The title is a slightly edited copy of the title a discussion article in JAMA, that discusses some important issues in IRB oversight of clinical research. (Kass NE, et al. Making the Ethical Oversight of All Clinical Trials Fit for Purpose. JAMA. 2024), and is part of a special issue including a renewed version of the Declaration of Helsinki World Medical Association Declaration of Helsinki: Ethical Principles for Medical Research Involving Human Participants. JAMA. 2024.and several other discussions.

Rethinking the definition of clinical research, and how we approach the essential protections of research subject is clearly needed.

In my NICU, if I have a 3 week old preterm baby who develops signs of septic shock with poor perfusion and a low blood pressure, in general my approach would be to give a bolus of normal saline (10 mL/kg) over a fairly short time, then to start a norepinephrine infusion, and usually start low dose steroids very early. I would start norepinephrine at about 0.1 mcg/kg/min, often the response is slow, and I may increase to as much as 0.4 mcg/kg/min before, often, being able to back off fairly quickly as the steroids and antibiotics have their effect.

How much of that approach is evidence-based? Just about none of it. I have published about the use of norepinephrine (Rizk MY, et al. Norepinephrine infusion improves haemodynamics in the preterm infants during septic shock. Acta Paediatr. 2018;107(3):408-13) and using steroids early (Altit G, et al. Corticosteroid Therapy in Neonatal Septic Shock-Do We Prevent Death? Am J Perinatol. 2018;35(2):146-51). But those small observational cohorts will never tell us if outcomes were improved or worsened by the approaches.

Were those publications “research”? We described our practice and the outcomes, in the hope that others will evaluate their own practice and compare with ours, and we had local ethics approval to go through the databases and publish our results. In another NICU, they might routinely give multiple fluid boluses, followed by dopamine, and only use steroids as a last resort. Also without a good evidence base. As each baby is so different to the last, observational studies have limited value. It is, indeed, very hard to imagine how comparing these approaches and determining best practice is possible, even with all the fancy statistical adjustment in the world, including the latest version of propensity score matching, unless large randomized trials are performed.

I can treat babies in this non-evidence-based way, and then compare with other approaches, or change my practice and then re-evaluate, with almost no oversight, and certainly no prior IRB approval. What if I wanted to do something much more scientifically valid, such as trying to standardize therapy as much as possible, then prospectively comparing the outcomes from my centre to other centres that have also tried to standardize their treatment approaches? Then, even if every baby gets exactly the same therapy as they would have done anyway, does this suddenly become a research study? The CNN, Canadian neonatal network, are actually doing such a study, in which centres who usually use norepinephrine for septic shock, and those who use dopamine try to standardize their initial treatment, and we will compare outcomes.

A much more valid study, and one which would cost much more, and be substantially more difficult to organise, gain approval, and perform, would individually randomise eligible babies to one approach or the other. An intermediate type of study, where NICU’s are randomized, has some advantages in terms of, for example, ensuring the team is familiar with the approach over time, and not requiring masked vials of study drug; but has major impacts on sample size. Either a cluster-randomized or individually randomized trial will require prior IRB approval.

One big question for an IRB is how to determine the risk of such a study. In the past, because there is a substantial risk of death, and of major complications, this should be labelled as a high risk study. The risks of a participant in the study are much greater than the risks of daily life, even of daily life of a preterm baby in the NICU. Surely, though, the risks of the research project should be a comparison of the risks of adverse outcome if you are in the trial compared to being not in the trial. What matters is whether there is additional risk over not being a study participant. Clearly, in the scenario I have described, the risks are identical to not being in the study. It should be considered as “no additional risk”.

The new Declaration of Helsinki does not go into any detail about which risks to include in evaluating a research project. It states that a project must be “preceded by careful assessment of predictable risks and burdens to the individuals and groups involved in the research in comparison with foreseeable benefits to them and to other individuals or groups affected by the condition under investigation.”

The discussion by Kass et al puts it this way: “Our specific recommendation here, with relevance for comparative effectiveness and other types of research on approved or widely used clinical approaches, is for oversight bodies to make their risk-level determinations, including whether a study qualifies as minimal risk, on the marginal or additional risks and burdens of research participation, relative to those of ordinary usual care.”

An accompanying editorial in JAMA points out how these proposals differ from current rules. According to OHRP current rules, “in general the reasonably foreseeable risks of research in a study include the already identified risks of the standards of care being evaluated as a purpose of the research when the risks being evaluated are different from the risks some of the subjects would be exposed to outside of the study”. In other words all possible risks of being a preterm baby in the NICU with septic shock undergoing intensive care are considered study risks. Which means that all possible adverse outcomes of the extremely high-risk situation are included, death, IVH, renal failure, and so on. That interpretation of the OHRP rules makes intensive care research just about impossible, as it requires the IRB to ascribe all the risks of being critically ill to the trial, even if there are no additional risks compared to not being in the study. I think that many IRBs already interpret the risks of a trial as being those additional risks, and I can only hope that this interpretation will become widespread.

Another question is about consent. Normally, I do not share with parents all the many the uncertainties regarding the best approach to septic shock, I do not ask them to make an informed choice between norepinephrine and dopamine, I don’t counsel them about the uncertain role and timing of steroids, and get them to sign a consent form. If I am planning to prospectively collect the data, with the explicit intention of comparing outcomes with dopamine centres, does that now require specific informed consent? Even though I am doing exactly what I would have done in any case? Indeed, so are my comparison centres, who use a different approach.

I think that this does not require specific informed consent, I am not even sure that we should obtain consent to contribute the baby’s data to a truly anonymous database for future comparative analysis.

What if I decide, for a while, to switch my approach to using dopamine, and collect the data, because I understand the uncertainty involved, and I think that dopamine is an equally valid choice? And my comparison centres will also switch to using norepinephrine. According to this new discussion article:

(Informed consent is required if)… the research removes from patients the ability to make a decision that may be meaningful to them—a decision that, in ordinary clinical care, they may have been invited to make.

They go on to note :some commentators support the view that consent can, ethically, be waived when risks are low and when consent is typically not sought for either approach being studied, others oppose complete waivers of consent requirements in some or all circumstances. Importantly, most of these critics of complete waivers support streamlined approaches to consent for these lower-risk, randomized studies.

In this situation, parents would not ordinarily be asked whether they prefer norepinephrine or dopamine, so the first criterion does not apply. However, my initial intervention choices are being determined by a written protocol, I think this should require that I tell the parents that their infant is getting a therapy determined by a protocol, a therapy which is, nevertheless, within the usual range of good clinical practice.

I don’t think it is appropriate to perform prospective comparative research, even with zero additional risk, without informing the parents. It seems to me to be hiding important information from the parents, and that is likely to create distrust.

I propose for such a study that consent should be altered to focus on informing parents of the important details of the study, and giving them an opportunity to opt-out of data collection. A 2017 article refers to this as “targeted consent”, whereby patients who are individually randomized to one of two reasonable, clinically appropriate treatments have a verbal information sharing session, and a brief targeted consent form outlining the patient’s rights. I don’t think that a cluster-randomized trial should be considered to be morally different, if all patients admitted in January have one approach, then in February we switch to another approach, from the patient’s point of view this is identical. Or if NICU A is randomized to dopamine, and NICU B is randomized to norepinephrine, and then switched over after 6 months. In each situation we are performing pre-planned research, and parents should know about it, and have the chance to opt-out of information collection, and even from non-essential aspects of the care protocol if there are any.

Of course, as always, if I have a valid reason to believe that this particular patient would benefit from norepinephrine, then I am morally obligated to do so, and treat them as a protocol violation, or a non-enrolled subject.

If I was to introduce levosimendan, however, a drug with almost zero use in the NICU, and no experience in the preterm infant with septic shock, then there would be a possibility of a major impact on important clinical outcomes, and the study should have completely different, and more extensive requirements and consent process.

After the dopamine and norepinephrine study finishes, it may be found that mortality is higher in one group or the other, or the incidence of a major complication differs between groups. I don’t think we can then, in retrospect, suggest that the risk of the trial changes.

This is what happened with the SUPPORT trial, where the 2 oxygen saturation target ranges, both considered to be acceptable within usual clinical practice were prospectively randomly compared. Outside of the trial and infant could have been treated with either of the 2 approaches, so, according to the standards that I am promoting here, this would be considered a trial with no additional risk. In fact, as we all know, we found a higher mortality with the lower target range, which created a storm, and was largely responsible for the OHRP guidance which I quoted above.

But, if it turns out that norepinephrine babies have a lower mortality in our study, even after correcting for every baseline imbalance possible, that does not change the a priori risk assessment of the study. To think otherwise requires some sort of magical thinking, that things which are unknowable because such a trial has never been done, should have been known before the facts existed.

Building cluster randomized comparative effectiveness trials on top of existing databases, such as the CNN, will allow us to improve therapy for future babies, with no additional risk to our current patients, and the chance of major benefits for the next cohort.