Liver paired exchange model aligns donor safety with operating-room realities
Living donor liver transplantation depends on a narrow balance: the graft must be large enough for the recipient, the donor’s remaining liver must be safe, and blood type must fit the exchange. When any of those checks fail, a willing donor can still be unusable for a particular recipient, even when the transplant center has the surgical expertise to proceed. This study takes that problem seriously and translates it into a decision-support framework built around liver paired exchange, or LPE, a strategy adapted from kidney exchange to widen access to living donor transplantation.
The model was built as an integer linear programming system in OPL/CPLEX and tested in a synthetic cohort of 20 incompatible donor-recipient pairs. Its design did not stop at biological compatibility. It also encoded the day-to-day constraints that often determine whether a transplant can actually happen: operating-room coordination, inter-centre travel distance of 250 km or less, and surgical team availability. In other words, the model attempted to mirror how transplant programs function, not how they look on paper.
One result stood out sharply. Six feasible matches were identified in the simulated pool, equal to 30% of the cohort. Under random allocation with the same clinical thresholds and one-to-one assignment rules, the baseline matched 12.5% of pairs, with a standard deviation of 4.3%. The gap matters because paired exchange programs are often limited not by the lack of potential donors, but by the difficulty of converting theoretical compatibility into operationally executable surgery.
Clinical thresholds cut into a narrow feasible pool
The model used three hard clinical filters before any exchange could be counted as valid: ABO compatibility, a graft-to-recipient weight ratio GRWR of at least 0.8, and future remnant liver volume FRLV of at least 30%. Those thresholds reflect core transplant physiology. If the graft is too small, the recipient risks small-for-size syndrome and graft failure; if the donor remnant is too small, donor safety is compromised by postoperative liver insufficiency. Blood group compatibility remains a separate gatekeeper because immunologic incompatibility can make an otherwise anatomically sound graft unusable.
The simulated cohort was generated from clinically realistic parameter distributions derived from published LDLT and LPE programs in Korea, India, and North America. Blood group distribution followed East Asian prevalence estimates: A at 30%, B at 25%, AB at 10%, and O at 35%. GRWR was sampled uniformly between 0.6 and 1.6, and FRLV between 25% and 50%, with a random seed set for reproducibility. Those values gave the optimization problem enough heterogeneity to resemble a real transplant pool without claiming to replicate any single center.
All six matched exchanges satisfied the ABO, GRWR, and FRLV rules, along with the logistical filters. The output therefore does more than maximize count; it identifies exchanges that remain feasible after the surgical, geographic, and staffing constraints are layered in. That combination is the practical heart of the paper: donor compatibility is necessary, but not sufficient, when living donor liver surgery depends on synchronized teams and protected donor margins.
Operational feasibility changes how exchange systems should be built
The study’s most useful contribution is not the raw number of matched pairs. It is the way the model formalizes operational feasibility inside the exchange engine itself. In paired exchange programs, a center may identify compatible donor-recipient combinations that collapse once travel time, room scheduling, and team coverage are considered. A transplant exchange that cannot be staged safely or coordinated across institutions is not a real exchange, no matter how elegant the matching graph appears.
To capture that reality, the model imposed a global resource constraint through an aggregate logistical complexity score. Each donor-recipient pairing carried a feasibility weight, and the total had to remain within maximum capacity for the assumed scenario. This kind of constraint is important because it prevents the optimization from producing mathematically valid but practically impossible chains. In transplantation, resource limits are not administrative afterthoughts; they shape whether a donor can be protected and whether a recipient can be scheduled before clinical deterioration closes the window for surgery.
The algorithmic backbone used a binary decision variable for each possible donor-recipient match, then maximized the total number of valid transplants while forcing each donor and recipient to appear in at most one exchange. The mathematical structure is straightforward, but its clinical implication is not. Each added constraint reduces the feasible space, and each reduction reflects a real transplant risk: donor remnant volume, graft adequacy, blood-type mismatch, or institutional capacity.
A single invalid exchange can waste time, tie up resources, and delay surgery for other incompatible pairs.
A decision-support tool, not a substitute for transplant judgment
The authors position the framework as support for multidisciplinary decision-making rather than an automated replacement for it. That distinction is critical in living donor liver transplantation, where anatomy, donor willingness, recipient urgency, and institutional readiness all require human judgment. An optimization model can help rank candidates and expose hidden exchange opportunities, but it cannot independently judge donor comfort, operative nuance, or center-specific risk tolerance.
This study also stays within the limits of simulation. The cohort was synthetic, and the matching process was built from published distributions rather than prospective patient-level data. The work therefore demonstrates feasibility of the method, not clinical effectiveness in a live transplant network. Even so, the transparency of the model is a strength. The appendices lay out the assumptions, thresholds, implementation tools, and solver architecture in enough detail that another group could reproduce the framework or adapt it to local conditions.
The most immediate bedside implication is simple: if paired exchange systems ignore logistics, they will overestimate what can be done. If they ignore donor and recipient thresholds, they may overpromise and underperform. A model that integrates both sides of the equation offers transplant teams a more realistic map of what can move from theoretical compatibility to a completed operation.
The next test is not mathematical elegance. It is prospective use in actual transplant programs, where the number of compatible matches, the burden of coordination, and the preservation of donor safety will determine whether this framework can move from simulation into daily surgical planning.