Eleven trials met the inclusion criteria for the review. Nine (n=559) were included in the primary analysis.
None of the 11 trials reported the maintenance of allocation concealment, one reported the use of blinding and nine allowed the conduct of an ITT analysis. There was no evidence of publication bias in the funnel plot.
Based on the 9 trials presenting ITT results, a statistically significant reduction in mortality rates was found with PN compared with EN (OR 0.51, 95% confidence interval, CI: 0.27, 0.97, P=0.04). No evidence of statistical heterogeneity was detected (P=0.50; I-squared 0%).
Six trials with complete follow-up reported on infectious complications. There was a statistically significant increase in infectious complications with PN compared with EN (OR 1.66, 95% CI: 1.09, 2.51, P=0.02). Some evidence of statistical heterogeneity was shown (I-squared 37.7%; P=0.16).
In subgroup analyses PN demonstrated a statistically significant advantage over delayed EN for the reduction of mortality (OR 0.29, 95% CI: 0.12, 0.70, P=0.006), with no evidence of statistical heterogeneity (P=0.60; I-squared 0%). The use of PN compared with early EN did not show a statistically significant effect on mortality. There were more infectious complications with PN compared with early EN, but this was not statistically significant and statistical heterogeneity was noted.
The benefit of a reduction in mortality with PN remained in sensitivity analyses that included the two trials (both comparing PN with late EN) with incomplete follow-up.