Liver Resection Modeling (LRM)

Introduction:

Surgical management of liver cancers is increasingly complex, with often extensive and iterative resections on pathological liver. This attitude is made possible by good selection and perioperative management of patients, as well as the regenerative capacity of the liver. However, even if extensive resections are routinely performed, the limit remains the risk of postoperative liver failure (PLF), especially in patients with chronic liver disease. Despite many known risk factors, some of which being preventable, PLF and / or postoperative decompensation (ascites) remain common complications (incidence > 5%) and PLF is one of the leading causes of postoperative death.

The difficulty lies in the choice of a treatment adapted to the oncological needs, with the balance between what is technically feasible and what will be functionally and metabolically tolerated. This balance is based on a surgical evaluation, based on objective factors (liver volumetry and biological tests for example) and on the surgeon's experience. This estimate is imperfect, the proof being the mortality at 3 months after hepatectomy which remains high, 5 to 7%. Moreover, the current criteria lead to an often-inappropriate selection. It therefore remains very difficult for a given patient (particularly cirrhotic) to accurately predict the risk of postoperative decompensation and some patients may sometimes be undertreated for fear of decompensation. In these patients, it is therefore a loss of chance.

It would be obviously possible to further improve these results and numerical tools must find their place in the medico-surgical algorithm. The investigators think that hemodynamic simulation could be used as a decision-making tool.

After hepatectomy, the post-resection portocaval gradient is one of the most relevant reflections for understanding the hemodynamic conditions of a liver and the risk of PLF but it is only available intraoperatively, after completion of the surgical procedure, and is therefore not used as a tool for selecting candidates for surgery. Moreover, for a given patient, it is not currently possible to anticipate the variation of portocaval gradient and therefore the risk of occurrence of PLF.

The hypothesis based on experimental results in animals, consists in proposing a model of intrahepatic and systemic hemodynamics pre / postoperative scale 0D, that is to say in the form of simplified electrical circuit. The investigators can thus predict the evolution of the portocaval gradient in order to better select patients who are candidates for liver resection.

Description of Acts and Proceedings Added by Research To perform 0D modeling on human data to predict evolution of post-resection hepatic/systemic flows and pressures.

• Intraoperative flow measurement: installation of a certified CE marked (93/42EEC) and certified ISO9001 / EN46001 sensor around the portal vein and / or the hepatic artery during the procedure in order to estimate blood flow. Acquisition time: 5 min / patient / measurement (1 pre-resection measurement, 1 post-resection measurement). No injection or vascular puncture will be necessary for this acquisition. The risk of trauma is almost zero (no reported cases). The probe will be sterilized between two patients.
• MRI flow: addition of a specific acquisition sequence performed during a "standard" MRI indicated in the preoperative assessment. No intravascular injection required, no irradiation (magnetic field only). Acquisition time: about 5 minutes. MRI performed in routine (unless contraindicated) to all patients operated for a hepatectomy for positive assessment (tumor characterization) and tumor extension.

Potential Benefits There is no individual benefit because no clinical decision will be made during this study on the findings of the specific examinations carried out. In addition, 0D simulations will be performed after surgery, retrospectively.

On the other hand, there is a collective benefit expected from a positive study because the surgeons will then have additional tools to develop their decisional algorithm and adapt their technical gesture. Benefits = avoid cases of PLF, limit the risk of postoperative death.

The foreseeable risks for patients are considered minimal because:

• Flow measurement is a routine procedure for some teams, with no reported risk since a simple probe / vessel contact is required. No vascular puncture or possible trauma. Constraint: additional anesthesia duration = 5 min / measurement
• The flow MRI has no specific risk since it is simply an "extra" acquisition sequence, performed following the other sequences, without injection of contrast medium. Constraint = MRI duration extended by approximately 5 minutes.
• The extension of the anesthesia, due to the measurement of the flow measurement, is considered as risk-free by our team of anesthesiologists because it is about surgical operations lasting usually> 6 hours, with a cardio-vascular monitoring and systematic neurological. No specific anesthetic drugs will be administered in addition to the usual operative / anesthetic practice.

Experimental plan Prognostic and prospective research (prediction of the occurrence of a potentially lethal risk for the patient).

This study will be non-comparative and uncontrolled because at the actual stage of development, The investigators only want to apply and transpose to the human digital tools of decision support (pilot study).

Description of the statistical methods planned.

Calculation of the number of subjects needed:

This study is comparable to a phase II study. In order to evaluate the predictive reliability of the portocaval gradient, it is necessary to proceed according to a Simon mini max design in two stages. The feasibility hypotheses are p0 = 70%, p1 = 90%, risk α = 5%, power = 90%. In the first stage, 18 patients will be included and analyzed. If the reliability (portocaval gradient predicted with a margin of error +/- 4 mmHg) is not objectified in at least 13 patients, the recruitment will be stopped and The investigators will rework the algorithm in depth, for example by modifying the input data (= clinical-biological parameters introduced in the model). The investigators will test the new version of the model on the same 18 first patients until validation in at least 13 patients.

If ≥13 patients have a reliable prediction, then 14 additional patients will be recruited. The total number will be 32 patients. To be considered potentially interesting, at least 26 patients will have to have a reliable prediction at +/- 4 mmHg ".

All statistical analyzes will be performed with STATA version 15.1 software (StataCorp LLC, TX, USA). A statistical analysis plan will be drafted and finalized before closure of the study that is to say the freezing of the base. The Statistical Analysis Plan will provide all detailed analyzes of the primary and secondary endpoints.

All statistical tests will be bilateral with a risk of the first species at 5%.

Raw analysis:

The raw analysis will be intent-to-treat analysis including recruited patients.

Methods planned for the analysis:

Main analysis:

Analysis of the primary endpoint will be made by intent to treat. The reliability analysis will be performed by a Chi 2 or Fisher exact test if the number is less than 5. No adjustment will be made.

Secondary analysis:

Secondary analyses will be performed in the intent-to-treat population. The binary criteria will be analyzed by a Chi 2 or Fisher exact test if the number is less than 5, the quantitative criteria by a paired t test.

Methods of analysis of missing data:

No analysis is planned for missing data Quantitative data will be expressed on average (+/- SD). A Mann-Whitney test will be used for continuous data and a Chi-2 (or Fisher if applicable) test for categorical data.