Intravenous Fish Oil and Pediatric Intestinal Failure-Associated Liver Disease: Changes in Plasma Phytosterols, Cytokines, and Bile Acids, and Erythrocyte Fatty Acids

Abstract

Background

Soybean oil emulsions (SO) are associated with intestinal failure–associated liver disease (IFALD); fish oil emulsions (FO) are used to treat IFALD. SO and FO differ with respect to their fatty acid and phytosterol content. In children with IFALD whose SO was replaced with FO, we aimed to (1) quantify changes in erythrocyte fatty acids and plasma phytosterols, cytokines, and bile acids and (2) correlate these changes with direct bilirubin (DB).

Design

This study enrolled IFALD children who received 6 months of FO. Blood samples were collected prior to FO, and after 2 weeks and 3 and 6 months of FO. The primary outcome was 3-month vs baseline biomarker concentrations

Results

At study initiation, the median patient age was 3 months (interquartile range [IQR], 3–17 months), and mean ± standard deviation DB was 5.6 ± 0.7 mg/dL (n = 14). Cholestasis reversed in 79% of subjects. Eicosapentaenoic and docosahexaenoic acid was greater than baseline (P < .001, all time points). Linoleic and arachidonic acid and sitosterol and stigmasterol were less than baseline (P < .05, all time points). Three- and 6-month interleukin-8 (IL-8) and total and conjugated bile acids were less than baseline (P < .05). Baseline IL-8 was correlated with baseline DB (r = 0.71, P < .01). Early changes in stigmasterol and IL-8 were correlated with later DB changes (r = 0.68 and 0.75, P < .05).

Conclusion

Specific fat emulsion components may play a role in IFALD. Stigmasterol and IL-8 may predict FO treatment response.

Introduction

Because of shortened and/or dysfunctional gastrointestinal tracts, children with intestinal failure are dependent on parenteral nutrition (PN) for fluids and nutrition for survival. Prolonged PN dependence is associated with intestinal failure associated liver disease (IFALD)^1–5. IFALD is manifested by a direct hyperbilirubinemia, elevated transaminases, and liver synthetic dysfunction. Children with gastrointestinal disorders, such as necrotizing enterocolitis, gastroschisis, volvulus, and intestinal atresias, and who develop short bowel syndrome are at high risk for IFALD^1–5. Once liver failure occurs, an isolated liver or liver-intestinal transplant is the only therapeutic option.

Intravenous (IV) lipids have been associated with IFALD^2–6. The only Food and Drug Administration (FDA) approved lipid emulsion in the United States for children is soy-based. We have published that approximately 75% of children with IFALD who receive a six months (mo) course of IV fish oil (FO) experienced cholestasis reversal in comparison to 6% of children who received soybean oil (SO)³. There are only a few human studies attempting to elucidate FO’s mechanism for cholestasis reversal^6,7. SO contains high concentrations of pro-inflammatory omega-6 fatty acids and phytosterols—both of which cause hepatic inflammation and biliary obstruction^6–9. When biliary flow is restricted, circulating bile acids and bilirubin increase. In contrast, FO contains anti-inflammatory omega-3 fatty acids and a negligible amount of phytosterols^3,4. This study’s purpose is to quantify changes in polyunsaturated fatty acids, phytosterols, cytokines, and bile acids in children who received FO for IFALD treatment.

In children with IFALD, we hypothesized that: 1) FO would increase the red blood cell (RBC) membrane content of omega-3 fatty acids, 2) FO would reduce the RBC membrane content of omega-6 fatty acids and plasma phytosterol, cytokine, and bile acid concentrations, 3) baseline biomarker measurements would correlate with baseline serum direct bilirubin (DB) concentrations, and 4) early biomarker changes would predict later DB changes.

Subjects and Methods

This is a prospective, observational study. The study’s primary outcome was the difference between 3 mo and baseline RBC polyunsaturated fatty acid percentages and plasma phytosterols, cytokines, and bile acids. Baseline was defined as the start of the study (prior to FO treatment). Secondary outcomes included correlations between biomarker measurements with DB, and early biomarker changes with later DB changes. Cholestasis reversal was defined as a serum DB < 2 mg/dL.

Written informed consent was obtained from all parents/legal guardians. The institutional review board at the University of California, Los Angeles approved the study. The FDA granted an Investigational New Drug Application (105,326) for FO. This clinical trial is registered at www.clinicaltrials.gov (NCT00969332).

Eligibility criteria for the FO treatment trial included a gastrointestinal disorder, > two wk of age, < 18 years of age, a serum DB ≥ 2 mg/dL on two consecutive measurements, anticipated PN course > 30 days, and > 60% of calories from PN³. Only subjects with blood samples at baseline and 3 mo were included. Subjects with a primary liver disease, inborn error of metabolism, seafood/egg allergy, hemorrhagic disorder, hemodynamic instability, comatose state, stroke, pulmonary embolism, myocardial infarction, diabetes, or fatal chromosomal disorder were excluded. Due to the risk for anemia, subjects with a weight less 1.5 kg were excluded.

SO (Intralipid®, Fresenius Kabi, Uppsala, Sweden) was replaced with FO (Omegaven®, Fresenius Kabi, Bad Homburg Germany) dosed at 0.5 g/kg/d IV for the first two days, then 1 g/kg/d IV thereafter over 8–24 hours for 6 mo³. FO was discontinued prior to 6 mo if the subject no longer required PN, underwent liver and/or intestinal transplantation, or developed an adverse complication attributed to FO.

The medical team dictated the management of the subjects’ intestinal failure. Generally, at our institution, serum DB and liver function tests are measured on a bi-weekly to monthly basis for children with IFALD.

Approximately 2.5 cc of blood was collected at baseline and during the study (2 wk and 3 and 6 mo). If the subject required a RBC transfusion or was suspected to have an infection, the blood collection was delayed by at least two days. If sepsis was confirmed, the blood collection was delayed for seven days. At the time of the blood collection, FO was discontinued for at least two hours. Samples were centrifuged to separate plasma from RBCs, and stored at −80ºC.

Erythrocyte Membrane Fatty Acids

Fatty acids are expressed as a percent of total fatty acids. RBC membranes were prepared from 0.2 ml of RBCs by cell lysis with water followed by membrane washing with phosphate buffered saline. Membranes were recovered by centrifugation at 20000xg for 20 minutes. Total fatty acids were converted to methyl esters, and were separated and quantified using an Agilent Technologies (San Diego, CA) 5890A series II gas chromatography (GC)¹⁰. Quantification was based on the recovery of a known quantity of the internal standard (tridecanoic acid, NuChek Preparation Inc., Elysian, MN) and on the response ratio of fatty acid standards purchased from NuChek Preparation Inc. (Elysian, MN). For quality control, a pooled RBC sample was used. This pooled RBC sample was analyzed with each batch of samples. Using this pooled RBC sample the following inter-assay coefficient of variation has been established: linoleic acid 2.9%, arachidonic acid (AA) 4.5%, and docosahexaenoic acid (DHA) 5.8%.

Plasma Phytosterols

Gas chromatography-mass spectrometry (GC-MS) was used to measure plant sterols. An internal standard (epicoprostanol; Steraloids, Newhaven, RI) was added to plasma, and calibrants were generated using standards (sitosterol and stigmasterol from Steraloids, campesterol from Avanti Polar Lipids, Alabaster, AL). Sterols were saponified by the addition of ethanol/KOH, incubated at 37°C for one hour and the aqueous phase was extracted. Concentrations of the trimethylsilyl ether derivatives of plant sterols were measured using GC (pulsed/splitless injection) performed with a ZB1701 column (Phenomenex, Torrance, CA) coupled to a mass spectrometer (Agilent GC 6890N and MS 5975; Santa Clara, CA). Mass spectra were collected in selected ion mode with m/z= 355 and 370 ions monitored for epicoprostanol internal standard (quantifying and qualifying ions respectively), m/z= 306 and 305 ions for sitosterol, m/z= 367 and 382 ions for campesterol and m/z= 394 and 484 ions for stigmasterol. Analyte concentrations were calculated across the range 0.04–8.0 mg/dL using calibration curves generated by performing a least-squares linear regression for peak area ratios plotted against specified calibrant concentration. The lower limit of quantification was determined as the lowest spiked concentration in matrix for which the signal-to-noise ratio was ≥ 5. The between-run reproducibility across the calibration curve was determined to be a relative standard deviation < 20% with an accuracy between 85–115%.

Plasma Cytokines

IL1-β, IL-6, IL-8, IL-10, TNF-α, and IFN-γ were measured using two multiplexed immunometric assay panels for human cytokines (Luminex® xMAP®, R&D Systems®, R&DSystems^TM, Minneapolis, MN). The antibody-conjugated beads are allowed to react with the sample and a secondary detection antibody to form a capture sandwich immunoassay. After the assay was completed, the assay solution is drawn into the Bio-Plex™ 200 Luminex array reader (BioRAD, Hercules, CA) to quantify the analytes.

Plasma Bile Acids

The isotope-dilution liquid chromatography-tandem mass spectrometry (LC-MS/MS) method for measurement of bile acids was used. Stable isotopes were purchased in the form of a kit to validate the measurement of human bile acids with an accuracy between 85–115% and relative standard deviation < 20% (Biocrates Life Sciences, Innsbruck, Austria). In brief, stable-isotope labeled internal standards were added to 10 μl unknown plasma or calibrant samples then subjected to proprietary filter-plate extraction. The filtrate was analyzed using a LC-MS/MS method with negative-mode electrospray ionization(ESI)-tandem mass spectrometry(MS/MS) data generated using an AB SCIEX 4000 QTRAP instrument (AB SCIEX, Foster City CA). The internal standard and calibration range for each bile acid were as follows: cholic acid, d5-cholic acid internal standard was used with a calibration range of 0.03–75 μM; chenodeoxycholic acid, d5-chenodeoxycholic acid internal standard was used with a calibration range of 0.02–30 μM; glycocholic acid, d5-glycocholic acid internal standard was used with a calibration range of 0.03–75 μM; glycochenodeoxycholic acid, d4-glycochenodeoxycholic acid internal standard was used with a calibration range of 0.02–10 μM; taurocholic acid, d5-taurocholic internal standard was used with a calibration range of 0.02–5 μM; and taurochenodeoxycholic acid, d5-taurochenodeoxycholic internal standard was used with a calibration range of 0.02–20 μM.

Statistics

Biomarker concentrations were compared using a paired t-test. Categorical variables were compared using the Fisher’s exact test. Continuous variables at the start and end of the study and for subjects who remained cholestatic vs. those whose cholestasis resolved were compared using the Wilcoxon rank sum test. Pearson correlations were used to analyze the relationship between biomarkers and routine laboratory tests. Since not all laboratory tests were measured at exactly two weeks, some values were aligned using local linear interpolation for comparison over time using autoregressive regression models. There was no extrapolation beyond time points were values where not observed. Interpolation was not used for biomarker data. Results are presented as a median (IQR) and mean (±SD) for biomarker data. A two-sided p-value <0.05 was considered statistically significant. R version 3.2.4 (R Foundation, Vienna, Austria) was used for statistical analysis.

Results

Subjects

Subjects were recruited from the University of California, Los Angeles Mattel Children’s Hospital from April 2011 to June 2014. Twenty-one subjects were treated with FO during this time. Seven subjects did not have blood available at the baseline and 3 mo study visit (Figure 1). Of the 14 subjects included in the study, the median time for follow-up was 168 days (168–169 days).

Figure 1.

Patient flow chart.

All subjects were PN-dependent since birth, and most subjects had a history of prematurity and short bowel syndrome. During the study, 42% (n=6) of the subjects were diagnosed with at least one episode of bacteremia (Table 1). PN intake at the beginning of the study was greater than PN intake at the end of the study (79 (60–86) vs. 63 (50–77) kcal/kg/d, p=0.03). There was a marginal, but statistically significant increase in enteral nutrition (0.2 (0–14) vs. 12 (7–44) kcal/kg/d, p=0.03). The lipid dose at the start of the study was comparable to the end of the study (SO, 1 (0.97–1.1) vs. FO, 1 (1–1.1) g/kg/d, p=0.2). The median SO dose over the three months prior to starting FO was 0.7 g/kg/d (0.4–1.3 g/kg/d).

Table I.

Subject Characteristics

Study Population	median (IQR) or % (n)
Age fish oil started (month)	3 (3, 17)
Age fish oil discontinued (month)	8 (8, 30)
Gestational age (weeks)	34 (32, 36)
Gender – Male	57 (8)
Race – White	86 (12)
Ethnicity – Hispanic	71 (10)
Primary gastrointestinal diagnosis
Intestinal atresia	29 (4)
Necrotizing enterocolitis	21 (3)
Gastroschisis	14 (2)
Volvulus	14 (2)
Total aganglionosis	7 (1)
Other	14 (2)
No. of gastrointestinal surgeries at the start of the study	3 (1, 5)
No. of gastrointestinal surgeries during the study	0 (0,1)
Small bowel length (cm)I	13 (8, 34)
Gastrointestinal continuity – Yes	71 (10)
Ileocecal valve – Yes	57 (8)
Urosodiol at the start of the study – Yes	58 (8)
Urosodiol during the study – Yes	29 (4)
No. of episodes of late onset sepsis during the study	0 (0,2)
Transplant	7 (1)
Cholestasis resolved	79 (11)

^In=12.

DB decreased over time (−0.3 mg/dL/wk, 95% CI (−0.4, −0.2), p<0.001) and cholestasis resolved in 79% of the subjects (n=11) at a median of 13 wk (10–16 wk). Aspartate (−9.0 IU/L/wk, 95% CI (−13, −6), p<0.001) and alanine aminotransferase (−6.9 IU/L/wk, 95% CI (−11 to −3), p=0.002) and triglyceride concentrations (−5.1 IU/L/wk, 95% CI (−7, −3), p<0.001) followed a similar trend (Table II). Only one subject was weaned from PN during the study. One subject received a multi-visceral transplant for portal hypertension complicated by gastrointestinal bleeding despite improving serum DB concentrations.

Table II.

Direct Bilirubin and Liver Function Tests with Fish Oil Treatment

Laboratory Values	Baseline	2 week	3 month	6 month
Direct bilirubin, mg/dL	5.6±0.7	5.6±0.9	2.1±0.5**	0.3±0.1**
Asparate aminotransferase, IU/L	218±39	195±27	116±24**	39±3**
Alanine aminotransferase, IU/L	174±39	162±28	97±17*	27±4
Alkaline phosphatase, IU/L	412±49	395±35	428±52	358±49
Triglyceride, IU/L	173±28	153±19	76±8	42±6

Each study visit is compared to baseline.

^*p<0.05,

^**p<0.01.

Data are reported as mean (±SD).

Biomarkers Concentrations

See Table III for number of blood specimens collected during the study.

Table III.

Number of Specimens at Each Study Visit

	Study Visit	Number of Samples
		Fatty acids	Phytosterols	Cytokines	Bile acids
Baseline	--	14	11	14	12
2 week	14 days (13–21 days)	13	10	12	11
3 month	2.9 mo (2.8–3.1 month)	13	12	14	13
6 month	5.3 mo (5.2–5.5 month)	11	9	11	10

Data are presented as a median (IQR). Mo, month.

The percentage of linoleic and AA in the RBC membrane at 2 wk, 3 mo, and 6 mo was less than baseline (p<0.001 for all). In contrast, 2 wk, 3 mo and 6 mo DHA and EPA percentages were dramatically greater than baseline (all > 100%, p<0.001 for all). The proportion of α-linolenic acid remained stable throughout the study. Two wk stigmasterol and sitosterol were 17–21% less than baseline (p<0.01 for both). At 3 mo, stigmasterol and sitosterol were at least 80% less than baseline (p<0.01 for both). By 6 mo, these phytosterols were 96% less than baseline (p<0.01 for both).

With respect to cytokines, 2 wk and 3 and 6 mo TNF-α concentrations were 9%, 14%, and 20% less than baseline (p<0.01, p=0.06 and 0.01, respectively). Three and 6 mo IL-8 were 64 and 83% less than baseline (p<0.001 for both). At 6 mo only, IL-10 were less than baseline (p=0.03). Lastly, at 3 and 6 mo, total cholic and chenodoxycholic acid and all conjugated bile acids were at least 65% less than baseline (p≤0.01 for all) (Table IV).

Table IV.

Biomarker Values at Each Study Visit

Biomarker	Baseline	2 week	3 month	6 month
Linoleic acid (%)	13±0.6	6.8±0.5**	4.5±0.7**	4.3±0.9**
α-linolenic acid (%)	0.3±0.1	0.3±0.1	0.3±0.1	0.2±0.1
Arachidonic acid (%)	15±0.5	12±0.4**	7.8±0.4**	7.1±0.7**
Eicosapentaenoic acid (%)	0.4±0.1	3.9±0.3**	6.3±0.3**	6.9±0.6**
Docosahexaenoic acid (%)	3.7±0.2	7.5±0.4**	11.9±0.4**	12.7±0.7**
Stigmasterol (mg/dL)	2.8±0.5	2.2±0.5**	0.4±0.1**	0.1±0.02**
Sitosterol (mg/dL)	14.2±2.3	11.8±2.3**	2.9±0.9**	0.6±0.2**
Campesterol (mg/dL)	2.5±0.5	2.2±0.5	0.7±0.2**	0.3±0.1**
TNF-α (pg/mL)	35±2	32±3 **	30±3	28±2**
IL-8 (pg/mL)	95±13	83±15	34±5**	16±2**
IL-10 (pg/mL)	3.7±0.6	3.6±0.5	2.8±0.5	2.3±0.4*
IL-1β (pg/mL)	4.3±0.6	3.0±0.8*	3.4±0.5	4.0±0.8
IL-6 (pg/mL)	19±5	14±2	11±1	10±2
IFN-γ (pg/mL)	1.6±0.0	1.0±0.2	1.8±0.4	3.8±1.5
Cholic acid (μmol/L)	0.3±0.2	0.4±0.4	0.7±0.4	1.4±0.9
Glycocholic acid (μmol/L)	22±6	16±5	3.2±0.6**	1.4±0.4**
Taurocholic acid (μmol/L)	9.8±1.9	5.6±1.6I	1.2±0.5**	0.2±0.1**
Total cholic acid (μmol/L)	27±6	17±5* I	5±1**	2±0.8**
CDA (μmol/L)	0.3±0.2	0.2±0.2	2.2±1	2.5±1*
GCDA (μmol/L)	21±4	14±3	6.7±3.2*	2.7±1.5*
TCDA (μmol/L)	13.3±3.4	12.5±4.2	1.6±0.5**	0.4±0.2**
Total CDA (μmol/L)	29±6	21±6	10±3**	4±2**

Each study visit is compared to baseline.

^*p<0.05,

^**p<0.01.

Data are reported as mean (±SD).

^IOutlier was removed. When outlier was included, the value was 8.9±3.5 μmol/L (p=0.7) for taurocholic acid, and 20±5 μmol/L (p=0.1) for total cholic acid. CDA, chenodoxycholic acid, GCDA, glycochenodoxycholic acid, TCDA, taurochenodoxycholic acid. Data is reported as mean (±SD).

Biomarker Correlations

Baseline DB was positively correlated with baseline IL-8 (r=0.71, p<0.01) (Figure 2A). Baseline stigmasterol demonstrated a similar, yet not statistically significant trend (r=0.60, p=0.05) (Table V). The converse was true for AA (r=−0.57, p=0.03) (Figure 2B). While there was no correlation between baseline serum asparate aminotransferase and baseline biomarker concentrations (data not shown), campesterol (r=0.75, p<0.01) and cholic (r=0.8, p<0.01) and chenodeoxycolic acid (r=0.8, p<0.01) were positively correlated with serum alanine aminotransferase. Baseline TNF-α (r=0.79, p=0.03) and IL-8 (r=0.78, p=0.03) were positively correlated with triglycerides, while AA (r=−0.85, p=0.02) and EPA (r=−0.87, p=0.01) demonstrated the opposite relationship. Of note, only sitosterols positively correlated with PN intake (kcal/kg/d) (r=0.63, p=0.04). Lipid dose was not correlated with phytosterols (data not shown).

Figure 2.

Correlation of biomarkers with direct bilirubin. A. Baseline arachidonic acid and direct bilirubin, B. Baseline IL-8 and direct bilirubin, C. Two week stigmasterol and 3 month direct bilirubin changes, D. Two week IL-8 and 3 month direct bilirubin changes.

Table V.

Biomarker Correlations with Serum Direct Bilirubin

	Baseline Direct Bilirubin	3 month Direct BilirubinI	6 month Direct BilirubinII
Linoleic acid	0.21 (05)	−0.30 (0.3)	0.03 (1)
α-linolenic acid	−0.07 (0.8)	0.23 (0.46)	−0.20 (0.7)
Arachidonic acid	−0.57 (0.03)	−0.44 (0.1)	−0.17 (0.8)
Eicosapentaenoic acid	0.02 (1)	0.26 (0.4)	0.43 (0.5)
Docosahexaenoic acid	−0.41 (0.1)	−0.33 (0.3)	0.17 (0.8)
Stigmasterol	0.60 (0.05)	0.68 (0.03)	0.50 (0.5)
Sitosterol	0.27 (0.4)	0.39 (0.3)	0.91 (0.1)
Campesterol	−0.12 (0.7)	0.06 (0.9)	−0.60 (0.4)
TNF-α	−0.09 (0.8)	0.38 (0.2)	−0.25 (0.6)
IL-8	0.71 (<0.01)	0.75 (<0.01)	0.37 (0.5)
IL-10	0.03 (0.9)	0.41 (0.2)	0.16 (0.8)
IL-1β	−0.07 (0.8)	0.25 (0.4)	−0.3 (0.6)
IL-6	−0.19 (0.5)	−0.07 (0.8)	−0.21 (0.7)
IFN-γ	0.38 (0.2)	0.31 (0.4)	0.14 (0.8)
Cholic acid	−0.38 (0.2)	−0.26 (0.4)	0.38 (0.5)
Glycocholic acid	−0.16 (0.6)	−0.32 (0.3)	0.62 (0.3)
Taurocholic acid	−0.04 (0.9)	−0.39 (0.3) III	0.06 (0.9)
Total cholic acid	−0.03 (0.9)	−0.23 (0.4) III	0.2 (0.7)
Chenodoxycholic acid	−0.35 (0.3)	−0.26 (0.4)	0.37 (0.5)
Glycochenodoxycholic acid	0.15 (0.6)	−0.25 (0.5)	−0.40 (0.5)
Taurochenodoxycholic acid	0.52 (0.08)	−0.29 (0.4)	−0.47 (0.4)
Total chenodoxycholic acid	0.45 (0.1)	0.09 (0.8)	−0.55 (0.3)

Correlations presented as Pearson correlation coefficients (p-value).

^ITwo week biomarker (2 week – baseline) and 3 month direct bilirubin changes (3 month – baseline).

^IIThree month biomarker changes (3 month – baseline) and six month direct bilirubin correlations (6 month – baseline).

^IIIOutlier was removed. When outlier was included, the results were r=−0.71 (p=0.02) for taurocholic acid, and r= −0.54 (p=0.04) and total cholic acid.

Two wk stigmasterol (r=0.68, p=0.03) and IL-8 (r=0.75, p<0.01) differences (2 wk – baseline) were positively correlated with the 3 mo DB differences (3 mo – baseline) indicating that early changes in these biomarkers may predict later changes in DB (Figure 2C, 2D). Three mo biomarker differences (3 mo – baseline) were not correlated with 6 mo DB differences (6 mo – baseline), although sitosterol demonstrated a non-significant correlation (r=0.91, p=0.09) (Table IV).

Two wk IL-8 concentrations were less in subjects whose cholestasis resolved vs. subjects whose cholestasis did not resolve (75±54 vs. 122±12 pg/dL, p=0.04). At 3 mo, the percentage of AA was greater in resolvers vs. non-resolvers (8±1.6 vs. 7±0.6%, p=0.03). Three mo linoleic (r=−0.6, p=0.04) and DHA differences (r=0.8, p<0.001) were correlated with cholestasis reversal, indicating that subjects with a larger linoleic acid decrease and larger DHA increase may be more likely to normalize their DB concentrations.

Discussion

In a group of children with IFALD who were treated with FO, we witnessed a dramatic change in polyunsaturated fatty acids, and reduction of circulating phytosterols, cytokines, and bile acids. While these changes represent a change in the composition of the lipid product, we believe this study provides information regarding the pathomechanism of IFALD, and the contribution of lipid emulsions to this disease. We speculate that FO’s reduced phytosterol load, in combination with a decrease in inflammation, which is partially driven by an increased omega-3 fatty acid provision, improves biliary flow in IFALD children^3,4,6–9.

Phytosterols are steroid compounds analogous to cholesterol and found in vegetable foods. Because only 0.4–5% of dietary phytosterols are absorbed in the intestine and humans lack the ability to synthesize phytosterols, plasma concentrations are low^11,12. However, when SO is administered with PN, it is infused into a patient’s bloodstream and contains 370–500 mg/L of phytosterols with greater than 70% being sitosterol. Phytosterol concentrations in control children have been reported at 0.4 mg/dL, while concentrations in IFALD children were as high as 17 mg/dL^13,14. Sitosterol was positively correlated with PN intake, but not SO dose. This is most likely because subjects were prescribed a low dose of SO.

Serum and hepatic phytosterol concentrations correlate with cholestasis, inflammation, and fibrosis on liver biopsy in children with IFALD^6,14,15. In mouse models, stimgasterol antagonizes the nuclear receptors, farnesoid X (FXR) and liver X receptor (LXR), which reduces hepatocyte expression of sterol, bile acid, and bilirubin transporters^8,9. Moreover, sitosterol inhibits cholesterol 7α-hydroxlase, the rate-limiting step in the conversion of cholesterol into bile acids. As a result, bile acids accumulate in hepatocytes and cause mitochondrial damage and cholestasis. Based on the assumption that high doses of intravenous phytosterols are hepatotoxic, it is hypothesized that a reduced SO dose protects against and treats IFALD^2,5,16. At the start of this study, 9 of the 14 subjects received ≤ 1 g/kg/d of SO and still developed advanced IFALD. Hence, a reduction in phytosterol load alone may not be sufficient to prevent or treat IFALD².

Studies indicate that phytosterols and inflammation act syngeristically to cause IFALD^8,9. The intestinal barrier in children with IFALD is compromised. As a result, toll-like receptor agonists, such as lipopolysaccharides, cross the intestinal mucosa, and enter into the portal circulation. Lipopolysaccharides, along with various cytokines, such as IL-6, TNF-α, and IL-1β, induce biliary cells to produce IL-8. IL-8 attracts neutrophils and T-cells to the liver, causing inflammation and fibrosis^17–19. Lipopolysaccharides also activate Kupffer cells, causing cholestasis^8.9. When compared to non-cholestatic patients with liver disease, cholestatic patients have higher IL-8 concentrations^17,18. In our study, baseline stigmasterol and IL-8 correlated with disease severity, and early changes in stimgasterol and IL-8 predicted later changes in DB. Hence, stigmsterol and IL-8 may be potential companion biomarkers for IFALD diagnosis and surveillance.

Similar to other FO studies, we observed a striking increase in DHA and EPA, and decrease in linoleic acid and AA²⁰. In a study by Le et al, serum linoleic and α-linolenic acid, and DHA and EPA concentrations changed immediately after FO initiation and remained stable thereafter²⁰. In contrast, in our study, the RBC membrane’s proportion of α-linoleic was unchanged. In addition, the percentage of linoleic acid appeared to decrease and DHA and EPA appeared to increase during the first 3 mo. These differences may be because serum samples reflect acute dietary changes, while the RBC samples reflect long-term dietary intakes. In both studies, AA gradually declined with FO²⁰.

This slow AA decline is most likely secondary to FO’s lack of linoleic acid, small provision of AA, and large EPA concentration^3,20. Because EPA competitively inhibits AA synthesis, one potential concern with FO is an AA deficiency and its affect on growth and neurodevelopment^16,21,22. Josephson et al examined plasma and brain fatty acid profiles in piglets infused with a short PN course with FO or low dose SO. The two groups had similar plasma essential fatty acid percentages. However, the EPA percentage was greater, and AA percentage and brain size was less in the FO group vs. the SO group. Despite these changes, AA brain content was unaffected by the emulsion type and DHA brain content increased in the FO group¹⁶. It is unknown if these changes affected the piglet’s cognition and behavior. Despite these concerns, FO studies have not reported growth deceleration or an essential fatty acid deficiency, measured by a triene:tetrane ratio^3,4,20.

Confirming previous publications, triglycerides improved over time in this cohort^3,4. Baseline triglycerides were positively correlated with IL-8 and TNF-α, while EPA demonstrated an inverse relationship. Studies have demonstrated that high omega-6:omega-3 ratios promote hepatic fat deposition, and that EPA protects against steatosis²³. There was a statistically significant, yet weak negative correlation between baseline AA and baseline DB and triglycerides. These results may be due to an increase in enteral AA consumption, sample size issue, variability in measurements, or other unknown reasons.

In contrast to rodent studies, piglet studies indicate that a lack of Vitamin E, not stigmasterol, promotes IFALD^8,9,24. In comparison to SO, FO contains a higher Vitamin E concentration, which helps prevent lipid perioxidation. Piglets infused with SO had higher DB and bile acid concentrations when compared to piglets who received SO plus α-tocopherol, FO alone, and FO plus phytosterols. Interestingly, there was no change in FXR hepatic expression in these groups²⁴. Due to restrictions in the amount of blood we could collect, we did not measure Vitamin E or markers of lipid peroxidation.

Because this study lacked a control group, these findings represent associations, and we cannot make conclusions regarding cause and effect. Additional limitations include the inability to include confounding variables in our analysis (sepsis, PN duration, prematurity, and low birth weight), small sample size, and biomarker measurements at only four time points¹. While FO is not FDA approved, it is used at our institution, in conjunction with a multi-disciplinary approach to intestinal failure, to avoid liver failure. As result, it was not possible to measure these biomarkers in children with IFALD who were receiving SO. It should be noted that subjects in this study would be considered high risk for liver failure considering their gastrointestinal anatomy, degree of PN dependence, high rate of sepsis, and the fact that only one subject was weaned from PN without a transplant.

At the time this study was designed, there was a lack of data on how fish oil alters the biomarkers of interest in this population. As a result, we were unable to power this study for a change in a biomarker. These time points were selected to reflect acute and chronic changes. We and others have previously demonstrated that reversal of cholestasis occurs at approximately 4–18 weeks of FO treatment for most subjects^3,4. Hence, our outcomes were to compare 3-month biomarker concentrations to baseline, and correlate these biomarkers with DB. We were unable to measure biomarkers for all subjects at each study visit due to a missed study visit or inability to collect sufficient blood. For these reasons, the power to detect specific changes and/or correlations decreased.

Conclusion

In a group of children with IFALD, intravenous fish oil treatment resulted in biochemical reversal of cholestasis and was associated with reduction in plasma phytosterols, cytokines, and bile acids and a change in the erythrocyte content of polyunsaturated fatty acids. IL-8 correlated with direct hyperbilirubinemia, and early changes in stigmasterol and IL-8 correlated with later changes in direct bilirubin. These biomarkers may play a role in IFALD and be indicators of treatment response.

Clinical Relevancy Statement.

Soybean-based lipid emulsions are associated with intestinal failure associated liver disease (IFALD), while fish-based lipid emulsions have been used to prevent liver failure. Soybean and fish oil differ with respect to their polyunsaturated fatty acid and phytosterol concentrations. In this study, fish oil reversed cholestasis in 79% of children. There was a significant change in the erythrocyte membrane’s fatty acid content and decrease in plasma phytosterol, cytokine, and bile acid concentrations. Stigmasterol and IL-8 correlated with disease severity, and early changes in these biomarkers predicted later changes in direct bilirubin. This study provides important information on the hepatoprotective effects of fish oil.

Abbreviations

IFALD

intestinal failure associated liver disease

PN

parenteral nutrition

SO

soybean oil

FO

fish oil

DB

direct bilirubin

RBC

red blood cell

wk

weeks

mo

months

AA

arachidonic acid

EPA

eicosapentaenoic acid

DHA

docosahexaenoic acid