Abortion and Mortality Rates – How not to do science.

Over on the other side of the big pond, the Sacramento Bee is carrying the following report of a newly published ‘study’ which looks at the ‘relationship’ between female mortality and first pregnancy outcome.

SPRINGFIELD, Ill., Sept. 5, 2012 — /PRNewswire/ — A new study of the medical records for nearly half a million women in Denmark reveals significantly higher maternal death rates following abortion compared to delivery. This finding has confirmed similar population studies conducted in Finland and the United States, but contradicts the widely held belief that abortion is safer than childbirth.

By linking records from Denmark’s fertility and abortion registries to death registry records, the researchers examined death rates following the first pregnancy outcome of all women of reproductive age in Denmark over a 30-year period, charting death rates at 180 days, 1 year, and in each of 10 subsequent years following each woman’s first pregnancy outcome. Significantly, higher rates of death were observed among women who aborted in every time period examined.

Overall, the study found that women who had first-trimester abortions had an 89 percent higher risk of death within the first year and an 80 percent higher risk of death over the full study period.

Scary, huh?

Well, no, not really – and for starters we should note that the article carries a very important and noteworthy disclaimer:

This section contains unedited press releases distributed by PR Newswire. These releases reflect the views of the issuing entity and are not reviewed or edited by the Sacramento Bee staff. More information on PR Newswire can be found on their web site. You can contact the service with questions or concerns here.

So the whole article is just a press release which the Bee has reprinted in full without any editing or checking – and credit where it’s due, the paper has been honest enough to make that clear to its readers, which makes a refreshing change from the usual ‘Daily Mail Reporter’ approach to regurgitating PR copy.

However, that aside what we have here is not evidence of significantly higher mortality rates associated with abortion, as the press release claims, but rather one of the worst, if not the worst, epidemiological study I’ve ever seen in my life, a study that is not just flawed but so poorly conceived and executed as to suggest that it is willfully shoddy in its construction, none of which should come as any surprise once one takes a look at its provenance.

According to Dr. David Reardon, an author of the new Danish record linkage study…

Yes, its our old ‘friend’ David C Reardon who, somewhat curiously, is still presenting himself as a ‘Dr.’, i.e. PhD, despite being outed as having obtained his ‘qualification’ from what was then an unaccredited institution (Pacific Western University, California) as long ago as 2004.

Reardon first emerged on the intellectual scene in 1987 with a book titled Aborted Women, Silent No More, a review of the “evidence” on abortion’s after-effects that included testimonies from women who had undergone post-abortion religious conversions. The next year, Reardon founded his own quasi-academic think tank, the Elliot Institute for Social Sciences Research. At the time, Reardon had a background in electronic engineering; he’s since acquired a PhD. in biomedical ethics from Pacific Western University, an unaccredited correspondence school offering no classroom instruction.

Reardon’s co-author on this paper is, inevitably, Priscilla Coleman of Bowling Green State University who, only this year, was royally caught out by researchers from UCSF and the Guttmacher Institute.

A study purporting to show a causal link between abortion and subsequent mental health problems has fundamental analytical errors that render its conclusions invalid, according to researchers at the University of California, San Francisco (UCSF) and the Guttmacher Institute. This conclusion has been confirmed by the editor of the journal in which the study appeared. Most egregiously, the study, by Priscilla Coleman and colleagues, did not distinguish between mental health outcomes that occurred before abortions and those that occurred afterward, but still claimed to show a causal link between abortion and mental disorders…

“This is not a scholarly difference of opinion; their facts were flatly wrong. This was an abuse of the scientific process to reach conclusions that are not supported by the data,” says Julia Steinberg, an assistant professor in UCSF’s Department of Psychiatry. “The shifting explanations and misleading statements that they offered over the past two years served to mask their serious methodological errors.”

The errors are especially problematic because Coleman later cited her own study in a meta-analysis of studies looking at abortion and mental health. The meta-analysis, which was populated primarily by Coleman’s own work, has been sharply criticized by the scientific community for not evaluating the quality of the included studies and for violating well-established guidelines for conducting such analyses.

While we’re on the subject of provenance, it would be churlish not to take a quick look at the journal, Medical Science Monitor, in which this ‘study’ has been published and a couple of things jump out, straight away, as interesting contextual issues. One is the journal’s publishing model, which it describes as follows:

Publishing model and authors fees. The submission and peer-review of manuscripts are free of charge. Authors are requested to pay US$ 1100 per article. This fee is requested after positive evaluation of a manuscript. Checks, bank wire and credit card online payments are accepted.

Checks and bank transfers should be directed to International Scientific Literature, Inc. 361 Forest Lane., Smithtown, NY 11787, U.S.A.

Although the journal does make full papers available free of charge on a seven-day ‘personal use’ license, it does charge $199 per article for educational use and a whopping $399 per article for commercial use.

While the journal doesn’t ask authors for payment up front, prior to peer review, the publishing model does raise questions about the standard of pre-publication peer review at the journal, given that the journal doesn’t get paid if it doesn’t publish, which could easily make it rather less selective in its editorial policy than other, far more prestigious journals.

Interestingly, Medical Science Monitor is also one of three journals that were suspended from the Thomson-Reuters impact factor rankings for 2011 due to their involvement in a ‘citation cartel’ which aimed to artificially boost the impact factor of one of the three journals, Cell Transplantation.

Almost all of those banned are excluded because of excessive self-citation, although three journals — Cell Transplantation, Medical Science Monitor and The Scientific World Journal — apparently worked together to cite each other and thus raise impact factors.  That “cartel” was originally reported by Phil Davis on The Scholarly Kitchen, and he has today posted a follow-up article on that ban. McVeigh says that this incident, which she calls “an anomaly in citation stacking”, is the only one of its kind that she has found.

Although the third journal in this alleged ‘cartel’ – The Scientific World Journal  – responded to this article by pointing out that a number of editors from Cell Transplantation had worked together to abuse their positions on other journals to boost its impact factor and that the offending papers had been retracted by the journal for violating its policy on citation manipulation the only response I can find from Medical Science Monitor is this,, from an article in the Wall Street Journal:

Mark Graczynski, executive publisher of Medical Science Monitor, says there was no effort between the journals to manipulate the IF. “It might just be coincidence that there’s an overlap of some editors,” he says.

However, the same article also notes that:

In April, Phil Davis, a publishing consultant who writes for a blog called The Scholarly Kitchen, noticed unusual citation patterns at Cell Transplantation.

In the blog, Mr. Davis noted that a review article published in another journal, Medical Science Monitor, had cited a total of 490 articles in the field, of which 445 were articles that had appeared in Cell Transplantation alone, in 2008 and 2009. Both those years were used to compute the 2010 impact factor for Cell Transplantation, and those citations apparently had an effect: the journal’s IF rose from 5.126 in 2009 to 6.204 in 2010, a jump of 21%.

Mr. Davis notes three of the four editors of the Medical Science Monitor review article were also on Cell Transplantation’s editorial board. Also, two MSM editors wrote a review in another journal, The Scientific World Journal, citing 124 papers. Of those, 96 were from Cell Transplantation in 2008 and 2009.

That’s apparent;y nothing more than a coincidence, huh?

Clearly, taking all this into account, our expectations for this latest paper should be extremely low but, in truth, when one looks at the actual paper one quickly finds that it fails to reach even that extremely modest standard.

So what of the actual ‘study’? Well, this is described in it’s abstract in the following terms:

Medical records for the entire population of women born in Denmark between 1962 and 1991 and were alive in 1980, were linked to death certificates. Mortality rates associated with first pregnancy outcomes (delivery, miscarriage, abortion, and late abortion) were calculated. Odds ratios examining death rates based on reproductive outcomes, adjusted for age at first pregnancy and year of women’s births, were also calculated.

So its a record linkage study and it produced the following results:

A total of 463,473 women had their first pregnancy between 1980 and 2004, of whom 2,238 died. In nearly all time periods examined, mortality rates associated with miscarriage or abortion of a first pregnancy were higher than those associated with birth. Compared to women who delivered, the age and birth year adjusted cumulative risk of death for women who had a first trimester abortion was significantly higher in all periods examined, from 180 days (OR=1.84; 1.11 <95% CI <3.71) through 10 years (1.39; 1.22 <95% CI <1.61), as was the risk for women who had abortions after 12 weeks from one year (OR=4.31; 2.18 <95% CI <8.54) through 10 years (OR=2.41; 1.56 <95% CI <2.41). For women who miscarried, the risk was significantly higher for cumulative deaths through 4 years (OR=1.75; 1.34 <95% CI <2.27) and at 10 years (OR=1.48; 1.18 <95% CI <1.85).

So, out of 463,473 women who had their first pregnancy between 1980 and 2004, 2,238 (o.4%) died during the same period – died of what, exactly?

The study doesn’t say because what it’s comparing is the outcome of women’s first pregnancies and mortality from all causes, which could easily include everything from homicides and road traffic accidents, to suicides and drug overdoses to deaths resulting from disease, congenital conditions and just sheer bad luck. Many, if not of the deaths included in these results will, quite obviously, be entirely unrelated to pregnancy, let alone to the outcome of a particular pregnancy.

Nevertheless, the data does show that women who terminated their first pregnancy did exhibit a higher mortality rate than women who carried their first pregnancy to term, as did women who miscarried, although not to the same degree as the abortion group, so this does require an explanation.

Reardon and Coleman offer up three possibilities:

There are at least three theories which may explain the differences in mortality rates observed. The first theory may be called the ”healthy pregnant woman effect”. This theory suggests that healthier women are more likely to be able to conceive and carry a pregnancy to term. Conversely, women who are unhealthy may be unable to conceive or may be more likely to have a spontaneous or therapeutic abortion.

This may well be applicable to women who conceive and either carry to term or spontaneously abort but its difficult to see how this will be relevant to induced abortion, where very few abortions are carried out on purely physical health ground.

The second theory is that pregnancy, especially one carried to term, produces health benefits which reduce the risk of death. For example, carrying a pregnancy to term is associated with physiological changes associated with a reduced risk of breast, ovarian, and endrometrial cancers. Live births may also contribute to psychological benefits, or at least behavioral changes and lifestyle choices associated with being a parent which improve health and/or reduce unhealthy or risk-taking behaviors.

As far as the cancer element is concerned, this is extremely implausible. Amongst the women included in the study, the average age of first pregnancy was 24-25 in the carried to term and miscarriage groups and only 20 or so in the two abortion groups ( early – i.e. pre 12 weeks gestation – and late) so the paper’s maximum follow-up period of ten years is too short a period for these women to accrue any of the protective effects associated with carrying a first pregnancy to term in relation to breast cancer. The oldest women included in the study were born in 1962, making them only 42 years of age by end of the period for data was available and at that age a woman’s annual risk of dying of breast cancer is only around 1 in 1000 so, at most, we’d expect this to account for only 2-3 of the 2,238 deaths included in the study.

So far as breast cancer and the ABC hypothesis is concerned, and deaths attributable to breast cancer in this study are likely to be found amongst women with a family history of breast cancer and, therefore, an elevated genetic risk of developing cancer, a risk that pregnancy increases rather decreases.

The changes is lifestyle/behaviour is a more promising and plausible hypothesis but this, of courser, offers only an indirect relationship between abortion and mortality risk – the abortion is not the cause of death, the risky behaviour/lifestyle is and, as social services departments will see every day, carrying a pregnancy to term is no guarantee of a lifestyle change.

The third theory is that pregnancy loss may contribute to physiological or psychological effects which increase risk of death. For example, abortion is associated with an increased risk of suicide, substance abuse, post-traumatic stress disorder, and a lower assessment of general health. In addition, it is notable that the elevated rates of mortality associated with pregnancy loss observed in this study are on the same order of magnitude as the elevated rates of mortality among women who experience the death of a child under 18 years of age. Factors common to both experiences may explain the effect observed in both groups.

Of the three studies cited in this passage as providing supporting evidence for this ‘theory’, two — Gissler (1997) and Reardon et al. (2004) were excluded from last year’s NCCMH systematic review of the evidence on induced abortion and mental health due to their use of inappropriate controls for prior mental health while the third – Ney 1994 – was excluded from the review for using inappropriate mental health measures. In regard to suicide, etc. the NCCMH review concluded that:

Studies that did not control for whether or not the pregnancy was planned or wanted, suggest that there are increased risks of receiving psychiatric treatment, suicide and substance misuse for women who have abortions compared with those who deliver a live birth. Findings for depression, anxiety disorders, suicidal ideation and PTSD did not indicate an increased risk.

In contrast, where studies controlled for whether or not the pregnancy was planned or wanted, there was insufficient evidence to determine whether or not there was an elevated risk of mental health problems, except for a small increase in possible self-harm in those having an abortion compared with the women who delivered an unplanned, but not unwanted pregnancy, and some evidence of lower rates of psychotic illness for women who had an abortion compared with those who delivered the pregnancy at full term.

There is, however, a fourth possibility here which, naturally enough, Reardon and Coleman fail to mention in their discussion of the paper’s results, one suggested by this observation from the NCCMH review:

Adequate control of confounding factors was shown to have an impact on the results, with previously significant findings no longer being significant when a range of confounding factors were accounted for. In essence, where studies controlled for multiple confounding factors (including the wantedness of the pregnancy), the risk of mental health problems following an abortion was comparable to the risk of mental health problems following a delivery. Consistent with this view, findings from both the APA and Charles reviews indicated that where studies were of better quality, controlling for previous mental health problems and accounting for other confounding factors, the risk of mental health problems was no greater following an abortion compared with a delivery.

So, in regards to mental health, better quality studies that included adequate controls for prior mental health and other potential sources of confounding, i.e. age, ethnicity, social class, etc. failed to produce statistically significant evidence of an association between induced abortion and subsequent mental health problem and, as sure as eggs is eggs, when we look at the limitations section of Reardon and Coleman’s paper we find the following admission:

Another limitation is that our analysis does not control for socioeconomic factors, marital status, psychological history, or other factors prior to first pregnancy which may affect the subsequent risk of death.

To put that into perspective, Reardon and Coleman’s paper using almost the exact same methodology as Gissler’s record linkage studies, albeit with the different national data set – Gissler took his data from Finland, Reardon and Coleman from Denmark – and the NCCMH review gave this assessment of the quality of Gissler’s study, which it rated as ‘very poor’:

Using the modified Charles review quality criteria, GISSLER1996 and GISSLER2005 were rated as very poor due to the lack of any control for previous mental health problems, a factor associated with higher suicide rates. Furthermore, the study failed to account for confounding factors such as how much the pregnancy was wanted, multiple pregnancy events, type of abortion (elective or medical) or any socioeconomic variables, which may be associated with both abortion and increased suicide risks.

Our fourth theory, or rather hypothesis, is therefore simply that the observed differences between mortality rates in the four groups in this study, carried to term, spontaneous abortion and early and late induced abortion, will cease to be statistically significant once the data-set used in this paper is reanalysed with the inclusion of appropriate controls for known sources of confounding.

In short, Reardon and Coleman’s paper provides no useful or usable evidence whatsoever and certainly does not contradict any ‘beliefs’ about the safety of abortion relative to childbirth, as the authors claim in their press release.

The only this paper has to teach us is that poor research methodologies consistently produce poor quality results, and when you find yourself confronted by researchers who consistently employ poor research methodologies, as is certainly the case when it comes to David Reardon and Priscilla Coleman, then one has to consider carefully whether what you’re dealing with is either rank incompetence or agenda-driven bias – and if you read any of my previous articles on Reardon and Coleman, there’ll be no prizes for guessing which of these two possibilities I favour, not least in light of the paper’s unintentionally hilarious closing statement:

The authors have no financial disclosures to declare and no conflicts of interests to report.

Really?

That was certainly not the view of Ben Goldacre and William Lee when the responding to a 2009 paper by Priscilla Coleman, which was published in the British Journal of Psychiatry, noting that the BJP subscribes to International Committee of Medical Journal Editors uniform requirements for declaration of conflict of interest, which requires authors to make a declaration of “any relevant non-financial associations or interests (personal, professional, political, institutional, religious, or other) that a reasonable reader would want to know about in relation to the submitted work”.

By those standards, Reardon and Coleman’s active involvement in anti-abortion activism should be considered to be a declarable interest and yet both have routinely claimed to have conflicting interests when submitting abortion-related ‘research’ papers for publication.

Whether this is any more (or less) dishonest that Reardon’s claim to have a PhD in Bioethics is something I’ll let you decide for yourself.

  • Pingback: Unity()

  • Pingback: Abortion Rights()

  • Pingback: Richard Parry()

  • Pingback: RHM Journal()

  • Pingback: Beth Sutton()

  • Martin

    “Medical records for the entire population of women born in Denmark between 1962 and 1991 and were alive in 1980” – is this your typo or is this really the most bizarre thing ever?  If you correct this, feel free to remove the comment

  • Pingback: Marika Gooch()

  • unity_ministry

    It’s a direct quote from the paper’s abstract, and I might also note that when quoting from the full paper I had to correct several basic spelling mistakes as well, so it would appear that Medical Science Monitor needs to hire a decent sub-editor as well as improve the standard of its pre-publication peer review process.

  • Pingback: Fiona Bloomer()

  • Jimmyhsands

    If you’ll forgive an off-topic observation, the reshuffle has led one blogger to revisit your archive.  I thought you might be interested.

     http://tcipost.com/?p=23561