They are at it again.
The people in the Public Citizen health research group don’t understand evidence based practice, they don’t understand clinical research and they don’t understand neonatology. Which doesn’t stop them from making a fuss about high quality important neonatal research, claiming that it is unethical, and that the consent forms and IRBs are again inadequate.
This time they are claiming that the TOP trial is unethical, and the consent forms and IRB approvals are inadequate. The TOP trial (Transfusion of Preterms) is a large multi-center RCT comparing 2 different treatment algorithms, it is designed to address a serious area of clinical uncertainty. The situation is very analogous to the oxygen saturation uncertainties prior to SUPPORT, COT and the BOOST trials.
Speaking of which John Lantos has published a vigorous defense of SUPPORT and the support investigators, even while a Nature editorial is much more ambivalent. I will quote the penultimate paragraph from the Nature editorial as I think it illuminates much of the misunderstanding of these trials.
Put yourself in the position of a parent with an extremely premature infant. Would you make the decision to enrol your child in the trial if the consent form stated in simple language that babies assigned to one group were more likely to go blind, and that those in the other were at a higher risk of getting neurodevelopmental disabilities? Equally, would you decide to enrol if the form spelled out that, if you do not take part, your own physician and institution might keep your infant in the middle of the range, trying to avoid either outcome? Perhaps you might, but you would do so with full knowledge of the attendant risks. The parents in this case could not do so.
My first response is to note that the babies did not have more blindness in one group or the other. There was also no difference in neurodevelopmental disabilities. The paragraph also fails to note that being in the middle of the range might well be worse than either of the other 2 ranges (they could theoretically have both a higher rate of RoP and more deaths), we actually don’t know, and the consent forms did note that infants outside of the study might be treated with any of the currently used target ranges. Which would include the lower range which is now known to increase mortality.
So what is the current situation with transfusion thresholds. They are extremely variable, from NICU to NICU and from neonatologist to neonatologist. There are no clear indications for transfusion in the majority of babies. Those who are actively bleeding and are hypovolemic are clearly appropriately transfused urgently, but that is the minority of transfusions in the NICU. Most are given simply because the hemoglobin number is low. But we don’t know what the best number is at which to transfuse, if there is one.
There are all sorts of reasons why transfusing at a higher or lower threshold might be preferable, stored blood has risks relating to ages of the cells, abnormalities of the cells due to the storage media, infections, etc. But they do increase oxygen carrying capacity and in cases where oxygen transport might be limited this could affect outcomes.
The members of Public Citizen writing to the Secretary of the HHS in the USA note the 2 previous relevant trials. One is the trial by Ed Bell, which enrolled 100 babies between 500 and 1300g birth weight with a primary outcome which was the number of transfusions received. Of note some of the babies in each group (about half) had already been transfused before being randomized. One of the secondary outcomes, and there were 17 of them, was different between the groups, that is, the combined incidence of grade 4 hemorrhage and PVL, which was more common in the restricted transfusion group 6 babies vs 0. Now usually we include all serious IVH (grade 3 and 4) when talking about serious brain injury. There were actually more grade 3 hemorrhages in the liberal transfusion group, 8 vs 1. So using the usual definition of serious brain injury, it was actually more frequent in the liberal transfusion group: 8 babies vs 11.
According to the wording of the Public Citizen document, this is evidence which suggests poorer outcome in the conservative group. Er, no. It is evidence which is rather unreliable which might possibly mean something and needs to be confirmed with an adequately powered trial.
The authors of the letter do note that the long term follow up of Bell’s trial was actually better in the restricted transfusion group. But they don’t like that finding so they criticize the low follow up rate, while not making any criticisms of the selection of the outcome variable in the initial publication.
The second trial was one I was involved in, the PINT trial. That was an RCT in 10 NICUs in Canada, the US and Australia, of restricted versus liberal transfusion regimens in 451 babies under 1 kg birth weight. That study showed no differences between any clinical outcomes, the primary composite outcome, which included brain injury, was not different between groups. Also the components of that outcome, specifically brain injury on ultrasound were not any different (there were actually slightly more babies with liberal transfusions who had brain injury).
The follow up publication from that study included nearly all the surviving babies, and showed no significant differences in the outcomes. There were a few more babies with a delay in their development (Bayley2 MDI at 18 to 21 months which was <70) , 38/156 (24.4) with restricted transfusions vs 29/165 (17.6) with liberal transfusions. This was not significant, and in any case Bayley scores at this age have little predictive power for outcomes of clinical importance. An unplanned post hoc analysis showed that more restricted infants had a Bayley which was below -1SD.
So we clearly have NO data to support transfusing at any particular hemoglobin threshold: the 2 trials which are analogous to TOP had some minor differences in outcomes, neither of which was confirmed by the other trial. What does this mean? Well in the best of worlds that would mean that we should mount a larger multi–center trial, which adequate power to determine if there is any difference, comparing transfusion threshold practices which are within current limits of practice. That is exactly what we have in TOP.
We are finally doing some of those trials that we have needed for a very long time, and this trial, planned to have 1800 infants of less than 1000g is sorely needed to give us some evidence base for future practice.
What on earth do public citizen have against that? Their arguments are about the study design and about the consent forms.
The study design argument, after vaguely stating that there are ‘many features of the protocol which raise ethical concerns’ they only mention one feature which they think makes the study unethical.
A. Lack of a control group. They state that usual clinical care is that
‘decisions regarding the level of hemoglobin at which to transfuse blood would be individualized, based on multiple clinical factors.’
Just like their lack of understanding of how saturation ranges were chosen, they completely misunderstand how we decide to transfuse a baby. As I mentioned above, most transfusions are given because the complete blood count shows a hemoglobin below a certain level. That level may be fully protocolized in some units, or it may be according to individual physician preference. But in 2 NICUs in the same city exactly the same baby will get transfused according to different thresholds. When the on-call doctor changes, the decision to transfuse may change. This is actually entirely reasonable as we have little evidence on which to base practice, and what evidence we have shows no effect of different transfusion thresholds.
According to Public Citizen transfusions are currently given according to individual patient factors including:
Current level of anemia
Active bleeding or coagulopathy
The degree of supplementary oxygen required
Level of respiratory support (e.g. intubation, positive pressure ventilation, nasal cannula)
Age of the baby
Reticulocyte count (count of new red blood cells)
The need for medication to help the heart pump blood (inotropic support) and
Major comorbidities, such as heart disease or sepsis
Other factors sometimes taken into account that support transfusion include
Lactic acidosis
Increasing episodes of apnea (stopping breathing)
Persistent tachycardia (abnormally fast heart rate)
Persistent tachypnea (fast breathing) and
Poor weight gain
At least some of that is true, as a description of current practice. Unfortunately almost none of it is evidence based, some of it makes no physiologic sense, and those parts which are clinically important are allowed anyway in TOP, or are exclusion criteria.
In detail:
Current level of anemia: Yes we do this, but we don’t know when we should, that’s what this trial is about.
Active bleeding or coagulopathy: transfusions are allowed for this in TOP
The degree of supplementary oxygen required and the Level of respiratory support (e.g. intubation, positive pressure ventilation, nasal cannula): Yes we do this, and the TOP algorithms include whether the child is on respiratory support, however, in fact it makes no sense. If your saturation is 92%, then what difference does it make to your transfusion requirement if you are in room air or ventilated with 40% oxygen? The cardiac function should perhaps affect transfusion requirements, but not pulmonary function.
Age of the baby. We don’t know if, or when, transfusion thresholds should change, but they do, arbitrarily, in the TOP protocols.
Reticulocyte count (count of new red blood cells). I don’t use retic counts for when to transfuse.
The need for medication to help the heart pump blood (inotropic support): Babies who are in shock will be allowed to get transfusions,
Major comorbidities, such as heart disease or sepsis: Babies with heart disease are ineligible, with sepsis can get transfusion if they are thought to need them.
Other factors sometimes taken into account that support transfusion include
Lactic acidosis: I don’t know if anyone uses lactates as a sign to transfuse, except maybe for babies in shock (see above)
Increasing episodes of apnea (stopping breathing). There is no good evidence that apnea is affected by transfusions, there was no effect on apnea in the PINT trial.
Persistent tachycardia (abnormally fast heart rate) I don’t think anyone transfuses for this alone.
Persistent tachypnea (fast breathing): no evidence to support this.
Poor weight gain: no evidence.
Public Citizen think there should be a ‘usual care control group’. By which they mean a 3rd group which is treated according to whatever the hell you feel like. I don’t know how on earth they think that will help.
Yes if there was an evidence-based ‘best therapy’ which was known to be better than other algorithms, then a standard-of-care group would be required ethically. Just as now, if someone wants to do another trial of oxygen saturation targeting, the control group would have to be the high saturation limits as studied in SUPPORT, COT and the BOOST2 trials.
By the arguments that Public Citizen are making a large proportion of clinical research is unethical. Any study that compares to protocols of care is unethical if there isn’t a 3rd group which is: carry on doing whatever you feel like.
That is what I meant by saying at the start of this long post, that they don’t understand evidence-based practice. Our care for our patients should be based on some evidence about what is the preferred approach. If we currently have enormous variations in practice, no reason for choosing one approach rather than another, it is ethically preferable to perform the randomized trial comparing two protocols of care than just continuing to guess what might be best for our patients, using our prejudices and our cognitive biases. Including a treatment arm which is non-evidence based, in which patients get treated according to the whims of the doctor is a terrible, unethical, way to perform research.
I will quote from John Lantos regarding the belief by OHRP and Public Citizen that we should exaggerate the risks of our research projects:
Why would they require this? The ideas that research is risky compared to non-validated therapy and that care by protocol is inferior to care by individualised clinical judgment have been around for a long time. They used to be widely held by doctors and criticised by bioethicists as unjustifiable medical paternalism. William Silverman, a pioneer of neonatology and a staunch advocate of better clinical studies, was familiar with such arguments. He identified them as a belief in ‘mystical certainty’ rather than an acceptance of ‘scientific uncertainty.’
Public Citizen are of the opinion that being treating according to the doctor’s gut feeling is somehow safer than being in a clinical trial. We know that is not true, it is an argument from medical infallibility, and it is ridiculous.
Neonatal Research 