Potential conflicts of interest: none.
Waitlist Mortality Decreases with Increased Use of Extended Criteria Donor Liver Grafts at Adult Liver Transplant Centers
Article first published online: 23 APR 2007
American Journal of Transplantation
Volume 7, Issue 5, pages 1265–1270, May 2007
How to Cite
Barshes, N. R., Horwitz, I. B., Franzini, L., Vierling, J. M. and Goss, J. A. (2007), Waitlist Mortality Decreases with Increased Use of Extended Criteria Donor Liver Grafts at Adult Liver Transplant Centers. American Journal of Transplantation, 7: 1265–1270. doi: 10.1111/j.1600-6143.2007.01758.x
- Issue published online: 23 APR 2007
- Article first published online: 23 APR 2007
- Received 28 June 2006, revised 31 October 2006 and accepted for publication 21 January 2007
- extended criteria donor;
- liver transplant;
- marginal donor;
- waitlist mortality
Extended criteria donor (ECD) liver allografts are often allocated to less severely ill liver transplant (LT) candidates who are at a relatively lower risk of pretransplant mortality, but it is not clear that the use of ECD allografts will decrease center waitlist mortality (WLM). Individual patient data from the UNOS OPTN database (2002–2005) were aggregated to obtain center-specific data. Deceased donor allografts with any of the following characteristics were defined as ECDs: from a donor with any of the criteria described by the New York State Department of Health Workgroup; or 12+ h of cold ischemia. Multivariate regression was used to examine the relationship between WLM and ECD, non-ECD and LDLT use after adjusting for candidate severity of illness. A total of 3555 ECD transplants, 11,660 standard criteria donor (SCD) transplants, and 717 LDLTs were performed at 100 centers during this period. The model demonstrated that SCD and ECD LTs were inversely correlated with a center's WLM (β=−0.242 and −0.221, respectively; p ≤ 0.003 for each). LDLTs did not significantly reduce WLM (β=−0.048, p = 0.55). In summary, increasing ECD liver allograft use significantly decreased WLM at US centers. Policies encouraging the increase used of ECDs would further reduce WLM.
extended criteria donor
living donor liver transplant
Studies examining the posttransplant outcomes of candidates who receive extended criteria donor (ECD) liver transplants abound, and many have surmised that a more aggressive use of ECDs would decrease the waitlist mortality (WLM) of adult and/or pediatric patients (1–4). Yet firm evidence demonstrating an association between the use of these organs and decreases in WLM is lacking. Indeed, the only US-based study to claim an association between ECD use and WLM was by Renz and colleagues from the New York Presbyterian Hospital (5). Transplant clinicians at this center regularly reviewed the clinical condition of all candidates on their center's waiting list on a weekly basis and determined which might be appropriate recipients for ECD liver allografts. ECD allografts were used for a given candidate if the recipient was among those thought to be an appropriate candidate. The authors noted a 77% increase in the number of transplants performed and a 50% decrease in the number of transplant candidates removed from the waiting list because of death or deterioration after a policy advocating the aggressive use of ECDs was implemented at their center. Although these changes are impressive, the authors could not ensure that other factors correlated with number of transplants performed or with WLM were comparable or adjusted for in the two time periods being compared.
Several European studies have examined the relationship between attempts at expanding the donor pool through the use of split-liver transplantation and mortality among pediatric liver transplant candidates. Most claim that the use of split-liver techniques and/or living donor liver transplantation has significantly decreased or even eliminated mortality among pediatric candidates awaiting liver transplantation (1,6–8). However, these studies also used historical control populations composed of candidates at the same institution and, like the previously described study by Renz and colleagues, did not take into consideration other factors that may have affected waitlist survival and changed over time. It is therefore also difficult to say how much impact these techniques have truly had on decreasing WLM among pediatric patients.
The use of ECDs varies widely among centers, and to date it has been unclear if use of ECD liver allografts results in a decrease in WLM at a given transplant center after adjustment for other factors. The main reason for this difficulty is the fact that the baseline probability of getting a transplant—which depends on local and regional donor liver supply, the severity of illness of candidates, and other candidate factors such as blood type—also varies widely among centers. Evaluating the association between these strategies and candidate mortality or transplant probability has therefore been challenging.
In the current study, multiple linear regression and other econometric methods were used to evaluate the association between a transplant center's use of ECD liver allografts and the that center's WLM after adjusting for the severity of illness of the center's candidates and the number of non-ECD transplants performed relative to the size of the center's waitlist. ECDs are not always allocated to the candidates at highest risk for death while awaiting liver transplantation, however, and it is quite possible that center use of ECDs has no significant impact on WLM.
Materials and Methods
The study sample consisted of all US transplant centers with adult (18+ years of age) liver transplant candidates awaiting liver transplantation between February 2002 and June 2005. The beginning date for the study period was chosen based on the implementation of the Model for End-Stage Liver Disease (MELD). The MELD model, which is the basis for ranking liver transplant candidates, was implemented at all US transplant centers in February of 2002. Individual patient data that were current as of November 2005 were made available through the United Network for Organ Sharing. These data were then aggregated to obtain center-level data. Only transplant centers that performed more than 10 liver transplants during the study period were included in the analysis.
Working Definition of ECD
An ECD liver allograft was, for the purposes of this study, defined as a deceased donor allograft having any of the characteristics identified by the New York State Department of Health Workgroup (9), namely:
- • derived from a donor ≥60 years of age whose cause of death was either anoxia, cerebrovascular accident or other nontrauma cause;
- • derived from a black donor ≥60 years of age;
- • derived from a black donor ≥50 years of age whose cause of death was either anoxia, cerebrovascular accident or other nontrauma cause;
- • split-liver or partial liver allograft from a donor that is any of the following;
- ^ ≥40 years of age;
- ^ <40 years of age and black; or
- ^ <40 years of age and whose cause of death was either anoxia, cerebrovascular accident or other nontrauma cause
- • derived from a donation after cardiac death that is any of the following;
- ^ ≥40 years of age;
- ^ <40 years of age and black; or
- ^ <40 years of age and whose cause of death was either anoxia, cerebrovascular accident or other nontrauma cause
Cold ischemia time (defined as the length of time that the donor liver graft is preserved on ice in between procurement and transplantation) has also been associated with an increase in the relative risk of posttransplant graft failure. While cooling the organ to 4°C does slow cellular metabolism considerably, the complete absence of blood flow during this period does introduce risk of injury from cellular anoxia (10–13). The preservation solutions used to store the liver during this period of cold ischemia have improved over the past several decades, and cold ischemia times of up to 12 h are usually well tolerated. Prolonged periods of cold ischemia still appear to be injurious to the preserved liver, and while the ‘threshold’ cold ischemia time at which livers are at risk for significant injury is not well defined, most clinicians would prefer using grafts with <12 h of ischemia (11). For this reason, we also included cold ischemia time of >12 h as among the characteristics defining an ECD liver.
All deceased donor allografts used in transplantation that did not fit the above definition of an ECD were classified as standard criteria donor (SCD) allografts. All liver allografts derived from living donors were classified as living donor liver transplant (LDLT) allografts.
Theoretical model and hypothesis testing
In general, adult waiting list mortality depends mainly on a candidate's pretransplant severity of illness and access to liver transplantation (Equation 2):
Pretransplant severity of illness is measured by either the MELD score or the Status 1 designation and access to liver transplantation may be measured by the number of standard criteria and expanded criteria donor allografts available per waiting list candidate. Specifically, this relationship may be described as:
where the following variables were included in the analysis:
- • WLMi= waitlist mortality rate; the number of adult liver transplant candidates that are removed from the waiting list of centeri because of death or deterioration at centeri during the study period divided by the total number of candidates added to the waitlist at centeri during the study period.
- • MELD2i= the number of adult liver transplant candidates at centeri that have a MELD score in the second (second-lowest) quartile for scores divided by the total number of candidates added to the waitlist at centeri during the study period.
- • MELD3i= the number of adult liver transplant candidates at centeri that have a MELD score in the third (second-highest) quartile for scores divided by the total number of candidates added to the waitlist at centeri during the study period.
- • MELD4i= the number of adult liver transplant candidates at centeri that have a MELD score in the fourth (highest) quartile for scores divided by the total number of candidates added to the waitlist at centeri during the study period.
- • stat1i= the number of adult liver transplant candidates at centeri that are listed as UNOS Status 1 at the time of removal from the list during the study period divided by the total number of candidates added to the waitlist at centeri during the study period.
- • SCDi= the number of liver transplants performed using standard criteria donor liver allografts performed at centeri during the study period divided by the total number of candidates added to the waitlist at centeri during the study period.
- • ECDi= the number of liver transplants performed using ECD liver allografts (as defined above) at centeri during the study period divided by the total number of candidates added to the waitlist at centeri during the study period.
- • LDLTi= the number of living donor liver transplants performed at centeri during the study period divided by the total number of candidates added to the waitlist at centeri during the study period.
Because of the parsimonious nature of the model and the clinical importance of the few variables that were included in the model, all variables remained in the model regardless of the individual statistical significance of their coefficients.
The main objective was to test the following hypotheses using an α level of 0.05 against a one-sided alternative:
H0: The use of ECD liver allografts do not significantly affect an adult liver transplant center's WLM (i.e. β6= 0)
HA: The use of ECD liver allografts significantly decrease an adult liver transplant center's WLM (i.e. β6 < 0).
The median and range were used to describe the distribution of variables in the model. Spearman's rho was used to characterize the degree of correlation between two variables. The relationship between center WLM and the independent variables listed above was investigated using multivariate linear regression. Coefficient estimates were obtained using ordinary least squares (OLS). Individual and group significance of these coefficients were tested using t and F tests, respectively. Multicollinearity among variables was detected by calculating the variance inflating factor. The presence of heteroskedasticity was assessed using the Breusch-Pagan-Godfrey test. The Ramsey RESET test was used to detect if important variables had been omitted from the theoretical model. Intercooled Stata version 8.0 (StataCorp, College Station, TX) was used for all statistical analyses, and a p-value of <0.05 was considered statistically significant. Permission to perform this study was granted by the Baylor College of Medicine Institutional Review Board.
Descriptive analysis of variables of interest
Exactly 100 US adult liver transplant centers did 10 or more liver transplants during the 2002–2005 study period. These 100 liver transplant centers performed a total of 15,932 liver transplants. These liver transplants included 11,660 liver transplants using SCD liver allografts (73.2% of total OLTs), 3,555 liver transplants using ECD liver allografts (22.3% of total OLTs), and 717 transplants using living donor liver transplants (4.5% of total OLTs).
Waitlist mortality, calculated as the percentage of adult liver transplants added to the waiting list during the study period that died or deteriorated prior to receiving a liver transplant, had a median value of 15.3% and ranged from 0% to 32.2% (Figure 1). The median number of liver transplants performed using ECD allografts per waitlist candidate was 0.09 and ranged from 0 to 0.36 (Figure 2). The median and range values of other variables of interest are summarized in Table 1.
|Variable||Median value (range)|
Cross-sectional multivariate regression analysis
There appeared to be some degree of correlation between ECDi and WLMi. Spearman's rho, quantifying the degree of correlation between the two variables, was −0.303 (p = 0.002). A scatterplot of ECDi and WLMi, shown in Figure 3, demonstrates the inverse relationship between these two variables. All variables specified in Equation 1 as being theoretically related to WLDi were then included in a linear regression model. The R2 value for the model was 0.47. OLS estimates of the model coefficients are listed in Table 2. As expected, the coefficients were positive and increasing in magnitude for the MELD2i, MELD3i, MELD4i and stat1i variables, demonstrating an increasing WLM with increasing numbers of candidates with high MELD scores or with status 1 designation. Testing the group significance of these four variables with an F-test yielded a F-value of 2.63 (p = 0.05).
|Variable||Coefficient||Robust standard error||95% CI||p-Value|
In addition, the coefficients for the SCDi and ECDi, were −0.242 and −0.221, respectively, demonstrating that the increased use of both SCD allografts and ECD allografts is associated with significant reductions in center WLM (p ≤ 0.003 for each). In addition, testing the hypothesis that β5=β6 (the coefficients for SCDi and ECDi, respectively) yielded an F-value of 0.11 (p = 0.77), suggesting that the coefficients for these two variables are not significantly different. In other words, the use of ECDs are associated with as much of a decrease in center WLM as are SCDs. Finally, the estimate of the coefficient for LDLTs was not significantly different from zero. This suggests that center use of LDLTs is not associated with decreases in WLM.
Assessment for multicollinearity, heteroskedasticity and omitted variables
The overall amount of multicollinearity present in the model was measured by calculating the variance inflating factor. The calculated value of the variance inflating factor was 2.7, indicating acceptable levels of multicollinearity. The amount of heteroskedasticity in the model was assessed with the Breusch-Pagan-Godfrey test, which yielded a χ2-value of 5.25 and a corresponding p-value of 0.02. This suggested that some heteroskedasticity was present; because of this, robust standard errors are reported and were used in all significance tests. Finally, the Ramsey RESET test was performed to test the null hypothesis that no missing variables were omitted from the model. The test calculated an F-value of 0.66, corresponding to a p-value of 0.58. Thus there is no evidence that important variables were omitted from the model.
The main objective of this study was to test the significance of the coefficient for ECDi, the variable representing the number of liver transplants performed at centeri using ECD liver allografts divided by the number of waitlist candidates at that center, in a multivariate regression model in which the independent variable was WLM. The model adjusted for the severity of illness of candidates at a given center as well as the number of liver transplants performed per waitlist candidate using SCD allografts and living donor liver allografts. Rejection of the null hypothesis—which stated that the coefficient for the ECDi variable would be equal to zero—would provide some evidence that the use of ECD may actually reduce the number of deaths among adult candidates awaiting liver transplantation at US liver transplant centers.
The results of the analysis yielded an OLS estimate of −0.221 for the coefficient of the ECDi variable, suggesting that WLM will be decreased by about 1% for every additional four ECD liver transplants per 100 waitlist candidates. The 95% confidence interval of this estimate ranged from −0.365 to −0.078 and corresponded to a p-value of 0.003. This provided sufficient evidence to reject the null hypothesis and accept the alternative hypothesis. In other words, after adjusting for the pretransplant severity of illness of the center's candidates and the number of SCD and living donor liver transplants performed, US adult liver transplant centers that used more ECD allografts tended to have a lower mortality among waitlist candidates at their center. Indeed, the fact that the estimates of the coefficients for the SCDi and ECDi variables did not differ significantly suggests that the use of ECD liver allografts is as effective in reducing mortality among adult liver transplant candidates as SCD liver allografts.
This represents the first clear evidence of the impact of ECD use on WLM at US adult liver transplant centers. Many previous authors have surmised that a more aggressive use of ECDs would decrease WLM (1–4), and it is entirely plausible that the increased availability of liver allografts resulting from the use of ECDs would decrease WLM. Yet ECD liver allografts are often allocated to less severely ill transplant recipients that have a lower risk of pretransplant mortality, so it could not be tacitly assumed that the use of these allografts would necessarily decrease WLM.
It may not be surprising to find that there is no significant relationship between the extent of living donor liver transplants performed at a center and the center's WLM. Previous studies have demonstrated that living donor liver transplant recipients tend to be less severely ill than recipients of deceased donor donor allografts (5, 14). The timing of liver transplantation using a liver allograft from a deceased donor donor is dependent on the regional death and organ donation rates as well as the severity of illness of a candidate relative to the other waitlist candidates in his or her region. In most centers in the United States, liver allografts are not allocated to candidates until the patient's liver disease is quite severe. In contrast, the timing of living donor liver transplantation depends only upon the availability of a donor that is healthy and willing. Candidates in whom the eventual need for liver transplantation is recognized are often able to identify potential donors earlier in the natural history of their liver disease. The elective nature of the procedure allows transplant clinicians to perform the living donor liver transplant earlier in the course of the candidate's disease, and, as a result, these candidates are typically less severely ill than recipients of deceased donor liver allografts. Living donor liver transplantation clearly helps the recipients who receive the living donor allografts. While one might also think that the use of living donor allografts might decrease WLM by removing candidates from the waiting list (and thus ensuring less competition for the limited number of deceased donor allografts), no evidence for this is provided by the current study.
Some limitations should be noted. Although the recent publication by Feng et al. (15) has made significant strides in quantifying donor risk, no well-accepted definition of an ECD donor liver has yet been established (9). Because no standard definition of ECD livers exists, the willingness or reluctance to use donor livers with certain characteristics will vary among centers. Certain centers, for example, might be much more willing to accept livers from donors after cardiac death than other centers. The definition of an ECD liver used in this study was selected based mainly on the findings of a workgroup on ECD donor livers (9). In addition, livers with cold ischemia times of >12 h were considered ECD livers. Data on percent steatosis was not available and thus could not be included in the definition. Nonetheless, we feel that the definition used in the manuscript adequately represents the working definition used at most centers. As the US experience with ECD donor livers increases and the methods for quantifying donor risk become more accurate, however, future studies might alter the definition of ECD donor livers or classify ECD donor livers into multiple categories based on degree of risk.
We also chose to include only those US liver transplant centers that performed 10 or more total OLTs during the study period. We excluded these because the lower volume centers introduce variability into the analysis. After excluding low-volume centers, the median number of OLTs performed at centers included in the study was 124, and 90% of included centers performed more than 35 OLTs. Excluding additional low-to-moderatve volume centers may have further decreased variability but would have resulted in a study sample less representative of US liver transplant centers as a whole. For this reason we decided against a higher minimum number of OLTs, but we acknowledge that the threshold of 10 OLTs was arbitrary.
Another limitation is the potential endogeneity between WLM and the use of ECD liver allografts among the centers in this study. It is possible that a transplant center's decision to use ECD liver allografts is determined, at least in part, by its WLM—specifically, that increasing WLM might motivate a center to utilize more ECDs for liver transplantation, resulting in some degree of direct correlation between the WLMi and ECDi variables. Such endogeneity might spuriously bias the estimate of the ECD coefficient obtained by OLS estimates towards the null. While the presence of endogeneity has not been ruled out by this study, the results suggests that any endogeneity present did not bias the estimate of the coefficient towards the null enough to prevent the estimate of the coefficient of the ECDi variable from being significantly different from zero (and thus leading one commit a type II error).
Finally, we made a clear decision to focus on center-level outcomes in this study. Transplant clinicians owe allegiance to individual patient for whom they care, but in many ways they are also trying to maximize the outcomes to a population, namely transplant candidates. The use of ECDs may disadvantage the few individual recipients that receive them while benefiting the transplant candidate population as a whole. Thus, the use of ECDs to lower WLM may come at a price: would it be better to increase the availability of OLT in the US through the liberal use ECD donor livers and decrease WLM to 5% but lower the overall one-year posttransplant survival to 70%? Or to be much more selective, use optimal donors and few ECD donor livers, allowing WLM to remain between 10–15% while achieving 90% 1-year posttransplant survival? Ultimately, this is a question of whether the transplant clinician's primary goal is to decreasing WLM (and maximize the benefit to the population of liver transplant candidates at the cost of individual liver transplant recipients) or increasing posttransplant survival (and maximizing the benefit to individual liver transplant recipients at the cost of the liver transplant candidate population) (16,17). The ‘intent-to-treat’ analysis, introduced by Dr. Richard Freeman (18), is a novel approach and a major advancement in quantifying the results of allocation decisions. This approach treats pretransplant (waitlist) and posttransplant survival as equivalent, analyzing the outcomes of patients from the time they are added to the liver transplant waiting list regardless of whether they undergo liver transplantation. It should be recognized, however, that while science can aid clinicians in making allocation decisions, there is no scientific means to weigh the relative values of pre- and posttransplant mortality. Decisions on how to achieve the ideal balance between pretransplant mortality and posttransplant outcomes and the extent to which ECD liver allografts should be used in liver transplantation will need to be based on a dialogue between researchers, policy makers, national transplant organizations such as UNOS, transplant clinicians, and individual patients. The transplant community has gained some experience with the use of ECD donor livers, but it is not yet enough to allow for firm decisions to be made at this point in time. Until the donor selection strategies and the posttransplant outcomes associated with ECD donor livers are better understood, transplant centers should be allowed to use ECD donor livers under peer-reviewed research protocols, allowing the most effective (or most agreeable) strategy or strategies to emerge.
The authors would like to acknowledge The Methodist Foundation for financial support and the United Network for Organ Sharing for the data used in this study. The results of this study were presented in thesis format to the faculty of the University of Texas School of Public Health in March of 2006 in partial fulfillment of requirements for the degree of Master of Public Health.