Brain-related outcome measures in trials recruiting critically-ill children

Abstract

Purpose of review:

Randomized controlled trials leading to innovations that improve outcomes in acute life-threatening illnesses in children are scarce. A key issue is how we refocus research on outcomes that matter and are more relevant to those making emergency decisions, and those involved with managing and living with the late-outcome. We have used information from recent trials in critically ill children – in particular those illnesses without any primary neurologic involvement – in order to develop an approach to brain=related outcomes that will maximize child and family benefit from research.

Recent findings:

Fifteen recent pediatric critical care trials illustrate four types of brain-related outcomes assessment: 1) death or organ-system-failures – as illustrated by studies in systemic illness; 2) neurological and neuropsychological outcomes – as illustrated by the glycemic control studies; 3) cognitive outcomes – as illustrated by a sedative trial; and 4) composite outcomes – as illustrated by the therapeutic hypothermia studies.

Summary:

The 15 research trials point to five areas that will need to be addressed and incorporated into future trial design, including use of: 1) neurologic monitoring during intensive care unit admission; 2) post-discharge outcomes assessments; 3) strategies to improve retention in long-term follow-up; 4) child- and family-centered outcomes; 5) core outcomes datasets.

Introduction

Adequately-powered multicenter randomized controlled trials (RCTs) leading to innovations that improve brain-related outcomes after pediatric critical (i.e., acutely life-threatening) illnesses are vitally important, but scarce.⁽¹⁾ One problem is that decisions about study design – such as choice of primary and secondary outcomes – need to be balanced by scientific validity, clinical relevance, and feasibility (e.g., subject retention and financial considerations). In order to optimize trial success. In a scoping review of RCTs involving pediatric critical care patients, Duffett et al found that one-third of 248 studies were stopped prematurely because of trial futility or difficulties with recruitment.⁽²⁾ When we examine the most common primary outcomes used in these RCTs we find that close to one-fifth had physiological or laboratory measurements as the primary endpoint, with only a few trials evaluating clinical efficacy, mortality, or long-term outcomes.⁽²⁾

A key question for the field is where do we go from here? How do we refocus the primary and secondary research outcomes such that they are relevant to those making emergency decisions, and those involved with managing and living with late-outcomes in severely ill children? In regard to relevancy, it is clear that mortality rates have declined to the level of making RCTs with this outcome unfeasible. Recently, however, co-opting family preferences and perspectives on late-outcome has led to increasing recognition of brain-related sequelae of critical illness like Post-Intensive Care Syndrome (PICS), which goes far beyond simply using dichotomous categories in RCTs (e.g., alive versus died, good outcome versus poor outcome, or favorable outcome versus unfavorable outcome).^{(3**,4,5,6*,7**)} Rather, in these late-outcomes remember it is all about the brain; that is, neurological insult and injury, whether present on admission or acquired during the hospital stay, is the major cause of death and disability in critically ill children. Therefore, more detailed neurological outcomes assessments are essential, even in apparently non-neurological studies since these outcomes may also disclose important, unexplored long-term health effects of an intervention.^(8,9)

Background

The objective of this review is to discuss approaches to outcomes assessment that are relevant for future RCTs in critically ill children. We have therefore used information from 15 recent multicenter trials in the critically ill child – in particular those illnesses without any primary neurologic involvement – in order to develop an approach to brain-related outcome that will maximize child and family benefit from research.^{(10**,11,**12–24)} The 15 trials were published between 2011–2018 (Table 1). The studies included for our analysis were found using a Pubmed literature search and the website http://picutrials.net/. Together, they comprised three multicenter observational studies and 12 RCTs (phase II, III, or IV studies). The medical conditions studied in these trials were sepsis, acute respiratory failure, congenital heart disease, and post-cardiac arrest intensive care.

Table 1.

Patient characteristics and brain-related outcomes used in pediatric critical care studies (presented in reverse order, most recent first)

STUDY	CENTERS	n	POPULATION	INTERVENTION	1° OUTCOME	2° OUTCOMES	BRAIN-RELATED OUTCOMES
Long-Term Outcomes after Protocolized Sedation versus Usual Care in Ventilated Pediatric Patients (RESTORE-COG)10**	31	1,073	Acute respiratory failure	None (Parent study: sedation protocol vs. usual care)	Functional status (POPC, PCPC) 6m; Health-related quality of life (Ped QL or ITQOL) 6m; Post-traumatic stress disorder (CPSS), 6m	-	Cognitive function (PCPC) worsened from baseline to follow-up in 11% and improved in 6%
Effect of a Pediatric Early Warning System on All-Cause Mortality in Hospitalized Pediatric Patients (EPOCH)11**	21	144,539	Hospitalized children	PEWS vs. usual care	Mortality, HD	Significant clinical deterioration, HD	-
Tight Glycemic Control in Critically Ill Children12	35	713	Hyperglycemia in ICU (non-cardiac)	Tight vs. conservative glucose control	ICU free days, 28d	Incidence of HAI and hypoglycemia; 90d Mortality; PELOD, ventilator free days 28d	New seizure: 1.4% vs. 2.9%, (TGC, UC), p=.20
Therapeutic Hypothermia after in-Hospital Cardiac Arrest in Children (THAPCA)13	37	329	In-hospital cardiac arrest	Hypothermia vs. Normothermia	Survival with favorable neurobehavioral outcome (VABS), 1y	Mortality, 1y; Change in neurobehavior function Safety	Survival with favorable outcome: 36% vs. 39%, (hypothermia, normothermia); RR, 0.92; 95%CI, 0.67 to 1.27, p=.63
Early versus Late Parenteral Nutrition in Critically Ill Children14	3	1,440	ICU > 24h	Parenteral nutrition within 24h vs day 8	New infection in ICU; ICU LOS	Mortality, 7d; ICU, hospital, LOS; Hypoglycemia, liver lab abnormalities; ICU readmissions; Organ support days	-
Protocolized sedation vs usual care in pediatric patients mechanically ventilated for acute respiratory failure (RESTORE)15	31	2,449	Acute respiratory failure	Sedation protocol vs. usual care	Mechanical ventilation days	Respiratory failure recovery time; LOS, Mortality, HD; Adverse events; Sedative exposure; Withdrawal frequency	Pain score: median, 50% [IQR, 27%–67%] vs. 23% [0%–46%] of days, (Intervention vs. Control), p<.001 Agitation: 60% [33%–80%] vs. 40% [13%−67%], (Intervention vs. Control), p=003
Therapeutic Hypothermia after Out-of-Hospital Cardiac Arrest in Children (THAPCA)16	38	295	Out-of-hospital cardiac arrest	Hypothermia vs. Normothermia	Survival with a favorable neurobehavioral outcome (VABS), 1y	Mortality 1y; Change in neurobehavioral function; Safety	Survival with favorable outcome: 20% vs. 12%; (hypothermia, normothermia); relative likelihood, 1.54; 95% CI, 0.86 to 2.76, p=.14
Simultaneous Prediction of New Morbidity, Mortality, and Survival Without New Morbidity From Pediatric Intensive Care17	7	10,078	PICU admission	None	Morbidity (FSS), HD; Mortality, HD	-	Unadjusted morbidity rates: 4.6%, site range, 2.6–7.7%
Dopamine Versus Epinephrine as First-Line Vasoactive Drugs in Pediatric Septic Shock18	1	120	Septic shock	Dopamine vs. epinephrine infusion	Mortality, 28d	HAI; Vasoactive medication use; Multiple organ dysfunction	-
Control of hyperglycaemia in paediatric intensive care (CHiP)19	13	1,369	Mechanical ventilation and vasoactive medication	Tight vs. conservative glucose control	Days alive and free of mechanical ventilation, 30d	ICU LOS, d; Mortality; Organ support days; Blood infection; Antibiotics >10d; RBC transfusion; PELOD; Readmissions; Hospital cost; Hypoglycemia; Hospital and community health service utilization, 1y	Seizure requiring medication: 3.3% vs. 2.2, (TGC, UC), Mean difference (95% CI) 1.15 (0.77 to 2.98)
Pediatric calfactant in acute respiratory distress syndrome trial20	24	110	MV + ARDS with direct lung injury	Calfactant vs. placebo	Mortality, 90d	Ventilator free days, 28d; LOS; Change in oxygenation; Adverse events, HD	-
Critical Illness Stress-Induced Immune Suppression (CRISIS) Prevention Trial21	8	293	ICU >72h	Immunonutrition vs. placebo	Days to nosocomial infection or clinical sepsis, up to ICU d 28	Rate of nosocomial infection or clinical sepsis per 100 PICU d; Antibiotic-free days; Incidence of prolonged lymphopenia; Serum prolactin, zinc, and selenium levels; Mortality, 28d; Adverse events	-
Safe Pediatric Euglycemia in Cardiac Surgery (SPECS)22	2	980	Congenital heart disease	Tight vs. conservative glucose control	Rate of HAI, up to ICU 30d or 48 h post ICU	Mortality, 30d and HD; Mechanical ventilation days; LOS; Organ failure Hypoglycemia	Seizures: <1% vs. 1%, (TGC, UC), p=.34
Intensive Insulin Study in PICU Patients23	1	700	ICU admission	Tight vs. conservative glucose control	Intelligence, 3y	Visual-motor integration, attention, motor coordination, executive functions, memory, behavior, 3y	Poor outcome (death or severe disability precluding neurocognitive testing): 19% vs. UC 18%, (TGC, UC); risk-adjusted OR 0.93, 95% CI 0.60–1.46, p=.72. Clinical neurological evaluation score (range, 0–8): 1 (0–2) vs. 0 (0–1), p=.09. Full-scale IQ (range, 45–155): 84 (72–95) vs. 89 (76–100), p=.35. Visual-motor integration (range, 0.9–20): 8 (7–10) vs. 9 (7–10), p=.41. Reaction time dominant hand, msec: 895 (746–1115) vs. 895 (739–1089), p=.66. Motor coordination (No. of taps in 10 s), Dominant hand: 20 (17–28) vs. 24 (19–29), p=.17. Memory span (forward): 7 (6–9) vs. 5 (5–11), p=.67. Behavior (by proxy), T score, CBCL–total problems, (range, 24–100): 53 (44–61) vs. 51 (45–58), p=.42. Brief hypoglycemia in TGC group not associated with worse neurocognitive outcome
Fluid Expansion As Supportive Therapy in critically ill African children (FEAST)24	6	3,141	Febrile with shock	Fluid bolus vs. none	Mortality, 48h	Shock, 48h; Adverse events, 4w; Mortality, 4w; Neurologic sequelae, 4 and 24w	Neurologic sequelae: 2.2%, 1.9%, and 2.0% (albumin-bolus, saline-bolus, control), p=.92. Increased intracranial pressure: 1.5% vs. 1.7% vs. 1.1%. Neurologic sequelae: 2.2% vs. 1.9% vs. 2.0%

Where: ARDS, acute respiratory distress syndrome; CPSS, Child PTSD Symptom Scale; FSS, Functional Status Scale; HAI, healthcare associated infection; HD, hospital discharge; IQ, intelligence quotient; LOS, length of stay; OR, odds ratio; PCPC, Pediatric Cerebral Performance Category and POPC, Pediatric Overall Performance Category; Ped QL, Pediatric Quality of Life Inventory, Version 4.0 Generic Core Scales (PedsQL); PELOD, Pediatric Logistic Organ Dysfunction; PEWS, Pediatric Early Warning System; ITQOL, Infant and Toddler Quality of Life Questionnaire-97; VABS, Vineland Adaptive Behavioral Scale; PICU, pediatric intensive care unit or ICU for intensive care unit; RR, relative risk; TGC, Tight Glucose Control; UC, usual care; Times with hour (h), day (d), weeks (w) and year (y); VABS, Vineland adaptive behavioral scale; vs, versus; 95%CI. 95% confidence interval; 1° and 2°, primary and secondary, respectively.

Study outcomes

The primary end-points used in the 15 studies were: death in five;^{(11**,17,18,20,24)} child functional health in four;^{(10**,13,16,23)} duration of organ support or pediatric intensive care unit (PICU) stay or days free from support in PICU in four;^{(12,14,15,19)} and, event occurrence, event rate, or time to event in two.^(21,22) A composite of these outcomes were used in four of the 15 studies.^{(10**,12,13,16)} Of the four studies with child functional outcomes assessed after hospital discharge, two were RCTs,^(13,16) one was an observational study,^(10**) and one was an ancillary observational study of RCT recruits that had hospital-based outcomes.⁽²³⁾

We now use these 15 studies to illustrate four types of brain-related outcomes that have been used in pediatric critical care trials: 1) death or organ-system-failures – as illustrated by studies in systemic illness; 2) neurological and neuropsychological outcomes – as illustrated by the glycemic control studies; 3) cognitive outcomes – as illustrated by a sedative trial; and 4) composite brain-related outcomes – as illustrated by the therapeutic hypothermia studies; and

Death or organ-system-failures as outcomes in studies in systemic illness

Studies assessing mortality as their primary outcome were all RCTs testing hospital- or PICU-based interventions. Groups were compared, for example: at 48 hours, in a trial of fluid bolus for septic shock;⁽²⁴⁾ at 28 days, in a trial of dopamine versus epinephrine for septic shock;⁽¹⁸⁾ at 90 days, in a trial of calfactant for respiratory failure;⁽²⁰⁾ or, at hospital discharge, in a study of using the pediatric early warning system (PEWS) to identify clinical deterioration.^(11**) Death rate for children enrolled in the fluid study ranged from 7.3 to 14.0%, but death was an uncommon event in the PEWS study. Only the inotrope study, performed at a single center, found a significant difference in favor of epinephrine when using the primary outcome of death. Taken together, death, and all-cause mortality, as an endpoint has four challenges for research in pediatric critical care. First, in modern PICU practice death is an infrequent event (~2.5%) and even for conditions in which it is currently higher than average, survival is improving over time, thereby limiting statistical power in future studies. Second, the expected impact of an intervention on death (say >10% reduction) may decrease the sample size needed, such rate reductions are unrealistic. However, a lesser impact in terms of mortality reduction may still be clinically relevant, but would the greater sample size needed for study be achievable? Third, variability in the timing of mortality assessment captures different situations. There are various reasons why a later death after critical illness occurs, and these are not often captured in RCTs. For example, death may result from an illness that is unrelated to the intervention or illness being studied, or it may be the result of withdrawal of life-supporting treatment. Last, outcomes other than survival are valid and important to patients and families.^{(6*,7**,9,25–27)}

The advantage of duration of organ support and length-of-stay on the PICU as primary outcome measures are that they assure almost full follow-up rates that maintain trial statistical power and may be the most validated and timely measures to capture the early effects of a particular study intervention. This reasoning can also be applied to the choice of outcome of time-to-(in-hospital)-event used in two of the studies in Table 1. The Critical Illness Stress-Induced Immune Suppression (CRISIS) prevention and Safe Pediatric Euglycemia in Cardiac Surgery (SPECS) trials, testing the utility of immunomodulating medications and tight glucose control against placebo or usual care, respectively, evaluated time to nosocomial infection and rate of nosocomial infection as primary outcomes.^(21,22) As a surrogate for brain-related outcome, we know that nosocomial infections are associated with increased length-of-stay and thus, potentially, an increased risk of PICS.⁽²⁸⁾

Neurological and neuropsychological outcomes used in the glycemic control studies

Tight glucose control has been the topic of a few PICU multicenter RCTs since the publication of the groundbreaking single-center trial in 2009 that showed glucose-control resulted in decreases in length of stay, systemic inflammation, and death.⁽²⁹⁾ However, the glucose-control with insulin group experienced more hypoglycemia (25% versus 1%) compared to usual care, leading to concerns for adverse long-term neurological impact,^(30,31) which were explored in the four-year follow-up of 569 of the 700 children originally enrolled.⁽²³⁾ The authors found that serum levels of two brain injury biomarkers – Neuron Specific Enolase (NSE) and S100b – were of no significance on the day of admission; NSE and S100b levels did not differentiate between trial-groups and neither did they increase after episodes of hypoglycemia. However, infants and children experiencing hypoglycemic episodes did have increased severity-of-illness and higher levels of both biomarkers on admission.⁽³²⁾ In regard to later neuropsychological outcomes on follow-up, there were no differences in intelligence quotient (IQ) – the primary outcome – or any of the secondary outcomes (i.e., death or severe disability, visual motor integrity, attention, motor control, executive function, memory, behavior, and neurologic examination) between trial groups.^(23,33) Although it is unsurprising that critically ill children had worse neurocognitive performance compared to healthy children, it is not clear whether functional disabilities were pre-morbid since three-quarters of the children enrolled in the original study had congenital heart disease with many in a post-operative state, a cohort at high risk of neurocognitive impairment.^(34,35) Of note, subsequent multicenter trials testing tight glucose control did not lead to differences in the hospital-based primary outcomes in children, for example: in congenital heart disease, the rate of nosocomial infection; in respiratory failure or shock, days alive and free of mechanical ventilation; and, in critically ill children with hyperglycemia, ICU free days.^{(12,19,22,36)} Again, there were more episodes of hypoglycemia in the intervention group in each study, but the later RCTs did not have post-hospital discharge neurodevelopmental assessments.

Cognitive outcomes as used in the sedative trial

Common barriers to including patient brain-related outcomes assessments after hospital discharge in research programs are the substantial resources, time, and costs that are required for retention in the clinical study. A successful strategy used by some large interventional trials with PICU or hospital-based primary outcomes was to assess their hypothesis at time points relevant to the intervention (within the episode of hospital admission) and use ancillary studies to support the later post-discharge outcomes. For example, the Randomized Evaluation of Sedation Titration for Respiratory Failure RESTORE) trial tested whether a nurse-driven sedation protocol or usual care led to decreased duration of mechanical ventilation as the primary outcome.⁽¹⁵⁾ At the time of study, the effect of pain and exposure to sedatives (along with the effect of drug withdrawal and delirium) on neurocognitive outcomes was relatively unexplored in the PICU population – a situation that has now been improved.^(37–39) For example, RESTORE-cognition^(10**) was an independently funded study ancillary to the original RESTORE trial and tested the post-hospital functional and health-related quality of life outcomes for more than 1,000 children enrolled in RESTORE. RESTORE-cognition had a 79% follow-up rate and, importantly, found that new morbidities at 6 months after illness was common, but there were no differences in outcomes between randomized groups. Twenty percent and 11% of children had worsened Pediatric Overall Performance Category (POPC) and Pediatric Cerebral Performance Category (PCPC) scores, respectively, while 6% of children had improved PCPC scores. The study reported a 30% overall frequency of post-traumatic stress disorder, a condition known to impact long term child and family functioning.⁽⁴⁰⁾

Composite brain-related outcomes as used in the therapeutic hypothermia studies

The Therapeutic Hypothermia After Pediatric Cardiac Arrest (THAPCA) RCTs (whether in-hospital or out-of-hospital event) used, after a detailed analysis of available metrics, a composite primary outcome of survival with favorable neurological outcome as defined by a threshold score for the Vineland Adaptive Behavioral Scale (VABS-II) at 12 months post-arrest.^(13,16,41) Using a composite of brain-related outcomes should be beneficial when studying experimental interventions that may affect both mortality and morbidity.⁽⁴²⁾ Thus, in THAPCA, targeted temperature management (TTM) at hypothermia rather than near-normothermia level was hypothesized to improve the composite of both survival and functional outcomes for survivors.⁽⁴¹⁾ This combined approach may also increase the rate of occurrence of events-of-interest, thereby leading to more power and the need for smaller (and achievable) sample size; but, the interpretation of results may be less clear than when using a single measure, since one component of the outcome-metric may be accounting for a greater part of any association identified. Results from both THAPCA trials did not show any difference between using the two levels of TTM. However, the absolute risk reduction for the composite outcome and secondary mortality outcome in the out-of-hospital RCT were 8% and 9%, respectively, which may be interpreted by some as clinically relevant in a condition with high mortality.⁽¹⁶⁾ A post-utilization survey by the THAPCA team revealed that most clinicians are now practicing TTM using near-normothermia level of temperature management in children surviving cardiac arrest, and some are reserving TTM-hypothermia for certain cases – despite knowledge of the trial results.^(43**)

In regard to brain-related outcomes, is also noteworthy that THAPCA used individual sites to conduct baseline (pre-arrest) assessments. Then, at 1-year follow-up, centrally-located and trained personnel coordinated the assessments from both reports by local caregivers and from in-person reviews by experts based in the trial coordinating center. The follow-up rate was >90%. The THAPCA investigators found high rates of neurobehavioral morbidity in surviving children who did not have any significant baseline dysfunction across all functional domains tested in the IQ and VABS-II measures.⁽⁴⁴⁾ In addition, many of the children designated as demonstrating “favorable outcome” by a priori definitions had significant morbidity at 12 months.^(45**) Finally, family functioning was explored in families with children surviving out-of-hospital cardiac arrest at one year.⁽⁴⁶⁾ The key finding was that family dysfunction was associated with having children with high-risk of neurologic disability at one year. Family dysfunction is also associated with worse child outcomes following critical illness such as in traumatic brain injury.⁽⁴⁷⁾

Future considerations

Pediatric critical care and illness research programs, whether focused on the brain-injured population or not, will need to incorporate innovative ideas and approaches to brain-related outcomes. There are five such areas that are covered in the literature, and have been used in the design of some of the pediatric trials described in Table 1, including: 1) neurologic monitoring during PICU admission; 2) post-discharge outcomes assessments; 3) strategies to improve retention in long-term follow-up; 4) child- and family-centered outcomes; and 5) core outcomes datasets.

Neurologic monitoring during illness

Neurological morbidity, high prevalence of co-morbid neurological conditions, and the need to assess long-term outcomes in critically ill children, mean that we should explore the use of brain-related serum biomarkers in our clinical research (see above section “Neurological and neuropsychological outcomes used in the glycemic control studies”). In addition to these brain-tissue biomarkers, electroencephalography may be useful. For example, quantifying duration of seizures in high-risk, critically ill post-operative congenital heart disease infants, or children after cardiac arrest resuscitation – along with measuring serum brain-tissue-injury biomarkers – may help monitor impending central nervous system injury.⁽⁴⁸⁾

Post-discharge outcomes

Some RCTs have had incredible success in achieving high retention rates for post-hospital discharge outcomes at a year or later. However, this target is an enormous challenge for many research programs. Exploring whether an earlier time point for assessment serves as a surrogate for later long-term outcomes would increase trial feasibility and cost-effectiveness. This approach was addressed in a THAPCA-related publication in which the 3-month VABS-II scores were found to be strongly correlated with 12-month VABS-II scores.⁽⁴⁹⁾ Hence, in future post-cardiac arrest studies using the earlier time point will hopefully mean that fewer trial recruits are in the “lost to follow-up” category. This approach may not be generalizable to all studies but we should consider validation for other PICU categories such as sepsis, systemic illness and acute respiratory failure. However, in congenital heart disease studies it seems likely that nothing can replace the long-term outcome endpoint, since these infants (often newborns) will have significant developmental progression and a longer recovery period may be necessary to assess accurate functional health outcomes.

Strategies to improve trial-retention

A systematic review has shown that the number of unique trial retention strategies used in studies is correlated with higher retention rates; the authors recommend using at least 5 different strategies to maximize trial retention.⁽⁵⁰⁾ As an example, the THAPCA trials were particularly successful in their rate of participant retention, and it is interesting to note that the investigators used a multifaceted approach to post-hospital discharge outcomes assessment by employing a few highly-engaged trained personnel who worked at the trial center where they were given operational goals for the research (see above section “Composite brain-related outcomes as used in the therapeutic hypothermia studies”).

Child and family-centered outcomes

Patient- and family-centered outcomes in RCTs are becoming recognized as more important by both investigators and funding agencies. However, few trials have used this approach.⁽⁵¹⁾ Patient/Family-reported outcomes are also increasingly used to inform best practice and clinical care since such shared decision-making improves quality of care and eventual outcomes.^(51–53)

Core outcome datasets

The idea here is that all “stakeholders” in pediatric critical care trials – along with best practice methodologies outlined in this review –inform the development of core outcome datasets that can be used in future studies. A pediatric critical care core outcome dataset is currently in development (see http://www.comet-initiative.org/studies/details/1131). One advantage of standardizing brain-related outcomes in PICU trials is that consistency in reporting of outcomes – sometimes with more specific variables for certain patient cohorts – will enable better systematic review of results that can then be used in clinical decision-making.⁽⁵⁴⁾

Conclusion

In summary, despite the great need for long-term comprehensive brain-related outcomes in pediatric critical care research, few multicenter RCTs incorporate and report them. Research investigators, funding agencies, and other stakeholders will need to work to find creative solutions towards this mission.

Key Points.

• Adequately-powered multicenter randomized controlled trials leading to innovations that improve brain-related outcomes after pediatric critical illnesses are vitally important, but scarce.

• Recent pediatric critical care trials have used four types of brain-related outcomes assessment: death or organ-system-failures; neurological and neuropsychological outcomes; cognitive outcomes; and, composite outcomes.

• Pediatric critical care and illness research programs, whether focused on the brain-injured population or not, will need to incorporate innovative ideas and approaches to brain-related outcomes.

• The recent literature points to five areas that will need to be addressed and incorporated into future trial design, including use of: neurologic monitoring during intensive care unit admission; post-discharge outcomes assessments; strategies to improve retention in long-term follow-up; child- and family-centered outcomes; core outcomes datasets.