This study describes the advantages and disadvantages of several forms of enteral nutrition for patients with severe head injury (Glasgow Coma Scale Score [GCS], <12). Included in the study are nasoenteric nutrition delivery using blind, endoscopic, percutaneous endoscopic gastrostomy (PEG) and PEG with jejeunostomy (PEG/J), and open jejeunostomy tube placement methods. These methods are compared with parenteral delivery of nutrition. The study constituted a retrospective analysis of the success rate of early enteral feedings by blind, endoscopic PEG and PEG/J and by open jejeunostomy placement of small-bowel feeding tubes for 57 patients with severe head injury. The delivery cost of enteral nutrition per intensive care unit day was compared to the delivery cost of parenteral nutrition per intensive care unit day in the same group of patients. Fifty-three percent of patients were adequately maintained nutritionally with nasoenteric delivery alone and did not require parenteral feeding. The average number of days for initiation of either enteral or parenteral feedings was 1.8 +/- 0.2 days from injury [standard error of mean (SEM); range, 0-10 days]. An average of 3.3 days (range, 0-23 days) was required for feeding tube placement in all patients. For 70% of patients, tube placement was completed within 48 h after injury. Full-strength, full-rate enteral feedings were achieved by a mean of 4.9 days after injury. A total of 128 feeding tubes were placed while the patients were in the intensive care unit (ICU; 2.2 +/- 0.2 tubes per patient). Blind placement of feeding tubes into the small bowel was rarely achieved without repositioning. Endoscopic tube placement into the duodenum was achieved in 50% of patients, into the jejunum for 33% of patients, and into the stomach for 18% of patients. While in the intensive care unit, patients received an average of 77 +/- 2% of their measured energy expenditure (range, 57-114%). Eleven percent of patients experienced severe gastrointestinal problems. Other problems were associated with the inability to achieve or maintain access: dislodged tubes (30%), clogged or kinked tubes (21%), and mechanical access problems (7 %). Seventy-one percent of patients in barbiturate coma were able to tolerate early nasoenteric feedings. Aspiration pneumonitis occurred equally among patients fed nasogastrically and those fed nasoenterically. The overall aspiration rate was 14%. The cost of acute enteral feeding was $170 per day and that for parenteral feeding, $308 per day. We conclude that blind transpyloric feeding tube placement is difficult to achieve in patients with severe head injury; endoscopically guided placement is a better option. Endoscopic feeding tube placement most consistently allows for early enteral nutritional support in severe head injured patients. Limitations include the inability to establish and/or maintain enteral access, increased intracranial pressure, unstable cervical spinal injuries, facial fractures, and dedication of the physician to tube placement and monitoring.