Just back to blogging from a few days at a very successful conference in Geneva, organized by Annie Janvier. At the conference I gave a presentation about prediction of long term outcomes in preterm infants. Some of the issues that I raised will be the subjects of a series of posts over the next couple of weeks, so I will be discussing the usefulness of ultrasounds, MRIs, EEGs, amplitude integrated EEGs, clinical examinations and clinical complications and anything else that has been investigated as possibly being of value in predicting long term outcomes among preterm infants.

A few general observations to start with:

Most reports that you can analyze to determine the reliability of a particular test or finding are flawed, and flawed in one very important way. The usually do not report the number of  patients with a particular finding during their hospital course who either died, or had active therapy withdrawn or withheld before they reach follow up. So for example we may have information about the infants with intraparenchymal hemorrhage who reached 2 years of age, but we often do not know about those, who may be a substantial proportion, who acutely died or had palliative care. This could be important, if those who had care withdrawn or died were systematically different from those who had on-going care.

Indeed I think it is likely that there are such systematic differences. I think it is likely that in some centers many or even most infants with intraparenchymal hemorrhage were offered redirection of care. Meaning that the parents of those who refused will have had a different background or different motivations to the rest. I think it is likely that in other centers withdrawal of active care has been offered selectively, meaning that those with less severe lesions are more likely to have survived. I think it is likely that in other centers redirection of care is not the primary consideration of the neonatologists, and it is parents who make the suggestion, again the characteristics of the parents will differ between those who survive and those who do not.

You will note that I use a phrase 3 times in that paragraph that I rarely use in anything I publish “I think it is likely”, I try to be as evidence-based as possible, but in this instance I cannot be. I am not aware of data about differential characteristics of patients and families whose babies survive with significant brain injuries (or significantly abnormal EEG or anything else) and those who die. Or whether the brain injuries differ between survivors and non-survivors.

I think in the future that studies of early prediction of outcomes of preterm infants should report the total cohort, those who had the findings of interest, those who died, those who had care redirected and those who survive. They should report the findings on the particular test of each category of patient to demonstrate whether survivors really represent the group of all those who had that finding. If not we could be really misled by the data.