SUMMARIES OF LETTERS TO MEDICAL JOURNALS
Chris Kahlenborn's Letter to the Lancet Regarding Beral et al. 2004
In his letter, Kahlenborn criticized a review of the research by Oxford scientists.  He said the Oxford team "failed to accurately portray the world's literature."
Oxford scientists cited a prospective Swedish study by Lindefors-Harris et al. to support their argument that prospective studies are more reliable than retrospective studies.  At one time, Lindefors-Harris et al. claimed to have found credible evidence of recall bias. They made the preposterous claim that they'd found evidence that women report more abortions than they really had. If it were true that more patients than healthy women overreport their abortions, then findings from research relying on women's reports of their abortions - retrospective research - would be false.
The Swedish group, however, withdrew their absurd claim in 1998 after receiving criticism from two teams of scientists. [3,4,5] They admitted that the computerized medical records of abortions they used might be incomplete.
Prospective research has severe limitations because women tend not to report their abortions and often travel at a distance to get abortions that are more anonymous. The lesson to be learned is that even prospective research can be seriously flawed if medical records of abortion histories are deficient.
Kahlenborn said Oxford scientists failed to "critically examine the quality of the prospective studies" and that over one-third of their data represent unpublished studies which other scientists "will never be able to double-check."
Patrick Carroll's Letter to the Lancet Regarding Beral et al. 2004 
In his letter, Carroll said the Lancet's credibility is "impaired by the failure throughout the world to fully record induced abortions." Referring to a prospective study on British women by Goldacre et al., Carroll asserted that the authors "found only a small fraction of the women that could have been expected to have had abortions from what is known from national incidence data." 
Carroll added that if a British study can have so many unrecorded abortions, then under-recording must be even more prevalent in countries that legalized abortion after Great Britain did in 1967. Carroll said many of the abortions relevant to the Oxford paper, which were performed in other countries, were illegal. Therefore, women in other countries will be more reluctant to admit to having had abortions than British women will be.
Prospective research isn't the "gold standard" in epidemiological research that some say it is because medical records are deficient in capturing all abortions. Carroll concluded that a better way to conduct an investigation of the ABC link is to use national data that can be correlated to successive age groups.
Carroll's research relied on historic national data reporting breast cancers and abortions in England and Wales.  He used the data to make forecasts of breast cancer trends. England and Wales have exceptionally thorough registrations of breast cancers and abortions. Therefore, nearly all abortions are captured by official statistics. He found that abortion is the "best predictor of English breast cancer trends and, in particular, the increase in breast cancer incidence."
Joel Brind's Letter to Cancer Epidemiology Biomarkers and Prevention Regarding Mahue-Giangreco et al. 2003
In his letter, Joel Brind criticized a California study by Mahue-Giangreco et al. 2003, which found no increased breast cancer risk associated with abortion. 
Brind noted that Mahue-Giangreco et al. had the highest patient refusal rate of all of the ABC studies (11.5%). What this means is that a significant number of eligible breast cancer patients refused to participate as subjects in the study. Brind observed, "This high rate of patient refusal may reflect the sensitive nature of induced abortion, and the consequent reluctance of women to report it."
Brind argued that abortion advocates in the scientific community have often theorized that healthy women are more reluctant to report their abortions than are breast cancer patients, but credible evidence of recall bias has never been provided. Additionally, scientists have repeatedly demonstrated that it doesn't exist. [4,8,9]
What would cause so many women to refuse to participate as study subjects in an ABC study? Brind offered a very reasonable answer - their desire to keep their abortions secret. It makes little sense to follow abortion advocates' line of reasoning - that women who don't want to talk about their abortions choose to join ABC studies, but then they lie to researchers about their abortion histories. That is the faulty line of reasoning that abortion advocates are proposing when they argue that recall bias is a problem plaguing retrospective research.
Brind took his argument one step farther and asked his readers to consider how scientists' findings might be impacted if the number of women with declared abortions are underrepresented in the patient group. He remarked that the prevalence of abortion among patients included in the study should have been 30% higher. That would accurately reflect the prevalence of abortion in California during the years that most of the abortions in the study were performed. If the correct number of abortion-positive breast cancer patients had participated in the study, then the risk elevation would have been 20%.
1. Beral V, Bull D, Doll R, Peto R, Reeves G. Lancet (2004) 363:1007-16.
2. Lindefors-Harris et al. (1991) Am J Epidemiol 134:1003-1008.
3. Meirik et al. (1998) J Epidemiol Community Health 52:209.
4. Daling et al. (1994) J Natl Cancer Inst 86:1584-92.
5. Brind et al. (1998) J Epidemiol Community Health 52:209-11.
6. Mahue-Giangreco M et al. (March 2003) Cancer Epidemiology Biomarkers and Prevention, Vol. 12, 209-214.
7. Goldacre MJ, et al. J Epidemiol Community Health (2001) 55:336-337.
8. Howe et al. (1989) Int J Epidemiol 18:300-4.
9. Tang et al, Case-control differences in the reliability of reporting a history of induced abortion, Am J Epidemiol, 2000, 151:1139-43.