Dr. Kopans and Dr. Logan-Young and colleagues question looking at the results of breast cancer screening by age groups. Subgroup analyses are appropriate when looking for clinically important differences, the hypothesis precedes the analysis, few hypotheses are tested, the comparison is within studies, differences are consistent across studies, and differences are plausible [1]. In completed studies, confidence intervals, rather than power calculations, determine how large an effect could have been missed by chance alone. A meta-analysis of the six reported trials that included women 40 to 49 years old found a relative risk of 0.99 (95% CI, 0.74 to 1.32) after 7 years of follow-up [2]. Dr. Kopans cites a combined analysis of four Swedish trials showing a statistically insignificant 13% mortality rate reduction at 12 (not 8) years of follow-up [3]. Even if substantiated by other studies when they achieve 12 years of follow-up, the benefit in younger women would be substantially slower and smaller than that seen after screening older women.

The controversy surrounding breast cancer screening in women under the age of 50 raises the question about the appropriateness of recommending screening without strong, direct evidence of benefit. We all want screening to work, for breast cancer, lung cancer, prostate cancer, or cardiovascular disease, but is not the burden of scientific proof on those who would recommend screening?

We appreciate the reminder by Drs. Ross and Gerber that when results of preventive maneuvers are expressed in terms of absolute risk and numbers to screen, perception of benefit is smaller. This is so because even "common" health conditions, such as breast cancer, are really rather uncommon at any point in time. Like Dr. Wright, we noted that not all trials showed statistically significant mortality reductions in women over 50 years of age. Results in two of the trials were significant; a meta-analysis of six trials found a relative risk of 0.66 (95% CI, 0.55 to 0.79) [2]. The Canadian trial in women ages 50 to 59 does not contradict this because that trial was a study of the incremental effect of mammography over a carefully performed, standardized clinical breast examination. We appreciate Dr. Giliberti's reminder about the need to perfect training in the clinical breast examination.

We referred the specific criticisms of the Canadian studies to the authors for their response, which is published below.