Enteral nutrition: a hard look at some soft evidence.

Those who read the medical literature should understand the principles of evidence-based medicine. Even randomized trials can contain design or interpretative flaws that allow bias to produce, or exaggerate the size of, beneficial effects. Such problems beset the literature of enteral nutrition (EN). Investigators who have compared EN with parenteral nutrition (PN) have alleged that EN produces fewer adverse events, but such studies do not assess the absolute value of either therapy, and data exist suggesting that PN causes net harm. Trials comparing EN with no nutrition therapy have not yielded convincing evidence of efficacy because the study designs have failed to use methods to prevent bias from interfering with the observations. This same problem exists with trials that have assessed volitional feeding programs (eg, oral supplements). Thus, although systematic reviews have alleged that EN benefits patients undergoing surgery, patients in the critical care unit, patients with liver disease, and patients with pancreatitis, the presence of bias limits any positive conclusions. As a manifestation of this issue, when the various trials are separated into studies with high and low risks of bias, those with low risks have not shown any benefit. EN has been accepted and implemented despite the lack of convincing scientific support of efficacy.