Does mammographic density predict survival in women with invasive breast cancer? The need to account for potential confounding from cancer stage and overdiagnosis

•This editorial highlights relevant epidemiological issues related to the recently published original research paper [[1]Boyd N.F. Guo H. Martin L.J. Sun L. Stone J. Fishell E. et al.Mammographic density and risk detection of breast cancer.N Engl J Med. 2007; 356: 227-236Crossref PubMed Scopus (1749) Google Scholar].•Sturesdotter L, Larsson A-M, Zackrisson S, Sartor Investigating prognostic value mammographic tumor appearance in women with invasive cancer: The Malmö Diet cancer study. Breast. 2023; 70:8–17. High on mammography is a well-established factor for diagnosis Scholar,[2]Boyd Rommens J.M. Vogt K. Lee V. Hopper J.L. Yaffe M.J. as an intermediate phenotype cancer.Lancet Oncol. 2005; 6: 798-808Abstract Full Text PDF (1) Scholar]. There interest whether might also predict how aggressive is, including likelihood causing death. If features at time were found have value, this added histopathology, staging, other clinical features, then information could utility management decisions. For example, choice treatment sequencing be impacted, such decision use neoadjuvant chemotherapy (see Fig. 1). In issue Breast, Dr Sturesdotter colleagues investigate first part question (do value) over 1100 diagnosed who participating Cancer study [[3]Sturesdotter A.-M. S. study.Breast. 70: 8-17Abstract (0) Using large population-based sample women, they present robust evaluation potential associations between findings subsequent dying from cancer, separate analysis screen-detected clinically detected (stratified analysis) adjustment many confounders. Unfortunately, data not available stage which important confounder that known associated (cancers dense breasts tending later stage) mortality. Lack unmeasured confounders may bias estimates association appearances long-term prognosis. Figure explores possible effects unadjusted case screen cancers. Unmeasured explain authors’ finding increased mortality high where was (HR 1.45, CI 0.87 2.43). scenario, acting marker higher through its previous screening masking early [[4]Tagliafico A.S. Houssami N. Towards consensus managing population screening?.Breast. 69: 422-423Abstract Early-stage cancers are liable overdiagnosis – lesions histopathology (not false positive) but would never caused symptoms or death if left undetected untreated [5Barratt A. Overdiagnosis screening: 45 year journey shadowy idea acknowledged reality.Bmj. 2015; 350: h867Crossref (54) Scholar, 6Thomas E.T. Del Mar C. Glasziou P. Wright G. Barratt Bell K.J.L. Prevalence incidental precursor autopsy studies: systematic review meta-analysis.BMC Cancer. 2017; 17: 808Crossref (32) 7Glasziou P.P. Jones M.A. Pathirana T. A.L. Estimating magnitude Australia.Med Aust. 2020; 212: 163-168Crossref (62) 8Ryser M.D. Lange Inoue L.Y.T. O'Meara E.S. Gard Miglioretti D.L. al.Estimation U.S. Breast cohort.Ann Intern 2022; 175: 471-478Crossref (25) Women low more likely overdiagnosed although has yet been demonstrated empirically. This uneven distribution cause apparent density. Both less prominent setting Furthermore, both tend favour effect, adds confidence no group real. method measurement important. Density measured using “qualitative” 4 category Imaging Reporting Data System (BI-RADS, used current study) diagnosis, single Scholar], changes measurements predictive [[9]Mokhtary Karakatsanis Valachis Mammographic risk: meta-analysis.Cancers. 2021; 13Crossref (12) Scholar,[10]Jiang Bennett Rosner B.A. Colditz G.A. Longitudinal change each risk.JAMA 9: 808-814Crossref However, there considerable inter-rater intra-rater variability determining defined ‘dense tissue’ BIRADS compared their outcomes ‘fat involuted’ ‘moderately dense’ 1–3 combined. Other studies tended define 3–4, comparison 1–2. Consistency definitions allow results studies. With transition digital [[11]Farber R. Wortley Jacklyn Marinovich M.L. McGeechan al.Impact full-field versus film-screen meta-analysis.J Natl Inst: Inst. 113: 16-26Crossref quantitative volumetric possible, decrease estimates, especially repeated measures same technology available. A recent woman subsequently [[10]Jiang significantly attenuated ipsilateral (in diagnosis) control (who did develop cancer), providing incremental prediction baseline alone It way provide cancer. overall negative Malmo offer some reassurance while tissue it affect prognosis once diagnosed. should interpreted caution due possibility breast-specific health beyond diagnosis. Future seeking hypothesis include consideration ways account overdiagnosis.