Friday, May 21, 2010

Clomid and autism

A recent spate of news stories has reported in glaring headlines that researchers have linked the use of fertility drugs like Clomid with autism. "Studies Link Fertility Treatments to Autism," "IVF, Infertility Drugs Might Boost Autism Risk," and "Infertility Drugs Raise Autism Risk?"

Gaaaaah. Et tu, News Media, otra vez? Was the MMR-autism debacle not a lesson learned?

There are two studies involved here, both presented at the IMFAR conference, neither yet published. The first is based on the Nurses' Health Study II and involves a large cohort of women, 3,985 of them, 111 of whom responded on a questionnaire that they have a child with autism. I'll leave out for now what one could infer about autism prevalence (at least among the children of nurses) from these data and just address some issues. The Time.com article delves into caveats about the study--in the sixth paragraph. But these caveats are significant.

For anyone who's reading this who's a woman who took Clomid or any other infertility drug and had a child with autism, please take heart in the following and try to avoid any self flagellation. For any woman who is taking Clomid right now or using other infertility therapies, please read this and take heart. You're using these therapies. You know the risks. Presumably, you accept them. Try not to add to your worry.

1. This study was based on answers to questionnaires. The answers, given retrospectively, were not confirmed with clinical data. They're just what the respondents recall or state for the record. These types of data are notoriously unreliable.

2. The absence of clinical information--and apparently related questions on the questionnaires--means that the researchers also had no data on the many potential confounding variables that could form a link to autism. That indeed have been shown to be linked to it: preterm birth, multiple birth (maybe), and low birth weight.

3. These data are from a paper presented at a conference. They are not published. That means they have yet to be peer reviewed. For public consumption, that also ought to mean that they don't exist yet, or, at the very least, the caveats need to be given earlier than the sixth graf. Like maybe in the lede graf?

4. I don't see any mention of inclusion of other potential effective variables possibly associated with infertility treatment in the data, including the age of the father, which has been linked to autism, as well.

In other words, a large number of variables known to be associated with infertility and infertility treatments were not considered in this analysis. All it shows is that there was an increased rate of autism among women who took fertility drugs. It does NOT demonstrate that fertility drugs are the cause of that increase. The researchers themselves are aware of this, and it is not their fault that the data they've justifiably presented at a conference have been transmogrified into a howl of alarm for public consumption. In the Time.com article, the commentary from the lead epidemiologist on the study clarifying the situation occurred about 10 paragraphs down:
Epidemiologist Kristen Lyall, who led the Harvard study, cautions that even if further research should confirm a link between infertility drugs and autism, any additional treatment-related risk appears to be small: among women whose average age was 35 when they had their first child, there was a 4% risk of having a child with autism for those who had taken fertility drugs, compared with 2% for those with no drug exposure. The increase in risk was even smaller among a younger subset of women.
Note that the correlation they identified was relatively decreased in younger women. That points, in my mind, to factors other than the infertility drugs themselves being involved here.

At the same conference a team from Israel reported a correlation between in vitro fertilization (IVF) and autism. Again, what they're talking about here is the finding that with the presence of one, the presence of the other increased. These data are described in one news media report as "even more preliminary" than the Harvard data. This report describes low birth weight and IVF as being more common among the histories of children with autism vs. their non-autistic peers.

Correlation is always problematic and inferences from it can be quite iffy. The rate of West Nile virus infection and the rate of ice cream consumption both increase during the hotter months. That doesn't mean that ice cream consumption causes West Nile. But the two phenomena do share other factors in common, including warming temperatures that drive both. IVF and the use of infertility drugs also have factors in common, including higher rates of multiple births, early births, and low birth weight. In the absence of consideration of these strongly relevant factors, the associations identified here between infertility therapies and autism are as relevant as that between ice cream and West Nile.

Correlation, as many of us often hear, does not mean causation.

Back to the news media. I wonder how many women they just scared shitless, drove into a guilt-induced slump with those headlines? Women who right now are using these therapies probably felt ill on reading those headlines, and not from the Clomid for once. Women who used these therapies before conceiving children later diagnosed with autism likely resurrected those "what did I do wrong?" self-questionings that seem to haunt parents of children with developmental differences.

What would have been better, more accurate? How about "Hint at link between autism and fertility treatments, but cause is unknown" or "Unpublished findings hint at autism-infertility treatment link, cause not known"? I know. Those aren't nearly as clickable as "Hey! That fertility drug you took before you had that autistic kid may have done that to him!"

Come on, news media. Remember the Wakefield paper! Viva la verite!