Continuing progress toward improved lung assessment

Lung transplantation is an effective therapy for end-stage lung disease.¹ As in other types of solid organ transplants there are limitations to the number of organs available for transplantation; therefore, efforts to expand the donor pool through use of extended criteria and circulatory death donors have been advocated.^2-8 More recently, use of ex vivo lung perfusion (EVLP) has further expanded the potential donor pool;⁹ however, challenges facing all donor expansion strategies are fundamentally the same. Criteria for organ use can be subjective, vary between centers⁶ and surgeons and do not reliably predict good outcome. EVLP is perhaps the best illustration of this subjectivity since there is a limited number of variables with standardized criteria for safe organ use; furthermore, despite these criteria being followed, a significant minority of recipients continue to experience poor outcomes (primary graft dysfunction (PGD), prolonged mechanical ventilation, prolonged length of stay and mortality). Methods for improving decision making by decreasing subjectivity have the potential to improve access, use and outcome in candidates receiving lung transplants.

The manuscript by Sage et al in this issue of the Journal presents an objective scoring system for outcome prediction after EVLP transplant, using the equally weighted normalized product of IL-6 and IL-8 protein and ΔPO₂ calculated during perfusion. Reported predictive accuracy was excellent (87% [95% CI:83-92] and divided organs into a 2x2 matrix for conceptualization and decision making (Figure 1). The implications of this figure and this work are certainly far-reaching and expand on objective biomarker enhanced donor organ assessment strategies.¹⁰

Figure 1.

Adapted from Sage et al.

Let us first consider the experience of many centers that use EVLP as part of their routine clinical practice. It is not uncommon to perform an EVLP transplant in which all standard measures were met and yet the recipient develops severe PGD, perhaps requiring ECMO salvage. This experience results in reevaluation of the donor, the EVLP run, the conduct of the operation and the candidate where, unsatisfyingly, no smoking gun is ever identified or a few weak possibilities are considered. Given the evidence presented by the Toronto group, this scenario is likely the result of transplanting organs with reasonable gas exchange and unknown but significant inflammatory lung injury (Figure 1 red box), demonstrating the appeal of making real-time EVLP diagnostics. Alternatively, it might be inferred that until a real-time diagnostic is readily available (or cost is negligible) that a simple alternative to avoid poor outcomes after EVLP transplant is to just use a ΔPO₂ cutoff of ~430 mmHg as organs above this threshold appear to be less vulnerable to inflammatory injury. However, it must be noted that choosing this strategy will unfortunately result in lower conversion rates and potentially greater cost to transplant programs.

Perhaps a more intriguing application of this predictive classifier is the potential for precision medicine approaches in lung transplant. An archetype of biomarker directed treatment decision making was first described in breast cancer where HER2 expression (protein) guided trastuzumab use (intervention) over 20 years ago.¹¹ Sage and coauthors similarly present a protein based characterization (normalized product of cytokines IL-6 and IL-8) for decision making for an intervention (transplant). Both cytokines have been demonstrated to be highly associated with inflammatory associated lung injury, the TLS₂ defines an endotype (subtype of a disease defined by a distinct pathophysiological mechanism) which may be amenable to pharmacologic intervention. Stipulating the assumption that the TLS₂ defined endotype is required for a therapeutic to be effective, the practical next step is to utilize a drug-diagnostic co-development model in which inclusion criteria is based TLS₂. Using this approach has been successful in oncology trials with surprisingly small numbers of patients because of greater trial efficiency.¹¹ The results presented are encouraging and amenable to this approach. Ideally, identification of an effective agent could reduce inflammation and shift Group D standard criteria donors to Group C (shift from red to blue box) or improve oxygenation (shift from red to brown box) or both improve oxygenation and inflammation (shift from red to green box). All of these scenarios expand the donor pool and the last would also optimize transplant outcomes as well.

Unfortunately, there are potential hidden consequences of this predictive classification system and limitations to generalizability. While interventions to move organs up and/or to the right expand the donor pool, it is entirely possible for standard Group D organs (red box) to be declined for transplant by some centers to avoid risk to patients and excessive resource requirements for severe PGD. Whether this will be significant will need to be defined by clinical utility studies; however, as illustrated by Figure 1 there could be a net loss rather than gain of organs for transplant (more organs in the red box than the blue box). It is unclear how the TLS₂ will perform in bilateral EVLPs where one organ is clearly unusable as these organs were excluded. It is also unclear how the classifier will perform in ECMO bridge to transplant and critically ill candidates since these patients were also excluded but likely to result in reductions in discriminant ability of the TLS₂ given the higher risk recipient characteristics. Limitations notwithstanding, the Toronto group present an intriguing and exciting classifier that has the potential to once again disrupt current practice. Future studies on real-time integration and target therapeutics are needed and have the potential to further expand the donor pool and improve lung transplant outcomes.