Predictors of failure of fish-oil therapy for intestinal failure–associated liver disease in children^,

Abstract

Background: Parenteral fish-oil (FO) therapy is a safe and effective treatment for intestinal failure–associated liver disease (IFALD). Patients whose cholestasis does not resolve with FO may progress to end-stage liver disease.

Objective: We sought to identify factors associated with the failure of FO therapy in treating IFALD to guide prognostication and referral guidelines.

Design: Prospectively collected data for patients treated with FO at Boston Children’s Hospital from 2004 to 2014 were retrospectively reviewed. Resolution of cholestasis was defined as sustained direct bilirubin (DB) <2 mg/dL, and treatment failure as liver transplantation or death while DB was >2 mg/dL as of July 2015. Demographics, laboratory values, and medical history at FO therapy initiation were compared between patients who achieved resolution of cholestasis and those who failed therapy.

Results: Among 182 patients treated with FO, 86% achieved resolution of cholestasis and 14% failed therapy. Patients who failed therapy had median (IQR) lower birth weight [1020 g (737, 1776 g) compared with 1608 g (815, 2438 g); P = 0.03] and were older at FO initiation [20.4 wk (9.9, 38.6 wk) compared with 11.7 wk (7.3, 21.4 wk); P = 0.02] than patients whose cholestasis resolved. Patients who failed therapy had more advanced liver disease at therapy initiation than patients whose cholestasis resolved, as evidenced by lower median (IQR) γ-glutamyltransferase [54 U/L (41, 103 U/L) compared with 112 U/L (76, 168 U/L); P < 0.001], higher DB [10.4 mg/dL (7.5, 14.1 mg/dL) compared with 4.4 mg/dL (3.1, 6.6 mg/dL); P < 0.001], and a higher pediatric end-stage liver disease (PELD) score [22 (14, 25) compared with 12 (7, 15); P < 0.001]. A PELD score of ≥15, history of gastrointestinal bleeding, age at FO initiation ≥16 wk, presence of nongastrointestinal comorbidities, and mechanical ventilation at FO initiation were independent predictors of treatment failure.

Conclusions: Most infants with IFALD responded to FO therapy with resolution of cholestasis, and liver transplantation was rarely required. Early FO initiation once biochemical cholestasis is detected in parenteral nutrition–dependent patients is recommended. This trial was registered at clinicaltrials.gov as NCT00910104.

INTRODUCTION

Parenteral nutrition (PN)⁷ is a lifesaving therapy for children with intestinal failure caused by insufficient bowel length or function (1, 2). PN affords infants who would otherwise die of starvation or malnutrition a chance to grow and develop. With time, intestinal adaptation may occur, allowing children to absorb nutrients from an enteral diet such that PN can be discontinued (3). Long-term PN treatment is limited by potentially serious complications, including liver disease. Up to 75% of infants who require PN for 60 d develop intestinal failure–associated liver disease (IFALD) (4). In the early 2000s, infants with IFALD followed for ≥2 y had a 26% transplant and 27% mortality rate (4).

IFALD is characterized initially by intrahepatic cholestasis, which can occur within 2 wk of initiating PN, and then by progressive fibrosis and ultimately cirrhosis over the course of several months (5, 6). Without transitioning to full enteral feedings, cirrhosis may progress to end-stage liver disease, with death from sepsis or portal hypertension (4, 7, 8).

Parenteral fish-oil (FO) therapy has been shown to be a safe and effective treatment for IFALD, with successful resolution of biochemical cholestasis and avoidance of liver transplantation in 85% of infants (9–13). However, PN-dependent patients in whom cholestasis is not resolved with FO therapy may progress to end-stage liver disease, requiring liver transplantation or resulting in death. To date, there are no methods to our knowledge for predicting which infants will fail to respond to FO therapy. Thus, we aimed herein to identify early patient factors associated with subsequent failure of FO therapy to guide prognostication and patient referral guidelines.

METHODS

Patients

We retrospectively reviewed prospectively collected data from children with IFALD treated with parenteral FO (Omegaven; Fresenius Kabi) at Boston Children’s Hospital (BCH) between 1 January 2004 and 31 December 2014. The study protocol was approved by our institutional review board. IFALD was defined as serum DB >2 mg/dL for 2 consecutive weeks with no other causes of liver disease. FO monotherapy was administered intravenously at a dose of 1 g · kg⁻¹ · d⁻¹ as previously described in accordance with the study protocol at BCH for the compassionate use of FO through the Food and Drug Administration (9). Patients whose cholestasis did not resolve with FO therapy were identified and defined as treatment failures. Resolution of cholestasis was defined as sustained DB <2 mg/dL. Treatment failure was defined as undergoing liver or multivisceral transplantation or death while the DB was >2 mg/dL as of 31 July 2015. Patients who initiated FO therapy at another institution were excluded from analysis because of the inconsistent availability of baseline laboratory values and medical history. Patients currently under active therapy for cholestasis were also excluded (defined as DB >2 mg/dL as of 31 July 2015).

Study endpoints

The primary aim of this study (NCT00910104; clinicaltrials.gov) was to determine predictors of treatment failure of parenteral FO in children treated for IFALD from 2004 to 2014. Patient demographics, anthropometrics, hospital transfer status, laboratory values, and medical history at initiation of FO therapy were compared between patients who achieved resolution of cholestasis with FO therapy and those who failed therapy.

Baseline weight, length, and head circumference measurements were used to calculate age- and sex-adjusted z scores for 175 subjects, with growth measurements obtained [weight-for-age z score, length-for-age z score (LAZ), and head circumference-for-age z score] with the use of methods appropriate to baseline age: Olsen et al. (14) if <40 wk corrected gestational age (n = 71), WHO if aged <2 y (n = 92), or CDC (15) if aged >2 y (n = 12). To determine the presence of cholestasis, severity of hepatic injury, and status of hepatic synthetic function, the following laboratory values were recorded from the medical record: total bilirubin, DB, γ-glutamyltransferase, aspartate aminotransferase, alanine aminotransferase, international normalized ratio (INR), albumin, and platelet count. From these data, the pediatric end-stage liver disease (PELD) score was calculated as PELD = 4.8[ln(total bilirubin)] + 18.57[ln(INR)] – 6.87[ln(albumin)] + 4.36 (if aged <1 y) + 6.67 (if growth failure), where growth failure was defined as a length or weight z score <−2 per United Network for Organ Sharing guidelines. The PELD score is used to measure illness severity in the allocation of livers to pediatric candidates (≤11 y) and is a validated predictor of mortality from end-stage liver disease; however, it cannot distinguish irreversible from reversible liver disease (16–18). Two patients aged >11 y at FO initiation were excluded from the analysis of independent predictors of treatment failure.

Charts were reviewed for medical history and severity of illness at the time of FO initiation. Medical and surgical diagnoses were recorded. Patent ductus arteriosus was recorded if it required medical or surgical treatment. Lung disease of prematurity was recorded if a diagnosis of respiratory distress syndrome, chronic lung disease, bronchopulmonary dysplasia, or pulmonary hemorrhage was made in a patient who was born prematurely and required a surfactant, intubation, or prolonged intubation. Intraventricular hemorrhage was recorded for grade ≥3 (19). Nongastrointestinal comorbidities were categorized as pulmonary, cardiovascular, neurologic, genetic or chromosomal, hematologic or oncologic, endocrinologic, renal or genitourinary, and traumatic (data not shown). History of gastrointestinal bleeding at any time before FO therapy initiation was recorded. Mechanical ventilation, vasopressor treatment, and sepsis at the time of FO therapy initiation were recorded. History of mechanical ventilation, vasopressor treatment, and sepsis that had resolved by the time FO therapy was initiated were not recorded.

Statistics

Categorical data were reported as frequencies (percentages) and compared across groups by Pearson’s chi-square test or Fisher’s exact test when the expected cell count was <5 in ≥1 cell. Unadjusted continuous data were reported as medians (IQRs) and compared across groups by Wilcoxon’s rank-sum test. Logistic regression was used to determine independent predictors of treatment failure and was restricted to 8 variables, with P < 0.10 on unadjusted analysis (birth weight, PELD score, age and LAZ at FO initiation, mechanical ventilation, history of gastrointestinal bleed, sepsis, and nongastrointestinal comorbid condition). Birth weight, PELD score, and age at FO initiation were not linear in the logit and were dichotomized. Penalized likelihood estimation was used to overcome the potential for sample bias caused by the small number of subjects with treatment failure (20, 21). A final model obtained by stepwise regression resulted in 5 predictors: a PELD score ≥15 at FO initiation, history of gastrointestinal bleeding, initiation of FO at ≥16 wk of age, presence of nongastrointestinal comorbidities at FO initiation, and mechanical ventilation at FO initiation. This model fit well as assessed by the Hosmer-Lemeshow statistic (P = 0.65) (22). Multiple imputation was used to investigate the sensitivity of the results to missing data. Results from both the complete-case and multiple imputation models were reported as ORs with 95% CIs. The conventional threshold of P < 0.05 was used to establish statistical significance. The presence of gastrointestinal comorbidities at FO initiation was borderline significant by multiple imputation and is included in Table 1. All tests of significance were 2-sided, and all data were analyzed with SAS version 9.3 (SAS Institute Inc.).

TABLE 1.

Independent predictors of treatment failure¹

			Complete case (n = 146)		Multiple imputation (n = 180)
Characteristic	Cholestasis resolved (n = 127)	Treatment failure (n = 19)	OR (95% CI)	P	OR (95% CI)	P
PELD2 score ≥15	30 (24)	13 (68)	5.7 (1.8, 17.9)	0.003	6.9 (2.4, 20.2)	0.0004
History of gastrointestinal bleeding	22 (17)	8 (42)	5.1 (1.5, 17.7)	0.01	4.5 (1.5, 13.6)	0.009
Age ≥16 wk	46 (36)	13 (68)	6.9 (1.9, 25.0)	0.003	4.9 (1.5, 15.3)	0.007
Mechanical ventilation	16 (13)	7 (37)	6.7 (1.5, 29.5)	0.01	4.2 (1.3, 14.0)	0.02
Nongastrointestinal comorbidity	62 (49)	14 (74)	3.2 (0.9, 11.7)	0.08	3.5 (1.0, 11.9)	0.05

¹Values are n (%) unless otherwise indicated. Results are from logistic regression and are shown for the complete-case model and after imputation to replace missing covariate data. INR, international normalized ratio; PELD, pediatric end-stage liver disease.

²PELD = 4.8[ln(total bilirubin)] + 18.57[ln(INR)] – 6.87[ln(albumin)] + 4.36 (if aged <1 y) + 6.67 (if growth failure).

RESULTS

Outcomes of parenteral FO therapy

A total of 188 patients treated with parenteral FO for IFALD at BCH from 1 January 2004 to 31 December 2014 were identified (Figure 1). Six patients were excluded because they transferred care or were weaned off PN before cholestasis had been resolved but did not fail FO therapy (i.e., undergo transplant or die). Of the remaining 182 patients, 156 (86%) achieved resolution of cholestasis. Twenty-six (14%) patients failed therapy [8 transplantations (6 multivisceral and 2 liver) and 18 IFALD-associated deaths]. Of note, although they were not deemed treatment failures, 9 patients whose cholestasis resolved subsequently died of nonhepatic causes. Liver failure with sepsis was the most common cause of death among treatment failures, whereas sepsis was the most common cause among those who reversed on FO but subsequently died (Supplemental Table 1, Table 2).

FIGURE 1.

Patient population (n = 212). Transplants included liver and multivisceral transplants. Three subjects who died after transplant were counted as transplants and not deaths. BCH, Boston Children’s Hospital; DB, direct bilirubin; FO, fish oil; IFALD, intestinal failure–associated liver disease.

TABLE 2.

Causes of death

	n (%)
Never reversed (treatment failure)	18
Liver failure and:
Sepsis	5 (26)
Redirection of care in setting of respiratory failure	3 (17)
Redirection of care in setting of intracranial hemorrhage	3 (17)
Redirection of care in setting of poor neurologic prognosis	2 (11)
Redirection of care in setting of multiple congenital malformations	2 (11)
Hemorrhage	2 (11)
Chronic graft-versus-host disease	1 (6)
Cholestasis resolved with fish oil but patient subsequently died	9
Sepsis	3 (33)
Sudden cardiac arrest at home	2 (22)
Metastatic neuroblastoma	2 (22)
Respiratory failure or arrest	1 (11)
Unknown	1 (11)

A total of 118 patients (65%) were transferred from another institution with pre-existing IFALD compared with 64 patients (35%) who developed IFALD at BCH. A higher proportion of patients transferred to BCH with IFALD-failed FO therapy than those who developed IFALD at BCH; however, this difference was not statistically significant [20/118 (17%) compared with 6/64 (9%)].

Influence of patient demographics

Patients were 57% male and 69% Caucasian (Table 3). The median gestational age was 32 wk (IQR: 26, 36 wk), and the median birth weight was 1475 g (IQR: 800, 2305 g). The median age at PN initiation was 1 d (IQR: 0, 2 d), and the median age at FO initiation was 12.7 wk (IQR: 7.7, 24.9 wk). Patients who failed FO therapy had a lower birth weight [1020 g (IQR: 737, 1776 g) compared with 1608 g (IQR: 815, 2438 g); P = 0.03] and birth length [36.2 cm (IQR: 32.0, 41.5 cm) compared with 42.0 cm (IQR: 34.5, 47.0 cm); P = 0.05] and were older [20.4 wk (IQR: 9.9, 38.6 wk) compared with 11.7 wk (IQR: 7.3, 21.4 wk); P = 0.02] at FO therapy initiation than patients whose cholestasis resolved. There was no statistically significant difference in gestational age or age at initiation of PN between patients who failed FO therapy and those who responded.

TABLE 3.

Unadjusted predictors of treatment failure¹

Characteristic	Unknown	Overall (n = 182)	Cholestasis resolved (n = 156)	Treatment failure (n = 26)	P
Males	0	104 (57)	88 (56)	16 (62)	0.62
Caucasians	0	126 (69)	111 (71)	15 (58)	0.17
Hispanics	2	35 (19)	31 (20)	4 (15)	0.57
Birth gestational age, wk	9	32 (26, 36)	32 (26, 36)	31 (26, 34)	0.30
Birth weight, g	17	1475 (800, 2305)	1608 (815, 2438)	1020 (737, 1776)	0.03
Extremely low birth weight (<1 kg)	17	57 (35)	47 (33)	10 (48)	0.18
Birth length, cm	80	41.7 (34.0, 46.5)	42.0 (34.5, 47.0)	36.2 (32.0, 41.5)	0.05
Age at PN start, d	6	1.0 (0.0, 2.0)	1.0 (0.0, 2.0)	1.0 (0.0, 2.0)	0.59
Age at FO start, wk	7	12.7 (7.7, 24.9)	11.7 (7.3, 21.4)	20.4 (9.9, 38.6)	0.02
≤14 d FO therapy	3	6 (3)	4 (3)	2 (8)	0.18
Ever exclusive enteral feeding	2	33 (18)	27 (17)	6 (24)	0.41
Any prior operations	1	173 (96)	149 (96)	24 (96)	1.00
Stoma2	3	88 (49)	77 (50)	11 (44)	0.58
Duodenostomy	3	9 (5)	9 (6)	0 (0)	0.36
Jejunostomy	3	42 (23)	38 (25)	4 (16)	0.34
Ileostomy	3	37 (21)	31 (20)	6 (24)	0.66
Colostomy	3	4 (2)	3 (2)	1 (4)	0.46

¹Values are n (%) or medians (IQRs) unless otherwise noted. P values are from Pearson’s chi-square or Fisher’s exact tests if categorical and Wilcoxon’s rank-sum test otherwise. FO, fish oil; PN, parenteral nutrition.

²Stoma location, if present, at time of FO therapy initiation (most proximal if multiple).

Influence of baseline anthropometry

There was no statistically significant difference in weight-for-age z score [−1.41 (IQR: −2.80, −0.10) compared with −1.45 (IQR: −2.38, −0.60); P = 0.82] at the time of FO therapy initiation between patients who failed therapy and those who responded (Figure 2). Patients who failed FO therapy had a lower LAZ [−2.88 (IQR: −4.49, −1.64) compared with −1.85 (IQR: −2.85, −0.93); P = 0.05] and head circumference-for-age z score [−2.17 (IQR: −3.51, −0.88) compared with −1.36 (IQR: −2.45, −0.40); P = 0.06] at the time of FO therapy initiation than patients whose cholestasis resolved.

FIGURE 2.

Baseline anthropometric data. Sample sizes for cholestasis resolved and treatment failures, respectively, were n = 151 and n = 24 for WAZ, n = 125 and n = 21 for LAZ, and n = 122 and n = 18 for HCZ. Shown are box-and-whisker plots, where each box represents the IQR, the horizontal line represents the median, and the vertical whiskers extend ≤1.5 times the IQR. P values are from Wilcoxon’s rank-sum tests. HCZ, head circumference-for-age z score; LAZ, length-for-age z score; WAZ, weight-for-age z score.

Influence of baseline laboratory values

Cholestasis resolved in 96% (89/93) of patients with DB 2–4.9 mg/dL at FO initiation, 86% (49/57) with DB 5–9.9 mg/dL, 62% (13/21) with DB 10–14.9 mg/dL, and 45% (5/11) with DB ≥15 mg/dL (P < 0.0001 by Cochran-Armitage test for trend; data not shown). Patients who failed FO therapy had more advanced liver disease at the time of FO initiation than patients who did not fail therapy, as evidenced by higher DB [10.4 mg/dL (IQR: 7.5, 14.1 mg/dL) compared with 4.4 mg/dL (IQR: 3.1, 6.6 mg/dL); P < 0.0001]; lower platelet counts [67 × 10³/μL (IQR: 46, 104 × 10³/μL) compared with 194 × 10³/μL (IQR: 125, 301 × 10³/μL); P < 0.0001]; higher INR [1.27 (IQR: 1.16, 1.61) compared with 1.13 (IQR: 1.05, 1.21); P < 0.0001]; lower γ-glutamyltransferase [54 U/L (IQR: 41, 103 U/L) compared with 112 U/L (IQR: 76, 168 U/L); P = 0.0002]; higher aspartate aminotransferase [200 U/L (IQR: 101, 354 U/L) compared with 115 U/L (IQR: 70, 174 U/L); P = 0.003] and alanine aminotransferase [131 U/L (IQR: 67, 246 U/L) compared with 75 U/L (IQR: 38, 152 U/L); P = 0.03]; and a higher PELD score [22 (IQR: 14, 25) compared with 12 (IQR: 7, 15); P < 0.0001] (Figure 3). There was no statistically significant difference in serum albumin at FO initiation between patients who failed FO therapy and those who responded [2.8 g/dL (IQR: 2.5, 3.3 g/dL) compared with 3.1 g/dL (IQR: 2.6, 3.5 g/dL)].

FIGURE 3.

Laboratory values at FO initiation. Sample sizes for cholestasis resolved and treatment failures, respectively, were (A) n = 156 and n = 25 for direct bilirubin; (B) n = 138 and n = 24 for AST, n = 146 and n = 23 for ALT, and n = 142 and n = 23 for GGT; (C) n = 138 and n = 20 for PELD; (D) n = 156 and n = 25 for platelets; and (E) n = 152 and n = 25 for INR. Shown are box-and-whisker plots, where each box represents the IQR, the horizontal line represents the median, and the vertical whiskers extend ≤1.5 times the IQR. P values are from Wilcoxon’s rank-sum tests. PELD = 4.8[ln(total bilirubin)] + 18.57[ln(INR) − 6.87[ln(albumin)] + 4.36 (if aged < 1 y) + 6.67 (if growth failure). ALT, alanine aminotransferase; AST, aspartate aminotransferase; FO, fish oil; GGT, γ-glutamyltransferase; INR, international normalized ratio; PELD, pediatric end-stage liver disease.

Influence of baseline medical history

Intestinal failure was caused by short bowel syndrome in 155 (85%) patients, motility disorder in 10 (5%), absorptive disorder in 4 (2%), and miscellaneous other reasons in 13 (7%) (data not shown). Gastroschisis was included as an etiology of short bowel syndrome, although its morbidity is often caused by motility dysfunction. The most common diagnosis resulting in intestinal failure was necrotizing enterocolitis (43%). There were no statistically significant differences between treatment failures and nonfailures with regard to the diagnoses resulting in intestinal failure. Most patients (96%) underwent ≥1 surgical procedure before FO initiation, and approximately half (49%) had a stoma present at the time of FO initiation. There were no differences in stoma presence or location at the time of FO initiation between patients who failed FO therapy and those whose cholestasis reversed (Table 3).

Forty-four percent of patients who failed parenteral FO therapy had a history of gastrointestinal bleeding before FO initiation compared with 17% who did not fail therapy (P = 0.003). Patients who failed FO therapy were more likely to be mechanically ventilated or septic at the time of FO initiation than patients who did not fail [11 (42%) compared with 27 (18%), P = 0.007; 7 (29%) compared with 9 (6%), P = 0.002] (Figure 4). There was no statistically significant difference in vasopressor requirement at the time of FO initiation between patients who failed FO therapy and those who responded [2 (8%) compared with 5 (3%)].

FIGURE 4.

Severity of illness at FO initiation. Sample sizes for cholestasis resolved and treatment failures, respectively, were n = 147 and n = 26 for the percentage of patients who were mechanically ventilated, n = 149 and n = 25 for those requiring vasopressors, and n = 145 and n = 24 for those who were septic at the time of FO treatment initiation. FO, fish oil.

Nongastrointestinal comorbidities were more prevalent among patients who failed treatment than those whose cholestasis resolved [21/26 (81%) compared with 83/156 (53%); P = 0.009]. Patients who failed FO therapy were more likely to have a neurologic [6 (23%) compared with 12 (8%); P = 0.03], genetic/chromosomal [5 (19%) compared with 10 (6%); P = 0.04], hematologic/oncologic [4 (15%) compared with 7 (4%); P = 0.05], or endocrinologic [5 (19%) compared with 4 (3%); P = 0.004] comorbidity present at FO initiation than patients who did not fail therapy (Supplemental Table 2).

Independent predictors of treatment failure

A multivariate analysis revealed that a PELD score ≥15, history of gastrointestinal bleeding, age ≥16 wk, presence of nongastrointestinal comorbidities, and mechanical ventilation at the time of FO initiation were independent predictors of treatment failure (Table 1). No other effects, including normal birth weight (≥2500 g), LAZ, and sepsis at the time of FO initiation, were independently predictive of treatment failure after adjusting for the aforementioned variables.

DISCUSSION

Long-term PN is lifesaving for patients with intestinal failure but is associated with potentially serious complications, including IFALD. Historically, the only effective treatment options for IFALD have been the discontinuation of PN, lipid reduction, and liver transplantation (23–26). The introduction of parenteral FO therapy has changed the way IFALD is managed, allowing most treated children to experience resolution of cholestasis and avoid progression to end-stage liver disease (13). However, FO therapy is not broadly available in the United States because it is not approved by the Food and Drug Administration and must be given under a compassionate-use policy at approved institutions (27). For this reason, there is a need for establishing recommendations for practitioners at institutions without access to FO therapy to guide transfer decisions.

Risk factors for the development of IFALD have been described (28–30), and patients with DB >5.8 mg/dL have been reported to be at a greater risk for progression to end-stage liver disease (31, 32). Our study protocol at BCH requires initiating parenteral FO when DB is >2 mg/dL. However, at institutions without expertise managing IFALD or access to FO, the optimal time for transfer remains unclear. Although practitioners can agree that transfer should be pursued before the onset of irreversible liver disease, this point can be difficult to determine. Furthermore, initiating FO therapy requires patients to be enrolled in a formal study with close long-term follow-up, placing considerable demands on patients and families. It is therefore of great importance to be able to accurately provide prognostic information and to predict the likelihood of response to therapy. However, there are currently no available data to our knowledge to inform these difficult treatment decisions and to set expectations for providers and caregivers.

In this study of 182 PN-dependent patients with IFALD treated with parenteral FO therapy, 86% experienced resolution of biochemical cholestasis and 14% failed therapy, with a 4% transplant rate and 10% mortality rate. Of the 18 IFALD-associated deaths, 8 occurred after redirecting care for reasons unrelated to their liver disease. Patients who failed FO therapy were older and had more advanced biochemical and clinical liver disease at the time of FO initiation than patients who responded to therapy. FO initiation at ages ≥16 wk, a PELD score ≥15, history of gastrointestinal bleeding, presence of nongastrointestinal comorbidities, and mechanical ventilation at the time of FO initiation were independent predictors of treatment failure.

Other reports of treatment with FO monotherapy have reported transplant rates of 0–10% and mortality rates of 0–25% (9–13, 33, 34). The largest of these reports included a cohort of 97 patients with IFALD and reported resolution of cholestasis in 86% of patients, no transplants, and death in 14% of patients (13). That study reported a mortality rate of 35% with a baseline DB >10 mg/dL compared with 6% with an initial DB <5 mg/dL; however, it excluded patients who had a congenital diagnosis with a lethal prognosis. Although patients were not excluded based on comorbidities, this is consistent with our treatment failure rate (transplant plus mortality) of 44% in patients with DB ≥10 mg/dL at FO initiation and 4% in patients with DB <5 mg/dL at FO initiation.

Most patients who failed FO therapy for the treatment of IFALD were severely ill, both acutely and chronically, as a result of and oftentimes unrelated to their liver disease at the time at which FO was started. However, despite this, many severely ill patients with substantial laboratory derangements and comorbidities were still able to achieve resolution of biochemical cholestasis with FO therapy, suggesting that practitioners and caregivers may be cautiously optimistic that a treatment response is possible. Of the patients whose cholestasis resolved, 24% had a PELD score ≥15 at the time of FO initiation (Table 1), and 17% had a history of gastrointestinal bleeding. Cholestasis resolved in 56% of patients with DB ≥10 mg/dL at FO initiation and in 45% of those with a baseline DB ≥15 mg/dL. These results suggest that the reversibility of IFALD cannot be reliably determined and support the practice of offering FO therapy to all PN-dependent patients with IFALD, even those with a markedly elevated bilirubin and PELD score.

Septic episodes have been shown to contribute to the development of IFALD (35, 36). However, the presence of sepsis at the time of FO initiation did not predict treatment failure. Sepsis was the most common cause of death in patients whose cholestasis never resolved and those whose cholestasis successfully resolved with FO therapy but subsequently died. Three patients who died of sepsis after resolution of cholestasis were not deemed treatment failures despite the inability to prove that these patients’ sepsis was completely unrelated to liver disease. The sources of their infections were multiple enterocutaneous fistulae and recurrent intra-abdominal abscesses, necrotizing fasciitis of the abdominal wall, and secondary peritonitis after the placement of a ventriculoperitoneal shunt, all of which were not thought to be related to possible liver disease. Other groups reporting outcomes in patients with IFALD have reported similar findings (4, 11, 13). This highlights the need for vigilant central venous catheter care and early detection and treatment of possible infections in patients without IFALD, with active liver disease, and those whose biochemical cholestasis has resolved with FO. A high level of suspicion for possible central line-associated bloodstream infections in PN-dependent patients presenting with a variety of complaints, including gastrointestinal symptoms (emesis, increased ostomy output) and lethargy, is necessary given that patients may present without fever or leukocytosis. One cohort study of 60 pediatric intestinal failure patients who received home PN with a rate of 1.5 infections/1000 catheter days found that whereas patients with culture-proven infections frequently presented with fever, 15% of patients were afebrile (37).

There are several important limitations to this study. We were unable to assess the duration of bilirubin elevation or frequency of bacterial infections before FO initiation because these measures were not part of the prospective protocol. These data would have helped to better assess the severity of IFALD at FO initiation (38). In addition, these findings reflect the treatment of patients at a single institution. Although BCH to our knowledge has the largest experience with the use of FO for the treatment of IFALD and has shared its protocol with other institutions, protocol management, practice preferences, and referral patterns are not generalizable. Finally, because this study was designed to identify baseline characteristics associated with the failure of FO, these results do not account for diagnoses made after FO initiation such as the subsequent development of necrotizing enterocolitis or identification of chromosomal abnormalities.

Despite these limitations, this study demonstrates that most PN-dependent infants with IFALD will respond to FO therapy with resolution of biochemical cholestasis and avoidance of liver transplantation. Critically ill patients with clinical and laboratory evidence of advanced liver disease are at a higher risk of FO therapy failure. However, many patients with similarly advanced liver disease and a high acuity of general illness responded to FO, suggesting that the likelihood of treatment failure cannot be reliably predicted. Early initiation of FO therapy once cholestasis is detected is recommended. There are no definitive tests or factors to our knowledge to identify which patients have irreversible liver disease, suggesting that all PN-dependent patients with IFALD should be referred to specialized centers that care for this complex disease.