Randomized controlled trials are the bedrock of evidence-based medicine. If a treatment has a good theoretical rationale, and preclinical data showing efficacy, the only way to prove efficacy in the human is to randomise patients to the treatment, compared to an alternative, which should usually be some sort of standard therapy, and compare clinically important outcomes. In order to be reliable, randomisation should be masked, which means that, once the patient is enrolled in the trial, and prior to pressing the “randomise” button, the investigators are unaware of which group they will be enrolled.
Masking of the actual intervention is not always possible, for example when comparing different modes of assisted ventilation, and there is less empirical evidence that it makes a difference to results. In particular, in neonatal research I don’t think there is any comparative evidence that shows whether the results of masked trials are systematically different to unmasked trials of the same intervention. The placebo effect, or the improvement in outcomes of control groups, is often misinterpreted as being evidence that the human body has great powers of self-healing, some people even talk about “harnessing the placebo effect”. But, when the outcomes of interest are objective outcomes, the majority of the placebo effect is in fact, “regression to the mean”, or simply that extreme findings usually become less extreme with time, and that most patients recover from most illnesses.
As an example, an uncontrolled trial of a drug (or any intervention) for apnoea will usually show an improvement in apnoeic spells, for a number of reasons. Babies tend to be enrolled in studies when their apnoea is troublesome, and they will therefore, usually, have fewer apnoeas after enrolment. In addition, in this particular example, apnoeas get better with time, so any trial without controls will tend to show improvement over time. But there is really no reason to think that treating babies with a placebo will have any more effect on apnoeic spells than simply not treating them with anything, as long as an objective measure of apnoea is used. Uncontrolled trials of medications for hypotension, as another example, will enrol babies who have blood pressures lower than average; overall such babies will subsequently have higher blood pressures, even if the drug has no effect. But having no treatment, compared to having a placebo infusion will not change that occurrence, both non-treatment and a placebo will have identical effects.
In studies with objective outcomes therefore, one could question the importance of masking the intervention. In my Cochrane review of inhaled nitric oxide for term and late preterm infants, as one example, the outcome “death or ECMO” is very similar between the masked and the unmasked trials. There were a few of both, and I compared the RR and confidence intervals between the masked and the unmasked trials, the results being very similar with the RR for the outcome “death or ECMO” being 0.66 for the masked studies, and 0.7 for the unmasked trials.
This question becomes extremely important when the intervention is a parenterally administered medication. In babies with no IV access in place, or when the medication must be given by another route (IM, subcutaneous…) the tendency in older publications was to give placebo injections, which inevitably create pain. For example, in a trial of erythropoietin prophylaxis published in 1994, control babies received placebo subcutaneous injections 3 times a week for up to 6 weeks. It seems to me to be highly unlikely that subcutaneous saline has any impact on erythropoiesis, not even a “placebo effect”, so the up to 18 painful injections were completely unnecessary. The more recent trial of Juul et al (and some older trials by Ohls and colleagues) used placebo injections for the intravenous phase of the trial, and when an IV was no longer in place, they avoided placebo subcutaneous injections by using sham procedures, in which curtains were drawn around the bed, and a bandage placed where the injection would have been.
This may be inconvenient, compared to just supplying vials with masked information on them and giving the unknown contents by injection, and it may be more costly, but the huge advantage of not inflicting pain on control babies must surely be worth it. A recent article in Acta Paediatrica discusses this issue, and also concludes that placebo injections are neither necessary nor ethically acceptable.
One recent article, which describes a potentially important improvement in RSV prophylaxis, was this one Griffin MP, et al. Single-Dose Nirsevimab for Prevention of RSV in Preterm Infants. N Engl J Med. 2020;383(5):415-25. The authors randomized preterm infants not eligible for RSV prophylaxis in their home countries, to receive either nirsevimab or placebo, 969 received active drug and 484 were randomized to have an intramuscular injection of saline. IM injections hurt. We should only give an IM injection to a newborn infant if there is some benefit to them. The primary outcome of the study was RSV infections requiring medical assistance, which were dramatically reduced from 46 (9.5%) to 25 (2.6%), hospitalisations from RSV were also reduced, from 4% to just under 1%.
Nirsevimab is potentially a significant advance in RSV prophylaxis, a single injection appearing to provide protection for the entire RSV season. Nevertheless, 481 infants received an intramuscular injection of saline. There is no possible benefit to the infant of this painful procedure. The published protocol notes that the blinding was performed at each individual centre, therefore there was an individual who was unblinded at each participating centre. The unblinded individual could easily have been the healthcare worker giving the injection, who could have performed a sham procedure on the control babies.
Even if blinding of the intervention is considered essential (and I hesitantly suggest that it was not, surely RSV infections would be identical in an open-label untreated control group and a masked control group) the blinding could have been maintained by placing an adhesive dressing on the thigh of the control babies, rather than subjecting them to a painful IM injection.
Another recent example is this Rosenfeld WN, et al. Stannsoporfin with phototherapy to treat hyperbilirubinemia in newborn hemolytic disease. J Perinatol. 2022;42(1):110-5, full term babies with a diagnosis of hemolytic jaundice were randomized to stannsoporfin or control, with the primary outcome being changes in serum bilirubin concentration. The 30 control babies received IM saline. I can think of no good reason for subjecting the control babies to the pain of the placebo; surely the lab tech analysing the serum for bilirubin concentrations will not be influenced by knowing which group the infant was in? Even if it was thought that other important secondary outcomes might be influenced by knowing which group the infant was assigned to, the intervention could equally well have been masked by a sham procedure without painful injection. But the only secondary outcomes listed all depend on the serum bilirubin concentrations. There is a plan to perform long term neurodevelopmental outcome evaluation in the infants; I guess it was thought to be just about feasible (and I would challenge that assumption) that knowledge of treatment group could have an impact on neurological or developmental outcomes. Even if this is the reason for maintaining masking of the intervention, such masking does not require intramuscular placebo injections.
Surely it is time to abandon additional unnecessary pain in research participants. We could start with banning placebo skin-breaking injections. Studies in newborn infants, who obviously don’t know themselves which group they are in, could be performed unmasked if the primary outcome variable is objective. If there is some subjectivity in the determination of the major outcomes than masking can be maintained by the use of sham injections.
A major problem is the way painful procedures are evaluated by ethics review committees. One of the worst studies in terms of pain inflicted on the neonatal participants was another RSV prophylaxis trial, the MAKI trial, where infants were subjected to either monthly palivizumab, or monthly IM placebo injections, to a maximum of 5 intramuscular injections of saline. Unusually, this trial was also the source of an article trying to justify its ethical approval. That article concluded
The Institutional review board (IRB) concluded the study has high clinical relevance because the benefit of 50% chance of protection by palivizumab outweighs the risk of side adverse events due to intramuscular administration of placebo.
It is actually impossible to argue with that conclusion, the study was indeed of high clinical relevance, and the “risk of side adverse events” from up to 5 IM saline injections is negligible. But only if you think that pain is not an adverse event. If you include pain as an adverse event the “risk” of adverse events was 100%.
The authors try to justify the use of the IM placebo, without ever mentioning pain, as follows “A placebo controlled control group was necessary because the primary objective will depend on parent-reported daily scores of wheezing along with information from parent-reported questionnaires”. Firstly, I question that rationale, is there any reason to believe that parents would provide biased scores of daily wheezing based on whether the child actually had a placebo injection compared to being enrolled in an untreated control group? Even if there were some evidence of such an effect, the placebo injections could have been replaced by sham injections.
The book from the Institute of Medicine “Ethical conduct of clinical research involving children” has a chapter “Defining, Interpreting, and Applying Concepts of Risk and Benefit in Clinical Research Involving Children” describing how to determine risks, and tries to define “minor increase over minimal risk”, in research with children as research participants. It includes a table which illustrates the hidden way in which pain is taken into account. The table lists “routine history taking” and a “complete neurological examination” as procedures with minimal risks, which we surely cannot argue with. But in the same category is included “venepuncture/fingerstick/heelstick”.
From a purely “risk” point of view, if pain is not considered a risk, then I guess that makes sense, but surely examining a baby and sticking a needle into them should be considered differently? The table also includes, as a minor increase over minimal risk, a lumbar puncture. Lumbar puncture is an extremely low risk procedure in the otherwise stable newborn, why is it given a higher risk status? Is it because we know it hurts, a lot? The only place pain is mentioned in that table is for two other “minor increase over minimal risk” procedures: skin punch biopsy and bone marrow biopsy, where “topical pain relief” is added as part of the name of the procedure. One might wonder why pain relief is not mentioned for heelstick or for lumbar puncture. In another part of the book it is stated “children should always be given the option to receive a topical anesthetic to reduce needle-stick pain”, but I can find no mention of routine analgesia prior to painful procedures in the newborn. The only mention of intramuscular injections is that they are more risky in children with hemophilia
A new publication from Ruth Grunau and the group in Vancouver who have performed amazing research into the adverse impacts of pain, and how to minimize it, has just appeared. McLean MA, et al. Association of Neonatal Pain-Related Stress and Parent Interaction With Internalizing Behaviors Across 1.5, 3.0, 4.5, and 8.0 Years in Children Born Very Preterm. JAMA Netw Open. 2022;5(10):e2238088. This study examined child behavioural patterns at the ages mentioned in the title, and determined the association with painful procedures in the neonatal unit. There was a clear correlation, after correcting for multiple other factors including gestational age, between having more painful experiences and having more internalizing behaviours, across all of those ages. They also showed that a more supportive, positive, and less stressed family environment could mitigate those impacts.
This study, among many others, emphasizes that we need to do all we can to reduce pain in the neonatal period, and any additional avoidable pain should be prohibited. This must include the use of placebo injections in research, which can always be avoided.