Introduction

Nutritional support is considered an essential component of the management of critically ill patients [1], with Clinical Practice Guidelines (CPG) emphasizing early (within 24–48 h of ICU admission) normocaloric enteral nutrition (daily caloric intake estimated to match 80–100 % of energy expenditure) [2–5]. It is generally believed that early normocaloric enteral nutrition attenuates the catabolic response of acute illness, reduces disease severity, attenuates the pro-inflammatory and anti-inflammatory immune response, diminishes complications, decreases ICU and hospital length of stay (LOS) and favorably impacts patient outcome [2–5]. Observational cohort studies suggest that underfeeding critically ill patients with the accrual of a large caloric deficit (difference between required and delivered calories) is associated with worse patient outcomes [6–11]. However, a number of recent randomized controlled trials (RCTs) have failed to demonstrate an improvement in the outcomes of critically ill patients receiving a normocaloric feeding protocol as opposed to a strategy of intentional hypocaloric feeding. The objective of this systematic review and meta-analysis was to determine whether nutritional support targeting recommended caloric goals improves the outcomes of ICU patients as compared to an intentional hypocaloric approach.

Methods

Identification of trials and outcome variables

Our aim was to identify all relevant RCTs that compared a normocaloric (targeting 80–100 % of daily energy expenditure) with an intentional hypocaloric feeding strategy. We included studies that compared normocaloric feeding with a protocol that used “tropic feeding” as well as studies that compared normocaloric feeding versus permissive underfeeding (<70 % of daily energy expenditure). We excluded studies in which parenteral nutrition was the main source of nutrition. We limited the search strategy to adult ICU patients. There was no restriction on language, publication year or setting. Both authors independently searched the National Library of Medicine’s MEDLINE database for relevant studies in any language published from 1966 to July 2015. The MEDLINE (OVID) search strategy is depicted in Fig. S1. In addition, we searched the Cochrane Register of Controlled Trials (CENTRAL) and the bibliographies of all selected articles and review articles for other relevant articles. This search strategy was done iteratively, until no new potential citations were found. We performed and reported this meta-analysis according to the guidelines proposed by the PRISMA group [12, 13]. The outcome measures of interest were infectious complications, hospital mortality, ICU LOS and ventilator-free days (VFDs).

Data extraction and quality assessment

Both reviewers independently assessed eligibility of articles identified in the initial search strategy for inclusion in the review. They discussed those papers deemed potentially eligible, independently extracted data using a standardized data abstraction form, and assessed studies’ methodological quality using the risk of bias assessment tool from the Cochrane handbook for randomized trials [14]. Disagreements between the reviewers were resolved by consensus. Missing data were kindly provided by the primary investigator when available [15, 16].

Data analysis

Summary data is presented as sample size, time to enrollment in study, patient characteristics, caloric goals, the average amount of calories and protein delivered and the outcomes of interest in each group of patients. Studies were grouped according to whether a trophic feeding or permissive underfeeding strategy was used. When sufficient data elements were available and, when appropriate, meta-analytic techniques were used to summarize the data. We used the random-effects model (more conservative) using Review Manager 5.3.4 (Cochrane Collaboration, Oxford, UK) and considered p ≤ 0.05 (two-sided) as significant. We report binary outcomes as odds ratios (OR) and continuous outcomes as weighted mean differences (WMD). Summary effects estimates are presented with 95 % confidence intervals (CI). We assessed heterogeneity between studies for each outcome using the Cochran Q statistic [17], with p ≤ 0.10 indicating significant heterogeneity [18], and I 2 with suggested thresholds for low (25–49 %) moderate (50–74 %) and high (>75 %) values. We performed a Funnel plot to determine publication and study bias [19, 20].

Results

Our initial search strategy identified 191 possible studies for inclusion in this analysis. A total of 133 studies were excluded after screening and, of the 58 full text articles assessed for eligibility, 54 were excluded. In addition to the four studies identified by our primary search, two additional studies were identified from the bibliographies of the primary articles identified and review articles. The results of the search strategy are depicted in Fig. S2. The six studies included in the systematic review and meta-analysis are summarized in Table 1. In all, 2517 patients were enrolled in these six studies. The average sample size was 420 (83–1000) patients. Two studies evaluated trophic feeds (n = 1200) [15, 16], while four studies evaluated permissive underfeeding (n = 1317) [21–24]. The risk of bias assessment of the included studies in depicted in Fig. S3. None of the studies blinded patients or their healthcare providers. Blinding of outcome assessment was not reported in any of the studies included.

Table 1 Characteristics and nutrition data of studies included in meta-analysis
Full size table

The demographics of the patients were similar across studies with a mean age of 53 ± 5 years and a mean BMI of 29.1 ± 1.5 kg/m2. The APACHE II score was reported in five studies with a mean of 24 ± 5. The nutrition provided and outcomes data are listed in Table 2. In the trophic studies, 1359 ± 83 Kcal (77 % of goal) was provided in the normocaloric group as compared to 350 ± 70 Kcal (20 % of goal) in the trophic group during the first week of hospitalization. Patients in the trophic group did not receive supplemental protein, receiving 54 and 11 g, respectively. In the permissive underfeeding studies, patients in the normocaloric group received 1246 ± 126 kilocalories (72 % goal) as compared to 910 ± 73 kilocalories (49 % of goal) in the permissive underfeeding group (see Table 2). Three of the permissive underfeeding studies provided additional protein in the hypocaloric group to achieve similar protein intake [21–23]. In the trophic studies, patients in the intentional hypocaloric arm were underfed for 6 days (at which point they were transitioned to target nutritional goals if they remained in the ICU) [15, 16]. In the permissive underfeeding studies, the intentional hypocaloric group were underfed for the duration of the patients ICU stay or until day 14 [22].

Table 2 Nutrition received and outcome data of studies included in meta-analysis
Full size table

Infectious complications were reported in all six studies; summary data demonstrated no difference in the risk of infectious complications (OR 1.03; 95 % CI 0.84–1.27) with minimal heterogeneity between studies (I 2 = 16 %) (see Fig. 1). The funnel plot revealed no evidence of publication bias (Fig. S4). Hospital mortality was reported in all six studies; overall, there was no difference in the risk of death between the intentional hypocaloric and normocaloric group (OR 0.91; 95 % CI 0.75–1.11, I 2 = 8 %) (see Fig. 2); however, there was a trend towards a lower mortality favoring hypocaloric feeding in the permissive underfeeding subgroup (OR 0.80; 95 % CI 0.62–1.02, p = 0.07, I 2 = 0 %). ICU length of stay was reported in four studies, with no overall difference between the intentional hypocaloric and normocaloric groups (mean difference 0.05 days; 95 % CI 1.33–1.44 days; I 2 = 37 %) (see Fig. 3). VFDs were reported in three studies, and in each of these there was no significant difference in VFDs between the intentional hypocaloric and normocaloric group (see Table 2). VFDs to study day 28 was reported in two studies [15, 16], while VFD to study day 90 was reported in one study [22]. Due to the characteristics of the data, it was not possible to pool these end-points to determine the summary statistics.

Fig. 1
figure 1

Comparison of the risk of hospital acquired infections/sepsis for patients receiving standard feeds as compared to hypocaloric feeds grouped by tropic feeds and underfeeding. Weight is the relative contribution of each study to the overall treatment effect (odds risk ratio and 95 % confidence interval) on a log scale assuming a random effects model

Full size image
Fig. 2
figure 2

Comparison of the risk of hospital mortality during study period for patients receiving standard feeds as compared to hypocaloric feeds grouped by tropic feeds and underfeeding. Weight is the relative contribution of each study to the overall treatment effect (odds risk ratio and 95 % confidence interval) on a log scale assuming a random effects model

Full size image
Fig. 3
figure 3

Comparison of length of hospital stay (LOS) for patients receiving standard as compared to hypocaloric feeds grouped by tropic feeds and underfeeding. Weight is the relative contribution of each study to the overall treatment effect (weighted mean difference with 95 % confidence interval)

Full size image

Discussion

The results of our meta-analysis demonstrated no benefit of targeting estimated caloric requirements (80–100 % of daily energy expenditure) as compared to intentional hypocaloric enteral nutrition in the acute phase of critical illness. Furthermore, the studies by Rice et al. [15, 16] suggest that trophic nutrition combined with limited protein intake does not adversely affect patient outcomes. These findings challenge conventional wisdom and current Clinical Practice Guidelines [3–5]. It is noteworthy that intentional hypocaloric nutrition did not increase hospital mortality and infectious complications nor decrease VFDs. It has been suggested that hypocaloric nutrition may impair immune responsiveness and increase infectious complications. However, limiting nutritional intake stimulates autophagy which is an important host defense mechanism against intracellular pathogens [25–27]. Critical illness is characterized by stress hyperglycemia which appears to be an evolutionary preserved response to provide energy at a time of crises [28]. “Occult overfeeding”, when exogenous calories are delivered on top of the non-inhibitable endogenous production of glucose by gluconeogenic organs [29], may result in metabolic dysequilibrium with adverse effects. Critical illness is associated with the rapid loss of skeletal and diaphragmatic muscle mass, and this hampers attempts at weaning patients from mechanical ventilation and is associated with muscle weakness up to 5 years post-ICU discharge [30, 31]. Aggressive nutritional support with increased provision of protein has been postulated to limit this catabolic response and therefore preserve muscle mass [32]. The finding that VFDs were not decreased with hypocaloric feeding suggests that underfeeding does not accelerate loss of muscle mass. This finding is further supported by the studies of Needham et al., who followed the patients enrolled in the EDEN study for up to a year post-hospital discharge [33, 34]. Muscle strength, muscle mass and 6 min walk distance did not differ between the trophic and full feeding groups at 6 and 12 months [33]. Furthermore, the age and sex adjusted physical function domain of the SF-36 instrument at 12 months did not differ between the two groups [34].

The results of our meta-analysis are consistent with two additional randomized controlled trials that did not meet the inclusion criteria for our meta-analysis [35, 36]. These studies tested the hypothesis that a “volume-based” feeding strategy that enhanced the delivery of calories and protein would improve patient outcomes. Heyland et al. performed a prospective, cluster randomized controlled trial to determine the effect of the PEP uP protocol combined with a nursing educational intervention on the outcomes of 1059 critically ill mechanically ventilated patients in 18 ICUs [35]. Although the proportion of prescribed protein and energy delivered by enteral nutrition was significantly greater (p = 0.004) in the intervention ICUs compared to the control ICUs, none of the outcome measures investigated differed significantly between the control and intervention ICUs. Braunschweig et al. randomized 78 patients with acute lung injury to a volume based or standard nutritional protocol [36]. The percent of energy needs received per day averaged 85 % in the volume-based group as compared to 55 % in the control group. The trial was terminated prematurely by the Data Safety Monitoring Board due to a significantly higher risk of hospital death in the volume-based group (40 vs. 16 %, p = 0.02).

The reasons for the lack of benefit of normocaloric over hypocaloric feeding in ICU patients are not entirely clear. It is possible that normocaloric enteral nutrition does not alter the catabolic process or immune response associated with acute critical illness. Anorexia is an evolutionary preserved response during acute illness which may be protective [37, 38]. Nutrient deprivation promotes autophagy and this may play a key role in promoting host defences and the immune response to intracellular pathogens [27, 39]. In all the studies included in this meta-analysis, patients received continuous rather than intermittent enteral nutrition. Intermittent enteral nutrition is more physiologic than continuous feeding and could possibly explain the lack of benefit of normocaloric nutrition provided as a continuous infusion [40–43]. Finally, it is possible that the combined trials do not have the power to rule out the presence of a clinically meaningful effect.

The meta-analysis reported here combines data across studies in order to estimate treatment effects with more precision than is possible in a single study [12]. The main limitation of this meta-analysis is that the patient population and feeding protocols were not the same across studies. It should be recognized that, while all the studies enrolled “typical” ICU patients and did not exclude patients based on age or BMI, the mean age of the patients’ was 55 years and their BMI was 29.1 kg/m2. Furthermore, the duration of hypocaloric nutrition was limited to either 7 days, 14 days or the duration of ICU stay; it is likely that prolonged intentional hypocaloric nutrition (once the acute critical illness has resolved) is neither safe nor desirable. It has been suggested that “trophic feeds and permissive underfeeding cannot be considered safe or indicated in older, higher risk ICU patients as it appears to increase mortality and impair long-term quality of life” [44]. There are, however, no prospective studies that support this contention. Additional prospective studies targeting this patient population would be required to resolve this issue. It is important to emphasize that this meta-analysis compared the provision of different levels of energy and not protein. It may be that the quantity, quality and mode of delivery of protein may be more relevant for outcome than the level of energy supplied.

Conclusions

This meta-analysis failed to demonstrate an outcomes benefit from normocaloric as opposed to intentional hypocaloric nutrition in heterogenous ICU patients during the acute phase of their illness. The optimal strategy for providing calories and protein in critically ill patients has yet to be determined.