Jaundice And Autism

Earlier this month, a study was published in Pediatrics that suggested that there might be a relationship between neonatal jaundice and autism.  After reading the study several times, I think I can best sum it up as a little weak and very confusing.

In this study, researchers examined the electronic health records of at all 733,826 children born in Denmark between 1994 and 2004 and collected data on a variety of factors.  These factors not only included whether the child had jaundice or one of several developmental disorders but also included data on their birth weight, gestational age, fetal presentation, apgar score, demographic information about the parents, and other factors.  Or, in short, they collected a lot of data.

The researchers found that almost 5% of the children in the study had a diagnosis of neonatal jaundice noted in their medical records.  As would be expected, these children were more likely to be born premature (36 weeks or earlier) by a margin of 43% to 4.4%, more likely to be underweight - 53% to 14%, and more likely to have an irregular fetal presentation - 29% to 17%.  I don't know if any of these relationships are statistically significant as the study does not say, but I don't think that really matters as none of these relationships are controversial.

Moving onto the autism side of things, the study found a total of 1,721 children who had a diagnosis "in the spectrum of disorders of psychological development".  This spectrum included disorders of speech, motor, and social development as well as autism and intellectual disability (ID).  In this group, there were 577 children that had a diagnosis that put them on the autism spectrum and 1,239 children that had intellectual disability.  There were 74 children in this group that had both intellectual disability and autism.

The researchers crunched the numbers and found some rather peculiar results.  Jaundice was found to be associated with autism but only in full term infants.  To make matters more confusing, this association partially disappeared when the children were separated out by the time of year that they were born (summer vs winter) and whether they were first born or not.  Children with jaundice who were born in the summer or who were first born did not have an increased risk of autism, while children born in the winter or were not first born did.

As I said, I find this result to be rather confusing.

The lack of association between jaundice and autism in premature children is odd.  Children with jaundice are much more likely to be premature and premature children are more likely to have autism.  The numbers found in this study bear that out - 43% of the children with jaundice were premature and prematurity is more common in the children with autism (7.8%) than the general population (6.3%, 4.4% in those without jaundice).  Intuitively it seems like there should have been a relation there, but there wasn't.

The authors try and explain away the time year effect by exposure to sunlight and hence vitamin D levels, but, while that sounds good, there is no demonstrated correlation between vitamin D and autism.  Similarly, they try to explain the order of birth effect by differences in hospital discharge policies and maternal antibodies in second and later pregnancies.  I don't find any of these reasons convincing as I like to see actual data behind assertions rather than just magical thinking.  Although, to be fair, some studies have hinted that there might be a seasonal bias with autism.

I have to question how significant the association found is.  It is one thing to say that there is an increased risk but the actual significance of the association rests on the count of cases.  So I would have liked to see the actual number of children who had autism and jaundice but the only figures provided are hazard ratios (HR).

Not to get too technical, but this hazard ratio, which is basically a relative risk, provides an estimate of how much higher the (relative) risk of the outcome (autism in this case) is if the child had exposure to what was being measured (jaundice) when compared to those who were not.  So in this case, an HR of 1 would mean equal risk while an HR of 2 would mean that the children with jaundice were twice as likely to have autism than the children who didn't have autism.

The overall unadjusted HR for all children with autism was found to be 1.84 while that number dropped to 1.56 after being adjusted for maternal smoking, apgar scores, mother's citizenship, birth weight, and congenital malformations.  So we aren't talking about a large increase of risk.  But perhaps more telling is the ranges of these values.

At the risk of getting a bit technical again, the study says that the 95% confidence intervals for these figures are 1.05 to 2.69 and 1.05 to 2.30, respectively.  This confidence interval basically means that we would be 95% certain that the "real" value of the HR would be between these values.  These values represent a rather large range - from no increased risk to about 2.5 times the risk - which suggests that the increased risk are based on relatively few cases.  A result based on a fewer cases or, more appropriately, on a a small difference in the relative number of cases, would be weaker than a result based on a larger number of cases or a larger difference.

A crude example here would be if the unexposed group had 1 case in a group of 10 while the exposed group had 2 cases in a group of ten.  In theory, the exposure group would have twice the risk but in practice that risk could shift drastically if you looked at a larger sample.

And, speaking of relatively few cases, consider the total number of children with autism in the study.  In ten years only 577 children with autism were born in Denmark?  That would equate to a rate of about 7.8 per 10,000 which is extremely low when compared to the 91 per 10,000 recently found in this country.  This isn't data from the 80s or early 90s when you would expect to see this rate of autism, this data was from the late 90s and almost halfway through the 2000's.  It it possible that the 577 number would increase over time as the children aged (the youngest children in the study would have only been a little over 3 when the data was collected), but still, if you considered all of the children found to have a disability included in this study, we are still only talking about 23.5 per 10,000.

Either there are simply fewer cases in this country or the majority of children with autism have been missed.  If it is the first reason, then I would have to wonder at how applicable these results would be the world and, more importantly, why researchers aren't trying to figure out what is different about this country.  If it is second, then the association found would likely be flawed as well.

So, as I said at the beginning, the results from this study are confusing.  If I had to guess, I would think that the association found wasn't real, but that is just my opinion.  The only way to know for certain would be to do more studies.

References

Maimburg RD, Bech BH, Væth M, Møller-Madsen B, Olsen J. Neonatal Jaundice, Autism, and Other Disorders of Psychological Development. Pediatrics. 2010 Oct 11. [Epub ahead of print] PubMed PMID: 20937652. doi: 10.1542/peds.2010-0052