An observational trial by Jim Kutsoguiannis, Cathy Alberda, Daren Heyland et al published in the December 2011 Critical Care Medicine showed no difference in 60-day survival among critically ill, mechanically ventilated patients started on parenteral nutrition (TPN) "early" (within 48 hours of ICU admission) or "late" (after 48 hours).
Only 258 patients were included in the early/late TPN arms (another 2,562 were able to take enteral nutrition; not surprisingly, they did a lot better), so it's hard to conclude much.
One very interesting finding was that the enteral nutrition group had the lowest percentage of patients attaining "adequacy" of calories and protein as determined by a dietician/nutritionist -- with an absolute 20% fewer attaining "adequacy" of nutrition. Yet those patients had an absolute ~7% higher chance of being discharged home alive within 60 days (~28% vs. ~35% 60-day mortality).
A 2011 randomized trial in NEJM suggested a strategy of providing sugar water and liquid vitamins and minerals for 7 days before starting TPN was slightly better than providing "early" TPN to critically ill well-nourished patients who could not tolerate enteric feedings in sufficient amounts. Few U.S. ICUs use the liquid vitamin preparation they used in that European trial, so (if the vitamin water did anything) the direct extrapolation of its findings to U.S. ICUs is debatable.
The EDEN trial recently published in JAMA by the ARDSNet investigators, could detect no difference in survival or other clinical outcomes among 1,000 well-nourished patients on mechanical ventilation for acute lung injury/ARDS who received either trophic (trickle) enteric feedings or full enteric feedings for 7 days.
All of which suggests that in most critically patients who are not malnourished, the nutrition strategy chosen has little effect on clinical outcome, and any reasonable attempt to provide adequate nutrition is okay.
Kutsogiannis J et al. Early use of supplemental parenteral nutrition in critically ill patients: Results of an international multicenter observational study. Crit Care Med 2011;39:2691-2699.