Supplement 1. Introduction and Methods

Supplement 1. Introduction and methods.

Introduction

Existing guidelines recommend initiating enteral nutrition (EN) within the first 24-48 h after intensive care unit (ICU) admission if patients are unable to eat, provided that their gastrointestinal (GI) tract is functioning and haemodynamics are stable [1,2,3] without clearly defining “functioning” and “haemodynamics”. Accordingly, reasons to delay EN are not uniformly defined. The rationale against full EN in the early phase of critical illness is that the inflammation-driven endogenous production of calories cannot be suppressed with nutrition, whereas probably harmful suppression of autophagy occurs [4,5]. On the other hand, observational studies report that greater calorie and protein debt is associated with worse clinical outcomes [6,7]. Studies report conflicting results with early supplemental parenteral nutrition (PN) for circumstances where EN does not reach energy targets [8,9,10].

Available guidelines define early EN as nutrition started within 24-48 h of ICU admission [1-3,11]. The general consensus is that early EN, at least at a trophic dose, is of benefit [12]. Early EN contributes to the maintenance of gut barrier and gut derived immunity [13].

Doig and colleagues [14] suggested that EN initiated within 24 hours of ICU admission might be associated with improved survival, but meta-analyses of studies that started EN later than this failed to show significant differences in mortality [3, 13,15,16].

Despite several existing clinical nutrition guidelines, there is a lack of recommendations as to when to commence EN, or at least when to challenge the GI tract for the tolerance of EN, in patients with certain hemodynamic instability, or specific abdominal pathologies. As a consequence, EN may be unnecessarily delayed, which risks a worse patient outcome.

Our objective was to provide evidence-based guidelines for early EN in critically ill patients, focusing on specific clinical conditions that are frequently associated with delayed EN. We did not address caloric and protein requirements, the time to reach targets, the type of EN, the different routes to provide EN and the timing of supplemental or full PN.

Group members

This task was formulated and completed by the Working Group on Gastrointestinal Function within the Metabolism, Endocrinology and Nutrition (MEN) Section of European Society of Intensive Care Medicine (ESICM). A methodologist was engaged from outside the Working Group. The group meetings were held during Annual Congresses of ESICM to plan for literature review and development of tables for evidence synthesis. The group continued working via email, teleconferences, and online webinars.

Management of conflict of interest

There was no industry input into guidelines development. No industry representatives were present at any of the meetings. Any possible issues of conflict of interest were discussed, none of them was considered relevant regarding the present guidelines because specific formulae are not mentioned. However, most of the group members received funding for research or fees for lecture or advise from nutritional companies. These specific nutrition-related conflicts of interest are reported separately in Supplement 8.

Methods

Questions Development

We performed a systematic review and meta-analysis comparing early EN vs. early PN and early EN vs. delayed EN. The rationale for this approach was to first determine whether EEN is advisable in the heterogeneous cohort of critically ill patients and, thereafter, to subsequently investigate whether there are exceptions to our general recommendations. Based on consensus among the authors reached during the Working Group meetings and e-mail discussions, we predefined conditions in which EN is frequently delayed and structured each question in PICO format (population, intervention, control, outcome), prioritising the outcomes based on their importance to patients [17]. The conditions defined were related to 1) major organ dysfunction and metabolic derangements outside the GI tract (shock, respiratory failure, acidosis, hypothermia), 2) GI dysfunction or function at risk (pancreatitis, abdominal trauma and/or surgery, bowel ischemia/abdominal compartment syndrome, GI bleeding, diarrhoea, fistulas) and 3) perceived increased aspiration risk not necessarily related to specific GI problems (prone position, neuromuscular blocking agents, neurological pathology, large gastric aspirate volume).

Definition of “Early EN”

The cut-off time for “early” EN was not predefined to avoid exclusion of valuable data. Instead, a critical appraisal of identified studies for Question number 1 (general ICU population) was undertaken and following review of this evidence base (see Supplement 5) we defined “early” EN as EN commenced within 48 hours of admission. “Early” EN in included studies refers to the initiation of any enteral nutrition, independent of the constituents or amount of EN.

Outcome Prioritization

We focused on patient-centred outcomes, which included mortality and infectious complications. We did not perform formal rating of the outcomes, instead, we selected outcomes based on the guideline group discussion and input.

When mortality was reported at different timelines in eligible studies, we pooled the longest follow-up period at which mortality was reported.

Search Strategy

Following group agreement of the questions and outcomes, we formulated a search strategy for each question. For each question, two reviewers independently searched Pubmed, MEDLINE and Web of Science databases from inception to December 2015. We applied no date restrictions, but restricted language to English. Abstracts without full-text were excluded. Additional relevant papers were scanned for related articles and reference lists of assessed papers for further full texts. Only studies addressing EN were extracted into evidence tables, studies with early oral intake were not considered. For each sub-question, two reviewers screened titles and abstracts of potentially eligible studies. Procedure of identification of relevant papers is presented in Supplement 2. The aim of each search was to identify the highest quality evidence available. Eligibility criteria were defined as 1) study patients with respective condition requiring admission to the ICU (traumatic brain injury); 2) study groups defined according to respective study question (EEN vs. early PN or EEN vs. delayed EN, irrespective of specific feeding formula or dose); 3) outcome reported. If available, we included individual RCTs that met eligibility criteria and performed a meta-analysis. If no eligible RCTs were identified, we searched for observational studies relevant to the question.

Key words and results for literature searches for different sub-questions are presented in Supplement 3, Pubmed search formulas in Supplement 4, and PRISMA Checklist [18] in Supplement 9.

Analysis

After individual RCTs were identified and relevant data were available, we generated pooled estimates for each outcome using meta-analytic techniques. We used random-effect model and the method of DerSimonian and Laird to generate pooled estimates across studies [19]. Weights were estimated using Mantle-Haenszel method in random-effects model. If fewer than three studies were included in a meta-analysis, we used a fixed-effect model [20]. We used risk ratio (RR) to report effect estimates for dichotomous outcomes along with 95% CI. We used the chi squared and I2 tests to assess statistical heterogeneity [21], we considered a chi squared p-value <0.1 or I2 >50% as statistically significant heterogeneity. We analysed data using RevMan software (Review Manager, version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014).

Risk of bias assessment

We used the Cochrane Risk of Bias Tool to assess risk of bias in randomised controlled trials (RCTs) [22]. Studies were classified into high or low risk of bias, based on the assessment of following categories: selection, comparability, and outcome.

Quality of Evidence Assessment

The methodologist, with input from group members, assessed the quality of evidence for each outcome in each question using the GRADE (Grading of Recommendations Assessment, Development and Evaluation) methodology [23].

Evidence profiles were created for each question with pooled estimates, making judgments on risk of bias, inconsistency, indirectness, imprecision and other possible considerations. We downgraded the quality of evidence for risk of bias due to lack of or inappropriate blinding when the outcome was subjective and the results could be influenced by bias. We accepted some degree of indirectness related to the intervention and population across different studies, we judged the impact of this variation to be minimal and did not downgrade the quality of evidence for indirectness. Due to small number of published studies for some questions, optimal assessment of publication bias was not feasible. However, we limited the bias by performing a comprehensive search and included experts in the field that could identify important non-published studies.

We considered that observational studies evaluating early EN were all intrinsically biased, because patients who are less severely ill are more likely to receive and tolerate early EN. Accordingly, although all observational studies were critically assessed during the study process and helped to form expert opinion, we did not formulate our recommendations based solely on observational studies.

We used strict criteria before using expert opinion to formulate our recommendations according to the following: a) RCTs were not available or the results did not allow decision in favour or against early EN; b) very low quality observational studies were identified and critical appraisal of the results suggested inconclusive or potentially misleading results; c) there is large need for advice to clinicians in practice. For the purpose of this guideline, expert opinion is considered as very low quality evidence, therefore, we did not issue any strong recommendations based on expert opinion.

Taken together, if randomised controlled trials were available we gave an evidence-based recommendation; if not, our recommendations were based on expert opinion (very low quality of evidence).

Evidence to Decision Framework

Over six Webinars we used the Evidence to Decision (EtD) framework [24] to facilitate the formulation of the recommendation. In each webinar the methodologist facilitated the discussion between working group members on each of the components of EtD framework. The framework takes into consideration the following factors: quality of evidence, magnitude of benefit and harm, balance between desirable and undesirable consequences, patients preferences, resources, cost, feasibility, and acceptability.

Definitions on grading the certainty of evidence and explanation of consequences of recommendations are provided in Tables 1 and 2 in this Supplement.

Table 1. Certainty of evidence

Rating / Definition
High / A / High confidence that the true effect lies close to that of the estimated effect.
Moderate / B / Moderate confidence in the estimated effect. The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.
Low / C / Low confidence in the estimated effect. The true effect may be substantially different from the estimated effect.
Very Low / D / Very low confidence in the estimated effect. The true effect is likely to be substantially different from the estimated effect

Table 2. Consequences of strong vs. conditional recommendations for different stakeholders

Strong Recommendation (GRADE 1) / Conditional Recommendation (GRADE 2)
For patients / Most individuals in this situation would want the recommended course of action and only a small proportion would not. / The majority of individuals in this situation would want the suggested course of action, but many would not.
For clinicians / Most individuals should receive the recommended course of action. Adherence to this recommendation according to the guideline could be used as a quality criterion or performance indicator. Formal decision aids are not likely to be needed to help individuals make decisions consistent with their values and preferences. / It needs to be recognized that different choices will be appropriate for different patients, and that for each patient a management decision consistent with her or his values and preferences must be reached. Decision aids may well be useful helping individuals making decisions consistent with their values and preferences. Clinicians should expect to spend more time with patients when working towards a decision.
For policy makers / The recommendation can be adapted as policy in most situations including for the use as performance indicators. / Policy making will require substantial debates and involvement of many stakeholders. Policies are also more likely to vary between regions. Performance indicators would have to focus on the fact that adequate deliberation about the management options has taken place.

Patients preferences

We did not identify any studies addressing the values and preferences of patients in the clinical context of interest. We did not include patient representative on this guideline. However, working group members considered and discussed the values of preferences of patients from their point of view. The discussion was focused on the trade-off between benefit and harms or burden of the interventions. The larger the variability in values and preferences, the more likely we were to issue a weak recommendation.

Resources and cost

The costs and resources needed for EN vs PN were not formally analysed, but were discussed during the development of this guideline and included in expert opinion. Decisions in favour of EN were supported, but never based solely on the evaluation that fewer resources are needed for EN when compared with PN. However, we discussed the potential resources required for each intervention as this contributed to the clinical relevance of the recommendation.

Balance between desirable and undesirable effects

Finally, the working group members considered all the above-mentioned factors and decided if the desirable consequences outweigh the undesirable consequences and how confident they were about this balance. The more confident we were about the magnitude and direction of benefit, the more likely we were to issue a strong recommendation, and vice versa. We used the GRADE working group methods to guide this process [25].

For all our final recommendations we aimed to reach consensus within the group. Therefore, to allow input from all group members, we organised two electronic Delphi rounds [26]. After the first round, an anonymous summary of results of the previous round as well as the reasons that were provided for disagreement with proposed recommendation were sent to all co-authors. At the same time, a new questionnaire with only the questions where consensus was not reached, was also distributed. If consensus was not reached after the second Delphi round, a predefined level of 80% agreement was considered necessary to give a recommendation [27].

Description of methodology for Question 1: Should we recommend early EN in critically ill adult patients?

We identified 30 RCTs. We performed separate analyses to compare “Early EN vs. Early PN” and “Early EN vs Delayed EN” using mortality and infection outcomes. For mortality, any time period was included but “hospital mortality” was preferred. For infection, we preferred “any infection”, followed by “pneumonia”. Some studies reported only specific infections (e.g. pancreatic infections), which were analysed under “any infections” outcome.