The impact of broader regional sharing of livers: 2-year results of “Share 35”

General

This study used data from OPTN. The OPTN data system includes data on all donors, wait-listed candidates, and transplant recipients in the United States, submitted by the members of OPTN, and it has been described elsewhere (https://optn.transplant.hrsa.gov/data/about-data/). The HRSA, US Department of Health and Human Services, provides oversight to the activities of the OPTN contractor.

For the majority of analyses, data were provided for 2 eras. The pre-implementation era included data from 6/18/2011 to 6/17/2013, whereas the post-implementation era included data from 6/18/2013 to 6/18/2015. Each period consists of 730 days. Unless otherwise noted, all MELD/PELD scores were those used for allocation (ie, they include additional exception points).

All analyses were performed using SAS, version 9.3 (SAS Institute, Cary, NC) and were based on OPTN data as of September 4, 2015. All P values are 2-sided.

Waiting List, Transplants, and Liver Discards

Waiting-list demographics were obtained from the OPTN transplant candidate registration form. Cold ischemia time (CIT) of the donor liver was obtained from the OPTN transplant recipient registration form. The donor risk index (DRI), a measure of donor quality, was calculated according to Feng et al.8 Larger values of the DRI are associated with donor livers of less quality. Organ travel was based on the distance between the donor hospital and the transplant center and was calculated using the longitude and latitude of each institution. Data on livers that were not recovered and livers that were recovered for transplant but not transplanted were obtained from the OPTN deceased donor registration form. Tests for differences in the proportions of patient demographics between the post- and pre-implementation eras were based on the chi-square distribution. Tests for differences in continuous variables were based on the t distribution or the Wilcoxon rank sum test. Testing for the difference in the variance of the median allocation MELD/PELD score between eras was performed using the F test. The trend in the number of MELD/PELD 35+ candidates on the waiting-list snapshot over time was estimated using a linear spline with 1 knot on the date of the Share 35 implementation.

Waiting-list Outcomes

Removals for death as well as “too sick” were treated as deaths. However, if a candidate was removed from the list for too sick and that same candidate was discovered on a subsequent listing, that candidate was not considered to have been removed for too sick (recoded to “other removal”). Records for candidates who were multiply listed were combined into 1 registration. Prevalent candidates on the list at the beginning of each era had delayed entry into the risk set using left truncation.

Candidates who had not yet experienced any event were censored as of the end of the applicable period. For each registration, a candidate could experience at most 1 event. The covariate used in the analysis was the allocation MELD/PELD for each candidate at the start of the period (for prevalent candidates) or the initial allocation MELD/PELD for incident candidates. The laboratory MELD/PELD was used for status 1 candidates and candidates who were inactive at the start of the period. The analysis was performed independently for each era using all candidates, and an overall analysis was used to assess formally the impact of the era. A separate analysis was performed for each era excluding those candidates who at some point entered the MELD/PELD 35+ category.

For candidates initially entering the MELD/PELD 35+ category in each era, transplant rates and death rates within 90 days were estimated using a cumulative incidence analysis10 to account for competing risks. In this analysis, the pre-implementation era cohort included candidates entering MELD/PELD 35+ between 6/18/2011 and 90 days before 6/17/2013. The pre-implementation era cohort was shortened to avoid overlap with the post-implementation era data. Similarly, the post-implementation era cohort included candidates entering MELD/PELD 35+ between 6/18/2013 and 90 days before 6/18/2015. Confidence intervals for the competing risks outcomes were calculated by the method described by Aalen.11

Posttransplant Outcomes

Unadjusted patient and graft survival was assessed using the Kaplan-Meier method.12 Adjusted patient and graft survival was estimated using Cox regression modeling.13 For these analyses, all transplants in the pre-implementation era cohort were used and transplants from 6/18/2013 to 03/13/2014 were used in the post-implementation era cohort to allow for sufficient follow-up. In the pre-implementation era cohort, 98% of recipients were followed for a minimum of 12 months. In the post-implementation era 6-month cohort, approximately 65% of recipients were followed up to that point, with 80% having follow-up information as late as 350 days after transplant. Only primary transplants were considered in the analysis of patient survival, whereas all transplants were used in the analysis of graft survival. For the patient survival analysis, survival time was calculated from the first transplant to the latest follow-up for the recipient, regardless of the number of transplants. Patient survival time was censored at the latest follow-up if the recipient was alive, and graft survival time was censored at the latest follow-up if the liver was functioning. All observations were censored after 365 days. Cox regression models of patient survival were adjusted for recipient age, whether the recipient had a MELD/PELD exception score at the time of transplant, and the DRI, which is calculated using CIT and several factors related to donor quality. In addition to the factors used in the models for patient survival, Cox regression models of graft survival included a factor for previous transplant. Tests for differences in unadjusted survival were performed using the log-rank test.14

Transplants

There were 12,989 deceased donor liver transplants in the post-implementation era, compared to 12,137 in the pre-implementation era. This was an increase of 852 transplants. As shown in Fig. 1, this increase was consistent across all regions. Of the 12,989 transplants, 26.5% were performed in the MELD/PELD 35+ group in the post-implementation era. Of these, 12.9% had MELD/PELD exceptions. In the pre-implementation era, 18.5% of transplants were performed in the MELD/PELD 35+ group, and of these, 11.4% had exceptions. The overall median MELD/PELD scores at transplant also increased from 27 to 28, with scores in the post-implementation era ranging from 25 to 35 across regions. Figure 2A,B shows that the variance in the median MELD/PELD at transplant also increased, from 14.4 in the pre-implementation era to 16.8 in the post-implementation era (P = 0.56). The demographics of recipients were relatively unchanged between eras, as shown in Table 1.

The level of regional sharing also increased from 20.8% to 32.0%. The number of liver-intestine transplants increased from 83 to 154. There was a slight increase in the number of liver-kidney transplants (936 versus 1118), increasing from 7.7% to 8.6% of all deceased donor liver transplantations.

Organ Travel Distance, CIT, and DRI

As shown in Table 2, the overall median organ travel distance increased from 58 to 82 miles in the post-implementation era (P < 0.001). The variance in this metric by DSA increased substantially (P = 0.003) in the post-implementation era, driven largely by the changes in region 5 (Figs. 3A,B). Little or no change was seen in CIT (P = 0.30), although there were a few more outliers in the post-implementation era. There was an overall very slight increase in the DRI (P = 0.001). Interestingly, however, the DRI for regional shares showed a marginal decrease, indicating slightly better organs were being used in these recipients (P = 0.02).

Livers Not Used

In the pre-implementation era, there were 2235 donors from whom a liver was not recovered. This was 13.7% of all donors in the period. In the post-implementation era, there were also 2235 donors from whom a liver was not recovered (12.9% of donors; P = 0.04). It should be noted that there were 870 more donors in the post-implementation era. Whole liver discards (recovered for transplant but not transplanted) were also reduced, falling from 1355 (10.3%) to 1338 (9.5%; P = 0.04). Although most regions experienced a decrease in the discard rate, there were some regions that experienced an increase (Fig. 4).

Waiting-list Dynamics

Table 3 shows the number of waiting-list additions by era and allocation score at listing. Over 600 additional candidates were added to the waiting list in the post-implementation era, with 58% in the MELD/PELD 15-34 category, and over 39% in the MELD/PELD 35+ group. Figure 5 shows the number of candidates with MELD/PELD allocation scores of 35+ on a waiting-list snapshot during the study period. Although there were some fluctuations from month to month, there was an increasing trend in the number of these candidates that was enhanced with Share 35 from approximately 100 in the pre-implementation era months to more than 120 in the post-implementation era months.

Pretransplant Outcomes

There were 37,656 candidates on the waiting list during the pre-implementation era, of which 5798 (15.4%) were removed because of death or being too sick. In the post-implementation era, there were 37,929 candidates on the waiting list, of which 5903 (15.6%) were removed for death or being too sick. Because of the net increase in more medically urgent patients during the post-implementation era, waiting-list mortality adjusted for starting MELD is appropriate when comparing the results across eras.

Overall adjusted waiting-list mortality rates and mortality rates for adults, pediatrics, and by ethnicity are shown in Figs. 6–9. To show the difference in adjusted mortality rates by era, each figure shows the expected waiting-list mortality over time for hypothetical candidates with a MELD/PELD score of 15 at the beginning of each era. The overall adjusted mortality rate was slightly lower in the post-implementation era (Fig. 6), although this rate was found to be within the confidence limits of the pre-implementation era mortality rates (Fig. 7). In addition, the overall adjusted hazard of mortality was not significantly different by era (P = 0.70). There were no differences in mortality rates across eras for adult and pediatric candidates (Fig. 8). Likewise, no differences in mortality rates across eras were found by ethnicity (Fig. 9). Although African American and Hispanic candidates appeared to fare slightly better under Share 35, their adjusted hazard of mortality was not significantly different across eras (P = 0.45 for African Americans and P = 0.44 for Hispanics). In a separate analysis, excluding candidates who entered the MELD/PELD 35+ category in each era, waiting-list mortality rates were nearly identical by era (Figs. 10 and 11).

Figure 12A,B shows the results of a competing risks analysis from the time of the candidate's first entry into the MELD/PELD 35+ category in both eras. Although these candidates were captured in the overall analysis, the outcomes for this specific group of candidates during both eras were of interest. Candidates in the post-implementation era were more likely to be transplanted within 90 days (66% versus 59%; P < 0.05) and were less likely to die while waiting (25% versus 32%; P < 0.05). In the post-implementation era, 92% of the candidates who entered the MELD/PELD 35+ category and underwent transplantation received their transplants while remaining in the MELD/PELD 35+ category, compared to 85% of the candidates in the pre-implementation era. In both eras, approximately half of the candidates who entered the MELD/PELD 35+ category and died did so while remaining in the MELD/PELD 35+ category.

For liver-intestine candidates, the probability of receiving a combined liver-intestine transplant within 12 months increased from 30.2% in the pre-implementation era to 46.8% in the post-implementation era (P < 0.05). Death rates within 12 months fell slightly, but not significantly, from 12.4% to 11.4%.

Posttransplant Outcomes

Figure 13 shows the comparison of 12-month patient survival for the primary deceased donor liver transplants performed in each era. There were no significant differences in survival (pre- versus post-) for either the overall group or the MELD/PELD 35+ group. Likewise, there were no significant differences in graft survival by era (Fig. 14) for either the overall group or the MELD/PELD 35+ group. Cox regression models comparing patient and graft survival also showed no significant differences by era (data not shown).