Article Text

This article has a correction. Please see:

Download PDFPDF

Dilemmas in undertaking research in paediatric intensive care
  1. Hari Krishnan Kanthimathinathan1,
  2. Barnaby R Scholefield2
  1. 1Paediatric Intensive Care Unit, Evelina London Children's Hospital, London, UK
  2. 2Paediatric Intensive Care Unit, Birmingham Children's Hospital, Birmingham, UK
  1. Correspondence to Dr Barnaby R Scholefield, Consultant Paediatric Intensivist, Paediatric Intensive Care Unit, Birmingham Children's Hospital, Steelhouse Lane, Birmingham B4 6NH, UK; barney.scholefield{at}


Providing evidence-based interventions for infants and children is important in paediatric intensive care, where decision making impacts most acutely on morbidity and mortality. However, despite the major progress of medicine in the 21st century, we still lack this evidence for majority of the decisions we make. In this article, we explore and suggest possible solutions for several dilemmas faced by paediatric intensive care researchers. These include ethical dilemmas such as validity of informed consent, use of deferred consent, balancing risk versus benefit and methodological dilemmas such as how to generate high-quality evidence with low-patient volume, choice of valid outcome measures and how best to use research and researchers’ networks.

  • Intensive Care
  • Paediatric Practice
  • Evidence Based Medicine
  • Ethics
  • consent
View Full Text

Statistics from


Conducting research in paediatric intensive care (PIC) is challenging but remains essential to deliver high-quality patient care. Paediatric intensive care units (PICUs) manage critically ill children of all ages from preterm neonates to adolescents. In 2012, PICUs in the UK admitted nearly 20 000 children accounting for 120 000 bed days with >96% survival.1 However, robust evidence base for many standard practices, interventions and medications is lacking. In this article, we explore some of the reasons why PIC research is unique and imperative. Recruitment of critically ill ‘minors’ in a stressful, intensive care setting is just the first hurdle. PIC researchers are faced with several ethical, logistical and methodological dilemmas in the conduct of research in critically ill children. We review these issues and discuss potential solutions for the future.

Why is research in PIC necessary?

Children are not small adults and PIC is neither the smaller version of adult intensive care nor the larger version of neonatal intensive care. Anatomical, physiological and developmental differences exist resulting in important pathophysiological variations between these age groups. Using cardiac arrest as an example, in adults, this is usually coronary artery disease related with sudden onset; in neonates, this is rare but may be related to perinatal events; whereas in children, this often follows respiratory insufficiency, hypoxia and acidosis. All these factors affect injury severity, reversibility, prognosis and treatment options. Age-related diseases (bronchiolitis, croup, febrile convulsions, etc) and congenital syndromes are unique to children. Children metabolise medications differently leading to conditions such as Propofol infusion syndrome and Aspirin-related Reye's syndrome. Their host response to triggers such as infection is different due to maturity of immunological and hormonal systems. In addition, their resilience to certain diseases can be remarkable. In the developed world, mortality from sepsis in children remains less than 10% compared with 30–40% in adults.

These significant differences inhibit direct extrapolation of adult or neonatal evidence to PIC population. Using the earlier example of cardiac arrest, the evidence for neuroprotective effect of therapeutic hypothermia in neonates and adults cannot simply be assumed to apply to paediatric patients. History reveals other examples of conflicting results between different age groups. Randomised controlled trials (RCT) of extra-corporeal membrane oxygenation (ECMO) in the 1980s for respiratory failure showed benefit in neonates but not in adults and early studies of activated protein-C in sepsis showed survival benefit in adults but not in children (RESOLVE trial) (table 1). Research in PIC is essential to develop effective, age-group-specific interventions.

Table 1

Summary of randomised controlled trials referred to in the article

The dilemmas

With such an overwhelming requirement for research in PIC, the question is why has this not happened? Perception of clinical trials as experiments has led to a misleading distinction being made between clinical practice and clinical research.2 Clinicians, regulatory agencies and public are comfortable using untested interventions on children as standard practice, rather than enrol them in a clinical trial to test them. A double standard exists whereby treatments given outside clinical trials are less stringently reviewed than within the trial.3 Prescription rates of off-label or unlicensed medications up to 80% have been described.4 Such practice, while accepted as routine in children, may be dangerous. It is a sobering fact that a recent systematic review of paediatric traumatic brain injury (TBI) trials did not find robust evidence to make any level 1 recommendation.5

This does not mean replicating all adult or neonatal studies in paediatrics. Presence of significant differences between children, neonates and adults such as those mentioned above does not mean that the innumerable obvious similarities among different age groups of same species cannot be exploited to the advantage of children. In fact, evidence from studies in adults may be used to avoid unnecessary studies in children. Use of healthy adult volunteers for pharmacokinetic or bioequivalence studies of paediatric formulations, use of existing adult efficacy data and paediatric clinical safety data prior to conducting small-scale clinical trials in children have been described.6 Regulatory agencies such as the European Medications Agency (EMA) share this view of performing paediatric trials when necessary and avoiding unnecessary studies.7 Since 2007, all applications for marketing authorisation for medicines in the European Union (EU) are encouraged (with patent extension incentives) to include studies carried out in children.

Informed consent

Nuremberg code (1949), perhaps the most important statement of ethics, stated that ‘the voluntary consent of the human subject is absolutely essential’.8 However, this cannot be applied to children. This was amended in the ‘Declaration of Helsinki’ (1964) to allow parents or guardians of children to consent on their behalf. While desirable, concerns exist about the process of informed consent obtained from parents of PIC patients. High levels of anxiety in these parents may impair their ability to comprehend concepts such as randomisation and provide true informed consent.9 ,10 Studies have shown widespread parental misconceptions after informed consent process. Where they mistakenly believed randomisation is a method of rationing access to expensive new technology when it was assessing effectiveness; that children would receive direct clinical benefit from the trial, when there may be none; and that participation ensured access to medications, when it was not the case.11 Parents of children in life-threatening condition are more likely to consent to trials with hope of finding ‘a miracle cure’.12 Some of these pitfalls may be avoided by initiating the consent process before PIC admission. This may be possible in case of elective cardiac surgery for example. In such cases, studies have shown that parents preferred to be informed about trials at least a month in advance.13 The Control of Hyperglycaemia in PIC (CHiP) trial obtained provisional consent for its cardiac surgical population preoperatively and then confirmed it after admission (table 1).14

A qualitative study of consent forms used in trials found incongruent and inconsistent practices relating to statement of risks and benefits, disclosure of research results and incidental findings.15 Statement of risks and benefits of participating in RCT in the consent forms may be liable to contradictory interpretations. The Office of Human Research Protection in the United States found that the conduct of Surfactant, Positive Airway Pressure and Pulse Oximetry Trial (SUPPORT) trial was in violation of the regulatory requirements for informed consent and failed to describe the ‘reasonably foreseeable risks of blindness, neurological damage, and death’.16 ,17 While the researchers viewed both saturation targets (85–90% and 90–95%) as within what was then standard practice, the regulatory authorities interpreted this differently. Similar uncertainties exist in PIC. Patient and parent involvement in every step of trial process including trial design, content, design of information leaflets and consent forms is encouraged and may avoid or resolve such problems. Medicines for Children Research Network (MCRN), for example, has set up Young Person's Advisory Groups at local and national level to work with researchers and other organisations involved with research. Independent agencies such as Nuffield council on bioethics have expertise in such issues and may help facilitate dialogue between stakeholders.


Where possible, children with sufficient maturity are required to give assent. Medical Research Council guidelines on conduct of medical research in children state that ‘if the child is able to give assent to decisions about participation in research, the investigator must obtain that assent in addition to the consent of the legally authorised representative’.18 Similar requirements for assent exist across the world and are encouraged as best practice; however, definitions vary. Problems can also develop when the parent and child provide incompatible answers, rendering either the consent or assent meaningless.19 However, in most circumstances in PICU, the child is too unwell to participate in the consenting or assenting process.

Deferred consent

There are situations in PICU, such as resuscitation from respiratory or cardiac arrest, where informed consent before intervention may not be possible. An ethically justifiable case could be made for research without waiting for informed consent, when the potentially beneficial research intervention cannot be delayed. Delays due to informed consent may even be detrimental. Clinical randomisation of an antifibrinolytic in significant haemorrhage (CRASH-2) trial randomised tranexamic acid or placebo after adult trauma.20 Not all hospital ethics committees approved waiver of informed consent, necessitating a delay obtaining informed consent for some patients. A comparison study using this trial data showed that delay due to consenting process reduced the number of patients who may have benefitted and increased risk of death when compared with hospitals that waived informed consent.21 Similar arguments could be made for septic shock, TBI and postcardiac arrest trials in PICU with limited therapeutic windows.

Three possible solutions have been suggested: deferred consent, proxy consent by a third party and retrospective consent.22 Deferred consent is the most commonly advocated. In 2001, EU directive on clinical trials stated that ‘consent must be given on behalf of a minor prior to research commencing’. Significant difficulties arose in emergency and intensive care research when this directive was made a law in the UK. However, this was amended in ‘The Medicines for Human Use (Clinical Trials) and Blood Safety and Quality (Amendment) Regulations 2008’.23 The amendment permits deferred consent while the minor requires urgent treatment; urgent action is required for the trial; and informed consent is not reasonably practicable, provided that an ethics committee has given its approval. Deferred consent has recently been used in paediatric studies in the UK (Catheter Infections in Children (CATCH) trial) and in Africa (Fluid Expansion as Supportive Therapy (FEAST) trial).24–27

Most parents support the concept of deferred consent.28 Bereaved parents expected to be told about the trial in cases where deferred consent was used, although this was not the case with non-bereaved families. Ethical dilemmas relating to early mortality in a trial involving deferred consent include whether to approach parents for informed consent after death of a child already included and whether data from such a child be analysed. FEAST study investigators chose not to seek consent after death of a child to avoid parental self-blame among other reasons.27 Disregarding data obtained from such children introduces the risk of systematic bias wherein the population with high early mortality are excluded. It is also against the principle of intention to treat analysis. Studies that explored the impact of not using such data on validity of results recommend including such clinical data, provided data confidentiality requirements are met and doing so does not harm the interests of families.29 Regulatory requirements pertaining to this are unclear and ethics committees have assessed trial protocols on a case-by-case basis.

Death and bereavement

Standards for management of death and bereavement during or after PIC research are unclear. The Bereavement and Randomised Controlled trials (BRACELET) study analysed mortality during neonatal and PIC trials and policies relating to bereaved parents.30 None of the 50 RCT protocols documented a policy relating to care of bereaved parents following enrolment of their child. Few trials had developed information leaflets for bereaved parents and is certainly an important area in need of further studying and development of standards.

Risk versus benefit

NHS England research governance framework states that ‘if there are any risks to participants, the risks must be in proportion to the potential benefit. Unless the risk to them is negligible, it is unethical to involve minors, in research that could have no therapeutic benefit for the group involved’.31 While being a safeguard to minimise and prevent unnecessary harm or pain in children, it also poses a challenge. Trials that do not directly benefit children may be useful for other children in the future. Pharmacokinetic, efficacy and safety data for even commonly used analgesics in PICU are scant for young infants, and studies may be useful. However, placebo-controlled studies, the norm in adults in this setting, are not feasible in children due to research governance rules. A Food and Drug Administration scientific workshop developed consensus guidelines suggesting use of small sample sizes, crossover design, advanced statistical methods and multicentre consortia as a way forward.32 Studies of biomarkers of infection or organ dysfunction may require participants to undergo serial blood sampling. In children without existing indwelling catheters, could this be considered negligible risk or be done without inflicting pain? Is it more acceptable to extrapolate adult data and incorporate such tests into standard practice without rigorously studying them? What could be considered as ‘negligible’ or ‘minimal risk’ and how to measure proportionality of risk versus benefit is debatable.33 Precise definitions and consensus are required. Participating in research may sometimes result in unexpectedly worse outcomes such as in the respiratory syncytial virus vaccine trial in 1960s.34 There is no substitute for careful consideration of research ethics committees and robust data monitoring committees providing interim analyses in these situations.

Randomised controlled trials

Well-conducted RCT has been widely accepted as the ‘gold-standard’ of trial designs to generate evidence for or against an intervention. However, the conduct of RCT poses a number of challenges when applied to PIC patients. Concerns about RCT design include problems with disconcerting parents with statement of uncertainty over best intervention, and the possibility of parental disappointment with treatment allocation. In addition to admission and mortality rates in PIC being low (table 2), the case mix of admitted children is extremely heterogeneous. These result in small eligible pool of children for PIC trials, large sample size requirement to detect significant mortality differences and resultant long recruitment periods. This can lead to underpowered, negative RCTs. Therapeutic hypothermia for paediatric in-hospital cardiac arrest (THAPCA-IH) trial has not yet completed recruitment a decade after study conceptualisation.35 Technological advances, evolution of standard therapy and shifts in clinicians’ perception of equipoise during such long periods also add to the difficulties in performing RCTs.

Table 2

Population differences between PICU and other ICUs

A review of PIC RCTs noted that the quality of the evidence informing paediatric recommendations in the ‘Surviving Sepsis guidelines’ was low and only 3 (14%) of the paediatric recommendations were supported by high/moderate evidence compared with 41 (54%) in adults.36 This review found that PIC RCTs were generally small, single-centred and primarily measured short-term laboratory and physiological outcomes. There were only 3 RCTs with >1000 children and 12 with ≥300 children. A useful database of all RCTs in PIC population may be found at

Alternative study designs

Alternative randomisation designs such as Zelen, adaptive, step wedge and cluster randomisation have been used in the past. Zelen or prerandomisation is a design where consent is sought only from parents of children who have already been randomised to receiving experimental intervention. While this avoids uncertainty and disappointment in parents who do not receive the experimental intervention, several disadvantages including denial of choice, denial of information and overselling of allocated treatment have also been pointed out.37 ‘Play-the-winner’ adaptive randomisation strategy may have been used by Bartlett for his neonatal ECMO trial due to concerns about number of deaths to achieve a statistically significant result in conventional RCT designs among other reasons.38 However, this led to heavily skewed randomisation with 11 patients randomised to ECMO, all of whom survived, and only one to the conventional arm, who died. UK clinicians chose to conduct a conventional RCT for ECMO in neonatal respiratory failure in the 1990s, despite earlier positive trials that used adaptive randomisation, due to concerns about skewed distribution and study design.

Bayesian study design may improve the power of PIC trials or reduce the sample size by leveraging on existing adult data.39 Alternatively, the study population may be used to test two different hypotheses simultaneously by using a 2×2 factorial design as in the SUPPORT study.16 Registry-based randomised trial design may complement the strengths and address the weaknesses of randomised clinical trials and registry-based comparative effectiveness trials.40 This study design has been hailed as the ‘next disruptive technology in clinical trial design’ because it has the rigour of randomisation, but for fraction of cost and length of recruitment as it uses existing high-quality observational registry-based clinical data. While this design is being used in adult cardiovascular research, it may be a solution to PIC trials in the future. High-quality data registries such as Extracorporeal Life Support Organisation (ELSO) registry, Paediatric Intensive Care Audit Network (PICANET), Australia & New Zealand Paediatric Intensive Care (ANZPIC) registries have existed for over a decade. In fact, PICANET collaborated with investigators of CHiP trial and reduced the overall trial-specific data burden for recruiting centres.41 Although questions about this design such as blinding, consent procedures and data quality remain, they may not be insurmountable.

Research into paediatric TBI poses many of the methodological problems described above. Modelling of development of intracranial hypertension and outcome found that >320 children per arm of a trial would be needed for an intervention to show reduction in mortality. In the UK, recruitment would take >6 years.42 A multinational study has problems relating to standardisation of care. The investigators of Approaches and Decisions in Acute Paediatric TBI (ADAPT) trial propose conducting observational studies using ‘comparative effectiveness research’ methodologies to determine important associations between therapies and outcomes by using statistical methods to control for confounders.43 Advanced statistical adjustment methods such as propensity score matching are increasingly used to eliminate the effect of confounding variables in observational trials in adults. However, its role in PIC research is unclear.


Co-enrolling children into multiple trials increases pool of eligible study participants. Several factors such as safety, validity and interaction between proposed therapies need to be considered when co-enrolling children.24 Some UK PICUs that had the opportunity did not co-enrol children into CATCH and CHiP trials, citing concerns including overloading parents with information. However, consent rates in the co-enrolling units were similar to those that did not. Parental attitudes towards co-enrolment in PICU are unknown, although evidence from neonatal research is supportive.

Outcome measures

As PICU mortality is very low (table 2),1 it is difficult to show further reduction in mortality as an outcome measure. Several investigators have therefore used surrogate outcome measures as the primary outcome. For example, the Safe Paediatric Euglycaemia after Cardiac Surgery (SPECS) study and CHiP investigators used healthcare-associated infections and ventilator-free days as primary outcome measures, respectively44 ,45; the Trial of Drotrecogin Alfa in Children (RESOLVE) study used a composite outcome measure called ‘composite time to complete organ failure resolution’46 and Transfusion Requirements in PICU (TRIPICU) study used new or progressive multiple-organ dysfunction syndrome.47 While this strategy may be valid and acceptable in some trials, widespread use of proxy short-term outcome measures is perhaps fraught with danger. If reduction of intracranial hypertension was used as primary outcome measure in Decompressive Craniectomy in TBI (DECRA) and Hypothermia Paediatric Head Injury Trial (HyP-HIT), we may have concluded the opposite as both interventions reduced intracranial hypertension, but resulted in worse patient outcomes.48 ,49

Role of networks

Several national and international networks of PIC investigators have been set up during the last two decades to tackle some of these challenges (table 3).

Table 3

Examples of paediatric intensive care (PIC) research networks

Networks provide improved access to bigger pool of study participants and expertise in grant application. They promote dissemination and standardisation of good practices, ensure widespread applicability of the results, and minimise need for duplication. Other research agencies such as MCRN and Scottish medicines for children network in the UK provide valuable trial expertise. Offering infrastructure support, improved coordination and maintaining quality of RCTs. The European network of paediatric research at the EMA, an international network of research networks, offers efficient internetwork and stakeholder collaboration, support to trial recruitment and may enter into discussion with ethics committees on issues relevant to multinational research in children.

Trial regulations are notoriously fragmented across the world. International Conference on Harmonisation brought drug regulatory agencies and pharmaceutical companies together to streamline regulatory assessment process; prevent duplication; reducing drug development times and resources. Good clinical practice guidelines drafted by the International Conference on Harmonisation of technical requirements for registration of pharmaceuticals for human use50 is an excellent example of solutions with potentially worldwide applicability.


PIC research is essential. Several dilemmas relating to ethics and methodology remain. Further qualitative research is required to study problems with informed consent, assent and deferred consent processes from various stakeholders’ perspectives. Consultation and dialogue with stakeholders, collaborative networking, harmonisation of regulatory frameworks, use of novel statistics and trial designs may help solve many of these dilemmas relating to PIC research. In spite of good outcomes in modern PIC, we owe it to the children to continue conduct of good quality research and ensure dissemination of evidence-based practices that are fit for the future.


Dr Kevin P Morris for his helpful comments at draft stages of the article.


View Abstract


  • Contributors HKK: literature review, initial draft of article and multiple revisions. BRS: design, content of the article, literature review and multiple revisions.

  • Competing interests None.

  • Provenance and peer review Commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Atoms
    R Mark Beattie
  • Clarification
    BMJ Publishing Group Ltd and Royal College of Paediatrics and Child Health