Commonly used tools: computer models underestimate breast cancer risk.(Gynecology)

LOS ANGELES -- Mathematical models commonly used to predict a woman's risk of developing breast cancer profoundly underestimated that risk in a British study of 3,170 subjects tracked for 5 years for evidence of cancer after a baseline mammogram.

The computerized Breast Cancer Risk Assessment Tool--commonly known as the Gail model--which is published on the National Cancer Institute's Web site and is frequently used by physicians and genetic counselors to educate individual women about their breast cancer risk, was among the least accurate of five models tested by Dr. D. Gareth Evans of St. Mary's Hospital in Manchester, England.

"The Gail model underestimated by a factor of two," Dr. Evans said at the annual meeting of the American Society of Human Genetics. "It only predicted half of the cancers ... that actually occurred."

The validation study differed from some other analyses of risk models in that it looked prospectively at subjects after assessing their risk factors, rather than retrospectively calculating accuracy in women enrolled in unrelated studies.

Dr. Evans and his associates obtained a complete set of risk factors on 3,170 women, 1,933 of whom underwent annual mammograms as part of the study. The other women were tracked via a comprehensive cancer registry for at least 5 years after a mammogram at baseline showed no sign of cancer.

A total of 64 women developed breast cancer.

The Gail model missed the mark most dramatically by underestimating the breast cancer risk faced by women with a very common risk factor: one first-degree relative with breast cancer.