Abortion Discussion, Part the Fourth

Igotadose wrote: »

Mutter...grumble...

I'm in an FB group, "gaelige amhain,' where we discuss various things about Irish, in Irish. Helpful to me as I'm learning, I have the Irish language ability of a slightly conversant 5 year old, but I can copy/paste into Google Translate with the best of them.

Well guess what. The pro-life phonies have shown up whining about the law changes coming to Northern Ireland. They pop up whenever there's a rally in favor of their pile of lies. My problem is, my Irish isn't good enough to refute them. If yours is, or know someone who might help, here's the group:
https://www.facebook.com/groups/166677873392308/

I would suggest contacting Gaeil ar son Rogha, they were the pro-choice group for Irish speakers during the referendum. They would have plenty of people who are well able to refute their arguments in Irish.

Just to expand briefly on uptherebels point about the giannacare citation for the benefit of the readers here, the only positive evidence that giannacare provide in support of the services they provide is a reference to a 2011 paper published in the British Journal of Psychiatry by Priscilla Coleman. However, this study is about as textbook an example of flawed and unethical science as its possible to get.


And here's why.


A brief introduction to the paper. This paper was published in 2011 and purports to be a systematic review and meta-analysis of the relationship between abortion and mental health. It cites 22 studies and finds that:


"Women who had undergone an abortion experienced an 81% increased risk of mental health problems, and nearly 10% of the incidence of mental health problems was shown to be attributable to abortion."

We don't have to dig very deeply to find problems with this paper.


Problem 1 - Data misuse
The first indication of a problem with this paper is in the methodology where the author states:

"the sample comprised 22 studies, 36 measures of effect and 877 181 participants (163 831 experienced an abortion)"

This is a dishonest approach to methodology. A meta-analysis is designed to synthesise data from small independent studies in order to see if there's a larger effect. The key word here is independent. If you've got 22 studies then the maximum measures of effect you can have is 22, not 36. Also the 22 studies included in the review actually stem from 14 datasets so there should only be 14 measures of effect. Anything beyond that is double counting.


Problem 2 - Ethical considerations
In the paper the author declares no conflicts of interest. However this is doubly wrong. First, the author fails to mention in the paper that of the 22 papers included in the review, she is the author of 11 of the papers. This is a breach of established guidelines on conducting a meta-analysis. The Cochrane Collaboration guidelines on conducting a meta-analysis state:

“there should be an independent assessment of eligibility and risk of bias by a second author [of the meta-analysis] with no conflict of interest.”


However, Coleman is the sole author of the paper.


Second, she is also a prominent anti-abortion campaigner who has previously spoken about using journals as a platform for campaigning. She has stated:

"We need to develop organised research communities to continue the research, apply for grants, recruit young academics, critique data produced by pro-choice researchers, challenge politically biased professional organisations, train experts to testify and disseminate cohesive summaries of evidence."


Problem 3 - Selection criteria
There are several problems with the paper which fall under the category of selection criteria, i.e. deciding which papers to include in the study.
First, there is a general lack of transparency. While the author includes the databases searched and some general criteria, she fails to include the search terms and also fails to detail why certain papers which fulfil the criteria were still excluded from the analysis.
Second, the author fails to acknowledge or comment on the existence of publication bias. The idea of a systematic review is to examine what the individual papers in a field suggest when examined as a single entity. But the prerequisite for that is that you have a level playing field to begin with, i.e. that negative studies haven't been left unpublished. However, in this instance, Ben Goldacre and William Lee have demonstrated that there is in fact significant publication bias in this field. This should require a large caveat in the interpretation of the results but there is none.

Third, the author fails to account for the quality of the included studies. Three of the studies included were rated Poor for methodology by the Royal College of Psychiatrists. Further, in their own systematic review of the literature in 2011, 13 of the studies included in the Coleman paper were excluded because of methodological issues. Five studies were excluded because of inappropriate control of prior mental health (i.e. the papers in question didn't find out whether or not the women involved already had mental health problems before abortion). Four studies were excluded because they failed to compare groups properly (e.g. women who had abortions were compared against women who had both planned and unplanned full-term pregnancies). One study was excluded because it had overlapping groups (i.e. women who had had both an abortion and an unplanned full-term pregnancy). One study was excluded because it contained no relevant data and finally two studies were excluded because of inadequate controls relating to how the data itself was measured.

Finally, in addition to including poor quality studies in the paper, the author also fails to acknowledge that a 2008 literature review which the author references highlighted the relationship between study quality and health outcomes:

"The highest quality studies had findings that were mostly neutral, suggesting few, if any, differences between women who had abortions and their respective comparison groups in terms of mental health sequelae. Conversely, studies with the most flawed methodology found negative mental health sequelae of abortion."


Problem 4 - Reporting errors
There are two issues here, one minor and one much more significant but they are interconnected.

First, the author of the paper misleadingly reports the risk in the paper abstract as relative risk increase. You see, there are three ways that you can report on risk in a study. You can report the relative risk increase as the author has done here, you can report the objective risk increase or you can report the natural frequency.

For example, let's say that you're studying the risk of increase in heart attacks from coffee consumption. You find that drinking more than four cups of coffee a day leads to a 50% increase in risk. Now, let's say that out of every 100 people, 10 people get heart attacks. This is the natural frequency. So you can report your risk as an increase in the natural frequency, so in our hypothetical example, a 50% increase in risk means that an extra 5 people per 100 will get heart attacks (15 per 100 vs. 10 per 100). You can also report the objective risk increase. So in our hypothetical study, the baseline risk is 10%. The relative increase is 50% so the objective increase is just 5%.

Reporting only the relative risk increase is misleading because it fails to take into account baseline risk. For example, last month the Irish Independent ran the following irresponsible and thoroughly clickbait-y headline:

"C-section babies 'are one-third more likely to develop autism' but cause still unknown."

This headline is completely misleading because it makes the problem much more significant than it really is. The reality is that (as reported by Reuters), the objective risk of autism increased from 0.8% to just over 1%.



The second reporting problem is that the author also cites the population attributable risk results. This is a measure of how many people would not experience the outcome if the intervention were taken away. This means that this is the percentage of women who would not experience mental health issues if they didn't have an abortion which the author quotes as 10%. However, a population attributable risk is only useful when you have already demonstrated a causal relationship. The author of this study hasn't even demonstrated a firm correlation nevermind causation. Further, the author's method for calculating PAR is designed for outcomes which are rare among the population which is true for none of the mental health outcomes included in the paper.

Other problems
We have covered the main problems with the paper but there are other problems which are either more technical, more minor or more boring depending on your perspective and which we need not cover in more detail here. These include the author not adhering to her stated analysis criteria, the problem with multiple comparisons, problems with adjusting for confounding variables, problems with weighting of effect sizes and problems with applying appropriate controls to both the subjects and the effects measures.

So, getting back to the current topic:

splinter65 wrote: »

I don’t follow at all. Where is the dishonesty?
http://giannacare.ie/

You want to know where the dishonesty is? Fine.



It's dishonest to cite a paper on your website which has already been debunked by multiple authors (there have been 10 published rebuttals at last count) and which has been described by the Royal College of Psychiatrists as:


"“cannot be regarded as a formal systematic review.”



It's even more dishonest to describe the science on the issue as "undeniable" when all you've got is one discredited paper.



It's also dishonest to cite risks of abortion without quantifying those risks in terms of numbers. For example, Gianna Care cite death as a risk of abortion. While that is true, without providing the statistics on this risk Gianna is being misleading and scaremongering. Just to show what I mean, the death rate from general surgical procedures in the US is approximately 1.3%. The death rate from abortions in the US is approximately 0.0014%. In 2008, for example, there were 825,564 abortions performed in the USA. In that period there were only 12 deaths from abortions or complications thereof. So abortion is 1800 times less risky than any standard hospital operation.

Predicting Risk of Death in General Surgery Patients on the Basis of Preoperative Variables Using American College of Surgeons National Surgical Quality Improvement Program Data

Abortion Surveillance 2009


That and the other details linked to by DubInMeath show the dishonesty of Gianna Care. I hope that helps to clear things up.