By Erik Ness

A Race-Tailored Risk Calculator

A new tool may better predict breast cancer risk in black women

By Erik Ness

While surfing the internet, most of us ignore pop-up windows. But one in particular deserves close attention: a caveat from the National Cancer Institute (NCI) website’s Breast Cancer Risk Assessment Tool, which estimates risk of developing invasive breast cancer. Visitors to the tool’s webpage (www.cancer.gov/bcrisktool) get this warning: If you’re not a white woman, the results of the risk calculator may not be valid.

Check “Asian or Pacific Islander” or “American Indian or Alaskan Native” or “Hispanic” or “Unknown” and the program warns that its predictions are uncertain. Most unsettling was the warning that appeared, until recently, if users were African-American: “Assessments for black women may underestimate the chance of breast cancer and are subject to greater uncertainty than those for white women.”

That warning was removed when a more accurate model for predicting breast cancer in African-American women was recently incorporated into the tool.

The previous version of the risk calculator was based on a model developed by Mitchell Gail, a biostatistician at the NCI’s Division of Cancer Epidemiology and Genetics. The Gail model records a few simple breast cancer risk factors, including a woman’s current age, her age at first menstruation, her age at first childbirth, the number of her first-degree relatives with breast cancer, and whether or not she has had a breast biopsy. Because the tool was developed and tested on a primarily white population, its predictive value for other ethnic groups has been questioned.

But now Gail has revised his model for African-American women, after working with a research team that included Joseph Costantino, a biostatistician at the National Surgical Adjuvant Breast and Bowel Project Biostatistical Center at the University of Pittsburgh. The researchers used data from the Women’s Contraceptive and Reproductive Experiences (CARE) study, which was designed to examine lifetime hormone use and breast cancer risk. Gail and Costantino identified about 1,600 African-American women in CARE diagnosed with breast cancer and a similar-size control group of CARE women without breast cancer, then combined their data on all of these women with breast cancer data from the NCI’s comprehensive Surveillance, Epidemiology and End Results (SEER) database. The final model was validated against data from another 14,059 African-American women who participated in a large study called the Women’s Health Initiative.

The findings, which were published online in the Nov. 27, 2007, Journal of the National Cancer Institute, show that a woman’s age when she had her first child, which is a pretty good predictor of breast cancer risk in white women, is barely a factor in African-American women. “The only risk factor that had the same effect, quantitatively, was age of menarche,” says Gail, who was surprised by the differences.

The CARE model finds that African-American women who are 45 or older face a higher risk of breast cancer than the original Gail model predicts. “Roughly twice as many [of these] women might want to consider taking an intervention to prevent breast cancer as would have been recommended before,” says Gail.

The new risk calculator is a bittersweet development to Lucile Adams-Campbell, an epidemiologist and the director of the Howard University Cancer Center in Washington, D.C.  She says she’s happy the new model has been developed but that it points out the weakness of studies like the STAR trial, one of the largest breast cancer prevention clinical trials. “We should not have used the Gail model as a determinant of eligibility for African-American women,” she says.

Gail knows of no work in an advanced stage that recalibrates his model for other ethnic groups. “I think such models need to be developed and will be,” he says.