Two recent articles [1, 2] and your editorial in Annals help bring into focus a central issue in clinical care today—namely, the balance between individual care and population-based care. Naylor and colleagues [1] note that the way that results of randomized, controlled trials are presented can affect clinicians' perceptions of therapeutic effectiveness. Presenting data in terms of relative risk reduction impressed clinicians more than using either absolute risk reduction or the "number needed to be treated" to avoid one adverse outcome. Coleman and colleagues [2] report that the use of mammography in older women increased dramatically in two surveys separated by 3 years. Finally, in the accompanying editorial you noted recent work suggesting that women under 50 years of age did not benefit from breast cancer screening efforts, and that because women 50 to 74 years of age clearly did, the failure of women in this age group to be screened could be equated with poor clinical care.

Advocates of mammography screening base their recommendations on the observed 20% to 39% reduction in the relative risk for death in screened compared with unscreened populations. If one were to report the same data as absolute risk reduction and as the number needed to be tested, those benefits would be less compelling, from 0.12% and 883 [3] to 0.02% and 5000 [4], respectively. Clearly, women at average risk are much less likely to personally benefit from screening mammography than is generally assumed. Stated another way, if the risk for dying from breast cancer for a 50-year-old woman over 20 years is 1.04% [5], then we might expect a reduction in risk to 0.63% attributable to screening mammography. Similar dilemmas are posed by other low-yield public health policy recommendations such as cholesterol, mild hypertension, or colorectal cancer screening. With each pronouncement, the population grows more anxious and medicalized, costs increase, and few individuals benefit while many endure burdens of testing and treatments.

There is a tension between the single patient seeking optimal care and a population-based risk reduction strategy. Randomized, controlled trials expressed in relative risk terminology frequently encourage mass screening or treatment as a standard of care. We believe that your editorial accurately reviewed the relevant randomized, controlled trials but overstated the contribution of such data to clinical significance and good clinical care. Combining the single-patient focus of the work by Naylor and colleagues [1] with the population focus of Coleman and colleagues' study [2] would have enriched the debate on mammogram screening and provided a creative step forward in this area of critical concern to both physicians and patients alike.

More than knowing and applying current guidelines for preventive care, internists need to know how to interpret existing data for individual patients without the powerful framing effects alluded to by Naylor and colleagues [1]. Identifying high-risk patients more likely to benefit from screening interventions or those at lower risk who would prefer aggressive screening takes time and considerable clinical skill. We submit that this is just what our patients want from their personal physicians.