A preliminary RCT in babies undergoing therapeutic hypothermia (Wu YW, et al. High-dose Erythropoietin and Hypothermia for Hypoxic-Ischemic Encephalopathy: A Phase II Trial. Pediatrics. 2016). They randomized 50 babies with HIE who were being cooled to 1000 units/kg on days 1,2,3,5, and 7, or placebo. The primary outcome is reported as being 1 year outcomes.
The registration document says something quite different. The registration record states that the primary outcome is “Markers of organ function- The investigators will monitor organ function and adverse events until hospital discharge from the neonatal intensive care unit”. That record also states that the secondary outcome is the Alberta Infant Motor Scale, and the “other outcomes” include the Warner Initial Developmental Evaluation of Adaptive and Functional Skills (WIDEA) which is a questionnaire filled in by parents about a young child’s functional outcomes.
The published article, in contrast, states that the primary outcome was “12 month neurodevelopmental outcome assessed by (1) AIMS and (2) the Warner Initial Developmental Evaluation (WIDEA), The WIDEA was also administered at 6 months of age. Moderate to severe neurodevelopmental impairment at 12 months was defined as an AIMS score less than the fifth percentile for age, or a WIDEA score >2 SDs below the mean based on normative data from typically developing infants at 12.9 months of age.”
I was very confused by this part of the document for a while, but I think the phrase “The WIDEA was also administered at 6 months of age” is just in the wrong place, and should have been put later in that paragraph. They don’t explain why they used 12.9 months, and a threshold of 76.4, for the WIDEA test. I don’t know anything about that test, and can’t find a lot of information on-line.
I also don’t know if the planned analysis was a comparison of the mean scores, or the proportion of babies with the composite “moderate to severe neurodevelopmental impairment”.
The registration document also says nothing about MRI.
There were 42 babies who survived and were seen in follow up; there were some differences in motor scores, both the mobility part of the WIDEA and the AIMS, which were improved in the group that received erythropoietin. But there was no difference in the proportion who were classified as “moderate to severe”. There were also some differences in the MRI brain injury scores, showing less cerebellar and subcortical abnormalities.
Encouraging, but by no means definitive, we need more data before this becomes routine. Fortunately there are some in progress, so we shouldn’t have to wait too long. Also good news is that it seems to be a very low toxicity intervention, even in this group of intensive care dependent babies. Once we are sure of efficacy the only thing to stop its use would be the cost, which is not huge.
This study also points out the importance of registration of trials. Trial registration was initially conceived of as a way to reduce publication bias, it is essential to register trials before they are started, and to ensure that the eligibility criteria and outcomes are accurately recorded. If there is a good reason for changing those criteria or outcomes after registration, this should be explained in the text of the publication. Many trials are published and the outcomes don’t match, which can raise all sorts of questions. Trial registration is not just an extra chore for investigators, but an important part of the performance of a clinical trial, and is, in fact, one of the less onerous parts. It doesn’t take a lot of effort to get an accurate registration document in place.
One of the things that the Cochrane Collaboration has introduced into the “risk of bias” evaluation is an evaluation of whether the reported outcomes match the planned outcomes, for which the registration documents, or a published protocol if it is available, are used. For older trials it is often impossible to find the information, but more recent, registered trials often show some differences between the planned and the reported outcomes, which is rarely well explained.