This review investigated whether early enteral nutrition offers any advantages over early parenteral nutrition in hospitalised patients. No difference was found between enteral nutrition and parenteral nutrition for hospital mortality, although parenteral nutrition is associated with an increased risk of complications. Despite the risk of publication and language bias, the conclusions are likely to be reliable.

To assess whether early enteral nutrition (EN) offered any advantages over early parenteral nutrition (PN) in hospitalised patients.

PubMed and MEDLINE were searched from 1966 to 2002; the search terms were reported. The reference lists of identified articles were checked and relevant journals were handsearched. Non-English language articles were excluded.

Study designs of evaluations included in the review

Randomised controlled trials (RCTs) were eligible for inclusion.

Specific interventions included in the review

Studies comparing early EN with early PN, defined as the institution of supplemental nutrition within 96 hours of hospitalisation, intensive care admission or surgery, were eligible for inclusion. Studies in which patients were treated with immunonutrition or glutamine were excluded.

Participants included in the review

Studies of adult hospitalised patients were eligible for inclusion. The included studies involved adults hospitalised for inflammatory bowel disease, head or neck cancer, transplant, trauma, head injury, pancreatitis, gastrointestinal tract surgery, or malnourishment.

Outcomes assessed in the review

Studies that reported hospital mortality and/or complications or hospital length of stay were eligible for inclusion.

How were decisions on the relevance of primary studies made?

The authors did not state how the papers were selected for the review, or how many reviewers performed the selection.

Study quality was assessed using standard methods for RCTs: an assessment of the method of randomisation; inclusion and exclusion criteria; the use of intention-to-treat analysis; and the reporting of complications, cointerventions and follow-up details. Quality scores were not derived. Three investigators assessed study quality in an unblinded fashion.

Two investigators extracted the data onto standardised extraction forms, and any disagreements were resolved by consensus. Mortality and complications were recorded as the number of patients experiencing the event, and the risk difference (RD) between EN and PN was derived. Length of stay was recorded as the mean with standard deviation, and the weighted mean difference between EN and PN was derived.

How were the studies combined?

RDs and weighted mean differences between EN and PN were pooled using both fixed-effect (Mantel-Haenszel) and random-effects (DerSimonian and Laird) models, although only the random-effects results were reported. A cumulative meta-analysis was performed to study possible time trends. Publication bias was assessed using funnel plots, Begg's rank correlation test, Egger's regression test, and Duval and Tweedie's 'trim and fill' method.

How were differences between studies investigated?

Differences between the studies were assessed using the Q-statistic and by calculating I-squared (I2). In addition, subgroup analyses were used to assess whether differences in outcome were affected by reason for hospitalisation (medical, surgical or trauma), and a meta-regression was used to assess the effect of the pre-specified factors of age, albumin level on admission, and time to initiate treatment.

Thirty RCTs (n=2,430) were included.

There was no evidence of a difference between EN and PN for hospital mortality; the overall RD was 0.6% (95% confidence interval, CI: -1, 2.2, P=0.4). The individual study results were mostly homogeneous (I2=0%). Similar results were seen in the medical, surgical and trauma subgroups.

Hospital length of stay was significantly reduced in patients receiving EN than in those receiving PN; the overall reduction was 1.2 days (95% CI: 0.38, 2.03, P=0.004). There was some heterogeneity between studies (I2=51%), but subgroup analyses were not performed because of the small number of studies. A similar result was also seen for length of stay in the intensive care unit (ICU; 4 studies), with an overall reduced length of stay for EN patients of 1.4 days (95% CI: 0.37, 2.4, P=0.008).

PN was associated with a significant increase in the risk of infective complications; the overall RD was 7.9% (95% CI: 4, 11.8, P=0.001). Similar results were seen in all subgroups. PN was also associated with a significant increase in the risk of noninfective complications (RD 4.9%, 95% CI: 0.3, 9.5, P=0.04) and a significant decrease in the risk of diarrhoea (RD -8.7%, 95% CI: -13.6, -3.8, P=0.001). There was no evidence of a difference between EN and PN with respect to technical complications; the overall RD was 4.1% (95% CI: -1.8, 9.9, P=0.2). Significant heterogeneity (I2=64%) was observed only for technical complications.

The meta-regression (results not reported) found no evidence that age, albumin level on admission, or time to initiate treatment were related to the outcomes. The cumulative meta-analysis did not show any effects of time. There was no evidence of publication bias (P=0.89 Egger's test; P=0.9 Begg's test). The trim-and-fill method suggested one missing study, although the mortality results remained unchanged.

There was no mortality effect with the type of nutritional supplementation. EN significantly reduced complication rates, although this needs to be interpreted in the light of missing data and heterogeneity. The enthusiasm that early EN, compared with early PN, would reduce mortality appears misplaced.

The search had several limitations which the authors discussed: only one electronic database was searched, no attempt was made to find unpublished studies, and language bias might have been introduced by the inclusion of only English language papers. However, publication bias was assessed using several methods, one of which suggested that one study was missing. The inclusion criteria were clearly defined and the studies were quality assessed, although there was little discussion of study quality and its potential impact on the conclusions. The authors did not state how relevant studies were selected, but at least two reviewers extracted the data and performed the quality assessment, thus reducing the potential for bias. The methods used to combine the studies in a meta-analysis were appropriate and a thorough investigation of heterogeneity was undertaken. The subgroup analyses seemed sensible. The authors presented a clear discussion of the potential limitations of the review and their conclusions are likely to be reliable.

Practice: The authors did not state any implications for practice.

Research: The authors stated that large, multicentre studies focusing on select subgroups (ICU patients, head trauma, gastrointestinal surgery) and specific outcomes (complications, and hospital and ICU length of stay) would further illuminate the understanding of the role of early nutritional support.

Unit Trust Funds.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.