Najm 2017.

Methods	Design: single‐centre blinded RCT in infants < 28 weeks' gestation Setting: NICU of Sahlgrenska University Hospital in Gothenburg, Sweden Study enrolment: 4 April 2013 to 22 September 2015 I. Allocation concealment: yes II. Blinding of intervention: no III. Blinding of outcome measurement(s): yes (for data analysts and screening ophthalmologists) IV. Complete follow‐up: yes
Participants	Inclusion criteria: 90 infants < 28 weeks' gestation admitted to NICU Exclusion criteria: major congenital malformations
Interventions	Group 1 (n = 41): MOFS‐LE; 20% SMOFlipid Group 2 (n = 37): OS‐LE; 20% ClinOleic PN was initiated as soon as possible after birth with a standard solution containing Vaminolac and 10% glucose (total protein content 2 g/100 mL) aiming at 80–90 mL/kg/day of the resulting solution during the first 24 hours. Lipid solution (ClinOleic or SMOFLipid) was normally started 6–12 hours after birth at a rate of 1 g/kg/day with daily increases up to 2 g/kg/day. Enteral nutrition used either maternal or donor breast milk with individualised fortification based on results from breast milk analysis using a commercial bovine milk fortifier. Daily intakes of FAs arachidonic acid, EPA and DHA (mg/kg/day) were prospectively registered from birth during the first 2 weeks of life. The parenteral lipid dosing strategy was to deliver 2 to 3 g/kg bodyweight every 24 hours. The FA compositions of SMOFlipid and ClinOleic were analysed by gas chromatography
Outcomes	Primary outcome: ROP classified as no ROP or ROP stage 1, 2, 3, or 3+. Other outcomes: serum LC‐PUFA (DHA, EPA and arachidonic acid) profiles, ROP, BPD, NEC, PDA, sepsis and growth between birth and 36 weeks. Cbil blood level of > 50 μmol/L for ≥ 2 weeks at any time during the follow‐up, unrelated to sepsis was consider significant. Data presented for 78 infants. 12 infants died. The cause of death (and thereby partial outcomes) were presented for the deceased infants in the supplementary table.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Randomization was in blocks of 20 infants, adjusting for gestational age." Comment: low risk (by author consensus (VK, MM))
Allocation concealment (selection bias)	Low risk	Quote: "The treating nurse/doctor received the randomisation online." Comment: possibly done.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Quote: "The type of lipid emulsion given was not blinded." However, blinding was done for the analysts and the screening ophthalmologists. Comment: the LE were not blinded, which could potentially introduce possibility of performance bias. However, most of the outcomes were objective based on study methodology for blood sampling at predefined time points. The review authors (by consensus; MM, VK) considered that the risk of performance bias affecting the study was unclear.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Type of lipid emulsion was blinded for data analysis and the screening ophthalmologists." Comment: blinding was reported for ROP. We did not consider that there would be risk of material bias for most objective outcomes including cholestasis, sepsis, etc.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: data were only presented for 78 infants who survived the study period with the gestation. Cause and the day of death of deceased infants (6 in each arm) was presented in the supplementary table from which additional data were extracted but not used for primary meta‐analysis. A sensitivity analysis was performed including and excluding the data of sepsis and NEC from the reported causes of death which did not change the results of these outcomes. The deceased infants would have died prior to the ROP screening (primary outcome) could be done.
Selective reporting (reporting bias)	Unclear risk	Quote: "The study was approved by the Regional Ethical Board, Gothenburg (Dnr 303‐11) (Clinical trial NCT02760472)." Comments: the protocol was published on the ClinicalTrials.gov website after the trial had stopped recruiting according to the information from the ClinicalTrials.gov website. The first record on the ClinicalTrials.gov was on 30 March 2016 whereas the trial stopped on 22 September 2015 as per the authors. The fact that the protocol was published on clinicaltrials.gov after the study completed makes assessment for the selective reporting difficult.
Other bias	Low risk	None detected.