Thursday, April 2, 2015

Bizarre Things Purported to Cause Autism: Hormonal Contraceptives

EXECUTIVE SUMMARY: An article published last winter in Medical Hypotheses suggests that long-term use of hormonal contraceptives might raise the likelihood of one's future children being born autistic. There is no good reason to believe this; the only things the article offers as potential reasons for it are 1) a very tenuous temporal relationship between widespread contraceptive use and a later (much later) rise in autism prevalence; 2) a misinterpreted bit of information largely pertaining to fertility testing; and 3) "What If?" scenarios involving epigenetics. I am worried about potential political fallout, to the further detriment of children and teenagers' sex education in the US, should this idea be uncritically popularized.

Of all the myriad forms anti-feminist backlash can take, anti-contraceptive fearmongering is probably the most irritating to me.

So when I stumbled across this appalling article in Medical Hypotheses while I was casting about for items to populate my list of proposed causes of autism, I knew I would have to write a post specifically addressing this one.

The author, Kim Strifert, says that it may be possible that taking combined oral contraceptives (the ones that include both an estrogen, typically ethinyl estradiol, and a progestin, like levonorgestrel, norethindrone, or any of these others) over a long period* of time can change the conditions inside the ova and perhaps trigger epigenetic modification of genomic DNA.

I'd love to go into more detail about how this is supposed to work, but I can't, because no such detail is provided in the article:
Given the lack of research on the effects of oral contraceptive use on progeny, it is impossible at this point to specify a mechanistic link between oral contraceptives and autism. This fact represents a meaningful limitation to the hypothesis presented in this review. However, recent research already suggests that the current understanding of the pharmacology of oral contraceptives may be over-simplified. It has been proposed, for example, that epigenetic side-effects of pharmaceuticals may be involved in the etiology of cancer, heart disease, neurological and cognitive disorders, obesity, infertility, and sexual dysfunction [7][PDF]. It has also been suggested that epigenetic assays be incorporated into the safety assessment of all pharmaceutical drugs, which might lead to new mechanistic insights in the future [7]. Finally, new evidence is emerging that oral contraceptive use directly and deleteriously affects both the ovaries and the ova [8]. Thus, we are at a point where some concrete mechanistic hypotheses may be achievable in the near future.
What she does offer in support of her hypothesis is summarized here:
  • Temporal correlation between use of oral contraceptives and increased prevalence of ASD.
  • To date no definitive cause or contributing factors for increase in ASD prevalence has been established.
  • Oral contraceptives disrupt the endocrine system -- COC's are endocrine disruptors.
  • Oral contraceptives directly and deleteriously affect both the ovaries and ova.
  • Likely effects of oral contraceptives on progeny are an open question.
  • The called-for [by Dr. Roy Hertz**] further study and controlled follow-up of the possible transgenerational effects of oral contraceptive use has not been executed.
I will address these one at a time.

The first bullet point, the temporal correlation between use of oral contraceptives and autism prevalence, is true enough, but there are temporal correlations between lots of unrelated things. There's a whole mess of 'em archived for educational and comedic effect at Tyler Vigen's website Spurious Correlations. A cursory look at that website should be all the explanation you need of why this is not very good evidence that the two variables actually have anything to do with one another.

And even if we ignore that aspect of it, the time frames of the two trends -- increasing oral contraceptive use over time and increasing prevalence of autism over time -- don't quite overlap the way we would expect them to if there really were any sort of cause-and-effect relationship between them.

If there were such a relationship, we would expect a graph showing autism prevalence rates to be sort of a muted echo of a graph showing oral contraceptive use -- lower numbers overall, and trailing by maybe 5-10 years, but roughly the same shape, with peaks corresponding to peaks in the contraceptive-use graph.

Instead of that, we see two graphs of very different shapes, with their peaks separated by a much wider margin than would be possible if Strifert's hypothesis were true.

I cannot find comprehensive data on how many women used The Pill for every year between 1960, when it was approved, and now, but I can cobble together enough information from several different places to sketch a rough outline.
A very, very loose sketch of oral contraceptive usage in the US between 1960 and 2010, stitched together from numbers provided by PBS, the CDC (PDF, PDF), the Kaiser Family Foundation and the Guttmacher Institute
I can't vouch for the strictest accuracy of this picture, but the overall shape -- up, up, up, then slowing down, then a dip, then holding steady at about 10 million -- I'm reasonably confident in. 

By way of contrast, here is a graph of autism prevalence rates from 1985 to 2012 that Emily Willingham made:
Autism prevalence, 1985-2012, by Emily Willingham

As you can see, the shapes are very, very, very different! The graph of oral contraceptive use starts shooting upwards immediately, and then slows down its rate of increase, and then hits its peak and flattens out, while the graph of autism prevalence rises very slowly before it finally begins to build up steam. Like a logarithmic vs. an exponential*** curve...

They both trend upward, but other than that, they have little in common.

Even more important than the shape is the timing. Oral contraceptive use seems to hit its peak around the mid-to-late-1970s, while autism prevalence has perhaps not even peaked yet. It has a long, slow rise with what appear to be two inflection points, one around 2001 and another at about 2005.

Twenty-five years is an awfully long lag time between getting on the Pill and having young children. My mom could easily have been taking the Pill in the mid-1970s, and she had me thirty years ago! And I was diagnosed in 1989 or 1990, when the graph of autism prevalence was still hugging the x-axis.

The women making up that first big wave of oral contraceptive users are probably more likely to be the grandmothers of the children making up the present Autism Epidemic than they are to be their mothers.

The second bullet point is not really an argument in support of a causal relationship between contraceptive use and autism in progeny -- it's just saying, "Well, we don't know what causes autism, so why not entertain my idea?"

The third bullet point -- that hormonal contraceptives are endocrine disruptors -- is kind of a tautology. Of course they are endocrine disruptors -- they wouldn't be able to suppress ovulation if they were not.

Pointing out that hormonal contraceptives are endocrine disruptors is about as helpful as running up to someone and announcing that the Tylenol they are about to take will interfere with their body's natural warning system by dulling their sensations of pain. They know that; that is why they're taking the pill in the first place!

Finally, let's return to the first passage I quoted, which is the first mention of bullet point #4, "[o]ral contraceptives directly and deleteriously affect both the ovaries and ova." The source Strifert gives for this statement -- the "new evidence" that is "emerging" -- is this article on the biotechnology news website

What that article says is very different from what Strifert seems to think it says.

Prolonged hormonal contraceptive use does indeed shrink the ovaries, lower the levels of anti-Müllerian hormone in the blood, and reduce the number of ovarian follicles at a certain stage of maturation. Those two things are considered reliable indicators of how many egg cells might be left within the ovary.

(Here is a literature review comparing the relative merits of those two biomarkers)

However, the researchers who discovered this -- Dr. Kathrine Birch Petersen and her team -- do not think these effects are permanent, and are mostly concerned with making sure people getting their fertility tested after just coming off of birth control get an accurate estimate of their ovarian reserves. Dr. Birch Petersen is especially worried that this temporary suppression of ovarian-reserve markers by long-term contraceptive use might mask a naturally low ovarian reserve, something a person wishing to become pregnant would want to know about.

Here she is, quoted in the Bioscience Technology article:
Birch Petersen's team does not believe these changes [decreased ovarian volume, lower AMH levels, lower antral follicle count] are permanent. But as a result of the study, she said, women in the Pre-conceptional Care Program who have been on the contraceptive pill are now advised that their ovaries may look older and smaller, and may possess only a few small antral follicles, with low levels of AMH for a time after stopping. They are told this likely does not affect future fertility for most women. 
But it could matter for women undergoing premature menopause. Naturally diminished ovarian reserves could be masked by the above. It is therefore possible ovarian reserve assessment should be repeated after stopping birth control pills. 
"Worldwide, 160 million women are on the pill," Birch Petersen told Bioscience. "One percent will go into early menopause before the age of 40. The pill can mask the symptoms of early menopause, and this is why women should consider repeating the tests after six months [off the pill], if they have a low ovarian reserve." 
Next up, says Birch Petersen: "To examine what happens with the ovarian reserve parameters after one, three, and six months."
And here is a short press release published last year on the Clinical Endocrinology News website:
Oral contraceptives do more than prevent unwanted pregnancy. They also make it harder to gauge a woman's ovarian reserve, based on data from 833 women aged 19-46 years seen at a single Danish fertility clinic. 
Study findings suggested that an accurate measure of a woman's ovarian reserve can occur only after she has been off an estrogen-containing [oral contraceptive], probably for at least 3 months, Dr. Kathrine Birch Petersen reported at the annual meeting of the European Society of Human Reproduction and Embryology. 
The impact of estrogen-containing OC use on reducing ovarian volume was especially pronounced in women under 30, the reduction increased with longer durations of OC use, and the ability of OC's to mask a woman's actual ovarian reserve was strong enough to potentially conceal a true case of premature ovarian insufficiency, said Dr. Birch Petersen, an ob.gyn. at the Fertility Assessment and Counseling Clinic at Righospitalet in Copenhagen. 
"When we see a woman on an OC with impaired ovarian reserve, we would presume [based on these new findings] that her real ovarian reserve was about 30% higher than what we measure. We would advise her to be retested after she was off her OC for about 3 months," Dr. Birch Petersen said during a press conference before her presentation at the meeting. 
The study included the first women seen at the clinic since it opened in 2011, excluding those who were pregnant or failed to supply adequate information. The cross-sectional cohort included 240 women on estrogen-containing OC and 593 women with natural cycles. 
The analysis focused on three parameters: blood level of anti-Müllerian hormone (AMH), antral follicle count (AFC), and ovarian volume. The multivariate, linear regression analysis adjusted for age, body mass index, smoking, age of maternal menopause, maternal smoking during pregnancy, preterm birth, and duration of OC use. 
The analysis showed that compared with the women with natural cycles, those on an OC had a 19% relative reduction in their average blood level of AMH, a 16% reduction in average AFC, and a 47% relative reduction in average ovarian volume. The women on an OC also had smaller antral follicles. All three differences were statistically significant. 
Seeing an effect from an estrogen-containing OC on all three measures makes sense because of their interrelatedness. The antral follicles produce AMH, and a reduction in antral follicle number as well as size would shrink the ovarian contents and result in reduced volume. These results would not occur in women on a progestin-only OC, she said.
This additional context makes it clear to me that the changes wrought on the ovary by estrogen-containing contraceptives are indeed temporary.

And here is one more thing I wonder -- why autism? Why would the kind of epigenetic interference Strifert seems to be postulating only result in one type of developmental disability? I know autism is a broad category, but still -- why would only development of the nervous system be affected? Why not all aspects of fetal development?

(A cynic's response to that question would be, because autism is the biggest cultural bogeyman with which to threaten prospective parents. See also: the anti-vaccine movement.)

I would not be opposed to any of the further research into contraceptive safety, uterine and ovarian physiology, or potential alternative methods for contraception!

Even though I do not think it likely at all that what she suspects is true, I think that only good could come of additional efforts to develop even safer methods of contraception that work in multiple different ways.

But what I do not want to happen, which I strongly suspect would be a more likely near-term consequence of Strifert's article gaining widespread attention, is for her warnings to be hyperbolized (instead of the question "what is the effect of long-term hormonal contraceptive use on the ova?" it would become the statement "using hormonal contraceptives means that your children will be born with developmental disabilities later on") and be taught to teenagers in sex education classes as yet another reason they should eschew any and all birth control methods. Sex education classes in the US already lie to children about the efficacy of condoms and tell them that total sexual abstinence is the only way to protect themselves, so adding one more lie to the list is hardly unthinkable.

And looking at the path Strifert's ideas have taken into the blogosphere, this impression -- that, far from echoing her call for more research and development of new contraceptive methods and for improved safety of existing methods, the blogs reporting on her article merely urge their readers to eschew hormonal methods of contraception entirely.

The only blog post I could find dealing with this specific article is this one, by Dr. Kelly Brogan, MD, who despite her medical training seems to reject all of modern, conventional medicine in favor of diet and lifestyle changes.

There are other posts that express a similar idea -- that birth control pills can cause autism in one's future children -- but credit a different source and propose a different mechanism by which it happens (usually gut bacteria rather than the epigenetic explanation loosely sketched by Strifert); those blogs are about an even mix of anti-vaccine, alternative medicine blogs that advocate rejection of all conventional medicine and Catholic blogs that reject all methods of contraception except Natural Family Planning and abstinence.

I do not think it is irrelevant that those are the types of the blogs on which this particular Bizarre Autism Hypothesis has appeared.

*Pun intended

**In a 1966 report by the U.S. Food and Drug Administration's Advisory Committee on Obstetrics and Gynecology; that report, physically archived at the University of Michigan, has been digitized and can be viewed online here.

***Or, what is more likely, a logistic curve. Populations cannot grow indefinitely.