Those following daily are beginning to get confortable with the jargon (I hope). For ease of newcomers following along , please consult the glossary of terms that I’ve written to make the terminology very understandable. Also, consult the post that explains the essential background.
A very busy weekend, so here in #5 is the article that should have been published on Saturday. I will catch up by later today, so as to keep my word about a paper/editorial per day.
The issue of recall bias stems from a paper we shall review this week by Rookus and van Leeuwen in the Netherlands. In the past week we have seen Palmer and Rosenberg refer to this study as the source of “reporting bias”, or “recall bias” as it is otherwise known. Before we delve into the Rookus and van Leeuwen study, we need to set the stage for understanding it and its assertion by considering a response to their study by Dr. Joel Brind of the City University of New York, along with fellow authors Vernon M. Chinchilli, Walter B. Severs, and Joan Summit-Long. Their response was a letter to the editor in:
Journal of the National Cancer Institute, April 16, 1997; Vol. 89, No. 8, 588-589.
Brind, et al., note that the Dutch paper reports a much greater relative risk (RR) of 14.6 {meaning 13-fold increase, which is gigantic} among women having had abortions from the Roman Catholic southeastern region of the Netherlands, compared to the more secular western region with a RR of 1.6 (30% increased risk). Clearly Roman Catholicism does not, when coupled with induced abortion, raise a woman’s risk of BC. How then to explain the whopping 13-fold increased risk in Catholic women, when the 30% increased risk in the more secular region is more in line with the similar data from around the world?
The answer resides in how the statistical analysis is performed. While a 95% CI was generated for these numbers, it is widely regarded in statistics that the actual sample size needs to be sufficiently large. For example, If I read in today’s paper that 66% of Americans are satisfied with the economic status of the country, I would become suspicious. If I then read that only three people were surveyed, my deep suspicions would be confirmed. The sample size was too small.
This is in fact what Brind et al., assert about the Dutch study. Only 13 women were included in the Catholic region of the country. In their own words, Brind, et al..:
“…this apparently huge difference was obtained by limiting the analysis to parous women only under the age of 45 years old, a subset containing only 13 subjects exposed to induced abortion in the southeast {Catholic region}. It is not prudent to make such a strong claim based on such a small sample, regardless of statistical significance.
“To bolster their claim, Rokus and van Leeuwen also compared self-reports with prescribers’ records of oral contraceptive use in the two regions. They found a slight but significant tendency for southeastern control subjects, compared with western control subjects, to underreport the duration of their oral contraceptive use However, since the authors found no evidence of reporting bias between case patients and control subjects (who had been matched for region), reporting bias could not logically be held accountable for the observed positive association between induced abortion and breast cancer.”
It is noteworthy to highlight the fact that the authors restricted the upper age limit to an age when many breast cancers are not yet detectable (45 yrs.), which serves to further skew the data.
When I was a psychology major as an undergraduate, we were made to take courses in statistics, quantitative analysis, research design, and tests and measurements. St. John’s University was rigorous in its emphasis on mastery of understanding valid vs. flawed research design and our ability to grasp the meaning of valid vs. invalid statistical analyses. The graduate program in biology was even more rigorous in this regard. It pays off abundantly when schools place such heavy emphasis in these areas, and young pro-lifers contemplating working in the pro-life field are well advised to take this coursework in college, regardless of their major, as Brind, et al. have demonstrated repeatedly that the great area of deception is here in the statistical arena.
As we go along, we shall see more of Brind’s exemplary work in exposing the shoddy research design and statistical analysis that served as the foundation for this fiction of reporting bias.
This October, please consider $upporting the following who desperately need our $upport to get the truth out*:
Breast Cancer Prevention Institute
Coalition on Abortion/Breast Cancer
*I have no institutional affiliation or membership with either group. Karen Malec and BCPI have been great resources for me, utterly generous with their time and resources.
and this supports an asserted link between abortion and breast cancer how?
Cranium,
A consistent finding in the literature, even research done by pro-aborts such as Palmer and Rosenberg, shows an average 30%-40% increased risk of BC in women having abortions prior to FFTP. Brind et al. have uncovered a major flaw in the research that gave rise to the much-echoed “recall bias”.
It doesn’t support the ABC link. It destroys the vehicle used to deny the relevance of the statistics which demonstrate the existence of a link. In so doing, it restores confidence in the methodology used that consistently gives rise to an association.
When you refer to “it” in the 2nd paragraph, what are you referring to?
“It” doesn’t support the ABC link.
“It” destroys the vehicle used to deny the relevance…..
“It” restores confidence in the methodology….
Sorry, but I am confused. Would you mind clarifying…I see that you wrote this at 3 am. Gosh, you’re a trooper.
Rebecca, see how the paragraph reads now. I’m sorry for the confusion.
Thanks
Hi Gerard,
Are you saying that the research that gave rise to the much-echoed “recall bias” doesn’t support the ABC link and that it destroys the vehicle used to deny the relevance of the statistics which demonstrate the existence of a link?
What is the “vehicle” used to deny the relevance of the statistics that demonstrate the existence of a link?
Rebecca,
The evidence does not exist to support the data used to promote the idea of recall bias.
Perfect! Thank you.