Researcher questions new study suggesting benefits of breast screeningBMJ 2018; 363 doi: https://doi.org/10.1136/bmj.k4790 (Published 13 November 2018) Cite this as: BMJ 2018;363:k4790
All rapid responses
Re: Researcher questions new study suggesting benefits of breast screening: Researcher’s questions reflect a failure to carefully read the study
Kalager and Jørgensen both fail to realize that no previous publication has measured the incidence of fatal breast cancer. Kalager claims that the methodology we used in our recent paper published in CANCER (1), i.e., the incidence rate of fatal breast cancer” is “really incidence-based mortality”, but she is mistaken. In incidence-based mortality, the X-axis represents the year of death, making correlation with exposure at the time of incidence subject to considerable bias. On the contrary, in our novel method, the X-axis represents the year of diagnosis for breast cancers that become fatal within 10 or 20 years, making possible a direct comparison of screening exposure with non-exposure at the time of breast cancer incidence. As we noted, improvements in therapy do not account for our results because in each comparison year, women received the state-of-the art therapy appropriate for their stage-at-diagnosis, regardless of how their cancer was detected.
Kalager and Jørgensen also raise the issue of healthy screenee bias. Each also failed to notice that our analysis adjusted for selection bias. But it also is the case that healthy screenee bias has little relevance to a service screening program in which more than 85% of all women aged 40-69 participate. How could the entirety of 85% of the population have the super healthy attributes mentioned, as well as being “well-edicated [sic] and affluent?”
We counted the number of women who died of breast cancer within ten years, not “survived” as Jørgensen mistakenly claims. While it also is easy to be dismissive by invoking lead-time bias, he should be aware that most women who are destined to die from breast cancer will have died before 20 years of follow-up. Thus, lead-time bias also is not a concern. Jørgensen states that “the authors do not seem to have taken overdiagnosis into account”. It should be obvious that overdiagnosis is utterly irrelevant when studying fatal cancers, as we pointed out in the Introduction of our article: “Overdiagnosis is not an issue when studying fatal cancers, because an overdiagnosed breast cancer, by definition, cannot ever be fatal.”
Jørgensen’s citation of the IARC Handbook of Cancer Prevention vol. 7 is outdated, as the 2016 update (in which three of us participated) clearly states that “There is sufficient evidence that screening women aged 50–69 years by mammography reduces breast cancer mortality. This evaluation is supported by randomized controlled trials of efficacy of mammography screening and by observational studies of effectiveness of both invitation to and attendance at service mammography screening. Women aged 50–69 years invited to service mammography screening have, on average, a 24% reduced risk of mortality from breast cancer. Women aged 50–69 years who attend service mammography screening have, on average, about a 40% reduced risk of mortality from breast cancer.” (2)
Finally, with respect to his final remarks, as usual, when someone has lost the scientific argument they fall back on accusations of conflict of interest.
1 Tabar L, Dean PB, Chen TH, et al. The incidence of fatal breast cancer measures the increased effectiveness of therapy in women participating in mammography screening. Cancer 2019;125:515-23.
2 IARC Working Group on the Evaluation of Cancer-Preventive Strategies. Breast Cancer Screening. Vol 15. Lyon, France: IARC Press; 2016, p 467.
Competing interests: Dr. Tabar reports personal fees from General Electric Healthcare for preparing automated breast ultrasound teaching cases, and honoraria and travel costs to General Electric Healthcare speaking engagements; personal fees from Mammography Education, Inc. in his capacity as president of the company, which organizes CME courses on breast imaging; and personal fees from Three Palm Software for consultation related to breast imaging interpretation, all outside the submitted work. The other authors made no disclosures.
Mette Kalager is correct that there is nothing new about this study or its design and that comparing outcomes for attendees with that of non-attendees is problematic. As there is no reliable way to know the influence of this well-known source of bias also called the healthy screenee bias, it cannot be corrected for reliably, especially as it likely varies with context and over time. I believe Angela Raffle and Muir Gray said it best when writing that the screenees are the "healthy, well-edicated, affluent, physically fit, fruit and vegetable-eating non-smokers with long lived parents" [Raffle/Gray: Screening. Oxford University Press 2007].
But there are (at least) two further well known problems.
The study seems to have counted the number of women who survived their breast cancer at 10 years in attendees and non-attendees, respectively. As breast cancer should work by advancing the time of diagnosis, more screenees will be alive 10 years after their diagnosis, regardless if the earlier diagnosis led to improved prognosis or not (lead time bias). They simply live longer with the cancer label, but necessarily not longer in absolute terms . As we do not have reliable knowledge about lead-time, which is also context dependent and likely to vary over time, this cannot be corrected for with any greater confidence either.
Lastly, but most importantly, the authors do not seem to have taken overdiagnosis into account. Overdiagnosed tumors invariably have an excellent prognosis, as they are defined as those detected through screening that would never have caused health problems or have been detected without screening. There are many such cases of overdiagnosis; 3 for every 1 woman who has her life extended according to the Marmot review . Of course, the diminishing group of researchers who believe overdiagnosis does not exist or is a very minor problem would be unlikely to see this as a major flaw of the present study.
These problems were well recognised already in 2002 when the IARC/WHO issued their Handbook of Cancer Prevention vol. 7 (breast cancer. This noted that "observational studies based on individual screening history, no matter how well designed and conducted, should not be regarded as providing evidence of an effect of screening" (pp. 91). I think this is absolutely true and very well put. The second author of the present paper was a member of the Secretariat for that Handbook.
It should be noted as a CoI that, not only did the American Cancer Society, who actively promote breast screening, fund this study. The journal Cancer in which the present journal is published is also "a journal of the American Cancer Society" (https://onlinelibrary.wiley.com/page/journal/10970142/homepage/productin...), and the corresponding author is "Vice President, Cancer Screening, American Cancer Society" (https://onlinelibrary.wiley.com/page/journal/10970142/homepage/productin...).
Competing interests: No competing interests