Observational Studies on Hypocaloric Nutrition in Critically Ill Patients

Over the past decade, several observational studies have examined the association between energy intake and clinically important outcomes in critically ill patients. Nine observational studies have demonstrated that caloric debt or feeding less than goal calories is associated with worse clinical outcomes. However, four additional studies found contrasting results, indicating that providing close to goal calories has adverse effects in critically ill patients. Different methodological approaches may account for the inconsistent conclusions across these observational studies. Consequently, we conducted a large observational study to evaluate the association between caloric intake and clinical outcome. Our data, using a pooled dataset of 7872 mechanically ventilated critically ill patients from 352 ICUs from within 33 countries and including only those patients who remained in the ICU for at least 96 hours, showed that the result is highly dependent on the statistical methods used. When the most robust statistical method (i.e., excluding patients who permanently progressed to oral feeding within 4 days and basing the 12-day average proportion of prescribed calories received only on ICU days before permanent progression to oral intake, in addition to adjusting for evaluable days and covariates) was applied, we observed that 60-day hospital mortality in patients receiving more than two thirds of their caloric prescription was significantly lower than patients receiving less than one-third of their caloric prescription (odds ratio [OR], 0.67, 95%; confidence interval [CI], 0.56 to 0.79; P ≤ .0001). Furthermore, these results indicated that providing approximately more than 80% of prescribed calories was associated with the optimal clinical outcome. Similar results were observed in a more recent analysis restricted to 2270 patients with an ICU admission diagnosis of sepsis. Thus, on the basis of these large-scale “real world” observational studies, one would conclude that “underfeeding” the critically ill patient (including patients with sepsis) is not appropriate. As stated in our prior publication, though, “the causal association between nutritional intake and outcome cannot be definitively established by any observational study. No perfect adjustment is available despite our best efforts to account for the confounding effects of the duration of artificial nutrition.” To truly find the answer, we need to seek evidence from randomized controlled trials (RCTs).

Randomized Controlled Trials of Intentional Underfeeding

Over the past four years, three RCTs comparing intentional hypocaloric and full feeding early in the course of ICU stay have been published. The first of these, by Arabi et al., adopted a 2 × 2 factorial design to examine the effect of permissive underfeeding (i.e., 60% to 70% of calculated energy requirements) compared with feeding to goal calories (i.e., 90% to 100% of calculated energy requirements) and of intensive insulin therapy compared with conventional insulin therapy on the clinical outcomes of critically ill patients. A total of 240 predominantly medical (83%) patients with the mean age of 51 years and body mass index (BMI) of 28.5 kg/m ² were enrolled in this single-center study. Those allocated to the permissive underfeeding group received on average 59.0 ± 16.1% of their goal calories compared with 71.4 ± 22.8% in the target feeding group ( P < .0001 although the target goal in the second group was not reached). The primary outcome of 28-day mortality was not significantly different between the two groups (18.3% vs. 23.3%, relative risk [RR] 0.79; 95% CI, 0.48 to 1.29; P = .34), but hospital mortality was significantly lower in the permissive underfeeding group compared with the target feeding group (30.0% vs. 42.5%: RR, 0.71; 95% CI, 0.50 to 0.99, P = .04). No other differences were observed. Arabi et al. are currently repeating this trial in multiple centers, the results of which are anticipated in the near future (Current Controlled Trials Register Number: ISRCTN68144998).

In the second single-center pilot RCT conducted by Petros et al., 100 critically ill patients who were predicted to require artificial nutrition for at least 3 days were randomized to receive either early (within 24 hours of ICU admission) full feeding or hypocaloric feeding (50% of their estimated caloric requirements based on 25 kcal/kg/day regimen). Patients allocated to the hypocaloric feeding group received on average 42.6% of their caloric requirements whereas patients allocated to the full feeding group received on average 75.5% of goal calories ( P = .0001). The primary endpoint was the rate of nosocomial infections during ICU stay. Significantly more infections were detected in patients in the hypocaloric group compared with the full feeding group (12 of 46 [26.1%] vs. 6 of 54 [11.1%], P = .046). No other differences were observed in clinical outcomes.

In another single-center pilot RCT, 83 patients with a mean age of 52 years and a BMI of 30.5 kg/m ² , who were admitted to a surgical ICU and projected to require artificial nutrition for more than 48 hours, were randomized to hypocaloric (i.e., 50% of calculated caloric requirements) or full feeding (i.e., 100% of calculated caloric requirements). Patients randomized to the hypocaloric arm received significantly fewer calories compared with patients in the full feeding arm (983 [standard deviation [SD], 61] vs. 1338 [SD, 92] kcal; P = .019). The primary outcome was the proportion of patients acquiring infection. No significant differences were observed (RR, 70.7% [29 of 41] in the hypocaloric arm and 76.2% [32 of 42] in the full feeding arm; P = .57, adjusted OR, 0.82; 95% CI, 0.28 to 2.39). No differences were observed in other infectious or clinical outcomes.

Given the disparate and preliminary nature of the results of these three studies, it is prudent to wait for the results of the larger multicenter trials before making conclusions regarding early intentional underfeeding.

Randomized Controlled Trials of Trophic Feeds

Two RCTs, conducted by the same research group aimed to test the hypothesis that initial trophic EN (i.e., provision of small volume of EN) would decrease gastrointestinal complications and improve outcomes. In the first single-center study, 200 patients (mean age, 54 years; BMI, 28.7 kg/m ² ) with acute respiratory failure expected to require mechanical ventilation for at least 72 hours were enrolled and allocated to receive either full-energy EN (EN initiated at 25 mL/hr within 12 hours of randomization and advanced every 6 hours until goal rate achieved [within 1 to 2 days]) or trophic EN (EN initiated at 10 mL/hr and advancing to full-energy EN on study day 6). For study days 1 to 5, patients in the full-energy group received significantly more calories than patients in the trophic EN group (1418 ± 686 kcal/day vs. 300 ± 149 kcal/day, P ≤ .001). Overall, there was a trend toward less gastrointestinal intolerance (26.5% vs. 39.2%; P = .08), less gastric residual volumes of more than 300 mL (2.1% vs. 7.5%; P ≤ .001), and less diarrhea (19.1% vs. 24.1%; P = .08) in the trophic EN group. No differences were observed in clinical outcomes or infectious complications.

The second, more recent RCT was a large, multicenter study conducted in 44 ICUs in the United States. The research team adopted a 2 × 2 factorial design with the intention of also evaluating the effectiveness of omega-3 fatty acid supplementation (the Early versus Delayed Enteral Nutrition [EDEN] study). A total of 1000 patients with a mean age of 52 years, a BMI of 30 kg/m ² , and a diagnosis of acute lung injury (ALI) were randomized to receive the same full-energy or trophic EN interventions as described for the previous single-center RCT. Patients randomized to the full-energy EN group received significantly more calories in the first 5 days than patients in the trophic EN group (1300 ± 82 vs. 400 ± 25 kcal/day, P = .001) and achieved the goal rate within 1.3 ± 1.2 days compared with 6.7 ± 1.8 days ( P = .001). Overall, there was a significantly lower incidence of gastric residual volumes of greater than 300 mL (2.2% vs. 4.9%, P ≤ .001) and a trend toward less diarrhea (16.5% vs. 18.7%, P = .16) and vomiting (1.7% vs. 2.2%, P = .05) in the trophic EN group. The authors reported no significant differences in ventilator-free days, infections, 60-day mortality, physical function, cognitive performance, and other outcomes at 6 and 12 months. There was a trend, though, toward improved 6-minute walk test scores with full feeding, and more patients who received trophic feeding were admitted to a physical rehabilitation facility (57 [23%] vs. 30 [14%]; P = .01).

Therefore data from these two trials of trophic feeding do not indicate that it improved clinical outcomes, but it may reduce gastrointestinal complications. The absence of harm from trophic feeds in the initial 6 days of ICU stay may reflect the underpowered nature of these studies, and early trophic feeding may negatively affect long-term recovery and physical function.

Randomized Controlled Trials of Early Enhanced Enteral Nutrition

We identified seven trials designed to answer the questions “Is enterally providing more calories compared with fewer calories during this early phase beneficial?” Four RCTs that have linked increased energy intake from EN begun early in the course of critical illness with improved patient-centered outcomes, one RCT evaluated a bundle comprising active supervision of nutrition provision together with delivery of near target energy requirements determined by repeated energy measurements, and two cluster RCTs evaluated the effects of an enhanced feeding protocol intended to increase EN delivery.

The first RCT by Taylor et al. investigated the effects of early enhanced EN on clinical outcomes in 82 mechanically ventilated patients with severe head injury randomized to receive either standard early EN or enhanced early EN. Enteral feeding was started within 24 hours of the injury in both groups. In the control group, patients received EN starting at 15 mL/hr, which was increased incrementally as tolerated according to a predefined protocol. In the intervention group, patients received EN starting at the rate that would meet their full energy requirements. During the first week after head injury, patients in the enhanced EN group received significantly more calories than patients in the control group (59.2% vs. 36.8% of caloric goal, P ≤ .001). There was a trend toward improved neurologic outcome 3 months after injury in the intervention group (proportion with good neurologic recovery 25 of 41 [61%] vs. 35 of 41 [85%]; P = .08), but this difference was not apparent at 6 months, suggesting that the aggressively fed group had a faster time to recovery. Patients in the intervention group also had fewer overall complications, including infections, up to 6 months after the initial injury (37% vs. 61%; P = .046). There was no difference in mortality (12.2% in the intervention group and 14.6% in the control group), although the study was not adequately powered for this endpoint.

The second RCT comparing the use of early enhanced EN to standard early EN was conducted by Desachy et al. One hundred patients admitted to two ICUs were enrolled and randomized to either initiate EN within 24 hours at goal rate (i.e., to achieve a caloric intake of 25 kcal/kg) or to initiate EN within 24 hours at 25 mL/hr with gradual increase to goal rate. Patients in the study group received significantly more calories than the control group (1715 ± 331 kcal/day vs. 1297 ± 331 kcal/day, P ≤ .001), achieving on average 95% of their energy needs compared with 76% in the control group. The incidence of high gastric residual volumes of more than 300 mL was greater in the early enhanced EN group ( P = .04). There was no difference in the mortality, hospital, and ICU length of stay or incidence of adverse events necessitating withdrawal of EN.

The third single-center RCT, conducted by Braunshweig et al., aimed to evaluate the influence of intensive medical nutrition therapy in patients with ALI. A total of 78 patients, the majority of whom were well nourished, were enrolled and randomized to receive either intensive administration of EN (>75% of goal calories) or standard care. Patients in the intervention group received 84.2% of goal calories compared with 55.4% in the control group ( P ≤ .0001). The trial was stopped early because of safety concerns surrounding the significantly higher mortality rate in the intensive EN group (40% vs. 15.6%, P = .017).

The fourth, and most recent, trial, conducted in five ICUs in Australia, randomized 112 mechanically ventilated patients who were expected to require EN for more than 2 days to receive a concentrated 1.5-kcal/mL EN solution or a standard 1-kcal/mL EN solution. Study EN was provided for the duration of the patients’ ICU stay up to a maximum of 10 days. Patients allocated to the concentrated EN group received significantly more calories than patients in the standard EN group (2040 ± 578 vs. 1504 ± 573 kcal, P ≤ .001). The study was not powered to detect differences in adverse events or clinical outcomes. There was a trend, though, toward longer 90-day survival in the concentrated EN group ( P = .057).

The fifth RCT evaluating the “dose” of nutritional support was a multicenter cluster-RCT of algorithms for critical care enteral and parenteral therapy in 14 Canadian ICUs (ACCEPT [algorithms for critical-care enteral and parenteral therapy]). This trial evaluated the impact of evidence-based feeding algorithms on nutrition practices and patient outcomes. Four hundred and ninety-nine patients, ages 16 years or older, who were expected to stay in the ICU at least 48 hours were enrolled in the study. An intensive educational program was provided at the sites assigned to the intervention group. ICUs assigned to the control group did not receive any of the interventions. Patients at the intervention hospitals received significantly more days of EN per 10 days (6.7 vs. 5.4 days; P = .042), had a significantly shorter length of hospital stay (25 vs. 35 days; P = .003), and demonstrated a trend toward reduced mortality (27% vs. 37%; P = .058). Length of ICU stay was not different between the two groups, though (10.9 vs. 11.8 days; P = .7). Admittedly, it is difficult to understand how such a small difference in the dose of EN is associated with such large changes in clinical outcomes.

To confirm these observations, Doig and colleagues performed a complex, multifaceted intervention in 27 community and teaching hospitals in Australia and New Zealand. The trial involved 18 different strategies to change nutrition practice, including an evidence-based feeding algorithm. ICUs randomized to receive the intervention participated in a 2-day guideline development conference that included an educational workshop on the use of the 18 interventions to be used to implement the new guidelines. The study found that EN was initiated earlier in patients from intervention ICUs (0.75 vs. 1.37 days; P < .001) and patients achieved the caloric goal more often (6.10 vs. 5.02 days per 10 fed-patient days; P = .03). In addition, more patients were never fed in the control ICUs (28.2 vs. 5.7%; P ≤ .001). No significant differences were observed in any of the measured clinical outcomes, though.

Finally, a single-center pilot RCT conducted in Israel aimed to determine if tight caloric control improved hospital survival. One hundred and thirty patients were randomized to have their nutritional requirements guided either by repeated resting energy expenditure (REE) measurements or by a single, initial weight-based measurement. Although the mean REE was not different between the study and control groups, the mean energy delivered was significantly higher in the study group (2086 ± 460 vs. 1480 ± 356 kcal/day; P = .01). An intention-to-treat analysis showed a trend toward improved mortality in the study group (32.3% vs. 47.7%; P = .058). However, the study group also had a longer duration of mechanical ventilation (16.1 ± 14.7 days vs. 10.5 ± 8.3 days; P = .03) and ICU stay (17.2 ± 14.6 days vs. 11.7 ± 8.4 days; P = .04) as well as more infectious complications (37 vs. 20; P = .05). These discrepant results need to be further explored.