Projected 20- and 30-Year Outcomes for Pediatric Liver Transplant Recipients in the United States

Abstract

Background:

Observed long-term outcomes no longer reflect the survival trajectory facing pediatric liver transplant (LT) recipients today. We aimed to use national registry data and parametric models to project 20- and 30-year post-transplant outcomes for recently transplanted pediatric LT recipients.

Methods:

We conducted a retrospective cohort study of 13,442 first-time pediatric (age <18) LT recipients using 1987 to 2018 Scientific Registry of Transplant Recipients data. We validated the proposed method (ie, to project long-term patient and graft survival using parametric survival models and short-term data) in 2 historic cohorts (1987–1996 and 1997–2006) and estimated long-term projections among patients transplanted between 2007 and 2018. Projections were stratified by graft type, recipient age, and indication for transplant.

Results:

Parsimonious parametric models with Weibull distribution can be applied to post-transplant data and used to project long-term outcomes for pediatric LT recipients beyond observed data. Projected 20-year patient survival for pediatric LT recipients transplanted in 2007 to 2018 was 84.0% (95% confidence interval 81.5–85.8), compared to observed 20-year survival of 72.8% and 63.6% among those transplanted in 1997 to 2006 and 1987 to 1996, respectively. Projected 30-year survival for pediatric LT recipients in 2007 to 2018 was 80.1% (75.2–82.7), compared to projected year survival of 68.6% (66.1–70.9) in the 1997 to 2006 cohort and observed 30-year survival of 57.5% in the 1987 to 1996 cohort. Twenty- and 30-year patient and graft survival varied slightly by recipient age, graft type, and indication for transplant.

Conclusions:

Projected long-term outcomes for recently transplanted pediatric LT recipients are excellent, reflective of substantial improvements in medical care, and informative for physician-patient education and decision making in the current era.

Liver transplantation (LT) is the optimal treatment for children with end-stage liver disease (ESLD) and current short-term outcomes are excellent. Recent registry data indicate 90% 1-year graft survival and 82% 5-year graft survival for pediatric LT recipients (1). These outcomes have steadily improved over the last 15 years largely due to improvements in survival during the immediate post-transplant period (ie, first 30 days), especially for recipients of living-donor and deceased-donor split grafts (2). Advances in the surgical management of partial grafts, optimization of donor selection, and medical management have led to substantially improved patient and graft survival over several decades (3).

Children with ESLD and their parents frequently seek out information regarding long-term outcomes following transplant (4). A limited number of single centers have reported on their long-term experience and provided estimates of patient and graft survival beyond 10 years. One analysis of 806 pediatric LT recipients reported a 20-year patient and graft survival of 69% and 53%, respectively, whereas a study of 808 pediatric LT recipients (ages 3–18) reported 18-year patient survival of 65% (64% for those ages <3) (5,6). Although these empiric results provide some information to families about the likelihood of long-term survival for children awaiting a transplant, interpretation is limited in that these results reflect outcomes from surgeries performed in the 1980s and do not properly incorporate subsequent advances in surgical technique and postoperative management that have yielded better short- and mid-term outcomes.

To account for improvements in short-term post-transplant outcomes and provide relevant projections of long-term post-transplant outcomes to patients and their families, we used parametric models to project 20- and 30-year outcomes among recently transplanted pediatric LT recipients for whom only 10 years of follow-up data were available.

METHODS

Data Source

This study used data from the Scientific Registry of Transplant Recipients (SRTR). The SRTR data system includes data on all donors, waitlisted candidates, and transplant recipients in the United States, submitted by the members of the Organ Procurement and Transplantation Network and has been described elsewhere (7). The Health Resources and Services Administration, US Department of Health and Human Services, provides oversight to the activities of the Organ Procurement and Transplantation Network and SRTR contractors. The interpretation and reporting of these data are the responsibility of the author(s) and in no way should be seen as an official policy of, or interpretation by, the SRTR or the US Government. This study used deidentified data and was exempted by the Johns Hopkins School of Medicine Institutional Review Board with a waiver of informed consent (NA_00042871).

Study Population

We studied 13,442 pediatric (age <18) first-time LT recipients who received a whole (n = 8784), split (n = 2984), or living (n = 1674) donor transplant from October 1, 1987 to April 30, 2018. We excluded 1811 pediatric retransplant recipients and 18 pediatric recipients of whole living-donor liver transplants, in which domino transplantation was likely performed. Pediatric LT recipients were categorized into 3 cohorts based on calendar year of transplant: 1987 to 1996, 1997 to 2006, and 2007 to 2018. We compared recipient (eg, age, race/ethnicity, sex, diagnosis, weight, and the allocation pediatric end-stage liver disease/model for end-stage liver disease (PELD/MELD), defined as the greater score between the calculated score or the score with exception points), donor (eg, graft type, age, cause of death, donation after cardiac death), and transplant characteristics (eg, cold ischemia time, national sharing) between recipients within each cohort using Kruskal-Wallis tests (nonparametric) for continuous variables and χ² tests for categorical variables; medians and interquartile ranges (ie, 25th and 75th percentiles) were reported for continuous variables.

Patient and Graft Survival

To estimate post-transplant patient survival, we followed recipients from date-of-transplant until date-of-death or administrative censorship on April 30, 2018. To estimate graft survival, we followed patients from date-of-transplant until the earliest of death, retransplant, graft loss, or administrative censorship.

Projecting 20- and 30-Year Survival

Our objective was to use a parametric model to create 20- and 30-year projections of graft and patient survival following LT in the 2007 to 2018 cohort given improvements in short-term outcomes. Unlike the semiparametric Cox proportional hazards model, parametric models assume an underlying distribution of hazard over time and allow for projection beyond the observed data if the hazard follows the selected distribution (8,9). Thus, although only 10 years of follow-up have been observed among those transplanted in 2007 to 2018, we aimed to use an appropriately selected parametric model to project 20- and 30-year patient and graft survival based on 10 years of observed data.

Model Selection and Validation

To validate our proposed method, we used 2 cohorts for which 20-year outcomes (1997–2006 cohort) and 30-year outcomes (1987–1996 cohort) have been observed. We censored all of the data at 10 years post-transplant to mirror the availability of data in the 2007 to 2018 cohort; in other words, we acted as though we only had 10 years of follow-up in the 2 historic cohorts. We fit 3 unadjusted parametric models (generalized gamma [GG], lognormal, and Weibull) to the 10 years of data in each of the 2 cohorts and estimated the expected 20- and 30-year patient survival from each model (10–12). We compared observed to expected survival at 20 and 30 years for each model. Exploratory analyses suggested the fit of observed data to the parametric models was poorest during the first 2 years post-transplant. To improve our model fit, we excluded the initial 2 years of post-transplant data from the parametric models, including only patients who survived to 2 years, and again censored all patients at 10 years post-transplant. We fit each of the 3 parametric models to the left-truncated 8 years of data in the 1987 to 1996 and 1997 to 2006 cohorts and compared observed to expected 20- and 30-year post-transplant survival among those who survived to 2 years post-transplant. We report the absolute difference between observed and expected survival at 20 and 30 years for each model, and ultimately found that the Weibull model based on 8 years of data post-transplant most closely approximated the observed survival distribution (Table, Supplemental Digital Content 1, http://links.lww.com/MPG/B754).

Projected Long-term Outcomes

Based on our model selection, we used the 2007 to 2018 cohort, excluded those who did not survive to 2 years post-transplant, and modeled survival using a parametric model with Weibull distribution. We projected 20- and 30-year patient and graft survival in the 2007 to 2018 cohort based on the Weibull model parameters. To obtain 95% confidence intervals (CIs), we bootstrapped the entire population with replacement and repeated model estimation over 100 iterations. To graphically illustrate long-term survival, we plotted the Kaplan-Meier curves (nonparametric, observed) alongside the projected survival probabilities and 95% CIs. To reincorporate outcomes from within the first 2 years post-transplant, we multiplied the projected survival probabilities by observed survival at 2 years and plotted the product.

Projections by Subgroup

To project long-term outcomes based on relevant criteria at the time of transplant, we used a parsimonious adjusted parametric model with Weibull distribution to estimate the associations between post-transplant mortality and graft type, recipient age, and diagnosis. To validate this method, we ran adjusted parametric models among the 1987 to 1996 and 1997 to 2006 cohorts using 8 years of post-transplant data, estimated expected survival, and evaluated the absolute difference between observed and expected survival per subgroup at 20 and 30 years post-transplant (Table, Supplemental Digital Content 2, http://links.lww.com/MPG/B754). To improve stability of long-term projections and increase the number of recipients within each subgroup, we combined diagnoses with nonstatistically significant differences in mortality: biliary atresia (BA)/metabolic disease and non-BA/nonmetabolic disease, which included acute hepatic necrosis, malignancy, and other/unknown diagnoses. To obtain 95% CIs, we bootstrapped the entire population with replacement and stratification by age, graft type, and diagnosis and repeated the model estimation over 100 iterations. We report 10-year observed, 10-year expected, absolute difference between 10-year observed-expected, 20-year projected, and 30-year projected patient and graft survival among recently transplanted pediatric LT recipients by age (<1, 1–5, 6–10, 11–17), graft type (whole, split, and living donor), and diagnosis (BA/metabolic, non-BA/nonmetabolic). The absolute difference at 10 years should be used to inform variability that may be expected given the sample size of a particular subgroup.

Statistical Analysis

All statistical tests used a 2-sided α of 0.05. CIs are reported using the method of Louis and Zeger (13), as previously reported. All analyses were performed using Stata/SE 15 (StataCorp, College Station, TX).

RESULTS

Study Population

We studied 13,442 pediatric LT recipients, of whom 3787, 4467, and 5188 were transplanted in 1987 to 1996, 1997 to 2006, and 2007 to 2018 cohorts. Median age at transplant remained consistent over time at 2 years of age (P = 0.9) (Table 1). Between 1987 and 1996 and 2007 and 2018, the percentage who received a whole deceased-donor graft decreased (82.1%–59.0%), a split deceased-donor graft increased (11.0%–28.4%), and a living-donor graft increased (6.9%–12.6%). BA was the primary indication for LT for 51.1%, 41.1%, and 40.2% of recipients in each era, respectively.

TABLE 1.

Characteristics of 13,442 pediatric liver transplant recipients, stratified by era of transplant

	1987–1996 N = 3787	1997–2006 N = 4467	2007–2018 N = 5188	P
Age at transplant, median (IQR)	2 (0, 8)	2 (0, 9)	2 (0, 8)	0.87
Age categories, y				<0.001
<1	28.0%	31.9%	29.4%
1–5	38.9%	33.0%	37.8%
6–10	14.0%	13.1%	12.8%
11–17	19.1%	22.0%	19.9%
Male sex	49.0%	46.8%	49.7%	0.015
Primary diagnosis				<0.001
Biliary atresia/hypoplasia	51.1%	41.1%	40.2%
Metabolic disease	12.6%	12.2%	16.0%
AHN	11.9%	14.0%	12.3%
Malignancy	2.4%	5.3%	9.7%
Other/unknown	22.0%	27.4%	21.9%
Allocation PELD/MELD, median (IQR)	N/A	19 (9, 30)*	27 (17, 35)	<0.001
Recipient weight, median (IQR)	11.9 (7.2, 26)	12 (7.2, 30.6)	12.4 (7.9, 27.6)	<0.001
Weight categories				<0.001
<10 kg	44.2%	43.6%	40.3%
10–35 kg	37.5%	34.4%	40.0%
>35 kg	18.3%	22.0%	19.7%
Graft type				<0.001
Whole deceased-donor LT	82.1%	58.5%	59.0%
Split deceased-donor LT	11.0%	24.6%	28.4%
Living-donor LT	6.9%	16.9%	12.6%
Donor age, median (IQR)	6 (1, 19)	14 (3, 27)	13 (2, 23)	<0.001
Donor cause of death				<0.001
Anoxia	16.8%	16.6%	31.7%
Stroke	15.0%	12.9%	8.3%
Head trauma	38.8%	49.3%	43.7%
Other/unknown	29.3%	21.2%	16.3%
Donation after cardiac death	0.1%	0.4%	0.6%	<0.001
PHS infectious risk	0.0%	0.9%	6.8%	<0.001
Cold ischemia time, med (IQR)	9 (5, 12)	6 (3, 9)	6 (5, 8)	<0.001
Sharing				<0.001
Local	39.3%	45.7%	34.2%
Regional	28.7%	37.5%	48.1%
National	32.1%	16.8%	17.7%

AHN = acute hepatic necrosis; IQR = interquartile range; LT = liver transplant.

^*As available for the 49.6% of people transplanted after implementation of PELD/MELD (ie, March 1, 2002) in the 1997 to 2006 era.

Patient and Graft Survival

There were 1497 (1987–1996 cohort), 1102 (1997–2006 cohort), and 454 (2007–2018 cohort) deaths observed. In each respective era, the 5-year patient survival was 73.7%, 83.2%, and 91.2% and the 10-year survival was 70.3%, 79.7%, and 88.3% (log rank P < 0.001) (Fig. 1A). Similarly, in each respective era, the 5-year graft survival was 64.0%, 75.2%, and 85.8% and the 10-year graft survival was 59.4%, 70.4%, and 81.6% (log rank P < 0.001) (Fig. 1B).

FIGURE 1.

Observed and projected patient (A) and graft (B) survival following pediatric LT, stratified by era of transplant: 1987 to 1996 (n = 3787), 1997 to 2006 (n = 4467), and 2007 to 2018 (n = 5188). CI = confidence interval; LT = liver transplant. Projected survival based on parametric model with Weibull distribution and the observed data between the two dashed lines (2–10 years post-transplant).

Model Selection

Observed 20-year patient survival was 72.8% and 63.6% in the 1997 to 2006 and 1987 to 1996 cohorts; observed 30-year patient survival was 57.5% in the 1987 to 1996 cohort. Based on the GG, lognormal, and Weibull models, the absolute difference between observed-to-expected 20-year patient survival was 2.3%, 2.8%, and 1.0% in the 1997 to 2006 cohort and 2.2%, 2.9%, and 1.0% in the 1987 to 1996 cohort, respectively (Table, Supplemental Digital Content 1, http://links.lww.com/MPG/B754). The absolute difference between observed-to-expected 30-year survival was 5.2% (GG), 6.7% (lognormal), and 2.5% (Weibull) in the 1987 to 1996 cohort. The parametric model with Weibull distribution based on 8 years of left-truncated post-transplant data was identified as most closely approximating observed long-term outcomes in the 2 historic cohorts.

Projected Patient Survival

In the 2007 to 2018 cohort, projected 10-year patient survival (88.5%) overestimated observed 10-year survival (88.3%) by 0.3% (95% CI 0.0%−0.8%) (Table 2). Projected 20-year survival for pediatric LT recipients transplanted in 2007 to 2018 was 84.0% (95% CI 81.5–85.8), compared to observed 20-year survival of 72.8% and 63.6% among those transplanted in 1997 to 2006 and 1987 to 1996. Projected 20-year survival for the 2007 to 2018 cohort should be considered alongside a potential overestimation of 1.0% given that observed in the 2 historic cohorts. Projected 30-year survival for pediatric LT recipients transplanted in 2007 to 2018 was 80.1% (95% CI 75.2–82.7), compared to projected 30-year survival of 68.6% (66.1–70.9) in the 1997 to 2006 cohort and observed 30-year survival of 57.5% in the 1987 to 1996 cohort. Projected 30-year outcomes should be considered alongside a potential 2.5% (95% CI 0.1–5.5) overestimation given that observed in the 1987 to 1996 cohort.

TABLE 2.

Projected 10-, 20-, and 30-year patient survival and graft survival among pediatric liver transplant recipients transplanted in each era

Patient survival
	10-Year outcomes			20-Year outcomes			30-Year outcomes
Era	Observed	Projected	Absolute difference	Observed	Projected	Absolute difference	Observed	Projected	Absolute difference
2007–2018	88.3%	88.5% (87.8%–89.3%)	0.3% (0.0%–0.8%)	–	84.0% (81.5%–85.8%)	–	–	80.1% (75.2%–82.7%)	–
1997–2006	79.7%	79.7% (79.1%–80.4%)	0.0% (0.0%–0.1%)	72.8%	73.6% (72.0%–75.1%)	1.0% (0.0%–2.5%)	–	68.6% (66.1%–70.9%)	–
1987–1996	70.3%	70.3% (69.4%–71.1%)	0.0% (0.0%–0.2%)	63.6%	64.4% (62.5%–66.1%)	1.0% (0.1%–2.5%)	57.5%	59.6% (56.6%–62.2%)	2.5% (0.1%–5.5%)
Graft survival
	10-Year outcomes			20-Year outcomes			30-Year outcomes
Era	Observed	Projected	Absolute difference	Observed	Projected	Absolute difference	Observed	Projected	Absolute difference
2007–2018	81.5%	81.8% (80.6%–82.7%)	0.4% (0.0%–1.0%)	–	75.1% (71.9%–77.3%)	–	–	69.1% (64.2%–73.1%)	–
1997–2006	70.3%	70.3% (69.4%–71.0%)	0.0% (0.0%–0.1%)	62.2%	62.7% (61.1%–64.3%)	0.8% (0.0%–2.5%)	–	56.7% (54.4%–59.0%)	–
1987–1996	59.5%	59.5% (58.4%–60.2%)	0.0% (0.0%–0.1%)	52.0%	51.7% (49.6%–53.2%)	0.7% (0.0%–3.1%)	46.4%	45.5% (42.3%–47.8%)	1.3% (0.2%–5.8%)

Projected survival based on crude parametric survival model with Weibull distribution.

Projected Graft Survival

Projected 20-year graft survival for pediatric LT recipients transplanted in 2007 to 2018 was 75.1% (95% CI 71.9–77.3), compared to observed 20-year graft survival of 62.2% and 52.0% in the 1997 to 2006 and 1987 to 1996 cohorts (Table 2). Projected 20-year graft survival should be considered alongside a potential 0.7% to 0.8% overestimation. Projected 30-year graft survival for those transplanted in 2007 to 2018 was 69.1% (95% CI 64.2–73.1), compared to projected 30-year graft survival of 56.7% (95% CI 54.4– 59.0) in the 1997 to 2006 cohort and observed 30-year graft survival of 45.5% in the 1987 to 1996 cohort. Projected 30-year outcomes should be considered alongside a potential 1.3% (95% CI 0.2–5.8) overestimation.

Projections by Subgroup

Across subgroups, the 20-year absolute difference ranged from 1.1% to 13.1% and 0.8% to 23.1% in the 1997 to 2006 and 1987 to 1996 cohorts, respectively, with smaller subgroups showing greater variability (Table, Supplemental Digital Content 2, http://links.lww.com/MPG/B754). The 30-year absolute difference ranged from 1.7% to 23.6% in the 1987 to 1996 cohort. The variability and sample size of each subgroup should be considered alongside long-term projections for the 2007 to 2018 cohort. Within the 2007 to 2018 cohort, there was a statistically significant interaction between age and graft type in the risk of post-transplant mortality (P = 0.001). Among pediatric whole LT recipients, those 1 to 5 and 6 to 10 years old were at 34% and 70% reduced risk of mortality when compared to those younger than 1 year old (Figure 2A, Table, Supplemental Digital Content 3, http://links.lww.com/MPG/B754). Among pediatric split LT recipients, there were no statistically significant differences across age (P ≥ 0.1) (Figure 2B and 2C). Among pediatric living-donor LT recipients, there was a trend toward increased risk of mortality among older recipients versus those younger than 1 year old; however, this was not statistically significant (P ≥ 0.1). Non-BA/nonmetabolic disease was associated with a 2.2-fold higher risk of mortality versus BA/metabolic disease. Based on this model, Supplemental Digital Content Table 4a and 4b (http://links.lww.com/MPG/B754) shows projected 20- and 30-year patient and graft survival given recipient age, diagnosis, and graft type. For example, recipients younger than 1 year with BA/metabolic disease have projected 30-year patient survival of 86.3%, 83.5%, or 92.9% depending on graft type (whole, split, or living donor), and these projections may be subject to 0.6% to 1.5% over- or underestimation given the variability observed at 10 years.

FIGURE 2.

Observed (solid lines) and projected (dashed lines) for patient survival by pediatric recipient age, diagnosis, and graft type for whole deceased donor liver transplant (A), split deceased donor liver transplant (B), and living donor liver transplant (C). CI = confidence interval; DDLT = deceased donor liver transplant; LT = liver transplant.

DISCUSSION

In this nationally representative study of 13,442 pediatric LT recipients transplanted between 1987 and 2018, projected 20-year survival for those transplanted in 2007 to 2018 was 84.0% (95% CI 81.5–85.8), compared to observed 20-year survival of 72.8% and 63.6% among those transplanted in 1997 to 2006 and 1987 to 1996. Projected 30-year survival for pediatric LT recipients in 2007 to 2018 was 80.1% (75.2–82.7), compared to projected 30-year survival of 68.6% (66.1–70.9) in the 1997 to 2006 cohort and observed 30-year survival of 57.5% in the 1987 to 1996 cohort. Twenty- and 30-year patient and graft survival varied slightly by recipient age, graft type, and indication for transplant.

Improvements in projected long-term survival for pediatric recipients can be traced to reductions in mortality and graft failure in the immediate post-transplant period. Several studies have shown that broader implementation of newer surgical techniques, such as the approach to anastomosis of the vena cava and hepatic artery, yields better outcomes (4). Improved understanding of risk factors for vascular thromboses and better postoperative care management (eg, anticoagulation) have also correlated with improved graft survival (14–17). Changes in immunosuppression over recent decades, predominantly from cyclosporine to tacrolimus, have also yielded improvements in patient and graft survival secondary to fewer episodes of rejection and better prevention of side effects including post-transplant lymphoproliferative disorder (6,18). Finally, preoperative care management, such as nutritional optimization and control of complications of ESLD, likely yields better outcomes in the immediate post-transplant period (19).

Prior studies suggest that recipient age, graft type, and diagnosis are associated with short- and mid-term patient and graft survival, and we observed similar variations in long-term outcomes (1,20–23). Our finding that LT recipients with BA and metabolic disorders have better outcomes than individuals with malignancy or acute hepatic necrosis is consistent with reports showing variations in short-term outcomes between these indications (20,21,24). Our long-term projections for patients with metabolic disease may also help families balance the risks of pursuing medical management when feasible versus the option of LT. We also identified an interaction between age and graft type in post-transplant outcomes. Among whole LT recipients, better survival was projected for individuals between 1 and 10 years, compared to recipients younger than 1 year and older than 10 years, whereas outcomes for split and living-donor LT recipients did not vary by age group. This finding likely represents careful candidate selection for individuals receiving technical-variant grafts. We used this model to individualize projections and inform what patients with specific characteristics may expect with regard to long-term outcomes; however, each projection should be considered alongside potential variability driven by sample size within each subgroup.

This analysis makes novel use of parametric modeling and applies it toward a robust surgical registry to project long-term outcomes beyond the observed data. Parametric models have been used to estimate survival among adult and pediatric liver transplant populations; however, to the best of our knowledge parametric models have not been used to project long-term survival among recently transplanted patients. In other patient populations, there have been applications of parametric models with varied methods to project outcomes beyond observed data (25–27). Li et al (28) used flexible parametric models to predict median and mean survival for 12,307 adult kidney transplant recipients given current practices and particular patient profiles in the United Kingdom. Cabarrou et al (29) used parametric competing risks regression to estimate 20- and 25-year outcomes among 4761 patients with breast cancer based on 10 years of data in France. Keogh et al (30) applied flexible parametric models to data on 10,428 patients in the UK cystic fibrosis registry to project survivorship and age at the time of death beyond observed data. Jackson et al (31) fit parametric models with Weibull distribution to a birth cohort of 13,115 patients in Ireland and projected median cystic fibrosis survivorship beyond the observed data. Each varied application, including that of the present study, has the potential to provide relevant approximations of long-term outcomes and illustrate improvements in survival not yet represented by observed data. All should be considered in conjunction with expected variability that is often determined by sample size.

One important limitation is that models can never thoroughly describe observed data. Our projections tended to overestimate patient and graft survival by approximately 0.7% to 2.5%. Although we consider this bias to be fundamentally small, outcomes may change if future innovations affecting long-term outcomes are implemented. For example, better understanding of machine perfusion may allow for improved recovery of marginal organs, whereas better understanding of immunosuppression management and promotion of tolerance may minimize rejection (32,33). Such innovations will likely improve long-term outcomes for pediatric LT recipients, in which case we may have underestimated projected survival. Another limitation is the small sample size for specific subgroups leading to 0.3% to 11.9% overestimation in 10-year projected patient survival for the 2007 to 2018 cohort. Survival projections across subgroups should be interpreted cautiously, and the range of potential over- or underestimation should be provided when communicating with families. In addition, we developed a parsimonious model based on recipient age, graft type, and indication as these are well-established contributors to outcomes in pediatric LT, but other variables, such as weight, PELD/MELD score, donor age, cause of death, donation after cardiac death, or cold ischemia time likely contribute to outcomes as well. Finally, we used a standard parametric model with Weibull distribution, whereas flexible parametric models may have more closely approximated unobserved outcomes. Despite these limitations, we validated our approach within 2 historic cohorts and found the overall potential variability to be low and directly related to sample size.

In conclusion, there have been substantial improvements in pediatric LT outcomes over the last 30 years, such that observed long-term outcomes are no longer relevant. Projected long-term outcomes for recently transplanted pediatric recipients are excellent and can better inform shared decision making between providers, patients, and families regarding expectations of pediatric LT in the current era.

What Is Known

• Studies of long-term outcomes (ie, >10 years) in pediatric liver transplantation have been reported from single centers based on their historical data, and range from 53% to 69% patient survival.

• Improvement in short-term outcomes following pediatric liver transplantation has been reported, likely due to improved surgical technique and immunosuppression.

What Is New

• Applying parametric models to a nationally representative registry, we demonstrate significant improvement in long-term (ie, 20- and 30-year) patient and graft survival compared to historical data.

• Projected long-term outcomes following split liver transplantation are similar to outcomes for whole liver transplantation.

• These data may better inform clinical practice and patient education for newly listed and transplanted children.