Reply to The incidence of fatal breast cancer measures the increased effectiveness of therapy in women participating in mammography screening
Journal
Cancer
Journal Volume
125
Journal Issue
12
Pages
2130-2131
Date Issued
2019
Author(s)
Abstract
Giannakeas and Narod erroneously claim that our findings are due to unrecognized selection bias. They believe that a substantial fraction of breast cancer deaths in the group that did not attend screening in a given invitation year could be attributable to these women having been diagnosed with breast cancer in previous years. Their speculation is mistaken. We have not studied unrefined mortality, which is biased by deaths from cancers diagnosed over a number of years before and including the incident year and is based on the date of death. Instead, we have studied the incidence of fatal breast cancers, which measures deaths from cancers diagnosed only in the incident year. Our novel method is thus based on the date of diagnosis and not on the date of death. We have explained this clearly in our article.1 With our method, for each year, incident cancers are classified according to screening exposure. To be included in the comparison group in a given year, a woman would have to have breast cancer diagnosed in that same year during our 39-year screening period. The 39-year segment of our study period included only breast cancer cases diagnosed in each of those years. Thus, Giannakeas and Narod’s example of women with previously diagnosed cancers is irrelevant to our data. Indeed, avoidance of the error of including deaths from previously diagnosed cancers was one of the motivations for the development of our method. That said, we always take seriously the possibility of a selection bias in observational studies. In our article,1 we adjusted the effect on 20-year fatal cancers for potential selection bias, and this made a very small difference, with a significant 47% reduction with screening (95% confidence interval, 37%-56%) being reduced to a significant 45% reduction (95% confidence interval, 33%-55%). In addition, because we are examining the incidence rate of fatal cancers, raising the issue of overdiagnosis is irrelevant because no woman dies of an overdiagnosed breast cancer. As we wrote in our publication, “overdiagnosis is not an issue when studying fatal cancers, because an overdiagnosed breast cancer, by definition, cannot ever be fatal.”1 Finally, Giannakeas and Narod express skepticism that a breast cancer mortality reduction greater than those observed in randomized trials could be achieved with organized, population-based service screening. In fact, the International Agency for Research on Cancer’s recent systematic review of the experimental and observational evidence for the effectiveness of mammography screening concluded that women aged 50 to 69 years who attend breast cancer screening have on average an approximately 40% reduced risk of mortality from breast cancer.2 This work was supported by the American Cancer Society through a gift from the Longaberger Company’s Horizon of Hope Campaign (project NHPDCSGBR-GBRLONG). László Tabár reports personal fees from General Electric Healthcare for preparing automated breast ultrasound teaching cases and honoraria and travel costs for General Electric Healthcare speaking engagements; personal fees from Mammography Education, Inc, in his capacity as president of the company, which organizes continuing medical education 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.
Publisher
John Wiley and Sons Inc.
Type
letter