Winham, Stacey J.

Abstract Background: The Breast Cancer Surveillance Consortium (BCSC) model predicts invasive breast cancer risk in women with Benign Breast Disease (BBD), and recently incorporated BBD histology into the model. The BCSC has been validated in the Mayo Mammography Health Study, but has yet to be examined in Mayo Clinic’s BBD cohort. The Benign Breast Disease to Breast Cancer (BBD-BC) model predicts risk of both invasive and in situ breast cancer. Here we compare the performance of the BCSC and BBD-BC in Mayo Clinic’s BBD cohort. Methods: Eligible women underwent a breast biopsy with benign findings at the Mayo Clinic between years 1997-2001 and had a 4-view screening mammogram within six months of biopsy. Clinical BI-RADS density assessments using the 4th edition American College of Radiology were available on all mammograms, and were coded as almost entirely fat, scattered fibroglandular densities, heterogeneously dense, and extremely dense. Risk at 5 and 10 years of invasive cancer only (BCSC model) or both invasive and in situ cancer (BBD-BC model) were estimated. In situ cancers were censored at time of diagnosis for both models. Concordance statistics, i.e. model discrimination (higher is better), for each model were calculated using a Cox proportional hazards model with predicted risk as the sole predictor, and were compared using permutation tests. The ratio of total predicted (sum of predicted risk) to observed number of invasive breast cancers was used to assess model calibration. Calibration was not formally compared across models due to differences in how DCIS was treated in the development of each (BCSC censored; BBD-BC event). Results: 999 women met inclusion criteria. BI-RADS density was categorized as fatty in 46 (4.6%), scattered densities in 372 (37.2%), heterogeneously dense in 423 (42.3%), and extremely dense in 158 (15.8%). 62 invasive cancers occurred over a median 13.3yrs of follow-up, with 16 (25.8%) and 48 (77.4%) of the cancers occurring with-in 5 and 10 years. The concordance of the BCSC at 5 years was 0.550 (95% CI 0.409—0.691), compared to 0.719 (95% CI 0.578—0.860) for the BBD-BC (p-value=0.005). At 10 years the BCSC concordance was not significantly different from the BBD-BC, at 0.624 (95% CI 0.542—0.706) and 0.662 (95% CI 0.580—0.744), respectively (p-value=0.306). The BCSC over predicted the number invasive cancers in the BBD cohort at 5 years (predicted-to-observed=1.54; 95% CI 1.01—2.70), but was well calibrated at 10 years (1.06; 95% CI 0.83, 1.47). Conclusions: The BCSC model performed reasonably well in the BBD Cohort 10 years post-biopsy, but overestimated risk at 5 years. Additional study is needed to improve models for prediction of breast cancer risk among women with BBD. Citation Format: Ryan D. Frank, Celine M. Vachon, Stacey J. Winham, Robert A. Vierkant, Marlene H. Frost, Derek C. Radisky, Daniel W. Visscher, Amy C. Degnim. Validation of the BCSC model within the Mayo Benign Breast Disease Cohort [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 5299. doi:10.1158/1538-7445.AM2017-5299

Abstract High grade serous ovarian cancer (HGSOC) is a complex disease in which initiation and progression have been associated with gene mutation, DNA methylation changes, genetic variation, and epigenetic processes. Variation in several susceptibility regions and cancer-typical global methylation patterns have been observed in HGSOC; however, this knowledge has not been compelling in understanding HGSOC intiation or progression. As ingetration of genomic, epigenomic, and transcriptomic data has increased mechanistic understanding in other cancers, we hypothesized that tumor methylation alone or in combination with germline genetic variation influences tumor gene expression in HGSOC. We examined three nested models using an Elastnic Net (ENET) penalized regression method while adjusting for somatic copy number (CNV): a) germline genotype and tumor DNA methylation (full model), b) genotype only, and c) DNA methylation only. We included 339 cases from The Cancer Genome Atlas (TCGA), 54 cases from Mayo Clinic, and 78 cases from the Australian Ovarian Cancer Study (AOCS). Genotyping and copy number calls on germline DNA, expression, methylation and copy number on somatic samples were collected and analyzed on different platforms separately at each study site. We excluded genes with low overall expression and thus analyzed a total of 11,922 genes available in three datasets ( Ensembl IDs, 500kb window up- and downstream). In general, combining genomic data in HGSOC did not reveal a role for germline genetic variation in altering gene expression. However, in methylation only models 79 genes were associated with differential expression in the TCGA cases (permutation multiple testing adjusted p-val <0.05), in the Mayo cases (unadjusted p-val <0.05) and AOCS cases (unadjusted p-val <0.05). A known tummor suppressor (FBXW7) was associated with differential expression in the three datasets at p-val <0.01. This work demonstrates the feasibility, utility, and statistical power of ENET gene-level analyses incoporating maximal genomic information. Citation Format: Yanina Natanzon, Madalene Earp, Julie M. Cunningham, Kimberly R. Kalli, Stacey J. Winham, Sebastian M. Armasu, Melissa C. Larson, Chen Wang, David D. Bowtell, Dale W. Garsed, Ellen Goode. Omics data integration analysis in high grade serous ovarian cancer: results from three studies [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 3368. doi:10.1158/1538-7445.AM2017-3368

Abstract Introduction: In females, X-chromosome inactivation (XCI) epigenetically silences transcription of one copy of the X chromosome. Which chromosome is silenced is randomly selected, and is tissue- and cell-specific. While some genes are known to escape XCI under normal conditions, aberrant XCI patterns are thought to occur in female-specific cancers, although the role of XCI in ovarian tumorigenesis and progression is largely unknown. The process of XCI is complex, and integration of gene expression, DNA methylation, and copy number data can inform the XCI status of individual genes and chromosome-wide XCI patterns for individual patients. Methods: We evaluated gene- and chromosome-level patterns of XCI by integrating RNA sequence, copy number alteration, genotype, and DNA methylation data to study XCI escape patterns in tumor samples from 99 ovarian cancer patients. We measured allele-specific expression (ASE) for 397 X-linked genes to identify the active alleles for each tumor. Combining ASE data with knowledge of copy number status, we used a Bayesian beta-binomial mixture model to estimate which genes escaped XCI for each patient, and validated our findings using DNA methylation data. To assess global XCI patterns, we performed cluster analyses on the ASE and methylation data, after adjusting for loss of heterozygosity. We examined the relationship between the clusters and clinical factors, including overall survival and time to recurrence. Results: DNA promoter methylation demonstrated inverse regional correlations with ASE. Cluster analyses using ASE and methylation data demonstrated evidence of two tumor clusters, representing normal XCI and global XCI dysregulation. The dysregulated XCI cluster (N=52) was associated with lower X-inactive specific transcript expression as expected (p<0.01). Patients with XCI dysregulated tumors were higher grade, stage, serous histology and were sub-optimally debulked (p<0.05). These patients also had shorter overall survival time (HR=1.87, p=0.02) and time to recurrence (HR=2.34, p<0.01), although associations were attenuated after covariate adjustment. In 45 tumor samples with sufficient data, we observed escape patterns largely consistent with previous reports of multiple tissue types. When comparing tumor to normal ovarian tissue, eight genes (CXorf23, CXorf36, BRWD3, ELF4, SLITRK4, GABRE, CLCN4, SH3BGRL) showed putative escape in the tumor and two genes (RBBP7, OFD1) showed discrepant tumor inactivation. Conclusions: We identified discrepant gene-level XCI tumor classifications compared to normal tissue and identified a group of patients with chromosome-wide XCI dysregulation associated with worse clinical prognosis. This provides evidence of the role of XCI in ovarian cancer and highlights the need to integrate multiple genomic data types to study XCI. Citation Format: Stacey J. Winham, Nicholas B. Larson, Sebastian M. Armasu, Zachary C. Fogarty, Melissa C. Larson, Kimberly R. Kalli, Kate Lawrenson, Simon Gayther, Brooke L. Fridley, Ellen L. Goode. Integrative analyses of gene expression, DNA methylation, genotype and copy number alterations characterize X-chromosome inactivation in ovarian cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 2420. doi:10.1158/1538-7445.AM2017-2420

2 Background: Women with atypical hyperplasia (AH) on breast biopsy have an aggregate increased risk of breast cancer (BC), but accurate personalized risk prediction is desirable to facilitate individual clinical management decisions. Currently used models provide poor BC risk prediction for women with AH. Our goal was to develop and validate an improved risk prediction model for women with AH. Methods: From a cohort of 13,538 women with benign breast disease from 1967-2001, pathology review confirmed 699 with AH. Clinical risk factors and histologic features of the tissue biopsy were recorded, and BC events were ascertained from study questionnaires, tumor registry, and review of medical records. Using a lasso approach, 23 variables were assessed for model inclusion. Lasso-identified features were then fit in a Cox regression model to estimate BC risk. Model discrimination was assessed with C-statistics in the model-building set and in a separate external validation set. Calibration was assessed by comparing observed to predicted breast cancer counts. Results: The model-building set comprised 699 women with 142 BC events (median follow-up 8.1 years), and the external validation set comprised 461 women with 114 BC events (median follow-up 11.4 years). The final model included three covariates: age at biopsy, age squared, and number of foci of AH. Model performance was good, with a C-statistic of 0.622 (SE = 0.027) in the model-building set and 0.594 (SE = 0.029) in the external validation set. The model is well-calibrated, with observed to expected numbers of BCs nearly equal across all post-biopsy follow-up years. Conclusions: We propose a new model for predicting BC risk in women with AH based on age at biopsy and number of foci of atypia. This model provides absolute risk estimates for women with AH, has good discriminatory ability, is well-calibrated, and validates in an external cohort.

18 Background: Benign breast disease (BBD) is diagnosed in 1-2 million women/year in the US, and while these patients are known to be at substantially increased risk for subsequent development of breast cancer, existing models for risk assessment perform poorly at the individual level. Here, we describe a DNA-microarray-based transcriptional model for breast cancer risk prediction for patients with sclerosing adenosis (SA), a lesion found in ¼ of all BBD patients that is characterized by epithelial proliferation, disordered acinar architecture, and stromal fibrosis. Methods: A training set was developed from 86 patients diagnosed with sclerosing adenosis (SA), of which 27 subsequently developed cancer within 10 years (cases) and 59 remained cancer-free at 10 years (controls). A diagonal linear discriminate analysis (DLDA)-prediction model for prediction of cancer within 10 years (SA TTC10) was generated from transcriptional profiles of formalin fixed, paraffin-embedded (FFPE) biopsy-derived RNA. This model was tested on a separate validation case-control set composed of 65 SA patients. Results: The SA TTC10 gene signature model, composed of 35 gene features, achieved a clear and significant separation between case and control with receiver operating characteristic area under the curve of 0.913 in the training set and 0.836 in the validation set. Conclusions: Our results provide the first demonstration that benign breast tissue contains transcriptional alterations that indicate risk of breast cancer development, and that essential precursor biomarkers of malignancy are present many years prior to cancer development. Furthermore, the SA TTC10 gene signature model, which can be assessed on FFPE biopsies, constitutes a novel prognostic biomarker for patients with SA.

Abstract Background: Reductions in breast density with tamoxifen and aromatase inhibitors may be an intermediate marker of treatment response. We compare changes in volumetric breast density among breast cancer cases using tamoxifen or aromatase inhibitors (AI) to untreated women without breast cancer. Methods: Breast cancer cases with a digital mammogram prior to diagnosis and after initiation of tamoxifen (n = 366) or AI (n = 403) and a sample of controls (n = 2170) were identified from the Mayo Clinic Mammography Practice and San Francisco Mammography Registry. Volumetric percent density (VPD) and dense breast volume (DV) were measured using Volpara (Matakina Technology) and Quantra (Hologic) software. Linear regression estimated the effect of treatment on annualized changes in density. Results: Premenopausal women using tamoxifen experienced annualized declines in VPD of 1.17% to 1.70% compared with 0.30% to 0.56% for controls and declines in DV of 7.43 to 15.13 cm3 compared with 0.28 to 0.63 cm3 in controls, for Volpara and Quantra, respectively. The greatest reductions were observed among women with ≥10% baseline density. Postmenopausal AI users had greater declines in VPD than controls (Volpara P = 0.02; Quantra P = 0.03), and reductions were greatest among women with ≥10% baseline density. Declines in VPD among postmenopausal women using tamoxifen were only statistically greater than controls when measured with Quantra. Conclusions: Automated software can detect volumetric breast density changes among women on tamoxifen and AI. Impact: If declines in volumetric density predict breast cancer outcomes, these measures may be used as interim prognostic indicators. Cancer Epidemiol Biomarkers Prev; 26(6); 930–7. ©2017 AACR.