Intestigate the effect of trial quality on overall conclusions surrounding use of parenteral nutrition (PN) in critical illness.

Characteristics of Included Articles

Patient characteristics

• PN and EN did not contain supplemental glutamine, arginine or other immune-enhancing ingredients
• Published in the English language prior to April 30, 2003
• Patients were recruited in ICU and the majority were receiving invasive mechanical ventilatory support
• Average ICU length of stay (LOS) was longer than two days
• Severity of illness was reported and commensurate with patients being critically ill
• Patients were suffering from a condition usually requiring care in ICU (burns over 40% to 50% of total body surface, multi-trauma requiring urgent laparotomy).

Methodological article quality

Reporting of three key methodological components including:

• Intent-to-treat analysis
• Maintenance of allocation concealment during randomization
• Appropriate use of blinding (participants, investigators, outcome adjudicators).

Papers were excluded if they did not meet criteria for patient inclusion, article quality or had more than 10% loss of patients to follow-up.

Search Procedures

• Database search included Medline and EMBASE
• Reference lists of overviews and evidence-based guidelines hand-searched
• Contacted experts and industry representatives to ensure trials were not missed
• The final closure data for search process was April 30, 2003.

Study Quality

• Patients recruited into trials were critically ill, as defined per inclusion criteria
• Positive methodological quality, as defined per inclusion criteria (presence of intent-to-treat analysis, maintenance of allocation concealment during randomization and appropriate use of blinding, less than 10% of patients lost to follow-up).

Populations Included

Critically ill (included patients with burns over 40% of body).

• Infectious complications (sepsis, pneumonia, UTI, wound infections, peritonitis)
• Mortality
• Timing of feeding on clinically meaningful outcomes.

Method for Synthesis of Information

• Meta-analysis with fixed-effect model, using odds ratio metric
• Herterogeneity assessed with X² statistic and I² measure
• Formal test of treatment-sub-group interactions.

• 465 articles were identified and retrieved as potentially relevant
• 329 RCTs were reviewed for more detailed evaluation
• 124 RCTs were considered for meta-analysis
• 22 RCTs compared EN to PN
• 11 RCTs met all criteria for inclusion.

Key Findings

• There were six studies for early EN with 224 subjects; five studies for late EN with 121 subjects
  • When PN compared to early EN (started within 24 hours), no effect on mortality (OR, 1.07; 97% CI, 0.39 to 2.05; P=0.89)
  • Mortality was significantly less with PN compared to delayed EN (start time of greater than 24 hours; OR, 0.29; 95% CI, 0.12 to 0.70; P=0.006).
• For infectious complications, there were three studies that reported a number of patients with infections (N=136) and three studies reporting a total number of infections (N=89 subjects): Use of PN resulted in a trend toward more infectious complications compared to early EN (OR, 1.47; 95% CI, 0.90 to 2.38; P=0.12).

Other Findings

Infectious complications in early vs. delayed EN was not analyzed due to inadequate studies to include in the meta-analysis: Two studies.

Although there is an association with increased infectious complications with PN use, PN should be used in patients for whom EN cannot be initiated within 24 hours or ICU admission or injury.