Survival Benefit of Split-Liver Transplantation for Pediatric and Adult Candidates

Abstract

Patient and graft survival are similar following whole-liver transplantations (WLTs) versus split-liver transplantations (SLTs) among pediatric and adult recipients, yet SLTs are rarely used. We sought to determine the survival benefit associated with accepting a splittable graft offer for SLT versus declining and waiting for a subsequent offer using 2010 to 2018 Scientific Registry of Transplant Recipients (SRTR) data on 928 pediatric and 1814 adult liver transplantation candidates who were ever offered a splittable graft. We compared eventual mortality, regardless of subsequent transplants, between those patients who accepted versus declined a split liver offer with adjustments for Pediatric End-Stage Liver Disease/Model for End-Stage Liver Disease (MELD) scores, diagnosis, and weight among pediatric candidates and matching for MELD score, height, and offer among adult candidates. Among pediatric candidates ≤7 kg, split liver offer acceptance versus decline was associated with a 63% reduction in mortality (adjusted hazard ratio [aHR], _0.170.37_0.80 [P = 0.01]; 93.1% versus 84.0% 1-year survival after decision). Within 1 year of decline for those ≤7 kg, 6.4% died and 31.1% received a WLT. Among pediatric candidates >7 kg, there was no significant difference associated with acceptance of a split liver offer (aHR, _0.631.07_1.82 [P = 0.81]; 91.7% versus 94.4% 1-year survival after decision). Within 1 year of decline for those >7 kg, 1.8% died and 45.8% received a WLT. Among adult candidates, split liver offer acceptance was associated with a 43% reduction in mortality (aHR, _0.390.57_0.83 [P = 0.005]; 92.2% versus 84.4% 1-year survival after decision). Within 1 year of decline for adult candidates, 7.9% died and 39.3% received a WLT. Accepting split liver offers for SLT could significantly improve survival for small children and adults on the waiting list.

Split-liver transplantation (SLT) provides an opportunity to significantly increase the supply of organs, thereby decreasing waitlist time and waitlist mortality for liver transplant candidates. Furthermore, outcomes following SLT are increasingly understood to be similar to those following whole-liver transplantation (WLT). Recent evidence suggests that, since 2010, pediatric recipients have experienced a 1-year patient survival rate of 95% and a 1-year graft survival rate of 90%, irrespective of whether the individual receives a split-liver or whole-liver graft.⁽¹⁾ An analysis of adult recipients demonstrated improvements since 2009, such that the risk of graft failure was similar across graft types through the first 5 years after transplant, with no evidence of early graft failure as seen in previous eras.⁽²⁾

Despite improvements in survival associated with SLT, in 2017 these grafts were used for only 18% of pediatric liver transplants (108 of 599 pediatric liver transplants) and 1.2% of adult transplants (87 of 7483 adult liver transplants), whereas 47 pediatric and 2511 adult candidates died while on the waiting list or were too sick for transplant.⁽³⁾ Furthermore, greater use of SLT provides an opportunity to increase access for transplant for 2 disadvantaged groups in organ allocation: younger children and smaller (ie, female) adult candidates. Young children have the highest waitlist mortality of any age group and might benefit greatly from increased access to organs for SLT.⁽³⁾ At the same time, female candidates and adult candidates with smaller statures are disadvantaged in size matching with whole deceased donor livers and might benefit from greater opportunities to accept a secondary graft that had been split.^(4-7) Despite this, the potential survival benefit following these decisions—first to split a graft for a child and second to accept a split graft for an adult—remains unclear.

To better understand the survival benefit associated with accepting versus declining a split liver offer for SLT and to inform decision making for liver transplant candidates and providers, we used national registry data to (1) characterize overall risk of mortality associated with accepting versus declining a split liver offer for pediatric and adult liver transplant candidates and (2) understand the next major outcome facing candidates who accepted or declined the offer.

Patients and Methods

DATA SOURCE

This study used data from the Scientific Registry of Transplant Recipients (SRTR). The SRTR data system includes data on all donor, waitlist candidates, and transplant recipients in the United States submitted by the members of the Organ Procurement and Transplantation Network (OPTN). The Health Resources and Services Administration, US Department of Health and Human Services provides oversight to the activities of the OPTN and SRTR contractors. This data set has previously been described elsewhere.⁽⁸⁾ This study used deidentified data and was exempted by the Johns Hopkins School of Medicine Institutional Review Board (NA_00042871).

SPLIT LIVER OFFERS

To study split liver offers, we first used the SRTR to identify 726 donors who were used in 1309 SLTs for 710 pediatric and 599 adult recipients between December 25, 2009, and April 30, 2018. We used the match run data to identify all candidates who were ever offered these 726 donors prior to their use in an SLT (n = 8834; Fig. 1). We refer to all offers on behalf of these donors as split liver offers because each donor graft was eventually used in an SLT. Offers that were never accepted by anyone (ie, discarded) were not included in the analysis. Offers reported as an acceptance that did not result in a transplant for the accepting patient were considered declines. Offers reported as a decline that were found to have resulted in a transplant for the declining patient were considered acceptances. Using offer sequence and age distribution of accepting candidates, we made the assumptions that initial split liver offers for each donor graft were made for the left-lateral segment (90.2% of those who accepted were pediatric candidates) and offers made after the first acceptance were made for the right trisegment (92.3% of those who accepted were adult candidates).

FIG. 1.

Pediatric and adult study population. Pediatric cases and controls not matched on MELD/PELD or height. Adult cases and controls matched on (1) the same graft offer, (2) ±2 MELD points, and (3) ±15 cm in height.

PEDIATRIC STUDY POPULATION

Among candidates who had the opportunity to accept a split liver offer, we identified 928 pediatric (<18 years old), liver-only transplant candidates of whom 617 were offered, accepted, and received transplants with the left-lateral segment, and 381 were offered and declined the split liver offer. We excluded 56 pediatric candidates who received transplants with the right trisegment and 37 who were simultaneously listed for kidney transplant. Of the 617 pediatric candidates who accepted and received transplants with the split liver offer, 70 (11%) had previously declined a different split liver offer. We compared demographic and clinical characteristics of pediatric candidates who ever accepted (n = 617) with those who were offered but never accepted a split liver offer (n = 311). For each split liver offer, we calculated donor-to-candidate weight ratio and absolute age difference in years to better understand the role of size mismatch in the decision to decline the split liver offer.

ADULT STUDY POPULATION

Among adult candidates who had the opportunity to accept a split liver offer, we identified 7958 adult, liver-only candidates of whom 442 were offered, accepted, and received transplants with the right trisegment and 7516 who were ever offered and declined the split liver offer. We compared demographic and clinical characteristics of adult candidates who ever accepted with those that were offered but never accepted a split liver offer. Of the 7516 who declined, 52 later accepted a different split liver offer. These 52 candidates were only included among those who accepted in the descriptive analysis, reducing the number who declined to 7464 in Table 1. We calculated donor-to-candidate weight ratio and absolute age difference in years to better understand the role of size mismatch in the decision to decline the split liver offer.

TABLE 1.

Characteristics of Adult Liver Transplant Candidates Who Declined (and Never Accepted) or Ever Accepted a Split Liver Offer: 2010 to 2018

Characteristic	Declined (n = 7464)	Accepted (n = 442)	P Value
Age at offer, years	58 (11)	59 (13)	0.97
Allocation MELD score at offer*	22 (12)	27 (9)	<0.001
Status 1	0.1	0.0	0.999
Exception points, ever granted	46.7	55.9	<0.001
Months from listing to offer	8.8 (17.6)	7.5 (14.1)	<0.001
Race/ethnicity			<0.001
Caucasian	66.5	62.9
African American	7.7	6.6
Hispanic	19.7	17.9
Asian America	4.5	11.5
American Indian	1.0	0.5
Native Hawaiian/Pacific	0.5	0.7
Islander/Multi-racial
Female sex	36.1	56.8	<0.001
Indication for transplant			<0.001
Hepatitis C	31.6	25.8
Nonalcoholic steatohepatitis	10.6	6.8
Hepatocellular carcinoma	13.8	15.4
Alcohol-related cirrhosis	16.6	14.5
Other	27.4	37.6
Hepatitis C virus antibody+	38.0	31.4	0.01
Height, cm*	170.2 (15.2)	165.1 (14.7)	<0.001
Weight, kg	82.6 (24.9)	70.9 (21.5)	<0.001
Donor-to-candidate weight ratio	0.8 (0.3)	1.0 (0.3)	<0.001
Donor-to-candidate age difference, years	37 (18)	35 (19)	0.02
On life support	0.8	0.5	0.44
Previous abdominal surgery	38.9	40.0	0.64
Portal vein thrombosis	6.0	4.3	0.15

NOTE: Data are provided as median (IQR) or percentage. Among those who declined more than 1 offer, characteristics from the first decline are reported.

^*Characteristics of candidates who accepted versus declined the livers prior to matching on candidate height and allocation MELD score are reported.

PEDIATRIC SURVIVAL BENEFIT

To characterize the survival benefit associated with accepting a split liver offer for pediatric candidates, we compared patient survival between pediatric candidates who accepted (n = 617) versus ever declined (n = 381) a split liver offer. The 70 pediatric patients who declined and later accepted a different split liver offer were treated as having a time-varying exposure: followed from the initial decline in the risk set of those who declined, censored at the date of acceptance, and followed from date of acceptance in the risk set of candidates who accepted a split liver offer. Pediatric candidates accepted the split liver offer after a median of 1 offer and thus we were unable to match pediatric candidates who accepted versus declined the same split liver offer. Instead, we compared mortality between pediatric candidates who accepted versus declined any split liver offer during the study period and used Cox proportional hazards regression to determine the association between split liver offer acceptance and patient mortality in a multivariable framework.

The goal of this analysis was to characterize the natural history or cumulative risk following the decision to accept or decline the split liver offer. In other words, acceptance includes perioperative risks or risks with an SLT, whereas declining includes risk of additional time spent on the waiting list and perioperative risks of a possible subsequent transplant. Inherent and potential risks associated with each decision fail to satisfy proportional hazards in the Cox proportional hazards regression model. By using Cox proportional hazards regression, we determined the average of these risks following the initial decision and used the overall accumulated risk to compare mortality. We also estimated mortality risk associated with split liver offer acceptance in the absence of the option of living donation. Thus, we censored pediatric candidates who received a living donor liver transplantation on the date of transplant. Pediatric candidates were followed from the date of offer until date of death, acceptance of a different split liver offer, living donor liver transplantation, or administrative censorship on April 30, 2018, irrespective of removal from the list or a subsequent deceased donor transplant. In other words, if they received a deceased donor graft or were removed from the waiting list for any other reason, they were still followed until mortality or administrative censorship. Pediatric candidates who declined multiple split liver offers were followed from their initial decline.

To account for clinical differences between pediatric candidates who accepted versus declined split liver offers, in the regression model we adjusted for Pediatric End-Stage Liver Disease (PELD) or Model for End-Stage Liver Disease (MELD) score, age, and weight at offer; indication for transplant; and status 1. To determine if the benefit of accepting a graft for SLT varied by pediatric candidate weight, we used exploratory data analysis and locally weighted scatter-plot smoothing curves to assess the functional form of weight and mortality risk and determine how the risk of mortality associated with split liver offer acceptance varied across pediatric weight. We identified a cut-point at 7 kg and tested the interaction between the acceptance of a split liver offer and candidate weight in the risk of mortality. We determined life-years gained from accepting a split liver offer by calculating the difference between the area under the survival curves associated with acceptance versus decline. To better understand the consequences of each decision, we report 1-year and 5-year cumulative incidences of subsequent outcomes following the decision to accept or decline a split liver offer during the study period. To understand how increasing follow-up time might affect the association between split liver offer acceptance and mortality and because time-varying hazard ratios (HRs) can introduce bias, we calculated average HRs over increasing follow-up times at 1 year, 2 years, and 6 years after decision.

ADULT SURVIVAL BENEFIT

To characterize the survival benefit associated with accepting a split liver offer for adult candidates, we began with the study population of adult candidates who accepted and received transplants with the right trisegment (n = 442 acceptances [n = 442 individual acceptors]) versus declined the split liver offer (n = 9138 declines [n = 7516 individual decliners]). To account for differences in organ quality, adult disease severity, and body size, we matched adult candidates who accepted the split liver offer with those who (1) declined the same split liver offer earlier in the match run, (2) were within ±2 allocation MELD points, and (3) were within ±15 cm in height of the adult who accepted the offer. After matching, the analysis included 242 cases (n = 242 acceptors) matched to 1572 controls (n = 1466 individual decliners), with a median of 3 controls (interquartile range [IQR], 7) per case. Adult candidates were excluded if they accepted the split liver offer at the top of the match run and had no controls who had previously declined the same split liver offer (n = 110). Candidates who declined and later accepted a different split liver offer (n = 52) were considered to have a time-varying exposure (ie, censored at the date of split liver offer acceptance and followed from date of acceptance in the risk set of candidates who accepted the split liver offer). Adult candidates included as a matched control for more than 1 case were treated as independent controls. To calculate mortality risk, adult candidates were followed from the date of offer until eventual date of death, different split liver offer acceptance, or administrative censorship on April 30, 2018, irrespective of removal from the waiting list or a subsequent transplant. We used Cox proportional hazards regression with stratification across matched candidates to determine the association between mortality and accepting a split liver offer.

The goal of this analysis was to compare eventual mortality following the initial decision to accept or decline the split liver offer, incorporating any subsequent transplants, improvements, or removals from the list that may have followed the decision. We included age at offer, sex, and diagnosis in the regression model to account for additional differences in candidate characteristics. To determine whether the association varied across candidate factors, we tested the interactions between age at offer, diagnosis, and sex and split liver offer acceptance. We determined life-years gained from accepting a split liver offer by calculating the difference between the area under the survival curves associated with acceptance versus decline. To better understand the consequences of each decision, we report the 1-year and 5-year cumulative incidences of subsequent outcomes following the decision to accept or decline a split liver offer during the study period. Given that 2 candidates with similar allocation MELD scores might differ by exception point status, we repeated the analysis with further matching on exception points. Acceptance might vary based on location of the pediatric candidate who accepted the left-lateral segment; thus, we characterized the survival benefit of accepting a split liver offer when in the same organ procurement organization (OPO) as the pediatric candidate and when in a different OPO as the pediatric candidate. To understand how increasing follow-up time might affect the association between split liver offer acceptance and mortality and because time-varying HRs can introduce bias, we calculated average HRs over increasing follow-up times at 1 year, 2 years, and 6 years after decision.

SENSITIVITY ANALYSES

To determine whether censoring for living donor liver transplantation introduced bias in the pediatric analysis, we performed a sensitivity analysis following pediatric candidates who declined the split liver offer until eventual death or administrative censorship irrespective of WLT, living donor liver transplantation, or removal from the list. We performed a sensitivity analysis excluding pediatric and adult candidates who were temporarily too sick for transplant or refused the offer for donor size from each survival analyses. Similarly, given that pediatric patients categorized at status 1A might improve and that status 1B pediatric patients might have less urgency, we performed a sensitivity analysis excluding status 1A and 1B candidates from the pediatric analysis. To determine whether excluding adult candidates lacking available matched controls introduced bias, we compared baseline characteristics and mortality between adult candidates with (n = 242) and without (n = 208) matched controls. To determine whether excluding unmatched controls introduced bias, we compared baseline characteristics and mortality between matched (n = 1466) and unmatched controls (n = 6050). To determine whether allowing adult candidates who declined multiple offers to serve as independent matched controls for multiple cases introduced bias, we performed a sensitivity analysis in the adult cohort using 1-to-1 matching without replacement.

STATISTICAL ANALYSIS

To compare patient characteristics between those who accepted versus declined split liver offers, we performed χ² tests for categorical variables, Fisher’s exact tests for categorical variables when expected cell counts fell below 5, and Wilcoxon rank sum tests for continuous variables. We used Cox proportional hazards regression to determine the association between split liver offer acceptance and mortality with adjustment and matching in the pediatric and adult models, respectively. All tests were performed for a 2-sided alternative hypothesis at a significance level of α = 0.05, except when using a χ² test of homogeneity to compare patient characteristics in Tables 1 and 2. Confidence intervals (CIs) are reported using the method of Louis and Zeger, as previously reported.⁽⁹⁾ All analyses were performed using Stata/SE 15 (StataCorp, College Station, TX).

TABLE 2.

Characteristics of Pediatric Liver Transplant Candidates Who Declined (and Never Accepted) or Ever Accepted a split liver offer: 2010 to 2018

Characteristic	Declined and Never Accepted* (n = 311)	Ever Accepted (n = 617)	P Value
Age at offer, years	2 (7)	1 (3)	<0.001
Age, years			<0.001
<1	26.4	37.6
1-5	42.4	48.6
6-10	17.7	10.2
11-17	13.5	3.6
Female sex	47.9	47.5	0.90
Race/ethnicity			1.00
Caucasian	47.6	46.2
African American	16.4	17.2
Hispanic	27.7	27.9
Asian American	5.1	5.7
American Indian	1.9	1.8
Native Hawaiian/Pacific Islander/	1.3	1.3
Multi-racial
Indication for transplant			<0.001
Biliary atresia	25.4	39.8
Metabolic disease	25.7	13.6
Hepatoblastoma	10.9	11.7
Other†	37.9	34.9
Weight, kg	11.9 (14.9)	9.3 (7.9)	<0.001
Weight ≤7 kg	21.0	29.0	0.01
Allocation MELD/PELD score at offer	30 (16)	35 (15)	0.17
Exception points, ever granted	67.2	58.0	0.007
Status 1	42.1	42.6	0.88
Months from listing to split liver offer	2.6 (5.7)	1.7 (3.6)	<0.001
Donor-to-candidate weight ratio	5.2 (5.3)	6.9 (5.1)	<0.001
Donor-to-candidate age difference, years	14 (9)	17 (11)	<0.001

NOTE: Data are provided as median (IQR) or percentage.

^*Among those who declined more than 1 offer, characteristics at the time of the first decline are reported.

^†Acute hepatic necrosis, cirrhosis, familial cholestasis, biliary hypoplasia, cholecystic liver disease, and other.

Results

PEDIATRIC STUDY POPULATION

Pediatric candidates who accepted a split liver offer (n = 617) were a median of 1 year old ([IQR, 3] versus 2 years old [IQR, 7] among those who declined) and 39.8% had biliary atresia (25.4% of decliners), and they had a median PELD/MELD score of 35 ([IQR, 15] versus 30 [IQR, 16] among decliners; Table 2). Although pediatric candidates who accepted the split liver offer were more likely to have low weight (29.0% ≤7 kg versus 21.0%; P < 0.001), the size distribution of these groups largely overlapped (Fig. 2). The donor-to-candidate weight ratio and age differences were higher among those who accepted versus declined the split liver offer: weight ratios of 6.9 (IQR, 5.1) versus 5.2 (IQR, 5.3; P < 0.001) and age differences of 17 (IQR, 11) versus 14 (IQR, 9; P < 0.001). Those who accepted were a median of 1 offer (IQR, 2; range, 1-446) down the match run, and those who declined were a median of 2 offers (IQR, 4; range, 1-443) down the match run. Among the 381 pediatric patients who declined a split liver offer, reasons for decline included donor size (42.5%), donor age/quality (27.0%), and temporarily too sick for transplant (12.1%). Decline codes were missing for 13.1% of declines (Table 3).

FIG. 2.

Distribution of weight (kg) and height (cm) of pediatric liver transplant candidates who accepted (triangle) and declined (circle) split liver offers. Symbols are shown with transparency to illustrate increased density in areas of overlap. Each shape represents a candidate who had the opportunity to accept a split liver offer. Axes are on a log scale so that differences at low values can be observed.

TABLE 3.

Decline Codes for Pediatric Candidates Who Ever Declined a split liver offer and Matched Adult Candidates Who Declined the split liver offer

Decline Code	Pediatric (n = 381)	Adult (n = 1572)
800 (improved)	6 (1.6)	1 (0.1)
801 (temporarily too sick)	46 (12.1)	63 (4.0)
825 (operational/transplant center)	—	1 (0.1)
830 (donor age/quality)	103 (27.0)	177 (11.3)
831 (donor size)	162 (42.5)	314 (20.0)
832 (donor ABO)	1 (0.3)	1 (0.1)
833 (donor social history)	4 (1.1)	—
836 (anatomical defect)	1 (0.3)	28 (1.8)
837 (organ-specific donor issue)	1 (0.3)	140 (8.9)
898 (other)	7 (1.8)	137 (8.7)
Missing decline code	50 (13.1)	684 (43.5)

NOTE: Data are provided as n (%).

ADULT STUDY POPULATION

Adult candidates who accepted the split liver offer (n = 442) had a median age of 59 years ([IQR, 13] versus 58 years among decliners [IQR, 11]) and 62.9% were Caucasian (66.5% of decliners), 11.5% were Asian American (4.5% of decliners), 56.8% were female (36.1% of decliners), 25.8% had hepatitis C virus as their primary indication for transplant (31.6% of decliners), and 6.8% had nonalcoholic steatohepatitis (10.6% of decliners). Those who accepted had a median MELD score of 27 ([IQR, 9] versus 22 [IQR, 12] among decliners; Table 1). When stratified by sex, those who accepted the split liver offer were smaller than those who declined (female: median 64.9 kg [IQR, 18.6] versus 73.4 kg [IQR, 22.1], P < 0.001; male: median 76.7 kg [IQR, 21.8] versus 87.5 kg [IQR, 23.6], P < 0.001; Figs. 2 and 3). The donor-to-candidate weight ratio was marginally, but not clinically, higher among those who accepted versus declined: 1.0 (IQR, 0.3) versus 0.8 (IQR, 0.3; P < 0.001). Similarly, donor-to-candidate age differences were not clinically different between those who accepted versus declined: 35 (IQR, 19) versus 37 (IQR, 18; P = 0.02). Those who accepted were a median of 22 offers (IQR, 37; range, 3-546) down the match run. Among the 1572 matched controls who declined, the reasons for decline included donor size (20.0%), donor age/quality (11.3%), and temporarily too sick for transplant (4.0%). Decline codes were missing for 43.5% of declines (Table 3).

FIG. 3.

Distribution of height (cm) and weight (kg) of adult female and male liver transplantation candidates who accepted (triangle) and declined (circle) split liver offers. Symbols are shown with transparency to illustrate areas of increased density.

SURVIVAL BENEFIT AMONG PEDIATRIC CANDIDATES

Among pediatric candidates who accepted versus declined the split liver offer, survival was 92.1% for both groups 1 year after decision and 89.6% versus 88.2% 5 years after decision (log-rank P = 0.56). split liver offer acceptance was associated with a 22% reduction in risk of mortality, although this was not statistically significant (adjusted HR [aHR], _0.500.78_1.21; P = 0.29). However, mortality risk associated with acceptance of a split liver offer varied by pediatric candidate weight (P value of test for interaction = 0.03).

Among pediatric candidates ≤7 kg, split liver offer acceptance was associated with a 63% reduction in mortality (aHR, _0.170.37_0.80; P = 0.01; Fig. 4; Table 4). This reduction in mortality translated to 0.94 life-years gained (11.3 months) for those who accepted the split liver offer for 8.3 years of follow-up. The survival rate 1 year after decision was 93.1% versus 84.0% (Fig. 3). The adjusted association between split liver offer acceptance and mortality for those ≤7 kg was 0.35 (95% CI, 0.16-0.80; P = 0.01), 0.35 (95% CI, 0.16-0.77; P = 0.009), and 0.37 (95% CI, 0.17-0.80; P = 0.01) after 1, 2, and 6 years of follow-up, respectively. Within 1 year of split liver offer decline among those ≤7 kg, 6.4% died without a transplant, 31.1% received a WLT, 39.0% received a different SLT, and 7.5% received a living donor liver transplantation (Table 5). Within 1 year of split liver offer acceptance, 5.7% died, 5.2% received retransplants, and 89.1% were still living with the accepted graft. Those patients ≤7 kg who accepted versus declined were similar in age, race, sex, exception points, and weight distribution (P values > 0.3). Those who accepted were more likely to have higher MELD/PELD scores (35 [IQR, 15] versus 30 [IQR, 20.5]), more likely to have biliary atresia (75.4% versus 55.4%), and less likely to have metabolic disease (3.9% versus 18.5%) compared with those patients ≤7 kg who declined (P values < 0.02). These factors were included in the adjusted analysis and do not fully explain differences in mortality.

FIG. 4.

Pediatric patient survival following the decision to accept versus decline a split liver offer stratified by candidate weight. Pediatric candidates followed from the date of decision until death or administrative censorship on April 30, 2018, irrespective of subsequent deceased donor SLT or WLT or removal from the waiting list. Pediatric candidates were censored for living donor liver transplantation.

TABLE 4.

Pediatric Candidate Factors Associated With Mortality Following the Decision to Accept Versus Decline the split liver offer

Pediatric Characteristic	aHR	P Value
Accepted versus declined, ≤7 kg	0.170.370.80	0.01
Accepted versus decline, >7 kg	0.631.071.82	0.81
Per year of age	0.930.991.06	0.77
Per unit of PELD/MELD	1.001.021.04	0.04
Status 1	1.333.9611.80	0.01
Diagnosis
Biliary atresia	Reference	—
Metabolic disease*	0.340.781.75	0.54
Hepatoblastoma	0.811.814.05	0.15
Other†	1.021.873.43	0.04

NOTE: Mortality risk associated with the decision to accept varied by pediatric weight (P interaction = 0.03).

^*Metabolic disease includes alpha-1-antitrypsin deficiency, Wilson’s disease, hemochromatosis, glycolytic storage disease, hyperlipidemia, tyrosinemia, primary oxalosis, hyperoxaluria, and maple syrup urine disease.

^†Other includes acute hepatic necrosis, cirrhosis, familial cholestasis, biliary hypoplasia, and cholecystic liver disease.

TABLE 5.

The 1-Year and 5-Year Cumulative Incidences of the Next Major Outcome Following the Decision to Decline or Accept the Split Liver Offer Among Pediatric and Adult Liver Transplant Candidates

	Cumulative Incidence, %
	Pediatric Candidates≤7 kg		Pediatric Candidates>7 kg		Adult Candidates
Outcomes among those who declined	1 Year	5 Years	1 Year	5 Years	1 Year	5 Years
Died without a transplant	6.4	6.4	1.8	3.0	7.9	12.4
WLT	31.1	32.9	45.8	48.3	39.3	47.7
Different SLT	39.0	39.0	20.6	22.2	1.3	1.6
Living donor liver transplantation	7.5	9.1	3.9	3.9	1.0	1.3
Removed from waiting list (deteriorated)	1.1	1.1	1.1	1.1	6.5	9.8
Removed from waiting list (improved)	2.2	8.0	7.4	11.0	1.9	5.5
Removed from waiting list (other)	2.2	3.5	5.0	8.2	7.7	15.6
Still waiting	10.5	0.0	14.5	2.4	34.3	6.2
Outcomes among those who accepted
Died	5.7	7.3	7.4	9.9	6.5	16.9
Retransplanted	5.2	7.1	5.5	7.6	3.8	3.8
Living with accepted SLT	89.1	85.6	87.1	82.5	89.7	79.3

Among candidates >7 kg, acceptance of a split liver offer was not associated with a statistically significant difference in the risk of mortality (aHR, _0.631.07_1.82; P = 0.81). Survival 1 year after decision was 91.7% versus 94.4% (Figs. 4 and 3). The adjusted association between split liver offer acceptance and mortality for those >7 kg was 1.37 (95% CI, 0.72-2.60; P = 0.96), 1.20 (95% CI, 0.66-2.16; P = 0.55), and 1.07 (95% CI, 0.63-1.82; P = 0.81) after 1, 2, and 6 years of follow-up, respectively. Within 1 year of split liver offer decline among those patients >7 kg, 1.8% died without a transplant, 45.8% received a WLT, 20.6% received a different SLT, and 3.9% received a living donor transplantation. Within 1 year of split liver offer acceptance, 7.4% died, 5.5% received retransplants, and 87.1% were still living with the accepted graft. The 5-year cumulative incidences are shown in Table 5.

SURVIVAL BENEFIT AMONG ADULT CANDIDATES

Among adult candidates who accepted versus declined the split liver offer (after matching on offer, MELD score, and height), survival was 92.2% versus 84.3% 1 year after decision and 81.9% versus 65.7% 5 years after decision (log-rank P < 0.001; Fig. 5). After further adjustment for age, sex, and diagnosis, split liver offer acceptance was associated with a 43% reduction in mortality (aHR, _0.390.57_0.83, P = 0.005; Table 6). The adjusted association between split liver offer acceptance and mortality for matched adult candidates was 0.49 (95% CI, 0.29-0.85; P = 0.01), 0.50 (95% CI, 0.32-0.79; P = 0.003), and 0.54 (95% CI, 0.36 = 0.80; P = 0.002) after 1, 2, and 6 years of follow-up, respectively. We calculated 1.03 life-years gained by acceptance versus decline. Within 1 year of split decline, 7.9% died without a transplant, 39.3% had received a WLT, 1.3% received a different SLT, and 1.0% received a living donor liver transplantation. Within 1 year of split liver offer acceptance, 6.5% of adult candidates died, 3.8% received retransplants, and 89.7% were still living with the accepted SLT. The 5-year cumulative incidences are shown in Table 5. We did not observe an interaction between split liver offer acceptance and sex (P = 0.36), age at offer (P = 0.98), diagnosis (P = 0.25), or MELD score at offer (P = 0.34). Although the association did not vary statistically across MELD scores, a survival benefit was observed with lower MELD scores. For adults with MELD scores <22 (n = 756), 22 to 28 (n = 586), and >28 (n = 432), acceptance was associated with 57%, 55%, and 31% reductions in mortality, respectively: MELD scores <22: aHR, _0.200.43_0.91 (P = 0.03); MELD scores 22 to 28: aHR, _0.240.45_0.86 (P = 0.02); and MELD scores >28: aHR, _0.370.69_1.28 (P = 0.2; P interaction = 0.34). When matching on offer, MELD score, height, and exception points, there were 207 cases and 1085 matched controls. Within this smaller matched population, accepting a split liver offer was associated with a 27% reduction in mortality compared with declining the same split liver offer (aHR, 0.63; 95% CI, 0.42-0.95; P = 0.03). When considering OPO, the point estimates of the association remained similar to the overall findings from the main model, but the stratified findings were not statistically significant. Specifically, among matched adults offered the right trisegment in the same OPO as the pediatric candidate who accepted the left-lateral segment, there was a 29% reduction in mortality associated with acceptance versus decline (aHR, 0.62; 95% CI, 0.36-1.04; P = 0.07). Among adults offered the right trisegment in a different OPO as the pediatric candidate, there was a 47% reduction associated with acceptance versus decline (aHR, 0.53; 95% CI, 0.28-1.007; P = 0.05).

FIG. 5.

Adult patient survival following the decision to accept versus decline a split liver offer. Adult candidates matched on the split liver offer, MELD score at offer, and height. Adult candidates followed from the date of decision until death or administrative censorship on April 30, 2018, irrespective of subsequent deceased donor SLT or WLT, living donor liver transplantation, or removal from the waiting list.

TABLE 6.

Adult Candidate Factors Associated With Mortality Following the Decision to Accept Versus Decline the Split Liver Offer

Adult Characteristic	aHR	P Value
Accepted versus declined	0.390.570.83	0.005
Per year of age	1.011.021.04	<0.001
Male sex	0.690.911.21	0.59
Race/ethnicity
Caucasian	Reference	—
African American	0.560.821.23	0.34
Hispanic	0.680.901.18	0.45
Other*	0.771.131.68	0.53
Diagnosis
Hepatitis C virus	Reference	—
Nonalcoholic steatohepatitis	0.821.191.72	0.36
Hepatocellular carcinoma	0.731.011.39	0.97
Alcohol-related cirrhosis	0.821.121.53	0.49
Other	0.760.991.30	0.96
Height, per cm	0.970.991.01	0.17
Weight, per kg	1.001.011.02	0.001

^*Includes Asian American, American Indian, Alaskan, Hawaiian Native, and other.

SENSITIVITY ANALYSES

Inferences were unchanged in a sensitivity analysis that did not censor for living donor liver transplantation among pediatric candidates (aHR of acceptance versus decline among ≤7 kg, _0.180.40_0.86; among >7 kg, _0.641.09_1.86). When excluding the 46 pediatric candidates who were too sick to transplant, overall inferences remained unchanged: aHR ≤7 kg, 0.40 (95% CI, 0.17-0.91; P = 0.03) and aHR >7 kg, 1.35 (95% CI, 0.73-2.48; P = 0.33). After further exclusion of 162 pediatric candidates who refused for donor size, there was no longer a statistically significant association between split liver offer acceptance and mortality among those patients ≤7 kg: aHR, 0.63 (95% CI, 0.18-2.25; P = 0.48). When excluding status 1A (n = 135) and 1B candidates (n = 287), 575 pediatric candidates remained. In this subpopulation, the point estimate indicating the association between accepting the split liver offer and mortality was similar to the overall model, yet CIs widened to no longer be statistically significant for candidates ≤7 kg (aHR, 0.41; 95% CI, 0.15-1.11; P = 0.08) and for those candidates >7 kg (aHR, 1.72; 95% CI, 0.70-4.25; P = 0.24). When excluding 63 adult candidates who were temporarily too sick and 314 who refused the offer for donor size, inferences remained unchanged: aHR associated with acceptance, 0.57 (95% CI, 0.39-0.83; P = 0.004). Adult cases with matched controls had slightly lower allocation MELD scores (median, 26 [IQR, 11] versus median, 28 [IQR, 11]; P = 0.002), more likely to have nonalcoholic steatohepatitis (8.7% versus 5.0%), more likely to have alcohol-related cirrhosis (18.3% versus 11.6%), and less likely to have hepatocellular carcinoma (11.1% versus 19.0%; P = 0.03), but were similar in age, sex, and race/ethnicity to adult cases without matched controls (P values > 0.1). There was an elevated but not statistically significant risk of mortality among cases with controls (aHR, _0.711.21_2.06; P = 0.48). Adult matched controls had similar allocation MELD scores (22 [IQR, 12] versus 22 [IQR, 13]) and median age (59 years versus 58 years) but were slightly more likely to have had Hispanic/Latinx ethnicity (24.1% versus 18.7%), be female (41.8% versus 34.9%), and have a history of exception points (53.5% versus 45.1%) than adult unmatched controls. Matched versus unmatched controls had a similar risk of mortality following the decision to decline: aHR, 1.07 (95% CI, 0.96-1.18; P = 0.24). When matching adult candidates in a 1-to-1 framework without replacement, inferences remained unchanged from the main analysis (aHR of acceptance versus decline, _0.230.43_0.80).

Discussion

In this study of 1814 matched adult candidates and 928 pediatric candidates with the opportunity to accept a split liver offer while on the waiting list in the United States between 2010 and 2018, we found a clinically and statistically significant survival benefit associated with acceptance and subsequent SLT among pediatric candidates ≤7 kg (63% reduction in mortality) and adult candidates (43% reduction in mortality). For pediatric candidates >7 kg, we did not observe a statistically significant difference in the survival benefit associated with accepting a split liver offer. This lack of survival benefit observed among pediatric candidates >7 kg might be driven by the low cumulative incidence of death while on the waiting list (1.8%) and high cumulative incidence of subsequent WLTs (45.8%) within the first year of split liver offer decline. By contrast, pediatric candidates ≤7 kg and adults candidates faced 6.4% and 7.9% cumulative incidence of death while on the waiting list and only 31.1% and 39.3% cumulative incidence of a WLT within the first year of split liver offer decline, respectively.

There is emerging consensus that outcomes following SLT are similar to WLT for pediatric and adult recipients, especially in recent years. A recent study by our group demonstrated that pediatric recipients from 2010 to 2015 receiving whole or split grafts had similar patient and graft survival rates, and recipients of living donor grafts had improved outcomes, further supporting technical mastery over vascular and biliary complications that historically occur at higher rates with partial grafts.⁽¹⁾ Similarly, studies of adult recipients have shown similar patient and graft survival rates between WLT and SLT in recent years.^(2,10) However, these studies do not address the decision that candidates and health care teams are likely to experience: to accept the split liver offer today or wait for a different deceased donor offer that might never come. By comparing candidates who were ever offered eventually transplanted split liver offers, we attempted to exclude candidates who did not have the opportunity to accept the split liver offer and include those who had the opportunity and were theoretically able to accept the split liver offer to better understand the survival benefit associated with these grafts.

Our finding that there exists a survival benefit for accepting a split liver offer for pediatric recipients ≤7 kg, compared with larger children, augments previous work from our group on the association between graft type and posttransplant survival in pediatric liver transplantation.⁽¹¹⁾ We showed that smaller children had similar graft survival irrespective of whether they received a WLT or SLT, whereas larger children had slightly improved graft survival following WLT. Thus, it is not surprising that earlier transplantation using a split graft confers a survival advantage compared with continuing to wait for a whole graft. In addition, although it might make sense for patients to decline split liver offers driven by size mismatch concerns, we found that pediatric candidates who accepted the split liver offer had slightly higher donor-to-candidate weight ratios than those who declined split liver offers. Thus, size mismatch does not fully explain the decline of split liver offers.

In this analysis, we use Cox proportional hazards regression to characterize the natural history or accumulated risk following the decision to accept or decline the split liver offer. We followed candidates who declined the split liver offer until eventual mortality even if they received a WLT or were removed from the waiting list for other reasons after declining the split liver offer. The mortality risks facing those who accepted the split liver offer (perioperative and postoperative risks, changes in quality of life) are different from risks facing those who declined the split liver offer (additional time on the waiting list without a transplant, potential perioperative and postoperative risks later if they received a different transplant). These risks are also occurring at different times with respect to the decision to accept or decline the split liver offer, and thus our approach fails the proportional hazards assumption of Cox regression. Given that an overall average hazard radio (HR) can be misleading or make varied risks between groups difficult to identify and that time-varying HRs can introduce bias, we calculated average HRs for increasingly longer periods of follow-up to illustrate changes in the distribution of risk and minimize bias.⁽¹²⁾ The average or accumulated risk incorporates the different risks facing those who accepted and those who declined the split liver offer, and with each model based on increasing follow-up time we observed minimal changes in the aHRs. Thus, the survival benefit (reduced mortality) associated with the acceptance of split liver offers observed for adult candidates and the subgroup of children ≤7 kg incorporates the varied risks facing those who accepted and declined these offers.

Despite the potential for greater use of SLT to reduce the consequences of organ shortage, the procedure remains underused. Less than 5% of potentially splittable livers, as designated by strict criteria, are used for SLTs, and approximately 1% of all adult recipients use these grafts.^(13,14) Similarly, a recent analysis of offers by Ge et al. found that the second portion was offered a median of 12 times before it was accepted and that use of split livers is concentrated by a very small number of centers.⁽¹⁴⁾ Valentino et al. created a simulation based on deceased donor organs that were suitable for splitting and evaluated the consequences on the waiting list for receiving an SLT compared with WLT including waitlist times and outcomes.⁽¹⁵⁾ These authors found that children who received a split liver in their simulation had shorter waitlist times and decreased risk of death compared with those who did not. Although our findings further support these observations, increased acceptance of SLTs in the present study would not have led to increased use but, rather, earlier transplantation, as we only studied split liver offers that were all eventually accepted by a pediatric or adult candidate.

As the organ shortage continues to be a major driver of increased waitlist morbidity and mortality, small children have the highest waitlist mortality of any age group. One-third of these waitlist deaths occur in centers that have demonstrated experience in SLT.⁽¹⁴⁾ It is also well established that adult female and small-stature candidates are disadvantaged in liver transplantation.⁽⁷⁾ Although the implementation of the MELD system has helped to reduce disparities with respect to access to organs, separate studies have shown that women actually had decreased access following implementation.^(17,18) Women also have higher waitlist mortality and greater need for pretransplant hospitalization.^(6,19) In this context, Ge et al. showed that smaller height was 1 of the largest drivers of acceptance of the second portion of an SLT and that the majority of recipients were female.⁽¹⁶⁾ Expansion of SLT therefore represents a great opportunity to reduce disparities in access to organs not only for children but also for women.^(4,5)

The limitations of our study merit discussion. We were unable to incorporate meaningful outcomes such as biliary and vascular complications that might influence patient health and quality of life. It is possible that candidates who accepted the split liver offers will have greater survival benefit but face greater morbidity associated with other complications. Such a trade-off may make the decision to use a split graft less clear. However, for adult candidates, frequency of vascular thromboses and biliary complications has abated such that they now occur in similar rates among recipients of either graft type.^(10,20,21) In addition, the study might be subject to confounding by indication as candidates for whom physicians felt comfortable accepting an offer for a split graft might be different from those for whom it was declined. The possibility exists that the favorable outcomes following SLT for small children and adults were the result of careful candidate selection on the part of health care teams. In the adult analysis, we attempted to mitigate this bias through matching on disease severity, but we acknowledge that some candidate characteristics remain uncaptured by these metrics. The results should thus be considered in the context of potential unmeasured confounding, as we are unable to fully account for uncaptured differences between patients who accepted versus declined the split liver offers. By identifying patients who had the opportunity to accept a graft for SLT yet declined as controls we attempted to select the most approximate comparison group to those who had the opportunity to accept and accepted the graft. Our sensitivity analyses excluding those patients who declined for “temporarily too sick” and “donor size” attempted to exclude those who should not have received a split liver offer, although it is notable that decline codes are largely incomplete and reasons for acceptance and decline are not well captured. To minimize differences in organ quality that might drive offer decision making, we compared adult recipients who received the same offer. However, we did not have a sufficient sample size to match on offer in the pediatric analysis. Although matching restricted our sample size in the adult analysis, sensitivity analyses found that unmatched and matched controls had similar outcomes, indicating that excluding unmatched controls did not alter our overall findings. Cases with controls had slightly elevated mortality risk when compared with cases without controls that were excluded, indicating that our overall findings might be an underestimation of the difference in mortality observed between those patients who accepted versus declined. Despite these limitations, we used nationally representative data to study difficult acceptance decisions facing adult and pediatric candidates on the waiting list.

In conclusion, although the choice of accepting a graft to be used as a split liver may be difficult for transplant teams, the decision to accept was associated with substantially improved survival for both small children and adult recipients. Greater efforts should therefore be made to promote the use of SLTs for populations in whom they can provide the greatest benefit.

Abbreviations

aHR

adjusted hazard ratio

CI

confidence interval

HR

hazard ratio

IQR

interquartile range

MELD

Model for End-Stage Liver Disease

OPO

organ procurement organization

OPTN

Organ Procurement and Transplantation Network

PELD

Pediatric End-Stage Liver Disease

SLT

split-liver transplantation

SRTR

Scientific Registry of Transplant Recipients

WLT

whole-liver transplantation