Beyond survival: a closer look at lead-time bias and disease-free intervals in mammography screening

In the ever-evolving landscape of early breast cancer detection, routine mammography screening has long been a subject of both praise and controversy. Amidst the intricate tapestry of statistical analyses, 1 challenge that often casts a shadow on mammography’s benefits is lead-time bias, where routine mammography has been criticized for possibly overinflated survival benefits because of this bias. In this issue of the Journal, de Munck et al. (1) delve into how to address lead-time bias, adjusting for it while concurrently exploring breast cancer outcomes beyond survival.

Lead-time bias occurs when an asymptomatic breast cancer is detected by routine screening mammography at an earlier time point in its natural history than when it would have been detected clinically. For example, a patient may survive for 15 years after a screen-detected diagnosis compared with surviving only 10 years after a clinically detected diagnosis. This longer survival time, however, may simply reflect earlier diagnosis if the time of death for the patient is unchanged. This additional “lead time” may not be adequately considered in survival analyses from mammography clinical trials. The result, if measured from the time of asymptomatic screen detection, is that the survival time may appear artificially inflated for women undergoing screening vs not undergoing screening.

The authors use a method to correct for lead-time bias described by Duffy et al. (2) and a conservative mean tumor sojourn time of 4.3 years for the preclinical screen-detectable period in the Dutch population, resulting in maximum adjusted lead times of 2.0 to 6.1 years for their study cohort. In addition to presenting a novel analysis in which they adjust for lead-time bias, de Munck et al. (1) also examine differences in disease-free intervals based on women with breast cancer detected by mammography vs detected clinically.

The disease-free interval is an important but less examined outcome, defined as the time span where women cured from their primary breast cancer do not experience locoregional recurrence, distant metastases, or contralateral breast cancer. Disease-free intervals can serve as a marker for the quality of life of women who are cured of their primary cancer but at increased risk of breast cancer recurrence. Moreover, unlike all-cause mortality, which includes many causes of death that mammography cannot affect, use of the improved disease-free interval can more directly be tied to routine mammography screening and early detection.

The researchers used robust data from the population-based Netherland Cancer Registry, with up to 10 years of follow-up for women diagnosed with breast cancers between 2005 and 2008. The method of detection was recorded for every breast cancer; thus, the researchers were able to allocate women into screen-detected and clinically detected cancer groups (the latter further subdivided into interval cancers, defined as breast cancers diagnosed within 24 months after a negative screening mammogram result, and non–screen-related breast cancers). The study authors concluded that women with mammography screen-detected breast cancers had statistically significantly improved disease-free intervals compared with women with clinically detected interval breast cancers (hazard ratio = 0.77, 95% confidence interval = 0.68 to 0.87) and compared with women with clinically detected, non–screen-related breast cancers (hazard ratio = 0.76, 95% confidence interval = 0.66 to 0.88).

This analysis adds to the literature on the potential benefits of routine mammography in multiple ways. First, the authors stratified their breast cancer population by the primary method of detection. Clinically detected interval breast cancers are known to be more aggressive, with a worse prognosis than screen-detected breast cancers (3). Second, the study investigators compare their results adjusted for lead-time bias with results of an analysis correcting for patient and tumor confounders instead of lead-time bias using the same dataset. This comparative analysis demonstrated no change in association between method of primary breast cancer detection and the disease-free interval based on the 2 methods for adjusting for confounding factors. Third, the use of disease-free interval rather than overall survival as a primary outcome acknowledges benefits of current early breast cancer detection and treatment approaches, which have collectively improved overall breast cancer–related survival. Because of the strong interaction between early detection and early treatment, the benefits of mammography may be partly explained by early treatment (4). As the study authors point out, however, early treatment is not possible without accurate early detection from routine screening.

Finally, this study highlights the importance of collecting data on the method of breast cancer detection. Although many European countries collect details of the method of detection, the United States does not. Currently, US breast cancer databases, including the National Cancer Institute’s Surveillance, Epidemiology, and End Results program, do not require method of detection to be recorded. As a result, the American College of Radiology Breast Commission recently called for a change in data-collection practices to capture method of detection for all breast cancers diagnosed (5). Doing so may be challenging given the siloed nature of mammography data in the United States, but these data would help address several contentious issues, including questions on the relative benefits of mammography and treatment in reducing breast cancer morbidity and mortality, understanding differences in tumor molecular signatures for screen-detected vs interval cancers, and addressing disparities in breast cancer screening outcomes across different US subpopulations (5).

The analysis presented by de Munck et al. (1) sheds light on the often-overlooked domains of lead-time bias and disease-free intervals, offering a nuanced perspective on the possible impact of routine mammography. With more robust data-collection practices, the potential benefits of routine screening beyond mortality may be better teased out from long-term screening data from different population-based programs. Future studies should examine mammography benefits and harms from the perspectives of risk of recurrence and disease-free interval and by method of primary breast cancer detection. Such analyses would help all mammography screening stakeholders clarify the benefits vs harms of routine mammography screening, especially in the United States. The American College of Radiology’s push for breast cancer registries to properly record method of detection for all breast cancers diagnosed will certainly aid health-care professionals, policy makers, and researchers in this future work.

Data availability

No new data was generated or analyzed for this editorial.

Author contributions

Christoph I. Lee, MD, MS (Conceptualization; Writing—original draft; Writing—review & editing), Joann G. Elmore, MD, MPH (Conceptualization; Writing—original draft; Writing—review & editing).

Funding

No funding was used for this study.

Conflicts of interest

The authors report no conflicts of interest related to this article. C.I.L. reports personal fees from the American College of Radiology for journal editorial board work and textbook royalties from Oxford University Press; McGraw Hill, Inc; and UpToDate, Inc, all unrelated to this manuscript. J.G.E., a JNCI associate editor and author of this editorial, was not involved in the editorial review of the manuscript or decision to publish the editorial.