Breast Cancer Risk Prediction Research Articles

Cost-effectiveness of using artificial intelligence versus polygenic risk score to guide breast cancer screening.

BackgroundCurrent guidelines for mammography screening for breast cancer vary across agencies, especially for women aged 40–49. Using artificial Intelligence (AI) to read mammography images has been shown to predict breast cancer risk with higher accuracy than alternative approaches including polygenic risk scores (PRS), raising the question whether AI-based screening is more cost-effective than screening based on PRS or existing guidelines. This study provides the first evidence to shed light on this important question.MethodsThis study is a model-based economic evaluation. We used a hybrid decision tree/microsimulation model to compare the cost-effectiveness of eight strategies of mammography screening for women aged 40–49 (screening beyond age 50 follows existing guidelines). Six of these strategies were defined by combinations of risk prediction approaches (AI, PRS or family history) and screening frequency for low-risk women (no screening or biennial screening). The other two strategies involved annual screening for all women and no screening, respectively. Data used to populate the model were sourced from the published literature.ResultsRisk prediction using AI followed by no screening for low-risk women is the most cost-effective strategy. It dominates (i.e., costs more and generates fewer quality adjusted life years (QALYs)) strategies for risk prediction using PRS followed by no screening or biennial screening for low-risk women, risk prediction using AI or family history followed by biennial screening for low-risk women, and annual screening for all women. It also extendedly dominates (i.e., achieves higher QALYs at a lower incremental cost per QALY) the strategy for risk prediction using family history followed by no screening for low-risk women. Meanwhile, it is cost-effective versus no screening, with an incremental cost-effectiveness ratio of $23,755 per QALY gained.ConclusionsRisk prediction using AI followed by no breast cancer screening for low-risk women is the most cost-effective strategy. This finding can be explained by AI’s ability to identify high-risk women more accurately than PRS and family history (which reduces the possibility of delayed breast cancer diagnosis) and fewer false-positive diagnoses from not screening low-risk women.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12885-022-09613-1.

136P Development of breast cancer risk prediction models using the UK Biobank dataset

Personalised BC risk models based on modifiable factors can help to increase BC awareness. Aims for this project: to review BC risk models, to review the published BC risk factors, to assess the BC risk factors of UK Biobank, to explore the effects of “healthier lifestyles” with different genetic predispositions, to develop BC risk prediction models.

Overlap of high-risk individuals predicted by family history, and genetic and non-genetic breast cancer risk prediction models: implications for risk stratification

BackgroundFamily history, and genetic and non-genetic risk factors can stratify women according to their individual risk of developing breast cancer. The extent of overlap between these risk predictors is not clear.MethodsIn this case-only analysis involving 7600 Asian breast cancer patients diagnosed between age 30 and 75 years, we examined identification of high-risk patients based on positive family history, the Gail model 5-year absolute risk [5yAR] above 1.3%, breast cancer predisposition genes (protein-truncating variants [PTV] in ATM, BRCA1, BRCA2, CHEK2, PALB2, BARD1, RAD51C, RAD51D, or TP53), and polygenic risk score (PRS) 5yAR above 1.3%.ResultsCorrelation between 5yAR (at age of diagnosis) predicted by PRS and the Gail model was low (r=0.27). Fifty-three percent of breast cancer patients (n=4041) were considered high risk by one or more classification criteria. Positive family history, PTV carriership, PRS, or the Gail model identified 1247 (16%), 385 (5%), 2774 (36%), and 1592 (21%) patients who were considered at high risk, respectively. In a subset of 3227 women aged below 50 years, the four models studied identified 470 (15%), 213 (7%), 769 (24%), and 325 (10%) unique patients who were considered at high risk, respectively. For younger women, PRS and PTVs together identified 745 (59% of 1276) high-risk individuals who were not identified by the Gail model or family history.ConclusionsFamily history and genetic and non-genetic risk stratification tools have the potential to complement one another to identify women at high risk.

Association of polymorphisms in four low-penetrance genes (ACYP2, CCNE1, ERCC5 and lincRNA CASC22) with breast cancer risk in a Vietnamese population

In breast cancer (BC), four low-penetrance genes (ACYP2, CCNE1, ERCC5 and lincRNA CASC22) are involved in cell cycle arrest, DNA damage repair and long non-coding RNA-mediated regulation is commonly known as causative genes of tumorigenesis. A functional polymorphism in these genes may cause cancer by altering the genes' structure and function. Single Nucleotide Polymorphisms (SNPs) have been discovered to be capable of predicting cancer risk, hence can be used as carcinogenesis markers. To date, the correlation of four SNPs (rs12621038, rs3218038, rs751402 and rs12325489) with BC risk in different populations has been revealed by numerous researches. This study examines the association between these SNPs and the risk of developing BC in Vietnamese women. These four SNPs were genotyped by HRM assay. The differences in allelic and genotypic frequencies between cases and controls were analyzed to determine their relationship with BC risk. While rs12621038, rs3218038 and rs751402 showed no relation to BC, there was an association between rs12325489 and BC risk. In both codominant and overdominant models, the lincRNA CASC22 rs12325489 was consistently associated with BC risk (TT vs. CC: OR = 0.63, 95% CI = 0.44 to 0.90, P = 0.01 and TC vs. TT + CC: OR = 0.61, 95% CI = 0.43 to 0.84, P = 0.003; respectively). These results indicated that rs12325489 could be a potential biomarker for predicting BC risk in Vietnamese women.

Comprehensive Study for Breast Cancer Using Deep Learning and Traditional Machine Learning

Comprehensive Study for Breast Cancer Using Deep Learning and Traditional Machine Learning

Predicting breast cancer risk in a racially diverse, community-based sample of potentially high-risk women.

Identifying women with high risk of breast cancer is necessary to study high-risk experiences and deliver risk-management care. Risk prediction models estimate individuals' lifetime risk but have rarely been applied in community-based settings among women not yet receiving specialized care. Therefore, we aimed: (1) to apply three breast cancer risk prediction models (i.e., Gail, Claus, and IBIS) to a racially diverse, community-based sample of women, and (2) to assess risk prediction estimates using survey data. An online survey was administered to women who were determined by a screening instrument to have potentially high risk for breast cancer. Risk prediction models were applied using their self-reported family and medical history information. Inclusion in the high-risk subsample required ≥20% lifetime risk per ≥1 model. Descriptive statistics were used to compare the proportions of women identified as high risk by each model. N=1053 women were initially eligible and completed the survey. All women, except one, self-reported the information necessary to run at least one model; 90% had sufficient information for >1 model. The high-risk subsample included 717 women, of which 75% were identified by one model only; 96% were identified by IBIS, 3% by Claus, <1% by Gail. In the high-risk subsample, 20% were identified by two models and 3% by all three models. Assessing breast cancer risk using self-reported data in a community-based sample was feasible. Different models identify substantially different groups of women who may be at high risk for breast cancer; use of multiple models may be beneficial for research and clinical care.

Association of contralateral breast cancer risk with mammographic density defined at higher-than-conventional intensity thresholds.

Mammographic dense area (MDA) is an established predictor of future breast cancer risk. Recent studies have found that risk prediction might be improved by redefining MDA in effect at higher-than-conventional intensity thresholds. We assessed whether such higher-intensity MDA measures gave stronger prediction of subsequent contralateral breast cancer (CBC) risk using the Women's Environment, Cancer, and Radiation Epidemiology (WECARE) Study, a population-based CBC case-control study of ≥1year survivors of unilateral breast cancer diagnosed between 1990 and 2008. Three measures of MDA for the unaffected contralateral breast were made at the conventional intensity threshold ("Cumulus") and at two sequentially higher-intensity thresholds ("Altocumulus" and "Cirrocumulus") using the CUMULUS software and mammograms taken up to 3 years prior to the first breast cancer diagnosis. The measures were fitted separately and together in multivariable-adjusted logistic regression models of CBC (252 CBC cases and 271 unilateral breast cancer controls). The strongest association with CBC was MDA defined using the highest intensity threshold, Cirrocumulus (odds ratio per adjusted SD [OPERA] 1.40, 95% CI 1.13-1.73); and the weakest association was MDA defined at the conventional threshold, Cumulus (1.32, 95% CI 1.05-1.66). In a model fitting the three measures together, the association of CBC with Cirrocumulus was unchanged (1.40, 95% CI 0.97-2.05), and the lower brightness measures did not contribute to the CBC model fit. These results suggest that MDA defined at a high-intensity threshold is a better predictor of CBC risk and has the potential to improve CBC risk stratification beyond conventional MDA measures.

A Case Study in Breast Density Evaluation Using Bioimpedance Measurements.

(1) Background: As breast cancer studies suggest, a high percentage of breast density (PBD) may be related to breast cancer incidence. Although PBD screening is one of the strongest predictors of breast cancer risk, X-ray-based mammography evaluation is subjective. Therefore, new objective PBD measuring techniques are of interest. A case study analyzing the PBD of thirteen female participants using a bioimpedance-based method, the anomalies tracking circle (ATC), is described in this paper. (2) Methods: In the first stage, the breast bioimpedance of each participant was measured. Then, the participant breast density was determined by applying a mammogram just after the breast bioimpedance measurement stage. In the third stage, the ATC algorithm was applied to the measured bioimpedance data for each participant, and a results analysis was done. (3) Results: An ATC variation according to the breast density was observed from the obtained data, this allowed the use of classification techniques to determine the PBD. (4) Conclusions: The described breast density method is a promising approach that might be applied as an auxiliary tool to the mammography in order to obtain precise and objective results for evaluation of breast density and with that determine potential breast cancer risk.

Machine Learning Models for Predicting Breast Cancer Risk in Women Exposed to Blue Light from Digital Screens.

Nowadays, there is a growing global concern over rapidly increasing screen time (smartphones, tablets, and computers). An accumulating body of evidence indicates that prolonged exposure to short-wavelength visible light (blue component) emitted from digital screens may cause cancer. The application of machine learning (ML) methods has significantly improved the accuracy of predictions in fields such as cancer susceptibility, recurrence, and survival. To develop an ML model for predicting the risk of breast cancer in women via several parameters related to exposure to ionizing and non-ionizing radiation. In this analytical study, three ML models Random Forest (RF), Support Vector Machine (SVM), and Multi-Layer Perceptron Neural Network (MLPNN) were used to analyze data collected from 603 cases, including 309 breast cancer cases and 294 gender and age-matched controls. Standard face-to-face interviews were performed using a standard questionnaire for data collection. The examined models RF, SVM, and MLPNN performed well for correctly classifying cases with breast cancer and the healthy ones (mean sensitivity> 97.2%, mean specificity >96.4%, and average accuracy >97.1%). Machine learning models can be used to effectively predict the risk of breast cancer via the history of exposure to ionizing and non-ionizing radiation (including blue light and screen time issues) parameters. The performance of the developed methods is encouraging; nevertheless, further investigation is required to confirm that machine learning techniques can diagnose breast cancer with relatively high accuracies automatically.

BREAst screening Tailored for HEr (BREATHE)-A study protocol on personalised risk-based breast cancer screening programme.

Routine mammography screening is currently the standard tool for finding cancers at an early stage, when treatment is most successful. Current breast screening programmes are one-size-fits-all which all women above a certain age threshold are encouraged to participate. However, breast cancer risk varies by individual. The BREAst screening Tailored for HEr (BREATHE) study aims to assess acceptability of a comprehensive risk-based personalised breast screening in Singapore. Advancing beyond the current age-based screening paradigm, BREATHE integrates both genetic and non-genetic breast cancer risk prediction tools to personalise screening recommendations. BREATHE is a cohort study targeting to recruit ~3,500 women. The first recruitment visit will include questionnaires and a buccal cheek swab. After receiving a tailored breast cancer risk report, participants will attend an in-person risk review, followed by a final session assessing the acceptability of our risk stratification programme. Risk prediction is based on: a) Gail model (non-genetic), b) mammographic density and recall, c) BOADICEA predictions (breast cancer predisposition genes), and d) breast cancer polygenic risk score. For national implementation of personalised risk-based breast screening, exploration of the acceptability within the target populace is critical, in addition to validated predication tools. To our knowledge, this is the first study to implement a comprehensive risk-based mammography screening programme in Asia. The BREATHE study will provide essential data for policy implementation which will transform the health system to deliver a better health and healthcare outcomes.

Improving breast cancer risk prediction with epigenetic risk factors

Improving breast cancer risk prediction with epigenetic risk factors

Health influenced by genetics: A first comprehensive analysis of breast cancer high and moderate penetrance susceptibility genes in the Tunisian population.

Significant advances have been made to understand the genetic basis of breast cancer. High, moderate and low penetrance variants have been identified with inter-ethnic variability in mutation frequency and spectrum. Genome wide association studies (GWAS) are widely used to identify disease-associated SNPs. Understanding the functional impact of these risk-SNPs will help the translation of GWAS findings into clinical interventions. Here we aim to characterize the genetic patterns of high and moderate penetrance breast cancer susceptibility genes and to assess the functional impact of non-coding SNPs. We analyzed BRCA1/2, PTEN, STK11, TP53, ATM, BRIP1, CHEK2 and PALB2 genotype data obtained from 135 healthy participants genotyped using Affymetrix Genome-Wide Human SNP-Array 6.0. Haplotype analysis was performed using Haploview.V4.2 and PHASE.V2.1. Population structure and genetic differentiation were assessed using principal component analysis (PCA) and fixation index (FST). Functional annotation was performed using In Silico web-based tools including RegulomeDB and VARAdb. Haplotype analysis showed distinct LD patterns with high levels of recombination and haplotype blocks of moderate to small size. Our findings revealed also that the Tunisian population tends to have a mixed origin with European, South Asian and Mexican footprints. Functional annotation allowed the selection of 28 putative regulatory variants. Of special interest were BRCA1_ rs8176318 predicted to alter the binding sites of a tumor suppressor miRNA hsa-miR-149 and PALB2_ rs120963 located in tumorigenesis-associated enhancer and predicted to strongly affect the binding of P53. Significant differences in allele frequencies were observed with populations of African and European ancestries for rs8176318 and rs120963 respectively. Our findings will help to better understand the genetic basis of breast cancer by guiding upcoming genome wide studies in the Tunisian population. Putative functional SNPs may be used to develop an efficient polygenic risk score to predict breast cancer risk leading to better disease prevention and management.

Incorporating Polygenic Risk Scores and Nongenetic Risk Factors for Breast Cancer Risk Prediction Among Asian Women

Polygenic risk scores (PRSs) have shown promise in breast cancer risk prediction; however, limited studies have been conducted among Asian women. To develop breast cancer risk prediction models for Asian women incorporating PRSs and nongenetic risk factors. This diagnostic study included women of Asian ancestry from the Asia Breast Cancer Consortium. PRSs were developed using data from genomewide association studies (GWASs) of breast cancer conducted among 123 041 women with Asian ancestry (including 18 650 women with breast cancer) using 3 approaches: (1) reported PRS for women with European ancestry; (2) breast cancer-associated single-nucleotide variations (SNVs) identified by fine-mapping of GWAS-identified risk loci; and (3) genomewide risk prediction algorithms. A nongenetic risk score (NGRS) was built, including 7 well-established nongenetic risk factors, using data of 416 case participants and 1558 control participants from a prospective cohort study. PRSs were initially validated in an independent data set including 1426 case participants and 1323 control participants and further evaluated, along with the NGRS, in the second data set including 368 case participants and 736 control participants nested within a prospective cohort study. Logistic regression was used to examine associations of risk scores with breast cancer risk to estimate odds ratios (ORs) with 95% CIs and area under the receiver operating characteristic curve (AUC). A total of 126 894 women of Asian ancestry were included; 20 444 (16.1%) had breast cancer. The mean (SD) age ranged from 49.1 (10.8) to 54.4 (10.4) years for case participants and 50.6 (9.5) to 54.0 (7.4) years for control participants among studies that provided demographic characteristics. In the prospective cohort, a PRS with 111 SNVs developed using the fine-mapping approach (PRS111) showed a prediction performance comparable with a genomewide PRS that included more than 855 000 SNVs. The OR per SD increase of PRS111 score was 1.67 (95% CI, 1.46-1.92), with an AUC of 0.639 (95% CI, 0.604-0.674). The NGRS had a limited predictive ability (AUC, 0.565; 95% CI, 0.529-0.601). Compared with the average risk group (40th-60th percentile), women in the top 5% of PRS111 and NGRS were at a 3.84-fold (95% CI, 2.30-6.46) and 2.10-fold (95% CI, 1.22-3.62) higher risk of breast cancer, respectively. The prediction model including both PRS111 and NGRS achieved the highest prediction accuracy (AUC, 0.648; 95% CI, 0.613-0.682). In this study, PRSs derived using breast cancer risk-associated SNVs had similar predictive performance in Asian and European women. Including nongenetic risk factors in models further improved prediction accuracy. These findings support the utility of these models in developing personalized screening and prevention strategies.

Prediction of Short-Term Breast Cancer Risk with Fusion of CC- and MLO-Based Risk Models in Four-View Mammograms.

This study performed and assessed a novel program to improve the accuracy of short-term breast cancer risk prediction by using information from craniocaudal (CC) and mediolateral-oblique (MLO) views of two breasts. An age-matched dataset of 556 patients with at least two sequential full-field digital mammography examinations was applied. In the second examination, 278 cases were diagnosed and pathologically verified as cancer, and 278 were negative, while all cases in the first examination were negative (not recalled). Two generalized linear-model-based risk prediction models were established with global- and local-based bilateral asymmetry features for CC and MLO views first. Then, a new fusion risk model was developed by fusing prediction results of the CC- and MLO-based risk models with an adaptive alpha-integration-based fusion method. The AUC of the fusion risk model was 0.72 ± 0.02, which was significantly higher than the AUC of CC- or MLO-based risk model (P < 0.05). The maximum odds ratio for CC- and MLO-based risk models were 8.09 and 5.25, respectively, and increased to 11.99 for the fusion risk model. For subgroups of patients aged 37-49years, 50-65years, and 66-87years, the AUCs of 0.73, 0.71, and 0.75 for the fusion risk model were higher than AUC for CC- and MLO-based risk models. For the BIRADS 2 and 3 subgroups, the AUC values were 0.72 and 0.71 respectively for the fusion risk model which were higher than the AUC for the CC- and MLO-based risk models. This study demonstrated that the fusion risk model we established could effectively derive and integrate supplementary and useful information extracted from both CC and MLO view images and adaptively fuse them to increase the predictive power of the short-term breast cancer risk assessment model.

Mammographic breast density and breast cancer risk in the Saudi population: a case-control study using visual and automated methods.

This study aims to establish risk of breast cancer based on breast density among Saudi women and to compare cancer prediction using different breast density methods. 1140 pseudonymised screening mammograms from Saudi females were retrospectively collected. Breast density was assessed using Breast Imaging Reporting and Data System (BI-RADS) density categories and visual analogue scale (VAS) of 285 cases and 855 controls matched on age and body mass index. In a subset of 160 cases and 480 controls density was estimated by two automated methods, Volpara Density™ and predicted VAS (pVAS). Odds ratios (ORs) between the highest and second categories in BI-RADS and Volpara density grades, and highest vs lowest quartiles in VAS, pVAS and Volpara Density™, were estimated using conditional logistic regression. For BI-RADS, the OR was 6.69 (95% CI 2.79-16.06) in the highest vs second category and OR = 4.78 (95% CI 3.01-7.58) in the highest vs lowest quartile for VAS. In the subset, VAS was the strongest predictor OR = 7.54 (95% CI 3.86-14.74), followed by pVAS using raw images OR = 5.38 (95% CI 2.68-10.77) and Volpara Density ™ OR = 3.55, (95% CI 1.86-6.75) for highest vs lowest quartiles. The matched concordance index for VAS was 0.70 (95% CI 0.65-0.75) demonstrating better discrimination between cases and controls than all other methods. Increased mammographic density was strongly associated with risk of breast cancer among Saudi women. Radiologists' visual assessment of breast density is superior to automated methods. However, pVAS and Volpara Density ™ also significantly predicted breast cancer risk based on breast density. Our study established an association between breast density and breast cancer in a Saudi population and compared the performance of automated methods. This provides a stepping-stone towards personalised screening using automated breast density methods.

Weight is More Informative than Body Mass Index for Predicting Postmenopausal Breast Cancer Risk: Prospective Family Study Cohort (ProF-SC).

Our results suggest that the relationship between BMI and breast cancer could be due to a relationship between weight and breast cancer, downgraded by inappropriately adjusting for height; potential importance of anthropometric measures other than total body fat; breast cancer risk associations with BMI and weight are across a continuum.

Abstract ES7-3:/AI

Abstract The tools of AI offer promise to uncover and amplify data, previously hidden in images of the human body, and expand the field of radiomics to more accurately predict breast cancer risk across diverse patient populations. Although deep learning products trained to predict mammographic breast density and risk of breast cancer are now available, their performance in routine clinical settings is largely unknown. Since the creation of the Gail model in 1989, risk models have supported risk-adjusted screening and prevention, and their continued evolution has been a central pillar of breast cancer research. Mammographic breast density is now being incorporated into well-established clinical risk models, which are used to determine eligibility for supplemental imaging and other services for patients at increased risk. However, mammographic breast density assessment is subjective and varies widely between and within radiologists. For example, Sprague et al demonstrated considerable variability of qualitative Breast Imaging Reporting and Data System (BI-RADS) density assessments with a range of 6% to 85% of mammograms being assessed as heterogeneously or extremely dense across 83 radiologists. To limit variability and subjectivity, various automated breast density assessment methods were commercially developed, but clinical evaluation has yielded mixed results. Density is only one limited feature of any woman’s mammogram. Deep learning models can operate over the full resolution of mammogram images to assess a patient’s future breast cancer risk. Rather than manually identifying discriminative image patterns, machine learning models can discover these patterns directly from the data. Specifically, models are trained with full resolution mammograms and the outcome of interest, namely whether the patient developed breast cancer within five years from the date of the examination. Our recent work demonstrates that application of novel artificial intelligence applications to imaging data can significantly improve breast density assessment and breast cancer risk prediction. In addition, unlike traditional models, our DL models perform equally well across varied races, ages, and family histories and we have built a clinical platform which is currently in use to support implementation of our density and risk models into routine clinical care. From this clinical platform, we found that use of a deep learning model more accurately predicts mammographic breast density, reducing human variation and “overcalling” of breast density, and can help healthcare systems more appropriately utilize limited supplemental screening resources as well as provide patients with more accurate information regarding their breast density. We have also found a DL breast cancer risk model, generated in routine clinical mammography screening programs, can predict future breast cancer risk and can more accurately identify women who will most benefit from screening mammography. This DL score can be used in diverse clinical settings that currently depend on traditional risk scores, which are at best modest in performance. DL risk scores based on the mammogram alone can inform shared decision making with patients and their healthcare providers regarding more personalized risk reduction and early detection strategies. Citation Format: C Lehman, B Dontchos, LR Lamb./AI [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr ES7-3.

Abstract P2-10-03: Improving breast cancer risk prediction using a convolutional neural network-based mammographic evaluation in combination with clinical risk factors

Abstract Background: Accurate assessment of a woman’s individualized breast cancer risk is necessary to inform shared decision-making regarding screening and risk-reducing strategies. Recently, deep learning techniques, including convolutional neural networks (CNN), have shown better predictive potential for breast cancer risk compared to mammographic density (MD). We evaluated whether combining clinical factors in the Breast Cancer Surveillance Consortium (BCSC) model, including MD, with a novel CNN-based mammographic evaluation more accurately predicts breast cancer risk than the BCSC model alone in a cohort of racially/ethnically diverse women. Methods: We conducted a retrospective cohort study of 23,552 women, age 35-74 years, who underwent screening mammography from 2014 to 2018 at Columbia University Irving Medical Center in New York City. We extracted data from the electronic health record (EHR) on breast cancer risk factors (age, race/ethnicity, prior benign breast biopsy, first degree family history of breast cancer, and MD). From this cohort, we identified 206 women who developed breast cancer by linkage to the tumor registry. We calculated 5-year invasive breast cancer risk using the BCSC model. We applied CNN-based breast cancer risk model to full-field craniocaudal mammographic views of both breasts, with an output of a risk score (range, 0-1). We used logistic regression models with breast cancer status as the outcome and predictors including clinical factors only (BCSC model) or combined with CNN risk score (hybrid model). We compared the prediction performance of these models via area under the receiver operating characteristics curves (AUCs) based on the DeLong test. We also calculated each model’s AUC for subgroups of age and race/ethnicity. Results: Among 23,552 evaluable women, mean age was 55.9 years (standard deviation [SD], 9.5) with 27% non-Hispanic White, 9% non-Hispanic Black, 36% Hispanic, 5% Asian, and 23% Other/Unknown race/ethnicity. Four percent had a first-degree family history of breast cancer, 10% had a prior benign breast biopsy, 45% had heterogeneously or extremely dense breasts on mammography, and 22% met high-risk criteria based upon a 5-year invasive breast cancer risk 1.67% according to the BCSC model. Mean CNN risk score was higher among breast cancer cases compared to unaffected controls (0.477 vs. 0.466, p=0.077). We found that the hybrid model outperformed the BCSC model (AUC of 0.676 vs. 0.640, respectively; p=0.003). In subgroup analyses, the hybrid model more accurately predicted breast cancer risk compared to the BCSC model among women age&amp;lt;50 (AUC of 0.713 vs. 0.645, respectively; p=0.078) and age&amp;gt;=50 (AUC of 0.663 vs. 0.625, respectively; p=0.026); non-Hispanic Black women (AUC of 0.794 vs. 0.663, respectively; p=0.028) and Hispanic women (AUC of 0.666 vs. 0.621, respectively; p=0.060). Conclusion: Among women undergoing screening mammography, a hybrid model incorporating a CNN-based mammography evaluation with clinical factors from the BCSC model more accurately predicted breast cancer risk relative to the BCSC model alone, particularly among racial and ethnic minorities. Combined with clinical risk factors, our CNN model may be used to efficiently predict breast cancer risk and inform risk-stratified breast cancer screening and prevention strategies. Comparing prediction performance for breast cancer risk of the BCSC model vs. hybrid modelBCSC ModelHybrid ModelP-value*AUC95% CIAUC95% CIAll patients0.6400.602-0.6830.6760.640-0.7110.003Age (years)&amp;lt;500.6450.562-0.7280.7130.648-0.7780.078&amp;gt;=500.6250.579-0.6710.6630.623-0.7030.026Race/ethnicityNon-Hispanic White0.6880.623-0.7540.7040.643-0.7660.230Non-Hispanic Black0.6630.539-0.7880.7940.704-0.8830.028Hispanic0.6210.560-0.6860.6660.597-0.7330.060Asian0.6160.440-0.7910.6510.461-0.8410.694 Citation Format: Alissa Michel, Vicky Ro, Julia E McGuinness, Simukayi Mutasa, Richard Ha, Katherine D Crew. Improving breast cancer risk prediction using a convolutional neural network-based mammographic evaluation in combination with clinical risk factors [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P2-10-03.

Abstract P3-13-01: Association of polygenic risk score with 2 year risk of poor prognosis breast cancer

Abstract Introduction: Mammography reduces breast cancer mortality, however, there is controversy surrounding when and how often women should undergo mammography screening. Knowledge of short-term risk of developing breast cancer, particularly poor prognosis breast cancer, would help direct more intensive screening to those at highest risk. Polygenic risk scores (PRS) are emerging as a powerful tool to predict breast cancer risk, however, few studies have evaluated the associations of breast cancer PRS with short-term risk or with risk of poor prognosis breast cancers specifically. Methods: Using a mammography screening cohort at Massachusetts General Hospital, we identified women who had a negative mammogram (BI-RADS assessment 1 or 2) from 2006 to 2015. We linked the cohort to a research biobank to obtain genetic information. In addition, we recruited 205 patients who developed breast cancer within 2 years of a negative mammogram to provide a DNA sample. Women with a prior history of breast cancer, with breast implants, and who were not residents of Massachusetts were excluded. Samples were genotyped using the Illumina Multi-Ethnic GWAS/Exome SNP (MEGA) array. Genotypes were imputed using TOPMed (Version r2 2020). Patients whose saliva DNA samples had low concentration or that failed quality control procedures were excluded. Ancestry specific principal components were generated and used as a covariate. We generated the 313-SNP breast cancer PRS, the estrogen receptor positive (ER+) PRS and the estrogen receptor negative (ER-) PRS using established methods. (PMID: 30554720) Breast cancers were defined as poor prognosis if they were metastatic, had positive lymph nodes, were ER/PR+HER2- and &amp;gt 2cm or ER/PR/HER2- or HER2+ and &amp;gt 1cm. (PMID: 33169794) Logistic regression was used to estimate the odds ratios for standardized PRS measures, adjusted for age, breast density, race/ethnicity, year of screening, and ancestry principal components. Results: After exclusions, 308 breast cancers and 3329 non-cases were analyzed. Of the breast cancers, 86% were ER/PR+ (264/308) and 14% were ER/PR- (42/308), and 137 (44%) were poor prognosis. The majority of patients were non-Hispanic White (87%) and the mean age was 57 years and was similar for cancers and non-cancers. Cancer cases were more likely than non-cases to have higher breast density and a family history of breast cancer. First, we examined the overall breast cancer PRS, and found the PRS was significantly associated with breast cancer diagnosed within two years of a negative mammogram (OR=1.39, 95% CI 1.23-1.57, p &amp;lt 0.001). The PRS was also significantly associated specifically with diagnosis of poor prognosis disease (OR=1.21, 95% CI 1.01-1.45, p=0.037). In addition, the ER+ PRS was significantly associated with ER/PR+ breast cancer (OR=1.41, 95% CI 1.24-1.61, p &amp;lt 0.001), and the ER- PRS was significantly associated with ER- breast cancer (OR=1.48, 95% CI 1.08-2.02, p=0.015). Conclusion: Even after adjusting for breast density and other risk factors, breast cancer PRS was significantly associated with diagnosis of both breast cancer overall and poor prognosis breast cancer within two years of a negative mammogram. Furthermore, the subtype specific PRS were significantly associated with short-term risk of ER+ and ER- disease. These results suggest that PRS may be useful in guiding decisions about screening interval and supplemental screening, given the association of PRS with risk of poor prognosis disease in the short term. Table 1.Logistic Regression of PRS and cancer diagnosis within 2 years of a negative mammogramOR95% CIp-valueAll Cancers (N=308), Overall PRS1.391.23-1.571.78x10-7Poor prognosis (N=147), Overall PRS1.211.01-1.450.037ER+ cancers (N=264), ER+ PRS1.421.24-1.621.87x10-7ER- cancers (N=42), ER- PRS1.521.11-2.090.008 Citation Format: Anne Marie McCarthy, Alisa K. Manning, Sarah Hsu, Beverly Moy, Constance D. Lehman, Katrina Armstrong. Association of polygenic risk score with 2 year risk of poor prognosis breast cancer [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P3-13-01.

Abstract P2-10-09: Oxysterol profiling in breast adipose tissue identifies brassicasterol and 24-hydroxycholesterol as breast cancer risk predictors

Abstract Background: Hyperlipidemia and serum oxysterol levels are known to be associated with breast cancer risk. In murine model, altered oxysterol levels in adipose tissue were identified in obese mice. However, the significance of oxysterol levels in human adipose tissue and their association with breast cancer risk were largely unknown. In this study, we aimed to quantify the levels of different oxysterols in adipose tissue and to establish their association with breast cancer risk. Methods: The breast adipose tissue was collected from patients who underwent surgery at a tertiary medical center for benign breast disease or breast cancer. We have profiled oxysterol levels by liquid chromatography-mass spectrometry. Multivariate analysis was used to analyze the relationship of adipose tissue oxysterol levels and breast cancer risk. Result: Breast adipose tissue from 279 subjects (benign low risk, n=58; benign high risk, n=33, and breast cancer, n=188) collected between March 2009 and November 2014 were included in this analysis. Parity is a protective factor both for breast cancer in premenopausal and postmenopausal groups. Brassicasterol ≤0.28 ng/mg (OR=6.739, 95% C.I.=2.092-21.707, p=0.001) and age &amp;gt;40.06 years old (OR=3.774, 95% C.I.=1.328-10.722, p=0.013) in premenopausal patients; high log (7α-. hydroxycholesterol) levels (OR=6.690, 95% C.I.=1.915-23.369, p=0.003) and 24-hydroxycholesterol &amp;gt;0.92ng/mg (OR=4.539, 95% C.I.=1.153-17.866, p=0.030) in postmenopausal patients were associated with increased BC risk. Sitosterol ≤3.03ng/mg (OR=0.057, 95% C.I.= 0.016-0.205, p &amp;lt;0.0001) and brassicasterol level (OR=0.011, 95% C.I.= 0.001-0.222, p=0.003) were most important protective factors for BC in premenopausal and postmenopausal patients, respectively. Conclusion: Taken together, our study is the first cohort to prospectively profile oxysterol levels in breast adipose tissue. Brassicasterol and 24-hydroxycholesterol levels in adipose tissue predict breast cancer risk both in premenopausal and postmenopausal patients. Citation Format: I-Chun Chen, Wen-Hung Kuo, Fu-Chang Hu, Ming-Yang Wang, Yen-Shen Lu, Dieter Lütjohann, Ching-Hung Lin. Oxysterol profiling in breast adipose tissue identifies brassicasterol and 24-hydroxycholesterol as breast cancer risk predictors [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P2-10-09.