Folic Acid Increases the Risk of Autism aka the Dangers of Bad Data

As I talked about last time, another study on thimerosal was released last week.  Like some of the studies before it, this one found that not only was thimerosal not involved in causing autism but also that a higher exposure actually lowered the the odds of having autism.  Or in other words, being injected with a higher level of mercury actually greatly decreased your chances of having autism.

If that result seems a bit strange to you, you are not alone.  I don't think that many people would claim that this is actually a valid finding (although there are some) and the researchers acknowledged as much when they said that they "are not aware of a biological mechanism that would lead to this result."  Although, to be slightly glib, it would be one of the great ironies in recent history if thimerosal actually did actually protect against autism.

But still, the result reached statistical significance which means that it is unlikely to have happened by chance, and this isn't some fly by night paper were the statistics were in doubt.  The data in this paper was the result of a seven year effort and the published paper and supplemental material reaches almost 400 pages.

I think the answer to the puzzle is pretty obvious here and I am going to use another significant finding from the study to illustrate the point.  Buried on page 164 of Volume I of the supplemental material is the following statement -

"Use of prenatal vitamins containing folic acid was associated with a higher odds ratio of ASD"

This is also a statistically significant relationship.  If you look on page 169 of the volume, you will see that, if the mother took prenatal vitamins while pregnant, her children were more than twice as likely to have autism.  This result is based on the data that 96% of the mothers of children who have autism took folic acid while only 91% of the mothers of "typical" children did.  When you put these numbers into the models with the rest of the data - the same models used to determine the main no relationship result - out pops this strange relationship.

For those of you who don't know, folic acid is essential to many bodily functions, is used to help synthesize and repair DNA, and helps determine which parts of the generic material is active.  Pregnant women have been taking folic acid during pregnancy since at least 1993 because it can mostly prevent neural tube defects such as spina bifida.

Here we have yet another improbable result from this study.  So, what could be going on?

When you consider the facts that folic acid plays a role in maintaining and repairing genetic code and that autism seems to littered with many rare genetic mutations, it might be possible that taking too much folic acid during pregnancy could increase the chances of a child having autism.  However, even though the timing (1993 onwards) might agree with rising autism rates and there might be some kind-of plausible mechanism for the relationship, I don't think that this is likely.

Which leads to another possibility.  I am sure that most of you have heard the oft-repeated phrase about how correlation does not imply causation - this would be a classic example of it.  It is possible that folic acid use is highly related to another factor and it is this other factor that is causing autism.  If I had to guess, folic acid use during pregnancy could be be correlated with affluence.  Meaning that mothers who are more well-off are more likely to wait until later in life to have children and are more likely to do the "right" things during pregnancy - such as take folic acid.  But it is equally likely that this relationship could be flipped and affluence, which has been linked elsewhere to a higher risk of autism, is the dependant factor and it folic acid use that is the causative factor.  That is the problem with these sorts of relatinships, you are never quite sure what is a cause and what is simply along for the ride.

Another explanation would be that the association would happen by simple dumb luck.  Even with all of the robust methodologies and fancy statistics that researchers use, it is still possible for a significant result to appear that is due to simple random chance.  This is why it is very important that all scientific findings be replicated by other researchers using a different set of data.  The fact that thimerosal is actually protective has been suggested by other studies but I am not aware of any others finding a relationship with folic acid use.

Regardless, I think the most likely explanation for what is going on is the simplest - there is a problem with the study's data.  I would have to say that a problem with the underlying data is one of the most overlooked and under-considered problems out there.  If the data that the analysis and conclusions of the study are based on is biased or does not properly represent what you trying to study, everything that the data is based on is flawed.

Since this study found that a higher exposure to thimerosal decreases the chance of autism and that taking folic acid during pregnancy raises the chance of autism, I think it is safe to say that something is not quite right with the data in this study.

As the saying goes - garbage in, garbage out.