Introduction

While enteral nutrition (EN) is the preferred method of providing nutrition support, it is often interrupted in critically ill patients due to various reasons and often falls short of meeting nutrient needs. Suboptimal nutrition in critically ill patients is well documented [1, 2, 3, 4]. A recent survey of Canadian ICUs shows that patients on EN received on average less than 60% of prescribed calories and protein [5]. When EN fails to meet the prescribed rate of nutrition, practitioners often prescribe parenteral nutrition (PN) in combination with EN to achieve the estimated nutrient needs and to avoid nutritional deficits. There is great variation in practice among ICUs, ranging from only 4% of patients on combination EN + PN [3] to as high as 60–80% of patients, as seen in some European centers [4, 6]. However, the use of parenteral nutrition is not without its risks in critically ill patients [7]. When randomized trials comparing EN to PN were reviewed, EN was associated with significantly fewer infectious complications [relative risk (RR) 0.61, 95% confidence intervals (CI) 0.44, 0.84, p=0.003] than PN [8]. In recently developed Clinical Practice Guidelines for Nutrition Support, the use of EN over PN is strongly recommended in patients with an intact gastrointestinal tract [9].

The benefits of any nutrition intervention must be balanced with the potential risks and complications of that intervention. The purpose of this study was to systematically review and statistically aggregate those studies using combination EN + PN to those using EN alone in critically ill patients.

Methods

Four bibliographic databases (Medline, Embase, CINAHL, and the Cochrane Library) were searched from 1980 to 2003. Studies were selected for inclusion in the review if they were: randomized clinical trials in critically ill patients (defined as those cared for in an ICU environment who had urgent or life threatening complications) and compared combination EN + PN to EN alone. We assessed the methodological quality of all selected articles in duplicate, independently, using a scoring system we have used previously [7] (Table 1).

Table 1 . Methodological quality assessment criteria
Full size table

The primary outcomes of interest were mortality rate (ICU and hospital) and number of patients who developed infectious complications. Secondary endpoints included measures of nutritional intake. We combined data from all studies to estimate the common risk ratio and associated 95% confidence intervals for death and infectious complications. The common risk ratios and their confidence intervals were estimated using the random effects model of DerSimonian and Laird [10] as implemented in RevMan4.1 [11]. For evaluation of the treatment effect we considered p<0.10 to be supportive of a trend and p<0.05 to be statistically significant. For the test of heterogeneity we considered p<0.10 to be statistically significant.

Results

Study identification and selection

Of 11 studies identified [12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22] 5 met the inclusion criteria [13, 14, 15, 16, 17]. The details of the study design, population, methodological quality, and outcomes of the included studies are shown in Table 2. In all the studies patients had an intact gastrointestinal tract, and PN was started at the same time as EN; none of the studies evaluated outcomes of combination EN + PN in patients with feeding intolerance in whom PN was started subsequently. Other trials were excluded due to nonrandomization [18, 19, 20, 21, 22] and other interventions (i.e., compared EN + PN to PN, not to EN) [12].

Table 2 Randomized studies evaluating combined EN + PN in critically ill patients. Maximal score: 14 (for methodological quality assessment criteria see Table 1) (EN enteral nutrition, PN parenteral nutrition, TBSA total burn surface area)
Full size table

Effect of combination EN + PN on mortality

In three studies the group receiving combination EN + PN had a higher mortality rate than the group receiving EN alone, but this reached statistical significance in only one study (63% in the EN + PN group vs. 26% in the EN alone group, p<0.05) [14]. The largest study showed no effect on ICU or day 90 mortality [17]. When all five studies were aggregated, the meta-analysis showed that the use of combination EN + PN had no effect on mortality (RR 1.27, 95% CI 0.82–1.94, p=0.3; see Fig. 1). The test for heterogeneity was not significant (p=0.26). A subgroup analysis was conducted to determine the confounding effects of nonisocaloric regimes in the trials and the possible relationship between “overfeeding” and the effect on mortality. There were no significant differences in mortality when trials in which the EN + PN group received more calories than the EN group [13, 14, 17] (RR=1.30 95% CI 0.78–2.16, p=0.3) were compared to trials in which the EN + PN group received similar calories to the EN group [15, 16] (RR=1.31, 95% CI 0.29–5.82, p=0.7, p value for differences between subgroups=0.94).

Fig. 1
figure 1

Effect of combination EN + PN on mortality [13, 14, 15, 16, 17]. EN Enteral nutrition; PN parenteral nutrition; EN + PN combination enteral nutrition and parenteral nutrition; n number of persons who died in the group; N total number of persons in the group; RR relative risk; CI confidence intervals. (Reprinted with permission from the American Society for Parenteral and Enteral Nutrition from the Journal of Parenteral and Enteral Nutrition, Canadian Clinical Practice Guidelines for Nutrition Support in Mechanically Ventilated, Critically Ill Adult Patients, Vol. 27; no, 5, pp355-378, Fig. 8; Sept. 2003; the American Society for Parenteral and Enteral Nutrition does not endorse the use of this material in any form other than its entirety

Full size image

Effect of combination EN + PN on infectious complications, length of stay, ventilator days, and cost

In one trial of mixed ICU patients [16] a trend towards a higher incidence of bronchopneumonia was observed in the group receiving EN + PN vs. EN alone (6/12, 50%, vs. 3/12, 25%, p=0.085). There were no differences in infectious complications between the groups in a larger study in a similar population of ICU patients [17]. We aggregated these two studies and observed an overall estimate of RR 1.14 (95% confidence interval 0.66 – 1.96, p=0.6; see Fig. 2). Only two trials reported on length of stay and ventilator days [16, 17], and when these were aggregated, combination EN + PN had no effect. The cost of combination EN + PN was significantly higher than EN alone as demonstrated by Bauer et al. [17] (204±119 vs. 106±47 euros/patient per 7 days, p=0.0001) and Chiarelli et al. [16] (31 Italian lira per parenteral calorie vs. 19 per enteral calorie).

Fig. 2
figure 2

Effect of combination EN + PN on infectious complications. EN Enteral nutrition; PN parenteral nutrition; EN + PN combination enteral nutrition and parenteral nutrition; n number of persons with infectious complications in the group; N total number of persons in the group; RR relative risk; CI confidence intervals

Full size image

Effect of combination EN + PN on nutritional intake

Three of the four studies evaluating the effect of combination EN + PN on nutritional intake (see Table 3) reported that the combination group received significantly more calories than the group that received EN alone [13, 14, 17]. Improved visceral proteins were seen in patients receiving EN + PN compared to those receiving EN alone in two studies [16, 17].

Table 3 Studies on the effect of combination EN + PN on nutritional intake (both interventions began at the same time) (EN enteral nutrition, PN parenteral nutrition, BEE basal energy expenditure, GI gastrointestinal)
Full size table

Discussion

In many ICUs, particularly in Europe, up to 60% of patients [6] receive EN + PN in combination as a strategy to optimize nutrient delivery. In this review we found no supportive evidence for this practice with respect to improving clinical outcomes. This systematic review of all studies comparing EN + PN to EN in critically ill patients revealed that despite favorable effects on nutritional indices, when PN is started at the same time as EN, it has no beneficial effect on mortality, complications, or length of stay but does result in significant higher costs. It has been argued that the negative effects associated with combination EN + PN may be attributed to “overfeeding” when compared to EN alone, however, according to the subgroup analysis there were no differences in mortality that could be attributed to “overfeeding.”

This review, which focuses on studies in critically ill patients only, has a comprehensive search strategy and robust methodology. The results of the studies are homogeneous, making the interpretation more valid. However, there are several limitations to the studies included in the review. First, one should be careful in extrapolating the results to those populations of critically ill patients who are severely malnourished. In the largest study in this review [17] a majority of the ICU patients were well nourished (59%) and had a normal body mass index. In other studies, authors do not comment on the state of malnutrition. Second, there were only two trials that report infections, and given the paucity of data, only weak inferences can be made from the aggregated estimate. Third, the duration of the intervention in the trials varied from 4–7 days [17] to 2 weeks [14, 16]. Since the trials of longer duration reported a higher mortality in the EN + PN group than in the EN group [13, 14, 15], one could hypothesize that the longer exposure to PN resulted in higher mortality. Fourth, as illustrated in Table 2, the methodological quality of these studies varied significantly (range from 6–12, out of a maximum of 14 points). Furthermore, although two of the primary authors reported costs associated with the combined approach, we did not do a formal evaluation of the economic data in these primary trials. Hence the inferences from these costing data are limited.

It is also important to note that the studies included in this review did not address the effects of combination EN + PN in patients who were started on PN due to intolerance to EN or in those with a compromised gastrointestinal tract or who were malnourished. All the studies looked at the effects of starting PN with EN at the same time.

Our findings have many implications for future studies on combination EN + PN in critically ill patients. Well designed randomized trials in critically ill patients who are intolerant to enteral nutrition and/or have a compromised gastrointestinal tract and/or are malnourished need to be conducted to determine the effects of combined EN + PN on clinically important outcomes in these populations. The degree of malnutrition and weight loss also needs to be considered while conducting these studies to determine the effects of combination EN + PN on mortality and morbidity in malnourished patients. The optimal timing of PN in these patients and the effects of transitional feeding (i.e., weaning from parenteral nutrition to enteral nutrition) also need to be studied. Recent evidence from nonrandomized studies shows that the addition of “trickle feeds,” i.e., small amounts of enteral nutrition added to parenteral nutrition reduces the complications associated with parenteral nutrition in ICU patients [20, 21].

In conclusion, in critically ill patients (with an intact gastrointestinal tract) who are not malnourished, starting PN at the same time as EN appears to have no added clinical benefit and is associated with higher costs than EN alone. Given the signals of potential harm from PN in critically ill patients from randomized trials and in the absence of data suggesting a benefit from using combination EN + PN, based on the evidence that we have presented, we recommend that PN not be started at the same time as enteral nutrition in critically ill patients. There are insufficient data on when to start PN in the patient who is not tolerating adequate amounts of EN. Practitioners will have to weigh the safety and benefits of initiating PN in patients not tolerating EN on an individual case-by-case basis. We recommend that PN not be started until all strategies to maximize EN delivery (e.g., small bowel feeding tubes, motility agents) have been attempted.