Ethical pitfalls in neonatal comparative effectiveness trials

Evidence-based medicine has been embraced wholeheartedly, and rightly so, as the best approach for reducing clinical uncertainty and ensuring that patients receive treatment and care that are efficacious (i.e. they work) and effective (i.e. they work in real life). High-quality evidence comes from high-quality clinical research. It would hence be reasonable to assume that these two would form a closely integrated partnership. Alas, this is not yet the case. So many uncertainties in medical care relate to treatments and practices already widely in use. In neonatal medicine, for example, some of us use protein-carbohydrate fortification of human milk and some of us do not, some of us stop enteral feeds during blood transfusions whereas some of us do not, some of us reach for dopamine when blood pressure falls while some of us use dobutamine. For our patients, these uncertainties represent a lottery, the throw of the dice that determines whether they receive the treatment advocated by Dr. A or Dr. B. They deserve better than this. Randomization is considered the gold standard approach to eliminating the clinician bias that very often dominates the choice of treatments. Randomization reduces the influence on outcomes of confounding by unknown factors, and ensures that every patient has a fair and equal chance of receiving the best possible treatment when this is, in fact, not known. In an ideal world, every medical uncertainty would be addressed in this way. The evaluation of treatments that are in accepted use has been termed 'comparative effectiveness research', i.e. the comparison of existing healthcare interventions to determine which works best, for whom and under which circumstances. Recently a long-standing uncertainty, the optimum saturation target for preterm babies receiving oxygen was put to the test of randomization. The accepted standard-of-care saturation range of 85-95% has been used for a considerable time and its use is intended to avoid both levels of oxygen that are too low or too high. Investigators in the UK, Australia, New Zealand and the USA designed randomized controlled trials to provide more precise guidance, by determining whether targeting the lower end of the accepted range (85-89%) resulted in reduced retinopathy of prematurity when compared with the upper end of the accepted range (91-95%). Between 2004 and 2009, the US SUPPORT trial (Surfactant, Positive Pressure and Oxygenation Randomized Trial) recruited approximately 1,300 infants and showed that babies at the higher end of the recommended oxygen saturation range had a greater incidence of retinopathy of prematurity, but that, unexpectedly, babies at the lower end had a higher risk of death [1]. The data monitoring committees of the BOOST II (Oxygen Saturation and Outcomes in Preterm Infants) trials in the UK, Australia and New Zealand reviewed their interim data, confirmed the higher risk of death in babies randomized to the lower saturation range, and halted further recruitment [2]. Without the trials, the lower saturation target would have continued to be applied to many babies, and many would have died as a result. Though many uncertainties remain, the trials facilitated advances in care. However, in March 2013, the lead investigators for the SUPPORT trial were informed by the US 'Office for Human Research Protections' that they were 'in violation of the regulatory requirements for informed consent, stemming from the failure to describe the reasonably foreseeable risks of blindness, neurological damage and death' [3]. This extraordinary conclusion indicates that the US regulators considered the researchers to be at fault for failing to foresee an unexpected trial result, and for randomizing babies to receive oxygen within the accepted standard-of-care limits. The ruling further implies that the regulators consider that clinicians are acting ethically when they deliver an accepted but non-evidence-based treatment based upon their personal bias, but are acting unethically when they make the selection by randomization. Clearly, there is a gulf between the view of the medical profession and that of the regulators regarding the ethical and scientific validity of randomization as a means to select treatments in comparative effectiveness research aimed at reducing uncertainties in care. What are the ways forward? I suggest that, in order for medicine to advance, a paradigm shift is necessary, involving a deeper public (and regulator) understanding of randomization as the fairest approach to allocating treatments that are in wide and accepted use, but where the evidence base is actually uncertain, so that the chance of receiving the as yet unknown best treatment is unaffected by clinician bias, and where care is delivered along a clearly designed, closely monitored pathway. In practice, peer review, regulatory approval, patient involvement and the delivery of explanation and information would be the same as for research involving experimental treatments. The key difference would be that randomization would be the recommended default and patients would be offered the opportunity to opt out, rather than be invited to opt in. For neonatal medicine, this would reduce the risk of 'injurious misconception', where trial entry is inappropriately rejected by parents because of an exaggerated and disproportionate perception of risk [4] that is brought on or magnified by the burden of making decisions at this difficult and stressful time. Randomization to treatments that fall within accepted practice and are considered standard-of-care involves no research-related risks to participants, and as trial data can increasingly be extracted from electronic clinical records [5], the costs and burden of data collection placed upon clinical teams will be minimized and, ultimately, the resolution of uncertainties about treatment can be hastened. It should also be noted that this approach fulfils the four cardinal principles of research ethics, namely: autonomy, justice, beneficence and nonmaleficence as well as upholding the responsibility of all doctors to strive to reduce uncertainty in the care they provide to their patients [6].