EXPERIMENTAL AND CLINICAL TRANSPLANTATION

Key words : Cold ischemia time, Kidney allocation system, Organ distribution

A severe gap exists between the number of patients in need of an organ transplant and the availability of viable organs. Over 120 000 individuals are currently on national wait lists, but only approximately 30 000 transplants were performed in 2015. An average of 22 people die each day waiting.¹ In recent years, the United Network for Organ Sharing (UNOS) enacted changes in distribution policies for both kidneys and livers. For kidneys, the major recent changes to the Kidney Allocation System (KAS) enacted in December 2014 included a kidney donor profile index score and an estimated posttransplant survival score, which would provide a percentile that would be compared with a reference score of all adult patients on wait lists for a kidney.² In addition, the new KAS included a rule specifying that recipients with panel reactive antibody levels of 98% to 100% would share organs both on a regional and national scale. These changes to the KAS were enacted due to rising kidney discard rates, variability in kidney access, and inequities resulting from wait time calculations, with overall goals of increasing successful kidney transplant outcomes.

For liver transplant, the risk of early mortality is best expressed by the Model for End-Stage Liver Disease (MELD) score.³ A June 2013 rule called “Share 35” allowed for a shared wait list for local and regional patients with MELD scores of 35 or greater. These critically ill patients with MELD scores higher than 35 would receive priority in all local and regional match runs. As a result of this rule change, transplant recipients can be at a distance further from where the organ was procured. Allocation policies continue to evolve. As recently as December 2017, the UNOS/Organ Procurement and Transplantation Network (OPTN) Board of Directors approved policy amendments stating that, “Adult candidates who have a calculated MELD score of 32 or higher ... would be prioritized for organ offers.”⁴ Both the MELD system and Share 35 were designed to ensure organ allocation to the recipient with the most to gain without jeopardizing the patient’s life. Simulation analysis has demonstrated that broad share programs for liver transplants can reduce mortality from end-stage hepatic disease and reduce health care costs.⁵

Because organ procurement distance can theo-retically increase due to Share 35 and KAS, an increase in adverse outcomes due to higher potential cold ischemia time (CIT) may result. It has been established that each additional hour of CIT in renal transplant is associated with an increased risk of graft failure and death.6 In this study, we aimed to investigate the variables that play a role in liver and kidney transplant outcomes and how UNOS distribution rule changes such as KAS and Share 35 have affected these transplant outcomes.

Materials and Methods

This retrospective study sought to evaluate outcomes of transplanted organs vis-a-vis the recent organ distribution rule changes. Data containing kidney and liver transplant donor and recipient information since October 1, 1987 were collected by the OPTN. Only deceased-donor transplants were considered. Each organ transplant cohort was classified into 2 groups based on the distance from the donor hospital to the transplant center, that is, 0 and greater than 0. Upon completion of the initial survival analysis in distance-related outcomes, we then compared outcomes in the context of recent policy changes related to kidney and liver transplants. For kidney transplant outcomes, we compared qualifying outcomes for transplants before and after the December 2014 KAS rule change. For liver transplant outcomes, we compared qualifying outcomes for transplants before and after the June 2013 Share 35 policy change.

We selected risk factors for kidney transplantation using the stepwise method in the Cox proportional hazards model. Descriptive statistics of clinical and demographic characteristics were summarized using one-way analysis of variance and t test for continuous variables and chi-square test for categorical variables. For continuous variables, compliance with the normality assumption was tested using goodness-of-fit test, and Kruskal-Wallis rank test was performed when the normality assumption was violated. Because the goodness-of fit test showed that the normality assumptions for all continuous variables in the table were violated, P values in Table 1 were from the Kruskal-Wallis rank test.

Propensity score matching
Propensity score matching was a 3-step process. In step 1, we used a logistic model to calculate the propensity score for every patient. The propensity score for a patient is defined as the probability of the patient belonging to a certain group given the relevant covariates. In our case, the group is either distance of 0 or greater than 0. The model is logit(Pr[Z = 1|X]) = PS(X) = Xβ, where Z is 1 for distances equal to 0, Z is 0 for distances greater than 0, and X is a multidimensional vector of covariates that included recipient information and the Kidney Donor Risk Index (KDRI) values. In the model equation, β is the regression coefficient vector. Recipient information included, among others, previous transplant status, age, ethnicity, body mass index, and CIT. The KDRI score was based on donor information using organ-specific donor risk indices established by previous literature, which assumed average transplant variables (ie, 2 mismatches at the HLA-B locus, 1 mismatch at the HLA-DR locus, 20 hours of cold ischemia, and not an en bloc or double transplant).⁷

Thus, we established 2 logistic models for kidney transplantation, one including the CIT and the other without: (1) logit(Pr[Z = 1|X]) = PS(X) = β1 prev + β2 creat_trr + β3 bmi_tcr + β4 black + β5 age + β6 KDRI + β7 cold_isch_ki and (2) logit(Pr[Z = 1|X]) = PS(X) = β1 prev + β2 creat_trr + β3 bmi_tcr + β4 black + β5 age + β6 KDRI.Based on these variables, the logistic regression model calculated a conditional probability of having Z = 1 or Z = 0 given the recipient information and donor risk index values for each observation. We then utilized this conditional probability to do the propensity score matching.

In step 2, we used one-to-one propensity scores to find the paired observations. We used 1:1 propensity score matching to find the pairs of patients in the 2 groups (Z = 1 or Z = 0). For every member of group Z = 1, the propensity score found patients from group Z = 0 with the closest estimated propensity score and then formed a pair. From the dataset, we had 18 078 patients for Z = 1, which we then matched by finding the closest propensity score in the potential control group in which Z = 0. All remaining members from the control group were disregarded. The main purpose was to reduce the entire collection of the covariates to 1 composite variable by a logistic regression model.

In step 3, we performed a survival analysis for the propensity score-matched data. The survival analysis was conducted for matched data. Two models were used for the survival analysis, including the Cox model and the nonlinear mixed model.

Cox proportional hazards model
From our dataset, we found 18 078 paired obser-vations from step 2 and recorded the survival time Xi. From step 2, we selected the paired patients (that is, those having only difference between them of binary variable distance = 0 vs distance > 0). Because of this, our Cox regression results could be considered from the comparison of survival time with distance, and we were also able to observe directly the impact of distance on survival time.

In Table 2, P values are shown for 3 different disPs = 1 in the logistic model, which are based on distance (distance = 0 miles, distance ≤ 15 miles, and distance ≤ 18 miles), and these 3 values gave information about the paired sample size after propensity score matching, which were utilized in survival analysis. We chose 18 miles because it was the median transfer distance of locally shared organs.

Cox regression only
In Table 3, we used the Cox proportional hazards model directly, meaning that we put covariates directly in the hazards models to see whether the distance covariate is significant for recipient graft survival. In this case, we did not reduce the sample size by propensity score matching.

When we compared patients who received kidney transplants at the same hospital in which the organ was procured (identified as distance = 0) versus patients who received a transplant at a hospital other than where the organ was procured in (identified as distance > 0), the patients who received kidney transplants in the same hospital had overall survival outcomes similar to those patients who received a transplant at a hospital at any distance from the hospital where the organ was procured (distance > 0). This finding was determined after controlling for both patient risk factors and organ quality using propensity score matching.

We then determined that patients who received a transplanted kidney at a hospital in the vicinity of the site of organ procurement (vicinity defined as less than 15 miles away) showed significantly better survival outcomes than patients who received a kidney from a hospital greater than 15 miles away. This finding was even more striking when using 18 miles as the threshold rather than 15 miles. The 18-mile distance from the procurement hospital to the engraftment hospital is especially relevant to clinical practice because 18 miles was calculated to be the median organ transfer distance of locally shared organs in our dataset from the OPTN.

In Tables 4 and 5, we specifically included data analyses for transplants only before the December 2014 KAS change. With this new analysis, we determined that the significance that was previously established no longer existed after enactment of the rule change (both at 15 miles and 18 miles).

We also completed a similar analysis for data collected on liver transplants during the same timespan. After an initial analysis, we evaluated whether any changes were evident in comparing data from before and after the enactment of Share 35.

For categorical variables, we used the chi-square test to test whether the 2 groups had the same proportion; for the continuous variables, we first tested whether these variables fit the normal distribution by creating “QQPlot.” If the variable fit the normal distribution, we then used the t test to determine whether the mean difference between these 2 groups was 0. If it did not fit, we utilized the Wilcoxon rank sum test. Here, none of the continuous variables satisfied the normal distribution; thus, we used the Wilcoxon rank sum test. From Table 6, we noted that all variables in the 2 groups were signi-ficantly different (P < .05), except “Split_liver_graf” (P = .11). Thus, the propensity score method was best for analysis as it could make these groups fit in the same distribution.

Table 7 shows that the distance variable, based on the median distance of the locally shared organ (22 miles), significantly affected patient survival. Moreover, Table 8 also had the same conclusion.

Both models appeared to show a distance-related significance when we compared liver transplants from greater than 22 miles (the median distance for liver transplants in our dataset) versus less than 22 miles. Share 35 did not appear to have the same effect on outcomes as seen in the rule change on kidney transplants; however, there was a significant difference in outcomes even when accounting for Share 35.

After extensive analysis of a large dataset of kidney and liver transplant patients, statistical significance was found when distance between procurement and engraftment sites was greater than 15 miles. This significance increased even more so when distance was examined at median transplant distance (more than 18 and 22 miles for kidney and liver datasets, respectively). However, after we corrected for the recent KAS and Share 35 rule changes, statistical significance was no longer evident for kidney transplants, although the liver transplant data continued to show statistically significant differences in outcomes. This indicates that our data were not skewed by the fact that the distances increased after Share 35. The uptick in adverse events with respect to distance showed the effect of CIT on liver transplant outcomes. Recent work has quantified the median CIT for same-hospital liver transplants as 5.0 hours and median CIT for outside hospital transplants as 6.6 hours, attributing the prolongation of CIT mostly to in-hospital delays.⁸ Although we are in agreement that in-hospital delays play a significant role in CIT, we still believe that the most addressable cause of prolonged CIT is distance.

The KAS rule change allows for patients who have a high panel reactive antibody profile to have preference for kidney transplants across the country. After it was instituted, the average distance of transplant matching doubled from 300 to 600 miles. Initial analyses have confirmed an increase in average CIT in transplants that occurred after the enactment of the KAS rule change.⁹ These results may indicate that distance does not have as great of an effect on kidney transplants as first thought, except with regard to extreme distances. In fact, the data revealed that the December 2014 changes to KAS may have improved kidney transplant outcomes.

Regarding liver transplant data (both before and after Share 35), a multitude of policies and solutions are likely required to improve distance and CIT-related outcomes. Strategies that broaden organ sharing such as equalizing MELD score over a wider geography have been shown to be inefficacious.¹⁰ The most likely effective way would be to increase the supply of organs available for transplant, a readily obtainable goal as shown through numerous organ procurement organizations increasing yield in the past years.¹¹ However, to maximize the efficiency of transplant outcomes, there is always room for more improvement. One potential strategy for improving transplant outcomes outside of improving yield is tweaking the structure of UNOS organ-sharing districts to maximize sharing, such as through a recently proposed model that would superimpose concentric “neighborhoods” on top of the existing 11 UNOS regions.¹² Another potential strategy is the use of centralized, free-standing facilities for organ retrieval. These facilities, which would decrease the distance between procurement and engraftment locations, have been shown to allow safe transport to an independent site without major consequences to overall hemodynamic stability.¹³ Each transplant region should have at least 1 or preferably multiple free-standing facilities properly outfitted solely for organ procurement and engraftment so that travel time can be minimized while also allowing access to regular, experienced staff trained in transplant medicine. However, our outcomes after the rule change indicated that procurement sites should be centralized based on regional liver rather than kidney transplant needs. If both liver and kidneys are being procured from a single donor, then this donor should then be transported to the procurement center closest to the destination of the liver rather than the kidneys. If policies promoting these strategies were to be implemented by the lowest performing donation service areas, it would be a great first step toward a meaningful improvement in outcomes.

There are some weaknesses and certain limita-tions that our project cannot properly address. In minimizing CIT, there are issues that persist both on the donor procurement end (whether the organ is being procured in a traditional hospital or a free-standing facility) and on the end of the recipient hospital (administration, operating room rules, and so forth). Within each hospital, there are certain rules, regulations, and policies on organ transplantation that can vary from hospital to hospital. All of the “red tape” that exists between organ procurement and organ engraftment could potentially extend CIT of the liver even when both procurement and engraftment occur within the same hospital. To truly minimize the unnecessary lengthening of CIT, future studies should investigate data at both individual hospitals and free-standing procurement facilities to determine how to best reduce CIT effectively. In addition, CIT should be evaluated in the context of travel time, not just distance, because the duration spent traveling with a procured organ can vary in each individual case due to mode of transportation and local traffic patterns. An additional weakness is that our analysis did not consider extrarenal and extrahepatic organs, and further research should also investigate these. Finally, the effect of new rule changes needs to be regularly evaluated in future studies in the context of transplant distance and outcomes so that there is continuous work toward maximizing beneficial outcomes.

Liver transplants showed significant differences with regard to distance-related outcomes, even when accounting for the Share 35 rule change. For kidney transplants, a significant difference could not be established following the new KAS enacted in December 2014. One potential solution could be moving brain-dead organ donors to an independent, free-standing procurement facility to minimize the effects of CIT due to travel. The procurement facilities chosen should be centralized based on liver destination; donors should be transported to the facility that would minimize liver travel distance. Hospitals, administrators, surgical teams, and others should work together to minimize CIT and improve outcomes related to procurement and the transplant site.

Volume: 20 Issue: 3 March 2022

FULL TEXT