Understanding Mathematical Models for Breast Cancer Risk Assessment and Counseling

Abstract: Chemoprevention and prophylactic surgery are effective interventions for lowering breast cancer incidence. However, these approaches are associated with risks of their own. Accurate individualized breast cancer risk assessment is an essential component of the risk/benefit analysis that must take place prior to implementing either of these strategies. Several mathematical models for estimating individual breast cancer risk have been proposed over the last decade. The Gail model is the most generally applicable model; however, it neglects family history information in second-degree relatives, treats pre- and postmenopausal breast cancer the same, and ignores personal histories of lobular neoplasia. The Claus model is a better family history model, but it does not assign any special relevance to histories of bilateral breast cancer or ovarian cancer, and neglects all of the nonfamily history information accounted for by the Gail model. BRCAPRO is a Bayesian family history model that calculates individual breast cancer probabilities based on the probability that a family carries a mutation in one of the BRCA genes. Though its treatment of family history information is more thorough than the other models, it neglects the nonfamily history risk factors accounted for by the Gail model and may not appreciate familial clustering unrelated to BRCA gene mutation. A thorough understanding of the principles of risk analysis and the available mathematical models is essential for anyone wishing to perform intervention counseling. This review describes the basic components of risk analysis, explains how the mathematical models work and compares the strengths and weaknesses of the various models. CancerGene is a software tool for running all of these models. It may be obtained without charge at http://www.swmed.edu/home_pages/cancergene .     

Breast Cancer Risk Assessments Comparing Gail and CARE Models in African-American Women

Abstract:  The Gail model has been used to predict invasive breast cancer risk in women using risk factors of age, age at menarche, age at first live birth, number of first-degree relatives with breast cancer, and number of previous benign breast biopsies. However, this model underestimates breast cancer risk in African-American women. The Contraceptive and Reproductive Experience (CARE) model has been developed to replace the Gail model in predicting breast cancer risk in African-American women. In a sample of 883 women who participated in the breast cancer screening program at Howard University Cancer Center, we compared the breast cancer risk estimates from the Gail model and the CARE model. The mean 5-year breast cancer risk was 0.88% (Range: 0.18–6.60%) for the Gail model and 1.29% (Range: 0.20–4.50%) for the CARE model. Using the usual cutoff-point of 1.67% or above for elevated risk, there is a significant difference in the proportion of women with elevated breast cancer risk between the Gail and the CARE models (McNemar's test, p < 0.0001). For both models, there was a significant mean risk difference between those with and without a family history of breast cancer (Wilcoxon rank-sum test, p < 0.0001). Our results confirm the need for validation of the Gail model in African-Americans and diversity in research. Although these findings are not perfect and perhaps not definitive, they are additive in the discussions during counseling and risk assessment in African-Americans. Furthermore, these findings will be complemented by new technologies such as genomics in refining our ability to assess risk.     

The effectiveness of the Gail model in estimating risk for development of breast cancer in women under 40 years of age

Epidemiologic studies have provided information on risk factors for breast cancer. Gail and associates identified five risk factors using the Breast Cancer Detection Demonstration Project (BCDDP) population and developed a model to calculate a composite relative risk (RR). This model is commonly used to counsel women regarding their risk for breast cancer and was used by the National Surgical Adjuvant Breast Project (NSABP) for eligibility for the Breast Cancer Prevention Trial. Because the BCDDP population was composed almost entirely of women 40 years of age or older, our purpose was to evaluate the effectiveness of the Gail model in estimating the risk of breast cancer for women under 40 in the clinical setting. The Gail risk factors were assessed for 124 patients under the age of 40 treated for either ductal carcinoma in situ (DCIS) or invasive breast cancer at the Lahey Hitchcock Medical Center between 1983 and 1995. The RR was calculated using the Gail model. For comparison, two cohorts of women under the age of 40 were used: 107 randomly selected patients who underwent a breast biopsy because of a benign condition and 129 nurses from our institution who responded to a questionnaire that included reproductive and family history information as used in the Gail model. The RR calculated was the RR that existed at the time of the surgical consultation for a suspicious breast lesion. The Tarone-Ware method was used to analyze statistical significance of differences between distribution. Contingency tables were analyzed using Miettinen's modification of Fisher's exact test. No differences were found between the median RR for all groups. Only 2 of the 124 patients with breast cancer had a RR of 5 or more (the RR required to enter the Breast Cancer Prevention Trial). The distribution of age at menarche (AGEMEN) was the same for each group. No difference was found for the distribution of age at first live birth (AGEFLB) between those with breast cancer and those with a benign biopsy or the control group. The number of breast biopsies (NBIOPS) was higher in patients with a benign breast biopsy. No difference was found in the distribution of number of first-degree relatives with breast cancer (NUMREL). Overall the Gail model failed to differentiate those women about to have cancer diagnosed from two control populations. The Gail model is not useful in identifying immediate risk of breast cancer in women under 40 and should not be used for that purpose.     

Gail model risk factors: impact of adding an extended family history for breast cancer

Abstract:  An approach commonly used in estimating breast cancer risk is the Gail model. The objective of this study was to evaluate the feasibility and impact of adding extended family history as a new breast cancer risk factor into the Gail model. The data of the present study include cases with breast cancer and hospitalized controls recruited in the National Cancer Institute of Naples (southern Italy) between 1997 and 2000. We compared the first-degree relative (FDR) risk factor (standard Gail model) with the second-degree relative (SDR) information; and the FDR risk factor (standard Gail model) with the combination of FDR and SDR. We computed the c-statistic by comparing the risks found in our population to those in Gail-US population. The concordance for the model with FDR was 0.55 (95% CI 0.53–0.58), the model with SDR shows a modest but significant discriminatory accuracy (0.56, 95% CI 0.53–0.59), and the combination of FDR+SDR gave the concordance statistic of 0.57 (95% CI 0.54–0.60), indicating a good comparison between the two models. The results of our study show that extended family history information could be useful to improve the discriminatory power of the Gail model risk factors.     

Breast cancer risk assessment in patients who test negative for a hereditary cancer syndrome

BackgroundThe majority of women who undergo genetic testing due to a significant family history of breast cancer will receive a negative result. The purpose of this study was to calculate the lifetime risk of breast cancer in women undergoing genetic counseling who received an uninformative genetic test result.MethodsA retrospective chart review of mutation-negative women presenting to a cancer risk assessment clinic was performed. Lifetime risks of breast cancer were calculated using the Claus, Gail, and Tyrer-Cuzick risk assessment models.ResultsApproximately half (51%) of the women were classified as high-risk by at least one risk assessment model. The Tyrer-Cuzick model identified the highest proportion (43.2%) of patients as high-risk. Four percent (n = 4) of the sample was considered high-risk by all three models.ConclusionsMore than half (51%) of women who underwent genetic counseling and received an uninformative negative genetic test result had a significantly elevated risk for the development of breast cancer. It is, therefore, imperative that women do not conclude that a negative genetic test result represents a lack of risk.     

Cross‐sectional Study to Assess the Association of Population Density with Predicted Breast Cancer Risk

The Gail and CARE models estimate breast cancer risk for white and African‐American (AA) women, respectively. The aims of this study were to compare metropolitan and nonmetropolitan women with respect to predicted breast cancer risks based on known risk factors, and to determine if population density was an independent risk factor for breast cancer risk. A cross‐sectional survey was completed by 15,582 women between 35 and 85 years of age with no history of breast cancer. Metropolitan and nonmetropolitan women were compared with respect to risk factors, and breast cancer risk estimates, using general linear models adjusted for age. For both white and AA women, tisk factors used to estimate breast cancer risk included age at menarche, history of breast biopsies, and family history. For white women, age at first childbirth was an additional risk factor. In comparison to their nonmetropolitan counterparts, metropolitan white women were more likely to report having a breast biopsy, have family history of breast cancer, and delay childbirth. Among white metropolitan and nonmetropolitan women, mean estimated 5‐year risks were 1.44% and 1.32% (p < 0.001), and lifetime risks of breast cancer were 10.81% and 10.01% (p < 0.001), respectively. AA metropolitan residents were more likely than those from nonmetropolitan areas to have had a breast biopsy. Among AA metropolitan and nonmetropolitan women, mean estimated 5‐year risks were 1.16% and 1.12% (p = 0.039) and lifetime risks were 8.94%, and 8.85% (p = 0.344). Metropolitan residence was associated with higher predicted breast cancer risks for white women. Among AA women, metropolitan residence was associated with a higher predicted breast cancer risk at 5 years, but not over a lifetime. Population density was not an independent risk factor for breast cancer.     

Predictors of breast cancer development in a high-risk population

BACKGROUND: The purpose of this study was to investigate the strongest predictors of breast cancer in a high-risk population and to increase our understanding of the possible interactions between risk factors. METHODS: The Women At Risk High-Risk Registry provided the study population. The variables of interest included age at enrollment, presence of lobular carcinoma in situ (LCIS), atypical ductal hyperplasia (ADH), atypical lobular hyperplasia, family history of breast cancer, body mass index, and Gail scores (5-year high-risk > or =1.7%). Univariate and multivariate analyses were conducted with the Cox proportional hazards regression model and years of follow-up evaluation as the time scale. RESULTS: Out of 1553 high-risk women, 79 (5%) developed breast cancer during a median follow-up period of 5 years. Results from the multivariate Cox model demonstrated that FHBC (hazard ratio [HR] = 1.76; 95% confidence interval [CI], 1.05-2.97), ADH (HR = 1.90; 95% CI, 1.16-3.13), LCIS (HR = 1.71; 95% CI, .99-2.95), and a body mass index > or =30 (HR = 2.22; 95% CI, 1.14-4.35) were statistically significant predictors of breast cancer within this high-risk population. CONCLUSIONS: These results support current literature showing the synergistic increase in risk for patients with ADH, LCIS, and a positive family history of breast cancer. Obesity was also a strong predictor of breast cancer risk, which suggests that there may be a potentiating effect of obesity on other risk factors. Obesity may represent a modifiable risk factor, providing women with an opportunity to reduce their risk with lifestyle modification. Women with a strong family history of breast cancer or a diagnosis of ADH or LCIS may benefit most from risk-reduction strategies, chemoprevention, and surveillance.     

Diagnostic accuracy of the Gail model in the Black Women's Health Study

The Gail model is used to predict the risk of breast cancer in women of diverse race/ethnic groups for clinical trial protocols. However, this model has only been validated in US white women. Using a nested case-control study design, we evaluated the diagnostic accuracy of the original Gail model (GM) and that of the revised Gail model algorithm for blacks/African-Americans (GM-B) in the Black Women's Health Study (BWHS). Risk profiles were derived via a self reported questionnaire at the time of enrollment into the BWHS in 1995. Biennial questionnaires were obtained from the participants to determine the incident cases of breast cancer. The study of 725 breast cancer cases and 725 controls revealed that the 5-year risk of breast cancer based on the GM ranged from 0.2% to 15.4% among cases and 0.2% to 13.6% among the controls. Based on the GM-B, the 5-year risk of breast cancer ranged from 0.2% to 8.7% among cases and 0.2% to 7.2% among the controls. The sensitivities of the GM and GM-B model with the standard cutoff of 1.7% were 17.9% (95% CI: 15.9-19.9%) and 4.1% (95% CI: 3.0-5.2), respectively. Both the original and the modified version of the Gail model underestimate the risk of developing breast cancer in African-American women. More importantly, the modified Gail Model (GM-B) does a worse job at predicting the development of breast cancer for blacks than the original model (GM).     

Cohort study of women at risk for breast cancer and gross cystic disease

BACKGROUND: Gross cystic disease (GCD) is a common benign breast condition. Previous studies have reported conflicting results regarding the relationship of GCD and subsequent risk of developing breast cancer. This cohort study was conducted to investigate the association of GCD and breast cancer among women at high risk for breast cancer. METHODS: The Women At Risk Registry provided the study population. The variables of interest included age at enrollment, age at breast cancer diagnosis, body mass index (BMI), presence of lobular carcinoma in situ (LCIS), and Gail scores. Statistical significance was determined by calculating multivariable-adjusted rate ratios using Cox proportional hazards regression model with years of follow-up as the time scale. RESULTS: The study population included 1317 high-risk women, including 363 (28%) with GCD. The mean follow-up was 5.9 years for the GCD cohort, and 5.1 years for the non-GCD cohort (P < .001). The GCD and non-GCD groups differed by Gail score (P < .001), BMI (P < .01), presence of atypical hyperplasia (P < .001), presence of LCIS (P < .001), and family history of breast cancer (P < .001). Within the total population of 1317 women, 79 (6%) developed breast cancer; 28 (35%) out of the 79 had a prior history of GCD. Results from the Cox proportional hazards regression model showed a nonstatistically significant association of GCD and breast cancer (hazard ratio = 1.48, 95% confidence interval 0.88-2.51). The Kaplan-Meier overall survival estimate between the exposed and unexposed groups indicate that there are no differences in overall survival between the 2 groups (P = .5). CONCLUSIONS: These results do not support the contention that gross cystic disease is a significant risk factor for breast cancer.     

Mammographic screening of the high-risk woman

Annual screening mammography beginning at age 40 is recommended for the general population. For some women at high risk for developing breast cancer at a younger age, annual screening may be appropriate starting at an earlier age. These women include those with a personal history of breast cancer, nontherapeutic radiation to the breasts especially for Hodgkin's disease, BRCA positive women, women with a family history of a first-degree relative with breast cancer at a young age, and women with a biopsy diagnosis of lobular carcinoma in situ or atypical ductal hyperplasia. Women with a biopsy diagnosis of atypical lobular hyperplasia develop breast cancer after age 40 and do not need earlier screening, unless they have a family history of breast cancer. Although increasing a woman's risk for breast cancer, radial scar does not increase risk for women younger than 40 years old and therefore does not require screening at a young age.     

Health literacy and the perception of risk in a breast cancer family history clinic

Background

Informed consent is an essential component of medical practice, and especially so in procedural based specialties which entail varying degrees of risk. Breast cancer is one of the most common cancers in women, and as such is the focus of extensive research and significant media attention. Despite this, considerable misperception exists regarding the risk of developing breast cancer.

The Gail model predicts breast cancer in women with suspicious radiographic lesions

BACKGROUND: We sought to evaluate whether a woman's 5-year Gail risk adds to the predictive value of the Breast Imaging Reporting and Data System (BI-RADS) classification for the detection of breast cancer. METHODS: We performed a retrospective review of the BI-RADS classifications and pathology results for all image-guided needle breast biopsy examinations over a 3-year period at our institution. The 5-year Gail risk was calculated for eligible patients. Chi-square analysis was used to compare rates of malignancy based on Gail and BI-RADS scores. RESULTS: A total of 632 image-guided needle biopsy examinations were performed in 609 women. A total of 414 women had suspicious (BI-RADS 4) lesions and underwent 424 biopsy examinations. For this subset, women with a Gail risk of less than 1.7% had 21% malignant results, whereas those with a Gail risk of 1.7% or greater had 42% malignant results (relative risk, 1.94; 95% confidence interval, 1.45-2.66). CONCLUSIONS: The Gail model can stratify further the risk for breast cancer in women with suspicious breast imaging reports.     

Assessment of non-palpable mammographic abnormalities: comparison between screening and symptomatic clinics.

A retrospective study found that a breast screening clinic generated fewer localization biopsies for non-palpable mammographic abnormalities than a symptomatic clinic (3.36 versus 9.89 per 1000 mammograms, respectively) and that a greater proportion of such biopsies were malignant. This study determined the reason for this difference. There were 108 of 304 (35.5 per cent) and 17 of 130 (13.1 per cent) carcinomas in women attending the screening and breast clinics respectively (relative risk 2.72 (95 per cent confidence interval 1.70-4.34)). This difference was regardless of age. The characteristics of the mammographic abnormality, the Wolfe pattern, a family history of breast carcinoma, parity and age at first pregnancy were similar in both groups. Women attending the screening clinic were referred for localization biopsy after assessment by clinicians and radiologists at a joint clinic; there was no joint assessment for patients attending the breast clinic. The same staff attended both clinics, although the proportion of time spent at each varied. This study suggests that all women with a non-palpable mammographic abnormality should be reviewed at a joint assessment clinic before localization biopsy is recommended.     

Management of women at increased risk for breast cancer secondary to high‐risk proliferative lesions and family history of the disease

Women with breast biopsies showing high‐risk proliferative lesions such as atypical hyperplasia (AH) and lobular carcinoma in situ (LCIS) have an increased risk of developing breast cancer. Other factors including age, family history of breast cancer, and extent of AH may play a role in increasing breast cancer risk. In addition to women with AH, there is a subset of women with a positive family history of breast cancer, without a known germline mutation, which places them also at an increased risk for breast cancer. Clinical management, screening, chemoprevention, and surgical risk‐reduction are discussed in this review to inform the management of these high‐risk women. Advanced imaging technology, pharmacologic research into different targets, and innovations in breast reconstruction are changing the way in which patients are counseled of their individual risk.     

No useful role for fine needle aspiration as a marker for familial breast cancer

A study of a single fine needle aspiration taken from the upper outer quadrant of 228 asymptomatic women attending a family history clinic has been performed. No abnormalities were detected. This is not a useful screening tool in asymptomatic women at risk of breast cancer.     

Surgical management of the patient at high risk for breast cancer

The most controversial aspect of breast disease centres around the management of patients who have either a strong family history of breast cancer or a biopsy diagnosis of lobular carcinoma in situ or ductal carcinoma in situ. The current alternatives for patients who have two or more relatives with breast cancer consist of close follow-up or prophylactic total mastectomies and reconstruction. Invasive breast cancer in patients with lobular carcinoma in situ may occur in either breast and may be as high as 30% at 20 to 30 years. In these women it is reasonable to do a wide excision of the lobular carcinoma; in those without a family history, close follow-up is adequate. Intraductal cancer treated by biopsy only is associated with a 40% risk of cancer in the ipsilateral breast. Therefore, the usual management is total mastectomy. However, the information to support this therapy over a segmental resection has limited scientific validity. Because the cosmetic appearance after total mastectomy and reconstruction is not as good as that of the normal breast, this procedure must be employed cautiously and only with the total support of the patient and her husband or close family. Subcutaneous mastectomy for prophylaxis leaves behind macroscopic glandular tissue and, therefore, is not considered by many to be optimal management. A total mastectomy, preserving the skin and resecting all macroscopic breast tissue and nipple, is the treatment of choice if the procedure is deemed appropriate.     

Breast Cancer Chemoprevention among High‐risk Women and those with Ductal Carcinoma In Situ

Chemoprevention with the anti‐estrogens, tamoxifen, raloxifene, and aromatase inhibitors, reduce breast cancer incidence in high‐risk women; however, uptake has been poor (<5%) in the prevention setting. We assessed use of anti‐estrogens for breast cancer prevention, among high‐risk women seen at an academic breast center, to observe how uptake rates compare in this setting. We collected data on demographics, breast cancer risk factors, and health behaviors via self‐administered questionnaires and medical chart abstraction. Women eligible for chemoprevention with anti‐estrogens had a 5‐year predicted breast cancer risk according to the Gail model of ≥1.67%, history of lobular or ductal carcinoma in situ (LCIS/DCIS), and/or BRCA mutation. We dichotomized anti‐estrogen use as ever or never. Predictors of use were evaluated using multivariable log‐binomial regression. Of 412 high‐risk women enrolled, 316 (77%) were eligible for chemoprevention. Among eligible women, 55% were non‐Hispanic white, 29% Hispanic, 8% non‐Hispanic black, and 7% Asian. Women were grouped based upon their highest category of breast cancer risk (in descending order): BRCA mutation carriers (3%), DCIS (40%), LCIS (22%), and 5‐year Gail risk ≥1.67% (36%). Among those eligible for chemoprevention, 162 (51%) had ever initiated anti‐estrogen therapy (71% tamoxifen, 23% raloxifene, 5% aromatase inhibitor). Anti‐estrogen use was highest among women with DCIS (73%). In multivariable analysis, women with a 5‐year Gail risk ≥1.67% had approximately a 20% lower likelihood of anti‐estrogen use compared to women with DCIS (p = 0.01). In the primary prevention setting, excluding women diagnosed with DCIS, anti‐estrogen use was 37%. Multivariable analysis showed differences in uptake by education and potentially by race/ethnicity. Among high‐risk women seen at a breast center, anti‐estrogen use for chemoprevention was relatively high as compared to the published literature. Clinicians can support high‐risk women by effectively communicating breast cancer risk and enhancing knowledge about the risks and benefits of chemoprevention.     

A simple tool for identifying unaffected women at a moderately increased or potentially high risk of breast cancer based on their family history

Family history of breast cancer is a more important aspect of clinical management following the discovery of the genes BRCA1 and BRCA2. The authors developed a short questionnaire to categorize risk according to breast cancer history in close relatives. It was completed by 559 women attending for screening mammograms in Sydney, Australia. Twenty-three per cent reported a family history sufficient to be classified at a moderately increased or potentially high risk according to national guidelines (category II or III). Only 29 women (5%) made errors such that their risk category could not be determined. Validation of responses from 89 women, 44 from category II or III, found 100% agreement with classification after interview by a genetic counsellor. This questionnaire has the potential to accurately triage risk based on a woman's knowledge of her family history, and could be used in a variety of settings to identify women who may require further assessment, management and referral advice.