Reducing the risk of post-hepatectomy liver failure using a novel MRI method to predict post-operative function.

Liver cancer is often a fatal disease, and can only be cured via a surgical procedure called "hepatectomy", where the part of the liver containing the cancer is removed. In order to decide whether this surgery is possible, pictures obtained by a magnetic resonance imaging (MRI) scanner are used to work out the amount of healthy liver tissue that will remain after surgery. The challenge is to remove all of the cancerous tissue, but leave enough liver behind to keep the patient alive. If the remaining liver is thought to be too small then a procedure called "portal vein embolisation" can be applied, which will cause the remaining liver to grow rapidly before the surgery is actually performed. As soon as it has grown enough, the hepatectomy can be performed safely.

Despite procedures like this, about 1 in 10 patients are left after surgery with a liver that is too small. Those patients will suffer from post-operative liver failure, a condition that is associated with severe complications and can be fatal.

We believe that this problem is partly caused by the fact that current estimates of post-operative liver function are sometimes inaccurate. These methods assume that all parts of the liver work equally well, but this is not the case in patients with liver disease where certain parts may be injured and underactive, and others overactive. The effect can be significant: if the remaining liver function is overestimated then it is possible that not enough functioning liver tissue remains after surgery and liver failure may result. Alternatively, if the remaining liver's function is underestimated then the patient may be denied potentially life-saving surgery out of fear for developing postoperative liver failure.

Our group has recently developed a more accurate method to measure the remaining liver's function using standard MRI scanners, using fast imaging and advanced data analysis. The method can be added in patient's routine MRI scans, so there is no additional burden or extra scans for the patient. The function of the future remaining liver can be calculated from this data using a dedicated software tool developed in our group. This can then be used to decide whether the patient can undergo surgery, and whether portal vein embolisation is needed.

The objective of this study is to prove that this new imaging technique can have a significant impact on treatment decisions and eventually also patient outcome. We will do this by imaging patients that are considered for surgery with our new technique, and comparing this against the current method of volume measurement. These patients will be followed up after surgery to see if those that develop liver failure are exactly those where the remaining liver's function was overestimated. We will also assess how treatment decision can be improved in the future if our new technique is used for surgery planning. We will work out how many patients can benefit from our new technique, and how exactly this could be used in the future to improve the patient's outcome.

If this study confirms that our new method predicts the risk of post-operative liver failure, then this could lead to an improvement in patient outcomes in the future. In that case the next step would be to propose improved surgery planning using this new information, and test these in clinical trials. Eventually this may lead to a significant increase in the success rate of hepatectomy and in the number of patients that will be offered treatment with this potentially life-saving operation.

STUDY DESIGN

Prospective, single-center observational cohort study recruiting HCC patients considered for curative liver resection, and patients with colorectal liver metastases (CRLM) considered for major resection. They will be divided into 3 treatment groups: surgery (n=184), PVE followed by surgery (n=15), no surgery (n=47). DCE-MRI will be added to the preoperative MRI, as well as an ICG test on the same day. The PVE group will have an additional MRI after PVE. Patients will be followed up until 90 days after surgery.

DCE-MRI data will not be shared with the clinical care team and processed with blinding to patient history. The processing protocol will follow the same steps as in the pilot study, but a more practical semi-automated workflow will be created for this study. This will include automated segmentation facilitated by image registration for motion correction, and development of an open-access software plugin.

DATA ANALYSIS

Objective 1. The accuracy of DCE-MRI will be determined by comparing total liver function against the reference method ICG, and the results will inform statistical analysis. Functional bias of the FLR will be calculated as in the pilot study (figure 2), and the association with PHLF will be tested by univariable logistic regression analysis. A multivariable analysis will be performed to identify associations with other predictors.

Objective 2. Models will be developed to predict PHLF using cross-validation, including those based on standard logistic regression and generalised additive models to accommodate potential for nonlinearity and interactions. The model will be used to design (1) a risk assessment strategy that maximises PHLF reduction using a minimal panel of biomarkers; (2) a clinical trial.

Objective 3. In group B, function, volume and perfusion of the FLR, occluded segments and the tumor will be determined as a function of time. It will be tested whether temporal patterns are predictive of outcome.

Patients

Patients with primary or secondary liver cancer will benefit in several ways. Those that have been selected for surgery will have a lower risk of developing liver failure, increasing life expectancy and quality of life. A certain number of patients that are currently considered inoperable will be offered this potentially curative intervention. Conversely, some patients that are now considered operable but cannot survive on their remnant liver will be saved the burden of a heavy procedure and the ensuing complications of liver failure. The project will allow us to estimate the numbers in each of those groups.

Healthcare providers and NHS

A recent German study estimated the cost of a patient with PHLF at 60,000 GBP, that is 41,000 GBP more than patients with regular recovery. The cost of our novel functional imaging test is estimated at less than 100 GBP per patient (additional scan time + postprocessing). This means the healthcare provider will save almost 41k GBP per patient that does not go into PHLF. For the Leeds Teaching Hospitals NHS Trust (170 major resections per year), cutting the number of patients with PHLF by 50% amounts to a total annual saving of 488,000 GBP. Extrapolated to the population of the UK this leads to a projected 43m GBP annual savings for the NHS.

Surgeons

The study will provide liver surgeons with a new set of indices to be used in support of patient selection and surgery planning. If the study confirms that these new functional indices are able to improve the prediction of PHLF, they will be integrated in clinical trials and eventually in surgical routine. By providing a direct and accurate measurement of post-operative function they will simplify the decision process by removing confounding factors. As they can be integrated in routine staging MRI, they eliminate the need for separate diagnostic procedures to determine overall liver function, such as the 60min bedside LiMAx test or ICG measurements.

Radiologists and radiographers

Patients will be referred to radiology units, who will provide the test results as a new service. Data acquisition and data analysis will mostly be performed by radiographers, who will require dedicated training in slice positioning and software usage. Radiologists will be involved in quality control of the data and results, identification of segmental boundaries, lesions and incidental findings. They will integrate these tasks in their reporting on standard staging MRI.

Software developers (SME)

Integrating the new method in clinical practices will require the development of novel commercial software, and a service for customer support and user training. This will create a new market for independent software developers, usually SME's who provide software solutions for medical imaging (examples in the UK are www.bioxydyn.com and www.imageanalysis.org.uk). Moreover, the prototype software and data produced in the project will reduce their production costs and speed up development.

MRI scanner manufacturers (Siemens, General Electric, Philips)

They can also tap into this new market by providing integrated acquisition + software solutions for the clinical workflow. Compared to software from third party providers, this has the benefit of providing a tight integration between acquisition and analysis, albeit only for a single provider.

Contrast agent manufacturer (Bayer-Schering)

Manufacturers of contrast agents have an interest in quantitative analysis methods as they increase the usage of their product beyond merely visualising. This opens up new applications that scale up the market for the product. The contrast agent in this study is currently in routine use for applications such as liver cancer characterisation, but is not indicated for measurement of segmental liver function. The study will create a new indication in surgery planning and potentially wider applications in chronic liver disease.