Breast cancer risk calculator (based on the Gail model)

What the Gail model predicts

The model estimates the chance that a woman of a given age and set of risk factors will develop breast cancer over a set number of years (Gail 1989). It uses four risk factors besides age: age at first period, age at first live birth, number of past breast biopsies and number of first-degree relatives (mother, sisters, daughters) with breast cancer (Gail 1989). A finding of atypical hyperplasia on a biopsy raises the estimate further (NCI DCEG).

The version used here is the one that projects invasive breast cancer only, not DCIS. In 5,969 White women 35 and older from the placebo arm of the Breast Cancer Prevention Trial, it predicted almost exactly the number of invasive cancers that occurred (expected to observed ratio 1.03) (Costantino 1999).

How it is calculated

The model first works out your relative risk compared with a woman of your age with none of the risk factors. Biopsies count for less after age 50, so there is one relative risk before 50 and one from 50 on (Gail 1989). It then combines that with the breast cancer rate in women with no identified risk factors, which is the overall rate times 1 minus the attributable risk (Gail 1989). Deaths from other causes are taken into account, so the estimate is your chance of being diagnosed with breast cancer, not just a rate (Gail 1989).

All the numbers come from the National Cancer Institute's open BCRA software: the age-specific breast cancer and death rates for each group, the attributable risks and the relative risk coefficients (NCI DCEG). That software was written by US government employees, and its release does not mean the institute endorses this site (NCI DCEG). Our tests reproduce, to nine decimal places, every non-Hispanic example result printed in the software's own worked example (NCI DCEG).

A few answers are grouped the way the model groups them. For Black women, age at first birth is not part of the model and a first period at 11 or younger counts the same as 12 or 13; for Asian and Pacific Islander women, two or more affected relatives count the same as one (NCI DCEG). "Don't know" answers are scored as the lowest-risk choice, as the software does (NCI DCEG).

Versions for different groups

• White women: the original model, from the Breast Cancer Detection Demonstration Project, a screening study of 280,000 White women aged 35 to 74 (NCI).
• Black or African American women: built from 1,607 women with breast cancer and 1,647 without in the CARE study. In Women's Health Initiative data it matched observed cancers well, except that it underestimated risk in women who had had breast biopsies (Gail 2007).
• Asian and Pacific Islander American women: built from the Asian American Breast Cancer Study (589 women with breast cancer, 952 without), with separate rates for Chinese, Japanese, Filipina, Native Hawaiian, other Pacific Islander and other Asian women. In Women's Health Initiative data it underestimated risk by about 17% (Matsuno 2011).
• American Indian or Alaska Native women, or unknown background: the software uses the White rates and coefficients (NCI DCEG).
• Hispanic women: a version exists, built from the San Francisco Bay Area Breast Cancer Study (NCI). We could not reproduce the software's own printed examples for Hispanic women, so we do not offer it. Your clinician can run it for you.

What to do with the number

The US Preventive Services Task Force says women whose 5-year risk is 3% or higher are likely to gain more than they lose from medicines that lower breast cancer risk; it does not endorse one risk tool (USPSTF 2019). Our breast cancer prevention medicine check explains that recommendation. A higher result is a reason to talk with your clinician, not a reason to panic: a 5-year risk of 3% still means about 97 in 100 women like you would not be diagnosed in those 5 years.

If your family history includes breast cancer before 50, ovarian cancer, cancer in both breasts or breast cancer in a man, the BRCA family history check is the better first step: general tools like this one are not designed to find inherited risk. For when to start mammograms, see the breast cancer screening check.

Limits

• The tool is good at predicting how many women in a group will get breast cancer, but only modest at telling which individual women will (USPSTF 2019).
• In validation it tended to overestimate risk for women in the highest risk groups and underestimate it for women in the lowest (Costantino 1999).
• It may underestimate risk in Black women with past biopsies and in Hispanic women born outside the US (NCI).
• The original data came from women who had a breast exam about once a year, so the estimates fit women in regular screening best (Gail 1989).
• It does not use breast density, a detailed family tree, the father's side of the family, or genetic test results.
• It cannot give an accurate risk for women who have had breast cancer, DCIS or LCIS (NCI).

General safety advice (not part of the source studies): a new lump, skin dimpling, nipple changes or discharge needs a check with a doctor, whatever this calculator says.

Related tools: the breast cancer screening check, the BRCA family history check, the breast cancer prevention medicine check, the ovarian cancer screening check and all screening tests by age.