Assessment of the accuracy of the Gail model in women with atypical hyperplasia

Recalibration of the Gail model for predicting invasive breast cancer risk in Spanish women: a population-based cohort study

The Gail model for predicting the absolute risk of invasive breast cancer has been validated extensively in US populations, but its performance in the international setting remains uncertain. We evaluated the predictive accuracy of the Gail model in 54,649 Spanish women aged 45-68 years who were free of breast cancer at the 1996-1998 baseline mammographic examination in the population-based Navarre Breast Cancer Screening Program. Incident cases of invasive breast cancer and competing deaths were ascertained until the end of 2005 (average follow-up of 7.7 years) through linkage with population-based cancer and mortality registries. The Gail model was tested for calibration and discrimination in its original form and after recalibration to the lower breast cancer incidence and risk factor prevalence in the study cohort, and compared through cross-validation with a Navarre model fully developed from this cohort. The original Gail model overpredicted significantly the 835 cases of invasive breast cancer observed in the cohort (ratio of expected to observed cases 1.46, 95 % CI 1.36-1.56). The recalibrated Gail model was well calibrated overall (expected-to-observed ratio 1.00, 95 % CI 0.94-1.07), but it tended to underestimate risk for women in low-risk quintiles and to overestimate risk in high-risk quintiles (P = 0.01). The Navarre model showed good cross-validated calibration overall (expected-to-observed ratio 0.98, 95 % CI 0.92-1.05) and in different cohort subsets. The Navarre and Gail models had modest cross-validated discrimination indexes of 0.542 (95 % CI 0.521-0.564) and 0.544 (95 % CI 0.523-0.565), respectively. Although the original Gail model cannot be applied directly to populations with different underlying rates of invasive breast cancer, it can readily be recalibrated to provide unbiased estimates of absolute risk in such populations. Nevertheless, its limited discrimination ability at the individual level highlights the need to develop extended models with additional strong risk factors.

Personalized medicine and access to health care: potential for inequitable access?

Personalized medicine promises that an individual's genetic information will be increasingly used to prioritize access to health care. Use of genetic information to inform medical decision making, however, raises questions as to whether such use could be inequitable. Using breast cancer genetic risk prediction models as an example, on the surface clinical use of genetic information is consistent with the tools provided by evidence-based medicine, representing a means to equitably distribute limited health-care resources. However, at present, given limitations inherent to the tools themselves, and the mechanisms surrounding their implementation, it becomes clear that reliance on an individual's genetic information as part of medical decision making could serve as a vehicle through which disparities are perpetuated under public and private health-care delivery models. The potential for inequities arising from using genetic information to determine access to health care has been rarely discussed. Yet, it raises legal and ethical questions distinct from those raised surrounding genetic discrimination in employment or access to private insurance. Given the increasing role personalized medicine is forecast to play in the provision of health care, addressing a broader view of what constitutes genetic discrimination, one that occurs along a continuum and includes inequitable access, will be needed during the implementation of new applications based on individual genetic profiles. Only by anticipating and addressing the potential for inequitable access to health care occurring from using genetic information will we move closer to realizing the goal of personalized medicine: to improve the health of individuals.

Breast cancer risk assessment across the risk continuum: genetic and nongenetic risk factors contributing to differential model performance

Introduction

Clinicians use different breast cancer risk models for patients considered at average and above-average risk, based largely on their family histories and genetic factors. We used longitudinal cohort data from women whose breast cancer risks span the full spectrum to determine the genetic and nongenetic covariates that differentiate the performance of two commonly used models that include nongenetic factors - BCRAT, also called Gail model, generally used for patients with average risk and IBIS, also called Tyrer Cuzick model, generally used for patients with above-average risk.

Methods

We evaluated the performance of the BCRAT and IBIS models as currently applied in clinical settings for 10-year absolute risk of breast cancer, using prospective data from 1,857 women over a mean follow-up length of 8.1 years, of whom 83 developed cancer. This cohort spans the continuum of breast cancer risk, with some subjects at lower than average population risk. Therefore, the wide variation in individual risk makes it an interesting population to examine model performance across subgroups of women. For model calibration, we divided the cohort into quartiles of model-assigned risk and compared differences between assigned and observed risks using the Hosmer-Lemeshow (HL) chi-squared statistic. For model discrimination, we computed the area under the receiver operator curve (AUC) and the case risk percentiles (CRPs).

Results

The 10-year risks assigned by BCRAT and IBIS differed (range of difference 0.001 to 79.5). The mean BCRAT- and IBIS-assigned risks of 3.18% and 5.49%, respectively, were lower than the cohort's 10-year cumulative probability of developing breast cancer (6.25%; 95% confidence interval (CI) = 5.0 to 7.8%). Agreement between assigned and observed risks was better for IBIS (HL X₄² = 7.2, P value 0.13) than BCRAT (HL X₄² = 22.0, P value <0.001). The IBIS model also showed better discrimination (AUC = 69.5%, CI = 63.8% to 75.2%) than did the BCRAT model (AUC = 63.2%, CI = 57.6% to 68.9%). In almost all covariate-specific subgroups, BCRAT mean risks were significantly lower than the observed risks, while IBIS risks showed generally good agreement with observed risks, even in the subgroups of women considered at average risk (for example, no family history of breast cancer, BRCA1/2 mutation negative).

Conclusions

Models developed using extended family history and genetic data, such as the IBIS model, also perform well in women considered at average risk (for example, no family history of breast cancer, BRCA1/2 mutation negative). Extending such models to include additional nongenetic information may improve performance in women across the breast cancer risk continuum.

The role of chemoprevention in modifying the risk of breast cancer in women with atypical breast lesions

Women with atypical ductal hyperplasia (ADH), atypical lobular hyperplasia (ALH), lobular carcinoma in situ (LCIS), and severe ADH are at increased risk of breast cancer, but a systematic quantification of this risk and the efficacy of chemoprevention in the clinical setting is still lacking. The objective of this study is to evaluate a woman's risk of breast cancer based on atypia type and to determine the effect of chemoprevention in decreasing this risk. Review of 76,333 breast pathology reports from three institutions within Partners Healthcare System, Boston, from 1987 to 2010 using natural language processing was carried out. This approach identified 2,938 women diagnosed with atypical breast lesions. The main outcome of this study is breast cancer occurrence. Of the 2,938 patients with atypical breast lesions, 1,658 were documented to have received no chemoprevention, and 184/1,658 (11.1 %) developed breast cancer at a mean follow-up of 68 months. Estimated 10-year cancer risks were 17.3 % with ADH, 20.7 % with ALH, 23.7 % with LCIS, and 26.0 % with severe ADH. In a subset of patients treated from 1999 on (the chemoprevention era), those who received no chemoprevention had an estimated 10-year breast cancer risk of 21.3 %, whereas those treated with chemoprevention had a 10-year risk of 7.5 % (p < 0.001). Chemoprevention use significantly reduced breast cancer risk for all atypia types (p < 0.05). The risk of breast cancer with atypical breast lesions is substantial. Physicians should counsel patients with ADH, ALH, LCIS, and severe ADH about the benefit of chemoprevention in decreasing their breast cancer risk.

Evaluation of serum estrogen-DNA adducts as potential biomarkers for breast cancer risk

This study was conducted to determine whether the ratio of estrogen-DNA adducts to their respective metabolites and conjugates in serum differed between women with early-onset breast cancer and those with average or high risk of developing breast cancer. Serum samples from women at average risk (n=63) or high risk (n=80) for breast cancer (using Gail model) and women newly diagnosed with early breast cancer (n=79) were analyzed using UPLC-MS/MS. Adduct ratios were statistically compared among the three groups, and the Area Under the Receiver Operating Characteristic Curve (AUC) was used to identify a diagnostic cut-off point. The median adduct ratio in the average-risk group was significantly lower than that of both the high-risk group and the breast cancer group (p values<0.0001), and provided good discrimination between those at average versus high risk of breast cancer (AUC=0.84, 95% CI 0.77-0.90). Sensitivity and specificity were maximized at an adduct ratio of 77. For women in the same age and BMI group, the odds of being at high risk for breast cancer was 8.03 (95% CI 3.46-18.7) times higher for those with a ratio of at least 77 compared to those with a ratio less than 77. The likelihood of being at high risk for breast cancer was significantly increased for those with a high adduct ratio relative to those with a low adduct ratio. These findings suggest that estrogen-DNA adducts deserve further study as potential biomarkers for risk of developing breast cancer.

p16(INK4a) expression and breast cancer risk in women with atypical hyperplasia

p16, a nuclear protein encoded by the p16(INK4a) gene, is a regulator of cell-cycle regulation. Previous studies have shown that expression of p16 in tissue biopsies of patients with ductal carcinoma in situ (DCIS) is associated with increased risk of breast cancer, particularly when considered in combination with other markers such as Ki-67 and COX-2. Here, we evaluated how expression of p16 in breast tissue biopsies of women with atypical hyperplasia (AH), a putative precursor lesion to DCIS, is associated with subsequent development of cancer. p16 expression was assessed by immunohistochemistry in archival sections from 233 women with AH diagnosed at the Mayo Clinic. p16 expression in the atypical lesions was scored by percentage of positive cells and intensity of staining. We also studied coexpression of p16, with Ki-67 and COX-2, biomarkers of progression in AH. Risk factor and follow-up data were obtained via study questionnaire and medical records. Forty-seven patients (20%) developed breast cancer with a median follow-up of 14.5 years. Staining of p16 was increased in older patients relative to younger patients (P = 0.0025). Although risk of developing breast cancer was not associated with increased p16 expression, joint overexpression of Ki-67 and COX-2 was found to convey stronger risk of breast cancer in the first 10 years after diagnosis as compared with one negative marker (P < 0.01). However, the addition of p16 levels did not strengthen this association. p16 overexpression, either alone or in combination with COX-2 and Ki-67, does not significantly stratify breast cancer risk in women with AH.

Performing breast cancer risk assessments in a community setting

This article describes the implementation of a risk assessment program for women having screening mammography at a community center. The program used the National Cancer Institute's Breast Cancer Risk Assessment Tool to raise awareness in high-risk women. An evidence-based process is essential when implementing changes in clinical practice to overcome challenges and barriers.

Breast cancer risk assessment in women aged 70 and older

Although the benefit of screening mammography for women over 69 has not been established, it is generally agreed that screening recommendations for older women should be individualized based on health status and breast cancer risk. However, statistical models to assess breast cancer risk have not been previously evaluated in this age group. In this study, the original Gail model and three more recent models that include mammographic breast density as a risk factor were applied to a cohort of 19,779 Vermont women aged 70 and older. Women were followed for an average of 7.1 years and 821 developed breast cancer. The predictive accuracy of each risk model was measured by its c-statistic and associations between individual risk factors and breast cancer risk were assessed by Cox regression. C-statistics were 0.54 (95% CI = 0.52-0.56) for the Gail model, 0.54 (95% CI = 0.51-0.56) for the Tice modification of the Gail model, 0.55 (95% CI = 0.53-0.58) for a model developed by Barlow and 0.55 (95% CI = 0.53-0.58) for a Vermont model. These results indicate that the models are not useful for assessing risk in women aged 70 and older. Several risk factors in the models were not significantly associated with outcome in the cohort, while others were significantly related to outcome but had smaller relative risks than estimated by the models. Age-related attenuation of the effects of some risk factors makes the prediction of breast cancer in older women particularly difficult.

Independent association of lobular involution and mammographic breast density with breast cancer risk

Background

Lobular involution, or age-related atrophy of breast lobules, is inversely associated with breast cancer risk, and mammographic breast density (MBD) is positively associated with breast cancer risk.

Methods

To evaluate whether lobular involution and MBD are independently associated with breast cancer risk in women with benign breast disease, we performed a nested cohort study among women (n = 2666) with benign breast disease diagnosed at Mayo Clinic between January 1, 1985, and December 31, 1991 and a mammogram available within 6 months of the diagnosis. Women were followed up for an average of 13.3 years to document any breast cancer incidence. Lobular involution was categorized as none, partial, or complete; parenchymal pattern was classified using the Wolfe classification as N1 (nondense), P1, P2 (ductal prominence occupying <25%, or >25% of the breast, respectively), or DY (extremely dense). Hazard ratios (HRs) and 95% confidence intervals (CIs) to assess associations of lobular involution and MBD with breast cancer risk were estimated using adjusted Cox proportional hazards model. All tests of statistical significance were two-sided.

Results

After adjustment for MBD, having no or partial lobular involution was associated with a higher risk of breast cancer than having complete involution (none: HR of breast cancer incidence = 2.62, 95% CI = 1.39 to 4.94; partial: HR of breast cancer incidence = 1.61, 95% CI = 1.03 to 2.53; P_trend = .002). Similarly, after adjustment for involution, having dense breasts was associated with higher risk of breast cancer than having nondense breasts (for DY: HR of breast cancer incidence = 1.67, 95% CI = 1.03 to 2.73; for P2: HR of breast cancer incidence = 1.96, 95% CI = 1.20 to 3.21; for P1: HR of breast cancer incidence = 1.23, 95% CI = 0.67 to 2.26; P_trend = .02). Having a combination of no involution and dense breasts was associated with higher risk of breast cancer than having complete involution and nondense breasts (HR of breast cancer incidence = 4.08, 95% CI = 1.72 to 9.68; P = .006).

Conclusion

Lobular involution and MBD are independently associated with breast cancer incidence; combined, they are associated with an even greater risk for breast cancer.

Assessing women at high risk of breast cancer: a review of risk assessment models

Women who are at high risk of breast cancer can be offered more intensive surveillance or prophylactic measures, such as surgery or chemoprevention. Central to decisions regarding the level of prevention is accurate and individualized risk assessment. This review aims to distill the diverse literature and provide practicing clinicians with an overview of the available risk assessment methods. Risk assessments fall into two groups: the risk of carrying a mutation in a high-risk gene such as BRCA1 or BRCA2 and the risk of developing breast cancer with or without such a mutation. Knowledge of breast cancer risks, taken together with the risks and benefits of the intervention, is needed to choose an appropriate disease management strategy. A number of models have been developed for assessing these risks, but independent validation of such models has produced variable results. Some models are able to predict both mutation carriage risks and breast cancer risk; however, to date, all are limited by only moderate discriminatory accuracy. Further improvements in the knowledge of how to best integrate both new risk factors and newly discovered genetic variants into these models will allow clinicians to more accurately determine which women are most likely to develop breast cancer. These steady and incremental improvements in models will need to undergo revalidation.

Effect of changing breast cancer incidence rates on the calibration of the Gail model

PURPOSE

The Gail model combines relative risks (RRs) for five breast cancer risk factors with age-specific breast cancer incidence rates and competing mortality rates from the Surveillance, Epidemiology, and End Results (SEER) program from 1983 to 1987 to predict risk of invasive breast cancer over a given time period. Motivated by changes in breast cancer incidence during the 1990s, we evaluated the model's calibration in two recent cohorts.

METHODS

We included white, postmenopausal women from the National Institutes of Health (NIH) -AARP Diet and Health Study (NIH-AARP, 1995 to 2003), and the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO, 1993 to 2006). Calibration was assessed by comparing the number of breast cancers expected from the Gail model with that observed. We then evaluated calibration by using an updated model that combined Gail model RRs with 1995 to 2003 SEER invasive breast cancer incidence rates.

RESULTS

Overall, the Gail model significantly underpredicted the number of invasive breast cancers in NIH-AARP, with an expected-to-observed ratio of 0.87 (95% CI, 0.85 to 0.89), and in PLCO, with an expected-to-observed ratio of 0.86 (95% CI, 0.82 to 0.90). The updated model was well-calibrated overall, with an expected-to-observed ratio of 1.03 (95% CI, 1.00 to 1.05) in NIH-AARP and an expected-to-observed ratio of 1.01 (95% CI: 0.97 to 1.06) in PLCO. Of women age 50 to 55 years at baseline, 13% to 14% had a projected Gail model 5-year risk lower than the recommended threshold of 1.66% for use of tamoxifen or raloxifene but >or= 1.66% when using the updated model. The Gail model was well calibrated in PLCO when the prediction period was restricted to 2003 to 2006.

CONCLUSION

This study highlights that model calibration is important to ensure the usefulness of risk prediction models for clinical decision making.

Novel breast tissue feature strongly associated with risk of breast cancer

PURPOSE

Accurate, individualized risk prediction for breast cancer is lacking. Tissue-based features may help to stratify women into different risk levels. Breast lobules are the anatomic sites of origin of breast cancer. As women age, these lobular structures should regress, which results in reduced breast cancer risk. However, this does not occur in all women.

METHODS

We have quantified the extent of lobule regression on a benign breast biopsy in 85 patients who developed breast cancer and 142 age-matched controls from the Mayo Benign Breast Disease Cohort, by determining number of acini per lobule and lobular area. We also calculated Gail model 5-year predicted risks for these women.

RESULTS

There is a step-wise increase in breast cancer risk with increasing numbers of acini per lobule (P = .0004). Adjusting for Gail model score, parity, histology, and family history did not attenuate this association. Lobular area was similarly associated with risk. The Gail model estimates were associated with risk of breast cancer (P = .03). We examined the individual accuracy of these measures using the concordance (c) statistic. The Gail model c statistic was 0.60 (95% CI, 0.50 to 0.70); the acinar count c statistic was 0.65 (95% CI, 0.54 to 0.75). Combining acinar count and lobular area, the c statistic was 0.68 (95% CI, 0.58 to 0.78). Adding the Gail model to these measures did not improve the c statistic.

CONCLUSION

Novel, tissue-based features that reflect the status of a woman's normal breast lobules are associated with breast cancer risk. These features may offer a novel strategy for risk prediction.

Ki67: a time-varying biomarker of risk of breast cancer in atypical hyperplasia

Uncontrolled proliferation is a defining feature of the malignant phenotype. Ki67 is a marker for proliferating cells and is overexpressed in many breast cancers. Atypical hyperplasia is a premalignant lesion of the breast (relative risk approximately 4.0). Here, we asked if Ki67 expression could stratify risk in women with atypia. Ki67 expression was assessed immunohistochemically by digital image analysis in archival sections from 192 women with atypia diagnosed at the Mayo Clinic 1/1/67-12/31/91. Risk factor and follow-up data were obtained via study questionnaire and medical records. Observed breast cancer events were compared to population expected rates (Iowa SEER) using standardized incidence ratios (SIRs). We examined two endpoints: risk of breast cancer within 10 years and after 10 years of atypia biopsy. Thirty-two (16.7%) of the 192 women developed breast cancer over a median of 14.6 years. Thirty percent (58) of the atypias had >or=2% cells staining for Ki67. In these women, the risk of breast cancer within 10 years after atypia was increased (SIR 4.42 [2.21-8.84]) but not in those with <2% staining. Specifically, the cumulative incidence for breast cancer at 10 years was 14% in the high Ki67 vs. 3% in the low Ki67 group. Conversely, after 10 years, risk in the low Ki67 group rose significantly (SIR 5.69 [3.63-8.92]) vs. no further increased risk in the high Ki67 group (SIR 0.78 [0.11-5.55]). Ki67 appears to be a time-varying biomarker of risk of breast cancer in women with atypical hyperplasia.

[Atypical ductal and lobular hyperplasia of the breast]

Atypical hyperplasia represents 4% of all benign breast diseases. There are two different types: atypical ductal hyperplasia and atypical lobular hyperplasia. Aside columnar cell lesion. They represent an early stage of some forms of low grade carcinoma in situ and invasive carcinomas. Atypical hyperplasia is a benign lesion with intermediate carcinologic risk and the existence of a concomitant aggressive lesion should be suspected. When atypical lesion is found on a biopsy specimen, surgical excision is recommended especially in case of atypical ductal hyperplasia. A regular supervision is recommended.

Differences and similarities in breast cancer risk assessment models in clinical practice: which model to choose?

To show differences and similarities between risk estimation models for breast cancer in healthy women from BRCA1/2-negative or untested families. After a systematic literature search seven models were selected: Gail-2, Claus Model, Claus Tables, BOADICEA, Jonker Model, Claus-Extended Formula, and Tyrer-Cuzick. Life-time risks (LTRs) for developing breast cancer were estimated for two healthy counsellees, aged 40, with a variety in family histories and personal risk factors. Comparisons were made with guideline thresholds for individual screening. Without a clinically significant family history LTRs varied from 6.7% (Gail-2 Model) to 12.8% (Tyrer-Cuzick Model). Adding more information on personal risk factors increased the LTRs and yearly mammography will be advised in most situations. Older models (i.e. Gail-2 and Claus) are likely to underestimate the LTR for developing breast cancer as their baseline risk for women is too low. When models include personal risk factors, surveillance thresholds have to be reformulated. For current clinical practice, the Tyrer-Cuzick Model and the BOADICEA Model seem good choices.