We conducted an investigator-initiated, randomized, double-blind, pragmatic trial in 46 ICUs in Australia and New Zealand (Table S1 in the Supplementary Appendix, available with the full text of this article at NEJM.org) between June 21, 2016, and November 14, 2017. The trial, which was endorsed by the Australian and New Zealand Intensive Care Society (ANZICS) Clinical Trials Group, was designed by the management committee and conducted and analyzed by the investigators (see the Supplementary Appendix). The trial was funded by national peer-reviewed organizations. The funders had no role in the design or conduct of the trial; in the collection, analysis, or interpretation of the data; or in the approval of the manuscript for submission. In-kind support was provided by Fresenius Kabi Deutschland, which supplied both of the enteral nutrition formulations. Representatives from Fresenius Kabi Deutschland reviewed and provided feedback on the manuscript before submission; however, the authors on the writing committee wrote the manuscript, made the decision to submit it for publication, and vouch for the accuracy and completeness of the data, held by Monash University, and for the fidelity of the trial to the protocol, which has been published elsewhere20 and is available at NEJM.org. Ethics approval was provided by all relevant local institutional review boards (Fig. S1 in the Supplementary Appendix). An independent data and safety monitoring board provided trial oversight.
Patients 18 years of age or older in the ICU were eligible for inclusion if they were receiving invasive mechanical ventilation, were about to commence enteral nutrition, or had commenced enteral nutrition within the previous 12 hours and were expected to be receiving enteral nutrition in the ICU beyond the calendar day after randomization. Patients for whom the treating clinician considered the trial enteral nutrition formula or the rate of delivery to be clinically contraindicated or in whom death was deemed inevitable were excluded. A full list of the exclusion criteria is provided in the Supplementary Appendix.
Randomization and Treatment
Using permuted block randomization and variable block sizes with stratification according to site, we randomly assigned eligible patients in a 1:1 ratio to receive energy-dense or routine enteral nutrition. Concealment of the treatment assignments was maintained with a secure, Web-based randomization system, which was accessible 24 hours a day. Both the energy-dense enteral nutrition (1.5 kcal per milliliter, Fresubin Energy Fiber Tube Feed) and the routine enteral nutrition (1.0 kcal per milliliter, Fresubin 1000 Complete Tube Feed) were administered in identical 1000-ml bags (Fig. S2 in the Supplementary Appendix). The formulations were indistinguishable in color and packaging.19 The difference in calorie content between the energy-dense and routine formulations was shared between fat (58 g per liter in the energy-dense formulation vs. 27 g per liter in the routine formulation) and carbohydrates (180 g per liter vs. 125 g per liter); the protein content of the two formulations was similar (56 g per liter and 55 g per liter). Full product information is provided in Figure S3 in the Supplementary Appendix.
Administration of the trial enteral nutrition was commenced as soon as possible after randomization. The target rate for both groups was 1 ml per kilogram per hour and was based on the calculated ideal body weight (see the Supplementary Appendix).15,17,21 We recommended that the target rate be achieved within 48 hours after commencement of the trial nutrition. A clinician estimation of baseline energy requirements was not used to determine the target rate for the trial; however, when such an estimation was performed, we collected the information. To minimize the risk of overfeeding, the maximum target rate was 100 ml per hour; catch-up feeds were not permitted. Blood glucose concentrations of 180 mg per deciliter (10 mmol per liter) or less were recommended. All other aspects of management were handled according to local practice, including the rate at which trial nutrition was commenced and incremented, the method and frequency of measurement of gastric residual volumes, and strategies to increase delivery. If the treating clinician deemed supplemental parenteral nutrition necessary, the trial enteral nutrition was continued unless contraindicated.
The trial enteral nutrition was administered for up to 28 days or until the patient discontinued enteral nutrition, died, or was discharged from the ICU, whichever occurred first. In addition, the trial enteral nutrition was ceased if specific nutritional requirements developed, including the need for protein supplements; if the patient commenced oral nutrition; or if the trial enteral nutrition was no longer deemed to be in the patient’s best interest. Patients who were readmitted to the ICU within 28 days and still required enteral nutrition had feeding with their previously assigned formulation restarted.
The primary outcome was all-cause mortality within 90 days after randomization. Secondary outcomes included survival time (evaluated until day 90), 90-day cause-specific mortality, day 28 and in-hospital all-cause mortality, ICU-free and hospital-free days between randomization and day 28, the number of days free of organ support between randomization and day 28, and the percentages of patients receiving invasive ventilation, vasopressors, or new renal replacement therapy. Other secondary outcomes were the percentage of patients with positive blood cultures and the percentage receiving intravenous antimicrobial agents between randomization and day 28. On the basis of prerandomization variables, seven subgroups were predefined for the evaluation of the primary outcome: age (≥65 or <65 years), diagnostic subgroups (trauma, sepsis,22 a neurologic diagnosis, and treatment type [medical vs. surgical]), quintiles for the absolute risk of death based on the Australian and New Zealand Risk of Death Score after linkage to the ANZICS Center for Outcome Resource Evaluation (CORE),23,24 and body-mass index (BMI, the weight in kilograms divided by the square of the height in meters) according to the World Health Organization categories (<18.5, 18.5 to 24.9, 25.0 to 29.9, and ≥30.0).25
All analyses were conducted in accordance with our prepublished statistical analysis plan.26 On the basis of data from the TARGET feasibility study and the ANZICS CORE Adult Patient Database, we calculated that a sample of 3774 patients would provide 80% power to detect a difference of 3.8 to 4.3 percentage points in 90-day mortality, assuming a baseline mortality of 20 to 30%.19,23 A 6% sample size inflation to 4000 patients who could be evaluated allowed for anticipated losses during follow-up and for one interim analysis. The interim analysis was performed after completion of the day 90 follow-up of the first 1500 patients with the use of a twin-sided O’Brien–Fleming design and a two-sided P value of 0.005 and was reviewed by the data and safety monitoring committee.
All analyses were conducted with the use of a modified intention-to-treat principle.27 Per-protocol and as-treated sensitivity analyses in the modified intention-to-treat population were performed for the analysis of the primary and secondary outcomes (see the Supplementary Appendix). No imputation was used to estimate missing data, and analyses were based on all available data with numbers of available observations reported. The methods for calculating daily delivery of nutrition and gastrointestinal tolerance of enteral nutrition are described in the Supplementary Appendix. Continuous variables are reported as means and standard deviations or as medians and interquartile ranges. Categorical variables are reported as percentages. Between-group differences were analyzed with Student’s t-test or Wilcoxon rank-sum tests for continuous variables and a chi-square test for categorical variables and are reported as estimated mean difference, median difference (Hodges–Lehman estimate), or relative risk with 95% confidence intervals.
We report the relative risk and 95% confidence interval for death from any cause by day 90 using log-binomial regression with adjustment for site (random effect) and for predefined baseline covariates (age, Acute Physiology and Chronic Health Evaluation [APACHE] II score at ICU admission, BMI, country [Australia or New Zealand], sex, and ICU admission type [medical, elective surgical, or emergency surgical]) (fixed effects). The same unadjusted and adjusted analyses were performed for 28-day and in-hospital mortality. Modified Poisson regression with robust standard errors was used to estimate the relative risk when log-binomial models did not converge. Survival time, evaluated from randomization to day 90, is shown as a Kaplan–Meier curve and compared with the use of a log-rank test. Hazard ratios with 95% confidence intervals were obtained with the use of Cox proportional-hazards models. Numbers of ICU-free, hospital-free, and organ support–free days are reported as medians and interquartile ranges. The methods used for subgroup analyses are described in the Supplementary Appendix. Analyses were performed with SPSS Statistics software, version 22 or later (IBM), and Stata software, version 15.1 (StataCorp). No correction for multiplicity when conducting tests for secondary and other outcomes was predefined; the results are reported as point estimates and 95% confidence intervals. The widths of the confidence intervals have not been adjusted for multiplicity and should not be used to infer definitive differences in treatment effects between the groups.