Determining Optimal Outcome Measures in a Trial Investigating No Routine Gastric Residual Volume Measurement in Critically Ill Children

Background: Choosing trial outcome measures is important. When outcomes are not clinically relevant or important to parents/patients, trial evidence is less likely to be implemented into practice. This study aimed to determine optimal outcome measures for a trial of no routine gastric residual volume (GRV) measurement in critically ill children. Methods: A mixed-methods approach was used: a focused literature review, parent and clinician interviews, a modified 2-round Delphi, and a stakeholder consensus meeting. Results: The review generated 13 outcomes. Fourteen pediatric intensive care unit (PICU) parents proposed 3 additional outcomes; these 16 were then rated by 28 clinicians in Delphi round 1. Six further outcomes were proposed, and 22 outcomes were rated in the second round. No items were voted “consensus out.” The 18 “no-consensus” items were voted in a face-to-face meeting by 30 participants. The final 12 outcome measures were time to reach energy targets, ventilator-associated pneumonia, vomiting, time enteral feeds withheld per 24 hours, necrotizing enterocolitis, length of invasive ventilation, PICU length of stay, mortality, change in weight and markers of feed intolerance (parenteral nutrition administered), feed formula altered, and change to postpyloric feeds all secondary to feed intolerance. Conclusion: We have identified 12 outcomes for a trial of no GRV measurement through a multistage process, seeking views of parents and clinicians. ( JPEN J Parenter Enteral Nutr. 2020;00:1–8) from Nutrinia, outside the submitted work. The grant from Nutrinia in 2018 was for part of his salary to work as an expert advisor on a trial. J. Dorling was a member of the NIHR HTA General Board (from 2017 to 2018) and the NIHR HTA Maternity, Newborn and Child Health Panel (from 2013 to 2018). C. reports grants from Medical Research Council and the NIHR during the conduct of the study and grants from NIHR, Medical Research Foundation, Rosetrees Foundation and from Canadian Institute for Health Research outside the submitted work. He reports grants and personal fees from Chiesi Pharmaceuticals outside of the submitted work; the grant is for a research study and the personal fee was to attendance at an educational meeting. C. Gale is vice-chair of the NIHR Research for Patient Benefit London Regional Assessment Panel and has sat on the panel since 2016. F. V. Valla reports personal fees from BAXTER and personal fees from NUTRICIA outside the submitted work.


Introduction
Selecting the most appropriate outcome measures for a trial is increasingly recognized as important for trial evidence to be effectively implemented into clinical practice. 1 Ensuring trial outcomes are both clinically relevant and important to parents and patients is vital to optimize the impact of the research and to improve implementation and uptake into practice. 2 In addition, heterogeneity of clinical trial outcomes is a problem across clinical medicine but is more pronounced in relatively "small" specialties such as pediatric critical care, in which many trials are single centered 3 and underpowered. In such "small" specialties, synthesis of trial results is essential for evidence-based practice, but unfortunately, the pooling of trial results in meta-analyses is often not possible because of outcome heterogeneity. This work, as part of a larger funded feasibility study, sought to determine the optimal parent and clinician outcome measures for a future trial of no routine gastric residual volume (GRV) measurement to guide enteral feeding in pediatric critical care. The proposed future trial of no routine GRV measurement (compared with standard care involving regular GRV measurement) to guide enteral feeding in critically ill children was called the GASTRIC trial. 4 This was a National Institute for Health Research (NIHR)-funded mixed-methods feasibility study to determine whether it was possible to conduct a future trial in UK pediatric intensive care units (PICUs).

Methods
A mixed-methods study was undertaken between March 2018 and April 2019 as part of a larger NIHR Health Technology Assessment (HTA)-funded study (GASTRIC Reference16/94/02). The trial was registered on the ISRCTN registry (42110505). There were 4 phases involved in identifying the optimal outcomes: 1. A focused literature review to identify outcomes used in previous clinical trials and observational studies of GRV in any critically ill populations: including adults and preterm neonates. Databases were searched (Medline, CINAHL, ProQuest) in March 2018 by L.N.T. using the search terms: "gastric residual" and "gastric aspirate" in critically ill patients. For included studies, detailed data on each outcome measure used were extracted into a data collection form for summarizing. 2. Parent interviews: Qualitative researchers (L.R. and E.D.) recruited parents with experience of tube feeding in PICUs in the last 3 years though 4 routes. These included the use of social media, national contacts, word of mouth, and newspaper advertisements. Parents were excluded if they did not speak English. Once eligibility was confirmed, an interview date and time were scheduled. A participant information sheet, with a list of potential outcomes developed from the focused literature review, was emailed to parents to read prior to the interview. Based on previous research, it was anticipated that 10-15 parents would be recruited to reach data saturation point. 5 This is when the major themes identified in new data are recurring from previous interviews and no new major themes are being discovered in analysis. 6 Interviews explored views on the proposed GASTRIC trial. The last section of the interview included a series of questions to determine what outcomes parents felt would be important to measure in the proposed trial and then asked parents to rank them by importance. Full details of the outcome prioritization process are reported in a previous study conducted by members of our team. 7 Content analysis 8 was then used to identify outcomes to inform the subsequent Delphi study. 3. Delphi study: A modified 2-round Delphi method was used, 9 and an e-survey was developed (Supplementary File S1), derived from information gained from the literature review and the parent interviews.
The survey was pilot-tested for clarity and face validity with 10 individuals (a mix of nurses, doctors, and dietitians) and then tested again within the study team. It was then input into Delphi-Manager Software. 10  , and its associated study group PICS-SG). The target number of respondents was 40, and automated reminders were sent weekly via the survey software. Participants were allocated a unique identifier to allow identification of individuals completing all rounds of the Delphi exercise. Round 1 of the survey consisted of a succinct summary of the review findings by way of an introduction (previously used trial designs and primary and secondary outcomes), a question to establish respondents' stakeholder group, and a set of outcomes to score (Supplementary File S1). Participants were invited to score the importance of each of the outcomes listed using a 9-point Likert scale as recommended by the Grading of Recommendations Assessment, Development and Evaluation (GRADE) working group, 11 with 1-3 labeled "not important," 4-6 labeled "important but not critical," and 7-9 labeled "critical." Participants were able to provide comment on outcomes and were asked to suggest any additional outcomes that they thought were relevant (Supplementary File S1). Outcomes were listed alphabetically to avoid potential weighting of outcomes caused by the order in which they were displayed. For each outcome, the number of participants who scored the outcome and the distribution of scores (percentages of who had scored each outcome) were summarized by stakeholder group. At the end of round 1, the study management group reviewed the suggested additional outcomes and identified new distinct outcomes to add to the next round. In round 2, participants were presented with a summary of the results of round 1 (histograms of scores for each outcome and each stakeholder group) and asked to rescore these outcomes. They were also asked to score any additional outcomes suggested in round 1. Finally, they were asked to specify a single primary outcome. Outcome scores for each round were summarized using descriptive statistics and histograms, split by stakeholder group. Free text responses were listed. The round 2 scores were used to formulate consensus statistics for each outcome by stakeholder group and, overall, the percentage of respondents that scored 7, 8, or 9 (outcome rated critical); the percentage that score it 1, 2, or 3 (outcome rated unimportant); and consensus status ("consensus in," "consensus out," "no consensus") ( Table 1).

Literature Review
A focused review of the literature to identify previous trials and studies around GRV measurement in all critically ill populations was undertaken in March 2018. Six studies were identified that studied GRV, 1 in the pediatric population, 12 3 in critically ill adults, [13][14][15] and 2 in critically ill preterm neonates. 16,17 All the outcomes reported in these studies have been summarized into 14 outcomes ( Table 2). Two of these outcomes (days receiving parenteral nutrition and days of central venous line access) were relevant only for preterm infants and thus removed, and others (gastrointestinal morbidity) were separated into vomiting and diarrhea as separate outcomes, generating 13 items from the literature review. Figure 1 shows the flowchart detailing the generation of outcomes across all stages of the study.

Parent-Important Outcomes
Fourteen PICU parents were interviewed regarding 10 children (4 sets of parents were interviewed separately about their child), and 6 out of 10 children had had a previous neonatal unit admission. Interviews took place an average of 11 months (range, 8-37 months) after PICU admission. The children's cause of PICU admission varied: congenital heart disease (n = 4), sepsis (n = 2), airway (n = 1), and complications of congenital conditions (n = 3). The parents proposed 7 outcomes of importance; 4 of these were duplicate items generated from the literature review. The 3 nonduplicate parent-reported outcomes were long-term feeding issues, being "back to their normal self," and longerterm outcomes (after hospital discharge).

Delphi Study
A total of 16 outcomes were included in round 1 of the Delphi. In November 2018, 45 PICU healthcare professionals were invited to take part in the Delphi survey. Of these, 30/45 (67%) registered for the survey, 28/45 (61%) went on to score the 16 outcomes in round 1, and of these, 22/28 (79%) went on to complete round 2. Six distinct outcomes were suggested by respondents in round 1, making 22 outcomes to be scored in round 2. Sixty-four percent (14/22) of respondents changed their score for at least 3 outcomes from round 1 to round 2. "Consensus in" was achieved on only 4 outcomes by all 3 professional groups: incidence of gastrointestinal morbidity (vomiting); incidence of ventilator-acquired pneumonia (VAP); duration of time with no enteral feed per 24-hour period; and the time to achievement of estimated energy requirements. All outcomes scored in round 2 are shown in Table 3, together with the source of the outcome and consensus status. No outcomes were voted "consensus out." Respondents were asked their choice of primary outcome measure for a future trial, and 8/22 (36%) preferred time to achievement of energy targets, with 5/22 (23%) choosing incidence of VAP.

Consensus Meeting
Twenty-six PICU healthcare professionals (4 dietitians, 9 physicians, 12 nurses, and 1 general surgeon) attended the consensus meeting on April 2, 2019; some of these had participated in the Delphi study. Despite 2 parents agreeing to participate, no parents were able to attend on the day. The professionals represented 13 different UK PICUs out of a possible 27. At this meeting, there was discussion and subsequent voting on the 18 outcomes that did not reach consensus in the Delphi study. Using the same scoring criteria as the Delphi survey, 8 items were voted "consensus in," 6 were voted "consensus out," and 4 failed to reach any consensus ( Table 4). The choice of the 2 primary outcome measures, time to achieve estimated energy requirements and VAP, was presented and their feasibility discussed, but no voting took place on these. The final list of the 12 consensus outcomes for a future trial is presented in Table 5.

Discussion
This is one of the few studies in pediatric critical care to use a robust process to identify outcomes of importance for both  parents and clinicians for a future trial. 5,18 It is also the first to examine this in the context of a trial of no routine GRV measurement to guide enteral feeding in pediatric critical care. More recently, others have used this process in trials of fluids and temperature control in pediatric critical care. 5,18 The importance of choosing the most appropriate outcome measures for a clinical trial is increasingly recognized. 1 When trial outcomes are either not clinically relevant or not considered important to parents/patients, the evidence generated is less likely to be adopted and implemented into clinical practice. This is costly and inefficient and leads to research waste. 19 Increasing efforts are being made to robustly identify and use the most appropriate trial outcomes for all stakeholders, including patients and parents. To try to overcome these issues, the development of standardized core outcome sets (COSs) for patient populations is encouraged. 20 The Core Outcome Measures in Effectiveness Trials (COMET) initiative aims to do this by developing and agreeing on a set of outcomes known as a COS. 20 These are an agreed minimum set of outcomes that should be reported in all trials of a specific disease or population. 2 Of interest were the differences between clinicalsuggested and parent-suggested outcomes, especially with respect to the longer-term issues after intensive care discharge. Previous studies exploring prioritized outcomes by parents 5,18 have shown that parents who are interviewed shortly after a child's PICU admittance prioritized shorterterm outcomes, which were potentially most relevant to them at the time of being interviewed. However, parents in this study were interviewed on average 11 months after PICU discharge, and ongoing issues of importance to them were more highly ranked as outcomes.
The final list of 12 agreed outcomes for a future trial comprises nutrition outcomes (achievement of energy goals, time spent without feeds, gastrointestinal morbidity, and outcomes secondary to feed intolerance), adverse events (VAP and necrotizing enterocolitis [NEC]), and measures of morbidity (length of invasive ventilation and PICU length of stay). Nutrition outcomes are clearly important in a trial of any nutrition intervention, yet all previous adult trials were powered on adverse events (VAP). Conversely, previous neonatal trials, although measuring NEC (a condition far more common in preterm neonates), targeted nutrition outcomes (time to achieve full enteral feeds) as the primary outcome. Healthcare professionals are highly concerned about adverse events 21 and fear these potentially more than they value improved outcomes in any new interventional trial. However, the incidence of both VAP and NEC is very low in pediatric critically ill children; [22][23][24] therefore, powering any trial based on these outcomes would require such high numbers of eligible children to make a trial very difficult. When we asked clinicians to choose their preferred primary outcome measure, the nutrition outcomes of time and the achievement of target energy goals were most commonly selected. In pediatric critical care, the choice of outcome for trials of nutrition interventions is important, as mortality is too low to be feasibly used to power trials (unlike adult critical care). In addition, nutrition outcomes and their impact may be more important for children than for adults, given that children need to grow (as well as not lose weight) during critical illness. This becomes more important as the length of time spent critically ill increases.
Currently, no COS exists for pediatric critical care or for trials of nutrition interventions; however, adult work has begun to generate a COS for adult critical care nutrition trials. Future research to gain international consensus on core outcomes and incorporate these standardized outcomes in clinical nutrition and pediatric critical care trials is essential and is planned in collaboration with professional societies.
Our study has several strengths. The approach taken used a formalized approach and identified potential trial outcomes from previous research and from parents with experience of PICU care. In addition, the professionals who attended the consensus meeting were not all the same as those participating in the Delphi study, thus generating a broader perspective, although there was some overlap. However, there are also some limitations that warrant mentioning. Some of the PICU parents also had experience of neonatal care, which may have influenced their prioritization of longer-term outcomes and of the outcomes themselves. The sample size for the Delphi was limited to UK PICU professionals, specifically those with an interest in research, and was small, which may limit the generalizability of these results to the wider group of PICU health professionals. We targeted this group because this work was to determine the feasibility of a UK trial within a publicly funded healthcare system, and we therefore wanted to involve and engage with UK PICU clinicians about a future trial; there was, however, representation from the 3 main professional groups involved in enteral feeding. We did not include parents in the Delphi survey, primarily because we felt asking the parents at the interview was the most appropriate way to gain meaningful input; however, the lack of parent attendance at the consensus meeting was a limitation and may have led to the final consensus results being less representative of parent views. We utilized all means to get parents to attend (and payment for their time and travel was provided), but caring commitments on the day prevented parents who had expressed interest from attending the consensus meeting. Future work should consider providing childcare to enable parents to attend. Not having parents at this meeting is a limitation that may have impacted on some parent-focused outcomes being voted out. For the future trial, we will work further with our expert parent group to ensure that outcomes of importance to the parents are captured.

Conclusions
Given the increasing importance placed on determining the optimal outcome measures for trials, this paper describes a robust process to identify optimal outcomes for a future trial of no routine GRV measurement to guide enteral feeding in pediatric critical care. Using the most appropriate trial outcomes increases the likelihood that future trial results are both meaningful to parents/patients and implemented in clinical practice.