Are risk factors other than age useful for establishing screening intervals? No study has shown an association between a breast cancer risk factor and the length of the detectable preclinical phase. Thus, any attempt to decide on the basis of risk factors whether to be screened more or less frequently than recommended guidelines would be a decision based on the underlying probability of disease, not the underlying probability of having a longer or shorter sojourn time. It may also be a decision based on one’s concern about the risk of an abnormal mammogram and potential interventions that ultimately may be false-positives.
In this case, less frequent screening may reduce the risk of an avoidable intervention based on a false-positive result simply based on exposure to screening. However, that might not prove true, because specificity tends to be lower at the prevalent screen, and therefore it is by no means certain that less frequent screening reduces the odds of false-positive results. Here again, a woman would need to reconcile the greater odds of a false-positive examination compared with the lower odds of developing breast cancer and consider the relative benefits and potential costs associated with her decision.
On a population basis, risk-based screening for breast cancer has never been shown to have much potential to identify the majority of new cases by screening a subgroup of women at higher risk. In 1984, Solin et al. reported on the screening experience of 17,543 women, for whom data on eight risk factors were collected (including any family history, prior breast biopsy, menstrual history, pregnancy history, lactation, and hormone use). They concluded that more than one-half of the 246 cases diagnosed would not have been detected if the program had only screened women with either prior breast biopsy or family history and that more than 40% would have been missed if the program had been limited to any one of the listed risk factors.
In 1987, Alexander and colleagues reported similar findings from the analysis of Edinburgh trial data; by screening the 20% of the population at higher risk due to having had a prior biopsy or a family history in a mother or sister, only 30% of the first-round cancers would have been detected. When menopausal status or nulliparity/first birth after age 30 were included, the proportion of new cases that would have been detected in the first round increased to 29.6%. Again, concentrating on the more important risk factors would fail to identify the majority of prevalent cancers in an asymptomatic population.
Madigan et al. recently reported population-attributable risk estimates for breast cancer using data from the National Health and Nutrition Examination Survey Epidemiologic Follow-Up Study. Using well-established risk factors, such as later age at first live birth, nulliparity, higher family income, and family history of breast cancer in first-degree relatives, these risk factors were associated with approximately 41% of breast cancer cases in the United States In each of these instances, it is estimated that fewer than one-half of breast cancer cases would be identified by a risk-based screening strategy applying well-established risk factors.
Robert A. Smith and Carl J. D’Orsi
R. A. Smith: Cancer Screening, Department of Cancer Control, American Cancer Society, Atlanta, Georgia
C. J. D’Orsi: Diagnostic Radiology, University of Massachusetts Memorial Medical Center, Worchester, Massachusetts