Article Tools
ORIGINAL REPORTS
Cancer Prevention
Article Tools
Integrating Clinical and Polygenic Factors to Predict Breast Cancer Risk in Women Undergoing Genetic Testing
Abstract
Screening and prevention decisions for women at increased risk of developing breast cancer depend on genetic and clinical factors to estimate risk and select appropriate interventions. Integration of polygenic risk into clinical breast cancer risk estimators can improve discrimination. However, correlated genetic effects must be incorporated carefully to avoid overestimation of risk.
A novel Fixed-Stratified method was developed that accounts for confounding when adding a new factor to an established risk model. A combined risk score (CRS) of an 86–single-nucleotide polymorphism polygenic risk score and the Tyrer-Cuzick v7.02 clinical risk estimator was generated with attenuation for confounding by family history. Calibration and discriminatory accuracy of the CRS were evaluated in two independent validation cohorts of women of European ancestry (N = 1,615 and N = 518). Discrimination for remaining lifetime risk was examined by age-adjusted logistic regression. Risk stratification with a 20% risk threshold was compared between CRS and Tyrer-Cuzick in an independent clinical cohort (N = 32,576).
Simulation studies confirmed that the Fixed-Stratified method produced accurate risk estimation across patients with different family history. In both validation studies, CRS and Tyrer-Cuzick were significantly associated with breast cancer. In an analysis with both CRS and Tyrer-Cuzick as predictors of breast cancer, CRS added significant discrimination independent of that captured by Tyrer-Cuzick (P < 10−11 in validation 1; P < 10−7 in validation 2). In an independent cohort, 18% of women shifted breast cancer risk categories from their Tyrer-Cuzick–based risk compared with risk estimates by CRS.
Mammography and adjuvant treatment of early-stage disease are widely used for the mitigation of morbidity and mortality in invasive breast cancer.1 Effective prevention requires the identification of higher-risk individuals, which has typically involved family history evaluation, assessment of clinical and lifestyle factors, and testing for the presence of pathogenic variants (PVs) in a limited number of high- or moderate-risk breast cancer genes.2 Although guidelines for PV carriers have long recommended enhanced screening and discussion of surgical prevention measures, they do not adequately address the increased risk for unaffected PV-negative women with a strong family history of breast cancer. This clinical need has prompted recent proposals for more risk-adapted screening.3,4
Key Objective
Recently, single-nucleotide polymorphisms (SNPs) at multiple genomic locations have been shown to affect breast cancer risk. Polygenic risk scores that aggregate SNP-based risks explain considerable genetic breast cancer susceptibility. We examined confounding and interaction between an 86-SNP polygenic risk score and each clinical factor in the Tyrer-Cuzick model to inform the development of a combined clinical and polygenic risk score. The combined risk score was evaluated in two independent prespecified validation studies.
Knowledge Generated
The combined risk score showed improved stratification compared with the Tyrer-Cuzick model alone. This included downgrading and upgrading across risk categories.
Relevance
Professional societies recommend increased screening for women with a > 20% lifetime risk of developing breast cancer. Improved risk stratification from the combined risk score can help target the use of more intense screening strategies for unaffected women with truly elevated risk based on clinical and genetic factors.
Higher-risk women rely on risk prediction tools using clinical and demographic factors, such as the Tyrer-Cuzick model and others, for management guidance.5-10 The overall discriminatory accuracy of such prediction models, however, leaves room for improvement.11,12 Addition of single-nucleotide polymorphism (SNP)–based risk, aggregated into polygenic risk scores (PRSs), has been shown to enhance the discriminatory power of the Gail,13-16 BOADICEA,17-20 BRCAPRO,21 iCARE,22-24 and Tyrer-Cuzick models.21,25-28 In many of these studies, PRS-based risk is assumed to be independent of clinical risk, despite evidence for a genetic component in many of the conventional clinical risk factors.29,30 Family history in particular shows evidence for a partial overlap with polygenic risk, as addition of a 77-SNP score attenuated the odds ratio (OR) for family history by 12%-13% and attenuation increased to 21% for a 313-SNP score.31,32 In a separate report, a 75-SNP PRS was significantly associated with family history.20 Accurate risk prediction from a combined clinical and genetic risk model thus requires adjustment for shared risk contribution from genetic and clinical factors.
Here, we describe the development and validation of a strategy for integrating a SNP score with a validated clinical risk model, with attenuation for correlated effects. We used a previously validated 86-SNP PRS33 and the Tyrer-Cuzick clinical risk estimator to create a combined risk model adjusted for the shared genetic component related to family history.
These studies are based on two sample cohorts: a consecutive clinical testing cohort of patients submitted for hereditary cancer predisposition between June 1, 2017 and June 11, 2019 divided into sets for development and validation and one prospectively collected case-control validation study (validation 2). Clinical testing cohort data were subdivided according to successive time intervals to ensure the independence of development and validation activities (Fig 1); this included two development sets (development 1 and 2), one validation set (validation 1), and one clinical performance population. Genetic testing for all patients was performed at a Clinical Laboratory Improvement Amendments-approved and College of American Pathology-approved laboratory (Myriad Genetics, Inc., Salt Lake City, UT) by next-generation sequencing.34 Hybridization probes for the 86 SNPs in the PRS were included in the sequencing panel.33
FIG 1. Summary of independent study cohorts. *Development 2 was previously published by Hughes et al.33 CRS, combined risk score; DCIS, ductal carcinoma in situ; HBOC, hereditary breast and ovarian cancer; HNPCC, hereditary nonpolyposis colon cancer; LCIS, lobular carcinoma in situ; PRS, polygenic risk score; SNP, single-nucleotide polymorphism.
Women were eligible if they were of age 18-84 years, of European ancestry (Ashkenazi and non-Ashkenazi; self-reported), and negative for likely PVs or PVs in 11 breast cancer–related genes (BRCA1, BRCA2, TP53, PTEN, STK11, CDH1, PALB2, CHEK2, ATM, NBN, and BARD1). Patients were excluded as unaffected controls if they reported a history of ductal carcinoma in situ, lobular carcinoma in situ, hyperplasia, or unspecified breast disease. Women were excluded from the clinical testing cohort if they were submitted from states that disallow the research use of samples after completion of genetic testing. All studies were conducted with institutional review board (IRB) oversight (Quorum Review IRB #31713, 32556, 32608).
Development 1 included women from the clinical testing cohort unaffected by cancer of any type accrued between June 1, 2017 and August 11, 2017 (N = 5,489). Development 2 (N = 141,160) included 112,232 women unaffected by breast cancer and 28,928 breast cancer cases from the clinical testing cohort accrued between August 11, 2017 and January 11, 2019. Clinical characteristics of this set were previously reported.33
Validation 1 consisted of women in the clinical testing cohort accrued between June 1, 2017 and August 10, 2017 (N = 1,615). Cases included women with invasive breast cancer submitted for genetic testing within 1 year of diagnosis because of potential hereditary breast and ovarian cancer syndrome. Controls included unaffected women submitted for genetic testing because of possible hereditary nonpolyposis colorectal cancer syndrome as these patients have a breast cancer family history consistent with that of general population controls.
For validation 2, a consecutive series of women who presented to four breast cancer screening centers (Elizabeth Wende Breast Care, The Breast Center of NWA, Bethesda Health, and Cuda Women's Health Center/Cape Cod Healthcare) were prospectively recruited between February 6, 2017 and November 11, 2017 (N = 518). Cases included women with a pathologically confirmed first diagnosis of invasive breast cancer ≤ 12 months preenrollment or an incident breast cancer ≤ 6 months postenrollment. Unaffected controls had no history of breast disease through the end of study enrollment.
An independent set of clinical testing patients unaffected by breast cancer accrued between January 12, 2019 and June 11, 2019 (clinical performance population; N = 32,576) was used to assess risk stratification of the combined score.
The PRS was calculated as a linear combination of centered genotype calls for 86 confirmed breast cancer–associated SNPs (Data Supplement). Score composition information has been previously published.33
Family history of breast and ovarian cancer and personal risk factor data (Data Supplement) were retrieved from test request forms for clinical testing samples or from completed questionnaires collected as part of the prospective study. Absolute 5-year and remaining lifetime risk (RLR) estimates by Tyrer-Cuzick were calculated according to version 7.02.
We examined association between PRS and each clinical factor in the Tyrer-Cuzick model in development 1 (N = 5,489). For each factor, we conducted a simple linear regression with PRS as the dependent variable and the clinical factor as the independent variable. From these models, we examined regression coefficients, P values based on F-statistics, and Pearson correlation coefficients.
To calculate absolute risk incorporating Tyrer-Cuzick and PRS, we developed a Fixed-Stratified method (Data Supplement) to attenuate PRS after fixing the effects of confounded factors from the Tyrer-Cuzick model and to constrain risk separately within strata of the confounders. Briefly, women were stratified on the basis of breast cancer family history. Absolute risk for a woman in strata k was calculated as follows:
| (1) |
Simulation studies compared risks calculated by the Fixed-Stratified method with those from multivariable co-estimation and from univariable estimation of the effects of family history and PRS (Data Supplement).32
The combined risk score (CRS) was validated in two independent studies (Table 1): a clinical testing set (validation 1; N = 1,615) and the prospective case-control cohort (validation 2; N = 518). Validations were conducted according to prespecified statistical analysis plans using R version 3.5.3.35 P values were calculated from likelihood ratio chi-squared test statistics and reported as two-sided.
|
Primary analyses tested the CRS and Tyrer-Cuzick individually for association with breast cancer in age-adjusted logistic regression models (Data Supplement). Secondary analyses tested the incremental improvement of the CRS over Tyrer-Cuzick by including both risk estimators in the same age-adjusted model. Exploratory analyses tested calibration of the CRS by comparing average absolute risk estimates with those from the Tyrer-Cuzick model; with proper calibration, we expected to observe the same average risk for Tyrer-Cuzick as for the CRS among unaffected controls. These analyses were conducted separately for RLR and 5-year risk of developing breast cancer. Weighted logistic regression was used with weights for unaffected controls calculated such that average Tyrer-Cuzick RLR matched general population rates (Data Supplement).
Effect sizes of the 86-SNP PRS in validations 1 and 2 were calculated as ORs per standard deviation from multivariable logistic regression models adjusted for age, personal and family cancer history, and Ashkenazi ancestry and compared with those published previously.33 We tested for interaction between PRS and each clinical factor in Tyrer-Cuzick by constructing models with age, clinical factor, PRS, and an interaction term as predictor variables.
The clinical performance population (N = 32,576) was used to assess risk stratification using the CRS. RLR of breast cancer was calculated according to both the Tyrer-Cuzick model and the CRS, and classifications (increased [> 20%] or low [≤ 20%] risk) from both were compared.
We examined associations between the 86-SNP PRS and Tyrer-Cuzick model clinical factors in development 1 using linear regression models (Data Supplement). The 86-SNP PRS was significantly associated with family history, measured as either an affected first-degree relative (P = 2.0 × 10−9) or as a weighted count of affected relatives (P = 6.9 × 10−16). A marginal association with hormonal replacement therapy use was not significant after adjustment for multiple testing. The 86-SNP PRS was not correlated with any other Tyrer-Cuzick model clinical factor.
We evaluated the performance of the Fixed-Stratified method in a simulation study on the basis of previously published parameters for a 77-SNP PRS and family history.32 In the simulation, we matched published univariable and bivariable ORs for the 77-SNP PRS and family history and visually matched the published figure of cumulative absolute risk (Data Supplement). For women without family history, risk estimates based on the combined effects of the 77-SNP PRS and family history were accurate regardless of adjustment for confounding (Data Supplement). In contrast, for women with family history, the Fixed-Stratified method matched risks from multivariable co-estimation, whereas the unadjusted model overestimated risk (Data Supplement).
Validation 1 included 988 (61%) breast cancer cases and 627 (39%) unaffected controls (Table 1). Median age of hereditary cancer testing was 48 years [IQR 40 to 57]. Women with breast cancer tended to be older than unaffected controls. Overall, 362 (22%) patients reported breast cancer in ≥ 1 first-degree relative and 620 (38%) in ≥ 1 second-degree relative.
The prospectively collected case-control validation 2 included 256 (49%) breast cancer cases and 262 (51%) unaffected controls (Table 1). Median age at enrollment was 61 years (IQR 50-70). Nearly one third (29%) of patients reported breast cancer in a first-degree relative and 204 (39%) patients in ≥ 1 second-degree relative.
In both validations, RLR and 5-year risk estimates of the CRS and Tyrer-Cuzick were significantly associated with breast cancer (Table 2 and Fig 2). In a model with both risk predictors, the CRS added significant discrimination independent of that captured by Tyrer-Cuzick for both RLR (P < 10−11 in validation 1; P < 10−7 in validation 2) and 5-year risk (P < 10−11 in validation 1; P < 10−7 in validation 2; Data Supplement).
|
FIG 2. Discriminatory accuracy of CRS over Tyrer-Cuzick or PRS alone in validation 1 (A) and validation 2 (B). CRS, Tyrer-Cuzick, and PRS were evaluated separately in terms of likelihood ratio chi-squared test statistics from age-adjusted logistic regression models. In both validation studies, the CRS performed significantly better than either Tyrer-Cuzick or PRS at discriminating between women with and without invasive breast cancer. CRS, combined risk score; PRS, polygenic risk score.
Although all patients in validation 2 provided complete Tyrer-Cuzick risk factor questionnaires, data on Tyrer-Cuzick variables were incomplete for some patients in validation 1. Analyses were repeated in the subset of patients with complete information for all Tyrer-Cuzick risk factors with results similar to those from the full data set (Data Supplement).
Calibration of the CRS was examined by comparing average risk estimates with those from Tyrer-Cuzick in the control samples from either validation. The average RLR and 5-year risk estimates matched exactly in both validations (RLR 12.6%; 5-year 0.96%). Estimates were also consistent for patients grouped according to 5-year age bins (Fig 3). Concordance of mean CRS and Tyrer-Cuzick risk estimates indicates that the CRS was properly calibrated.
FIG 3. Evaluation of concordance between CRS (blue) and Tyrer-Cuzick (red). Estimates for average remaining lifetime (A and C) and 5-year risk (B and D) for unaffected controls in validation 1 (A-B) and validation 2 (C-D). Patients were grouped into 5-year age bins, and average risks were evaluated according to CRS and Tyrer-Cuzick. 95% CIs are also reported. CRS, combined risk score.
Effect sizes of the 86-SNP PRS calculated as ORs per standard deviation from multivariable logistic regression models were similar to previously published33 results in validation 1 (OR 1.57, 95% CI, 1.33 to 1.86) and validation 2 (OR 1.65, 95% CI, 1.37 to 2.00; Data Supplement). No evidence of interaction between the 86-SNP PRS and clinical factors in the Tyrer-Cuzick model in either of the two validations was observed (Data Supplement).
To illustrate how application of the CRS changes RLR estimates compared with Tyrer-Cuzick alone, we calculated RLR using Tyrer-Cuzick or CRS for an independent set of 32,576 unaffected women (clinical performance population). RLR estimates in this cohort using the CRS ranged from 0.01% to 74.0% (Fig 4A). This was a greater range than for RLR estimates from the Tyrer-Cuzick model for the same population (Fig 4B). Although Tyrer-Cuzick and CRS identified 35.5% and 33.0% of women as having > 20% RLR, respectively (Fig 4C), inclusion of SNP-based risk shifted risk categories for 18% of women. Of those with RLR > 20% by Tyrer-Cuzick alone, 29% were downgraded to ≤ 20% RLR by the CRS. Conversely, 12% of women with RLR ≤ 20% by Tyrer-Cuzick were upgraded to RLR > 20% by the CRS.
FIG 4. Distribution of CRS risk estimates in unaffected women. (A) Remaining lifetime risk (RLR) in a population of unaffected women (clinical performance population, N = 32,576; excluded women with ductal carcinoma in situ, lobular carcinoma in situ, hyperplasia, or unspecified breast disease) according to CRS with thresholds at 20% (increased) and 50% (high) RLR. (B) Scatterplot of RLR based on the Tyrer-Cuzick and CRS risk models for patients within the clinical performance population. (C) Distribution of patients above and below the 20% RLR threshold in the clinical performance population according to both the Tyrer-Cuzick and CRS models. Blue squares indicate patients with discordance between the scores (eg, the Tyrer-Cuzick model produced a score that indicated a patient had low RLR, but the same patient was determined to have increased RLR by the CRS model). CRS, combined risk score.
Clinical and epidemiologic risk factors have primarily been used to assess breast cancer risk in PV-negative women. Integration of these factors with SNP-based risk necessitates examination of both confounding effects and interactions between SNP-based and clinical risk. Here, we propose a novel Fixed-Stratified method that accounts for the confounding effect of family history. We applied this method to combine an 86-SNP PRS with the Tyrer-Cuzick risk model and tested the resulting CRS in two independent cohorts. Addition of the 86-SNP score significantly improved discrimination relative to the Tyrer-Cuzick model for predicting risk of breast cancer, and the CRS model showed excellent calibration across age groups. American Cancer Society guidelines recommend magnetic resonance imaging screening for women with > 20%-25% lifetime risk of breast cancer to improve early-stage cancer detection.36 In a clinical testing population, we showed that for 33% of PV-negative women, the CRS-estimated risk was > 20% and these women would qualify for consideration of enhanced surveillance. This includes 2,619 women (8%) who would have been categorized as < 20% using Tyrer-Cuzick–estimated risk alone.
Combinations of PRSs with the Tyrer-Cuzick model have been presented previously, generally without attenuation for confounding. An 88-SNP score was added to Tyrer-Cuzick under the assumption of independence in a nested case-control study.25 Although the SNP score added discrimination to the Tyrer-Cuzick prediction, it had poor calibration. In a larger data set, an 18-SNP PRS showed a small but significant correlation with the Tyrer-Cuzick model; however, its effect size for predicting risk was only marginally affected by adjustment with the Tyrer-Cuzick model.26 More recently, a 143-SNP PRS added to the Tyrer-Cuzick model attenuated the PRS effect, indicating an overlap between Tyrer-Cuzick and PRS.27 Combined risk associations of a 313-SNP PRS and classical breast cancer risk factors have also recently been evaluated, although these combined risks have yet to be validated.37 These examples highlight two issues. First, previous examinations of confounding often evaluated the correlations between a PRS and a model rather than individual clinical factors.21,25,26 This approach might have obscured associations between PRS and clinical factors with a stronger genetic component. Second, genetic overlap is more likely with a larger PRS since casting a wider net for cancer-associated SNPs is more likely to also capture a larger fraction of the genetic component of clinical risk. Here, we focused on correlations between the PRS and individual Tyrer-Cuzick variables rather than the Tyrer-Cuzick model as a whole. We identified confounding between PRS and family history and adjusted for the genetic overlap with family history. Others have examined the question of interaction between PRS and clinical factors and confirmed the appropriateness of a multiplicative model.37
Attenuation of shared PRS and family history risk has been described for the addition of a 313-SNP PRS to BOADICEA and the iCARE model.17,24,37 Segregation analysis was used to incorporate the PRS into BOADICEA. One limitation of that approach is approximating the distribution of the polygenic component with a binomial distribution; it is not clear how important this limitation is. The iCARE model is based on reducing the family history contribution by subtracting half of the PRS variance from the log(OR) for family history. However, the family history within this approach was binary (presence or absence of a first-degree relative with breast cancer). This limitation does not allow for optimal utilization of family history for patients with > 1 first-degree relative or those with second-degree relatives.
Limitations of the present study include a potential ascertainment bias in the clinical testing population cohorts. Qualification for genetic testing is often based on family history. It has been previously shown that this potential bias can be avoided by accounting for family history in a multivariable model.38,39 Consequently, all analyses presented here were conducted by multivariable co-estimation. In support of this approach, results from validation 1 (clinical testing samples) were similar to results obtained in validation 2, for which participants were prospectively collected and were unaffected by potential bias due to selection for hereditary cancer testing. Evaluation of the CRS in a prospectively collected, unselected population-based sample is being pursued.40 Future work to expand assessment tools such as Tyrer-Cuzick and the PRS for non-European ancestries will be required to apply these tools to other ancestry populations. Finally, information for Tyrer-Cuzick for the clinical testing populations was obtained from provider-completed test request forms. As such, the accuracy and/or completeness of this information cannot be verified and missing information could not be obtained retrospectively. However, subanalysis of validation 1 in women with complete Tyrer-Cuzick information showed similar results as the whole data set, indicating that missing information did not substantially affect the analysis.
This CRS model is suitable for reporting age-specific risk of developing breast cancer for unaffected women of European descent with or without significant family history. It is currently the only Tyrer-Cuzick–based model fully adjusted for the shared risk between SNPs and family history and is therefore less likely to overestimate risk in women with a family history of breast cancer. A CRS containing the 86-SNP PRS is commercially available for PV-negative women of European ancestry (Myriad Genetics, Salt Lake City, UT). Additional studies are needed to explore informative SNPs for non-European ancestries. Future evaluation of the CRS model with additional variables such as breast density (ie, Tyrer-Cuzick v8), lifestyle, and dietary factors is desirable. Combined clinical and genetic risk models improve breast cancer risk prediction and may result in better allocation of cancer risk–reduction resources, such as chemoprevention and enhanced imaging techniques, to women with the highest combined risk.
Conception and design: Elisha Hughes, Thaddeus Judkins, Shannon Gallagher, Johnathan M. Lancaster, Jerry S. Lanchbury, Alexander Gutin, Mark E. Robson
Collection and assembly of data: Elisha Hughes, Placede Tshiaba, Shannon Gallagher, Kathryn Dalton, Wade Hedegard, Carol A. Adami, Danna F. Grear, Jerry S. Lanchbury
Data analysis and interpretation: Elisha Hughes, Placede Tshiaba, Susanne Wagner, Eric Rosenthal, Benjamin Roa, Shannon Gallagher, Stephanie Meek, Wade Hedegard, Susan M. Domchek, Judy Garber, Jeffrey N. Weitzel, Allison W. Kurian, Jerry S. Lanchbury, Alexander Gutin, Mark E. Robson
Manuscript writing: All authors
Final approval of manuscript: All authors
Accountable for all aspects of the work: All authors
The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.
Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).
Employment: Myriad Genetics
Stock and Other Ownership Interests: Myriad Genetics
Employment: Myriad Genetics
Stock and Other Ownership Interests: Myriad Genetics
Employment: Myriad Genetics
Stock and Other Ownership Interests: Myriad Genetics
Patents, Royalties, Other Intellectual Property: Coauthor of patents held by Myriad Genetics, no royalties
Employment: Myriad Genetics
Stock and Other Ownership Interests: Myriad Genetics
Travel, Accommodations, Expenses: Myriad Genetics
Employment: Myriad Genetics
Stock and Other Ownership Interests: Myriad Genetics
Employment: Myriad Genetics
Leadership: Myriad Genetics
Stock and Other Ownership Interests: Myriad Genetics
Research Funding: Myriad Genetics
Patents, Royalties, Other Intellectual Property: Intellectual property held by employer Myriad Genetics
Travel, Accommodations, Expenses: Myriad Genetics
Employment: Myriad Genetics
Stock and Other Ownership Interests: Myriad Genetics
Employment: Myriad Genetics
Stock and Other Ownership Interests: Myriad Genetics
Speakers' Bureau: Myriad Genetics
Consulting or Advisory Role: Myriad Genetics
Speakers' Bureau: Myriad Genetics
Travel, Accommodations, Expenses: Myriad Genetics
Honoraria: AstraZeneca, Clovis Oncology, Bristol-Myers Squibb
Research Funding: AstraZeneca, Clovis Oncology
Consulting or Advisory Role: Novartis, GTx, Helix BioPharma, Konica Minolta, Aleta BioTherapeutics, H3 Biomedicine, Kronos Bio
Research Funding: Novartis, Ambry Genetics, InVitae, Myriad Genetics
Other Relationship: Susan G. Komen for the Cure, AACR, Diana Helis Henry Medical Foundation, James P. Wilmot Foundation, Adrienne Helis Malvin Medical Research Foundation, Breast Cancer Research Foundation, Facing our Risk of Cancer Empowered
Employment: Myriad Genetics, Regeneron
Leadership: Myriad Genetics, Regeneron
Stock and Other Ownership Interests: Myriad Genetics, Regeneron
Speakers' Bureau: AstraZeneca
Research Funding: Myriad Genetics
Other Relationship: Ambry Genetics, Color Genomics, GeneDx/BioReference, InVitae, Genentech
Employment: Myriad Genetics
Leadership: Myriad Genetics
Stock and Other Ownership Interests: Myriad Genetics
Patents, Royalties, Other Intellectual Property: I am an inventor on multiple patents filed by Myriad Genetics
Travel, Accommodations, Expenses: Myriad Genetics
Employment: Myriad Genetics
Stock and Other Ownership Interests: Myriad Genetics, Gilead Sciences
Consulting or Advisory Role: Change HealthCare
Research Funding: AstraZeneca, AbbVie, Pfizer, Merck
Travel, Accommodations, Expenses: AstraZeneca, Merck
Other Relationship: Research to Practice, Clinical Care Options, Physicians' Education Resource, Invitae, Pfizer
(OPTIONAL) Uncompensated Relationships: Merck, Pfizer, Daiichi Sankyo, Epic Sciences, https://openpaymentsdata.cms.gov/physician/612669/summary
No other potential conflicts of interest were reported.
| 1. | Nelson HD, Fu R, Cantor A, et al: Effectiveness of breast cancer screening: Systematic review and meta-analysis to update the 2009 U.S. Preventive Services Task Force Recommendation. Ann Intern Med 164:244-255, 2016 Crossref, Medline, Google Scholar |
| 2. | Daly MB, Pilarski R, Berry M, et al: NCCN Clinical Practice Guidelines in Oncology, Genetic/Familial High-Risk Assessment: Breast and ovarian (version 3.2019). NCCN Clinical Practice Guidelines in Oncology 2019; https://www.nccn.org/professionals/physician_gls/pdf/genetics_bop.pdf Google Scholar |
| 3. | Mukama T, Kharazmi E, Xing X, et al: Risk-adapted starting age of screening for relatives of patients with breast cancer. JAMA Oncol 6:68-74, 2019 Crossref, Google Scholar |
| 4. | Gierach GL, Choudhury PP, García-Closas M: Toward risk-stratified breast cancer screening: Considerations for changes in screening guidelines. JAMA Oncol 10.1001/jamaoncol.2019.3820 [epub ahead of print on November 14, 2019] Google Scholar |
| 5. | Amir E, Evans DG, Shenton A, et al: Evaluation of breast cancer risk assessment packages in the family history evaluation and screening programme. J Med Genet 40:807-814, 2003 Crossref, Medline, Google Scholar |
| 6. | Tyrer J, Duffy SW, Cuzick J: A breast cancer prediction model incorporating familial and personal risk factors. Stat Med 23:1111-1130, 2004 Crossref, Medline, Google Scholar |
| 7. | Antoniou AC, Cunningham AP, Peto J, et al: The BOADICEA model of genetic susceptibility to breast and ovarian cancers: Updates and extensions. Br J Cancer 98:1457-1466, 2008 Crossref, Medline, Google Scholar |
| 8. | Costantino JP, Gail MH, Pee D, et al: Validation studies for models projecting the risk of invasive and total breast cancer incidence. J Natl Cancer Inst. 91:1541-1548, 1999 Crossref, Medline, Google Scholar |
| 9. | Tung N, Domchek SM, Stadler Z, et al: Counselling framework for moderate-penetrance cancer-susceptibility mutations. Nat Rev Clin Oncol 13:581-588, 2016 Crossref, Medline, Google Scholar |
| 10. | Taylor A, Brady AF, Frayling IM, et al: Consensus for genes to be included on cancer panel tests offered by UK genetics services: Guidelines of the UK Cancer Genetics Group. J Med Genet 55:372, 2018 Crossref, Medline, Google Scholar |
| 11. | Louro J, Posso M, Hilton Boon M, et al: A systematic review and quality assessment of individualised breast cancer risk prediction models. Br J Cancer 121:76-85, 2019 Crossref, Medline, Google Scholar |
| 12. | Terry MB, Liao Y, Whittemore AS, et al: 10-year performance of four models of breast cancer risk: A validation study. Lancet Oncol 20:504-517, 2019 Crossref, Medline, Google Scholar |
| 13. | Gail MH: Value of adding single-nucleotide polymorphism genotypes to a breast cancer risk model. J Natl Cancer Inst 101:959-963, 2009 Crossref, Medline, Google Scholar |
| 14. | McCarthy AM, Armstrong K, Handorf E, et al: Incremental impact of breast cancer SNP panel on risk classification in a screening population of white and African American women. Breast Cancer Res Treat 138:889-898, 2013 Crossref, Medline, Google Scholar |
| 15. | Wacholder S, Hartge P, Prentice R, et al: Performance of common genetic variants in breast-cancer risk models. N Engl J Med 362:986-993, 2010 Crossref, Medline, Google Scholar |
| 16. | Zhang X, Rice M, Tworoger SS, et al: Addition of a polygenic risk score, mammographic density, and endogenous hormones to existing breast cancer risk prediction models: A nested case–control study. PLoS Med 15:e1002644, 2018 Crossref, Medline, Google Scholar |
| 17. | Lee A, Mavaddat N, Wilcox AN, et al: BOADICEA: A comprehensive breast cancer risk prediction model incorporating genetic and nongenetic risk factors. Genet Med 21:1708-1718, 2019 Crossref, Medline, Google Scholar |
| 18. | Lakeman IMM, Hilbers FS, Rodríguez-Girondo M, et al: Addition of a 161-SNP polygenic risk score to family history-based risk prediction: Impact on clinical management in non-BRCA1/2 breast cancer families. J Med Genet 56:581, 2019 Crossref, Medline, Google Scholar |
| 19. | Muranen TA, Greco D, Blomqvist C, et al: Genetic modifiers of CHEK2*1100delC-associated breast cancer risk. Genet Med 19:599-603, 2017 Crossref, Medline, Google Scholar |
| 20. | Muranen TA, Mavaddat N, Khan S, et al: Polygenic risk score is associated with increased disease risk in 52 Finnish breast cancer families. Breast Cancer Res Treat 158:463-469, 2016 Crossref, Medline, Google Scholar |
| 21. | Dite GS, MacInnis RJ, Bickerstaffe A, et al: Breast cancer risk prediction using clinical models and 77 independent risk-associated SNPs for women aged under 50 Years: Australian Breast Cancer Family Registry. Cancer Epidemiol Biomarkers Prev 25:359-365, 2016 Crossref, Medline, Google Scholar |
| 22. | Mealiffe ME, Stokowski RP, Rhees BK, et al: Assessment of clinical validity of a breast cancer risk model combining genetic and clinical information. J Natl Cancer Inst 102:1618-1627, 2010 Crossref, Medline, Google Scholar |
| 23. | Shieh Y, Hu D, Ma L, et al: Breast cancer risk prediction using a clinical risk model and polygenic risk score. Breast Cancer Res Treat 159:513-525, 2016 Crossref, Medline, Google Scholar |
| 24. | Pal Choudhury P, Maas P, Wilcox A, et al: iCARE: An R package to build, validate and apply absolute risk models. PLoS One 15:e0228198, 2020 Crossref, Medline, Google Scholar |
| 25. | Cuzick J, Brentnall AR, Segal C, et al: Impact of a panel of 88 single nucleotide polymorphisms on the risk of breast cancer in high-risk women: Results from two randomized tamoxifen prevention trials. J Clin Oncol 35:743-750, 2016 Link, Google Scholar |
| 26. | van Veen EM, Brentnall AR, Byers H, et al: Use of single-nucleotide polymorphisms and mammographic density plus classic risk factors for breast cancer risk prediction. JAMA Oncol 4:476-482, 2018 Crossref, Medline, Google Scholar |
| 27. | Brentnall AR, van Veen EM, Harkness EF, et al: A case–control evaluation of 143 single nucleotide polymorphisms for breast cancer risk stratification with classical factors and mammographic density. Int J Cancer 146:2122-2129, 2020 Crossref, Medline, Google Scholar |
| 28. | Evans DG, Brentnall A, Byers H, et al: The impact of a panel of 18 SNPs on breast cancer risk in women attending a UK familial screening clinic: A case–control study. J Med Genet 54:111, 2017 Crossref, Medline, Google Scholar |
| 29. | Fall T, Ingelsson E: Genome-wide association studies of obesity and metabolic syndrome. Mol Cell Endocrinol 382:740-757, 2014 Crossref, Medline, Google Scholar |
| 30. | He C, Murabito JM: Genome-wide association studies of age at menarche and age at natural menopause. Mol Cell Endocrinol 382:767-779, 2014 Crossref, Medline, Google Scholar |
| 31. | Mavaddat N, Michailidou K, Dennis J, et al: Polygenic risk scores for prediction of breast cancer and breast cancer subtypes. Am J Hum Genet 104:21-34, 2019 Crossref, Medline, Google Scholar |
| 32. | Mavaddat N, Pharoah PD, Michailidou K, et al: Prediction of breast cancer risk based on profiling with common genetic variants. J Natl Cancer Inst 107:djv036, 2015 Crossref, Medline, Google Scholar |
| 33. | Hughes E, Tshiaba P, Gallagher S, et al: Development and validation of a polygenic risk score to predict breast cancer risk. JCO Precis Oncol 4:585-592, 2020 Link, Google Scholar |
| 34. | Judkins T, Leclair B, Bowles K, et al: Development and analytical validation of a 25-gene next generation sequencing panel that includes the BRCA1 and BRCA2 genes to assess hereditary cancer risk. BMC Cancer 15:215, 2015 Crossref, Medline, Google Scholar |
| 35. | Team RC: R: A Language and Environment for Statistical Computing. Vienna, Austria, R Foundation for Statistical Computing, 2019 Google Scholar |
| 36. | Saslow D, Boetes C, Burke W, et al: American Cancer Society guidelines for breast screening with MRI as an adjunct to mammography. CA Cancer J Clin 57:75-89, 2007 Crossref, Medline, Google Scholar |
| 37. | Kapoor PM, Mavaddat N, Choudhury PP, et al: Combined associations of a polygenic risk score and classical risk factors with breast cancer risk. J Natl Cancer Inst doi: https://doi.org/10.1093/jnci/djaa056 [epub ahead of print on May 2, 2020] Google Scholar |
| 38. | Kurian AW, Hughes E, Handorf EA, et al: Breast and ovarian cancer penetrance estimates derived from germline multiple-gene sequencing results in women. JCO Precis Oncol 1-12, 2017 Medline, Google Scholar |
| 39. | Rothman KJ, Greenland S, Lash T: Modern Epidemiology. Philadelphia, PA, Lippincott Williams & Wilkins, 2008 Google Scholar |
| 40. | Kurian AW, Bernhisel R, Larson K, et al: Prevalence of pathogenic variants in cancer susceptibility genes among women with postmenopausal breast cancer. JAMA 323:995-997, 2020 Crossref, Medline, Google Scholar |
