et al. 3 have raised the long-standing question of the how frequently women should be screened for breast carcinoma. The current level of uncertainty regarding the best screening interval is reflected in the diversity of recommendations. In the U.S., the American Cancer Society, American Medical Association, and other major organizations recommend annual screening beginning by the age of 40 years. The National Cancer Institute recommends screening "every one to two years." Biennial screening is recommended in Canada, Australia, and much of Europe. In Sweden it is recommended every 18 months for women ages 40 -49 years and every 24 months for women age Υ 50 years, whereas in the United Kingdom screening is performed every 3 years. These diverse recommendations largely have arisen in isolation from any empiric guidance with regard to the effect of the screening interval on the ability of mammography to reduce deaths from breast carcinoma. The three recent articles 1-3 should help by framing the question concerning the proper screening interval in a form that is scientifically based, analyzable, and justifiable.
Hunt et al. 1 reported on the outcome of screening annually versus biennially in a screening program. Although women in their dataset were not assigned to annual versus biennial screening at the outset of the study, Hunt et al. were able extract estimates of the effect of various screening intervals by sorting women into those who returned to screening at either 10 -14-month intervals ("annual") or those who returned at 22-24-month intervals ("biennial"). They combined these data with information from the cancer registry regarding women found to have interval tumors, defined as tumors not detected at screening that were noted within either 1 year ("annual") or 2 years ("biennial") of a negative screening examination. Two findings from this analysis were statistically significant and noteworthy. First, Hunt et al. found that the median and mean greatest dimensions of the tumors observed in the annual group were smaller than the median