Breast Cancer Risk Assessment Models - How Do They Work and What Do They Do?

The various mathematical models that have been developed for cancer risk assessment are practically indispensable in today’s cancer clinics, both for calculating an individual’s lifetime risks for cancer or their likelihood of a hereditary cancer predisposition syndrome as well as helping clinicians to measure the risks vs. benefits when determining optimal screening and prophylactic interventions. But with so many models available to clinicians now, there are also just as many questions: which model(s) should I use? Is there a best model? How do they work?

We try to answer some of those questions for you below with a brief review of the hereditary breast risk models utilized within CancerGene Connect’s programs.


GAIL

What it tells you: The patient’s 5 year and lifetime risk (up to age 90) to develop breast cancer.

What it takes into account: Mostly non-genetic risk factors including: age, hormonal history (age at menarche, age at the time of birth of first child (or nulliparity)), number of past breast biopsies, number of breast biopsies showing atypical hyperplasia, race/ethnicity. Family history (limited).

What it leaves out: The Gail model will only take into account first degree relatives when considering family history of cancer and does not take age of onset of cancer into account. It does not take into account paternal family history. It treats pre- and post-menopausal breast cancers the same. It takes into account atypical hyperplasia, but ignores lobular carcinoma in situ (LCIS).

The original Gail model is based on a dataset compiled from the Breast Cancer Detection Demonstration Project (BCDDP) and is limited by the fact that it is older data collected (between 1973 and 1980) from a Caucasian-only population. The BCDDP was collected from women who were compliant in obtaining their annual mammogram and so may not be representative in non-compliant patient groups. An update to the model utilizing data obtained from the Contraceptive and Reproductive Experiences (CARE) study includes data from African American women. The updated dataset now also represents some Asian American and Pacific Islander women using data from the Asian American Breast Cancer Study (AABCS). However, it is important to note that while the model does age-adjust cancer rates for Caucasians and African Americans, women of any other ethnic background would be run as Caucasian. The version of Gail run by CancerGene Connect utilizes baseline breast cancer rates calculated from the 1992-1994 SEER data and probabilities of dying from non-breast cancer causes from the 1995 U.S. mortality data. [1]

CLAUS

What it tells you: The patient’s 5 year and lifetime risk to develop breast cancer. Remaining risk based on age is also calculated and displayed at 5-year increments using interpolation.

What it takes into account: Family history of breast cancer and ages of cancer diagnoses. The original model developed in 1991 only took into account family history of breast cancer; later versions took into account ovarian cancer data. An expanded Claus model developed in 2004 also take into account bilateral breast cancer and risks for three or more affected relatives. [2]

What it leaves out: Nonhereditary risk factors are not considered in the Claus model. Lifetime risk tables for some combinations of affected relatives are not given (such as mother and maternal grandmother) but can be approximated by using mother and maternal aunt combinations.

The Claus model estimates breast cancer risk based on which relatives are affected with breast cancer and the ages that they were diagnosed. The data used to develop this model is from Caucasian women only collected in the 1980s when there was a lower overall breast cancer incidence rate. There is often a discrepancy between the published risk tables and computerized Claus models; the tables do not make an adjustment for unaffected relatives while the computerized versions are able to account for the reduced likelihood for an autosomal dominant gene with an increasing number of unaffected women. The CancerGene Connect software calculates every possible combination of relatives and displays the Claus table that provides the greatest risk estimate for the patient. [1]

BRCAPRO

What it tells you: Probability that the patient carries a mutation in the BRCA1 or BRCA2 gene. The patient’s 5 year and lifetime risks for breast and ovarian cancer based on the calculated probability that they carry a BRCA1 or BRCA2 gene mutation.

What it takes into account: Family history (both affected and unaffected relatives, breast and ovarian cancer history, male breast cancer, and bilateral breast cancer) up to second degree relatives, breast pathology, and oophorectomy status. The BRCAPro model also considers ethnicity for each family member to allow for differing mutation allele frequencies in families of mixed ethnicity.

What it leaves out: Nonhereditary risk factors. Noninvasive breast cancer.

The BRCAPro model uses Bayes Theorem and incorporates various facets of both the maternal and paternal family history including ages of family members, both affected and unaffected family members, and published data on BRCA1 and BRCA2 mutation frequencies and cancer penetrance in mutation carriers to calculate the patient’s breast and ovarian cancer probabilities based on the probability that the family has a mutation in either the BRCA1 or BRCA2 gene.

TYRER-CUZICK

What it tells you: The Tyrer-Cuzick model calculates both the likelihood that the patient is carrying a hereditary breast cancer predisposition gene as well as the likelihood that the patient will develop breast cancer (10 year and lifetime risks) based on the probability of carrying the aforementioned gene mutation.

What it takes into account: Family history (affected and unaffected relatives, age at diagnosis, male breast cancer, half-siblings, cousins, and nieces), hormonal history (age at menarche, age at first birth, parity, age at menopause, hormone replacement therapy), benign breast disease, body mass index, age, and genetic factors.

What it leaves out: The Tyrer-Cuzick model does not leave much out which, in a way, can be a limitation in certain circumstances. It does not take into account the full family history of all unaffected family members. Tyrer-Cuzick has been found to overestimate risks in women with breast atypia. [3]

The Tyrer-Cuzick model uses a dataset partly derived from the International Breast Intervention Study (IBIS). The model allows for the possibility of multiple breast cancer predisposition genes of differing penetrance; not just BRCA1 and BRCA2.

BOADICEA

What it tells you: The patient’s lifetime risk of breast and ovarian cancer. The probability of a mutation in BRCA1 or BRCA2

What it takes into account:  Family history (including prostate and pancreatic cancers, age of diagnosis, affected and unaffected relatives), breast cancer pathology, and genetic factors.

What it leaves out: Nonhereditary risk factors.

The Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA) models the simultaneous effects of BRCA1 and BRCA2 mutations and assumes that the residual familial clustering of breast cancer is explained by a polygenic component. [4] BOADICEA does require dates of birth for everyone entered into the pedigree, which may not be common practice for many breast centers in the US. It does not take into account prophylactic surgery in the family history. The dataset used to develop BOADICEA was developed in the UK and so may not be entirely representative of the US population or other ethnicities.


It is important to remember when considering each model that no model is perfect and that there is no “one size fits all” model for every unique clinical situation. Oftentimes, you will want to consider multiple models that will each account for (and may leave out) various risk factors for your patient and their unique clinical and family history.  It is important to understand what each model takes into account – and equally as important to understand what each model does NOT take into account – when incorporating the models into your cancer risk assessment.

-Megan Frone, MS, CGC
Genetics Specialist, CancerGene Connect

 

CITATIONS

[1]   Euhus, D. Understanding mathematical models for breast cancer risk assessment and counseling. The Breast Journal. 2001, 7(4): 224-232

[2]  Claus, E., Risch, N., Thompson. W.D. The calculation of breast cancer risk for women with a first degree history of ovarian cancer. Breast Cancer Research and Treatment. 1993, 28: 115-120

[3]  Boughey, J., et al. Evaluation of the Tyrer-Cuzick (International Breast Cancer Intervention Study) model for breast cancer risk prediction in women with atypical hyperplasia. J Clin Oncol. 2010, 28(22): 3591-3596

[4]   BOADICEA. BWA v3. University of Cambridge Department of Public Health & Primary Care Centre for Cancer Genetic Epidemology. Web. http://ccge.medschl.cam.ac.uk/boadicea/. 1 Feb 2016.