Also, Agronomist, I in no way chase a "golden mean". Brauner was--in contrast to Howe--a study which took a control group that was already over age 50 and followed them to see who would get breast cancer in the ensuing 12 years. But anyone who had gotten breast cancer before she was old enough to enter the study was excluded from the study. It was, however, a valuable study in suggesting that if you have an abortion and don't get breast cancer during the next 30 years, you are no more likely to get it thenceforth than a woman who has not had an abortion. (In my published letter re: Brauner, you can see that my objection was to the extrapolation of this limited result to the generalization of there being no ABC link.) So the data available at this point seem to indicate an outcome window of between about 5 or 10 years, and up to about 30 years following abortion. But it is important to note that we are not looking for such a "golden mean" window out of our desire that there be one; rather, this seems to be emerging as characteristic of the actual risk profile.

Joel Brind, Howe et al was not prospective in any manner. It was a retrospective study. Individuals with breast cancer were identified and then linked to foetal deaths. Prospective studies enrol individuals and then follow them into the future. You should know the difference between a prospective study and a retrospective study, so I don’t understand why you keep claiming Howe et al. was a prospective study.

One point worth making about Howe et al. is that it found a significant effect of spontaneous abortions on breast cancer risk, something you claimed in the comments above did not happen.

So who has disproved the recall bias hypothesis? Don’t tell me it is Joel Brind? I think we have already established that Joel Brind is biased and his opinions are untrustworthy. Where is the data? It is well known that recall bias is a problem with case control studies and not just those involving abortion. It is why any case control study needs to be treated with some caution.

I will leave your comments about the golden mean for others to read. I think they illustrate quite clearly the problem you have.

Well said Agronomist.
I think you can safely rest your case : )

I'm disappointed in you, Agronomist, for I did not think you were one for mincing words re: unimportant distinctions. The key element in Howe is that the data were prospective, and therefore immune to the possibility of response bias.

Meanwhile, reporting bias was ruled out of Daling 1994, who reported a significant ABC link of OR = 1.5. In her study, she tested for bias by using a study population for cervical cancer and abortion. That population, with controls selected from the same general population and on the same basis of her breast cancer study, showed an OR of 1.0. If response bias were present, the cervical cancer population should also have shown an OR >1, but it did not.

Now back to Howe 1989, which is arguably the best evidence against response bias as the reason ORs >1 are consistently observed in ABC studies. You say she obtained a statistically significant association with spontaneous abortion as well, which means you only read the abstract. The Howe abstract is in error: Read the actual study. In table 3, the results are reported separately for induced and spontaneous abortion: Induced abortion: OR = 1.9, 95% confidence interval: 1.2 - 3.0; Spontaneous abortion: OR = 1.5, 95% confidence interval: 0.7 - 3.7, ergo, not statistically significant.

Joel Brind, when I first referred to Howe et al. I noted that it was not affected by recall bias. However, it is still a retrospective study and the fact that you have continually referred to it as a prospective study until I pulled you up, suggest to me that you are want to gild the lily when it suits your bias. What Howe et al. does suffer from is failure to identify confounders; hence these are not managed for.

Moving on to Daling, her study did not rule out reporting bias. In fact the study identified reporting bias among the subjects. 21% of the cases failed to report an abortion that was listed in the national registry as did 29% of the controls. That is reporting bias for you: failure to report an important event. More worryingly, in Daling’s study 7 times as many cases reported having abortions that were not listed in the national registry compared to controls. The most likely explanation is that these were abortions conducted by unlicensed practitioners and hence were illegal (even though abortion was legal in the state at the time). These represent 20% of the cases who reported having an abortion.

Still mincing words, I see. A case-control study that is nested in a prospective data base is also referred to as an historical prospective study, since the data are prospective. period. But since we at least agree that, because of the prospective nature of the data, the study is immune to the possibility of reporting bias (aka response bias), let's just stop beating this dead horse, shall we?

Yet your confusion re: the Howe study is now compounded by your confusing Daling (1994) with the 1991 study of Lindefors-Harris (not the Daling study), in which the interview records of subjects were compared with the national registry data on abortions. (There is no national abortion registry in the US, where Daling's population was drawn from.) You throw around numbers like "21%" and "29%" of case and control subjects having underreported abortions, which gives the impression that these numbers are significant. They are not. For a statistically significant result, the authors needed to compare underreporting among controls to "overreporting" among cases (which nonsense I discussed in an earlier comment). But even then, the gigantic ratio you tout of "7 times as many cases reported having abortions that were not listed in the national registry compared to controls" was literally owing to just 7 actual patients compared to one actual control subject!

You really need to read up on your abortion-breast cancer literature and get your facts straight, Agronomist. Or maybe just stick to agronomy.