Repa 2018.

Methods	Design: prospective, double‐blind RCT in ELBW infants Setting: level IV NICU of University Children's Hospital Vienna, Medical University of Vienna, Vienna, Austria Study enrolment: July 2012 to July 2015. Last follow‐up October 2015 I. Allocation concealment: yes II. Blinding of intervention: yes III. Blinding of outcome measurement(s): yes IV. Complete follow‐up: yes Both ITT and per‐protocol analyses were performed.
Participants	Inclusion criteria: ELBW < 1000 g admitted before 24 hours Exclusion criteria: cholestasis (Cbil > 1.5 mg/dL (25 mol/L)) before intervention, and higher‐order multiples infants with conditions associated with cholestasis independent of PN (i.e. infection with cytomegalovirus, HIV, hepatitis B or C, rhesus‐mediated haemolysis, cystic fibrosis, inborn errors of metabolism or primary liver diseases) were not eligible or excluded post‐randomisation.
Interventions	Group 1 (n = 110): MOFS‐LE; 20% SMOFlipid Group 2 (n = 113): S‐LE; 20% Intralipid Participants received full PN from birth using S‐LE (1 g/kg/day) were switched to study lipids after enrolment. Lipids were dosed up to 3 g/kg/day at the discretion of the attending physicians and reduced in relation to enteral nutrition (increased up to 20 mL/kg/day). Serum TG were measured at least weekly. Lipids were halted for 24 hours at TG levels > 400 mg/dL (4.5 mmol/ L) or downtitrated > 250 mg/dL (2.8 mmol/L). PN was stopped at 140–160 mL/kg/day of enteral feeds. Therapy adherence was calculated as the percentage study lipids were correctly provided; > 80% was considered highly adherent. Urodeoxycholic acid was administered to infants who developed cholestasis. Parenteral fish oil (Omegaven; Fresenius Kabi, Bad Homburg vor der Höhe, Germany) was permitted as rescue therapy (1 g/ kg/day) if Cbil was > 6 mg/dL (100 mol/L). Infants were followed until their 44th week' postmenstrual age, discharge or transfer to another hospital. All infants received probiotics and lactoferrin. Enteral feeds were provided every 3 hours; the median volume of a single feed per kg in the first week of life was calculated. For growth analysis (anthropometry with z score difference from birth to discharge), only survivors were analysed to avoid distortion of measurements by perimortal oedema.
Outcomes	Primary outcome: PNAC (Cbil > 1.5 mg/dL (25 mol/L) at 2 consecutive measurements). Peak levels of liver enzymes (ALT, AST,GGT, ALP during hospitalisation were identified. Blood sampling was performed weekly as long as PN was required and then every 7 to 14 days. Secondary outcomes: neonatal morbidities (death, duration of hospitalisation, ROP (any), and highest stage requiring treatment (severe ROP), culture‐confirmed sepsis, IVH III/IV, cystic PLVL, NEC ≥ IIa, focal intestinal perforation, abdominal surgery, days on mechanical ventilation, CLD, PDA requiring treatment, number of ibuprofen cycles or requiring surgical ligation, pulmonary hypertension, iNO/sildenafil treatment.
Notes	PNAC incidence of only 15.9% in the current trial, attributable to an accelerated weaning from PN compared with the planning phase (10 days). This shorter time on PN was an important study limitation and possibly related to the implementation of probiotics at the NICU before the start of the trial in 2010 and their preventive effect against NEC.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "… Participants were randomised using permuted blocks (ratio 1:1, block size of 4) and stratified according to sex and birth weight (< 750 vs ≥ 750 g) using a software, prepared by an independent statistician." Comment: done.
Allocation concealment (selection bias)	Low risk	Quote: "… an independent statistician, who kept the randomisation sequence concealed until the end of the study Comment: probably done, with author consensus
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "… Participants, healthcare providers, data collectors, and outcome adjudicators were blinded. A blinding team uninvolved in clinical decisions established the blinding code and masked the glass containers using opaque labels designated "Lipid A" or "Lipid B." Labels were resistant to detachment, in particular by 70% alcohol used during aseptic preparation. Neonatal nurses who prepared the study lipids for PN were part of the blinding team. Discarded containers were controlled for blinding integrity." Comment: well described
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "… Participants, healthcare providers, data collectors, and outcome adjudicators were blinded." Comment: well described in the study report
Incomplete outcome data (attrition bias) All outcomes	Low risk	Clinical outcomes were reported for all infants.
Selective reporting (reporting bias)	Low risk	The study protocol was available: NCT1585935: the review of the protocol indicated that the authors had not published the details of the secondary outcome measures of amplitude integrated amplitude integrated electroencephalography and visual evoked potentials at this stage. The review authors (VK, MM) were in consensus that risk of bias for selective reporting is low, as these secondary study outcomes were not a part of the current review's outcomes.
Other bias	Low risk	Not identified