Web site designed and maintained by Chris Cairns  © SICS EBM Group 2004                                  

Parenteral vs. enteral nutrition in critically ill patients

Citation/s: Parenteral vs. enteral nutrition in the critically ill patient: a meta-analysis of trials using the intention to treat principle.  Intensive Care Medicine 2005; 31: 12-23.  F. Simpson and G.S. Doig

Lead author's name and email:  Gordon Stuart Doig.  gdoig@med.usyd.edu.au

Three-part clinical question: When standard parenteral nutrition (PN) is compared to standard enteral nutrition (EN) in intensive care patients is there any evidence of benefit in terms of clinically meaningful outcomes?

Search Terms: Medline and EMBASE were searched using sensitive search statements to detect all controlled trials, overviews and evidence-based guidelines of primary feeding interventions in critically ill patients.  Reference lists of overviews and guidelines were hand searched.  Experts and industry representatives were contacted.

The Study:
Randomised controlled trials comparing standard PN to standard EN were reviewed.  Twenty two RCTs were identified and nine trials were eligible for inclusion in the meta-analysis on an intention to treat basis. Two further trials passed a sensitivity analysis for adequate follow-up (>90%) and were analysed in addition. 

The Study Patients: 

Control group (in 9 intention to treat (ITT) trials) received EN either early (within 24 hours of ICU admission: 6 trials) or late (after 24 hours after admission: 3 trials) (N =274; all analysed):
Experimental group: received parenteral nutrition after trial enrolment  (N = 285; all analysed)

The Evidence:

Comments:  There was a reduced overall mortality associated with the use of parenteral nutrition (NNT 21) but this effect was seen to result from the (a priori) subgroup comparing PN with late EN (NNT 4, p<0.006).  The effect persisted when the other 2 studies (also comparing PN with late EN,) were analysed, p<0.008.  There was a trend towards increased infection in the PN group but this was non-significant.  Also definitions of infections were unclear in the study groups – sometimes it was stated that a culture was positive but not that there was a clinically significant infection.  The pooled data looks at numbers of positive cultures from any site.

EBM questions:

1) Do the methods allow accurate testing of the hypothesis?:  Yes, but the these trials are not blinded RCTs, although full blinding in a feeding trial would be difficult to achieve.  Trials were small – total of 559 patients in 9 trials.  Only one trial contained any blinding and only one stated that it maintained allocation concealment.

2) Do the statistical tests correctly test the results to allow differentiation of statistically significant results?  Yes. An ITT analysis is known to provide the most conservative estimate of treatment effect and protects against attrition bias from non-random loss to follow-up. Heterogeneity was non-significant when looking at mortality but significant when looking at infections, such that it may not be appropriate to pool the infection results.  A component approach was used to look at the validity of trials.  With a different approach (random effects model) the benefit of PN remained. 

3) Are conclusions valid in light of the results?  Probably.  The conclusion that PN when compared to late EN leads to a lower mortality in critically ill patients is made on the basis of the results of 5 RCTs all of which show this as an outcome.  However without more information being available it is likely that the trials compared a group of patients who could have received either PN or EN, and not patients who could not be given EN early for clinical reasons, which is the group for which PN is usually considered and which may represent a different population. 

4) Did results get omitted, and why?  Trials with excessive loss to follow-up were not included.  Full reporting required follow-up of 95 % of patients for ITT analysis. Eleven out of 22 RCTs obtained by the search were omitted: 3 were excluded as they were part of a bigger trial, 5 because they included immune-enhanced EN or PN, one was pseudo-randomised, and 2 did not report outcomes on 12 and 21% of patients (10% taken as cut off).  
5) Did they suggest areas of further research?  Yes.  They suggest that further work should be done to look at methodological quality in trials in meta-analyses.  They also recommend a large multi-centre RCT addressing the use of standard PN in patients for whom standard EN cannot be started for at least 24 hours after initial injury or ICU admission.  This trial should contain a blinded assessment of clinically important infections and be powered to detect a difference in a clinically meaningful outcome.  It should also include an economic analysis of the use of PN compared to EN.
6) Did they make any recommendations based on the results and were they appropriate?  The authors say that previous reviews have not shown a benefit for PN vs EN but these have included trials of dubious quality.  This study excludes low quality trials that may have been biased against PN.  They recommend the use of PN in patients for whom EN could not be started within 24 hours of ICU admission or injury.  They do not recommend the use of PN in patients in whom EN can be initiated within 24 hours.
7) Is the study relevant to my clinical practice  Yes: there is a large variation in PN use across Scotland – some units use a lot of PN early and believe it should be given to supplement enteral.  Others use very little PN and there is a difference of opinion over when PN should be considered.

8) What level of evidence does this study represent? 1⁺

9) What grade of recommendation can I make on this result alone? A, with the reservation that although the RCTs are well conducted they are small and look at differing groups of critically ill patients..

10) What grade of recommendation can I make when this study is considered along with other available evidence?  Evidence based recommendations in the ACCEPT trial (CMAJ Jan 20,2004; 170(2)) also suggested early enteral nutrition within 24 hours with the early use of PN when enteral could not be initiated within 24 hours but these guidelines may have been based on some of the same trials as the evidence base is limited.

11) Should I change my practice because of these results? Although the results of this meta-analysis are based on a small number of small trials, better evidence is not available and we should consider starting PN within 24 hours in patients unable to start enteral feed within 24 hours as it may confer a mortality benefit. Evidence for increased infection rates with PN is weak.  All the enteral feeding in these trials (and in many other nutrition trials) was post-pyloric which may not be comparable to practice in all units.  However administering naso-jejunal feeding rather than naso-gastric feeding early in patients’ stay to optimise enteral absorption is a sensible strategy and is widely employed in many units throughout the world.

12) Should I audit my current practice because of these results?  We should look at fulfilment of nutritional requirements to examine whether our patients are falling behind their nutritional goals early in their ICU stay due to inadequate enteral feed.  However this meta-analysis does not address the question of whether PN should be commenced in patients who are not reaching their nutritional goals with EN, only patients who are not started on EN within 24 hours of ICU admission. It would seem likely that patients suffering inadequate enteral feeding should also benefit from the addition of parenteral nutrition but this conclusion cannot be drawn directly from these results.

Appraised by; Marcia McDougall, Consultant, Queen Margaret Hospital, Dunfermline, U.K.
Email: marcia.mcdougall@faht.scot.nhs.uk

Kill or Update By: June 2009

Reviewed and edited by BC & CC

Citation: EBM Critical Appraisals. Scottish Intensive Care Society EBM Group. 2007. McDougal M. Parenteral vs. enteral nutrition in the critically ill patient: a meta-analysis of trials using the intention to treat principle.  Intensive Care Medicine 2005; 31: 12-23.  F. Simpson and G.S. Doig

Printer friendly view

©SICS EBM 2007