The number of women in the United States today who are living into old age is steadily climbing as a result of aging of the “baby boom” generation, increases in life expectancy, decreases in rates of cardiovascular disease, improvements in medical care, and trends toward adoption of healthier, more active lifestyles. This so-called graying of America is coupled with cancer incidence rates that increase steadily with advancing age, peaking at approximately age 80 years. As a result of these parallel demographic and epidemiologic forces, over the coming years, more and more women are likely to survive to be at risk of developing cancer in their 80s. In fact, one third of all cancers in women and 23.5% of invasive breast cancers occur among women 80 years of age and older. This at-risk older population is very heterogeneous, with 25% of women who live to age 80 years being fairly robust and having a life expectancy of approximately 13 years, whereas 50% have a life expectancy of 8.6 years; the remaining 25% are frailer and will only live 4.6 years. The recent article by Badgwell et al on breast cancer screening among women 80 years and older and the controversy that has followed from its publication and media dissemination highlights several of the dilemmas facing clinicians, patients, and professional groups as they struggle to make rational, evidence-based decisions about optimal care for this diverse and growing older population. First and foremost of these difficulties is that women 80 years and older basically have not been included in mammography clinical trials—the gold standard of evidence about medical interventions. In the absence of clinical trial data, observational data are often used to provide evidence about the effectiveness of medical interventions in broader populations. Badgwell et al used the Surveillance, Epidemiology, and End Results–Medicare database to examine women 80 years and older with newly diagnosed breast cancer by past mammography use to provide indirect evidence regarding the value of screening in this age group. They found that women with more recent screening were diagnosed with smaller size tumors and had better 5-year breast cancer and all-cause survival than women with less screening. However, as pointed out by the authors themselves and their critics, these results may indicate screening benefits, or they may simply represent lead time, length biases, and selection factors. As noted by Berry et al in a letter about the article appearing in this issue of the Journal of Clinical Oncology, lead time biases result in an earlier diagnosis without any associated mortality benefit (ie, the woman is diagnosed earlier in the disease process, but that diagnosis and the treatment that follow do not alter the time that the woman was destined to die). In another observational study among older women that attempted to correct for lead time, screening seemed to reduce mortality for women younger than age 84 years but not among those older than 85 years. However, the lead time used, 1.25 years, may be an underestimate of the true lead time among older women, so this conclusion may also be overly optimistic. Other studies that attempt to correct for observational biases have not included women older than 74 years or have not separated results for women 80 years and older. Length bias (screen detection of the slowest growing tumors) results in screen detection of tumors with the most favorable prognosis, leading to overestimates of benefits. Thus length bias can be largely explained by the biologic characteristics of a tumor that govern its growth rate. For instance, in one large case series recently reported by Dong et al, biomarkers of tumor growth and aggressiveness were significantly related to mode of detection, with screen-detected tumors being more likely to have favorable prognosis markers than clinically detected tumors. Older women (age 60 years) were also more likely to have screen versus symptomatic detection, although the numbers and distribution of women in the 60 years age group was not presented. Of note, although mode of detection was an independent predictor of recurrence-free survival, it was not a significant predictor of breast cancer mortality. The authors concluded that it is impossible to separate length bias from any possible benefits of screening. The most extreme example of length bias is the situation in which a woman is diagnosed and treated for a screened detected breast cancer, but without screening, this slow-growing cancer would not have become clinically apparent before she died of, say, congestive heart failure a year later. This latter situation is sometimes referred to as overdiagnosis. Recently screened women in the study by Badgwell et al also had higher 5-year all-cause survival than those with less screening. This JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 27 NUMBER 4 FEBRUARY 1 2009