Background
Patients managed in the intensive care unit (ICU) must receive nutritional support to ensure that their energy and nutrient needs are met. There is evidence that ICU-acquired malnutrition is associated with worse patient outcomes [
1]. In observational studies, providing adequate amounts of calories and/or protein is associated with decreased mortality [
2,
3]. However, large randomised controlled trials (RCTs) aimed at determining the best timing, energy target, and route of administration of nutritional support either found no efficacy or produced apparently contradictory results [
4‐
9]. These discrepancies are an obstacle to developing consensual clinical nutrition strategies, and to adherence of physicians to clinical guidelines. One likely source of discrepancies is the variability in the outcomes selected to assess the efficacy of nutritional interventions in RCTs. This variability hinders the interpretation of results, comparisons of RCTs, and the development of strong evidence-based recommendations.
In theory, the benefits of adequate nutrition in critically ill patients should include lower risk of acquired infections including wound infections, pressure sores, and muscle mass loss; shorter duration of mechanical ventilation; shorter ICU and hospital stay; and perhaps lower mortality. Nevertheless, no accurate and robust data are available on the benefits of adequate nutrition or adverse effects of inadequate nutrition in critically ill patients. To obtain such data, the optimal outcome for assessing nutritional interventions must be determined. This is a challenging task, as shown by the broad range of outcomes used in RCTs. For example, the mortality rate is often used as the primary outcome in RCTs performed in critically ill patients, including those evaluating nutritional interventions. However, variability in responses to the intervention across patient subsets, with lower mortality in some subsets and higher mortality in others, may produce a net result of no effect. Although the ultimate goal of improving the management of critically ill patients is to improve survival several weeks after the ICU stay, whether the likelihood of achieving this goal can be increased by a short-term nutritional intervention is debatable. Developing a structured, clinically relevant, consensual, and validated RCT methodology may help to better define primary and/or secondary outcomes reflecting improvements in patient outcomes related to nutritional interventions for use in future RCTs.
The objective of this study was to obtain a clear picture of the outcomes used in published RCTs of nutritional interventions in the ICU, as a first step towards selecting optimal outcomes. To this end, we conducted a systematic review of the recent literature.
Discussion
The aim of this study was to describe the outcomes used in recent RCTs evaluating clinical nutrition interventions in critically ill patients. The most commonly used primary outcome was the complication rate, which was related to both the efficacy (infectious complications) and tolerance (metabolic complications and feeding intolerance) of the nutritional intervention. Mortality and duration of organ dysfunction were the primary outcomes in a quarter of the trials, whereas functional outcomes were only very rarely used.
The complication rate may seem relevant to an assessment of both efficacy - as adequate nutrition is believed to decrease the risk of infectious complications - and good tolerance with fewer metabolic and feeding complications. However, the complication rate reflects only limited and expected effects of nutritional interventions. In addition, the definition of complications is not standardised.
In the most recent and largest RCTs, mortality was widely used as the primary outcome. Mortality is an objective outcome that is easy to collect, free of interpretation bias, and clinically relevant. These advantages may increase the chances of obtaining funding. However, the use of six different mortality time points in the RCTs included in our analysis hampers comparisons across studies. Furthermore, mortality as the primary outcome is not necessarily associated with quality of care [
13]. A single and usually short-term nutritional intervention may be unlikely to improve long-term mortality in patients with severe critical illness. As with many RCTs in critically ill patients, those using mortality as the primary outcome showed no significant differences between groups. One possible explanation is that a nutritional intervention may have benefits in some patients but cause harm in others. Another is that predicted mortality varies widely across ICU patients included in RCTs.
Mortality and ICU and hospital length of stay are often used as outcomes in studies of critically ill patients. However, survivors of critical illness may experience profound physical, cognitive, and psychological impairments that may alter their long-term quality of life [
14]. Outcomes other than mortality that are likely to be important for patients, i.e., patient-important outcomes, were rarely used in the RCTs included in our review, although they allow a more accurate estimate of the long-term health burden of critical illness [
15]. Assessing functional outcomes has long been viewed as challenging, due to differences across patients in baseline parameters (before ICU admission) and to the absence of validated tools for evaluating post-ICU function. The emerging concept of frailty in critically ill patients refers to a limitation of physical and cognitive reserve that impairs the ability to cope with stressors, thereby increasing the risk of adverse events during the ICU stay. The frailty score categorizes patients according to their pre-ICU status and helps to assess the risk of lower survival and of poorer intermediate and long-term functional outcomes.
Considering functional outcomes in future trials may better define the benefits of nutritional interventions. Muscle strength can be assessed using dynamometry or the Medical Research Council Scale [
16] and muscle mass by ultrasound or computed tomography. Tools available for assessing physical function include the 6-min walking test, Chelsea Critical Care Physical Assessment Tool (CPAx) [
17], and Functional Status Score for the Intensive Care Unit (FFS-ICU) [
18,
19]. Functional self-sufficiency can be evaluated using instruments such as the Barthel Activities of Daily Living Index [
20] or the Functional Independence Measure [
21]. Quality-of-life scales may also be useful.
The RCTs included in our study used a broad diversity of outcomes, with six different categories, each containing several outcomes. Furthermore, one out of five RCT reports did not specify whether the outcomes were primary or secondary. This heterogeneity may reflect differences in study populations and interventions. Nutritional interventions cover a vast spectrum of treatments that differ in factors such as the energy target, composition and route of administration of the preparations, and timing of delivery. This last point is of special importance. Although the acute and chronic/recovery phases are challenging to define in clinical practice, similar nutritional interventions probably achieve different goals according to the individual patient’s metabolic response to the insult. Thus, the wide diversity in outcome measures identified in the present study can be ascribed to differences in research objectives, patient populations, and interventions.
Nevertheless, this heterogeneity of outcomes may complicate the interpretation and comparison of results across studies, thereby limiting the ability to conduct valid meta-analyses. It therefore indicates a need to develop a consensus about a minimum core outcome set (COS) for critical-care nutrition trials, to improve the consistency of outcome selection and measurement. The wide spectrum of outcomes identified in our study shows that much work will be needed to obtain such a consensus. In the field of paediatric critical care, the Core Outcome Measures in Effectiveness Trials (COMET) initiative is working towards identifying core outcome measures [
22]. As most studies in paediatric critical care are small and conducted at a single centre, the ability to combine data sets for analysis is of paramount importance. The COMET initiative will systematically describe the outcome measures used in these paediatric studies then conduct an international Delphi study to agree on a standard set of core outcomes for future trials. Our review was conducted as a preliminary to developing a set of valid, reliable, and feasible core outcomes for adults, with the goal of improving the quality of research results by decreasing study heterogeneity and outcome-reporting bias and by increasing the statistical power of meta-analyses. The development of a COS is supported by the COMET group and has been endorsed by the Cochrane Library, World Health Organisation, and Grading of Recommendations Assessment, Development and Evaluation (GRADE) working group [
23,
24]. The use of a COS for clinical nutrition research in ICUs may help to compare nutritional strategies, effectively pool data from different studies on the same condition, and encourage more complete reporting of outcomes.
When developing and applying a COS for critically ill patients and for specific nutritional interventions, it is important to consider the setting. For example, studies of nutrition in the ICU may include death as an essential outcome. However, as overall mortality declines with advances in intensive care, mortality may be low in all groups and therefore less relevant to comparisons of treatment strategies. Other outcomes, such as quality of life at discharge and muscle strength, may be more relevant. Therefore, once developed, the COS should be reappraised continually in the light of changes in practices, techniques, and patient outcomes. Furthermore, and importantly, the COS should be viewed as the minimum set of parameters that must be measured and reported. If parameters not included in the COS are relevant in the setting of a given study, they should also be measured and recorded. Such additional parameters may be particularly likely to exist in critical care, given the complexity of the conditions seen in ICU patients.
To harmonise outcome selection for studies of nutrition in the ICU, we are planning an international Delphi consensus process in which experts will be asked about outcome assessments appropriate for different study objectives, and patients will be asked about the outcomes they believe are most meaningful. Importantly, the present study did not focus on heterogeneity in outcome definitions but, instead, on the diversity of outcomes across trials. The main source of bias in the RCTs included in our study was absence of blinding of the patients and staff.
A potential limitation is that our study was confined to RCTs published in or after 2000. However, our goal was to capture the most recent practices in RCT design. In addition, very few RCTs published before 2000 focussed on nutrition in ICU patients. Our objective was to obtain an overview of the outcomes used and their definitions, regardless of the trial results, which our study was not designed to evaluate.