We agree with Drs. Whiteford and Whiteford that mammography detects many non–life-threatening malignancies and that, fortunately, it misses many more microscopic ones. Mammography is far from an ideal screening test. However, for clinically important cancer, the sensitivity and specificity of mammography compare favorably with those of other screening tests. Drs. Whiteford and Whiteford also claim that the incidence of breast cancer has tripled because of wider use of mammography and that the death rate from breast cancer has risen sharply. Actually, between 1973 and 1999, a period in which mammography rates increased substantially, the annual incidence of invasive breast cancer increased from 98.5 to 139.1 per 100 000 women (1). Over this period, age-adjusted breast cancer deaths decreased from 32.3 to 27 per 100 000 women.

We agree with Dr. Gøtzsche that classification of cause of death can be biased, but misclassification is not necessarily biased in favor of screening. Misclassification could as easily bias findings against a real mammography benefit by virtue of "sticky diagnosis" bias (2). In any case, rates of advanced breast cancer are not affected by classification of the cause of death, which is why we cite them as an independent measure of the impact of screening. We also agree with Dr. Gøtzsche and Drs. Whiteford and Whiteford that a full assessment of the harms of mammography should include the predictable harms of treatment, including the risk for undergoing tumorectomies and mastectomies because of overdiagnosis and overtreatment. This issue was discussed in our Appendix, in presentations to the U.S. Preventive Services Task Force, and in the full systematic evidence review on which our Annals article was based (3).

Because treatment options, radiation doses, and practice styles have changed substantially since the trials we examined were done, we did not see value in focusing on evidence of overtreatment and overdiagnosis. If treatment and surgeons have become more conservative over time, the harms of mammography might be less severe today than they were in the 1970s and 1980s. If cancer (particularly ductal carcinoma in situ) is treated too aggressively today, the harms might be greater. Unfortunately, the data were not suitable for assessing the actual harms of treatment in current practice, so we settled for informing the Task Force of these uncertainties. Better data are needed to assess the impact of overdiagnosis and overtreatment in today's practice settings.