The moral risk of methodological purity.

Babies feel pain.1 A baby's painful experiences can have long-term psychological effects.2 We have effective methods to relieve pain. It is morally imperative that we use them to reduce babies' pain and suffering as much as safely possible.3 But sometimes, we do not know how best to treat pain. Then, we have a moral imperative to carefully study different approaches to find the ones that are safest and most effective. Sometimes, research studies themselves can cause pain. In this issue, Theubo and colleagues present a (hypothetical) study in which the desire for ‘methodological purity’ led researchers to give sham intramuscular injections to critically ill babies who had been randomised to the placebo arm.4 They criticise this practice because, in their view, it causes pain that harms babies without any offsetting benefit. They acknowledge that the use of such a control group will increase the ‘methodological purity’ of the study. Methodological purity is a beautiful thing, but too much such purity can leave researchers morally tainted in other ways. The history of research ethics is full of stories about scientists causing harm because their moral sensibilities were numbed by their intense desire to do good. Often, the stories involve vulnerable populations such as racial minorities, citizens of low-income countries, women, people with disabilities or small children. Black men have had penicillin withheld so that scientists could study the natural history of untreated syphilis. Children with cognitive disabilities were deliberately infected with hepatitis to test preventive treatments. Elderly patients with dementia were exposed to ionising radiation to see what doses are carcinogenic. Babies are (still, today) subjected to painful procedures without adequate analgesia because of a once widely-held belief that babies could not feel pain. Such studies are still done. Barrington recently documented three such studies of analgesia for neonates that included a placebo arm. He suggested that they were unethical and should be retracted.5 I agree. In studies of pain control in neonates, there is no ethically defensible reason to study new interventions against placebo. But are there sometimes situations in which a painful sham procedure can be done (with analgesia) in order to increase the scientific rigour of the study? The Declaration of Helsinki once condemned the use of placebo in any circumstance in which a ‘proven intervention’ was available, recommending instead that studies should compare any innovative treatment with the proven effective alternative. But that position was revised. Now, the Declaration allows placebo controls when placebo ‘is necessary to determine the efficacy or safety of an intervention’ and when the use of placebo will not lead to ‘additional risks of serious or irreversible harm’.6 CIOMS has similar criteria, allowing placebo controls ‘When the use of an established effective intervention as comparator would not yield scientifically reliable results and the use of placebo would not add any risk of serious or irreversible harm to the subjects.’7 Our system of oversight for research involving human subjects is designed to be flexible and to allow experts to review each study based on its unique particulars. Let us assume that the hypothetical study described by Theubo et al was an important study done by researchers motivated by this ethical ideal. In the vignette, the study is described as ‘exciting’. Innovative treatment X ‘has been shown to improve lung function in a phase II study’. The researchers have moved on to a phase III study that includes the painful, sham, intramuscular injection for the placebo arm. The intervention is designed to both control for placebo effects and blind investigators to treatment assignment. To judge whether the use of a sham injection is ethically justifiable, we first must know more about the natural history of the disease and whether there is a standard treatment. We need to carefully review the results of the phase II study. We need to know the risks of the intervention.8 We need to know whether there are other drugs that target this disease in this population of babies? If so, it would be more appropriate to compare ‘treatment X’ to those drugs, rather than to placebo. The vignette is spare enough to allow different possibilities. The devil is in the details. A few years ago, some researchers debated the ethics of sham procedures or injections for those in the placebo arm of randomised trials. Cyna and colleagues argued that painful and potentially harmful sham interventions are unethical. They made it clear that they were not against placebo controls, but only against placebos that are ‘neither inert nor innocuous’. Sometimes, placebos involved sham operations or injections, procedures that can cause serious harm.9 As an example, they describe a study in which study participants in the placebo arm received a deep intramuscular injection.10 Kallmes and colleagues defended such studies. They opined that, in seeking the proper balance between the protection of research participants and scientific rigour, there are circumstances in which ‘…the use of an invasive procedure strictly as a control intervention in the setting of an RCT….is not only ethical but also essential for patient protection’. Kallmes et al go on to describe in detail some studies in which they think sham interventions are ethically justifiable. They argue that one must disentangle the effects of the natural history of the disease, the placebo effect or the anaesthesia used during invasive treatments. They stress that ‘the amount of risk to which the control group is exposed certainly can be customised as needed’. In other words, trade-offs between methodological purity and ethical imperatives are inevitable and demand case-based judgments about the proper compromises. The history of neonatology is rife with examples of treatments that were inadequately studied. That history should make us demand methodological rigour and accept that participation in research studies may protect participants from the harms of inadequately studied treatments. But the history of neonatology is also rife with examples of situations in which neonatal pain is ignored and babies are allowed to suffer needlessly. That history should predispose us to oppose any possibility of unnecessary untreated pain. The only way to know which moral taint to accept and which to avoid would be to have a more detailed description of the study and the study population. In evaluating such dilemmas, the burden of proof should always be upon the researchers to show why and how the scientific questions about safety and efficacy can only be answered by causing harm to research subjects. The bar should be set quite high. John D. Lantos