Background
Critically ill patients are at high risk of morbidity, mortality and prolonged care needs [
1]; implementing practical approaches to improve outcomes is of paramount importance [
2]. During critical illness, evolutionary survival mechanisms release energy from stored body tissues to fuel life-supporting tasks [
3]. The sacrifice from body stores is deleterious, and contributes to poor outcomes: when energy and protein is delivered to critically ill patients, this risk is ameliorated and recovery potential improved [
1,
4,
5].
Despite the potential benefits of nutrition therapy, feeding is stopped intermittently in 85% of critically ill patients due to essential procedures and symptoms of feed intolerance [
1]. These feed stops cause patients to meet only 40–60% of their energy and protein requirements, rather than allowing optimal delivery and meeting the minimum 80% recommended by clinical practice guidelines (CPGs) [
1,
6].
Most intensive care units (ICUs) worldwide use an hourly ‘rate-based’ feeding (RBF) approach, without strategies to rectify feed deficits. Evidence suggests changing from a rate- to volume-based feeding (VBF) approach helps mitigate these accrued deficits without increasing feed intolerance in medical, and some surgical ICU patients [
7‐
12], and by corollary, may improve clinical outcomes. Using the VBF approach, instead of prescribing an hourly feeding rate of, for example, 50 ml/h, a patient is prescribed 1200 ml/24-h period; systems are put in place to ensure the entire amount is delivered within 24 h.
CPGs encourage adoption of VBF in local ICU practice [
1,
6]. The Heyland group [
10‐
12], who named their VBF protocol: The
Enhanced Protein-Energy Provision via
the Enteral Route in Critically Ill Patients (PEPuP) protocol, openly share and encourage use of their resources to facilitate change in local ICUs, and endorse adapting their protocol to local context to facilitate implementation [
13].
Not all studies show improved outcomes when energy and protein goals are achieved. The VBF studies by Haskins et al. [
8], Taylor et al. [
9] and Heyland et al. [
11] reported no significant difference in LOS, mortality or ventilation was enabled by VBF. Some other studies evaluating overall nutritional delivery in the critically ill population [
14‐
17] suggest reductions in mortality, ventilation period and LOS may be achieved when patients meet their energy and/or protein goals. Others warn that meeting ICU feeding targets triggers complex biochemical processes which have the opposite effect: increasing mortality, ventilation duration and LOS [
18,
19], and recommend a less aggressive approach to feeding.
Three recent meta-analyses [
20‐
22] found no difference in mortality, ventilation or hospital- or ICU-LOS when patients did or did not meet their energy or protein needs. The randomised controlled trials included in analyses used various strategies to improve nutritional delivery, such as faster feed starts and rate increase to ameliorate deficits [
19], or managing gastric residual volumes or small bowel feeding. All meta-analyses found patients met sub-optimal levels of protein, which may be important. Emerging evidence suggests that with an increased turnover of up to 80% in critical illness, ICU patients have a much larger protein need than previously accepted [
23‐
27]. It is common for protein to be targeted secondary to total energy in this patient group and is considered a neglected area of research [
28,
29].
Deciding the optimal dose and timing of meeting the energy and protein needs of ICU patients remains controversial and subject to the impact of variable influences [
30]. ASPEN [
1] advise assessing for malnutrition risk, using either the Nutritional Risk Screening (NRS) 2002 or the Nutrition Risk in the Critically ill (NUTRIC) scores. NUTRIC was developed by Heyland et al. [
31] to identify which ICU patients benefit most from nutritional support. The score has been externally validated in large prospective observational trials and found to identify the sickest patients more likely to have increased morbidity and mortality [
32,
33]. Evidence suggests the NUTRIC score predicts energy and protein deficits in critically ill patients, but NRS does not [
34], and that energy- and protein-related improvements in mortality are greatest in those patients with longer stay, and at highest risk calculated using NUTRIC [
33,
35].
In January 2016, a Local Health Board (LHB) medical-surgical ICU multidisciplinary team commenced a prospective before-and-after study designed to compare nutritional delivery between RBF and VBF. The study was designed with the primary aim of confirming the hypotheses: that changing from RBF to VBF would significantly increase the percentage of prescribed feed volume, energy and protein delivered to adult critically ill patients, without altering feed tolerance. Secondary observations of interest included between-group comparisons of patients’ outcomes including mortality, length of ICU stay (LOICUS) and mechanical ventilation.
Methods
Permissions to undertake study
The study did not require informed patient consent: the system-level quality improvement initiative was designed to undertake a minimal-risk change in feed process which did not exceed the boundaries of standard clinical care, and could not take place practically if prior consent were required [
11,
36]. The LHB ‘Research and Development’ department consented to the work as a service evaluation project without need to pursue ethical review. The required University Healthcare Sciences and Medical Sciences Academics Ethics Committee approval was obtained before data analysis.
ICU characteristics
The adult, medical-surgical ICU is within a district general teaching hospital comprising 600–700 beds. Staffing is provided in a one-to-one nurse-to-patient ratio, and patients are overseen by a Consultant Intensivist. In the years spanning 2012–2015, quarterly ICU admissions were consistent at 181–204 patients, and average length of mechanical ventilation was 3.7–4.0 days [
37].
Recruitment
Data collection was undertaken prospectively in consecutively admitted, adult (≥ 18 years) patients who were mechanically ventilated for 72 h or more and fed for at least 48 h. The Local Health Board intensive care unit (LHB-ICU) does not currently use NUTRIC, and the 72-h duration was selected a priori as a method of sample restriction to define a level of disease acuity and longer stay.
Enteral feeding was commenced within 24 h of ventilation in stable patients. Only patients deemed clinically appropriate to receive full feeding by the medical or surgical team were included. VBF was undertaken from day 2 onwards, or once a patient was considered suitable to meet full-volume feeds. Patients initially nil by mouth, prescribed trophic feeding, or fed cautiously due to a risk of refeeding syndrome, were included if they were able to progress to full feeding within 72 h of ventilation.
Data was collected for up to 7 days, cessation of mechanical ventilation, death or ICU discharge, whichever occurred first.
Patients were excluded if they were pregnant, and/or were receiving parenteral or oral nutrition to limit the confounding effect of alternative nutritional support [
5].
Energy and protein requirements
The following principles were adhered to throughout the RBF and VBF periods.
Outside the dietitian’s working hours, the ICU used a ‘starter feeding regimen’ devised to closely meet the American Society for Parenteral and Enteral Nutrition (ASPEN) [1] energy and protein recommendations. It used a high protein, 1 kcal/ml feed containing 6.26 g of protein/100 ml for most patients (Osmolite HP [
38]), or an isocaloric, lower protein, renal-conserving [
39] feed for patients with established chronic kidney disease (Osmolite). The dietitian changed the feed prescription if required following assessment, using ASPEN [
1] or other relevant guidelines [
39,
40], and prescribing additional protein supplements when indicated (Additional file
1: Table S1).
Feed was progressed to target rate within 6 h of starting feed, unless prescribed trophic feeding, or considered at risk of refeeding syndrome, when feed targets were met gradually [
1].
The ‘PERFECT’ feeding protocol
The VBF protocol was adapted from PEPuP [
13] and entitled:
Protein & Energy Requirements Fed for Every Critically ill patient every Time (PERFECT); unlike PEPuP, baseline semi-elemental feeds, protein supplements and prophylactic prokinetics were not used.
The PERFECT toolkit instructed nurses how to increase feed rates, (maximum 150 ml/h) to compensate for feed-stops, and return to the initial goal rate at the beginning of the ICU 24 h period. For example, a patient prescribed 1200 ml would commence feeding at 50 ml/h at 0800 h; if the patient’s feed was off for 8 h, they had received 400 ml of feed prior, and there remained 8 h in the day on recommencing feeding, the deficit 800 ml (1200–400 = 800) would be given over 8 h at 100 ml/h (800/8). The 50 ml/h rate would recommence at 0800 h. A single end-of-day feed bolus up to 200 ml was given to replace remaining deficits. Boluses were not administered to jejunally fed patients.
Patients’ heads were elevated to 30–45° to reduce aspiration risk, and gastric-residual volume (GRV) was checked every 4–6 h. The ICU accepts and replaces GRVs up to 500 ml, with no change in feeding rate in the absence of other signs of intolerance.
Ward education was delivered by nurse-champions and the dietitian throughout December 2016 at daily and weekly team meetings. ‘How to’ booklets were kept at each bedside. One-to-one education and feedback was provided at the bedside, and continued ad hoc as required. Nursing daily documentation charts included an area to document feed deficits and corrections.
Data collection
Baseline RBF data was collected prospectively for 7 months from April 2016. PERFECT was implemented January 2017, and data again collected prospectively in consecutive admissions for 6 months.
Data included age, gender, weight, height and body mass index (BMI) in kg/m2; ideal body weight (IBW) if obese, daily energy and protein requirements, the calories and grammes of protein prescribed per kilogramme, hours without feed, and the feed-volume prescribed and delivered in millilitres. The mean daily percentage of prescribed feed-volume, protein and energy delivered (including energy from propofol) was calculated for each patient, based on minimum requirement. Each patient’s mean daily kilocalories per kilogramme and grammes of protein per kilogramme delivered were noted. A whole day of ‘0’ energy and protein delivery was included as 0% achieved.
Total episodes of witnessed vomiting (gastric contents external to mouth) and regurgitation (gastric contents within the mouth) were noted; the expression ‘vomit’ hereon includes both. Mean daily episodes for patients who vomited were calculated. Patients with diarrhoea were noted. Three or more daily liquid stools were classified as diarrhoea using the World Health Organization definition [
41], based on nurse perception of type 6–7 stools using the Bristol Stool Chart [
42].
Daily patient GRV (millilitres) and the amount replaced were recorded. Prior to commencing VBF, the ICU changed from using 8-French (Fr) and 10-Fr NGTs (used in the RBF period) to using 12-Fr tubes, which withdraw substantially more GRV [
43,
44], making the planned between-group comparison of aspirated volumes meaningless. Prokinetic prescription and the mean percentage GRV withdrawn and replaced per patient was compared between groups.
Subgroups of patients meeting < 80%, 80–89.9% and ≥ 90% of prescribed energy or protein were prepared for comparison, to explore any differences in clinical outcomes when patients achieved ‘over’ 80% of the ASPEN guideline recommendations, or specifically exceeded this.
Mean daily morning blood glucose and insulin requirement (mmol/L) per patient was noted, plus diabetes in past history as relevant to the frequency of hyperglycaemia [
9].
Clinical measures recorded from the Case Mix Program Database (coordinated by the Intensive Care National Audit & Research Centre) [
45] included the Acute Physiology and Chronic Health Evaluation II (APACHE-II) severity of illness score, advanced mechanical ventilation (days), ICU and 60-day hospital mortality (days) and length of ICU stay (LOICUS) (days: calculated from day 1 of ICU admission to when ‘ready for ICU discharge’ to account for delays in discharge); ICU admission diagnoses were summarised by surgery, respiratory, cardiovascular and ‘other’ (pancreatitis, gastrointestinal, neurological, sepsis, trauma, metabolic and haematological).
Statistical analysis
Power analysis
The primary outcomes of interest were energy and protein delivery and feed tolerance, while secondary observations of interest included ICU and 60-day mortality, ventilation period and LOICUS.
For the primary outcomes of interest, the improvements seen in protein and energy delivered to patients in the published VBF studies were classified as a medium-to-large effect size (0.70) for energy, and small-to-medium effect size (0.4) for protein. The G*Power 3 Power Analysis Program [
46], version 3.1.9.2, was used to conduct a priori analysis for one-tailed
t tests and Mann-Whitney
U using the estimated effect sizes, an
α-error level of 0.05 and an 80% power. A minimum sample requirement of 37 patients per group was noted.
Data management
IBM SPSS version 22 (IBM Corp., USA, 2013) [
47] was used in descriptive and inferential statistical tests unless otherwise stated. Statistical significance was accepted at the
α-error level ≤ 0.05; post hoc significance levels are cited in reporting.
Categorical variables are reported as counts and percentages. These were analysed for differences in proportional frequency between RBF and VBF groups using χ2 (chi-square) Test of Homogeneity, Test of Two Proportions, or Fisher’s exact test when cell counts were less than 5. Continuous variables are described by their means and standard deviations (±) when normally distributed, or by medians and interquartile range (IQR) when non-normally distributed. Mean and median group differences were compared using independent two-sample t tests for normally distributed data, or Mann-Whitney U for non-normal distributions.
Effect sizes are reported for percentage differences in the volume, protein and/or energy delivered between the RBF and VBF groups.
Equivalence between the VBF and RBF groups for mean episodes of vomiting was explored [
48,
49] using ‘two one-sided tests’ (TOST). NCSS version 11 (NCSS, LLC: USA, 2016) statistical analysis software was used with a pre-stated margin of equivalence of 20% of baseline [
48]. Combining both patient groups, binomial logistic regression was used to predict the probability of vomiting, adjusted for daily mean GRV, percentage feed volume delivered, and group.
Secondary outcomes of interest: 60-day survival, discharge and extubation rate were subjected to Kaplan-Meier and Cox Regression.
Kaplan-Meier for 60-day hospital survival used ICU admission date as start; censoring was based on hospital discharge alive or up to 60 days in hospital alive. For extubation rate analysis, patients who died on the ICU were excluded; day 1 of intubation was the starting point; extubation up to and including day 10 was the event, with ventilated patients thereafter censored. For LOICUS, censoring was undertaken after day 14.
Cox regression was adjusted for APACHE-II, group, and the percentage of energy or protein delivered; the covariate diagnosis of ‘respiratory disease’ was added to extubation-rate analysis [
50], and BMI 25–35 kg/m
2/< 25 and > 35 kg/m
2 to extubation-rate and mortality analyses.
Other
To identify predictors of increased mean morning blood glucose, insulin and propofol, multiple regression was adjusted for the percentage of prescribed energy delivered, APACHE-II, group, BMI and/or having diabetes.
Conclusion
The investigation found the PERFECT VBF feeding protocol significantly enhanced feed volume, energy and protein delivery to prolonged, mechanically ventilated patients in the LHB-ICU, without increasing feed intolerance. The exciting finding that enhanced protein delivery may improve ventilation is considered plausible, albeit requires further confirmatory study.
The approach is now embedded in daily practice on our ICU. As noted, patients in the PERFECT study were quite ‘typical’ of ICU admissions elsewhere, suggesting the findings will be useful to those working in similar medical-surgical ICUs considering adopting this approach. We recognise that our ICU patients met over 80% of protein and energy targets prior to commencing VBF: one might consider VBF unnecessary in such a cohort. As delineated in Fig.
2, using a VBF strategy significantly improved the likelihood of consistently achieving daily feed targets in all patients; therefore, we do consider this approach worthwhile in facilitating optimal energy and protein delivery regardless of baseline.
Large-scale research to demonstrate the safety and efficacy of VBF in other ICU populations has merit. That said, this analysis and the resulting discussion highlight that a unifying characteristic of many studies published thus far is a failure to optimise protein delivery: this may be key to improving ICU outcomes and suggests just doing ‘more of the same’ is not enough. Efforts at improving feeding in ICUs meeting low volumes are valuable, but these efforts should be aimed at increasing supplemented protein delivery, not just total energy.
The science which increasingly hints at patient groups more susceptible to the benefits of improved nutrient delivery, such as those achieving high NUTRIC scores, indicates a useful research target to trial VBF. Targeting such a group would help overcome the difficulty of ascribing credit to nutrition in improving outcomes in such a heterogeneous population. The evidence from NUTRIC studies so far does not show that feeding harms patients at low risk with short stays, and as it is often difficult to predict those who will have the longer ICU course, Compher et al. [
35] endorse continuing to optimise feeding in all patients, which this investigation has shown, is enabled by the PERFECT feeding protocol in our ICU. As noted, we do not calculate the NUTRIC score at present; this will be a recommendation for our Unit so we can be prepared to respond to emerging findings.
To optimise interpretation and generalisability, large, multicentre randomised controlled trials must be designed to measure outcomes related to improved protein delivery, using adequately powered samples for pre-specified effect sizes [
74], a priori determined patient outcomes, and subject to powerful statistical analysis.