulation of the hepatic circulation. Int. J. Artif. – Model. 176, 349–358. Organs 27, 222 230. Robinson, A.P., Duursma, R.A., Marshall, J.D., 2005. A regression-based equivalence Vasavada, B.B., Chen, C.L., Zakaria, M., 2014. Portal flow is the main predictor of test for model validation: shifting the burden of proof. Tree Physiol. 25, early graft dysfunction regardless of the grwr status in living donor liver — 903–913. transplantation a retrospective analysis of 134 patients. Int. J. Surg. 12, – Robinson, A., 2016. Package ‘Equivalence’. Technical Report. 〈https://cran.r-project. 177 180. fl org/web/packages/equivalence/equivalence.pdf〉. Vignon-Clementel, I.E., Figueroa, C.A., Jansen, K.E., Taylor, C.A., 2010. Out ow fl Rypins, E.B., Rosenberg, K.M., Sarfeh, I.J., Houck, J., Conroy, R.M., Milne, N., 1987. boundary conditions for 3D simulations of non-periodic blood ow and pres- fi Computer analysis of portal hemodynamics after small-diameter portacaval h- sure elds in deformable arteries. Comput. Methods Biomech. Biomed. Eng. 13, – 〈 〉〈 grafts: the theoretical basis for partial shunting. J. Surg. Res. 42, 354–361. 625 640. http://dx.doi.org/10.1080/10255840903413565 Go to ISI http:// 〉 Sainz-Barriga, M., Scudeller, L., Costa, M.G., de Hemptinne, B., Troisi, R.I., 2011. Lack WOS:000281851500012 . of a correlation between portal vein flow and pressure: toward a shared HPB

Kinetics of hepatic volume evolution and architectural changes after major resection in a porcine model

For Peer Review Only Journal: HPB

Manuscript ID HPB-2017-0611

Article Type: Original Article

Date Submitted by the Author: 22-Nov-2017

Complete List of Authors: Bekheit, Mohamed; El Kabbary General Hospital, Bucur, Petru; AP-HP Paul Brousse Hospital, Centre Hépato-Biliaire, Villejuif, France, Liver Surgery Audebert, Chloe; inria paris Standley, Elodie; Centre Hospitalier Regional Universitaire de Tours Vignon-Clementel, Irene; inria paris Vibert, Eric; Hopital Paul Brousse, Liver Transplantation and HPB Surgery

Keywords: Liver, Basic science < Liver

http://mc.manuscriptcentral.com/hpb Page 1 of 36 HPB

1 2 3 Kinetics of hepatic volume evolution and architectural changes after major resection in a 4 5 6 porcine model. 7 8 9 Mohamed Bekheit 1,2,3,4 Petru O. Bucur 1,2, Chloe Audebert 5,7, Elodie 10 11 MiquelestorenaStandley 6, Irene E. VignonClementel 5,7, Eric Vibert 1,2. 12 13 14 15 16 For Peer Review Only 17 Authors` affiliations 18 19 20 1. Centre HépatoBiliaire, 12 av. Paul Vaillant Couturier, APHP, Hôpital Paul 21 22 Brousse, 94800, Villejuif, France 23 24 25 2. Inserm Unité 1193, 12 av. Paul Vaillant Couturier, 94800, Villejuif, France 26 27 28 3. Ecole doctorale Innovation Therapeutique, Universite PaisSud, CHÂTENAY 29 30 MALABRY Cedex, France 31 32 33 4. Institute of Medical Sciences, University of Aberdeen, Aberdeen, UK 34 35 36 5. Centre de recherche Inria de Paris, Paris, France 37 38 39 6. CHRU Tours, Laboratoire D’anatomie et Cytologie Patholoiques, Tours, France 40 41 42 7. Sorbonne Universités, UPMC Univ. Paris 6, Laboratoire JacquesLouis Lions, 43 44 Paris, France 45 46 47 48 49 50 Corresponding author: 51 52 53 Eric Vibert, MD, PhD 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 2 of 36

1 2 3 12 Avenue Paul Vaillant Couturier 4 5 6 94804 Villejuif Cedex 7 8 9 Tel: 00 33 1 45 59 30 00 10 11 12 Fax: 00 33 1 45 59 38 57 13 14 15 EMail: [email protected] 16 For Peer Review Only 17 18 Authors` contribution: 19 20 21 Study concept and design: Mohamed Bekheit, Eric Vibert, Irene VignonClementel 22 23 24 Acquisition of data: Mohamed Bekheit, Petru Bucur, Elodie MiquelestorenaStandley, 25 26 Chloe Audebert. 27 28 29 Analysis and interpretation of data: Mohamed Bekheit, Chloe Audebert, Petru Bucur, 30 31 32 Elodie MiquelestorenaStandley, Irene VignonClementel 33 34 35 Drafting of the manuscript: Mohamed Bekheit, Chloe Audebert 36 37 38 Critical revision: Eric Vibert, Irene VignonClementel 39 40 Final approval: Eric Vibert, Irene VignonClementel 41 42 43 Source of Funding: This study was partially funded by the “Agence de la Biomedicine” 44 45 46 through its program of Research (AOR January 2009). Eric Vibert, Petru O. Bucur, Chloe 47 48 Audebert, Irene E. VignonClementel, and Mohamed Bekheit acknowledge funding by 49 50 the project (Agence Nationale de Recherche) ANR13TECS0006 (IFlow). 51 52 53 Keywords: Major liver resection, regeneration, kinetics, 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 3 of 36 HPB

1 2 3 Abstract: 4 5 6 Background: 7 8 9 This study aims to explore the rate of hepatic regeneration of the porcine liver following 10 11 major resection, highlighting estimates of the microarchitectural changes. 12 13 14 Methods: 15 16 For Peer Review Only 17 Nineteen Large White pigs had 75% resection with serial measurements of the hepatic 18 19 20 volume, density, blood flow, and architectural changes. 21 22 23 Results: 24 25 26 The growth rate, initially of 45% per day, then rapidly decreased and was accompanied 27 28 by a similar pattern of hepatic fat deposition. The architectural changes showed a 29 30 significant increase in the Ki67 expression (p<0.0001) in the days following resection 31 32 33 with a peak on the second day and nearly normalized at day7. The expression of CD31 34 35 increased significantly on the second and third days compared to the preresection 36 37 samples (p=0.03). Hepatic artery flow per liver volume remained at baseline ranges 38 39 during regeneration. Portal flow per liver volume increased after liver resection 40 41 42 (p<0.001), was still elevated at day1, then decreased. 43 44 45 Correlations were significantly negative between the hepatic volume increase at day3, 46 47 and the hepatic oxygen consumption and the net lactate production at the end of the 48 49 procedure (r=0.82, p=0.01, and r=0.70, p=0.03). 50 51 52 Conclusion: 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 4 of 36

1 2 3 The regeneration is highest in the first days, with a liver to body weight ratio that is back 4 5 6 to 50% of the preoperative value after three days to close to 100% volume regain at 1015 7 8 days. 9 10 11 12 13 14 15 16 For Peer Review Only 17 Background 18 19 20 The remarkable ability of the liver to regenerate has led to a significant advancement in 21 22 hepatic surgery. Currently, liver resection has been taken to an extreme state where 23 24 surgeons base their planning on the minimum possible residual volume to cope with the 25 26 27 significant disease burden (1). 28 29 30 Successful regeneration is a fundamental requirement for recovery following liver 31 32 resection, and the role of the sinusoidal endothelial cells is pivotal in this process (2). The 33 34 immediate increase in the portal pressure stimulates hepatocyte proliferation through a 35 36 cascade of intermediate factors mediated via the hepatocyte growth factor and the 37 38 39 vascular endothelial growth factor (3). 40 41 42 The equilibrium between the stimulation and inhibition of regeneration and subsequent 43 44 development of hepatic failure is not fully elucidated. It seems that there is an 45 46 exponential relationship between the sinusoidal cell stimulation through the shear stress 47 48 49 and the liver regeneration where at a certain threshold, excessive regeneration can 50 51 jeopardize liver function (4). 52 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 5 of 36 HPB

1 2 3 Despite that the residual hepatic volume is a recognized determinant factor for the 4 5 6 postoperative outcome (5), it is not the only one. The quality of the regenerating liver 7 8 parenchyma – along with the rate (4) of regeneration – plays a vital role in shaping the 9 10 operative strategy and the postoperative outcome (6). 11 12 13 The initial inductive angiogenesis phase orchestrates the regenerating parenchyma (2), 14 15 and the attainment of a functioning liver following massive resection is closely related to 16 For Peer Review Only 17 18 the architectural quality. The rapidly regenerating hepatocyte cluster around the sinusoids 19 20 causes portal hypertension and impairment of liver function during the regeneration 21 22 process (4). The functional and volume disparity noted after ALPPS (7), might be 23 24 25 attributed to an architectural disorganization. 26 27 28 Over the years, the porcine model has gained wide acceptance as a transitional model for 29 30 multiscale surgical research to provide a substantial knowledge that could be 31 32 implemented directly to humans. Previous studies addressed the porcine hepatic anatomy 33 34 (8) and surgical techniques for liver resection (9). 35 36 37 However, porcine recovery models have not comprehensively described the rate of the 38 39 40 hepatic volume increase following major resection along with the accompanying 41 42 architectural changes, although some studies attempted this tracing in a nonrecovery 43 44 model by direct weighing (10). The crucial knowledge of these changes should help to 45 46 47 optimize the modulation of hepatic hemodynamics to prevent early architectural damage 48 49 and enhance liver regeneration (11). Architectural changes during this initial period are 50 51 believed to influence the outcome of hepatic resection and the development of post 52 53 54 hepatectomy liver failure, which this study aims to highlight in the context of volume 55 56 gain. 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 6 of 36

1 2 3 Aim: 4 5 6 to describe the kinetics of regeneration after major resection of porcine liver and to 7 8 9 highlight the concomitant architectural changes. 10 11 12 Methods 13 14 Ethical approval: 15 16 For Peer Review Only 17 The study was approved by the regional committee of ethics in animal research, and by 18 19 20 the ministry of higher education and scientific research and ministry of agriculture and 21 22 fishing, according to the European Union directives. 23 24 25 Experimental study design: 26 27 28 The study aimed to follow the hepatic volume evolution after major liver resection in a 29 30 porcine model. Three protocols were developed to follow different timelines after liver 31 32 33 resection to minimize the suffering of the animals (summarized in Supplementary figure 34 35 2). Group 1 had a sevenday protocol with sacrifice seven days after liver resection. 36 37 Group 2 had a threeday protocol with relaparotomy at day1 and sacrificed at day3. 38 39 Group 3 had the same initial procedure then relaparotomy on the second postoperative 40 41 42 day and then animals were used for a different study. CT imaging was carried out 56 43 44 days before surgery and by the end of the first hour following surgery, then following 45 46 surgery (groups 1,2&3), the first (group 2), third (groups 1&2), seventh postoperative 47 48 rd 49 days (group 1), and the 23 day (group 3). 50 51 52 Animals: 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 7 of 36 HPB

1 2 3 Large White porcine animals were housed in individual pens with controlled temperature 4 5 6 and natural lighting. There was a period of conditioning before surgery of 7 days. 7 8 9 Setting: 10 11 The study was conducted at the CIRE platform, INRA Centre Val de Loire, Nouzilly, 12 13 14 France. 15 16 For Peer Review Only 17 Preoperative preparation: 18 19 20 On days of surgery, animals were given in their individualized cages 30 mg/kg ketamine 21 22 (Ketamin, Panpharma, France) and 0.03 mg/kg acepromazine (Calmivet, Vetoquinol, 23 24 France). 25 26 27 Anesthesia: 28 29 30 Each pig received 100 mg of xylazine 2% (Rompun, Bayer Healthcare) with 750 mg 31 32 33 ketamine for anesthesia induction followed by tracheal intubation (67 mm in size, 34 35 Portex, France). Subsequently, inhalational anesthesia was started using a 60% FiO2 with 36 37 2% Isoflurane (Isoflurane, Belmont, France) in assisted ventilation. The ventilator was set 38 39 up on volume control mode delivering 350400 ml at a rate of 1720 cycles/minute. 40 41 42 Animals received Fentanyl (100µg/2ml, Fentanyl Janssen) intramuscularly and 43 44 45 Crystalloid fluids at a rate of 2ml/kg/hfasting. Subsequently, Pancuronium bromide 46 47 (Pavulon, ScheringPlough), at a rate of 0.3mg/kg/h and fentanyl (Fentanyl Janssen 48 49 100µg/2ml), at a rate of 5g/kg/h were continuously perfused intravenously. Also, 500 50 51 52 1000ml, which was increased as required, were given during surgery. At the end of the 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 8 of 36

1 2 3 operation, the wound was infiltrated with ropivacaïne 150mg (Naropeine, AstraZeneca, 4 5 6 7.5mg/ml). 7 8 9 Cefotaxime 1g (Cefotaxime, Mylan) and gentamicine 80mg (Gentalline, Schering 10 11 Plough) were given intramuscularly once a day for 5 days. In addition, animals received 12 13 Pantoprazole 40mg/i.v (Inipomp, Nycomed) and enoxaparine 0.2 ml/S.C (Lovenox, 14 15 Sanofi Aventis) and 0.5mg/kg/b.i.d Nalbuphine (Nalbuphine Serb, 20mg/2ml). 16 For Peer Review Only 17 18 The same perioperative protocol was repeated at the time of sacrifice. Blood samples 19 20 rd 21 were collected before and after liver resection as well as on the 3 (for all groups) and 22 23 5th, and the 7th postoperative days (for groups 1 and 3). 24 25 26 Surgery: 27 28 29 Hepatic resection involved surgical removal of the left three lobes of the porcine liver 30 31 (left lateral, left medial, and the right medial lobes) (Supplementary Figure 1a). Resection 32 33 34 was performed under continuous hemodynamics monitoring as described below, and a 35 36 75% resection was executed using a Rummel tourniquet (supplementary Figure 1 bf). 37 38 The detailed description of the surgical technique is given in a previous study (9). 39 40 Surgery was based on the anatomical liver descriptions previously given (8,12). In groups 41 42 43 1 and 3, animals underwent a laparotomy on postoperative day one and two respectively. 44 45 46 Sacrifice: 47 48 rd th 49 On the 3 (group 2) or 7 (group 1) postoperative day, surviving animals were sacrificed 50 51 following a similar protocol to that on the day of surgery. Group 3 was used for a 52 53 different study where sacrifice took place on the 27th day after resection. The remnant 54 55 56 liver was weighed after euthanasia. 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 9 of 36 HPB

1 2 3 Data acquisition and analysis: 4 5 6 Imaging acquisition and postprocessing: 7 8 9 Abdominal CT scans, with an 80 ml (2 ml/kg) of iodinated contrast (Omnipaque TM, GE 10 11 Healthcare, Carrigtohil, Irland) injected through an intravenous catheter at a rate of 4 12 13 14 ml/s, were performed using a Somatom (Definition AS, Siemens, Forchheim, Germany). 15 16 The CT imageFor acquisition Peer was performed Review in 4 phases. VolumeOnly analysis was performed 17 18 using the Myrian® XPLiver 1.14.1 software (Intrasense, Montpellier, France). 19 20 21 Segmentation of the portal venous branches as well as the hepatic venous branches was 22 23 first performed. Subsequently, the liver parenchyma was anatomically segmented based 24 25 on the vascular trees. Liver volume/Body weight ratio (LVBWR) was calculated from the 26 27 28 following equation: liver volume in ml/body weight in kg. 29 30 31 CT scan estimation of the parenchymal density: 32 33 34 In an axial noncontrast CT scan, a region of interest (ROI) was selected in a 35 36 parenchymal zone away from major vessels. The Hounsfield Unit (HU) was measured 37 38 three times in three consecutive axial cuts – provided that it did not come across a major 39 40 vessel or outside the parenchyma – and the mean of these three measurements was 41 42 43 calculated and used for each animal. The whole procedure was repeated for each animal 44 45 on each temporal scan. The image analysis for this purpose was performed in ImageJ 46 47 (https://imagej.nih.gov/ij/). 48 49 50 Hemodynamic measurements: 51 52 53 The portal vein and the hepatic artery flow rates were continuously measured during the 54 55 56 surgery (n=14) as well as on postoperative day one (n=4), three (n=4) and seven (n=6), 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 10 of 36

1 2 3 (with Emka TECHNOLOGIES itf16USB usbAMP amplifier and iox2 acquisition 4 5 6 software). Flows were measured with a Transonic flowmeter (three channel perivascular 7 8 flowmeter T403PPP). The probe diameter for hepatic artery was 4mm or 6mm, and 9 10 10mm or 12mm for the portal vein. The detailed procedure for the hemodynamics 11 12 measurements was previously described in (10). The average of twenty seconds of a 13 14 15 stable signal was computed from the continuous measurements. 16 For Peer Review Only 17 18 Histological analysis: 19 20 21 Ki67 immunohistochemistry was used to assess proliferation index of the proliferating 22 23 cells according to Hammad et al. (13). A similar protocol – but samples were incubated 24 25 with CD31 antibody instead – was applied for the analysis of CD31 26 27 28 immunohistochemistry to examine the sinusoidal abundance in the different temporal 29 30 samples. Quantification of the Ki67 followed the same protocol as in Hammad (13), 31 32 whereas quantification of the CD31 expression was performed following a strategy 33 34 adapted from Wang (14). The intensity of the staining was given a score of 0, 1, 2, 3, 35 36 37 reflecting the intensity of staining into absent, weak, moderate, or strong intensity, 38 39 respectively. 40 41 Pathological scoring for the microarchitectural changes: 42 43 44 Pathological scores were given to specimens in standard hematoxylineosin and Masson 45 46 47 trichrome stains based on criteria adapted from Demetris et al. (15). Five criteria were 48 49 formulated: A= presence of inflammatory infiltrate and necrosis [subitems: neutrophil 50 51 infiltrate, lymphocytes infiltrate, and hepatocellular necrosis], B=sinusoidal dilatation and 52 53 54 congestion, C= ductular proliferation, D= steatosis [subitems: 55 56 macrovesicular/ballooning, microvesicular], and E=regeneration [subitems: pseudo 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 11 of 36 HPB

1 2 3 nodular, mitotic activity, and acinar formation]. Each item is graded from 0=no change, 4 5 6 1=mild, 2=moderate, to 3=severe, in samples taken from deep liver tissue for a total score 7 8 between 0 and 30 at each time point. Experienced, but blinded on the temporal factor of 9 10 the specimens, pathologist performed all pathological analyses. 11 12 13 Biochemistry: 14 15 16 The right internalFor jugular Peer vein and Review carotid artery wereOnly cannulated (8 and 5 Fr 17 18 respectively). 30mL of venous blood were sent to the central laboratory unit in universal 19 20 21 tubes for analysis. Blood samples were collected by venipuncture 20 minutes after the 22 23 induction of anesthesia – except for the one taken at the end of surgery – and after an 24 25 overnight fast. Samples for hematological analysis were collected in EDTA tube. For 26 27 28 chemistry analysis, samples were collected in fluoride tubes and buffered sodium citrate. 29 30 Serum was mixed with 0.8mg aprotinin, centrifuged at 2,000xg at 4 °C for 15 minutes 31 32 within 30 min after a puncture and incubated in an ice container till delivery. Delivery 33 34 time to the laboratory was within 45 minutes following centrifugation. 35 36 37 After midline abdominal incision, sequential blood samples (1 ml each) were taken the 38 39 40 right carotid artery, the right suprahepatic vein, and the portal vein from for gas analysis. 41 42 43 Hepatic oxygen consumption and net lactate production were calculated as 44 45 described elsewhere (16) using the following equations: 46 47 48 1 [2//100] = + − [Equation 1] 49 50 Hb 51 where = 10 1.34 2 + 0.003 2 , is the hemoglobin (in 52 53 g/dl), SiO2 is the O2 saturation, PiO2 is O2 pressure in plasma (in mmHg), Fi is 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 12 of 36

1 2 3 flow per 100g of liver weight (L/min/100gLW), and i denotes hepatic artery, 4 5 6 portal vein and hepatic vein blood. 7 8 2 [//100] = + − [Equation 9 10 2], where are the measured lactate concentrations. 11 12 13 14 A mathematical model for liver regeneration was used based on the following ordinary 15 16 differential equationFor to simulatePeer the Reviewevolution of the liver Only weight to body weight ratio 17 18 after liver resection: 19 20 21 = − [Equation. 3] 22 23 24 25 where is liver weight to body weight ratio, is the preoperative ratio and is the 26 27 regeneration characteristic time parameter. By solving the equation, the liver weight to 28 29 body weight ratio can be expressed as 30 31 32 = − exp −t/τ + [Equation 4] 33 34 35 with the postoperative liver to body weight ratio (corresponding to the time t=0). The 36 37 38 parameters and were fixed with the average of the data. The average and standard 39 40 deviation of the regeneration characteristic time in the population were estimated with 41 42 SAEM algorithm from Monolix software developed by Lixoft (17,18). Animals from all 43 44 45 the groups were used to estimate these population parameters. 46 47 48 Statistical analysis: 49 50 51 After normality testing, a summary of data was represented in mean ± standard deviation, 52 53 median, and range or percentages according to the variable type. 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 13 of 36 HPB

1 2 3 Appropriate statistical tests were selected, and the significance threshold was set at a p 4 5 6 value of 0.05. MedCalc Statistical Software version 14.8.1 (MedCalc Software bvba, 7 8 Ostend, Belgium), R version 3.3.0 (The R Foundation for Statistical Computing Vienna, 9 10 Austria. ISBN 3900051070, URL http://www.Rproject.org) and SPSS V.22 11 12 (IBM®SPSS®, Chicago, USA) were used for the analysis. 13 14 15 16 For Peer Review Only 17 18 Results 19 20 21 Nineteen Large White pigs (group 1: 11, group 2: 3, and group 3: 5) with an age of 2.5 22 23 24 +/ 0.8, 2.5 +/ 0.3, and 2.7 +/ 0.4 months (p>0.05) and a weight of 35.3+/5, 34.1 +/ 25 26 3.3, and 36.7 +/ 2.8 kg (p>0.05), respectively were included in this study. The mean 27 28 operative time was 4.6 ±0.9 (95% CI=45) hours. The mean time for hepatectomy was 29 30 31 1.23±0.76 (95% CI= .781.5) hours. The mean initial liver volume was 971.9ml and 32 33 95%CI=8801062ml. Following resection, the residual liver volume was 236.8ml and 34 35 95%CI=198–265ml. At day1, day3, day7, and, day23; the liver volume was 407ml 36 37 95%CI=173–626ml, 587ml 95%CI=504669ml, 631.5ml 95%CI=521–741ml, and 38 39 40 1219.5ml and 95%CI=985.91453 ml; respectively (figure 1 ae). The liver body weight 41 42 ratio followed the same pattern of change as the liver volume alone, and this indicates 43 44 that the changes in the liver volume were responsible for the substantial change in the 45 46 47 ratio. The LVBWR was 28.2 and 95% CI=26.3 30.2, 7.3 and 95% CI=6.7 8, 11.7 and 48 49 95% CI=8.514.9, 16.2 and 95% CI=14.717.8, 17.4 and 95% CI=15.619, and 30.2 and 50 51 95% CI=22.937.5 for before, after, day1, day3, day7, and day23. The mean HU at 52 53 54 day1 was 40±1.9 U, day2= 32±0.7 U, day3= 48±1.4, and day7=53±0.5 U (p<0.001) 55 56 (figure 1f). 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 14 of 36

1 2 3 After liver resection, portal vein flow decreased (p=0.02) but with a significant increase 4 5 6 of portal vein flow per liver volume (p<0.001). Then, the latter gradually decreased. On 7 8 postoperative day7 portal flow per liver volume was 54% larger than baseline (p=0.02). 9 10 Hepatic artery flow decreased after liver resection (p<0.001) and remained low. 11 12 However, hepatic artery flow per liver volume remained similar to baseline with a trend 13 14 15 to be lower at postoperative day3 by 41% and at day7 by 36% (figure 2). 16 For Peer Review Only 17 18 There was an initial significant increase (i.e., worsening) in the histological scoring for 19 20 the microarchitectural changes induced by resection towards the 2nd postoperative day, 21 22 extending over the 3rd postoperative day (figure 3af). These changes faded towards the 23 24 th 25 7 postoperative day. This difference was significant (p<0.001) except between samples 26 27 taken on the seventh day compared to those taken before surgery (p=0.39) (figure 3g). 28 29 30 There was a significant increase in the Ki67 expression (p<0.0001) in the days following 31 32 resection, which then peaked on the second postoperative period and nearly normalized at 33 34 day 7. The mean % of Ki67 positive cells in preoperative samples was 4.7±4.6 %, 35 36 37 5.7±5.3% in afterresection samples, 15.3± 23.2% in day1, 52.9±17.9% in day2, 38 39 42.4±23.7% in the 3rd postoperative days, and 7.8±7.2 % in the 7th postoperative day. The 40 41 expression of CD31 in the analyzed samples increased significantly on the second and 42 43 44 third days compared to the one prior to surgery specimens (p=0.03) (figure 4 a,b). 45 46 47 The estimated regeneration characteristic time was 6.4 +/ 1.4 days; assuming that the 48 49 liver to body weight ratio follows the dynamics described by eq. 3. Figure 5 illustrates the 50 51 mathematical model performance as opposed to the liver weight to body weight ratio 52 53 54 measurements over postoperative days. The regeneration was faster in the first days with 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 15 of 36 HPB

1 2 3 a maximum rate of 43% per day. According to the model, after 1015 days the liver to 4 5 6 body weight ratio was back to baseline values. 7 8 9 A summary of the biochemical parameters and their values measured before and after 10 11 resection is presented in Table 1. Correlation coefficient and twotail significance 12 13 between the hepatic volume increase at day3, and the hepatic oxygen consumption 14 15 before resection and the net lactate production at the end of the procedure were r=0.82 16 For Peer Review Only 17 18 (p=0.01) and r=0.70 (p=0.03) respectively. No other variable – among the studied – 19 20 showed a significant correlation. 21 22 23 24 25 26 Discussion 27 28 29 This study documents the speed of liver regeneration in this porcine model following 30 31 75% hepatic resection alongside the architectural changes in the early period of 32 33 34 regeneration. The estimated volume gains in the early postoperative days were almost 35 rd 36 double on the first postoperative day trending towards a plateau between the 3 and the 37 38 7th postoperative days. The remnant liver regains the preresection volume in almost two 39 40 weeks. The highlights of the architectural kinetics in the early phases of porcine hepatic 41 42 43 regeneration is a fundamental knowledge that would enable the basic understanding of 44 45 the disparity of the functional and volume gains following major resection (7). Few 46 47 studies have reported the early volume changes in a porcine model but not as early as day 48 49 50 1 (19). The authors used ultrasound scan for volume estimation; which is currently not 51 52 the standard practice. The molecular basis of these changes (20) implies that further 53 54 analysis was essential. 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 16 of 36

1 2 3 Redistribution of the portal flow through the remnant liver is responsible for a shear force 4 5 6 that subsequently contributes to liver regeneration (3). In our study, cell proliferation 7 8 estimated with Ki67 expression peaked at day2 and day3, following an increase of 9 10 portal vein flow per liver volume observed after resection and at day1. The amount of 11 12 portal flow diversion to the remnant liver was found to positively correlate with the 13 14 15 degree of hypertrophy and the mitotic index following portal vein ligation in a small 16 For Peer Review Only 17 animal model (21). 18 19 20 The absolute portal flow decreased after resection mainly due to blood loss and the 21 22 removal of the blood contained in the resected liver (22) , while the portal flow per unit 23 24 25 of liver volume increased by 275%. After that, there was a reduction of the portal flow 26 27 per unit of liver volume over time, but the latter remained higher than at baseline. On the 28 29 other hand, the absolute hepatic artery flow decreased, partially in response to a relative 30 31 32 increase in the hepatic resistance to flow after resection and partially due to blood loss 33 34 (22). Nonetheless, the hepatic artery flow per unit of liver volume remained at baseline 35 36 ranges, suggesting that regeneration is mostly triggered by the portal flow per volume 37 38 change. 39 40 41 Previous reports demonstrated that the cellular regeneration precedes vascular 42 43 44 regeneration during liver regeneration following partial hepatectomy (2). In our study, the 45 46 Ki67 peaked on the second day after the resection. The pattern of the CD31 scores 47 48 indicated that the increase in the stained sinusoids in the studied fields over days could be 49 50 51 related to the relative reduction in the edema and steatosis rather than true endothelial cell 52 53 proliferation. 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 17 of 36 HPB

1 2 3 The sinusoidal endothelial cell repopulation was found to occur over the first 48 hours 4 5 6 following rat liver transplantation (23), a situation that could be different in its kinetics 7 8 from the resection setting. The sinusoidal cell proliferation started at 48 hours and peaked 9 10 at 72 hours in a resection rat model (24). This variation indicates that there are different 11 12 kinetics of regeneration among the different models. 13 14 15 A more indepth analysis of these studies shows that the density of sinusoidal spaces 16 For Peer Review Only 17 18 decreases in the first days along with a remarkable increase in the intervening 19 20 parenchymal spaces prior to evident sinusoidal formation after the first 72 hours in rats. 21 22 This is particularly visible in figure 2 provided by Sato (24). The principles of sequential 23 24 25 cellular kinetics in the regenerating liver following partial resection were shown in a 26 27 study by Navarrate (25). Ready hepatocytes with polyploidy divide rapidly from bi 28 29 nucleated to mononucleated cells, and the mononucleated cells increment their ploidy 30 31 32 early with a parallel increase in the cell volume. This also consolidates the inductive and 33 34 proliferative angiogenesis sequence shown by Ding (2). 35 36 37 The capacitance of the liver could contribute up to 20% of the volume change in normal 38 39 circumstances. This mechanism is responsible for accommodating the increased portal 40 41 flow per liver volume after major resection. However, this is the case till a limit, with 42 43 44 regional variation within the parenchyma (26). Oedema and periportal exsanguination are 45 46 seen after partial hepatectomy in a small for size remnants but were also seen where the 47 48 liver functions were preserved (27). 49 50 51 In nonsmall for size remnant model, the development of the fullblown pathologic 52 53 54 picture described by Demetris (15) is not expected. In our model, we observed transient 55 56 changes similar to those changes but less severe. These changes are associated with 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 18 of 36

1 2 3 disruption of the sinusoidal basement membrane and hemorrhagic tracking in the peri 4 5 6 sinusoidal spaces. It probably results from the excess in sinusoidal pressure and failure of 7 8 the capacitance mechanism leading to stagnation and edema that could hypothetically 9 10 contribute initially to the volume increase. The relation between the pathological scores 11 12 and the hepatic functions has been linked to the microarchitecture changes in one of our 13 14 15 previous studies (11). 16 For Peer Review Only 17 18 There is evidence that the hepatic regeneration process is associated with transient 19 20 steatosis (28), which is mainly derived from peripheral adiposity (29). We observed an 21 22 increase in the steatotic index following resection in this model. Furthermore, this is 23 24 25 supported by the pattern of the HU change depicted in the CT scans. This suggests that 26 27 cell replication is not the only factor responsible for such rapid volume increase. 28 29 30 Steatosis has also been demonstrated to contribute to the volume increase in patients 31 32 receiving chemotherapy for colorectal cancer (30). The reduction of the Hounsfield Units 33 34 – i.e. density – corresponds to the transient development of hepatic steatosis. The mean 35 36 37 values of HU measured in our model at the first postoperative day indicate that fat 38 39 deposition occurs very early in the regeneration cycle and disappears over the course of 40 41 the first few days. This further reinforces the evidence of fast regeneration in the porcine 42 43 44 liver following resection. 45 46 47 The hepatic fat accumulation was found to start prior to the first wave of hepatocyte 48 49 proliferation and lasts during this phase and then fades with the subsequent three waves 50 51 of hepatocytes proliferation (31). This is demonstrated as well in this study by the HU 52 53 54 measures on the first day, which precedes the peak of the proliferative Ki67 index – i.e., 55 56 on the second day. 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 19 of 36 HPB

1 2 3 Hepatic steatosis has been linked to the oxidative mitochondrial activity and the hepatic 4 5 6 oxygen consumption (32). We found a significant inverse correlation between the hepatic 7 8 volume increase at the third postoperative day and the hepatic oxygen consumption after 9 10 resection. The negative correlation between the hepatic volume increase and the net 11 12 hepatic lactate production by the hepatic remnant liver at the end of the procedure is well 13 14 15 explained in that context, as previously studied by our group (33). The increased oxygen 16 For Peer Review Only 17 requirement by the remnant liver – as reflected by the increased lactate production – is an 18 19 expression of stress. Interestingly, Ozdogan and colleagues (34) demonstrated that the 20 21 22 supplementation of hyperbaric oxygen enhanced the volume and the quality of the 23 24 regenerating liver. Perhaps hepatic protection via administration of Nacetylcysteine 25 26 (NAC) could be an option to improve the metabolic performance of the liver – as was 27 28 29 found in the study from Torzilli group (35) – and improve hepatic volume gain in critical 30 31 cases. 32 33 34 The main limitation of this study is the lack of numeric quantification of the non 35 36 hepatocytes stained structures, although its inferences are consistent with the evidence 37 38 drawn from the existing literature. Furthermore, although most measurements were 39 40 41 performed on all animals for all days, only a relatively small number of animals were 42 43 studied for postoperative measurements. However, statistical significance was still found 44 45 as reported. 46 47 48 49 50 51 Conclusion 52 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 20 of 36

1 2 3 The porcine hepatic volume evolution following major resection is a fast process, with a 4 5 6 characteristic time of 6.4 days. The highest regeneration rates are in the first days, and 7 8 after three days the liver volume (normalized by body weight) recovered 50% of its 9 10 preoperative value and almost a 100% at 1015 days. Cell proliferation was found 11 12 maximum on the second postoperative day, following an immediate increase of portal 13 14 15 vein flow per liver volume that decreased again over days. 16 For Peer Review Only 17 18 19 20 21 Conflicts of interest: none 22 23 24 Acknowledgements 25 26 27 The authors would like to acknowledge the help of Hans Adriansen and Francois 28 29 LeCompte at the INRA, Tours, France, in imaging acquisition. The authors also thank Dr 30 31 Sebastian Paris from Nanobiotix®, Paris, France for supplying essential data for the 32 33 34 model generation. 35 36 37 38 39 Figure legends: 40 41 42 Figure 1: CT scan estimated liver volume in the perioperative period a) 3D 43 44 45 reconstruction of the total liver volume and the estimated residual hepatic volume, b) 46 47 Axial CT in a venous phase showing the actual residual volume following resection, c) 48 49 axial (noncontract enhanced) CT scan depicting the hepatic volume in the first 50 51 52 postoperative day, and d) further increase in the hepatic volume at day 3 postoperative is 53 54 shown in a contrastenhanced axial CT scan. e) Boxplot with dots representing the 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 21 of 36 HPB

1 2 3 hepatic volume evolution. LiverV is the whole liver volume, V is volume and D is the 4 5 6 day after surgery. f) Boxplot representing the change in the ROI density – measured by 7 8 the Hounsfield units. 9 10 11 Figure 2: Portal vein (left) and hepatic artery (right) flow per liver volume (in 12 13 ml/min/LiverVolume(ml)), before and after resection (D0B, D0A) and during liver 14 15 regeneration at day1 (D1), day3 (D3) and day7 (D7). 16 For Peer Review Only 17 18 Figure 3: Histopathological analysis under light microscopy of the porcine liver using 19 20 21 H&E x200: a) Liver sample prior to resection showing the absence of inflammation or 22 23 steatosis. b) Day 3 postoperative showing microvesicular steatosis (). 24 25 Ki67 x200: c) Liver sample prior to resection showing scanty Ki67 positive cells. d) At 26 27 28 day 3 postoperative there is an increase in the number of Ki67 positive cells. 29 30 CD31 x100 : e) Scanty staining of the sinusoidal cells in samples prior to resection, f) 31 32 Increase in positive staining of the sinusoids at day 3 postoperative, g) Boxplot 33 34 demonstrating the change in the pathological scores over time. 35 36 37 Figure 4: a) Boxplot representing the temporal change in the Ki67, b) Bar diagram 38 39 40 plotting the CD31 expression. Significance was estimated with a Twotailed test statistic 41 42 (Wilcoxon signed rank) at an αerror of 0.05 using SPSS V.22. 43 44 45 Figure 5: evolution over time of the liver weight to body weight ratio, with the 46 47 experimental measurements in red, the preoperative average value in continuous green 48 49 50 and +/ standard deviation (dashed), the mathematical model dynamics (eq. 4) in blue 51 52 with the average regeneration characteristic time (continuous) and its variation taken into 53 54 account (dashed). 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 22 of 36

1 2 3 4 5 6 Supplementary material: 7 8 9 Figure 1: a) Intraoperative photo demonstrating the 4 major hepatic lobes (caudate lobe 10 11 not shown). LL: left lateral lobe, LM: left medial lobe, RM: right medial lobe, RL: right 12 13 14 lateral lobe. bf) Passage of the Rummel tourniquet around the three left hepatic lobes to 15 16 control the inflowFor and outflow Peer in the sameReview time. The effectiveness Only is verified prior to 17 18 commencement of the resection via the colour change and the amount of bleeding during 19 20 21 the parenchymal dissection. 22 23 24 Figure 2: Summary of the protocol of the different groups. 25 26 27 28 29 30 References: 31 32 33 1. Kishi Y, Abdalla EK, Chun YS, Zorzi D, Madoff DC, Wallace MJ, et al. Three 34 hundred and one consecutive extended right hepatectomies: evaluation of outcome 35 based on systematic liver volumetry. Ann Surg. 2009;250(4):540–8. 36 37 2. Ding BS, Nolan DJ, Butler JM, James D, Babazadeh AO, Rosenwaks Z, et al. 38 Inductive angiocrine signals from sinusoidal endothelium are required for liver 39 regeneration. Nature. 2010;468(7321):310–5. 40 41 42 3. Niiya T, Murakami M, Aoki T, Murai N, Shimizu Y, Kusano M. Immediate 43 increase of portal pressure, reflecting sinusoidal shear stress, induced liver 44 regeneration after partial hepatectomy. J Hepatobiliary Pancreat Surg. 45 1999;6(3):275–80. 46 47 48 4. Ninomiya M, Shirabe K, Terashi T, Ijichi H, Yonemura Y, Harada N, et al. 49 Deceleration of regenerative response improves the outcome of rat with massive 50 hepatectomy. Am J Transplant. 2010;10(7):1580–7. 51 52 5. Lin XJ, Yang J, Chen XB, Zhang M, Xu MQ. The critical value of remnant liver 53 volumetobody weight ratio to estimate posthepatectomy liver failure in cirrhotic 54 patients. J Surg Res. 2014;188(2):489–95. 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 23 of 36 HPB

1 2 3 6. Abulkhir A, Limongelli P, Healey AJ, Damrah O, Tait P, Jackson J, et al. 4 Preoperative portal vein embolization for major liver resection: a metaanalysis. 5 6 Ann Surg. 2008;247(1):49–57. 7 8 7. Sparrelid E, Jonas E, Tzortzakakis A, Dahlén U, Murquist G, Brismar T, et al. 9 Dynamic Evaluation of Liver Volume and Function in Associating Liver Partition 10 and Portal Vein Ligation for Staged Hepatectomy. J Gastrointest Surg. 11 2017;21(6):967–74. 12 13 14 8. Bekheit M, Bucur PO, Wartenberg M, Vibert E. Computerized tomographybased 15 anatomic description of the porcine liver. J Surg Res. Elsevier; 2016; 16 For Peer Review Only 17 9. Bucur P, Bekheit M, Audebert C, VignonClementel I, Vibert E. Simplified 18 technique for 75% and 90% hepatic resection with hemodynamic monitoring in a 19 large white swine model. J Surg Res. 2016;209:122–30. 20 21 22 10. Iida T, Yagi S, Taniguchi K, Hori T, Uemoto S. Improvement of morphological 23 changes after 70% hepatectomy with portocaval shunt: preclinical study in porcine 24 model. J Surg Res. 2007;143(2):238–46. 25 26 11. Bucur PO, Bekheit M, Audebert C, Othman A, Hammad S, Sebagh M, et al. 27 Modulating Portal Hemodynamics With Vascular Ring Allows Efficient 28 29 Regeneration After Partial Hepatectomy in a Porcine Model. Ann Surg. 2017; 30 31 12. Court FG, WemyssHolden SA, Morrison CP, Teague BD, Laws PE, Kew J, et al. 32 Segmental nature of the porcine liver and its potential as a model for experimental 33 partial hepatectomy. Br J Surg. 2003;90(4):440–4. 34 35 13. Hammad S, Hoehme S, Friebel A, von Recklinghausen I, Othman A, BegherTibbe 36 37 B, et al. Protocols for staining of bile canalicular and sinusoidal networks of human, 38 mouse and pig livers, threedimensional reconstruction and quantification of tissue 39 microarchitecture by image processing and analysis. Arch Toxicol. 40 2014;88(5):1161–83. 41 42 14. Wang D, Stockard CR, Harkins L, Lott P, Salih C, Yuan K, et al. 43 44 Immunohistochemistry in the evaluation of neovascularization in tumor xenografts. 45 Biotech Histochem. 2008;83(34):179–89. 46 47 15. Demetris AJ, Kelly DM, Eghtesad B, Fontes P, Wallis Marsh J, Tom K, et al. 48 Pathophysiologic observations and histopathologic recognition of the portal 49 hyperperfusion or smallforsize syndrome. Am J Surg Pathol. 2006;30(8):986–93. 50 51 52 16. Bekheit M, Bucur P, Vibert E, Andres C, others. The reference values for hepatic 53 oxygen consumption and net lactate production, blood gasses, hemogram, major 54 electrolytes, and kidney and liver profiles in anesthetized large white swine model. 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 24 of 36

1 2 3 Transl Surg. Medknow Publications; 2016;1(4):95. 4 5 6 17. Monolix version 2016R1. Antony, France: Lixoft SAS [Internet]. 2016. Available 7 from: http://lixoft.com/products/monolix/ 8 9 18. Kuhn E, Lavielle M. Maximum likelihood estimation in nonlinear mixed effects 10 models. Computational Statistics \& Data Analysis. Elsevier; 2005;49(4):1020–38. 11 12 19. Liska V, Treska V, Mirka H, Benes J, Vycital O, Bruha J, et al. Immediately 13 14 preoperative use of biological therapy does not influence liver regeneration after 15 large resectionporcine experimental model with monoclonal antibody against 16 epidermalFor growth factor.Peer In Vivo Review (Brooklyn). 26(4):683–91. Only 17 18 20. Ding BS, Cao Z, Lis R, Nolan DJ, Guo P, Simons M, et al. Divergent angiocrine 19 signals from vascular niche balance liver regeneration and fibrosis. Nature. Nature 20 21 Publishing Group; 2013; 22 23 21. Lauber DT, Tihanyi DK, Czigány Z, Kovács T, Budai A, Drozgyik D, et al. Liver 24 regeneration after different degrees of portal vein ligation. J Surg Res. Elsevier; 25 2016;203(2):451–8. 26 27 22. Audebert C, Bekheit M, Bucur P, Vibert E, VignonClementel IE. Partial 28 29 hepatectomy hemodynamics changes: Experimental data explained by closedloop 30 lumped modeling. J Biomech. 2017;50:202–8. 31 32 23. Stolz DB, Ross MA, Ikeda A, Tomiyama K, Kaizu T, Geller DA, et al. Sinusoidal 33 endothelial cell repopulation following ischemia/reperfusion injury in rat liver 34 transplantation. Hepatology. 2007;46(5):1464–75. 35 36 37 24. Sato T, ElAssal ON, Ono T, Yamanoi A, Dhar DK, Nagasue N. Sinusoidal 38 endothelial cell proliferation and expression of angiopoietin/Tie family in 39 regenerating rat liver. J Hepatol. 2001;34(5):690–8. 40 41 25. MoralesNavarrete H, SegoviaMiranda F, Klukowski P, Meyer K, Nonaka H, 42 Marsico G, et al. A versatile pipeline for the multiscale digital reconstruction and 43 44 quantitative analysis of 3D tissue architecture. Elife. 2015;4. 45 46 26. Lautt WW. Regulatory processes interacting to maintain hepatic blood flow 47 constancy: Vascular compliance, hepatic arterial buffer response, hepatorenal 48 reflex, liver regeneration, escape from vasoconstriction. Hepatol Res. 49 2007;37(11):891–903. 50 51 52 27. Asencio JM, GarcíaSabrido JL, LópezBaena JA, Olmedilla L, Peligros I, Lozano 53 P, et al. Preconditioning by portal vein embolization modulates hepatic 54 hemodynamics and improves liver function in pigs with extended hepatectomy. 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 25 of 36 HPB

1 2 3 Surgery. 2017; 4 5 6 28. Brasaemle DL. Cell biology. A metabolic push to proliferate. Science (80 ). 7 2006;313(5793):1581–2. 8 9 29. Thevananther S. Adipose to the rescue: peripheral fat fuels liver regeneration. 10 Hepatology. 2010;52(6):1875–6. 11 12 30. Tani K, Shindoh J, Takamoto T, Shibahara J, Nishioka Y, Hashimoto T, et al. 13 14 Kinetic Changes in Liver Parenchyma After Preoperative Chemotherapy for 15 Patients with Colorectal Liver Metastases. J Gastrointest Surg. 2017; 16 For Peer Review Only 17 31. Zou Y, Bao Q, Kumar S, Hu M, Wang GY, Dai G. Four waves of hepatocyte 18 proliferation linked with three waves of hepatic fat accumulation during partial 19 hepatectomyinduced liver regeneration. PLoS One. 2012;7(2):e30675. 20 21 22 32. Ulicná O, Istvánová B, Valachová A, Brixová E. [Oxidative phosphorylation in 23 liver mitochondria after injury with carbon tetrachloride and during regeneration]. 24 Bratisl Lek Listy. 1994;95(9):402–7. 25 26 33. Vibert E, Boleslawski E, Cosse C, Adam R, Castaing D, Cherqui D, et al. Arterial 27 Lactate Concentration at the End of an Elective Hepatectomy Is an Early Predictor 28 29 of the Postoperative Course and a Potential Surrogate of Intraoperative Events. Ann 30 Surg. 2015;262(5):787–93. 31 32 34. Ozdogan M, Ersoy E, Dundar K, Albayrak L, Devay S, Gundogdu H. Beneficial 33 effect of hyperbaric oxygenation on liver regeneration in cirrhosis. J Surg Res. 34 2005;129(2):260–4. 35 36 37 35. Donadon M, Molinari AF, Corazzi F, Rocchi L, Zito P, Cimino M, et al. 38 Pharmacological Modulation of IschemicReperfusion Injury during Pringle 39 Maneuver in Hepatic Surgery. A Prospective Randomized Pilot Study. World J 40 Surg. 2016;40(9):2202–12. 41 42 43 44 45 Table 1: Summary of the hepatic functions and flows parameters. 46 47 48 Lactate Before After 49 50 51 (mmol/L) resection resection Day 1 Day 3 Day 7 52 53 54 Mean 1.77 3.61 2.37 2.16 4.01 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 26 of 36

1 2 3 Std. Deviation 1.04 1.91 .57 1.11 2.08 4 5 6 ASAT (Ul/L) 7 8 9 Mean 69.1 160 452.33 335.22 74 10 11 12 Std. Deviation 61.6 88.47 159.63 317.47 78.03 13 14 15 16 ALAT (Ul/L) For Peer Review Only 17 18 19 Mean 42.47 29.33 71.66 188.11 46.5 20 21 22 Std. Deviation 10.8 4.62 22.03 317.87 10.46 23 24 25 Total Bilirubin 26 27 28 (µmol/L) 29 30 31 Mean 4.94 4.83 7.33 6.44 6.83 32 33 34 Std. Deviation 3.2 2.5 2 5.3 3.4 35 36 37 Direct Bilirubin 38 39 (µmol/L) 40 41 42 Mean 1.53 2.72 4.66 3.36 2.36 43 44 45 Std. Deviation .73 1.54 2.08 4.80 1.13 46 47 48 GGT (Ul/L) 49 50 51 Mean 33.36 21.94 30 33.22 26.83 52 53 54 Std. Deviation 8.92 5.91 6.08 8.89 9.92 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 27 of 36 HPB

1 2 3 ALP (Ul/L) 4 5 6 Mean 273.68 315.05 431.66 343.88 137.33 7 8 9 Std. Deviation 52.5 80.9 92.26 161.3 50.83 10 11 12 Platelets (/mm3) 13 14 15 Mean 311.3 243.28 171.5 227.3 272.2 16 For Peer Review Only 17 18 Std. Deviation 120.6 121.3 2.1 124.4 111.4 19 20 21 Prothrombin 22 23 24 activity (%) 25 26 27 Mean 104.5 81 52 87.57 106.25 28 29 30 Std. Deviation 11.6 16.3 1.4 25.9 12.6 31 32 33 Ammonia 34 35 (µmol/L) 36 37 38 Mean 41.4 81.6 119.0 71.0 39.2 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 28 of 36

1 2 3 Std. Deviation 38.5 38.3 41.7 61.7 29.6 4 5 6 Hepatic artery 7 8 flow HAF 9 10 11 (L/min) 12 13 Mean 0.18 0.05 0.06 0.07 0.07 14 15 16 Std. Deviation For 0.07Peer Review0.02 0.02 Only 0.04 0.03 17 18 HAF per Liver 19 20 Volume 21 22 23 ( 24 25 ml/min/LV(ml) 26 27 ) 28 29 30 Mean 0.19 0.23 0.15 0.11 0.12 31 32 Std. Deviation 0.08 0.10 0.09 0.06 0.07 33 34 35 Portal vein 36 37 flow PVF 38 39 (L/min) 40 41 42 Mean 0.77 0.68 0.93 0.98 0.78 43 44 Std. Deviation 0.19 0.19 0.25 0.40 0.19 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 29 of 36 HPB

1 2 3 PVF per Liver 4 5 6 Volume 7 8 ( 9 10 ml/min/LV(ml) 11 12 13 ) 14 15 Mean 0.83 3.11 2.41 1.70 1.27 16 For Peer Review Only 17 18 Std. Deviation 0.27 0.19 0.93 0.60 0.37 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 30 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For Peer Review Only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 CT scan estimated liver volume in the peri-operative period a) 3D reconstruction of the total liver volume 44 and the estimated residual hepatic volume, b) Axial CT in a venous phase showing the actual residual 45 volume following resection, c) axial (non-contract enhanced) CT scan depicting the hepatic volume in the 46 first postoperative day, and d) further increase in the hepatic volume at day 3 postoperative is shown in a 47 contrast-enhanced axial CT scan. e) Boxplot with dots representing the hepatic volume evolution. Liver-V is 48 the whole liver volume, V is volume and D is the day after surgery. f) Boxplot representing the change in the 49 ROI density – measured by the Hounsfield units.

50 261x299mm (150 x 150 DPI) 51 52 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 31 of 36 HPB

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For Peer Review Only 17 18 Portal vein (left) and hepatic artery (right) flow per liver volume (in ml/min/LiverVolume(ml)), before and 19 after resection (D0-B, D0-A) and during liver regeneration at day-1 (D1), day-3 (D3) and day-7 (D7). 20 21 246x82mm (72 x 72 DPI) 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 32 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For Peer Review Only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 Histopathological analysis under light microscopy of the porcine liver using H&E x200: a) Liver sample prior 46 to resection showing the absence of inflammation or steatosis. b) Day 3 postoperative showing 47 microvesicular steatosis (◊). 48 Ki67 x200: c) Liver sample prior to resection showing scanty Ki67 positive cells. d) At day 3 postoperative 49 there is an increase in the number of Ki67 positive cells. 50 CD31 x100 : e) Scanty staining of the sinusoidal cells in samples prior to resection, f) Increase in positive staining of the sinusoids at day 3 post-operative, g) Boxplot demonstrating the change in the pathological 51 scores over time. 52 53 125x280mm (300 x 300 DPI) 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 33 of 36 HPB

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For Peer Review Only 17 18 19 20 a) Boxplot representing the temporal change in the Ki67, b) Bar diagram plotting the CD31 expression. Significance was estimated with a Twotailed test statistic (Wilcoxon signed rank) at an αerror of 0.05 using 21 SPSS V.22. 22 23 326x123mm (300 x 300 DPI) 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 34 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For Peer Review Only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 evolution over time of the liver weight to body weight ratio, with the experimental measurements in red, the 32 preoperative average value in continuous green and +/- standard deviation (dashed), the mathematical 33 model dynamics (eq. 4) in blue with the average regeneration characteristic time (continuous) and its 34 variation taken into account (dashed). 35 36 203x152mm (300 x 300 DPI) 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Page 35 of 36 HPB

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For Peer Review Only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 222x174mm (300 x 300 DPI) 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb HPB Page 36 of 36

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For Peer Review Only 17 18 19 234x68mm (300 x 300 DPI) 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 http://mc.manuscriptcentral.com/hpb Available online at www.sciencedirect.com ScienceDirect

Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 www.elsevier.com/locate/cma

Kinetic scheme for arterial and venous blood flow, and application to partial hepatectomy modeling

Chloe Audeberta,b,∗, Petru Bucurc,d, Mohamed Bekheitc,e, Eric Vibertc,e, Irene E. Vignon-Clementela,b,1, Jean-Fred´ eric´ Gerbeaua,b,1

a Inria Paris, 2 rue Simone Iff, CS 42112, 75589 Paris Cedex 12, France b Sorbonne Universites´ UPMC Univ Paris 6, Laboratoire Jacques-Louis Lions, 4 Place Jussieu, 75252 Paris Cedex 05, France c Inserm Unite´ 1193, Villejuif, France d CHRU, hopitauxˆ de Tours, Chirurgie Hepato-biliaire´ et Pancreatique,´ Transplantation Hepatique,´ Tours, France e AP-HP, Hopitalˆ Paul Brousse, Centre Hepato-Biliaire,´ Villejuif, France

Available online 22 July 2016

Highlights • The article introduces a kinetic scheme to solve the 1D Euler equations. • The scheme is shown to work well for both arterial and venous wall laws. • Vessel collapse is studied. • Liver resection of different extents is simulated with a closed-loop 1D–0D model. • Resulting waveform changes are successfully compared with real measurements.

Abstract

The article introduces a kinetic scheme to solve the 1D Euler equations of hemodynamics, and presents comparisons of a closed-loop 1D–0D model with real measurements obtained after the hepatectomy of four pigs. Several benchmark tests show that the kinetic scheme compares well with more standard schemes used in the literature, for both arterial and venous wall laws. In particular, it is shown that it has a good behavior when the section area of a vessel is close to zero, which is an important property for collapsible or clamped vessels. The application to liver surgery shows that a model of the global circulation, including 0D and 1D equations, is able to reproduce the change of waveforms observed after different levels of hepatectomy. This may contribute to a better understanding of the change of liver architecture induced by hepatectomy. ⃝c 2016 Elsevier B.V. All rights reserved.

Keywords: Kinetic scheme; Arterial flow; Venous flow; Vessel collapse; Surgery simulation

∗ Corresponding author at: Inria Paris, 2 rue Simone Iff, CS 42112, 75589 Paris Cedex 12, France. 1 Authors contributed equally. http://dx.doi.org/10.1016/j.cma.2016.07.009 0045-7825/⃝c 2016 Elsevier B.V. All rights reserved. C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 103

1. Introduction

Liver partial ablation surgery, namely partial hepatectomy, is necessary to treat some pathologies. In order to get a functional regeneration of the liver, the weight ratio of the remaining liver to the body must be at least 0.5% for a healthy human [1]. However, the liver ablation percentage needs sometimes to be higher, in presence of large tumors for example. Post-operative liver failure may then occur due to insufficient functional liver mass. When partial ablation is performed, the remaining liver experiences pressure and flow changes. The importance of the hemodynamics changes depends on ablation size, but their relationship remains unclear. Moreover, the remaining liver regeneration capacity seems to be impacted by the post-resection hemodynamics. A better understanding of the hemodynamics impact of hepatectomy might therefore help improve surgical practice. To contribute to this challenge, we adopt two approaches: one is based on animal experiments, the other on mathematical modeling and simulation. The present work shows that the simulations are able to reproduce, and possibly explain, some findings of the experiments. Experiments have been performed on pigs. This species is a good animal model for our problem since its liver to body weight ratio is close to human’s [2]. Pressure and flow in the main vessels of the liver have been recorded for different resection percentages. An interesting finding of these experiments was the following: at the resection time, waveform changes were observed repeatedly in the pressure and flow measured in the hepatic artery. These changes differ for 75% and 90% hepatectomy. Since it is hypothesized that there is a link between liver architecture and hemodynamics, and since liver architecture is important to understand liver regeneration, there is a strong interest in explaining these changes in pressure and flow waveforms. A mathematical model able to reproduce this phenomenon must satisfy several requirements. First, it has to be able to capture wave propagation. A network of vessels modeled by systems of the one-dimensional (1D) hyperbolic Euler equations is a natural candidate in this respect. The liver being perfused by both arterial and venous blood, the model should be able to address both kinds of vessels. In addition, since during surgery some vessels can be clamped, the model and the numerical scheme should be able to handle the limit of vanishing cross-section area. In this work, we propose to use a kinetic scheme, in particular because of its interesting capability to preserve the positivity of the cross-section area. This scheme was originally developed for the Saint-Venant shallow water equations. To our knowledge, this is the first time that it is used to model collapsible vessels. Second, keeping in mind that the liver receives about 25% of the cardiac output [3], hepatectomy may also influence the systemic circulation. It is therefore desirable to embed the network of 1D models within aclosed-loop model of the whole circulation, including the liver. To keep a moderate complexity, this compartment can be treated with zero-dimensional (0D) models, also known as lumped-parameter models, i.e. governed by ordinary differential equations. The paper is organized as follows. In Section 2, the hyperbolic equations are recalled and the kinetic scheme is described, along with the boundary and coupling conditions. The kinetic scheme is validated on benchmark cases, for both arterial and venous flows. In Section 3, the closed-loop 0D–1D model is presented and the effects of partial hepatectomy are studied numerically and compared with experimental observations. Section 4 ends the paper, with some conclusions and perspectives.

2. Kinetic scheme for arterial and venous blood flow

2.1. The Euler equations of hemodynamics

Blood flow in large vessels of the cardiovascular system can be represented with a collection of one-dimensional systems of nonlinear equations:

 + =  ∂t A ∂x (Au) 0 2 A (1) ∂t (Au) + ∂x (κ Au ) + ∂x p = Ag − f (A, A , u).  ρ 0 The first equation corresponds to mass conservation and the second to momentum conservation. x ∈ R denotes the + coordinate along the longitudinal axis of the portion of vessel, t ∈ R is the time, A(x, t) is the vessel cross-section area, u(x, t) is the mean velocity of blood through the corresponding cross-section, ρ is the fluid density assumed 104 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 constant, g denotes the gravity along the longitudinal axis, f (A, A0, u) is a friction term, and κ is a momentum-flux correction coefficient, assumed to be equal to 1 in this work. These equations have been used by many authors (e.g. [4–14] to only name a few). Many variants exist, for example in the treatment of dissipation, viscoelasticity, curvature, momentum-flux correction coefficient, etc. Here we choose the simplest form of these different components, adding complexity through gravity and dissipation as needed by the test cases. The mean pressure p(x, t) in a cross-section is related to the cross-section area through an algebraic constitutive law:

p(x, t) = P0(x) + ψ (A(x, t), A0(x), β(x)) , where A0(x) is a reference area, P0(x) is the pressure when A(x, t) is equal to A0(x), β(x) is a parameter representing the vessel stiffness, and ψ is a given function characterizing the “tube law”. The details regarding the tube law for this work can be found in [7,15] for arteries, and in [16–19] for veins. Eliminating the pressure from system (1) gives:  + =  ∂t A ∂x (Au) 0 2 A A   (2) ∂t (Au) + ∂x (Au ) + (∂Aψ)∂x A = Ag − f (A, A , u) − ∂x P + (∂A ψ)∂x A + (∂ ψ)∂x β .  ρ 0 ρ 0 0 0 β A In order to write the system in conservative form, the term ρ (∂Aψ)∂x A is reformulated:

 A(x,t)   A(x,t)  A 1 2 (∂Aψ)∂x A = ∂x a∂aψ(a, A0, β)da = ∂x c (a)da (3) ρ ρ εA0 εA0  = A where ε is a constant whose value will be discussed later, and c(A) ρ ∂Aψ is the wave speed. With this reformulation, system (2) becomes:

 ∂t A + ∂x (Au) = 0   A(x,t)   2 2 A  ∂t (Au) + ∂x (Au ) + ∂x c (a)da = Ag − f (A, A0, u) − ∂x P0 (4) εA0 ρ    + + (∂A0 ψ)∂x A0 (∂β ψ)∂x β .

When the reference cross-section area A0, the stiffness parameter β and the pressure P0 are assumed to be constant in space, the system reads:  + =  ∂t A ∂x (Au) 0   A(x,t)  + 2 + 2 = − (5) ∂t (Au) ∂x (Au ) ∂x c (a)da Ag f (A, A0, u).  εA0

2.2. The kinetic scheme

Many numerical methods have been used in the literature to address the solution of (5). In the arterial case, we refer to the recent overview presented in [20], where six different methods were compared: discontinuous Galerkin, locally conservative Galerkin, Galerkin least-squares finite element, finite volume, finite difference MacCormack, and a simplified trapezium rule method (STM). In the venous case, a Godunov scheme has been usedin[18], an ADER (Arbitrary Accuracy DERivative Riemann problem) scheme in [19], and a Runge–Kutta discontinuous Galerkin scheme in [21]. In the present work, a kinetic scheme is adopted for both arterial and venous flows. A motivation for this method, which was initially proposed for the Saint-Venant equations [22], is its capability to provably preserve the positivity of the cross-section area, which is especially relevant in collapsible vessels. To our knowledge, this is the first time this scheme is used in hemodynamics for collapsible vessels. It was recently used for arterial flow in [23]. A kinetic interpretation of system (5) is obtained by introducing a linear microscopic kinetic equation equivalent to the macroscopic model [24]. A real function χ defined on R is introduced. It is compactly supported and verifies C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 105 the following properties:  χ(−w) = χ(w) ≥ 0   = 2 = (6)  χ(w)dw w χ(w)dw 1. R R

Here, this function is defined by χ(w) = √1 1 √ , but other choices are possible [24]. A distribution function 2 3 |w|≤ 3 M(x, t, ξ) is introduced:

A ξ − u  M(x, t, ξ) = χ , γ γ with γ 2 = 1  A(x,t) c2(a)da. The wave speed c will be further specified in the next section for arterial and venous A εA0 blood flows. In the kinetic formalism, the variable ξ represents the microscopic particle velocity. Consider first the case without source terms. The equation verified by M and the system (5) are linked with the following result [24]: the functions A and u are solutions to the Euler equations (5), if and only if M(x, t, ξ) is solution to the kinetic equation

∂t M + ξ∂x M = Q(x, t, ξ), (7) where Q(x, t, ξ) is a collision term that satisfies:   Qdξ = ξQdξ = 0. R R The link between the microscopic density function and the macroscopic variable is given by the two relations:   Mdξ = A, ξ Mdξ = Au. (8) R R

n n Let ∆t and ∆x denote the time and space steps respectively. Let (Ai , ui ) denote an approximation of (A(xi , tn), = = n n u(xi , tn)), with tn n∆t and xi i∆x. The unknown (Ai , ui ) is solution to a finite volume kinetic scheme deduced n from the kinetic interpretation of the equations. Let Mi be the discrete particles density, defined by An ξ − un  n = n = i i Mi Mi (ξ) n χ n , γi γi

n 1 n  1  Ai 2  2 with γ = n c (a)da . Eq. (7) is approximated by an upwind scheme: i Ai εA0

∆t   n+1,− = n − n − n Mi Mi ξ M + 1 M − 1 , ∆x i 2 i 2

n n n n n n n with M = M 1 ≥ + M 1 ≤ . Then A and (Au) = A u are computed with (8): + 1 i ξ 0 i+1 ξ 0 i i i i i 2

 n+1    A  1 + − X n+1 = i = Mn 1, dξ. (9) i An+1un+1 ξ i i i R The kinetic scheme reads:

∆t n+1 = n − n − n Xi Xi (F + 1 F − 1 ), (10) ∆x i 2 i 2 106 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125   with F n =  ξ 1 Mn dξ. Given the function χ chosen above, the following integrals can be computed in closed + 1 ξ + 1 i 2 R i 2 form: √     ξ=max( −u ; 3) p A ξ − u 1 A  p+  γ ξ χ dξ = √ (ξγ + u) 1 √ + ξ=max( −u ;− 3) ξ≥0 γ γ 2 3 γ (p 1) γ √ − p = 1, 2 (11)     ξ=min( u ; 3) p A ξ − u 1 A  p+  γ ξ χ dξ = √ (ξγ + u) 1 √ = −u ;− ξ≤0 γ γ 2 3 γ (p + 1) ξ min( γ 3) which gives the expression of the flux F. In presence of a source term, we adopt the simple strategy of an explicit treatment:

∆t n+1 = n − n − n + n n Xi Xi (F + 1 F − 1 ) ∆t S(Xi ), (12) ∆x i 2 i 2 T where S(X n) = 0, g An − f (An, A , un) . i i i 0 i √ | n| + n ≤ Under the CFL condition ∆t maxi ( ui 3γi ) ∆x, following the same arguments as in [22] for the shallow n ≥ water equations, it can be proved that the scheme (12) preserves the positivity of the cross-section area, i.e Ai 0, if this property holds at time zero. A second order extension of (12) can be obtained with standard arguments (minmod flux limiter).

Remark 1. As mentioned above, A0 is assumed to be constant in each vessel. If A0 was space-dependent, for example to account for the vessel tapering, the source term should be carefully treated to obtain a numerical scheme that ensures the equilibrium at rest. A similar issue was addressed for the shallow water equations in [25] with a technique called “hydrostatic reconstruction”. To our knowledge, in the context of blood flow, this question was first addressed in[23] and named “dead man equilibrium”. It was also addressed in [26] for a different scheme.

Arteries and veins tube laws. The tube law for arteries [15] is defined by, √   ψ (A, A0, βa) = βa A − A0 , (13) with √ 4 π Eh0 βa = , (14) 3A0 where E is the Young’s modulus, and h0 the thickness of the tube. The wave speed is then defined by:

βa  c2 = A(x, t). (15) 2ρ

Thus, with the arterial tube law, c2 is integrable in A = 0 and we can choose ε = 0 in (3). The kinetic distribution A  ξ−u  A(x,t) β 1 = 2 = 1  2 = a 2 = 2 2 function M is defined by M(x, t, ξ) γ χ γ , with γ A 0 c (a)da 3ρ A 3 c and system (5) reads:  ∂t A + ∂x (Au) = 0    2 β 3 (16) ∂t (Au) + ∂x (Au ) + ∂x A 2 = Ag − f (A, A , u),  3ρ 0 √ √  with p(x, t) = P0 + βa A(x, t) − A0 . For collapsible tubes, like veins, we adopt the same tube law as in [17–19]:

 A m  A n ψ (A, A0, βv) = βv − , (17) A0 A0 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 107 where βv is an elasticity parameter. With m = 10 and n = −1.5, which are the values commonly used in the literature, the squared wave speed defined by

β   A m  A n c2 = v m − n (18) ρ A0 A0 is not integrable at A = 0. This question would deserve a special study. Here, for simplicity, we circumvent this difficulty by taking ε > 0 in (3): 1 β  A(x,t)   a m  a n γ 2 = v m − n da A ρ εA0 A0 A0 β  m  A m n  A n A  m n  = v − − 0 εm+1 − εn+1 . (19) ρ m + 1 A0 n + 1 A0 A m + 1 n + 1 The value of ε will be discussed in Section 2.4.

2.3. Boundary treatments

Characteristic variables. The characteristic variables are computed from the quasi-linear form of system (5). For arterial blood flow [15], the characteristic variables are:   β 1 β 1 W+ = u + 4 A 4 , W− = u − 4 A 4 . (20) 2ρ 2ρ For venous blood flow [19], the characteristic variables are:

 A c(a)  A c(a) W+ = u + da, W− = u − da. (21) A0 a A0 a In the following numerical examples, the characteristic variables are approximated with the trapezoidal rule for venous blood flow. Transmission conditions. In presence of a bifurcation, or a change in material properties, conservation of mass is = + = imposed: Qm Qd1 Qd2 , where Q Au denotes the flow rate, m the mother vessel and d1, d2 the two daughter = ρ 2 + vessels. Except in some specific cases detailed below, continuity of the total pressure PT 2 u p is also imposed: = = PT,m PT,d1 PT,d2 . These relations are complemented with the relations provided by the outgoing characteristics, as explained e.g. in [15]. The resulting system of nonlinear equations is then solved with a Newton method. Boundary conditions. Different types of boundary conditions are considered in the numerical examples. At the inlet of the open-loop models, either the pressure or the flow rate are imposed. At the outlet, either a constant pressure, or an absorbing boundary condition, or a coupling with a 0D model are used. For the absorbing boundary condition, the incoming characteristic variable is assumed constant in time. For the coupling with a 0D model, the differential equations are approximated with an implicit Euler scheme. Again, these relations are complemented with the information obtained from the outgoing characteristics. A Newton method and a parabolic linesearch algorithm are used to solve the resulting system of nonlinear equations.

2.4. Benchmark test cases

Arterial flow. Various benchmark test cases were proposed in [20] to compare six numerical methods for 1D blood flow models. Two representative tests are studied in the following paragraphs: a single pulse propagation, andanaortic bifurcation simulation. The kinetic scheme is compared to the results from [20]. For the two cases, the system (5) is solved, with a friction function defined by f (A, A0, u) = K f u(x, t), K f being constant, and gravity is neglected. Single pulse propagation The first test case is the (non-physiological) propagation of a pulse wave along atube, with an absorbing outlet boundary condition. Table 6 provides the parameters values. The inlet flow is imposed: 4 2 3 −1 Q0(t) = exp(−10 (t − 0.05) ) cm s . First, the test is performed with the first order kinetic scheme, and the 108 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125

Fig. 1. Comparison between time and space first order kinetic scheme results with ∆x = 0.1 cm, and ∆t = 1.0 10−4 s (blue), with ∆x = 0.01 cm, and ∆t = 1.0 10−5 s (red) and numerical results from [20] (dash black) for the inviscid single pulse propagation test. The pressure over space is represented for different time instants: 0.1, 0.3, 0.5, 0.7, 0.9, 1.1, 1.3, and 1.5 s. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) friction is neglected. In [20] all the numerical schemes give identical results for this benchmark. The kinetic scheme is here compared with the STM scheme. For ∆t = 10−4 s and ∆x = 10−1 cm the resulting pressure curves are shown in Fig. 1 (blue curve). An excessive numerical diffusion is observed, which is reduced when space and time steps are refined (Fig. 1 red curve), ∆t = 10−5 s and ∆x = 10−2 cm). The results obtained with the second order in space kinetic scheme, with ∆t = 10−5 s and ∆x = 10−1 cm, are plotted in Fig. 2(a). Fig. 2(b) shows the results obtained in the viscous case with the second order in space kinetic scheme (∆t = 10−5 s and ∆x = 10−1 cm). Table 7 summarizes the normalized errors for all presented simulations. With the second order kinetic scheme, the results are in excellent agreement with the STM scheme.

Aortic bifurcation The second arterial test is an abdominal aorta branching into two symmetric iliac arteries. The vessel parameters are in Table 8. Two three-element Windkessel models represent the rest of the systemic circulation and are coupled to the two 1D iliac arteries. The flow rate is imposed at the inlet.√ Space and time steps for the kinetic = −5 = | n| + n scheme are ∆t 5 10 s and ∆x 0.1 cm. The CFL number (∆t maxi ( ui 3γi )/∆x) remains around 0.63. Fig. 3 shows pressure, flow rate and radius change for the middle and the end points of the aorta, and the middlepoint of the iliac artery, compared to the results of 3D simulations and of the 1D scheme STM presented in [20]. In [20] errors with respect to 3D solution are computed. The errors are defined by:       1 n  P − P   1 n Q − Q RMS =   i i RMS =   i i EP  , EQ    (22) n Pi n max(Q j ) i=1 i=1 j       MAX  Pi − Pi  MAX  Qi − Qi  EP = max   , EQ = max   (23) i  Pi  i max(Q j )  j  max(P) − max(P) max(Q) − max(Q) ESYS = , ESYS = (24) P max(P) Q max(Q) min(P) − min(P) min(Q) − min(Q) EDIAS = , EDIAS = , (25) P min(P) Q max(Q) C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 109

(a) Inviscid blood.

(b) Viscous blood.

Fig. 2. Comparison between first order in time and second order in space kinetic scheme results with ∆x = 0.1 cm, and ∆t = 1.0 10−5 s and numerical results from [20] with inviscid (a) and viscous (b) blood for the single pulse propagation test. The pressure over space is represented for various times: 0.1, 0.3, 0.5, 0.7, 0.9, 1.1, 1.3, and 1.5 s. where Pi and Qi are the 1D simulation results at a given space point xi , i ∈ 1...N, Pi and Qi are the 3D solutions at the same space location. The errors for ∆P and ∆r are defined similarly. Table 9 presents the errors. Again, the kinetic scheme is in very good agreement with the other schemes presented in [20]. In that case, which is more physiological than the previous one, the first order kinetic scheme is sufficient to reach a good accuracy with reasonable discretization steps. Venous flow. After having been tested on arterial benchmarks, the kinetic scheme is now applied to venous flow, which is more challenging. For collapsible tubes, such as veins, the squared speed wave (Eq. (18)) is not integrable at A = 0. This difficulty is avoided by taking ε > 0 in (3). In the following numerical simulations, we took the value ε = 10−3. 110 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125

Fig. 3. Aorta bifurcation test case: kinetic first order scheme results with ∆t = 5.10−5 s and ∆x = 0.1 cm, and 3D and STM scheme results from [20], over one cardiac cycle.

We noticed that the solution was slightly sensitive to the value of ε, but this dependency is reduced when space and time steps are decreased. Jugular vein collapse The test of the “giraffe jugular vein” was proposed in [17,18], and used more recently 2 in [21,19]. A single vein is considered with length L = 200 cm, cross-section area A0 = 5 cm and material 2 property parameter βv = 50 dyn/cm . A constant flow rate is imposed at the inlet and a fixed cross-section areaat 2  A(t,x) the outlet. The value of gravity is g = 980.0 cm/s , and the friction is defined by f (A, A , u) = K f u(t, x), 0 A0 = 2 = + x = 3 with K f 0.96 cm /s. The initial conditions are A(x, 0) A0(0.2 1.8 L ) and u(x, 0)A(x, 0) 40 cm /s. 3 The boundary conditions are u(0, t)A(0, t) = 40 cm /s and A(L, t) = 2A0. In case of a super-critical inlet, 3 2 u(0, t)A(0, t) = 40 cm /s and A(0, t) = 0.3825 cm are the two imposed conditions. The system√ is solved with = = −4 | n| + n the first order kinetic scheme with ∆x 1.0 cm and ∆t 10 s. The CFL number (∆t maxi ( ui 3γi )/∆x) is around 0.06 for the chosen time and space steps. Here, 201 nodes are used to solve the problem, whereas 1000 nodes were used for the Godunov scheme in [18]. The results for α = A/A0 and the velocity u are plotted over the vessel length for t = 5.7 s and t = 50 s in Fig. 4. Gravity tends to empty the upstream part of the vessel, thus a super-critical flow appears at the inlet. The vessel cross-section area at the outlet is forced to remain equal to 2A0, hence the flow remains sub-critical at the outlet and a shock appears in the middle of the vessel. The position of the shock oscillates until the solution converges. Fig. 4 shows the solution at time t = 5.7 s and the stationary state (time t = 50 s). The obtained curves are very similar to the curves reported in [19,21,18]. The front position is x/L = 0.8 in [18], x/L = 0.72 in [21], x/L = 0.74 in [19]. With our numerical scheme, the front position is x/L = 0.74. Portal vein uncollapse To illustrate the robustness of the kinetic scheme, we propose a new benchmark test case mimicking the uncollapse of the portal vein. During the surgery described in the following section, the surgeons clamp the main vessels perfusing the organ to avoid blood loss. When the clamp is in place, the vessel is collapsed. Once C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 111

Fig. 4. Giraffe jugular vein test: (a) at time t = 5.7 s (black) and t = 50 s (red), simulation results with kinetic scheme (∆t = 10−4 s, ∆x = 1.0 cm). α = A (top) and velocity in m/s (bottom) are plotted over the vessel length. (b) schematic representation including gravity. (For A0 interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) sutures are done, they remove the clamp. The proposed test is mimicking the uncollapse of the portal vein, just after the clamp removal. We assume that a cross-section area of 0.5% of A0 corresponds to a collapsed vessel. The 1D blood flow equations (5) are solved with the vein constitutive law (18) in a single vessel, with length L = 6 cm, cross-section area 2 2 A0 = 0.8 cm and material parameter βv = 10 dyn/cm . The initial conditions are set to represent a collapsed vessel due to the clamp in the middle of the tube:  10.95 A(x, 0) = 1.1 − x A for x < 4L/10 4L 0 A(x, 0) = 0.005A0 for 4L/10 ≤ x ≤ 6L/10  10.95 (26) A(x, 0) = −1.6375 + x A for 6L/10 < x ≤ L 4L 0 A(x, 0)u(x, 0) = 0 cm3/s.

The inlet and outlet pressures are imposed, corresponding to a cross-section area of 1.1 A0. The pressure corresponding −3 to A = A0 is P0 = 1.05 mmHg, the blood density is ρ = 1 g cm , the friction term is f (A, A0, u) = K f u(t, x), 2 with K f = 0.96 cm /s. The gravity is neglected. The first order kinetic scheme is used, with ∆x = 0.05 cm and −4 ∆t = 10 s. Fig. 5 shows the quantity α = A/A0 for various time instants. At time t = 0 s, just after unclamping, the vessel is collapsed in the middle. Then, the vessel uncollapses and oscillates around the equilibrium position (see t = 2 s and t = 5 s in Fig. 5) to eventually reach a steady state (see t = 20 s in Fig. 5).

3. Application to hepatectomy

To our knowledge, only a few mathematical models were proposed in the literature to describe the hemodynamics impact of liver surgeries. In [27], a cast-based reconstruction of the rat liver vasculature was performed to compute the resistance in the different vascular trees. Various sizes of virtual resection were studied with a resistance model and two 90% resection techniques were compared. The results indicated a portal hyperperfusion after resection and demonstrated that probably better outcomes could be expected with one of the two resection techniques. In [28], a 3D simulation was performed in the portal vein after right lobe hepatectomy. The geometry, based on medical imaging data, included superior mesenteric and splenic veins merging in portal vein and three portal vein branches. Constant velocities boundary conditions were prescribed in the mesenteric and splenic veins and zero pressure was imposed 112 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125

Fig. 5. Uncollapse of portal vein simulation: α = A is plotted along the vessel at various time instants. A0 at the outlets. The right lobe hepatectomy was simulated changing the geometry. Similarly, for a two-lobe liver 0D model, driving conditions were kept unchanged before and after hepatectomy. The model proposed in the present work differs from these two approaches from several aspects that will be detailed below. Our strategy is to propose a model of moderate complexity which can be parameterized to match measurements, but with a sufficient level of realism to be able to capture non-trivial phenomena observed inanimal experiments.

3.1. A closed-loop model

To be able to consider waveform changes as a result – and not as an input – of the simulations, a closed-loop model is proposed including 1D and 0D compartments. Although this approach is not new, models of that sort, calibrated with experimental measurements, are not numerous in the literature. Closed-loop models, including 0D–1D–3D vessels, were proposed in [29] to study the impact of aortic insufficiency on the local hemodynamics of a cerebral aneurysm, and in [30] to study the effects of arterial and aortic valvular stenoses. Closed-loop 0D–1D models, including arteries and veins, were proposed in [31,32]. The latter article focused on head and neck, to study possible connections between the venous vasculature and a class of neurodegenerative diseases. The simulation results were compared to Phase-Contrast MRI flow data. The closed-loop model proposed in this work is represented in Fig. 6. The main arteries are modeled with the 1D Euler equations described above. The arterial and venous trees at each outlet, as well as the pulmonary circulation and vena cava, are modeled with three-element Windkessel models. The heart and the liver are also represented by 0D models. The main features of each of these compartments are now detailed. For the 1D models, the length, the cross-section area A0 and the bifurcation angles are estimated from CT- scans of the pigs which underwent the surgical operations (CT-scans were done with a Siemens Somatom AS definition 128 machine). At the bifurcation, the continuity of total pressure is enforced as explained inSection 2, except at the bifurcation between the abdominal aorta and the celiac trunk, and when the celiac trunk bifurcates into the hepatic and the splenic√ arteries. In these two bifurcations, the condition proposed in [7] is adopted: P = P − 2 sign(u )u2 2(1 − cos(α )), where α is the angle of the branches d with respect to the mother T,d1 T,m d1 d1 1 1 1 vessel. The elasticity parameters are computed with the following formula [6]:

Eh0 = a exp(br0) + c, (27) r0 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 113

Fig. 6. Schematic representation of the 1D–0D closed-loop model. 1D blood flow is simulated in the thick lines arteries while thin lines represent the 0D model connections. All RCR units and the liver are linked (thin arrows) to the vena cava (VC). where E is the Young’s modulus, h0 and r0 are the vessel thickness and radius when A = A0. This formula is scaled in order to obtain a pressure in the carotid artery which is similar to the one measured in pigs. The parameters are: a = 6.0 106 g s−2 cm−1, b = −22.53 cm−1 and c = 2.595 105 g s−2 cm−1. The values for each artery can be found in Table 2. = ∆P The total resistances for RCRs are computed with Rtot Q , by a combination of flow splits from [33], assuming pig and human flow splits are similar, and of available pressure and flow measurements. These total resistances are then separated into proximal Rp and distal Rd resistances, assuming the proximal resistance carries 10% of the total resistance in each RCR, within the ranges used in [34–36]. The total systemic capacitance is fixed at 4.10−4 cm5 dyn−1 as reported in [37] for pig circulation, further split based on the number of large arteries represented by each compartment according to [38]. Table 3 summarizes RCR parameter values. The liver model is based on the pig anatomy. The pig liver consists of three separate lobes and is perfused by venous blood, through the portal vein, and arterial blood, through the hepatic artery. The three lobes are represented by three 0D models in parallel, connected to the heart through the 1D models, and to the digestive organs through the venous input connected to the RCR models of the splenic and mesenteric arteries. The vascular tree sizes are assumed proportional to the perfused tissue mass. A larger vascular tree has a smaller resistance, therefore the lobar resistances of the hepatic artery tree, the portal vein tree and the liver tissue are assumed inversely proportional to the lobe mass. The lobar capacitances of the hepatic artery tree and the liver tissue are assumed proportional to the lobe mass. The proximal to total resistance ratios of the hepatic artery tree reflect the lobar architecture differences [27]. The values of the liver parameters are reported in Table 4. The functions governing the heart contraction come from the literature [39,40,30,29], but their parameters are adapted to the pig heart. Heart valves are modeled with logistic functions, in order to obtain smooth yet sharp transitions between open and closed states. The heart chamber equations read:

 dVi  = Qin,i − Qout,i dt (28) Pi = Ei (t)(Vi − V0i )  Qout,i = Gi (Pi − Pout,i )(Pi − Pout,i ), where i denotes either the right atrium (RA), right ventricle (RV), left atrium (LA) or left ventricle (LV); Vi and V0i are respectively the volume and unloaded volume of the heart chamber i; Qin,i and Qout,i are the incoming and outgoing 114 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125

Fig. 7. Carotid pressure over two cardiac cycles: measurement curves for four pigs (dashed lines) and 1D–0D closed-loop model simulated curve (solid line). flows; Pi is the heart chamber pressure; Pi − Pout,i is the pressure across the valve; Ei is the elastance function, αi defined by Ei (t) = Eai e(t) + Ebi with αi = 1 if i = RA, LA and αi = 0.5 if i = RV, LV , as in [41]. Eai and Ebi are the amplitude and baseline elastances respectively, and e is a normalized time-varying function of the elastance, defined as follows for the ventricles: 1   t   1 − cos π 0 ≤ t ≤ Tvc 2 Tvc     e(t) = 1 t − Tvc (29) 1 + cos π T c < t ≤ T c + T r 2 T v v v  vr 0 Tvc + Tvr < t ≤ Tcc, and for the atria:    1 t + Tcc − tar  1 + cos π 0 ≤ t ≤ tar + Tar − Tcc 2 Tar  0 tar + Tar − Tcc < t ≤ tac     e(t) = 1 t − tac (30) 1 − cos π tac < t ≤ tac + Tac 2 Tac     1 t − tar  1 + cos π tac + Tac < t ≤ Tcc, 2 Tar where Tcc is the duration of the cardiac cycle. The durations of the ventricular and atrial contractions and relaxations are denoted by Tvc, Tac, Tvr and Tar respectively; tac and tar are the times when the atria begin to contract and relax, respectively. The heart parameter values are given in Table 1. The valve conductance is described by the function = G0 = 5 −1 −1 = 2 G(∆P) 1+exp(−(∆P−d)) where G0 0.1 cm dyn s , and d 0.1 dyn/cm . Hepatectomy simulation in the 1D–0D closed loop model. The system of equations (5) is solved for the large arteries, with the first order kinetic scheme, with ∆x = 0.1 cm and ∆t = 10−4 s (see Table 5). Gravity is neglected 2 and the friction function is defined as f (A, A0, u) = K f u(x, t), with K f = 3 cm /s. The initial conditions are p(x, t = 0) = 45 mmHg and u(x, t = 0) = 5.0 cm/s. Before hepatectomy, the 1D–0D closed loop model is tuned with the available measurements. Given the variability between subjects, the parameters are not tuned to represent a specific animal but to obtain representative pressures and C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 115

(a) Pig 1. (b) Pig 2.

Fig. 8. Experimental measurements of hepatic artery pressure and flow rate during 75% hepatectomy for four different pigs, the dark lines indicating the clamping time.

(a) Pig 3. (b) Pig 4.

Fig. 9. Experimental measurements of hepatic artery pressure and flow rate during 75% to 90% hepatectomy for two different pigs, the darklines indicating the clamping time. 116 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125

(a) 75% hepatectomy simulation.

(b) 75% to 90% hepatectomy simulation.

Fig. 10. 1D–0D closed-loop model results: hepatic artery pressure (mmHg) and flow rate (L/min) during 75% simulated hepatectomy and 75%to 90% simulated hepatectomy, the dark lines indicating the time of simulated clamping. flow rates. Fig. 7 shows the measured carotid pressure curves over time for the four animals and the simulated curve. The first cardiac cycles at the left hand sideof Fig. 10 show the pressure and flow rates in the hepatic artery, tobe compared with the experimental measurements represented in Fig. 8. Given the intersubject variability, we considered that we reached a qualitative and quantitative agreement sufficient for our purpose. The influence of partial hepatectomy on these hepatic artery waveforms is then studied by simulating partial hepatectomy in the model. In pig surgery, partial ablation is done in two steps. In a first stage, two of the three liver lobes are removed, corresponding to approximately 75% ablation. Part of the remaining lobe is removed in a second step to reach a final ablation around 90%. The percentages of ablation are based on initial liver volume. Inthemodel, C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 117

(a) Before partial hepatectomy. (b) After 75% hepatectomy.

Fig. 11. Pressure GSF of hepatic arterial trees for the resistance and the capacitance. the first stage is taken into account by dynamically increasing the corresponding lobe resistances and decreasing the corresponding capacitances, to simulate the 75% hepatectomy:

R if t < T C if t < T R(t) = 75 C(t) = 75 (31) R exp(5(t − T75)) otherwise C exp(−5(t − T75)) otherwise where T75 is the time instant of the clamping, and R and C are the values before clamping. Then, to simulate the second part of the surgery, the remaining lobe mass is decreased, simulating a larger ablation resulting in a 90% hepatectomy. The remaining lobe mass is given by:  M if t < T90 M(t) =  r  (32) M 1 − r + otherwise ,  1 + exp(−5(t − T90)) where M is the initial lobe mass, T90 is the time of the second clamp, and r is the percentage removed in the remaining lobe. The parameters for the simulated hepatectomies are given in Table 4. Typical changes in waveform occurring at 75% hepatectomy have been observed in several animals; Fig. 8 shows four examples. The pressure amplitude goes up (between 5 and 10 mmHg). For the flow rate, although there is some variability in the pre-hepatectomy shape, after the clamping two characteristic changes can be observed besides the mean value decrease: the first peak is sharper, meaning the second peak is lower than before hepatectomy, and diastolic flow is at low values for longer. By contrast, no major changes in waveform have been observed for75% to 90% hepatectomy, as shown in the experimental curves for two different pigs in Fig. 9. Apart from a small mean flow decrease, in some pigs such as pig 4, the flow rate minimum that follows systole becomes lower thantheflowin diastole. The two hepatectomies are simulated one after the other with the 1D–0D model. Fig. 10 shows the simulated pressure and flow rate in the hepatic artery. For the 75% hepatectomy, the increase of pressure is well captured bythe model and the typical changes of the flow waveform are well reproduced. For the simulated 75% to 90% hepatectomy, the pressure does not change and a small decrease in the mean and minimum flow appear (Fig. 10(b)) as in the experimental curves (Fig. 9). Thus, the 1D–0D model is in good agreement with the experimental observations before and after clamping, both for 75% and 75% to 90% hepatectomies.

3.2. Discussion

It is quite remarkable that the 1D–0D model can predict the pressure and flow rate waveform changes for both 75% and 75% to 90% hepatectomies. This may be an indication that the waveforms are related to the liver architecture. 118 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125

(a) Before partial hepatectomy. (b) After 75% hepatectomy.

Fig. 12. Flow rate GSF of hepatic arterial trees for the resistance and the capacitance.

(a) Hepatic artery simulated pressure. (b) Hepatic artery simulated flow.

Fig. 13. Simulated pressure and flow in the hepatic artery, with reduced capacitance and increased resistance in each of the 3 lobes (blue)andwith previous parameters after the simulated 75% hepatectomy (red). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

To further understand this link, parameter sensitivity analysis can help to explain the changes in pressure amplitude and in flow waveform during the 75% hepatectomy. Generalized sensitivity functions (GSFs) analysis help identify correlations between parameters and the distribution over time of the information on parameters contained into the model outputs. Generalized sensitivity functions definitions are given in Appendix A.2. Details on sensitivity analysis can be found in [42–45]. By definition, a GSF starts at value zero and ends at value one. The increase in-between is not necessary monotonic; if important correlations between parameters exist oscillations occur. The time interval where the sharpest increase occurs is when most information on the parameter is contained into the model output. The GSF is computed before and after the simulated 75% hepatectomy. The GSFs of the total resistance and capacitance for flow and pressure in the hepatic arterial trees are plotted in Figs. 11 and 12. Before hepatectomy, pressure and flow are sensitive to resistance during the entire cardiac cycle. This result is expected as resistance impacts mean pressure and flow. The pressure is more sensitive tothe capacitance during its rising phase. The sharper increase of capacitance GSF after 75% hepatectomy indicates that the pressure amplitude is especially sensitive to capacitance. Between before and after 75% hepatectomy, capacitance is divided by approximately four. Before 75% hepatectomy, the flow is sensitive to resistance and capacitance during C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 119

Table 1 Parameter for heart, lungs and vena cava 0D models. Ea is the contraction function amplitude, Eb the contraction function baseline, Tc is the duration of contraction, Tr is the duration of relaxation, tc and tr are the times when the atria begin to contract and relax, respectively and V0 is the unstressed volume of the chamber. Rp and Rd are the proximal and distal resistances of the RCR model and C is the capacitance. Heart chamber Right atrium Right ventricle Left atrium Left ventricle 5 Ea (dyn/cm ) 80 750 200 1600 5 Eb (dyn/cm ) 110 100 400 350 Tc (s) 0.145 0.289 0.145 0.289 Tr (s) 0.145 0.128 0.145 0.128 tc (s) 0.68 – 0.68 – tr (s) 0.824 – 0.824 – 3 V0 (cm ) 4 10 4 5 5 5 5 Rp (dyn s/cm ) Rd (dyn s/cm ) C (cm /dyn) Lungs 53 53 0.03 Vena cava 10 10 0.004

Fig. 14. 1D arteries node number and arteries id; see Table 2 for their parameter values. the entire cardiac cycle. After 75% hepatectomy, the flow is more sensitive to the capacitance between 0.3 and 0.45s, corresponding to the sharp decrease in the flow curve. Thus, the change in parameters due to the 75% hepatectomy – resistance increases by around 75% and capacitance decreases by around 75% – seems to explain the changes in pressure and flow waveforms. To confirm this hypothesis, the pre-hepatectomy model is run but with hepatic artery resistance and capacitance parameters multiplied and divided by four respectively, as if each lobe was 75% smaller. The new simulations are compared with the previous ones after the 75% hepatectomy, in Fig. 13. Contrarily to pressures, the flow rates differ. Therefore, the change of global parameter values – total liver resistance and capacitance – can explain the change in pressure amplitude but it is not enough to obtain the sharp change observed in the flow waveform. 120 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125

Table 2 Parameters for the 1D vessels of the pig cardiovascular model. The first column contains the id number of the artery (Fig. 14), the second the name 2 of the artery, the third l its length in cm, the fourth its cross-section area A0 in cm from CT-scan, the fifth is the vessel elasticity coefficient β (defined in Eq. (14)) divided by the fluid density ρ, and the last one the number of elements used to discretize the vessel. 2 2 Id Name l (cm) A0 (cm ) β/ρ (cm/s ) Nel a Ascending aorta 3.5 2.54 2.07 105 35 b Brachiocephalic trunk 2.92 0.46 4.88 105 30 c Aortic arch A 0.36 0.39 5.34 105 5 d Right subclavian 8.5 0.20 7.96 105 85 e Right common carotid 11.8 0.12 1.24 106 120 f Left common carotid 11.9 0.15 9.79 105 120 g Aortic arch B 0.68 2.30 2.17 105 7 h Left Subclavian 12 0.31 6.01 105 120 i Thoracic aorta A 9.1 2.06 2.30 105 91 j Thoracic aorta B 9.5 1.43 2.75 105 95 k Thoracic aorta C 9.5 0.81 3.66 105 95 l Celiac trunk 0.66 0.29 6.22 105 7 m Hepatic artery 5 0.10 1.45 106 50 n Splenic artery 12.8 0.10 1.45 106 130 o Abdominal aorta A 1.7 0.80 3.69 105 17 p Mesenteric artery 3 0.36 5.57 105 30 q Abdominal aorta B 3.55 0.80 3.69 105 36 r Right Renal 3.65 0.18 8.63 105 37 s Abdominal aorta C 0.5 0.78 3.72 105 5 t Left renal 1.37 0.24 6.95 105 14 u Abdominal aorta D 8 0.57 4.36 105 80 v Right iliac 2.9 0.29 6.33 105 29 w Left iliac 2.8 0.34 5.74 105 28

The fact that changes in hepatic artery flow waveform during experiments are observed for 75% hepatectomy but not for 90% hepatectomy, can be explained by the change in architecture in the blood vessel trees. Indeed, in the first hepatectomy, two of the three liver lobes are removed, which leads to an important architecture change. For the second hepatectomy, the remaining lobe mass is decreased and the vessel tree architecture does not change as much. In the model, the simulation of the first stage corresponds to an impedance change from 3 RCRs in parallel to asingleRCR. For the second stage, the impedance remains the one of a single RCR; only the remaining lobe model parameters are changed, due to mass proportionality assumptions. In summary, the modeling choices linking the liver resistances and capacitances to the mass and to the lobar structure of the liver allowed us to reproduce the changes in the experimentally observed signals. Thus, the study and reproduction of hepatectomy with a model enable us to better understand experimental observations and propose a novel link between architecture and flow. Monitoring waveform changes during post-hepatectomy regeneration could thus be a surrogate for the underlying architectural changes, which are currently not possible to non-invasively quantify.

4. Conclusions

In this work, the kinetic scheme, mainly used for the Saint-Venant equations in the literature, was successfully adapted to blood flow models. This scheme proved to have a very good behavior for arterial and venous benchmark tests. In particular, its theoretical properties of positivity make it especially well adapted to simulate collapsible vessels. The scheme was then used to simulate complex behaviors occurring during liver surgeries. First, an idealized test representing the unclamping of the portal vein was proposed. Then the effects of partial hepatectomies on the hepatic artery pressure and flow waveforms were studied with a 1D–0D closed-loop model. Interestingly, the changes observed experimentally on pigs were correctly captured for different percentages of hepatectomy. To the best of our C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 121

Table 3 Parameters for the outlets RCR models. The first column contains the id of the artery, the second node id, the third Rp the proximal resistance, the fourth Rd the distal resistance, and the last one the capacitance. 5 5 5 Id Out Rp (dyn s/cm ) Rd (dyn s/cm ) C(cm /dyn) n 15 953 8584 2.0 10−5 p 16 864 7780 4.0 10−5 r 17 1189 10705 2.0 10−5 t 18 1196 10762 2.0 10−5 v 19 1655 1490 8.0 10−5 w 20 165 14907 8.0 10−5 h 21 1069 9622 6.0 10−5 f 22 1288 11592 4.0 10−5 e 23 1311 11806 4.0 10−5 d 24 1063 9566 6.0 10−5

Table 4 Liver 0D model parameters. First the different lobe masses are given and the ratio between proximal and total resistances in each lobe for the hepatic artery RCR model. Then, mass resistance and mass capacitance are given for hepatic artery tree, portal vein tree, tissue and hepatic vein tree, followed by the clamping parameters. r is the right lobe resected mass %.

Liver lobes Right lobe Middle lobe Left lobe Mass (g) 250 500 180 Rp/Rtot arterial tree 0.1 0.5 0.1 Arterial tree Portal vein tree Tissue + Hepatic veins

Mass resistance (g dyn s/cm5) 1.9 107 1.86 105 7.44 104 Mass capacitance (cm5/dyn/g) 3.0 10−8 – 1.5 10−5

Hepatectomy parameters 75% clamping time T75 (s) 8 90% clamping time T90 (s) 16 r (%) 35 knowledge, these experimental observations were never reported before in the literature. The capability of the model to represent this complex behavior allowed us to propose possible explanations of the observed phenomenon. Future work will be devoted to a finer characterization of the change of the liver architecture during hepatectomy, and to the adaptation of the model to humans.

Acknowledgments

This material is based upon work supported by the French National Agency for Research ANR-13-TECS-0006 iFLOW. The authors gratefully acknowledge Dr. Damiano Lombardi for assistance with implementation of the 1D models, and Dr. Jacques Sainte-Marie for his expertise in kinetic schemes. The authors are grateful to the INRA Plateforme CIRE (Nouzilly, France) staff for their technical assistance in surgeries and imaging, and to Mylene` Wartenberg for assistance in taking measurements.

Appendix

A.1. Parameter and error tables

In this appendix, tables of the 1D and 0D model parameters are summarized, along with precise errors for benchmark test results as referred to in the text. 122 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125

Table 5 1D–0D closed-loop simulation parameters, time and space discretization parameters, initial conditions for 1D part, P0, friction parameter and fluid density values.

Time step (s) 10−5 Mesh size (cm) 0.1 Initial pressure p(x, t = 0) (dyn cm−2) 6.0 104 Initial velocity u(x, t = 0) (cm/s) 5 −2 4 P0 (dyn cm ) 6.6 10 2 K f (cm /s) 3 ρ (g cm−3) 1.05

Table 6 Parameters for the single pulse propagation benchmark test case from [20]. Properties Values Length L 10 m 2 Cross-section area A0 π cm Initial velocity u(x,0) 0 cm/s Initial pressure P(x,0) 0 dyn cm−2 −2 Pressure P0 0 dyn cm Wall thickness h 0.15 cm Young’s modulus E 4.0 105 dyn cm−2 Elasticity parameter β 4.515 105 dyn cm−3 Blood mass density ρ 1.05 g cm−3 Blood viscosity µ 0 or 0.04 dyn cm−2 µ Friction term K f 22π ρ

Table 7 Normalized errors for the single pulse propagation test case; for cases 1, 2, 3 friction is neglected and case 4 is the viscous blood case. The = ∥ − ∥ ∥ ∥ = ∥ − ∥ ∥ ∥ normalized errors are defined by El2 Xkin X STM l2 / X STM l2 and E∞ Xkin X STM ∞/ X STM ∞ where X STM is the solution with the 1D STM scheme from [20] and Xkin is the solution obtained with the kinetic scheme. Case 1 presents the results of the first order kinetic scheme with ∆x = 0.1 cm, and ∆t = 1.0 10−4 s, case 2 of the first order kinetic scheme with ∆x = 0.01 cm, and ∆t = 1.0 10−5 s, case 3 of the first order in time and second order in space kinetic scheme with ∆x = 0.1 cm, and ∆t = 1.0 10−5 s, finally case 4 of the first order in timeand second order in space kinetic scheme with ∆x = 0.1 cm and ∆t = 1.0 10−5 s with a non-zero friction term.

Case 1 Case 2 Case 3 Case 4 Instants El2 E∞ El2 E∞ El2 E∞ El2 E∞ t = 0.1 s 0.036 0.043 0.017 0.007 0.007 0.012 0.007 0.012 t = 0.3 s 0.145 0.166 0.060 0.024 0.028 0.047 0.029 0.047 t = 0.5 s 0.230 0.251 0.104 0.040 0.049 0.073 0.049 0.077 t = 0.7 s 0.292 0.314 0.147 0.055 0.067 0.099 0.067 0.101 t = 0.9 s 0.343 0.364 0.189 0.069 0.083 0.121 0.083 0.122 t = 1.1 s 0.384 0.402 0.229 0.083 0.099 0.140 0.097 0.141 t = 1.3 s 0.419 0.438 0.267 0.096 0.111 0.158 0.111 0.158 t = 1.5 s 0.448 0.465 0.305 0.109 0.124 0.173 0.122 0.174

A.2. Generalized sensitivity functions

For those unfamiliar with the GSFs, we recall their definition [42,45]. Consider the model for the state vector x = [x1, x2,..., xL ]:

x˙i (t) = fi (t, x, θ) i = 1, 2,..., L (33) C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 123

Table 8 Parameters for the aortic bifurcation benchmark test case from [20]. Properties Aorta Iliac Length L 8.6 cm 8.5 cm 2 2 Cross-section area A0 2.3235 cm 1.131 cm Initial velocity u(x,0) 0 cm/s 0 cm/s Initial pressure P(x,0) 0 dyn cm−2 0 dyn cm−2 4 −2 4 −2 Pressure P0 9.46 10 dyn cm 9.46 10 dyn cm Wall thickness h 0.1032 cm 0.072 cm Young’s modulus E 5.0 105 dyn cm−2 7.0 105 dyn cm−2 Elasticity parameter β 4.671 105 dyn cm−3 9.3728 105 dyn cm−3 Blood mass density ρ 1.06 g cm−3 1.06 g cm−3 Blood viscosity µ 0.04 dyn cm−2 0.04 dyn cm−2 µ µ Friction term K f 22π ρ 22π ρ 2 −5 Windkessel proximal resistance Rp – 6.8123 10 dyn s cm 4 −5 Windkessel distal resistance Rd – 3.1013 10 dyn s cm Windkessel capacitance C – 3.6664 10−5 cm5 dyn−1

Table 9 Error for the aortic bifurcation test case with respect to 3D solution in percent as defined in [20].

Error Midpoint Aorta End point Aorta Midpoint Iliac RMS EP 0.39 0.42 0.45 RMS EQ 0.93 1.17 0.53 RMS E∆r 2.41 3.99 4.21 MAX EP 0.67 0.78 0.9 MAX EQ 2.81 3.64 2.09 MAX E∆r 3.87 6.74 7.25 SYS − − − EP 0.46 0.64 0.77 SYS − − − EQ 2.51 3.51 1.56 SYS − − − E∆r 3.72 6.61 7.03 DIAS EP 0.4 0.45 0.46 DIAS EQ 1.15 1.74 1.05 DIAS − − − E∆r 1.42 1.95 2.37

where θ = [θ1, θ2, . . . , θP ] is the model parameters vector and the dynamic model is represented with functions fi . The observation vector z = [z1,..., zM ], can be written as :

z(tn) = h(tn, θ) + ϵ(tn) n = 1, 2,..., N with h(t, θ) = H(x(t, θ)) (34) where H is the observation operator, h represents the noise-free observation vector and the vector ϵ(tn) is the noise on measurements at time tn. The noise vectors are assumed to be independent for all measurement times. Moreover, 2 all components of the noise vector are assumed independent with zero mean and σi (tn) variance associated with measurement zi (tn). The generalized sensitivity function for the parameter θk is defined by :

n M     1 g (t ) = (M−1∇ h (t , θ )) (∇ h (t , θ )) . k n 2 θ j i 0 k θ j i 0 k (35) i=1 j=1 σ j (t j ) 124 C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125

The sensitivity is computed around a reference parameter vector θ 0. The matrix M denotes the Fisher information matrix, defined as

N M   1 T M = ∇ h (t , θ ) ∇ h (t , θ ) 2 θ j i 0 θ j i 0 (36) i=1 j=1 σ j (t j )

References

[1] S. Truant, O. Oberlin, G. Sergent, G. Lebuffe, L. Gambiez, O. Ernst, F.-R. Pruvot, Remnant liver volume to body weight ratio 0.5%: a new cut-off to estimate postoperative risks after extended resection in noncirrhotic liver, J. Amer. College Surgeons 204 (1) (2007) 22–33. [2] H. Boxenbaum, Interspecies variation in liver weight, hepatic blood flow, and antipyrine intrinsic clearance: extrapolation of datato benzodiazepines and phenytoin, J. pharmacokinetics Biopharmaceutics 8 (2) (1980) 165–176. [3] W.W. Lautt, Regulatory processes interacting to maintain hepatic blood flow constancy: vascular compliance, hepatic arterial buffer response, hepatorenal reflex, liver regeneration, escape from vasoconstriction, Hepatology Research 37 (11) (2007) 891–903. [4] T.J. Hughes, J. Lubliner, On the one-dimensional theory of blood flow in the larger vessels, Math. Biosci. 18 (1) (1973) 161–170. [5] C. Guiot, P. Pianta, C. Cancelli, T. Pedley, Prediction of coronary blood flow with a numerical model based on collapsible tube dynamics, Amer. J. Physiol.-Heart Circ. Physiol. 258 (5) (1990) H1606–H1614. [6] M.S. Olufsen, Structured tree outflow condition for blood flow in larger systemic arteries, Am. J. Physiol. 276 (1 Pt 2) (1999) H257–H268. [7] L. Formaggia, D. Lamponi, A. Quarteroni, One-dimensional models for blood flow in arteries, J. Engrg. Math. 47 (3–4) (2003) 251–276. [8] S. Sherwin, L. Formaggia, J. Peiro, V. Franke, Computational modelling of 1D blood flow with variable mechanical properties and its application to the simulation of wave propagation in the human arterial system, Internat. J. Numer. Methods Fluids 43 (6–7) (2003) 673–700. [9] I.E. Vignon, C.A. Taylor, Outflow boundary conditions for one-dimensional finite element modeling of blood flow and pressure wavesin arteries, Wave Motion 39 (4) (2004) 361–374. [10] K.S. Matthys, J. Alastruey, J. Peiro,´ A.W. Khir, P. Segers, P.R. Verdonck, K.H. Parker, S.J. Sherwin, Pulse wave propagation in a model human arterial network: assessment of 1-D numerical simulations against in vitro measurements, J. Biomech. 40 (15) (2007) 3476–3486. [11] J. Alastruey, A.W. Khir, K.S. Matthys, P. Segers, S.J. Sherwin, P.R. Verdonck, K.H. Parker, J. Peiro,´ Pulse wave propagation in a model human arterial network: assessment of 1-D visco-elastic simulations against in vitro measurements, Journal of biomechanics 44 (12) (2011) 2250–2258. [12] X. Wang, O. Delestre, J.-M. Fullana, M. Saito, Y. Ikenaga, M. Matsukawa, P.-Y. Lagree,´ Comparing different numerical methods for solving arterial 1D flows in networks, Comput. Methods Biomech. Biomed. Eng. 15 (Suppl.) (2012) 61–62. [13] J. Keijsers, C. Leguy, W. Huberts, A. Narracott, J. Rittweger, F. van de Vosse, A 1D pulse wave propagation model of the hemodynamics of calf muscle pump function, Int. J. Numer. Methods Biomed. Eng. 31 (7). [14] L. Aslanidou, B. Trachet, P. Reymond, R.A. Fraga-Silva, P. Segers, N. Stergiopulos, A 1D model of the arterial circulation in mice, Altex 33 (1) (2016) 13. [15] L. Formaggia, A. Quarteroni, A. Veneziani, Cardiovascular Mathematics Modeling and Simulation of the Circulatory System, Vol. 1, Springer, 2009. [16] A.H. Shapiro, Steady flow in collapsible tubes, J. Biomech. Eng. 99 (3) (1977) 126–147. [17] T.J. Pedley, B.S. Brook, R.S. Seymour, Blood pressure and flow rate in the giraffe jugular vein, Philos. Trans. R. Soc. Lond. Ser. B Biol.Sci. 351 (1342) (1996) 855–866. http://dx.doi.org/10.1098/rstb.1996.0080. [18] B.S. Brook, T.J. Pedley, A model for time-dependent flow in (giraffe jugular) veins: uniform tube properties, J. Biomech. 35 (1)(2002) 95–107. [19] L.O. Muller,¨ G.I. Montecinos, E.F. Toro, Some issues in modelling venous haemodynamics, in: Numerical Methods for Hyperbolic Equations: Theory and Applications. An International Conference to Honour Professor EF Toro, 2012, pp. 347–354. [20] E. Boileau, P. Nithiarasu, P.J. Blanco, L.O. Muller,¨ F.E. Fossan, L.R. Hellevik, W.P. Donders, W. Huberts, M. Willemet, J. Alastruey, A benchmark study of numerical schemes for one-dimensional arterial blood flow modelling, Int. J. Numer. Methods Biomed. Eng. 31(10). http://dx.doi.org/10.1002/cnm.2732. [21] E. Marchandise, P. Flaud, Accurate modelling of unsteady flows in collapsible tubes, Comput. Methods Biomech. Biomed. Eng. 13 (2)(2010) 279–290. [22] E. Audusse, M.-O. Bristeau, B. Perthame, Kinetic Schemes for Saint-Venant Equations with Source Terms on Unstructured Grids, Research Report RR-3989, INRIA, Projet M3N, 2000. URL https://hal.inria.fr/inria-00072657. [23] O. Delestre, P.-Y. Lagree,´ A ‘well-balanced’ finite volume scheme for blood flow simulation, Internat. J. Numer. Methods Fluids 72(2)(2013) 177–205. [24] B. Perthame, Kinetic Formulation of Conservation Laws, Vol. 21, Oxford University Press, 2002. [25] E. Audusse, F. Bouchut, M.-O. Bristeau, R. Klein, B. Perthame, A fast and stable well-balanced scheme with hydrostatic reconstruction for shallow water flows, SIAM J. Sci. Comput. 25 (6) (2004) 2050–2065. [26] G.I. Montecinos, L.O. Muller,¨ E.F. Toro, Hyperbolic reformulation of a 1d viscoelastic blood flow model and ader finite volume schemes, J. Comput. Phys. 266 (2014) 101–123. [27] C. Debbaut, D. De Wilde, C. Casteleyn, P. Cornillie, D. Van Loo, L. Van Hoorebeke, D. Monbaliu, Y.-D. Fan, P. Segers, Modeling the impact of partial hepatectomy on the hepatic hemodynamics using a rat model, IEEE Trans. Biomed. Eng. 59 (12) (2012) 3293–3303. [28] H. Ho, K. Sorrell, A. Bartlett, P. Hunter, Blood flow simulation for the liver after a virtual right lobe hepatectomy, in: Medical Image Computing and Computer-Assisted Intervention–MICCAI 2012, Springer, 2012, pp. 525–532. C. Audebert et al. / Comput. Methods Appl. Mech. Engrg. 314 (2017) 102–125 125

[29] P.J. Blanco, R.A. Feijoo,´ A dimensionally-heterogeneous closed-loop model for the cardiovascular system and its applications, Med. Eng. Phys. 35 (5) (2013) 652–667. http://dx.doi.org/10.1016/j.medengphy.2012.07.011. [30] F. Liang, S. Takagi, R. Himeno, H. Liu, Multi-scale modeling of the human cardiovascular system with applications to aortic valvular and arterial stenoses, Med. Biol. Eng. Comput. 47 (7) (2009) 743–755. [31] J.P. Mynard, J.J. Smolich, One-dimensional haemodynamic modeling and wave dynamics in the entire adult circulation, Ann. Biomed. Eng. 43 (6) (2015) 1443–1460. [32] L.O. Muller,¨ E.F. Toro, A global multiscale mathematical model for the human circulation with emphasis on the venous system, Int. J. Numer. Methods Biomed. Eng. 30 (7) (2014) 681–725. [33] B. Lantz, J. Foerster, D. Link, J. Holcroft, Regional distribution of cardiac output: normal values in man determined by video dilution technique, Amer. J. Roentgenolo 137 (5) (1981) 903–907. [34] J.K. Raines, M.Y. Jaffrin, A.H. Shapiro, A computer simulation of arterial dynamics in the human leg, Journal of biomechanics 7 (1) (1974) 77–91. [35] I.E. Vignon-Clementel, C.A. Figueroa, K.E. Jansen, C.A. Taylor, Outflow boundary conditions for 3D simulations of non-periodic blood flow and pressure fields in deformable arteries, Comput. Methods Biomech. Biomed. Eng. 13 (2010) 625–640. http://dx.doi.org/10.1080/ 10255840903413565. [36] J.F. LaDisa, C. Alberto Figueroa, I.E. Vignon-Clementel, H. Jin Kim, N. Xiao, L.M. Ellwein, F.P. Chan, J.A. Feinstein, C.A. Taylor, Computational simulations for aortic coarctation: Representative results from a sampling of patients, J. Biomech. Eng. 133 (2011) 091008. http://dx.doi.org/10.1115/1.4004996. [37] P. Segers, N. Stergiopulos, N. Westerhof, P. Wouters, P. Kolh, P. Verdonck, Systemic and pulmonary hemodynamics assessed with a lumped- parameter heart-arterial interaction model, J. Engrg. Math. 47 (2003) 185–199. [38] D. Lombardi, Inverse problems in 1d hemodynamics on systemic networks: a sequential approach, Int. J. Numer. Method Biomed. Eng. 30 (2) (2014) 160–179. http://dx.doi.org/10.1002/cnm.2596. [39] H. Suga, K. Sagawa, Instantaneous pressure-volume relationships and their ratio in the excised, supported canine left ventricle, Circ. Res. 35 (1) (1974) 117–126. [40] G. Pennati, F. Migliavacca, G. Dubini, R. Pietrabissa, M.R. de Leval, A mathematical model of circulation in the presence of the bidirectional cavopulmonary anastomosis in children with a univentricular heart, Med. Eng. Phys. 19 (3) (1997) 223–234. [41] S. Pant, C. Corsini, C. Baker, T.-Y. Hsia, G. Pennati, I. Vignon-Clementel, Data assimilation and modelling of patient-specific single-ventricle physiology with and without valve regurgitation, Journal of Biomechanics (2015). http://dx.doi.org/10.1016/j.jbiomech.2015.11.030. [42] K. Thomaseth, C. Cobelli, Generalized sensitivity functions in physiological system identification, Ann. Biomed. Eng. 27 (5) (1999) 607–616. [43] P. Bai, H.T. Banks, S. Dediu, A.Y. Govan, M. Last, A.L. Lloyd, H.K. Nguyen, M.S. Olufsen, G. Rempala, B.D. Slenning, Stochastic and deterministic models for agricultural production networks, Math. Biosci. Eng. 4 (3) (2007) 373–402. [44] H. Miao, X. Xia, A.S. Perelson, H. Wu, On identifiability of nonlinear ode models and applications in viral dynamics, SIAM Rev. Soc.Ind. Appl. Math. 53 (1) (2011) 3–39. http://dx.doi.org/10.1137/090757009. [45] S. Pant, B. Fabreges,` J.-F. Gerbeau, I.E. Vignon-Clementel, A methodological paradigm for patient-specific multi-scale cfd simulations: from clinical measurements to parameter estimates for individual analysis, Int. J. Numer. Methods Biomed. Eng. 30 (12) (2014) 1614–1648. http://dx.doi.org/10.1002/cnm.2692. CE: A.U.; ANNSURG-D-16-00144; Total nos of Pages: 9; ANNSURG-D-16-00144

ORIGINAL ARTICLE

Modulating Portal Hemodynamics With Vascular Ring Allows Efficient Regeneration After Partial Hepatectomy in a Porcine Model

Petru O. Bucur, MD, MS,yôô Mohamed Bekheit, MD, MS,y Chloe Audebert, MS,z§ Amnah Othman, DVM,ôjjjj Seddik Hammad, DVM, PhD,ôjj Mylene Sebagh, MD, PhD, Marc-Antoine Allard, MD, MS, Benoıˆt Decante, MS,yy Adrian Friebel, MS,zz Elodie Miquelestorena-Standley, MD,§§ Dirk Drasdo, PhD,zzz Jan G. Hengstler, PhD,jj Irene E. Vignon-Clementel, PhD,z§ and Eric Vibert, MD, PhDy

pressure gradient was lower after hepatectomy in the MID-AVR group (P ¼ Objective: To investigate safety and efficacy of temporary portal hemody- 0.001). Postoperative serum bilirubin levels were lower in the MID-AVR namics modulation with a novel percutaneously adjustable vascular ring group (P ¼ 0.007 at day 5). In the MID-AVR group, the Ki67 index was (MID-AVR) onto a porcine model of 75% hepatectomy. significantly higher on day 3 (P ¼ 0.043) and the architectural derangement Background: Postoperative liver failure is a leading cause of mortality after was lower (P < 0.05). Morphometric quantification of the bile canaliculi major hepatectomy. Portal flow modulation is an increasingly accepted revealed a significantly lower number of intersection branches (P < 0.05) and concept to prevent postoperative liver failure. Nonetheless, the current intersection nodes (P < 0.001) on day 7 compared with the preoperative strategies have shortcomings. specimen, in the control group. These differences were not found in the ring Methods: Resection was performed under hemodynamic monitoring in 17 group. large, white pigs allocated into 2 groups. Eight pigs had ring around the portal Conclusions: MID-AVR is safe for portal hemodynamics modulation. It vein for 3 days with the aim of reducing changes in hemodynamics due to might improve liver regeneration by protecting liver microarchitecture. hepatectomy. Analysis of hemodynamics, laboratory, and histopathological parameters was performed. Keywords: liver regeneration, liver resection, portal flow modulation, Results: Percutaneous inflation, deflation, and removal of the MID-AVR posthepatectomy failure, small-for-size, vascular ring were safe. Two (25%) pigs in the MID-AVR group and 4 (45%) controls died (Ann Surg 2017;xx:xxx–xxx) before day 3 (P ¼ NS). A moderate increase of portal flow rate per liver mass after resection was associated with better survival (P ¼ 0.017). The portocaval he liver function is dependent on the integrity of the micro- T architecture that permits optimum exchange of metabolites 1 From the Unite´ INSERM 1193, Centre He´pato-Biliaire, Villejuif, France; yUnite´ between blood and hepatocytes. After resection, the portal flow INSERM3 1193, Villejuif, France; zINRIA Paris-Rocquencourt, Paris, France; through the sinusoidal network increases leading to stimulation of §Sorbonne Universite´s, UPMC University Paris 6, Laboratoire Jacques-Louis sinusoidal endothelial cells2 and initiation of liver regeneration.3 Lions, Paris, France; ôLeibniz Research Centre for Working Environment and Human Factors (IFADO), TU Dortmund University, Dortmund, Germany; Major hepatectomy or transplantation of small liver are, on the jjDepartment of Forensic Medicine and Veterinary Toxicology, Faculty of contrary, associated with disequilibrium between the portal flow Veterinary Medicine, South Valley University, Qena, Egypt; Pathology rate, which is excessively increased at the sinusoidal level, and the Department, AP-HP, Hoˆpital Kremlin-Bice`tre, Kremlin-Bice`tre, France; liver volume, leading to ‘‘barotrauma.’’4,5 yyCentre Chirurgicale Marie-Lannelongue, Experimental Surgery Unit, Le zz The volume and the quality of the future liver are important Plessis Robinson, France; Interdisciplinary Centre for Bioinformatics 6,7 (IZBI), University of Leipzig, Leipzig, Germany; §§CHRU Tours, Laboratoire determinants for this disequilibrium. Thresholds for portal flow D’anatomie et Cytologie Patholoiques, Tours, France; ôôService de Chirurgie rate per 100 g of liver8 and portal pressure9 were identified above Digestive, CHU Trousseau, Tours, France; jjjjLeibniz Institut fu¨r Analytische which the risk of postoperative liver failure is high.10 This phenom- Wissenschaften - ISAS e.V., Dortmund, Germany; and Molecular Hepatol- ogy - Alcohol Associated Diseases, Department of Medicine II, Medical enon is partially interplayed by the important reduction in the arterial Faculty Mannheim, University of Heidelberg, Germany. flow ‘‘dearterialization’’ of the remnant liver as a consequence of the This study was funded mainly by the ‘‘Agence de la Biomedecine’’ through its excess in portal flow rate through the sinusoidal network.11 program of Research (AOR 2009) and Socie´te´ Francophone de la Transplan- Healthy remnant liver volume superior to 20% of the theor- tation (Bourse IGL 2009). E.V., I.V.-C., D.D., P.O.B., M.B., C.A., and J.H. etical total liver volume12 and/or superior to 0.5% of the body weight acknowledge funding by project ANR-13-TECS-0006 (IFlow), Adrian Friebel 13,14 by Virtual Liver Network (German Bundesministerium fu¨r Bildung und is considered mandatory after major hepatectomy to keep a Forschung [BMBF]). balance between volume and flow to avoid postoperative liver failure. Disclosure: The authors disclose that the percutaneously adjustable vascular ring If more resection is anticipated, which is the case in many liver described in the manuscript is still investigational. Eric Vibert and Petru Bucur were consultants for MID, Dardilly, France. The other malignancies, preoperative portal vein occlusion, by embolization or authors declare no conflict of interest. ligation, might be necessary to induce regeneration of the future Supplemental digital content is available for this article. Direct URL citations liver.15 This approach may, however, increase the risk of cancer appear in the printed text and are provided in the HTML and PDF versions of progression.16,17 this article on the journal’s Web site (www.annalsofsurgery.com). Portal flow modulation was initially applied in living donor Reprints: Eric Vibert, MD, PhD, AP-HP, Hoˆpital Paul Brousse, Centre He´pato- 18 19 Biliaire, 12 Ave Paul VaillantCouturier, 94804 Villejuif Cedex, France. E-mail: liver transplantation. Partial portal flow diversion, splenic artery [email protected]. ligation,20 or splenectomy were proposed to reduce the incidence of Copyright ß 2017 Wolters Kluwer Health, Inc. All rights reserved. postoperative liver failure.21,22 These techniques do not allow precise ISSN: 0003-4932/16/XXXX-0001 DOI: 10.1097/SLA.0000000000002146 control of the portal flow rate and might have adverse effects. For

Annals of Surgery Volume XX, Number XX, Month 2017 www.annalsofsurgery.com | 1

Bucur et al Annals of Surgery Volume XX, Number XX, Month 2017

instance, excessive diversion of flow might have an equally delete- animals received pantoprazole 40 mg/i.v. (Inipomp, Takeda, Puteaux, rious effect on liver regeneration,22 to which it is essential.3 There- France) and enoxaparin 0.2 mL/S.C. (Lovenox, Sanofi Aventis, fore, it would be helpful to use a modulation technique with flexible Gentilly, France) and 0.5 mg/kg/b.i.d. nalbuphine (Nalbuphine and reversible control over the portal hemodynamics to tailor it to the 20 mg/2 mL, Serb, Paris, France). planned remnant volume. The same preoperative measures were repeated at the time of Toward that end, we developed an adjustable vascular ring sacrifice on the 7th postoperative day. Blood samples were collected ‘‘MID-AVR’’ to protect the hepatic microarchitecture from the initial before and after liver resection and on the third, fifth, and the seventh harmful barotrauma. The efficacy of the ring was assessed in terms of postoperative days. survival, liver function tests, liver regeneration, and changes in the microarchitecture. Surgical Procedure

MATERIALS AND METHODS Hemodynamic Measurements Median cervical incision and cannulation of the right Ethical Approval internal jugular vein with an 8 Fr and the right carotid artery with The study was approved by the regional committee of ethics of a 5 Fr Desivalve (Vygon, Ecouen, France) vascular cannula were animal research, and by the French Government authorities, comply- performed. ing with the European Union Directive No 2010/63/EU. Sternotomy was performed to place the TranSonic 20 mm (20 PAX, TranSonic, Ithaca, NY) transit time echo probe around the Animals origin of the ascending aorta for measuring the cardiac output. Subsequently, a midline abdominal incision was performed. Upon All animals received human care according to the criteria dissecting the hepatic hilum, 2 other flow meter probes, 14 and 4 mm, outlined in the ‘‘Guide for the Care and Use of Laboratory Animals’’ were positioned around the portal vein and the hepatic artery, prepared by the National Academy of Sciences and published by the respectively. Portal and vena cava pressures were measured by direct National Institutes of Health.23 Seventeen large white female pigs, puncture with a 24 G needle connected to a built-in electronic which underwent 75% liver resection, were randomized in blocks transducer in the anesthetic monitor. into 2 groups. The control (no MID-AVR) group included 9 animals Flow per unit mass was calculated as the recorded flow rate in and the ring group (MID-AVR) included 8 animals in which the the main portal vein divided by the liver weight multiplied by 100 vascular ring was positioned around the portal vein. The average age (mL/min/100 g of liver tissue), and the whole liver weight was of the included animals was 3 months 9 days and their mean weight estimated based on the fact that the left lateral, the left medial, was 32.9 5.3 kg. and the right medial lobes of the pig liver constitute approximately 75% of the whole liver weight.24 The whole liver weight is calculated Study Setting as the resected liver weight 100/75. Surgeries were performed at the experimental animal surgical unit at the Marie Lannelongue Center, Le Plessis Robinson, France. MID-AVR Positioning Preoperative Preparation The ring is silicone made (Fig. 1A, B); connected to a Animals were left fasting the night before surgery. On the day regulating valve via long tube. The 2 lips of the ring opening were of surgery, animals were given in their individualized cages 30 mg/kg fixed together with a fine nonabsorbable (polypropylene 8/0) suture ketamine (Ketamine, Panpharma, Luitre´, France) and 0.03 mg/kg that is readily broken upon overinflation of the balloon (Fig. 1C) acepromazine (Calmivet, Vetoquinol, Lure, France). (supplemental video 1, http://links.lww.com/SLA/B174 demonstrates the inflation process). The ring was placed around the portal Anesthesia vein (Fig. 1D) in the designated group and calibrated before starting All surgeries were performed under general anesthesia. Each liver resection. pig received 100 mg of xylazine 2% (Rompun, Bayer Healthcare, Once the ring was placed, the balloon was progressively Loos, France) with 750 mg ketamine for anesthesia induction fol- inflated with sterile saline solution with 0.1 mL steps. At each step lowed by tracheal intubation (6–7 mm in size, Portex, Smiths the flow rate in the portal vein and the pressure below and above the Medical, Rungis, France). Subsequently, inhalational anesthesia ring were measured during 4 to 5 minutes to ensure the stability of the effect. The target portal flow rate in the MID-AVR group was 50% of was started using a 60% FiO with 2% isoflurane (Isoflurane Bela- 2 its initial value, to limit the increase in the portal flow per the remnant mont, Piramal Healthcare UK, Morpeth, UK, France) in assisted 25 ventilation. liver mass after a 75% liver resection to around twofold. Pancuronium bromide (Pavulon, MSD France, Courbevoie, At the end of the surgery, the valve was fixed subcutaneously France), at a rate of 0.3 mg/kg/h and fentanyl (Fentanyl Janssen, on the xiphoid process with nonabsorbable sutures for percutaneous Janssen-Cilag, Issy-les-Moulineaux, France 100 mg/2 mL), at a rate control and extraction. of 5 mg/kg/h were continuously perfused intravenously. Crystalloid fluids were given at a rate of 2 mL/kg/h fasting in addition to 500 to Liver Resection 1000 mL, which was increased as required. At the end of surgery, Resection of the left lateral, left medial, and the right medial the wound was infiltrated with ropivacaı¨ne 150 mg (Naropeine hepatic lobes was done leaving in place the right lateral and the 7.5 mg/mL Astra Zeneca, Courbevoie, France). caudate lobes. During surgery animals were covered with heat blankets and At the end of the procedure, a central venous catheter was gastric aspiration through an orogastric tube was attempted if gastric placed in the internal jugular vein for postoperative fluid adminis- distension was observed. tration and blood samples withdrawal. Biopsies from the remnant Cefotaxime 1 g (cefotaxime, Mylan, Saint Priest, France) and liver lobe were taken before and 1 hour after liver resection. Pleuro- gentamicin 80 mg (Gentalline, MSD France, Courbevoie, France) mediastinal suction drain was placed at the end of the procedure and were given intramuscularly once a day for 5 days. In addition, removed on the first postoperative day.

Annals of Surgery Volume XX, Number XX, Month 2017 Modulating Portal Hemodynamics With Vascular Ring

FIGURE 1. The MID-AVR in its different shapes according to the degree of balloon inflation. A, The ring is closed while the balloon is completely deflated. B, The ring is closed while the balloon is inflated with small amount of saline. C, The ring tends to open while the balloon is inflated with large amount of saline. D, The MID- AVR is placed around the portal vein and the balloon on the inner surface is inflated with a small amount of saline, the portal vein shows moderate constriction.

Fifteen-minute Indocyanine Green Retention% each time point. An experimented pathologist (M.S.) assigned (ICG-R15) blindly all the values that were used for the scores. A second After resection, ICG-R15 was measured in serum from arterial pathologist (E.M.S.) familiar with pig liver pathology reviewed samples taken before injection of 0.5 mg/kg ICG (Infracyanine, independently all samples and established another set of values. SERB, Paris, France) and 2, 4, 8, and 15 minutes after the injection. Proliferation Index and 3D Morphometric Day 3 Postoperative Quantification of the Bile Canaliculi An ultrasound guided liver biopsy using an 18 Fr needle was Five-micron sections of formalin-fixed paraffin-embedded taken. In the MID-AVR group, removal of the ring was performed by liver tissues were processed according to previously published reopening the midline incision in the first 3 animals, to ensure that the protocols.27 Likewise, to three-dimensionally reconstruct and ana- ring was completely open and that the anchoring sutures ruptured. lyze the bile canalicular network, 100 mm liver slices were immu- Subsequently, it was removed through a small percutaneous incision nostained. Subsequently, Z-stacks (n ¼ 6–9 per group) were captured over the valve. by a 60-fold objective using a confocal laser scanning microscope (FV1000, Olympus, Hamburg, Germany). Image preprocessing was Sacrifice carried out by Autoquant-X3 Version X3.03 64 Bit Edition (Bit- On the seventh postoperative day, animals were sacrificed plane). Image segmentation for quantification of bile canaliculi was following a similar protocol to that at the day of surgery. The remnant achieved by TiQuant (www.msysbio.com/tiquant) using a voxel size liver was weighted after euthanasia. of 0.207 0.207 0.54 mm28 (Supplemental movies 2–4, http:// links.lww.com/SLA/B175, http://links.lww.com/SLA/B176, http:// Histological Analysis links.lww.com/SLA/B177). For morphometric quantification, bile Pathological scores were given to specimens in standard canaliculi were classified into branches and nodes (footnote supple- hematoxylin-eosin and Masson trichrome stains based on criteria mentary Table 4, http://links.lww.com/SLA/B173). adapted from Demetris et al.26 Five criteria were formulated: A ¼ cellular proliferation in clusters, B ¼ sinusoidal dilatation, Statistical Analysis C ¼ ballooning, D ¼ ductular proliferation, and E ¼ inflammatory After normality testing, summary of data was represented in necrosis. Each item is graded from 0 ¼ no change, 1 ¼ mild, mean standard deviation, median, and range or percentages 2 ¼ moderate, to 3 ¼ severe, in samples taken from deep and super- according to variable type. Odds ratio was reported to compare ficial liver tissue for a total score between 0 and 30 for the liver at mortality between both groups.

Bucur et al Annals of Surgery Volume XX, Number XX, Month 2017

Man-Whitney U test and chi-square tests were used to assess the difference between the control and the ring groups for non- parametric variables, whereas t test was used to calculate parametric variables. ANOVA with repeated measures and Friedman tests were used to compare the evolution of parameters for parametric and nonparametric data, respectively. Significance threshold was set at a P value of 0.05. MedCalc Statistical Software version 14.8.1 (Med- Calc Software bvba, Ostend, Belgium) was used for the analysis.

RESULTS The experimental design was set to study the effect of MID-AVR application, compared with no application, on liver regeneration after 75% resection in 2 groups of large white pigs. Hemodynamic, laboratory, and histopathological parameters were analyzed at different time points as indicated in the methods. Both groups had similar baseline parameters (supplementary Table 1, http://links.lww.com/SLA/B173). Ring Safety Application of the ring was easy and its removal was safe in all FIGURE 2. Scatter diagram showing animal mortality in both animals. In the first 3 animals, removal under visual control dem- groups stratified according to the change in portal flow per unit onstrated the efficiency of the overinflation to open the ring. Percu- liver mass. taneous removal in the subsequent animals was similarly safe and successful. Patency of the portal vein was confirmed in the surviving the contrary, 5 animals (71%) died when these values were outside animals by ultrasound on day 3 and by direct visualization at the range (P ¼ 0.017) (odds ratio ¼ 22, P ¼ 0.02) (Fig. 2). sacrifice. The main difficulty was closing the ring with such a fine Autopsy was performed for all premature deaths, and it was suture. But this was necessary for the opening of the ring by over negative for macroscopic explanation for the mortality. None of the inflation. pigs showed portal vein thrombosis. Survival Hemodynamic Measurements Six pigs (75%) in the MID-AVR group and 5 pigs (55.6%) in Portal flow rate after the ring placement, but before hepatec- the control group survived till day 7 (P ¼ 0.62). When portal flow per tomy, was 80 mL/min/g 100 g, 45% lower than its baseline value unit liver mass at the end of surgery was, however, within the range of (142 mL/min/g 100 g). The hemodynamics measured at the end 2.2 to 4 folds its baseline, only 1 animal died prematurely (10%). On of hepatectomy are presented in Table 1. In the ring group, the

TABLE 1. Different Parameters Measured After Liver Resection and on Day 7 Postoperative MID-AVR Group (n ¼ 8) Control Group (n ¼ 9) P Portal flow, mL/min 565 164 544 122 ns Portal flow for remnant liver mass, mL/min/g 100 g 372 97 359 95 ns Arterial flow, mL/min 92 49 64 60 ns Cardiac output, L/min 2.6 0.8 2.3 0.9 ns Central venous pressure, mm Hg 5.9 2.0 4.9 1.6 ns Portal pressure, mm Hg 7.5 2.4 9.3 2.3 ns Portocaval gradient, mm Hg 1.63 1.3 4.4 1.5 0.001 AST, IU/mL 169 60 197 110 ns ALT, IU/mL 37.3 8.1 42 6ns Platelets, 103/mL 348 101 312 31 ns Analyzed parameters at animal sacrifice Animal weight, kg 30 3.8 30 6.0 ns Liver weight, gy 507 77 478 37 ns Portal flow, mL/min 710 224 698 190 ns Portal flow per unit liver mass, mL/min/g 100 g 149 54 93 79 ns Hepatic arterial flow, mL/min 151 61 93 79 ns Cardiac output, L/min 2.7 1.2 2.8 1.3 ns Central venous pressure, mm Hg 4.7 1.7 4.8 1.1 ns Portal pressure, mm Hg 8.2 1.3 9.4 0.9 ns Portocaval gradient, mm Hg 3.5 1 4.6 1.7 ns % 15 min ICG retention 28 18 38 7ns AST, IU/mL 83 61 67 50 ns ALT, IU/mL 72 30 66 17 ns Platelets, 103/mL 375 30 435 133 ns ALT indicates alanine aminotransferase; AST, indicates aspartate aminotransferase. The weight of the remnant liver was estimated based on the fact that resection of the left and median lobes is nearly equal to 75%. yWhole liver weight is the sum of the resected liver weight and the calculated liver weight.

Annals of Surgery Volume XX, Number XX, Month 2017 Modulating Portal Hemodynamics With Vascular Ring

on day 5, than in the control group (3.8 vs 6.6 mmol/L, P ¼ 0.007; Fig. 4A). We observed slightly higher prothrombin activity values, after hepatectomy (Fig. 4B) in the ring group with no significant difference (P ¼ ns). For the rest of the parameters, no differences were observed. Table 1 summarizes the main findings after resection and at sacrifice.

Indocyanine Green Retention Percentage ICG-R15 was not significantly different between groups. The average retention percentage was, however, slightly lower on day 7 in the ring group compared with the control group (Fig. 4C).

Histopathological Results The overall changes in the total histopathological scores were significantly different between both groups (P ¼ 0.004). These differences were mainly in specimens after resection and on day 3 (P < 0.05 in both time points). These differences remained significant when we used the mean of values from our 2 pathologists (supplemental Tables 2 and 3, http://links.lww.com/SLA/B173). Sinusoidal dilatation was significantly lower in the ring group compared to the control at the end of surgery (P < 0.01) (Fig. 5 A–F). The Ki67-positive cells increased by 14.3% [95% confidence interval (CI) ¼ 10.8%–17.9%] at postoperative day 3 compared to before surgery and decreased again on day 7 with a difference in the activity of 1.5% (95% CI ¼2.8 to 0.2%) compared to baseline. Between the 2 groups, the percentage of Ki67-positive cells was higher on day 3 (P ¼ 0.043) in the ring group. The means of mitotic figures were not statistically different between groups, even if there were slightly less mitoses per field in the ring group on day 3 (1.14 vs 2, P ¼ 0.12) and more on day 7 (3.17 vs 1.8, P ¼ 0.5). In 3D morphometric quantification of the biliary structures, the number of intersection branches and the number of intersections nodes were significantly lower in the control group (but not in the FIGURE 3. A, Portocaval pressure gradient was significantly ring group) on day 7 compared to the values before resection [mean higher in the control group compared to the MID-AVR group difference ¼ 162 branch, 95% CI ¼ 44–279 intersection branch and only after hepatectomy ( P < 0.01). B, The liver weight before mean difference ¼ 130 node, 95% CI ¼ 13–248 intersection node and after resection and the regain on day 7 were not signifi- (supplemental Table 4, http://links.lww.com/SLA/B173)]. The cantly different between both groups. cumulative length of bile canaliculi was significantly lower on day 7 than that in preoperative specimen in both groups (P ¼ portocaval pressure gradient was significantly lower than in the 0.02) (Fig. 6C, D; supplemental movies 2–4, http://links.lww. control group (P < 0.01) (Fig. 3A). In a stepwise multiple regression com/SLA/B175, http://links.lww.com/SLA/B176, http://links.lww. analysis, which included the changes in the parameters of systemic com/SLA/B177). circulation and the presence or absence of the ring, only the last variable was a significant predictor of the change of portocaval DISCUSSION pressure gradient (P ¼ 0.001), whereas the systemic hemodynamic The present study presents the effects of a novel, adjustable, changes were not significant predictors. There was no difference in and less invasive technique for portal hemodynamics modulation on the pressure gradient at the postoperative day 7 between the 2 groups liver regeneration after 75% hepatectomy in a porcine model. (ie, 4 days after ring removal). The application of the ring was safe and was associated with better hepatic function represented in the lower bilirubin levels Changes in Liver Weight during the fifth postoperative day and in the less histological The estimated residual liver weight was not significantly derangement and changes in the temporal proliferation pattern different between the control and the ring groups (155 32 and indicated by a higher Ki67 index on day 3, whereas the liver mass 151 25 g, respectively, P ¼ 1). The residual liver weight increased on day 7 in both groups did not differ. Application of the ring was not significantly on day 7, P < 0.0001 (507 77 and 478 37 g in the easy, especially the closure with the fine suture. Its utilization seemed ring and the control groups, respectively; Fig. 3B). There was no safe in pigs, but liver position is different in humans and changes to significant difference in the liver weight between both groups at the adapt its position and avoid conflicts with other hepatic pedicle day of sacrifice, but the estimated gain was slightly higher in the ring elements might be necessary. group with no statistically significant difference (mean difference ¼ The proliferative activity seemed to occur later in the ring 353 68 vs 323 36 mL, respectively). group (more Ki67 activity and slightly less mitoses on day 3 and slightly more mitoses on day 7). Together with the higher patho- Laboratory Results logical scores these results indicate that the control group had a The MID-AVR was a significant influential factor on the temporally different cellular proliferation and a deranged micro- bilirubin level in repeated measures ANOVA (P ¼ 0.033). Post- architecture that could be attributed to the effect of barotrauma. operatively, serum bilirubin level was lower in the MID-AVR group Sinusoidal barotrauma is thought to be a leading mechanism of

Bucur et al Annals of Surgery Volume XX, Number XX, Month 2017

FIGURE 4. A, Total bilirubin level was higher in the control group than in the MID-AVR group, particularly on postoperative day 5 (P < 0.05). B, Prothrombin activity is not significantly different between both groups. C, The indocyanine green retention test (ICG) at 15 minutes was, after resection or on day 7, similar in the control group and in the MID-AVR group.

postoperative liver failure.5,20 In our study, each one of the animals The higher portocaval pressure gradient, which is a manifes- in the ring group with their portal flow within this range of 2.2- to tation of increased microvascular resistance,29 explains the more 4-folds survived. evident changes in the microarchitecture in the control group despite In the present study, the majority of animals at the end of that there was no difference in the calculated portal flow per unit surgery in both groups had an estimated portal flow per unit mass mass. Furthermore, the higher bilirubin and ICG-R15 level, which within the target range. The ring group, however, showed lower are typical manifestations of portal hyperperfusion,30 could be portocaval pressure gradient, which indicates that the intrasinusoidal attributed to the higher intrasinusoidal pressure in this group. Alone, pressure was lower in ring group. Indeed, the flow in ring group after the high bilirubin level was reported to be a sufficient laboratory ring positioning does not reflect the intraparenchymal resistance feature for the diagnosis of posthepatectomy liver failure.31 The or pressure because it reflects the resistance imposed by the ring aforementioned changes imply that the better hepatic function in positioning, unlike the case in the control group. The parenchyma in the ring group was attributed to the maintained microarchitectural the ring group continues to receive a lower flow compared with the integrity in this group. control group, which leads to lower intrasinusoidal pressure, smaller The relation between the portal pressure and the portal flow portocaval pressure gradient, and lesser architectural damage. There- rate is not strictly linear owing to the interplay of the hepatic fore, we conclude that the ring application helped in protecting the sinusoidal capacitance, which means that it would usually require hepatic microarchitecture in the initial phase after resection. more than doubling of the portal flow through the sinusoidal network

Annals of Surgery Volume XX, Number XX, Month 2017 Modulating Portal Hemodynamics With Vascular Ring

FIGURE 5. Biopsies taken from the control (left panel) and ring (right panel) groups: A, H&E, 200, day 7 after surgery, bridging necrosis, thick cords forming pseudonodules. B, H&E, 400, day 7 after surgery, cholangioductal proliferation. C, Trichrome, 100, day 7, dilatation and hemorrhagic destruction of the sinusoids with surrounding thick parenchymal cords. D, H&E, 200, day 7, normal architecture. E, H&E, 400, day 7, normal aspect portal pedicle with mild arterial dilatation. F, Trichrome, 100, day 7, conserved architecture with normal thickness cords and portal spaces.

for the portal pressure to slightly increase.32 Given the capacitance of by the small for flow syndrome.37,38 It targets the early protection of the liver as a reservoir for blood,33 the absence of increased porto- the architecture from the barotrauma inflicted by high portal flow caval pressure gradient in the ring group reflects the lower intra- during the inductive angiogenesis phase.2 sinusoidal pressure in spite of the relatively higher portal flow Several techniques have been proposed for modulation of the calculated from flow rate proximal to the ring in this group. portal inflow. Diversion of the portal flow through a partial porto- An increase in portal flow per liver mass is thought to be caval shunt would require a second intervention for the closure of the necessary for the stimulation of liver regeneration22 and is associated shunt.39 Splenectomy offered more benefit compared to the portal with increased cellular proliferation through proliferative gene flow diversion in terms of liver regeneration.38 This could be expression and apoptotic gene downregulation.34 On the contrary attributed to the negative effect of the excessive diversion on if there is an excessive increase in portal flow, liver functions are regeneration.40 Splenic artery embolization, as an alternative, deranged35 and suppression of liver regeneration paradoxically exposes the spleen to the risk of infarction.41 Those alternatives dominates, leading to liver failure.36 do not have the dynamic and minimal invasive advantages offered by The liver weight gain was lower in the control group than in the reversible modulation with an adjustable vascular ring, which the ring group. Furthermore, in the 3D quantification of biliary could be tailored to the extent of the planned liver resection. structures, all morphometric parameters had lower values in both In the present study, the application of the ring prevented the groups compared with specimens taken before resection, indicating a increase in portocaval pressure gradient and allowed a moderate higher hepatocyte density, as supported by the interpretation of the increase in portal flow per unit liver. This effect resulted in protection pattern of Ki67 index. These parameters were higher in the ring of the remnant parenchyma without compromising liver regener- group than in the control group implying that the regeneration in the ation. Comparing the 2 groups, only a few parameters, however, ring group was more balanced and organized, which might explain reached the statistical significance. the better hepatic function in this group. The full potentials of this novel technique could be more Hepatic inflow modulation is becoming an increasingly evident in a situation in which the increase in portal flow per unit accepted strategy for the reduction of the initial damage caused mass is higher. Here, the goal was to first test the ring safety in a

Bucur et al Annals of Surgery Volume XX, Number XX, Month 2017

FIGURE 6. Analysis of hepatocyte proliferation in regenerating pig livers and reconstruction and quantification of bile canalicular network in regenerating pig livers. A, There are increased numbers of Ki-67 labeled hepatocyte nuclei on day 3 compared to day 0, which decreased again on day 7 (scale bars are 50 mm). Overview images are included in the supplemental Figure 1, http://links. lww.com/SLA/B173.B,Quantificationof Ki-67 positive hepatocyte nuclei in 5 fields per specimen. Five to six pigs were used per time point, namely, 0, 3, and 7 days after partial hepatectomy. C, Examples of reconstructed pig livers. Blue (left column): nuclei, green (middle column): bile canalicular network and merge (right column). The corresponding reconstructions of control, 7 days after partial hepatectomy with MID- AVR and 7 days after partial hepatectomy withoutMID-AVRareshowninSupple- mental movies 2, 3, and 4, http://links. lww.com/SLA/B175, http://links.lww. com/SLA/B176, http://links.lww.com/ SLA/ B177, respectively. Scale bars are 50 mm. D, Length of bile canalicular network in a given volume is shown. The bile canaliculi length is not influenced by application of MID-AVR ring. P < 0.05 and P < 0.01.

nonextreme situation to have a better chance to see its effect on day 3 8. Vasavada BB, Chen CL, Zakaria M. Portal flow is the main predictor of early graft dysfunction regardless of the GRWR status in living donor liver and day 7. The present study demonstrates the high safety profile and transplantation—a retrospective analysis of 134 patients. Int J Surg. 2014; the potential efficacy of the MID-AVR. Therefore, we started a 12:177–180. human clinical trial (phase I/II) registered at clinicaltrials.gov under 9. Allard M-A, Adam R, Bucur P-O, et al. Posthepatectomy portal vein pressure the number NCT02390713. predicts liver failure and mortality after major liver resection on noncirrhotic liver. Ann Surg. 2013;258:822–829. CONCLUSIONS 10. Golse N, Bucur PO, Adam R, et al. New paradigms in post-hepatectomy liver failure. J Gastrointest Surg. 2013;17:593–605. The adjustable vascular ring ‘‘MID-AVR’’ applied around the 11. Michalopoulos GK. Liver regeneration after partial hepatectomy: critical portal vein is a safe, precise, reversible, and efficient mean to protect analysis of mechanistic dilemmas. Am J Pathol. 2010;176:2–13. the hepatic microarchitecture during the initial phase of liver regen- 12. Kishi Y, Abdalla EK, Chun YS, et al. Three hundred and one consecutive eration. It seems to delay slightly liver regeneration in a preserved extended right hepatectomies: evaluation of outcome based on systematic liver microarchitecture environment, which might result in better hepatic volumetry. Ann Surg. 2009;250:540–548. function over the course of regeneration. 13. Truant S, Oberlin O, Sergent G, et al. Remnant liver volume to body weight ratio >or ¼ 0.5%: a new cut-off to estimate postoperative risks after extended resection in noncirrhotic liver. J Am Coll Surg. 2007;204:22–33. REFERENCES 14. Mullin EJ, Metcalfe MS, Maddern GJ. How much liver resection is too much? 1. Reichen J. The role of the sinusoidal endothelium in liver function. News Am J Surg. 2005;190:87–97. Physiol Sci. 1999;14:117–121. 15. Van Lienden KP, Van den Esschert JW, De Graaf W, et al. Portal vein 2. Ding B-S, Nolan DJ, Butler JM, et al. Inductive angiocrine signals from embolization before liver resection: a systematic review. Cardiovasc Intervent sinusoidal endothelium are required for liver regeneration. Nature. 2010;468: Radiol. 2013;36:25–34. 310–315. 16. Hoekstra LT, Van Lienden KP, Doets A, et al. Tumor progression after 3. Niiya T, Murakami M, Aoki T, et al. Immediate increase of portal pressure, preoperative portal vein embolization. Ann Surg. 2012;256:812–817. reflecting sinusoidal shear stress, induced liver regeneration after partial 17. Al-Sharif E, Simoneau E, Hassanain M. Portal vein embolization effect on hepatectomy. J Hepatobiliary Pancreat Surg. 1999;6:275–280. colorectal cancer liver metastasis progression: lessons learned. World J Clin 4. Glanemann M, Eipel C, Nussler AK, et al. Hyperperfusion syndrome in small- Oncol. 2015;6:142–146. for-size livers. Eur Surg Res. 2005;37:335–341. 18. Troisi R, De Hemptinne B. Clinical relevance of adapting portal vein flow in 5. Asencio JM, Vaquero J, Olmedilla L, et al. Small-for-flow’’syndrome: shifting living donor liver transplantation in adult patients. Liver Transpl. 2003;9: the ‘‘size’’ paradigm. Med Hypotheses. 2013;80:573–577. S36–S41. 6. Eshkenazy R, Dreznik Y, Lahat E, et al. Small for size liver remnant following 19. Hou P, Chen C, Tu Y-L, et al. Extracorporeal continuous portal diversion plus resection: prevention and management. Hepatobiliary Surg Nutr. 2014;3: temporal plasmapheresis for ‘‘small-for-size’’ syndrome. World J Gastro- 303–312. enterol. 2013;19:5464–5472. 7. Adams RB, Aloia TA, Loyer E, et al. Selection for hepatic resection of 20. Lo C-M, Liu C-L, Fan S-T. Portal hyperperfusion injury as the cause of colorectal liver metastases: expert consensus statement. HPB (Oxford). primary nonfunction in a small-for-size liver graft-successful treatment with 2013;15:91–103. splenic artery ligation. Liver Transpl. 2003;9:626–628.

Annals of Surgery Volume XX, Number XX, Month 2017 Modulating Portal Hemodynamics With Vascular Ring

21. Kim J, Kim C-J, Ko I-G, et al. Splenectomy affects the balance between 32. Lautt WW. Regulatory processes interacting to maintain hepatic blood flow hepatic growth factor and transforming growth factor-b and its effect on liver constancy: vascular compliance, hepatic arterial buffer response, hepatorenal regeneration is dependent on the amount of liver resection in rats. J Korean reflex, liver regeneration, escape from vasoconstriction. Hepatol Res. Surg Soc. 2012;82:238–245. 2007;37:891–903. 22. Marubashi S, Sakon M, Nagano H, et al. Effect of portal hemodynamics on 33. Kjekshus H, Risoe C, Scholz T, et al. Methods for assessing hepatic distending liver regeneration studied in a novel portohepatic shunt rat model. Surgery. pressure and changes in hepatic capacitance in pigs. Am J Physiol Heart Circ 2004;136:1028–1037. Physiol. 2000;279:H1796–H1803. 23. Council NR. Guide for the Care and Use of Laboratory Animals. 8th ed. 34. Mueller L, Broering DC, Meyer J, et al. The induction of the immediate-early- Washington: National Academies Press; 2007. genes Egr-1, PAI-1 and PRL-1 during liver regeneration in surgical models is 24. Court FG, Laws PE, Morrison CP, et al. Subtotal hepatectomy: a porcine related to increased portal flow. J Hepatol. 2002;37:606–612. model for the study of liver regeneration. J Surg Res. 2004;116:181–186. 35. Nobuoka T, Mizuguchi T, Oshima H, et al. Portal blood flow regulates volume 25. Hessheimer AJ, Fondevila C, Taura´ P, et al. Decompression of the portal bed recovery of the rat liver after partial hepatectomy: molecular evaluation. Eur and twice-baseline portal inflow are necessary for the functional recovery of a Surg Res. 2006;38:522–532. ‘‘small-for-size’’ graft. Ann Surg. 2011;253:1201–1210. 36. Pan N, Lv X, Liang R, et al. Suppression of graft regeneration, not ischemia/ 26. Demetris AJ, Kelly DM, Eghtesad B, et al. Pathophysiologic observations and reperfusion injury, is the primary cause of small-for-size syndrome after histopathologic recognition of the portal hyperperfusion or small-for-size partial liver transplantation in mice. PLoS One. 2014;9:e93636. syndrome. Am J Surg Pathol. 2006;30:986–993. 37. Vasavada B, Chen CL, Zakaria M. Using low graft/recipient’s body weight 27. Hammad S, Hoehme S, Friebel A, et al. Protocols for staining of bile ratio graft with portal flow modulation an effective way to prevent small-for- canalicular and sinusoidal networks of human, mouse and pig livers, three- size syndrome in living-donor liver transplant: a retrospective analysis. Exp dimensional reconstruction and quantification of tissue microarchitecture by Clin Transplant. 2014;12:437–442. image processing and analysis. Arch Toxicol. 2014;88:1161–1183. 38. Umeda Y, Yagi T, Sadamori H, et al. Effects of prophylactic splenic artery 28. Friebel A, Neitsch J, Johann T, et al. TiQuant: software for tissue analysis, modulation on portal overperfusion and liver regeneration in small-for-size quantification and surface reconstruction. Bioinformatics. 2015;31:3234–3236. graft. Transplantation. 2008;86:673–680. 29. Bosch J, Groszmann RJ, Shah VH. Evolution in the understanding of the 39. Taniguchi M, Shimamura T, Suzuki T, et al. Transient portacaval shunt for a pathophysiological basis of portal hypertension: how changes in paradigm are small-for-size graft in living donor liver transplantation. Liver Transpl. leading to successful new treatments. J Hepatol. 2015;62(1 suppl):S121–S130. 2007;13:932–934. 30. Iguchi K, Hatano E, Yamanaka K, et al. Hepatoprotective effect by pretreat- 40. Wang X-Q, Xu Y-F, Tan J-W, et al. Portal inflow preservation during portal ment with olprinone in a swine partial hepatectomy model. Liver Transpl. diversion in small-for-size syndrome. World J Gastroenterol. 2014;20: 2014;20:838–849. 1021–1029. 31. Rahbari NN, Garden OJ, Padbury R, et al. Post-hepatectomy haemorrhage: a 41. Troisi R, Hesse UJ, Decruyenaere J, et al. Functional, life-threatening dis- definition and grading by the International Study Group of Liver Surgery orders and splenectomy following liver transplantation. Clin Transplant. (ISGLS). HPB (Oxford). 2011;13:528–535. 1999;13:380–388.

Available online at www.sciencedirect.com ScienceDirect

journal homepage: www.JournalofSurgicalResearch.com

Letter to the Editor The ideal porcine model for major liver resection: is there any yet?

Dear Editor, Hereafter, there will be an increased portal pressure that its We read with interest the letter to the editor by Dr Antonios degree is influenced by the sinusoidal network volume; hence Athanasiou and et al. The letter was in response to our the residual hepatic volume. Since both the artery and the published study on a simplified technique for major liver vein drain into the same space, the relative increase in the resection in a porcine liver.1 The study aimed at describing a portal flowdgiven its larger volume per unit of liver mass simple technique for 75% and 90% porcine liver resection. It is compared to the arterial flowdwill lead to reduction in the a pleasure to reply to the authors’ comments as we believe it is arterial flow in what is known as the hepatic artery buffering a good practice to openly exchange ideas and thoughts on system.2 Our observations are in general line with that, and it published studies. also depicts the relative difference of the effect of two Dr Athanasiou titled his letter as “The ideal porcine model different partial resection volumes on the hepatic for major liver resection”. It is clearly known that there is no hemodynamics. ideal when it comes to surgery and ideal remains only theo- In terms of simplicity of the technique, we have tried every retical. Scientists try iteratively to get closer to an appropriate other technique mentioned in the letter and more: Kelly, method rather than a perfect one. finger fraction, cavitron ultrasound surgical aspirator, and The authors attempted to criticize the methods and our other energy devices. The parenchymal dissection is not the interpretation of the results from different angles. First, the same as in human, and it is slightly more difficult with authors disagree with our observation of the hemodynamic ordinary tools adapted for human tissue. The main difference changes noted during the experiments. The summary of our is probably due to the different lobular architecture and findings were that in 75% resection, the portal flow was distribution of fibrous septa.5 reduced after resection by 8.13 28%. However, the portal Moreover, based on our experience that was built in years pressure increased by 20.1 51%. The hepatic artery flow of experiments, the young large white pigs do not tolerate the decreased by 63.86 26.3% as well as the pressure by 5 28%. pedicle clamping except for a brief time. Furthermore, the For 90% resection, the portal flow decreased by 33.6 12.6%, Pringle is prevented by our model of hemodynamic mea- and the pressure increased by 104 58%. The hepatic artery surement. Therefore, we needed to find a quicker and a more flow decreased by 88 7%, and the pressure decreased by simple way for liver resection than that described by Hori 5 14.8%. These statements are in a perfect line with the body et al.6 Individualized pedicle ligation within the parenchyma of literature concerned with this matter. was found well supported, and scissor cutting of the paren- Unfortunately, the authors quoted our results wrong stat- chyma was much quicker and aided by selective Rummel ing “This study concludes that the described technique is clamping to reduce the blood loss. simple and easily applicable for extended hepatectomy in The hemodynamic assessment and the postoperative care porcine model. Furthermore, the portal vein and hepatic protocol were included in the article, which were overlooked by artery flow decreased more after 90% hepatectomy than after the authors. We decided to include the histopathology data in a 75%, whereas the portal vein and hepatic artery pressure subsequent study focused on liver regeneration, and because it increased more following 90% compared to 75% resection.” would be less useful for the technical assessment, it was not Their interpretation to our data contradicts with our direct included herein. We know that the published technical articles statement concerning the arterial pressure and contradicts to focus on the technical aspects, and information further to that the physiological expectation that was demonstrated in is not usually included. This is because the article is devised to previous studies.2-4 describe a different way of performing a procedure with a Partial liver resection leads to increased flow/mass ratio, known effect.7 The situation is different from evaluating a which in turn alters the resistance to a point where the different physiological concept, which will definitely require capacitance of the hepatic sinusoidal plexus is saturated. such type of data as in the authors’ own study.8

DOI of original article: http://dx.doi.org/10.1016/j.jss.2016.11.022 282 letter to the editor

We used two-tailed paired t-test to assess the changes in the insufficient remnant or liver transplantation with a small-for- hemodynamic parameters following 75% resection, assuming size graft. Surg Today 2014;44:2201e7. the normality in this group, whereas Wilcoxon signed rank test 7. Court FG, Laws PE, Morrison CP, et al. Subtotal hepatectomy: a porcine model for the study of liver regeneration. J Surg Res was used for the same purpose in 90% resection. 2004;116:181e6. The difference in the number of experiments in both types 8. Athanasiou A, Papalois A, Kontos M, et al. The beneficial of resection has its justification within the project, which role of simultaneous splenectomy after extended could not be fully expressed in the article due to disclosure hepatectomy: experimental study in pigs. J Surg Res 2017; protected experiments. Moreover, we think that there is no 208:121e31. impact of this concealment on the presented data analysis and interpretation. Our model, all animals, was subjected to Mohamed Bekheit, MBChB, MSc Chir, different intraoperative evaluations (e.g., indocyanine green- MSc miniChir, MSc Biostat, MRCS, MRCPS, PhD* ebased experiments, intravital focal microscopy, blood com- Department of Surgery, Specialist in HPB and Minimal Invasive ponents testing.), which would render the survival rate Surgery, HPB Surgery Unit, Aberdeen Royal Infirmary, UK irrelevant to or independent from the technique. Department of Surgery, El Kabbary General Hospital, Egypt Honorary Clinical Associate, University of Aberdeen, UK references INSERM Unite 1193, Paul Brousse Hospital, Villejuif, France

Petru Bucur, MD, PhD 1. Bucur P, Bekheit M, Audebert C, et al. Simplified technique for INSERM Unite 1193, Paul Brousse Hospital, Villejuif, France 75% and 90% hepatic resection with hemodynamic monitoring Centre Hospitalier Universitaire, Tours, France in a large white swine model. J Surg Res 2017;209:122e30. 2. Lautt WW. Regulatory processes interacting to maintain hepatic blood flow constancy: vascular compliance, hepatic Eric Vibert, MD, PhD arterial buffer response, hepatorenal reflex, liver regeneration, INSERM Unite 1193, Paul Brousse Hospital, Villejuif, France escape from vasoconstriction. Hepatol Res 2007;37:891e903. Department of Liver Surgery and Transplantation, University of 3. Rocheleau B, Ethier C, Houle R, et al. Hepatic artery buffer Paris-Sud, Paul Brousse Hospital, Villejuif, France response following left portal vein ligation: its role in liver tissue homeostasis. Am J Physiol 1999;277:G1000e7. *Corresponding author. Specialist in HPB and Minimal Invasive 4. Smyrniotis V, Kostopanagiotou G, Kondi A, et al. Hemodynamic Surgery, Department of Surgery, Aberdeen Royal Infirmary, interaction between portal vein and hepatic artery flow in small- Foresterhill Health Campus, AB25 2ZN, e for-size split liver transplantation. Transpl Int 2002;15:355 60. Aberdeen, UK. Tel.: þ44 (0)12245 50519. 5. Ekataksin W, Wake K. Liver units in three dimensions: I. E-mail address: [email protected] Organization of argyrophilic connective tissue skeleton in porcine liver with particular reference to the “compound hepatic lobule”. Am J Anat 1991;191:113e53. 0022-4804/$ e see front matter 6. Hori T, Yagi S, Okamua Y, et al. How to successfully resect 70% ª 2017 Elsevier Inc. All rights reserved. of the liver in pigs to model an extended hepatectomy with an http://dx.doi.org/10.1016/j.jss.2017.03.004