Survival Outcomes Following Pediatric Liver Transplantation (Pedi‐SOFT) Score: A Novel Predictive Index

Abstract

A prognostic index to predict survival after liver transplantation could address several clinical needs. Here, we devised a scoring system that predicts recipient survival after pediatric liver transplantation. We used univariate and multivariate analysis on 4565 pediatric liver transplant recipients data and identified independent recipient and donor risk factors for posttransplant mortality at 3 months. Multiple imputation was used to account for missing variables. We identified five factors as significant predictors of recipient mortality after pediatric liver transplantation: two previous transplants (OR 5.88, CI 2.88–12.01), one previous transplant (OR 2.54, CI 1.75–3.68), life support (OR 3.68, CI 2.39–5.67), renal insufficiency (OR 2.66, CI 1.84–3.84), recipient weight under 6 kilograms (OR 1.67, CI 1.12–2.36) and cadaveric technical variant allograft (OR 1.38, CI 1.03–1.83). The Survival Outcomes Following Pediatric Liver Transplant score assigns weighted risk points to each of these factors in a scoring system to predict 3‐month recipient survival after liver transplantation with a C‐statistic of 0.74. Although quite accurate when compared with other posttransplant survival models, we would not advocate individual clinical application of the index.

Abbreviations

BAR score

Balance of Risk Score

CI

confidence interval

DRI

Donor Risk Index

GFR

glomerular filtration rate

HCC

hepatocellular carcinoma

HR

hazard ratio

ICU

intensive care unit

MELD score

Model for End‐Stage Liver Disease score

OPTN

Organ Procurement and Transplantation Network

OR

odds ratio

PELD

Pediatric End‐stage Liver Disease

ROC curve

receiver operating characteristic curve

SOFT score

Survival Outcomes following Liver Transplant score

SPLIT database

Studies of Pediatric Liver Transplantation database

SRTR

Scientific Registry of Transplant Recipients

UNOS

United Network for Organ Sharing

Introduction

There is a pressing need for an index that accurately predicts posttransplant survival in liver transplantation. The model would be instrumental to maintain appropriate center outcomes and could risk stratify our recipients for more intensely monitored postoperative care. An accurate model can also help to educate recipients and their families about the patient's individualized risk of mortality.

Because the Scientific Registry of Transplant Recipients (SRTR) risk assessment models are not applicable to individual clinical situations and have limited accuracy 1, there have been several attempts to construct applicable predictive indices for survival after liver transplantation in adults. The most notable indices are the SOFT and BAR scores, which report limited accuracy (C‐statistic of 0.7) 2, 3. As a result, their clinical impact has been limited, but they have stimulated discussion and may serve as the foundation for more applicable models. There has not been such a model in pediatric liver transplantation. Several studies have identified major risk factors for mortality after pediatric liver transplantation, but none have put forth a validated index to predict outcomes 4-10.

The primary objective of this analysis was to identify predictors of posttransplant mortality and construct an applicable predictive index. In an era when our pediatric recipients are becoming sicker 9, 10, this analysis intends to achieve the larger goal of risk stratification in pediatric liver transplantation.

Methods

Study population

We performed a retrospective analysis of UNOS de‐identified patient‐level data of all recipients of liver transplant between March 1, 2002 and December 31, 2012. Our analysis employed the liver registry with data collected by the Organ Procurement and Transplantation Network (OPTN). We included all transplant recipients 12 years and younger. Recipients over 12–18 years were not analyzed because their allocation is based on the MELD score rather than the PELD score. Donor and recipient characteristics were reported at the time of transplant. Follow‐up information was collected at 6 months and then yearly after transplantation. Patients undergoing combined or multivisceral transplants (n = 600), with the exception of liver kidney transplant (n = 79), were excluded from the study. Liver kidney transplant recipients were included because renal dysfunction requiring renal transplantation is often a sequela of liver disease in the sickest candidates. Furthermore, we wanted to investigate renal insufficiency as a risk factor. All patients were followed from the date of transplant until either death (n = 468) or the date of last known follow‐up (n = 4097); we analyzed 4565 recipients. This included 2509 recipients of whole allografts and 2056 recipients of technical variant grafts.

Statistical analysis

Data was analyzed using a standard statistical software package, Stata® 9 (Stata Corp, College Station, TX). Continuous variables were reported as a mean ± standard deviation and compared using the Student t‐test. Contingency table analysis was used to compare categorical variables. Results were considered significant at a p‐value of <0.05. All reported p‐values were two‐sided. The primary outcome measure was patient death. Time to death was assessed as time from the date of transplantation to the date of death. Kaplan–Meier analysis with log‐rank test and logistic regression were used for time‐to‐event analysis. Three‐month survival was the dependent variable and the risk factors were the independent variables in the logistic regression analysis. Risk factors that were significant in univariate analysis (p < 0.05) were included in the multivariate analysis. Patients lost to follow‐up (n = 117) or alive (n = 3979) on December 31, 2012 were censored at the date of last known follow‐up.

Risk factors

The recipient and donor risk factors considered in this analysis are listed in Table 1. Creatinine clearance was calculated using the updated Schwartz bedside formula: eGFR = 0.41 X height (cm)/Scr (mg/dL). Recipients with a creatinine clearance <30 and dialysis‐dependent recipients were grouped together because of similar results in separate univariate and multivariate analyses. Height and weight deficits were based on Center for Disease Control growth charts. Standard deviations were calculated based on published z scores. Status 1 included Status1A and Status1B. Threshold analysis determined 6 kg as the cut point for minimal recipient weight.

Table 1. Risk factors considered in univariate analysis

Donor risk factors	Odds ratio	p‐value	Recipient risk factors	Odds ratio	p‐value
Age <1 year	1.34 (0.93–1.94)	0.11	ABO incompatible	1.34 (0.70–2.59)	0.38
Age 1–2 years	1.03 (0.74–1.43)	0.85	Admitted to Hospital	1.22 (0.89–1.68)	0.22
Age 2–3 years	0.56 (0.28–1.15)	0.12	Admitted to ICU	3.22 (2.49–4.18)	0.01
Age 3–4 years	1.12 (0.60–2.10)	0.72	Age <1 year	1.22 (0.94–1.59)	0.13
Age 4–5 years	0.61 (0.25–1.51)	0.29	Age 1–2 years	0.91 (0.68–1.21)	0.51
Age 5–8 years	0.71 (0.41–1.24)	0.23	Age 2–3 years	1.48 (0.92–2.39)	0.10
Age 8–12 years	0.78 (0.46–1.33)	0.36	Age 3–4 years	0.99 (0.52–1.90)	0.98
Age 13–18 years	0.82 (0.56–1.20)	0.31	Age 4–5 years	0.64 (0.28–1.47)	0.29
ALT >500 IU/L	0.48 (0.07–3.50)	0.47	Age 5–8 years	0.96 (0.63–1.47)	0.87
AST >500 IU/L	0.82 (0.20–3.40)	0.79	Age 8–12 years	0.70 (0.45–1.08)	0.10
Bilirubin >2 mg/dL	0.82 (0.44–1.52)	0.53	Albumin <2.0 g/dL	1.02 (0.62–1.66)	0.95
Bilirubin 1–2 mg/dL	0.99 (0.71–1.39)	0.95	Albumin 2.0–2.5 g/dL	1.29 (0.93–1.80)	0.13
Cause of death anoxia	0.75 (0.55–1.02)	0.07	Albumin 2.5–3.0 g/dL	1.23 (0.92–1.64)	0.16
Cause of death cerebral vascular accident	1.33 (0.88–2.01)	0.18	Ascites	1.33 (0.98–1.78)	0.06
Cold ischemia time 0–6 h	0.87 (0.67–1.14)	0.32	Bilirubin <2 mg/dL	0.50 (0.36–0.69)	.01
Cold ischemia time 12–16 h	1.19 (0.66–2.17)	0.56	Bilirubin 8–16 mg/dL	1.12 (0.82–1.53)	0.48
Cold ischemia time 16–20 h	1.57 (0.37–6.73)	0.54	Bilirubin 16–32 mg/dL	1.07 (0.79–1.44)	0.67
Cold ischemia time over 20 h	2.18 (0.85–5.58)	0.10	Bilirubin >32 mg/dL	2.05 (1.39–3.03)	0.01
Deceased donor after cardiac death	0.72 (0.10–5.33)	0.75	Creatinine clearance <10	2.91 (1.12–7.58)	0.03
Diabetes mellitus	1.11 (0.27–4.68)	0.88	Creatinine clearance 10–20	3.25 (1.58–6.70)	0.001
Creatinine clearance <10	3.03 (1.04–8.84)	0.04	Creatinine clearance 20–30	4.71 (2.80–7.94)	0.001
Creatinine clearance 10–20	1.22 (0.53–2.82)	0.64	Creatinine clearance 30–40	2.78 (1.56–4.96)	0.001
Creatinine clearance 20–30	1.43 (0.66–3.13)	0.37	Creatinine clearance 40–50	1.36 (0.68–2.72)	0.38
Creatinine clearance 30–40	1.09 (0.47–2.51)	0.84	Diagnosis‐‐ Hepatoblastoma	0.54 (0.27–1.06)	0.07
Female	0.98 (0.76–1.27)	0.87	Diagnosis—Biliary Atresia	0.64 (0.49–0.85)	0.002
Live donor allograft	0.88 (0.60–1.31)	0.54	Diagnosis—Metabolic disorder	0.71 (0.45–1.12)	0.14
National allocation	1.20 (0.87–1.65)	0.26	Dialysis	2.22 (1.85–2.68)	0.001
Regional allocation	0.93 (0.72–1.21)	0.61	Encephalopathy	1.71 (1.30–2.23)	0.001
Technical variant allograft (cadaveric)	1.52 (1.17–1.97)	0.002	Height Deficit 1–2 SD
Warm ischemia time <30 min	0.82 (0.50–1.35)	0.44	Height Deficit >2 SD
Warm ischemia time 60–75 min	1.38 (0.74–2.59)	0.31	INR 2.0–2.5	1.22 (0.82–1.82)	0.33
Warm ischemia time >75 min	2.21 (1.17–4.19)	0.02	INR >2.5	2.40 (1.82–3.17)	0.001
			Life support	6.10 (4.68–7.97)	0.001
			Previous liver transplantation within 30 day interval	4.91 (3.38–7.14)	0.001
			Previous liver transplantation beyond 30 day interval	1.58 (0.95–2.64)	0.08
			PELD score 20–25	0.91 (0.61–1.34)	0.63
			PELD score 25–30	1.67 (1.16–2.41)	0.006
			PELD score 30–35	1.83 (1.19–2.79)	0.006
			PELD score 35–40	1.89 (1.13–3.18)	0.02
			PELD score 40–45	2.03 (1.01–4.09)	0.05
			PELD score 45–50	3.10 (1.29–7.45)	0.01
			PELD score >50	2.81 (1.32–6.01)	0.007
			Region 1	0.45 (0.16–1.22)	0.12
			Region 2	0.50 (0.31–0.82)	0.006
			Region 3	0.68 (0.44–1.07)	0.09
			Region 4	1.50 (1.05–2.16)	0.03
			Region 5	0.90 (0.64–1.24)	0.51
			Region 6	0.48 (0.15–1.52)	0.21
			Region 7	1.04 (0.67–1.62)	0.86
			Region 8	1.33 (0.87–2.02)	0.19
			Region 9	0.85 (0.51–1.43)	0.54
			Region 10	1.45 (0.98–2.15)	0.07
			Region 11	2.18 (1.41–3.38)	0.001
			Serum sodium <130 mEq/L	1.40 (0.77–2.56)	0.27
			Serum sodium 130–135 mEq/L	1.09 (0.78–1.52)	0.62
			Serum sodium 145–150 mEq/L	1.17 (0.65–2.07)	0.60
			Serum sodium >150 mEq/L	3.39 (1.98–5.78)	0.001
			Two previous transplants	7.38 (3.99–13.6)	0.001
			UNOS Status 1	2.48 (1.89–3.25)	0.001
			Weight <3 kg	3.47 (0.76–15.9)	0.11
			Weight 3–4 kg	2.50 (1.05–5.94)	0.04
			Weight 4–5 kg	1.20 (0.62–2.30)	0.59
			Weight 5–6 kg	1.70 (1.12–2.56)	0.01
			Weight 6–7 kg	1.35 (0.91–1.98)	0.13
			Weight 7–8 kg	0.95 (0.61–1.46)	0.82
			Weight 8–9 kg	0.68 (0.38–1.20)	0.18
			Weight 9–10 kg	0.84 (0.50–1.41)	0.50
			Weight Deficit 1–2 SD
			Wt Deficit >2 SD

• Bold = p < 0.05 in univariate analysis.

Missing variables

Multiple imputation with predicted mean matching was performed for the following incomplete predictors in the OPTN database: serum sodium (24.8%), cold ischemia time (8.9% missing), serum creatinine (8.5%), ascites (6.3%), encephalopathy (6.3%), recipient weight (4.0% missing), diagnosis (0.4%), PELD score (0.3%), recipient height (0.1% missing), albumin (0.1%) and Status 1 (0.1%).

Risk score

Logistic regression analysis determined the predictors of patient death at 3 months posttransplantation. Donor and recipient variables were first analyzed with univariate analysis and are listed in Table 1. Variables found to be significant in univariate analysis were then subjected to multivariate analysis. Points were assigned to each risk factor based on its odds ratio for patient death at 3 months. One point was awarded to each risk factor for every 10% increase in risk for death at 3 months.

We assigned three risk groups based on point totals. Model discrimination was assessed using the area under the receiver operating characteristic curve (ROC).

Results

Study population

The study population included 4565 patients. Analysis included 17 214 years‐at‐risk for the liver transplant recipients. Mean follow‐up was 3.8 years. Demographic and clinical characteristics are summarized in Table 2.

Table 2. Demographic characteristics of donors and recipients

Recipient characteristics	All recipients	Recipients of whole donors	Recipients of cadaveric technical variant allografts	Recipients of living donors
n	4565	2509 (55%)	1449 (32%)	607 (13%)
Age (years)	2.9 ± 3.6	3.5 ± 3.9	2.3 ± 3.0	2.0 ± 3.1
% Female	48.9%	48.0%	49.8%	49.9%
% African American	15.7%	16.9%	15.7%	10.7%
Height (cm)	84.6 ± 28.1	88.6 ± 29.8	80.4 ± 24.8	78.3 ± 25.5
Weight (Kg)	13.8 ± 11.1	15.6 ± 12.6	11.9 ± 8.5	11.3 ± 9.1
INR	2.0 ± 2.6	1.8 ± 2.3	2.1 ± 2.0	2.2 ± 4.5
Creatinine (mg/dL)	0.4 ± 0.7	0.5 ± 0.7	0.4 ± 0.6	0.3 ± 0.4
Lab PELD	14.1 ± 14.5	12.2 ± 14.0	16.2 ± 14.9	17.3 ± 14.2
Cause of liver failure
Biliary atresia	37.1%	35.1%	36.7%	46.3%
Acute liver failure	19.5%	16.3%	25.3%	18.3%
Metabolic liver disease	11.3%	13.2%	10.1%	6.3%
Hepatoblastoma	6.3%	6.6%	6.5%	4.6%

Donor characteristics	All donors	Whole cadaveric donors	Cadaveric technical variant allografts	Living donors
Age (years)	13.1 ± 13.2	5.1 ± 7.9	18.9 ± 11.2	31.8 ± 8.2
% Female	44.1%	44.4%	37.9%	57.7%
% African American	17.6%	20.8%	15.0%	10.4%
Height (cm)	127.6 ± 41.8	100.8 ± 32.9	159.5 ± 24.8	168.7 ± 10.4
Weight (Kg)	39.1 ± 28.0	20.7 ± 17.1	58.4 ± 21.6	72.2 ± 15.4
Bilirubin (mg/dL)	0.8 ± 1.6	0.7 ± 0.8	1.02 ± 2.4	NA
Creatinine (mg/dL)	0.7 ± 0.8	0.6 ± 0.7	1.0 ± 0.9	NA
Cold ischemia time (Hours)	7.0 ± 4.1	7.6 ± 3.7	7.3 ± 3.1	NA
Cause of death
Trauma	52.3%	49.9%	63.8%	NA
Anoxia	26.1%	37.1%	18.0%	NA

• BOLD = p < 0.05 compared to other groups.
• NA, not applicable to this group of patients; CVA, cerebrovascular accident.

Data entry rate

Most variables were well populated. Multiple imputation with predicted mean values was performed for missing variables.

Univariate and multivariate analysis

Table 1 lists all the risk factors that were considered. Risk factors that were significant in univariate analysis were then subjected to multivariate analysis. The risk factors that were significant in multivariate analysis are presented in Table 3. The most significant risk factors were two previous transplants (OR 5.9, CI 2.9–12.0), life support (OR 3.7, CI 2.4–5.7), dialysis or creatinine clearance <30 (OR 2.7, CI 1.8–3.8) and one previous transplant (OR 2.5, CI 1.8–3.7).

Table 3. Multivariate analysis

	Odds ratio	p–value	C.I.	Risk points
Donor risk factors
Cadaveric technical variant allograft	1.38	0.03	1.03–1.83	4
Recipient risk factors
Two previous transplants	5.88	> 0.01	2.88–12.01	49
One previous transplant	2.54	>0.01	1.75–3.68	15
Life support	3.68	> 0.01	2.39–5.67	27
Dialysis or creatinine clearance <30	2.66	> 0.01	1.84–3.84	17
Region 11	2.28	0.001	1.40–3.73
Region 4	1.67	0.01	1.11–2.36
Recipient weight <6 kg	1.62	0.01	1.12–2.36	6
Region 2	0.57	0.04	0.34–0.97
PELD score >50	1.67	0.31	0.63–4.38
PELD score 45–50	1.72	0.08	0.59–5.05
PELD 40–45	1.61	0.30	0.66–4.38
PELD 35–40	1.32	0.43	0.66–2.67
PELD 30–35	1.22	0.50	0.68–2.19
PELD 25–30	1.29	0.28	0.82–2.04
Serum sodium >150	1.71	0.08	1.95–3.08
INR >2.5	1.31	0.27	0.81–2.10
Warm ischemia time >75 min	1.29	0.47	0.65–2.67
Bilirubin >32 mg/ dL	1.28	0.65	0.76–1.54
Encephalopathy	1.09	0.50	0.82–1.52
Ascites	0.97	0.87	0.67–1.40
Bilirubin <2 mg/dL	0.88	0.52	0.60–1.29
Biliary atresia	0.83	0.28	0.60–1.16
ICU admission	0.78	0.27	0.50–1.08
Region 3	0.74	0.22	0.45–1.20
Status 1	0.71	0.11	0.47–1.08

Risk score

Table 3 summarizes donor and recipient risk factors and their assigned points. Table 4 presents the Pedi‐SOFT risk score with risk groups. Regions were not incorporated into the Pedi‐SOFT score because they did not increase the accuracy of the score (C‐statistic 0.74, CI 0.71–0.78 vs. C‐statistic 0.77, CI 0.74–0.80). Early retransplants (within 30 days of the primary transplant) and late retransplants (after 30 days) had statistically similar odds ratios (OR 2.65, CI 1.62–4.33 and OR 2.35, CI 1.35–4.08) in multivariate analysis (Table 5). Since there was no added score accuracy in dividing retransplantation by interval, this was not incorporated into the Pedi‐SOFT score. Population distribution and odds ratios are reported for each risk group. Figure 1 illustrates the Kaplan–Meier curves and life table analysis of 6‐month and 5‐year patient survival post–liver transplantation based on risk groups from the Pedi‐SOFT score. Using the Pedi‐SOFT score, the 3‐month patient survival of recipients with <25 points was 97%, 25–50 points was 83% and >50 points was 50%. The area under the receiver operating curve (C–statistic) for 3‐month survival using the Pedi‐SOFT score is 0.74 (0.71–0.78) (Figure 2A). The 3 month mortality as a function of the Pedi‐SOFT score is presented in Figure 2B. Table 5 presents actual clinical examples using the Pedi‐SOFT score.

Table 4. Risk index

Pedi SOFT score
Risk factor	Risk points	% of patients
Cadaveric technical variant graft	4	31.7
Recipient weight under 6 kg	6	11.9
Dialysis or creatinine clearance under 30	17	7.2
Life support	27	13.2
One previous liver transplant	15	9.5
Two previous liver transplants	49	1.1

Pedi SOFT risk groups
	Odds ratio (CI)	% of patients
<25 points	Ref	85.8
25–50 points	5.7 (4.2–7.6)	11.8
>50 points	14.9 (9.7–23.0)	2.4

Table 5. Specific combinations of risk factors and corresponding Pedi‐SOFT scores in actual patients from the database

Patient characteristics	Example 1	Example 2	Example 3	Example 4	Example 5
Age	6 years	7 years	9 months	8 months	1 year
Weight	31.5 kg	17.6 kg	4.9 kg	5.4 kg	5.9 kg
Graft type	Whole cadaveric	Live donor	Cadaveric technical variant graft	Cadaveric technical variant graft	Cadaveric technical variant graft
Dialysis or creatinine clearance <30	No	Yes	No	Yes	No
Life support	No	No	No	No	Yes
Number of previous transplants	0	0	1	0	1
Pedi‐SOFT score	0	17	25	27	52
Risk group	Low	Low	Moderate	Moderate	High

Figure 1

A and B: Kaplan–Meier curve of recipient survival by the Pedi‐SOFT score. Abscissa: Percentage recipient survival of total recipients. Ordinate: Months post–liver transplant. p‐value < 0.001 for each group by log rank test with reference to the low risk group.

Figure 2

A: The area under the Receiver Operating Curve when Pedi‐SOFT is used to predict 3‐month survival. B: 3‐month survival as a function of the Pedi‐SOFT score. Abscissa: 3‐month mortality. Ordinate: Pedi‐SOFT Score.

Infants and children

The Pedi‐SOFT score had statistically similar C‐statistics for 3‐month survival when applied to distinct age groups: <1 year = 0.70 (0.64–0.76) and 1–12 years = 0.77 (0.72–0.81) (Figure 3).

Figure 3

Kaplan–Meier curve of recipient survival by the Pedi‐SOFT score by age categories: Infants (>1 year) and children (1–12 years). Abscissa: Percentage recipient survival of total recipients. Ordinate: Months post–liver transplant. p‐value < 0.001 for each group by log rank test with reference to the low risk group.

PELD score

The PELD score is a poor predictor of posttransplant survival as demonstrated by the Kaplan–Meier survival curves in Figure 4. This is expected since the PELD score was modeled for waitlist mortality. The C‐statistic for 3‐month survival using the PELD score at transplant is 0.60 (0.55–0.63).

Figure 4

Kaplan–Meier curve of recipient survival by the lab PELD score at transplant for patients 0–12 years. Abscissa: Mortality percentage at 3 months. Ordinate: Score ranges for the PELD score. p‐value < 0.001 for each group by log rank test with reference to PELD <10 except PELD 10–20.

Discussion

There is much need for an accurate model to predict survival after pediatric liver transplantation. In an era when pediatric recipients are growing sicker 9, 10, an effective posttransplant survival model will further our understanding of the potential risk of mortality. This can be instrumental for programmatic outcomes and recipient expectations, and can also help tailor the intensity of our postoperative monitoring. The Pedi‐SOFT score predicts survival with a C‐statistic of 0.74. Although it is better than other posttransplant predictive indices 2, 3, 11, 12, we would assert that a C‐statistic of 0.80 would be necessary for wide scale clinical implementation. A germane example is the MELD score with a reported C‐statistic of 0.80 that has transformed liver allocation and our concepts of waitlist mortality in adults 13, 14. By incorporating center specific data not captured in the OPTN database, the Pedi‐SOFT score could serve as a foundation for a more accurate model. It can also be used to prognosticate large populations and risk stratify individual patients for closer postoperative monitoring with a longer ICU stay, for example. We have demonstrated that the Pedi‐SOFT score is much more accurate than the PELD score for forecasting survival at transplantation (Figures 1 and 4). We further demonstrated that the Pedi‐SOFT score can be applied to infants (<1 year) and children (1–12 years) (Figure 3).

The risk factors in the Pedi‐SOFT score are not new. In the literature, previous liver transplant 15 and life support 9, 10 are well‐known prominent risk factors for mortality. Recipient hemodialysis has been cited in several studies 9, 10. Very small recipients and technical variant grafts have also been cited in numerous studies as prominent risk factors 7, 10, 16-18; however, other single center experiences have contested this conclusion 18-21. What is new in this study is the constellation of these risk factors into an index to prognosticate survival in pediatric liver transplantation. Aside from the SRTR risk model, which cannot be applied to isolated patients and has limited accuracy 1, there has been no other proposed index for posttransplant survival in pediatric liver transplantation.

Some of the risk factors that have been cited in the literature were not found to be significant in our multivariate analysis. An analysis from the SPLIT database reported fulminant liver failure as one of the strongest risk factors for death 6. We identified fulminant liver failure by both the UNOS Status 1 label and the diagnosis of acute hepatic failure, but failed to find significance in our univariate analysis. This was also the case for height deficit of two standard deviations, a prominent risk factor in the split analysis that did not reach significance in our univariate analysis 6. A final risk factor cited in the SPLIT analysis was continuous pretransplant hospitalization 6. We found neither pretransplant hospital admission nor ICU admission to be significant in univariate analysis. Fulminant liver failure, height deficit and hospital admission were not significant in our analysis even if we forced them into our multivariate analysis. This may reflect differences between the SPLIT and OPTN databases or in the number of patients in the study. Our study had over five times the number of patients in the SPLIT analysis 6.

This analysis, which is the largest available in pediatric liver transplantation, has identified previous transplant, life support, renal insufficiency, recipient weight under 6 kilograms and cadaveric technical variant grafts as significant risk factors for short‐term survival after pediatric liver transplantation. The Pedi‐SOFT score assigns weighted risk points to these five risk factors to predict survival with a C‐statistic of 0.74. Although quite accurate when compared with other posttransplant survival models, it is not sufficiently accurate for wide‐scale and direct clinical application.

Limitations

Since the passage of the National Transplantation Act of 1984, data entry has been mandatory for all US transplant centers. Nevertheless, all patient registries often suffer from variability in data entry. We attempted to account for missing data with multiple imputation analysis. We conducted the analysis without imputation and did not find significant differences. The fact that center specific factors was not accounted for is another significant limitation. Finally, life support was found to be a very strong risk factor. This is a general entry that is subject to variable interpretation.

Acknowledgments

The authors would like to thank Ana María Rodríguez, PhD, a member of the Baylor College of Medicine Michael E. DeBakey Department of Surgery Research Core, for her editorial assistance during the preparation of this manuscript. This study was funded by the Cade R. Alpard Foundation.

Disclosure

The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.