Results of the search

The search yielded 1673 articles from medical literature databases and 88 clinical trial registry reports. We removed duplicates for a total of 1164 records. We excluded 1141 records based on titles and abstracts and assessed 23 full‐text articles for eligibility. We excluded 14 full‐text articles with reasons. Five studies were excluded after discussion between authors (VK, MM) (Characteristics of excluded studies table). We included nine studies in the review (Characteristics of included studies table). We reported the study selection process in the PRISMA flow diagram (Figure 1).

Figure 1

---

---

PRISMA flow diagram.

Included studies

Angsten 2002 was a double‐blind, randomised trial performed at the Department of Paediatric Surgery, University Children's Hospital, Uppsala, Sweden.

• Population: inclusion criteria were neonates of either sex with a gestational age of 36 to 41 weeks at birth, birthweight of 2.5 g to 4.0 kg, and a postnatal age of up to four days, needing surgical treatment and requiring TPN for a minimum of five days. Neonates needing surgery for various oesophago‐gastrointestinal malformations were assigned before surgery. Exclusion criteria were metabolic disease, renal disease, inflammatory disease, septic syndrome or a haematological disorder.

• Objectives: to determine whether MCT‐based LEs were more efficient in producing energy than pure LCT‐based LEs without adverse effects.

• Interventions: treatment group (n = 10) received Vasolipid, containing 50% MCT and 50% LCT and the control group (n = 10) received Intralipid. The total amount of fat infused on the first day was 2 g/kg/day and this amount was increased at a daily rate of 0.5 g/kg/day until an infusion rate of 3 g/kg/day was reached. Before the start of TPN, after five days of TPN and five hours after the LE ended, two blood samples (2.5 mL) were taken for measurement of free and total plasma carnitine concentrations, free fatty acids, fatty acid composition of plasma cholesterol esters, phospholipids, triglycerides and ketones as b‐hydroxybutyrate. A piece of adipose tissue weighing approximately 15 mg and a piece of muscle weighing approximately 30 mg were taken during surgery as an open biopsy with a scalpel.

• Outcomes: primary outcomes: fatty acid compositions in the plasma and adipose tissue and carnitine concentrations in the plasma and muscle. Secondary outcomes: heart rate, body temperature, bodyweight and clinical symptoms were recorded daily. Haematological (haemoglobin, leukocytes and platelets) and biochemical (aspartate aminotransferase, ALT, sodium, potassium, creatinine, ALP, albumin and bilirubin) parameters were measured on day 0, just before the first infusion of TPN started, and after termination of the study.

Ariyawangso 2014 was a prospective, open‐label and randomised study in surgical neonates conducted between 1 March 2013 and 30 November 2013 at Queen Sirikit National Institute of Child Health, Bangkok, Thailand.

• Population: surgical neonates with gastroschisis, omphalocoele, jejuno‐ileal atresia or duodenal atresia requiring PN for at least seven consecutive days were eligible to participate in the study. Infants were excluded if they had already received PN. Additional exclusion criteria included neonates with the following conditions: Down's syndrome, severe congenital malformations, patent ductus arteriosus (PDA), anuria due to circulation failure, liver/haemolytic diseases, hepatitis, thrombocytopenia, HIV, oxygen saturation (SO₂) less than 80% for more than two hours, severe acidosis, application of catecholamines, multiorgan failure, hypoxic‐ischaemic encephalopathy, shock and sepsis.

• Objectives: to assess the safety and efficacy of 20% SMOFlipid in surgical neonates receiving PN compared with an S‐LE (20% Intralipid).

• Interventions: 42 surgical infants requiring PN allocated to receive either SMOFlipid (n = 21) or Intralipid (n = 21). The LE was started parenterally at 0.5 g/kg/day on day 1 and was increased by increments of 0.5 g/kg/day daily up to 3.0 g/kg/day for term and 3.5 g/kg/day for preterm neonates, until it reached 50% of total energy intake. Blood samples were collected for assessment of primary and secondary outcomes on day 0, 8, 15 and 22 of the study if PN was continued.

• Outcomes: primary outcomes: safety outcomes including liver enzymes ALT, aspartate aminotransferase, ALP, GGT, total bilirubin, conjugated bilirubin, lipid profile, blood urea nitrogen, serum creatinine and haematological parameters. Secondary outcomes: efficacy outcomes including infant's weight, head circumference and length at study entry and exit were recorded.

Diamond 2017 was a multicentre parallel‐group RCT with 1:1 assignment comparing SMOFlipid with Intralipid for the prevention of progression of Intestinal failure associated liver disease (IFALD) in infants.

• Population: primary inclusion criteria were an infant aged less than 24 months with short bowel syndrome or intestinal failure who received substantial PN support (greater than 40% total calories) and was demonstrating early hepatic dysfunction (Cbil 17 μmol/L to 50 μmol/L (1 mg/dL to 3 mg/dL)) in the absence of sepsis. The study was included in the current review as the age of all the participants at study entry was less than 9 weeks. Exclusion criteria: sepsis or haemodynamic instability; coagulopathy (platelets ≤ 150 000/μL, or international normalised ratio (INR) ≥ 1.4); hypersensitivity to fish, egg or soy protein or to any of the active substances or excipients; current enrolment in another clinical trial involving a surgical or pharmacological intervention; serum Cbil greater than 50 μmol/L; hyperlipidaemia; treatment with intravenous N‐acetylcysteine or oral ursodeoxycholic acid; renal insufficiency; fluid imbalance and unstable medical conditions.

• Objectives: to examine whether SMOFlipid prevented progression of IFALD in PN‐ dependent infants with early IFALD (Cbil 17 μmol/L to 50 μmol/L (1 mg/dL to 3 mg/dL)).

• Interventions: participants received trial lipid for up to 12 weeks. Participants also ended the trial if they achieved full enteral tolerance prior to this time or if they developed progressive liver disease defined by a serum Cbil exceeding 100 μmol/L for greater than 14 days. The PN solution was formulated according to a nomogram that took into account the proportion of caloric intake received parenterally. Choice of feeds, timing of the adjustment of the enteral feeds and ratio of enteral to parenteral feeds were at the discretion of the treating physician.

• Outcomes: the primary end point for analysis was Cbil the week that the child received his or her last dose of trial lipid (at 12 weeks, at full enteral tolerance or on the development of progressive liver disease). Other data collected at baseline, weekly while on the trial and 4 weeks following trial completion (referred to as "post‐trial") included serum Cbil, serum albumin, liver enzymes, parenteral intake and enteral intake.

Magnusson 1997 was a double‐blind, RCT performed at the Department of Pediatric Surgery, University Hospital Uppsala, Sweden.

• Population: neonates undergoing surgery for various oesophago‐gastrointestinal malformations were assigned the day after surgery to either Intralipid (control group) or PFE 4501 (treatment group) in a double‐blind, randomised manner.

• Objectives: to compare Intralipid with PFE 4501 in neonates with special reference to changes in fatty acid composition in plasma and adipose tissue, plasma carnitine concentrations and tolerance during a short period of TPN.

• Interventions: 20 neonates received Intralipid (n = 10) or PFE (n = 10). Inclusion criteria were neonates of either sex needing surgical treatment, age up to and including seven days, and requiring TPN for a minimum of five days. The exclusion criteria were metabolic disease, renal disease, inflammatory disease, septic syndrome or a haematological disorder. The total amount of fat infused was 2 g/kg on the first day and increased at a daily rate of 1 g/kg until a rate of 4 g/kg/day was reached.

• Outcomes: primary outcomes: to examine the changes in fatty acid composition in plasma and adipose tissue, plasma carnitine concentrations and tolerance during a short period of TPN. Heart rate, body temperature, bodyweight and clinical symptoms were recorded daily. Haematological (haemoglobin, leukocytes and platelets) and biochemical (aspartate aminotransferase, ALT, sodium, potassium, creatinine, ALP, albumin and bilirubin) parameters were measured on day zero, just before the first infusion of TPN started, and after termination of the study.

Lam 2014 was a prospective, double‐blind, RCT conducted at a level III university‐affiliated NICU, one of three tertiary referral centres for neonatal surgery in Hong Kong.

• Population: inclusion criteria included Cbil 34 μmol/L or greater (2 mg/dL), likely to require PN for more than 2 weeks and informed parental consent. The exclusion criteria were major congenital or lethal chromosomal abnormalities, multiorgan failure or imminent death, and cholestatic jaundice secondary to other known causes, for example congenital or acquired TORCH (acronym for a group of diseases that cause congenital conditions; Toxoplasmosis, Other (such as syphilis, varicella, mumps, parvovirus and HIV), Rubella, Cytomegalovirus, Herpes simplex), syphilis, hepatitis B or C infection, biliary atresia, or other intra‐ or extrahepatic diseases obstructing bile flow.

• Objectives: to evaluate whether pure F‐LE could halt or reverse the progression of PNAC compared with soy‐based parenteral lipid preparation (SLP) and to assess the effects of pure F‐LE on liver function and physical growth.

• Intervention: infants were randomly assigned to receive either pure F‐LE (10% Omegaven; Fresenius‐Kabi AG, Bad Homburg vor der Höhe, Germany) or S‐LE (10% Intralipid; Fresenius‐Kabi AG, Uppsala, Sweden). Pharmacists not involved in the care of the infants prepared the lipids and the clinical and research teams were unaware of the randomisation during the study period. Infants randomised to the pure F‐LE arm received a starting dose of 0.5 g/kg/day and gradually advanced to the maximum of 1.5 g/kg/day at 0.5 g/kg/day increments every two days.

• Outcomes: the primary outcome was reversal of PNAC, defined as Cbil less than 34 μmol/L within four months after commencement of lipid treatment. The secondary outcomes were rate of change of weekly liver function tests, infant growth parameters (namely head circumference and bodyweight), blood lipid profile and number of episodes of late‐onset infection.

Larsen 2012 was a double‐blind, randomised, controlled clinical trial performed at The Stollery Children's Hospital Edmonton, AB, Canada. Study enrolment: November 2005 to March 2007. Larsen 2015 (an additional reporting of Larsen 2012) reported on lymphocytes, fatty acids and procalcitonin levels. Relevant data were extracted from both study reports.

• Population: Infants scheduled to have open heart surgery who were appropriate‐for‐gestational‐age and had Apgar scores greater than 7 at five and 10 minutes, and no clinical or microbiological evidence of infection preoperatively, were included in the study. Infants were randomised to receive one of two intravenous lipid preparations: the control group received an emulsion composed exclusively of S‐LE (Intralipid; Clintec Nutrition Company, Baxter, Deerfield, IL, USA), while the treatment group received an emulsion containing 50% MCTs and 40% LCTs from soybean oil and 10% fish oil (MLF 541, Lipoplus; B Braun Melsungen, Melsungen, Germany). Exclusion criteria: infants receiving immunosuppressive or anti‐inflammatory drugs (e.g. indomethacin, aspirin).

• Objectives: investigated the effects of providing an LE supplemented with ω‐3 lipids and MCT on plasma lymphocytes and procalcitonin concentrations in critically ill infants following open heart surgery with cardiopulmonary bypass (CPB).

• Interventions: 32 infants undergoing CPB and dependent on PN were randomised to receive either S‐LE (control, n = 16) or a 50:40:10 mixture of MCT, soybean oil and fish oil (treatment, n = 16). PN was administered for three days preoperatively and 10 days postoperatively.

• Outcomes: Fatty acids (eicosapentanoic acid, arachidonic acid, linolenic acid, linoleic acid, docosahexanoic acid), lymphocytes, granulocyte macrophage‐colony stimulating factor (GM‐CSF) and inflammatory markers were quantified at baseline; before surgery; and days 1, 7 and 10 after surgery. The measured inflammatory markers included: leukotriene B4, tumour necrosis factor‐α, IL‐1β, IL‐2, IL‐4, IL‐5, IL‐6, IL‐8 and IL‐10 interferon‐gamma. The study also reported on the clinical outcomes including length of stay, duration of ventilation and sepsis incidence.

Lima 1988 was a single‐centre RCT performed at the Departments of Child Health and Medical Biochemistry, University of Wales College of Medicine, Cardiff, UK.

• Population: inclusion criteria were nil by mouth, total serum bilirubin less than 100 μmol/L and tolerating intravenous nutrition. Infants had a variety of illnesses such as respiratory problems requiring ventilation and major neonatal abdominal surgery. The population included infants with gestations as low as 25 weeks.

• Objectives: to investigate how parenteral MCT was tolerated by newborns requiring PN. The clinical and biochemical effects of a 50% MCT/50% LCT‐based LE were compared with conventional 100% LCT (S‐LE) emulsion in terms of weight gain, plasma triglycerides, cholesterol, ketones, free fatty acids and glucose.

• Interventions: 51 infants were enrolled: group 1 (n = 26) received 50% MCT/50% LCT‐based LE and group 2 (n = 25) received the conventional 100% LCT emulsion. All neonates received similar parenteral infusates, apart from the LE. Amino acid solution‐Vamin 9 (KabiVitrum) was started on the third day of life in an increasing dosage schedule up to 2.1 g/kg/day of amino acids. LE was administered on day four, either as Lipofundin MCT/LCT (B Braun Medical) or as Intralipid (KabiVitrum), at initial doses of 0.5 g/kg/day for infants of less than 1.5 kg birthweight and 1 g/kg/day for infants 1.5 kg birthweight or more. LE was increased daily by 0.5 g/kg up to a maximum of 3 g/kg/day. The LEs were infused by the same route using a 'Y' connection tube near the infant for 20 hours per day. Blood samples (1 mL) were collected at 9 a.m. before the daily lipid infusion was started. Samples were taken daily up to the seventh day of lipid administration and then weekly until the infants were discharged from the study.

• Outcomes: clinical and biochemical effects including weight gain, plasma triglycerides, cholesterol, ketones, FFA and glucose.

Nehra 2014 was a randomised, controlled, double‐blind clinical trial at Boston Children's Hospital, Boston, MA, USA. This trial is registered at ClinicalTrials.gov (NCT 00512629). Study enrolment was July 2007 to June 2009.

• Population: inclusion criteria: neonates and infants (less than three months of age) with baseline conjugated bilirubin less than 1.0 mg/dL and a gastrointestinal disease requiring surgical intervention who were expected to be PN dependent for 21 days or more. Exclusion criteria: baseline INR greater than 1.5 (greater than 2 if one week of age or less) or triglyceride level greater than 400 mg/dL and infants with a haemolytic disorder, liver disease or shock requiring vasopressor support, extracorporeal membrane oxygenation, nitric oxide, or a combination of these; infants who had undergone an intestinal lengthening procedure.

• Objectives: to assess the safety and efficacy of an intravenous pure F‐LE in reducing the incidence of cholestasis in neonates compared with the traditional S‐LE in infants requiring surgery for gastrointestinal disorders.

• Interventions: S‐LE (Intralipid; n = 10) versus pure F‐LE (Omegaven; n = 9). Infants received standard medical, surgical and nutrition support. Both LEs were provided at 1 g/kg/day, and this dose was kept constant throughout the study period. All infants remained on their assigned LE until weaned from PN unless crossed over to the other study arm. Infants with persistently elevated conjugated bilirubin (greater than 2 mg/dL for two or more continuous weeks) were considered treatment failures and were crossed over to the other study arm.

• Outcomes: to determine whether the incidence of cholestasis, defined as a serum conjugated bilirubin greater than 2 mg/dL for two or more consecutive weeks, differed between the S‐LE and pure F‐LE. Secondary outcomes evaluated the safety and tolerability of the pure F‐LE compared with the S‐LE. Safety evaluation of the pure F‐LE involved assessment of mortality and adverse event rate including infections, cardiopulmonary arrest and incorrect LE dosing. Additional safety evaluation included an analysis and comparison of growth parameters, bleeding complications, transfusion requirements, laboratory studies and neurodevelopmental outcomes. Neurodevelopmental outcome was assessed by Bayley Scales of Infant and Toddler Development, 3rd edition (BSID‐III), a series of motor, cognitive, and language scales, at 6 and 24 months' corrected age.

Pereira‐da‐Silva 2017 was a single‐centre, double‐blind, RCT that compared the incidence of cholestasis using either SMOFlipid or MCT/SOY in neonates born at 34 weeks or greater gestational age undergoing major surgery. The study was performed at the NICU, Hospital Dona Estefania, Lisbon, Portugal. Recruitment was from August 2011 to February 2014.

• Population: inclusion criteria were infants at 34 or more gestational weeks who underwent surgery for a major anomaly of the digestive tract or of a congenital anomaly affecting the digestive tract (e.g. diaphragmatic hernia). Neonates were recruited within the first 48 postnatal hours, if PN had been initiated. Exclusion criteria were pre‐existing hepato‐biliary disease and abnormal liver functions within the first 72 postnatal hours.

• Objectives: to examine the effects of two LE: either 30% soybean oil, 30% MCT, 25% olive oil and 15% fish oil (SMOFlipid) or 50% MCT and 50% soybean oil n‐6 (MCT/SOY; MS‐LE) on the incidence of cholestasis in surgical term and near‐term neonates.

• Intervention: 49 infants received SMOFlipid (n = 22) or MCT/SOY (n = 27). All participants were scheduled to receive an individualised PN within the first 24 postnatal hours, including amino acids, glucose, electrolyte and vitamin PN solution plus lipids (using SMOFlipid or MCT/SOY). As customised PN was not available during the weekends, infants admitted over weekends received a standard solution containing only glucose, calcium and amino acids. Parenteral lipid intake was reduced to 0.5 g/kg/day to 1.5 g/kg/day if there was hypertriglyceridaemia (greater than 250 mg/dL), hyperglycaemia (greater than 150 mg/dL), unconjugated bilirubin greater than 12 mg/dL, acute phase of sepsis or pulmonary hypertension. If cholestasis was diagnosed, the parenteral intakes were limited to no more than 2 g/kg/day to 2.5 g/kg/day of lipids, 2 g/kg/day to 2.5 g/kg/day of amino acids and 12 mg/kg/minute of glucose. The same enteral feeding protocol was used in both groups. Minimal enteral feeding (10 mL/kg daily for five days) was initiated.

• Outcomes: primary outcome was incidence of cholestasis, defined initially as Cbil greater than 1 mg/dL (17.1 mmol/L). After the trial initiation, the primary outcome was modified (i.e. Cbil greater than 1 mg/dL (17.1 mmol/L) if total bilirubin was less than 5 mg/dL (85.5 mmol/L) or a Cbil greater than 20% of the total bilirubin if this was greater than 5 mg/dL). Secondary outcome was initially set as the severity of cholestasis, evaluated by the magnitude of the conjugated hyperbilirubinaemia and GGT greater than 225 IU/L. After the trial initiation, the definition was modified to include total ALP greater than 608 IU/L; elevated serum ALT and aspartate aminotransferase (AST), defined as greater than 55 IU/L in girls or greater than 60 IU/L in boys, as markers of PNALD.

Excluded studies

We excluded five studies after discussion among the authors (VK, MM) (Calkins 2014; Calkins 2017; Nehra 2013; Pichler 2014; Webb 2008). The full details of these excluded studies are provided in the Characteristics of excluded studies table. The full PRISMA flow chart of the review is provided in Figure 1.

Allocation

All included studies were described as randomised. However, six studies adequately described the method of the random sequence generation and were at low risk of bias (Diamond 2017; Lam 2014; Larsen 2012; Lima 1988; Nehra 2014; Pereira‐da‐Silva 2017). Three studies did not describe the randomisation procedure and were at unclear risk of bias (Angsten 2002; Ariyawangso 2014; Magnusson 1997).

Five studies described allocation concealment using the pharmacy/central allocation and were at low risk of bias (Diamond 2017; Lam 2014; Larsen 2012; Nehra 2014; Pereira‐da‐Silva 2017). In three studies, there was insufficient information regarding concealment of allocation (Angsten 2002; Lima 1988; Magnusson 1997). One study was at high risk for allocation concealment (Ariyawangso 2014). This was an open‐label study with blocked randomised design and authors did not describe measures to conceal allocation. Blocked randomised designs can be at particular risk that the blocks can be guessed.

Blinding

Five studies adequately described the blinding of the intervention and control LE as being identical (low risk; Diamond 2017; Lam 2014; Larsen 2012; Nehra 2014; Pereira‐da‐Silva 2017). Three studies were described as blinded but did not report how blinding was achieved (unclear risk; Angsten 2002; Lima 1988; Magnusson 1997). Two studies were assigned low risk for detection bias by author consensus as the reported outcomes were objective (Angsten 2002; Magnusson 1997). One study was described as an open‐label study and was assigned unclear risk of bias by author consensus as the reported outcomes were objective (Ariyawangso 2014).

Incomplete outcome data

Seven studies were at low risk of attrition bias or incomplete outcome data (Angsten 2002; Diamond 2017; Lam 2014; Larsen 2012; Lima 1988; Magnusson 1997; Pereira‐da‐Silva 2017). One study reported clinical outcomes for most infants but secondary outcome of neurodevelopment outcome had loss to follow‐up; however, the data for neurodevelopment were not used in the meta‐analysis (unclear risk; Nehra 2014). One study did not include data of all infants for the outcomes of liver enzymes (high risk; Ariyawangso 2014). Ariyawangso 2014 reported data on 24/41 (57%) enrolled infants. The study report mentioned that 18 neonates (n = 8 in the control group and n = 10 in the experimental group) had recovered by day 22 (the day of outcome investigation) and only 24 surgical infants remained in the study.

Selective reporting

Four studies provided trial registration information with published study details (Diamond 2017; Lam 2014; Larsen 2012; Nehra 2014) and were assigned low risk of bias for selective reporting. Trial registration information for one of the studies was dated after the trial completion and pre‐specified protocol or study details were not available to us. Therefore this study (Pereira‐da‐Silva 2017) was assigned unclear risk of bias for selective reporting. Reporting bias was unclear for the other four studies where protocols were not available. Some studies mainly reported biochemical outcomes (Angsten 2002; Larsen 2012; Magnusson 1997).

Other potential sources of bias

Three out of five studies that contributed data to the various outcomes of the review were terminated early and the risk of bias was not clear though it did further decrease the number of infants in the outcomes. The reasons for the study terminations included: non‐availability of LE preparation (Pereira‐da‐Silva 2017), parents refusing participation or agreeing for randomisation (Lam 2014), and low prevalence of cholestasis in the cohort (Nehra 2014).

Term and late preterm infants without PNALD or surgical conditions

No studies compared LE in term and late preterm infants without underlying cholestasis or surgical conditions.

Term and late preterm infants with surgical conditions

There were seven studies performed in infants with surgical conditions. Of these, four studies compared fish oil LE versus non‐fish oil LE (Ariyawangso 2014; Larsen 2012; Nehra 2014; Pereira‐da‐Silva 2017), and three studies compared alternative‐LE versus S‐LE (Angsten 2002; Lima 1988; Magnusson 1997).

Term and late preterm infants with cholestasis/PNALD

Infants with cholestasis was a distinct population group considered in this review. There were two eligible studies (n = 40) in infants with cholestasis with both studies comparing fish oil‐containing LE versus S‐LE (Diamond 2017; Lam 2014). No study compared F‐LE versus another F‐LE, alternative‐LE versus S‐LE or one alternative‐LE versus another alternative‐LE.

Fish oil LE versus non‐fish oil LE in infants with cholestasis (Comparison 2)

Resolution of PNALD/cholestasis (conjugated bilirubin less than 2 mg/dL) (outcome 2.1)

We predefined reversal of cholestasis in the review protocol as conjugated bilirubin less than 2 mg/dL due to the effect of intervention (Kapoor 2018). One randomised study used definition of reversal of cholestasis as conjugated bilirubin less than 2 mg/dL (Lam 2014). This study found that the cholestasis had resolved in most infants by the trial end point of four months. However, most infants in this study in the S‐LE group improved after they were on full enteral intake.

This study also described proportions of infants in both arms who had resolution of cholestasis while on trial PN which was considered for this outcome. There was no statistically significant difference between 10% pure F‐LE and 10% S‐LE (RR 5.60, 95% CI 0.34 to 93.95; typical RD 0.33, 95% CI –0.01 to 0.67; n = 16; Analysis 2.1; very low‐quality evidence).

PNALD/cholestasis (any definition) (outcome 2.2)

Two studies (n = 40) reported data on cholestasis (Analysis 2.2; Diamond 2017; Lam 2014). Both studies reported on infants with cholestasis at the end of PN or study end, though the primary outcomes in both studies were different. Diamond 2017 included infants with early hepatic dysfunction (Cbil ≥ 17 µmol/L up to 50 µmol/L) on two consecutive readings over seven days. This study looked at prevention of progression of PNALD in infants with cholestasis and provided data for infants whose conjugated bilirubin level was greater than 50 µmol/L at the study primary end point (Cbil in the week the infant received the last dose of PN, i.e. at 12 weeks, at full enteral tolerance or on development of progressive liver disease).

Lam 2014 included infants with cholestasis defined as conjugated bilirubin ≥ 2 mg/dL. The primary outcome for study by Lam and colleagues was the reversal of PNALD defined as Cbil level less than 34 µmol/L (2 mg/dL) within 4 months of the commencement of lipid treatment. This study also provided information on proportion of infants that recovered from PNAC (or remained cholestatic) while receiving PN.

In the subgroup comparison between MOFS‐LE and S‐LE, one study reported statistically significantly lesser cholestasis (conjugated bilirubin greater than 50 µmol/L) in the MOFS‐LE group compared to the S‐LE group (typical RR 0.39, 95% CI 0.14 to 1.10; typical RD –0.42, 95% CI –0.78 to –0.06; n = 24; Diamond 2017).

In the subgroup comparison between pure F‐LE and S‐LE, one study reported data with no statistically significant difference between groups (RR 0.69, 95% CI 0.43 to 1.13; RD –0.33, 95% CI –0.67 to 0.01; n = 16; Lam 2014).

In the meta‐analysis of both subgroups, there was statistically significantly less cholestasis in the F‐LE group compared to S‐LE group (typical RR 0.54, 95% CI 0.32 to 0.91; typical RD –0.39, 95% CI –0.65 to –0.12; 2 studies; n = 40; very low‐quality evidence). There was low heterogeneity for RR (24%) and no heterogeneity between the studies for RD (0%). There was no heterogeneity in the test for subgroup differences for RR or RD (I² = 0%) (Figure 3).

Figure 3

---

---

Forest plot of comparison: 2 Fish oil LE versus non‐fish oil LE in infants with cholestasis, outcome: 2.2 PNALD/cholestasis (any definition).

However, the very low number of participants in this outcome, with different Cbil cut‐offs in two included studies, one of which was terminated early, introduced further uncertainly about the outcomes.

Time to resolution of PNALD (outcome 2.3)

In the subgroup comparison between pure F‐LE and S‐LE, one study reported data with no statistically significant difference between groups (MD –27.00 days, 95% CI –70.91 to 16.91; n = 16; Analysis 2.3; Lam 2014).

Growth rate (outcome 2.4)

In the subgroup comparison between pure F‐LE and S‐LE, one study reported data with statistically significant effect in favour of pure F‐LE compared to S‐LE (MD 45.00 g/week, 95% CI 15.00 to 75.00; n = 16; Analysis 2.4; very low‐quality evidence; Lam 2014). It is worth noting that Lam 2014 used 10% Intralipid (S‐LE) preparation which is no longer recommended.

Head growth velocity (outcome 2.5)

In the subgroup comparison between pure F‐LE and S‐LE, one study reported data with no statistically significant difference between groups. The study authors reported the data as mean and 95% CIs and we imputed the SD using a normal distribution (rather than t distribution) to get similar results to the study authors (MD 0.16 cm/week, 95% CI –0.01 to 0.33; n = 16; P = 0.06; Analysis 2.5; Lam 2014).

Any sepsis (clinical or culture positive (or both)) (outcome 2.6)

Two studies (n = 40) reported data in a format that could be used for the meta‐analysis and specific definitions for sepsis were not provided in the study reports (Analysis 2.6; Diamond 2017; Lam 2014).

In the subgroup comparison between MOFS‐LE and S‐LE, one study reported data with no statistically significant difference between groups (typical RR 1.48, 95% CI 0.52 to 4.18; typical RD 0.15, 95% CI –0.24 to 0.53; n = 24; Diamond 2017).

In the subgroup comparison between pure F‐LE and S‐LE, one study reported data with no statistically significant difference between groups (typical RR 0.78, 95% CI 0.14 to 4.23; typical RD –0.06, 95% CI –0.49 to 0.37; n = 16; Lam 2014).

In the meta‐analysis of all subgroups, there was no statistically significant difference between groups in the meta‐analysis (typical RR 1.21, 95% CI 0.50 to 2.92; typical RD 0.06, 95% CI –0.23 to 0.35; 2 studies; n = 40; very low‐quality evidence). There was no heterogeneity among the subgroups for RR or RD (I² = 0%).

Hypertriglyceridaemia (outcome 2.7)

In the subgroup comparison between MOFS‐LE and S‐LE, one study reported data for this outcome. The report mentioned hyperlipidaemia. No definition was provided. There was no statistically significant difference between groups (typical RR 0.79, 95% CI 0.30 to 2.09; typical RD –0.1, 95% CI –0.49 to 0.29; n = 24; very low‐quality evidence; Analysis 2.8; Diamond 2017).

Hyperglycaemia (outcome 2.8)

One study reported data on hyperglycaemia with no statistically significant difference between groups. No definition was provided (typical RR 1.48, 95% CI 0.52 to 4.18; typical RD 0.15, 95% CI –0.24 to 0.53; n = 24; Analysis 2.8; Diamond 2017).

Conjugated bilirubin levels (outcome 2.9)

In the subgroup comparison between MOFS‐LE and S‐LE, one study reported data with statistically significant effect in favour of MOFS‐LE with lower Cbil values compared to S‐LE excluding an outlier in the data who had significant increase in the level of conjugated bilirubin with sepsis (MD –47.00 µmol/L, 95% CI –71.65, –22.35; n = 24; Analysis 2.9; low‐quality evidence; Diamond 2017). The study authors reported data in mean and 95% CI (as confirmed by the study author) for the distribution excluding the outlier. There was no statistically significant difference between groups when the outlier was included in the analysis. Study authors also performed analysis by including the conjugated bilirubin value for the outlying participant when the participant had improved and this analysis showed a statistically significant difference between groups.

In their study report the authors defined cholestasis as increased conjugated bilirubin levels not related to sepsis. Therefore, we presented the data without the outlier and described the analytical aspects reported by the study.

Gamma‐glutamyltransferase levels (outcome 2.10)

In the subgroup comparison between MOFS‐LE and S‐LE, one study reported data with statistically significant effect in favour of S‐LE (MD 115.00 IU/L, 95% CI 12.86 to 217.14; n = 24; Analysis 2.10; Diamond 2017). Diamond and colleagues interestingly found significantly higher GGT despite a trend towards lower ALP in the SMOFlipid group which the study authors could not explain.

Alanine aminotransferase levels (outcome 2.11)

In the subgroup comparison between MOFS‐LE and S‐LE, one study (Diamond 2017) (MOFS‐LE versus S‐LE; n = 24) reported data with no statistically significant difference between groups (MD –36.00 IU/L, 95% CI –155.41 to 83.41; n = 24; Analysis 2.11; Diamond 2017).

Alkaline phosphatase levels (outcome 2.12)

In the subgroup comparison between MOFS‐LE and S‐LE, one study reported data with no statistically significant difference between groups (MD –119.00 IU/L, 95% CI –240.01 to 2.01; n = 24; Analysis 2.12; Diamond 2017).

Triglyceride levels (outcome 2.13)

In the subgroup comparison between pure F‐LE and S‐LE, one study reported data with no statistically significant difference between groups (MD –0.40 mmol/L, 95% CI –0.85 to 0.05; n = 16; Analysis 2.13; Lam 2014). Test of heterogeneity was not applicable in the subgroup. No meta‐analysis was possible.

Rate of change of alanine aminotransferase (outcome 2.14)

In the subgroup comparison between pure F‐LE and S‐LE, one study reported statistically significantly lower rate of increase of ALT in the pure F‐LE group (1.1 IU/L/week, 95% CI –5.2 to 7.5 versus 9.1 IU/L/week, 95% CI 4.1 to 14.1; P = 0.02; MD –8.00 IU/L/week, 95% CI –14.79 to –1.21; n = 16; Analysis 2.14; Lam 2014).

Rate of change of conjugated bilirubin (outcome 2.15)

In the subgroup comparison between pure F‐LE and S‐LE, one study reported statistically significantly lower rate of increase of conjugated bilirubin in the pure F‐LE group (0.6 µmol/L/week, 95% CI ‐9.5 to 10.8 with pure F‐LE versus 13.5 μmol/L/week, 95% CI 5.4 to 21.6 with S‐LE; P = 0.03; MD –12.9 μmol/L/week, 95% CI –23.69 to –2.11; n = 16; very low‐quality evidence; Analysis 2.15; Lam 2014).

Death before discharge (outcome 2.16)

Two studies (n = 40) reported data with no study individually reporting any significant difference between the two groups (Analysis 2.16; Diamond 2017; Lam 2014). However, all the deaths in both studies were complicated or due to progressive liver disease.

In the subgroup comparison between MOFS‐LE and S‐LE, one study reported data with no statistically significant difference between groups (typical RR 0.39, 95% CI 0.02 to 8.69; typical RD –0.08, 95% CI –0.27 to 0.12; n = 24; Diamond 2017).

In the subgroup comparison between pure F‐LE and S‐LE, one study reported data with no statistically significant difference between groups (typical RR 0.16, 95% CI 0.01 to 2.88; typical RD –0.29, 95% CI –0.63 to 0.06; n = 16; Lam 2014).

In the meta‐analysis of all subgroups, there was no statistically significant difference between groups in the meta‐analysis (typical RR 0.24, 95% CI 0.03 to 1.87; typical RD –0.16, 95% CI –0.36 to 0.04; 2 studies; n = 40; very low‐quality evidence). There was no heterogeneity among the subgroups for RR or RD (I² = 0%) or for subgroup differences in RD (4.2%).

No study reported on outcomes of body composition, jaundice requiring treatment, phototherapy duration, duration of supplemental oxygen, home oxygen therapy, thrombocytopenia, time to development of PNALD, EFA deficiency or need for liver transplantation in infants with surgical conditions or infants with PNALD/cholestasis.