Borzotta 1994.

Methods	RCT, single‐centre, 2‐arm, parallel design
Participants	Total number of randomized participants: 59 Inclusion criteria Adults, 18 to 60 years of age, with head injuries and a GCS ≤ 8, coma persisting > 24 hours Exclusion criteria Spinal cord injury Pre‐existing metabolic disorders Renal failure Inflammatory bowel disease Neurological prognosis of rapidly fatal injury Primary diagnosis Severe head injury Baseline characteristics EN group Age, mean (SD): 26.2 (± 10.4) years Gender, M/F: 21/7 APACHE II, mean (SD): 15.7 (± 3.5) PN group Age, mean (SD): 28.9 (± 10) years Gender, M/F: 19/2 APACHE II, mean (SD): 14.9 (± 3.9) Country: USA Setting: level 1 trauma centre
Interventions	EN group n = 36; 8 losses; 28 analysed. Per‐protocol analysis (except for mortality) Details: jejunal tube placement, and gastrotomy tubes placed to drain stomach. Feeding started within 24 hours of randomization. Target rates calculated with Harris Benedict formula BEE + 50%. Initiated at 20% of target rate for 12 hours; 40% for 12 hours; 60% for 12 hours; 80% for 12 hours; then target rate. Formula was a Vivonex solution TEN (Norwich Eaton Pharmaceuticals, Inc, Norwich, NY, USA) consisting of 4.9 g/L glutamine, carbohydrates, fat, AAs, other minerals, and Travasol solution (Baxter Healthcare Corp, Deerfield, IL, USA) consisting of 3.21 g/L glutamine, carbohydrates, fat, AAs, other minerals. If extra protein was required then Travasol 10% was given to EN solution. At day 9 to 11, EN group converted from Vivonex TEN to Isotein HN via jejunal tube ("thus keeping both groups identical except for route") PN group n = 23; 2 losses; 21 analysed. Per‐protocol analysis (except for mortality) Details: central venous catheter placement. Feeding continued for 5 days and then attempts to convert to gastric feeding by any routes at the discretion of the clinician. Target rates calculated with Harris Benedict formula BEE + 50%. Feeding initiated 40% target rate of 24 hours; 60% for 12 hours, 80% for 12 hours, then target rate. Formula was an Isotein HN solution (Sandoz Nutrition Corp: Minneapolis, MN, USA) consisting of carbohydrates, fat, AAs, and other minerals. If extra protein was required then Travasol 10% was given to PN solution
Outcomes	Infections Nutrition‐related complications including hyperglycaemia (diagnosed by blood glucose level of > 180 mg/dL) and diarrhoea Mortality
Notes	Funding/declarations of interest: financial support from Norwich Eaton Pharmaceuticals, NY, USA. Study dates: July 1990 to December 1991
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Adequate sequence generation. Computer‐generated random number tables
Allocation concealment (selection bias)	Low risk	Computer‐generated numbers and we assumed that allocation was concealed from investigators
Blinding of participants and personnel (performance bias) All outcomes	High risk	No details and we assumed investigators made no attempts to blind personnel
Blinding of outcome assessment (detection bias) All outcomes (except mortality)	Unclear risk	No details
Blinding of outcome assessment (detection bias) Mortality	Low risk	No details. Lack of blinding unlikely to influence outcome data for mortality
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Some loss of participant data: 2 participants in EN group, 8 participants in PN group. Reasons for losses were explained and mortality data included these (death was one of reasons for loss)
Selective reporting (reporting bias)	Unclear risk	Clinical trials registration or prospectively published protocol was not reported. Therefore, not feasible to make judgement on selective outcome reporting bias
Baseline characteristics	Low risk	Appeared comparable, although we noted a large difference in sample size between groups which was unexplained
Other bias	Unclear risk	Some possible differences in nutritional formula. Unclear if this was likely to influence data