Donor assessment scores: Relevance and complete irrelevance†
Potential conflict of interest: Nothing to report.
1. Donor assessment scores can be used to prognosticate recipient outcomes but are often not clinically relevant.
2. The donor risk index, the survival outcomes following liver transplantation score, and the Donor Model for End-Stage Liver Disease score have specific advantages and disadvantages with respect to accuracy and ease of use.
3. The significance of the donor assessment is undermined by an allocation system that sometimes limits ideal donor-recipient matching and whose sole objective is the minimization of wait-list mortality instead of the benefit of transplantation. Liver Transpl, 2012. © 2012 AASLD.
Although multiple donor and recipient factors are known to contribute significantly to the overall success of liver transplantation, the decision to perform transplantation for an individual patient with any particular liver allograft often remains fraught with uncertainty. Although it is obvious to even the novice clinician that donor quality and recipient disease severity affect posttransplant outcomes, the judicious use of an allograft of marginal quality in a critically ill recipient remains a challenge for even the most advanced clinician. In effect, the ethereal art of matching donors and recipients is well appreciated by many but is mastered by very few.
In recent years, a number of investigators have attempted to transform the art of donor-recipient matching into a science by developing donor assessment scores that quantify the quality of liver allografts. The ultimate goal of these scoring systems is to assist transplant clinicians in the matching of appropriate recipients with donor organs and particularly those organs whose quality is more marginal. Although many of these scoring systems have become quite popular in academic and research arenas because of their ability to facilitate the comparison of a variety of donor-recipient outcomes, the utility of these scores in the clinical realm remains quite questionable. In other words, the very clinicians for whom these scoring systems have been developed do not frequently make use of them. Perhaps the most revealing criticism of these scores is that no donor assessment system has been embraced by the United Network for Organ Sharing to either modify liver allocation or predict graft outcomes for potential recipients during the DonorNet match run.
As we report in this article, these donor assessment scoring systems have not been widely accepted and applied in clinical practice for a multitude of reasons. First, we discuss concerns regarding the development and validity of certain well-known donor assessment scores. We subsequently present patient scenarios that showcase the practicality of these scores and contrast the simple approaches with the more cumbersome approaches. Next, we clarify why donor assessment scores cannot be widely applied in many parts of the country independently of their accuracy (ie, their predictive value) or accessibility (ie, their user-friendliness). Finally, we discuss the durability of these scoring systems in the ever-changing landscape of transplantation, and we conclude by highlighting the negligible impact that donor assessment scores have on the ultimate gamble of liver transplantation.
DONOR ASSESSMENT: THREE CONTRASTING SCHEMES
The prototypical donor assessment score is the donor risk index (DRI), which was developed by Feng et al.1 in 2006. With Cox regression models, these authors used data from approximately 20,000 patients in the Scientific Registry of Transplant Recipients database (1998-2002) to identify predictors of the time to liver graft failure. They identified 7 donor characteristics and 2 transplant factors (the cold ischemia time and national sharing) that were significantly associated with liver allograft failure (Table 1). The DRI can be calculated from these factors to predict 3-month, 1-year, and 3-year graft survival (Table 2), and it has been validated by numerous other centers.2-4
Table 1. Donor Parameters Included in the DRI Calculation
|Age (years)||1.0 (<40 years) to 1.65 (>70 years)|
|Race (African American versus white)||1.19|
|Height (per 10-cm decrease)||1.07|
|Cause of death: stroke||1.16|
|Cause of death: other||1.20|
|Donation after cardiac death||1.51|
Table 2. Graft Survival Outcomes: Single-Center Outcomes Versus Outcomes Predicted by the DRI
The advantages of the DRI include its relative simplicity and its accuracy. However, there are 3 major disadvantages to the DRI that limit its overall utility to the clinician who is considering a liver offer for an individual patient. First, the DRI does not account for the degree of macrosteatosis in the donor liver. In other studies of allograft quality, macrosteatosis has been shown to significantly affect posttransplant graft outcomes.5, 6 The magnitude of this effect is significant because the hazard ratio of 1.71 associated with allografts containing >30% macrosteatosis is comparable to the effect of an age > 70 years (hazard ratio = 1.65).1, 5 Although the architects of the DRI adjusted the regression model for the donor body mass index (BMI), there is little correlation between the BMI and the degree of steatosis.7 Second, the DRI calculation relies on an estimate of the cold ischemia time, which often cannot be predicted at the time of the liver offer when the decision to accept a particular organ for a given recipient is being made. Finally, a major limitation of the DRI in predicting posttransplant outcomes is the exclusion of recipient factors from the model. The lack of recipient factors serves as a reminder that the DRI is intended to identify donor factors predicting liver graft failure; its goal is not to comprehensively predict posttransplant outcomes.
With the incorporation of a number of recipient factors into its predictive equation, the survival outcomes following liver transplantation (SOFT) score was devised with the goal of matching donors and recipients at the time of a liver offer.8 More specifically, the objective of the SOFT score's authors was to devise a scoring system that would predict 3-month posttransplant recipient survival to complement the 3-month pretransplant mortality predicted by the Model for End-Stage Liver Disease (MELD) score. After performing a multivariate regression with data collected by the Organ Procurement and Transplantation Network from more than 21,000 liver recipients, the SOFT architects identified 15 recipient factors, 5 donor factors, and 1 operative factor that significantly affected posttransplant outcomes, and these factors are included in the final calculation of the score (Table 3). Ranges of points are assigned to these risk factors according to the odds ratios for patient death at 3 months, and patients can be grouped according to SOFT score ranges (Table 4). SOFT has been validated by other centers as well,9 and an example of a SOFT calculation is shown in Table 5.
Table 3. Factors and Points Awarded for Components of the SOFT Score
|Age > 60 years||4||Donor age of 10-20 years||−2||Cold ischemia time of 0-6 hours||−3|
|BMI > 35 kg/m2||2||Donor age > 60 years||3|| |
|Prior transplant||9||Donor cerebrovascular accident||2|
|2 previous transplants||14||Donor creatinine > 1.5 mg/dL||2|
|Prior abdominal surgery||2||National allocation||2|
|Albumin < 2.0 g/dL||2|| |
|Pretransplant intensive care unit stay||6|
|MELD score > 30||4|
|Pretransplant life support||9|
|Portal vein thrombosis||5|
|Portal bleed 48 hours before transplant||6|
Table 4. Risk Groups and Risks of Mortality Generated From SOFT Scores
|Low to moderate||6-15||2.16|
|High to moderate||16-35||5.82|
Table 5. Sample SOFT Calculation
|Age > 60 years||0||Donor age of 10-20 years||0||Cold ischemia time of 0-6 hours||0|
|BMI > 35 kg/m2||0||Donor age > 60 years||0|| |
|Prior transplant||0||Donor cerebrovascular accident||0|
|2 previous transplants||0||Donor creatinine > 1.5 mg/dL||2|
|Prior abdominal surgery||0||National allocation||0|
|Albumin < 2.0 g/dL||2|| |
|Pretransplant intensive care unit stay||6|
|MELD score > 30||4|
|Pretransplant life support||0|
|Portal vein thrombosis||0|
|Portal bleed 48 hours before transplant||0|
The primary advantages of the SOFT score are its comprehensive nature and accuracy. Unlike the DRI, the SOFT score was indeed designed to help match donor allografts with specific recipients, and it allows the establishment of risk limits for individual recipients in order to avoid futile transplants. However, the obvious problem with SOFT is its complexity, which is due to the sheer number of variables required for the calculation. Moreover, like the DRI, the SOFT score requires knowledge of the cold ischemia time, and it also does not account for the degree of steatosis in the donor liver.
The Donor Model for End-Stage Liver Disease (D-MELD) score eliminates the complexity of SOFT while still incorporating both donor and recipient factors into a predictive model.10 Because among all potential donor risk factors donor age is the predominant risk factor affecting posttransplant survival, the authors of D-MELD use the donor age and multiply this by the recipient MELD score to generate the D-MELD score. After analyzing Organ Procurement and Transplantation Network data from more than 17,000 liver transplant recipients, Halldorson et al.10 demonstrated that the D-MELD score correlated well with posttransplant survival, and they proposed a risk cap score of 1600 because of this score's ability to portend poor posttransplant survival. D-MELD has been validated in other patient cohorts,11 and examples of D-MELD calculations and actual outcomes are provided in Table 6.
Table 6. Graft Survival Outcomes: Single-Center Outcomes Versus Outcomes Predicted by the D-MELD Score
Although the simplicity of the D-MELD score appeals to many clinicians who are faced with the dilemma of determining which organ should go to which patient, its simplicity is also its greatest weakness. As discussed previously, the D-MELD score does not incorporate many other significant operative donor risk factors that affect transplant outcomes (eg, steatosis, previous number of transplants, and surgical complexity). Furthermore, among all potential recipient risk factors that affect posttransplant outcomes, the recipient MELD score is likely to be one of the least influential factors because investigators have shown that the MELD score does not correlate well with posttransplant outcomes.10, 12
BARRIERS TO THE UTILIZATION OF DONOR ASSESSMENT SCORES
Because of the significant clinical need for useful predictive tools that will assist with appropriate donor-recipient matching, it is surprising that clinicians have not widely adopted one of the numerous donor assessment systems described here or elsewhere in the literature. If clinicians appear ready to embrace a system that will simplify complex decision making, why do these donor scoring systems appear irrelevant outside of academic discussions?
A major barrier to the widespread utilization of these scoring systems in clinical practice is the simple fact that many transplant clinicians are unable to truly match a particular organ with a patient of their choice. This inability to match allografts and recipients is common among transplant centers in competitive organ procurement organizations, in which centers serve much larger populations with fewer available donors and patients undergo transplantation with higher MELD scores.13 In this situation, a center must use an offered allograft in the top patient in the match run or risk losing the allograft to another center in the organ procurement organization that also has a patient with a high MELD score on the wait list. Transplanting a particular liver into a more appropriate patient with a lower MELD score at a lower position on the wait list is simply not an option in this environment. Thus, many transplant clinicians who practice in competitive organ procurement organizations may find donor assessment tools to be relatively useless because of allocation nihilism; for these clinicians, the ability to choose particular organs for particular patients is theory, not reality.
There are potentially other practical reasons that donor assessment scores have had a negligible impact on the practice of transplantation. First, clinicians may be relatively undeterred by the potential for inferior outcomes despite the elevated hazard ratios that accompany their decisions. In other words, transplant physicians and surgeons recognize that the 50% increased relative risk of graft loss associated with a liver from a donor older than 60 years (hazard ratio = 1.53)1 equates to small differences in actual outcomes, and these inferior outcomes are still well within the realm of acceptable risk. Although an organ with a DRI of 1.55 may indeed fail earlier than an organ with a DRI of 1.2, the absolute difference in outcomes between these scores is essentially unremarkable (71% versus 75% 3-year survival). From this perspective, the only utility of donor assessment may be the identification of the extreme risk situation, that is, the organ of such poor quality that it should not be used in any recipient.
Similarly, donor assessment scores do not ultimately alter the unknown mortality risk inherent in declining any particular liver allograft for an individual patient on the list. Although donor assessment scores may allow clinicians to prognosticate how a particular patient will fare after transplantation, these scores do not affect the frequently desperate status of waiting patients. The knowledge that a patient may accrue a 50% risk of graft loss in 5 years may not alter a clinician's decision to proceed with transplantation when the clinician does not know when or if that patient will be offered a liver graft of better quality. Because of the current mortality rate on the wait list, the escalating disease severity of waiting patients, and the extreme donor shortage, a strategy of holding out for a better liver offer may be too risky for many recipients.
A number of donor assessment scores are validated prognostic tools that allow clinicians to predict the future for their transplant recipients. Although these tools all have distinct advantages and disadvantages in terms of accuracy and accessibility, their expansion into clinical practice is limited by a number of practical considerations. Because of the restrictions on donor-recipient matching in competitive organ procurement organization environments and the risk inherent in declining even organs of poor quality for desperately ill patients, donor assessment scoring systems in their current iterations and in the current allocation environment provide little real assistance for the clinician deciding whether or not to accept an organ offer for a particular patient. Nevertheless, the current scoring systems do identify a number of extreme risk situations and may aid clinicians in the development of a risk cap in order to avoid futile transplants.
To promote the utilization of donor assessment scores in the years to come and ultimately capitalize on their great potential, two broad interventions must be taken. First, scoring systems of the future must be appropriately redesigned to incorporate more advanced molecular assessments of donor quality. Technological advancements that improve our ability to assess donor quality at the cellular and genetic levels are currently evolving,14 and it is expected that the molecular profiling of donors and recipients will ultimately refine if not revolutionize donor-recipient matching in the future. In addition to predicting long-term posttransplant outcomes for any donor-recipient match, the discovery of novel biomarkers will undoubtedly challenge our definition of graft success because these assays will allow the earlier detection of allograft recovery and dysfunction.15, 16
However, the most effective intervention that would promote the routine utilization of donor assessment scores in clinical practice is a change in liver allocation. Although the assessment of donor quality provides important predictive data regarding recipient outcomes, the significance of this information is potently undermined by an allocation system whose sole objective is to limit wait-list mortality. In effect, the assessment of donor quality matters only in an allocation system that limits wait-list mortality and maximizes posttransplant benefits. Indeed, the assessment of donor quality in an allocation system that must accomplish both objectives would no longer be esoteric or academic in nature; it would become absolutely necessary.
In conclusion, the DRI has revealed the critical impact of donor quality on transplant outcomes. We dramatically underserve our patients if we choose to ignore the importance of this discovery. Although continually refining donor assessments will allow us to move beyond the DRI, we will excel beyond the DRI only when clinicians not only are convinced of the utility of donor assessments but also are empowered, if not mandated, by the organ allocation system to incorporate donor quality data into their clinical practice.