279219 Optimum Perfusion Duration for Machine-Perfused Rat Livers

Monday, October 29, 2012: 1:42 PM
Westmoreland East (Westin )
Sinem Perk1, Maria-Louisa Izamis2, Herman Tolboom3, Basak Uygun2, Francois Berthiaume4, Martin L. Yarmush2 and Korkut Uygun2, (1)Center for Engineering in Medicine, Massachusetts General Hospital, Harvard Medical School, Shriners Hospital for Children, Boston, MA, (2)Center for Engineering in Medicine (CEM) at Massachusetts General Hospital Harvard Medical School - Shriners Hospital for Children, Boston, MA, (3)Division of Cardiac and Vascular Surgery, University Hospital Zurich, Zurich, Switzerland, (4)Department of Biomedical Engineering, Rutgers, The State University of New Jersey, Piscataway, NJ

With the crisis in donor organ availability, recent studies on liver preservation focus on new, dynamic modalities such as machine perfusion to recover discarded donor organs to a functional state. Compared to static cold storage, machine perfusion provides oxygen and nutrition enabling recovery of the organ. Despite years of research however, little is known about key factors such as how to evaluate the degree of recovery during perfusion and what the ideal perfusion duration is to maximize the graft recovery and transplantation success. Development of such approaches are complicated by the fact that every donor liver is different and therefore it is very likely that the optimum perfusion time, as well as other operating parameters and specific treatments, may be different for every organ.

It is well known that the existing organ preservation systems are limited in duration that the success rate falls dramatically in time. While many studies demonstrate that hours of ex-vivo machine perfusion are necessary for resuscitation of a DCD organ, it is also well known that organ function deteriorates if it is kept in perfusion for too long. Accordingly, the organ viability and transplant success probability likely goes through an optimum, before which the recovery from ischemia is incomplete, and beyond which the organ begins to deteriorate due to damage by the perfusion system.

The aim of this work is the development of an approach that enables assessment of liver viability during perfusion, the prediction of clinical outcome post-transplantation, and the determination of the optimum duration of perfusion.

The ideal method in perfused graft viability assessment would be based on online measurements that can be performed accurately and quickly such as Blood Gas Analyses. The dynamic patterns of organ function during perfusion are evaluated and compared between successful and unsuccessful transplants in a manner that allows a statistical analysis of transplant success, providing an estimate for the likelihood of graft success to the surgeon as a decision support system for transplantation.

With the on-line data obtained during machine perfusion, a suitable methodology for such dynamic organ viability assessment is statistical process monitoring (SPM) based on partial least squares (PLS) methodology and its extension for dynamic processes: multi-way PLS (MPLS). Constructing a statistical model of liver dynamic response to perfusion enables in-silico predictions of the liver trajectory during perfusion. These predictions of the trajectory of the organ in the following hours can then be used to detect any emerging trends that indicate deterioration of organ viability, and therefore enable identifying the optimum perfusion duration, and potentially other operational variables, for each organ based on its function on-line during perfusion.

This work presents a novel approach and a two-step MPLS/PLS algorithm for the prediction of perfused liver survival, based on measurements taken during perfusions of rat livers for 6 and 12 hours, with recipients of the first group surviving and latter group mostly failing. Via an online MPLS analysis, blood gas data taken hourly is used to estimate the post-transplantation survival rates dynamically during perfusion, which are compared with the actual post-transplantation clinical outcomes for each animal. The results demonstrate that with the proposed methodology, graft viability can be monitored and transplant success rate can be predicted online during perfusion with cross-validated sensitivity and specificity over 91%.

Extended Abstract: File Not Uploaded