The Devil’s in the Anti-Abortion Footnotes (part 2)

As you may remember, during arguments before the Supreme Court last week, Texas Solicitor General Scott Keller stated that an amicus brief filed by Former Abortion Providers and the National Association of Catholic Nurses contested the settled fact that going through childbirth is more dangerous for a woman than having an abortion. Curious, I looked this amicus brief up, and yesterday I wrote my first post examining this brief. Today I’ll look at some of the studies in the brief’s footnotes, and tomorrow I’ll be looking at some new directions anti-abortion activists have taken in their efforts to exaggerate the purported risks abortion poses to women.

Let’s start with this section here:

That is some claim right there! Let’s check out the footnotes, shall we?

So I went to the link. It’s an article on the website lifeissues.net, not a scholarly study. It does claim what the amicus brief says it does, though it only offers that brief statement, supported by more footnotes. It is extremely shoddy work to quote an internet article in this way rather than going directly to the cited studies for the source, but I suppose the brief only claimed to be quoting “medical researchers” analysis, and not the scientific literature itself. Either way, here is the statement in the article:

Mifepristone abortion has 10 times more risk of death from infection than surgical abortion46,47,48 and 50 times more risk of death from infection compared to childbirth.49,50

Note that there is no analysis whatsoever. I think that’s what’s throwing me here. Usually, medical researchers analyzing scientific literature actually, you know, analyze. And while we’re at it, let’s take a moment to note that the internet article the amicus brief is citing for these claims is itself nine years old, and that all of the relevant citations are at least ten years old.

Regardless, here are the footnotes in question:

46 Harrison, op. cit. [Gary MM, Harrison DJ. Analysis of severe adverse events related to the use of mifepristone as an abortifacient. Annals of Pharmacotherapy 2006;40(2):191-7.]

47 McGregor J, FDA: Emerging Clostridial Disease Workshop, Transcript, May 11, 2006, Atlanta, GA. Available from: http://www.fda.gov/cder/meeting/clostridial/meeting_transcrript.pdf (accessed March 24, 2007). [working link]

48 Greene MF. Fatal infections associated with mife-ristone-induced abortion. N Engl J Med 2005:353(22):2317-8 [working link]

49 Harrison, op.cit.

50 Fischer M. Fatal toxic shock syndrome associated with Clostridium sordellii after medical abortion. N Engl J Med 2005:353:2352-60 [working link]

Remember that there are two claims at play—first, that medical abortions present ten times the risk of death from infection that surgical abortions do, and second, that medical abortions present fifty times the risk of death from infection that childbirth does. What do the four sources cited actually say?

First, the Gary and Harrison study, published in 2006, which is cited for both claims. Unfortunately, I couldn’t find a full-text version online or through my library, but I did find a summary. According to the description, this study examined “[s]ix hundred seven unique mifepristone AERs submitted to the FDA over a 4 year span.” In other words, two OBGYNs examined a database of 607 adverse reactions to mifepristone that were submitted to the FDA over the period of four years. What did they find? They found that:

The most frequent AERs were hemorrhage (n = 237) and infection (66). Hemorrhages included 1 fatal, 42 life threatening, and 168 serious cases; 68 required transfusions. Infections included 7 cases of septic shock (3 fatal, 4 life threatening) and 43 cases requiring parenteral antibiotics. Surgical interventions were required in 513 cases (235 emergent, 278 nonemergent). Emergent cases included 17 ectopic pregnancies (11 ruptured). Second trimester viability was documented in 22 cases (9 lost to follow-up, 13 documented fetal outcome). Of the 13 documented cases, 9 were terminated without comment on fetal morphology, 1 was enrolled in fetal registry, and 3 fetuses were diagnosed with serious malformations, suggesting a malformation rate of 23%.

And that’s all the summary gives us. What is clear from the above, though, is that the number of these adverse reactions that resulted in fatalities was low—the summary only mentions four. And this number would be low—remember that this study took place in 2006, and that between 2000 and 2011 there were a total of 14 deaths related mifepristone. (A paper published in 2005 stated that the FDA was aware of a total of 4 mifepristone deaths, which is in keeping with the four fatalities mentioned here.) It also appears that the author of this study analyzed adverse reactions to mifepristone, specifically, without analyzing other data for comparison (at the very least, any such analysis did not merit mention in the study’s summary of results).

How significant was the number of adverse reactions studied? By 2011, 1.52 million medical abortions had been performed. However, the rate was not equal across that period—in 2011, only 6% of abortions were medical abortions, whereas today nearly one quarter are. According to another study we’ll look at in a minute, there were 460,000 medical abortions performed between 2000 and the end of 2005. Using that number (which is high, given that Gary and Harrison were working with four years worth of adverse reactions, and this number represents five years of medical abortions) with a total of 607 reported complications, we find a 0.13% complication rate. The death rate, of course, was far, far lower than that, given how few of the reported complications resulted in deaths—only four. Remember that the maternal mortality rate ranged from 12.0 to 16.8 deaths per 100,000 live births during this same time, far, far higher than the proportional number of deaths associated with medical abortions during these years.

And yet, the amicus brief uses Gary and Harrison’s study as a citation for both its claim that medical abortions carry ten times the risk of death to infection as surgical abortions and its claim that medical abortions carry fifty times the risk of death to infection as childbirth. But perhaps this study is only supporting information, and the statistics themselves are included in the other citations? Let’s find out!

The only other citation for the claim that medical abortions create fifty times more risk of death due to infection than childbirth is Fischer’s 2005 study of four cases of fatal toxic shock following medical abortions. Remember that at this point, only four such deaths had been reported, so Fischer was working with a complete data set. Fischer states that:

These cases demonstrate that serious infection can occur after medically induced abortion, much as it can occur after childbirth, spontaneous abortion, and surgical abortion. However, available data suggest that the risk of such infection is low.

And that’s it. Fischer finishes by calling on doctors and medical providers to be aware of the symptoms of the specific type of toxic shock he focuses on in his analysis. There is nothing whatsoever in the article to suggest that the risk of death due to infection is higher following a medical abortion than it is following childbirth. And yes, I had access to the full text.

What about the claim that medical abortions come with a ten times higher risk of death due to infection than do surgical abortions? This claim has three citations, the first being the aforementioned Gary and Harrison article that studied reported complications due to medical abortions and did not make comparisons worth reporting in its summary results section (or mentioning in its summary methodology section). The second citation is for a talk given by James McGregor of the University of Southern California during a workshop on Clostridium sordellii. I read through the entire thing and the only thing at all related to claim this is used as a citation for is McGregor’s statement that surgical termination “appears to have a lower risk in terms of patient mortality” compared to the abortion drug. McGregor does not use comparative numbers in his talk, and his concern appears to be vaginal insertion of misoprostol, which is no longer how the drug is taken. It’s also worth noting that at least one participant in the workshop disagreed and argued against any connection between mifepristone, misoprostol, and clostridium sordellii.

The third and final citation, the study by Green, reads as follows:

These figures would suggest that the risk of death from infection [as a result of mifepristone] is less than 1 per 100,000. In the United States, the risk of death from any cause associated with attempting to carry a pregnancy to term is 8 to 10 times that.

The more appropriate comparison, however, is with the risk associated with other methods of inducing abortion. The overall maternal mortality rate associated with induced abortion in the United States is approximately 1 per 100,000. That overall rate is a “blended” rate including all the procedures performed in the United States at all gestational ages. The gestational-age–specific rate increases exponentially from 0.1 per 100,000 at 8 weeks’ gestation to 8.9 per 100,000 at 21 or more weeks’ gestation. Mifepristone is approved for the termination of pregnancies at less than seven weeks’ gestation. Therefore, the appropriate comparison is with a risk of 0.1 per 100,000 for surgical abortions performed at less than eight weeks’ gestation.

As tragic as the deaths of these young, healthy women are, they remain a small number of rare events without a clear pathophysiologic link to the method of termination. Patients should be informed of this risk before they consent to the procedure and should be vigilant for symptoms after the procedure. Providers must be aware of this potential complication and not be reassured by the absence of fever. Regulators should keep this rare complication in perspective and not overreact to scant data by prematurely foreclosing the only approved medical option for pregnancy termination.

Green is basing his analysis on the fact that the manufacturer reported that a total of 460,000 women had taken the abortion pill and the fact that the FDA was aware of four mifepristone-related deaths, which resulted in just under one death per 100,000 medical abortions. This number has held fairly constant—in 2011, the FDA reported that there had been 1,520,000 medical abortions and 14 misoprostone-related fatalities. Green’s point is that while this is the same as the fatality rate for abortions overall, it is greater than the fatality rate for other early first trimester abortions. But note that Green emphasizes that even this risk is still very, very low, and that the relatively higher risk should not lead governments to crack down on medical abortions. He uses terms like “rare complication” and “scant data” and urges regulators not to “overreact.”

Let’s talk about these numbers for a moment. As of when Green was writing, there had been 4 deaths from infection resulting from 460,000 medical abortions. By 2011 there were 14 deaths associated with 1,520,000 medical abortions, which sounds like a fairly constant rate, but only nine of those overall deaths were associated with infection. In other words, in the six years from 2005 to 2011, there were five additional deaths do to infection even as there were over twice as many medical abortions performed as in the five years from 2000 to 2005. This means that the rate of death due to infection went down by half after Green penned his article. Why? Well for one thing, Planned Parenthood changed how it administered the drug, switching from vaginal to oral usage, thus satisfying McGregor’s concern above. This is how it should be—medical fatalities should prompt us to ask what measures might have prevented those fatalities. And yet the amicus brief deceptively cites these old numbers from 2005 and 2006, when the rate of death due to infection associated with medical abortions was at its highest, and completely ignores any findings (or falling rates of death due to infection) from the last ten years.

By the way, note what Green does in his piece—he urges providers to inform patients of the risks and to make sure they’re knowledgeable enough to recognize signs of a problem and adequately advise their patients. In other words, he appears to genuinely care about promoting the health and needs of women who may seek medical abortions, and about finding ways to ensure that these procedures are safe. The authors of this amicus brief, not so much. Note, after all, that Green states that the risk of death due to medical abortion is a tenth of the risk of death associated with carrying a pregnancy to term, and yet the authors of this amicus brief use his study as a citation for a sentence that claims that the risk of death due to infection is 50 times greater in a medical abortion than in childbirth. If they actually cared about women’s health, they would note that this number is out of step with Green’s number by a factor of 500 and ask what this is so. Instead, what they’ve done is nothing short of willfully deceptive. And you know what? Good medical science is based on good data, not numbers cobbled together or pulled out of thin air to support an ideological cause.

I was actually not able to find a source for the assertion that the risk of death from infection due to a medication abortion is 50 times higher than the risk of death due to childbirth in any of the footnotes for that sentence. It’s true that I didn’t have the full text of Gary and Harrison’s study, but they were looking at 607 adverse reactions to mifepristone over the span of four years, only a small handful of which involved fatalities, and had no way to make comparisons between this data and data on adverse reactions to childbirth and could not have asserted any such rate based on the data they were examining. The closest I came to verifying anything in that sentence was Green’s statement that the rate of death associated with medical abortion might be ten times higher than the risk of death associated with an early surgical abortion, but again, he pointed out in the same section that the rate of death associated with childbirth is still ten times higher than the rate of death associated with a medical abortion, and the rate he posited plummeted in the years following the 2005 publication of his piece, which is not noted in the amicus brief.

It’s actually not scientifically possible for the amicus brief’s claim that medical abortions have 50 times the risk of death due to infection as childbirth to be true. Let’s imagine for a moment that the total death rate associated with a medical abortions is one death in 100,000 medical abortions, even though we know that this rate fell after 2005. The rate of death associated with childbirth is still at least ten times that, and potentially more than that based on other statistics (Greene’s numbers are low). A study we looked at in the last post in this series found that roughly 10% of all pregnancy-related deaths associated with live birth or stillbirth were due to infection, while the other 90% were due to other causes. Even if all deaths related to medical abortions were due to infection and Green’s posited rate in 2005 held constant—remember that this has not been the case—a medical abortion would carry the same risk of death from infection as carrying a pregnancy to term, not 50 times greater risk. And this does not, of course, count the non-infection other 90% of the risk of death related to carrying a pregnancy to term.

But you know what? The authors of this amicus brief must be counting on people not reading the footnotes. They must be counting on people hearing medical terms and seeing studies with medical-sounding names and assuming it all actually means what they say it does, even though it doesn’t. Personally, if I were the author of one of the studies cited in this amicus brief and presented as though they say things they don’t, I would be upset. I’m not sure that there’s any legal recourse—there probably isn’t—but what’s going on here is way than morally dubious. Still, in the next post in this series, which deals with even more expansive claims, we’ll find that even this level of moral dubiousness can be surpassed.

Note: Within 24 hours of posting this article, no fewer than three readers sent me the Gary and Harrison study, the only article in the footnotes examined here that I was unable to find full text for. I’ve written a follow-up post looking at this article, but the summary version is that the study does not make any effort to calculate a mortality rate for medical abortions or any attempt to compare this rate with the mortality rates for either surgical abortions or carrying a pregnancy to term. This means that none of the studies included as footnotes for the claim that medical abortions carry 50 times the rate of infection as carrying a pregnancy mention any such thing even peripherally.