Darmstadt 2005.

Study characteristics
Methods	Randomised controlled trial.
Participants	Preterm infants of gestational age at birth < 33 weeks and aged < 72 hours. Exclusions: Infants considered likely to die within 48 hours, infants with major congenital anomalies, infants requiring major surgery, infants with established skin infections.
Interventions	Massage (whole body apart from face and head) thrice daily for 14 days, then twice daily until discharge from hospital with: 1. Sunflower seed oil: N = 159 2. Aquaphor: N = 157 3. No emollient (control): N = 181
Outcomes	Invasive infection. Mortality.
Notes	Setting: Special Care Nursery, Dhaka Shishu Hospital, Bangladesh (1998 to 2003).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Manual generation of blocks of six with two assignments per block for all three of the groups, then computer‐generated sequence
Allocation concealment (selection bias)	Low risk	Sealed opaque envelopes
Blinding of participants and personnel (performance bias) All outcomes	High risk	Unmasked parents and caregivers
Blinding of outcome assessment (detection bias) All outcomes	High risk	Unmasked clinicians and investigators
Incomplete outcome data (attrition bias) All outcomes	Low risk	Near‐complete follow‐up assessment
Selective reporting (reporting bias)	Low risk	Unlikely
Other bias	Unclear risk	The report states that infants were randomised within "blocks of six with two assignments per block for all three of the groups". However, this process appears to be inconsistent with the allocated distribution of infants: 1. sunflower seed oil (N = 159), or 2. Aquaphor (N = 157), or 3. no emollient (control) (N = 181). The principal investigator of the trial has explained that this inconsistency may be due to two possible factors: 1. The trial originally had a fourth arm (emollient therapy with safflower oil) which was deemed unacceptable to parents and caregivers and removed from the trial when 24 infants had been randomised to the arms. These infants were not included in the final analyses. 2. The allocation sequence generation method was changed from the original manual process (selecting from a block of six envelopes) to a computer‐generated sequence that did not maintain the balance of assignments within blocks.