A population-level digital histologic biomarker for enhanced prognosis of invasive breast cancer

Breast cancer is a heterogeneous disease with variable survival outcomes. Pathologists grade the microscopic appearance of breast tissue using the Nottingham criteria, which are qualitative and do not account for noncancerous elements within the tumor microenvironment. Here we present the Histomic Prognostic Signature (HiPS), a comprehensive, interpretable scoring of the survival risk incurred by breast tumor microenvironment morphology. HiPS uses deep learning to accurately map cellular and tissue structures to measure epithelial, stromal, immune, and spatial interaction features. It was developed using a population-level cohort from the Cancer Prevention Study-II and validated using data from three independent cohorts, including the Prostate, Lung, Colorectal, and Ovarian Cancer trial, Cancer Prevention Study-3, and The Cancer Genome Atlas. HiPS consistently outperformed pathologists in predicting survival outcomes, independent of tumor-node-metastasis stage and pertinent variables. This was largely driven by stromal and immune features. In conclusion, HiPS is a robustly validated biomarker to support pathologists and improve patient prognosis.

A population-level computational histologic signature for invasive breast cancer prognosis

Breast cancer is a heterogeneous disease with variable survival outcomes. Pathologists grade the microscopic appearance of breast tissue using the Nottingham criteria, which is qualitative and does not account for non-cancerous elements within the tumor microenvironment (TME). We present the Histomic Prognostic Signature (HiPS), a comprehensive, interpretable scoring of the survival risk incurred by breast TME morphology. HiPS uses deep learning to accurately map cellular and tissue structures in order to measure epithelial, stromal, immune, and spatial interaction features. It was developed using a population-level cohort from the Cancer Prevention Study (CPS)-II and validated using data from three independent cohorts, including the PLCO trial, CPS-3, and The Cancer Genome Atlas. HiPS consistently outperformed pathologists' performance in predicting survival outcomes, independent of TNM stage and pertinent variables. This was largely driven by stromal and immune features. In conclusion, HiPS is a robustly validated biomarker to support pathologists and improve prognosis.

Racial Disparities in Flavored Tobacco Product Use, Curiosity, Susceptibility, and Harm Perception, National Youth Tobacco Survey 2019-2020

Introduction

Studies characterizing differences in youth flavored tobacco product use prevalence, curiosity/susceptibility, and harm perceptions by race and ethnicity are limited. This study comprehensively examines flavored tobacco product use and harm perceptions among U.S. middle and high school students, by race and ethnicity.

Methods

Data came from the 2019 (N=19,018) and 2020 (N=14,531) National Youth Tobacco Surveys (NYTS). Weighted prevalence estimates of flavored tobacco product use and curiosity, susceptibility, and harm perception are reported by race and ethnicity (non-Hispanic [NH] White, NH Black, Hispanic, or NH Other). t-Tests assessed differences in prevalence by years and racial/ethnic groups.

Results

Among youth with past 30-day tobacco use, use of most flavored tobacco products increased across all racial/ethnic groups; the largest increase was observed among Hispanic youth using other flavored tobacco products (30.3%). The group with the highest susceptibility to future electronic cigarette (e-cigarette) use was Hispanic students (42.3%). Hispanic students had the highest curiosity about and susceptibility to future use of cigarettes and cigars as well.

Conclusions

Increases in the use of and higher susceptibility to other flavored tobacco products, particularly among Hispanic youth, suggest a need for additional changes in environmental conditions and possibly targeted or tailored tobacco control interventions for Hispanic youth.

Implications

Given that flavored tobacco use is prevalent among youth and aggressively marketed more to racial/ethnic minority populations, it is important to understand how susceptibility and perceptions relate to tobacco use. Our results suggest a need for a better understanding of social and environmental factors that drive tobacco use behaviors and perceptions, particularly among Hispanic youth, to address the root causes of these differences and create more equitable tobacco control interventions.

Awareness and Use of Heated Tobacco Products Among Middle School and High School Students, United States, 2019-2020

PURPOSE

Heated tobacco products (HTP) heat-processed tobacco leaf into an aerosol inhaled by the user. This study assessed prevalence and correlates of HTP awareness, ever use, and current use among US middle and high school students.

METHODS

Data came from the 2019 and 2020 National Youth Tobacco Survey, a cross-sectional survey of US public and private, middle and high school students. HTP awareness, ever use, and current (past 30-day) use were assessed. Weighted prevalence estimates and adjusted prevalence ratios (aPR) were assessed overall and by sex, school level, race/ethnicity, and current other tobacco product use.

RESULTS

In 2019, 12.8% (3.44 million) of all students reported HTP awareness, increasing to 19.3% (5.29 million) in 2020 (p < .01). Ever [2019: 2.6% (630 000); 2020: 2.4% (620 000)] and current [2019: 1.6% (420 000); 2020: 1.4% (370 000)] HTP use did not significantly change from 2019 to 2020. Current e-cigarette users were more likely to report ever (2020 aPR = 1.79, 95% CI:1.23, 2.62) or current HTP use (2019 aPR = 5.16, 95% CI: 3.48, 7.67; 2020 aPR = 3.39, 95% CI: 2.10, 5.47) than nonusers. In both years, ever and current HTP use was more likely among current combustible (aPR range = 3.59-8.17) and smokeless tobacco product (aPR range = 2.99-4.09) users than nonusers.

CONCLUSIONS

HTP awareness increased 51% among US students during 2019-2020; however, HTP use did not significantly change during this period. Students who used other tobacco products were more likely to currently use HTPs. Estimates of HTP awareness and use provided serve as a baseline as future monitoring of these products is warranted.

IMPLICATIONS

Awareness of heated tobacco products (HTPs) increased among US youth from 2019 to 2020; however, HTP use did not change. These estimates of HTP awareness and use serve as a baseline for future surveillance of these products as their availability in the US increases.

Youth Indoor and Outdoor Exposure to Secondhand Smoke and Secondhand Aerosol

INTRODUCTION

Secondhand smoke and secondhand aerosol exposure are important public health concerns. This is the first study to present separate estimates of public indoor and outdoor secondhand smoke and secondhand aerosol exposure among U.S. youth.

METHODS

Data came from the 2020 National Youth Tobacco Survey, an annual cross-sectional survey of U.S. students in Grades 6-12. Self-reported past 30-day indoor and outdoor secondhand smoke and secondhand aerosol exposures were assessed separately. Weighted prevalence and adjusted prevalence ratios for each outcome were assessed among students overall and stratified by sex, school level, race/ethnicity, sexual orientation, and current tobacco product use; prevalence also was calculated among those who did not currently use tobacco.

RESULTS

Exposure to secondhand smoke and secondhand aerosol was reported by 60.6% (95% CI=58.7, 62.4) and 44.5% (95% CI=42.1, 46.9) of U.S. youth, respectively. Among all students, 37.6% (95% CI=36.0, 39.2) and 53.3% (95% CI=51.4, 55.2) reported indoor and outdoor secondhand smoke exposure, respectively; 34.9% (95% CI=32.9, 37.4) and 36.8% (95% CI=34.6, 38.9) reported indoor and outdoor secondhand aerosol exposure, respectively. After adjustment, female versus male students (adjusted prevalence ratio=1.15-1.30) and those who currently use versus do not use combustible tobacco products (adjusted prevalence ratio=1.15-1.36) were more likely to report exposure to all outcomes.

CONCLUSIONS

Approximately 1 in 2 students overall reported outdoor secondhand smoke exposure, and 1 in 3 students reported exposures to each indoor secondhand smoke, indoor secondhand aerosol, and outdoor secondhand aerosol. Separate estimates of indoor and outdoor secondhand smoke and secondhand aerosol exposure, along with data on correlates of exposure, provide information to support comprehensive indoor and outdoor smoke-free policies.

Use of Tobacco Products, Alcohol, and Other Substances Among High School Students During the COVID-19 Pandemic - Adolescent Behaviors and Experiences Survey, United States, January-June 2021

The COVID-19 pandemic has been associated with established risk factors for adolescent substance use, including social isolation, boredom, grief, trauma, and stress. However, little is known about adolescent substance use patterns during the pandemic. CDC analyzed data from the Adolescent Behaviors and Experiences Survey, an online survey of a probability-based, nationally representative sample of public- and private-school students in grades 9–12 (N = 7,705), to examine the prevalence of current use of tobacco products, alcohol, and other substances among U.S. high school students. Prevalence was examined by demographic characteristics and instructional models of the students’ schools (in-person, virtual, or hybrid). During January–June 2021, 31.6% of high school students reported current use of any tobacco product, alcohol, or marijuana or current misuse of prescription opioids. Current alcohol use (19.5%), electronic vapor product (EVP) use (15.4%), and marijuana use (12.8%) were more prevalent than prescription opioid misuse (4.3%), current cigarette smoking (3.3%), cigar smoking (2.3%), and smokeless tobacco use (1.9%). Approximately one third of students who used EVPs did so daily, and 22.4% of students who drank alcohol did so ≥6 times per month. Approximately one in three students who ever used alcohol or other drugs reported using these substances more during the pandemic. The prevalence of substance use was typically higher among non-Hispanic American Indian or Alaska Native students, older students, and gay, lesbian, or bisexual students than among students of other racial or ethnic groups, younger students, and heterosexual students. The prevalence of alcohol use also was higher among non-Hispanic White students than those of other racial or ethnic groups. Students only attending school virtually had a lower prevalence of using most of the substances examined than did students attending schools with in-person or hybrid models. These findings characterizing youth substance use during the pandemic can help inform public health interventions and messaging to address these health risks during and after the COVID-19 pandemic.

Use of Tobacco Products, Alcohol, and Other Substances Among High School Students During the COVID-19 Pandemic - Adolescent Behaviors and Experiences Survey, United States, January-June 2021

The COVID-19 pandemic has been associated with established risk factors for adolescent substance use, including social isolation, boredom, grief, trauma, and stress. However, little is known about adolescent substance use patterns during the pandemic. CDC analyzed data from the Adolescent Behaviors and Experiences Survey, an online survey of a probability-based, nationally representative sample of public- and private-school students in grades 9-12 (N = 7,705), to examine the prevalence of current use of tobacco products, alcohol, and other substances among U.S. high school students. Prevalence was examined by demographic characteristics and instructional models of the students' schools (in-person, virtual, or hybrid). During January-June 2021, 31.6% of high school students reported current use of any tobacco product, alcohol, or marijuana or current misuse of prescription opioids. Current alcohol use (19.5%), electronic vapor product (EVP) use (15.4%), and marijuana use (12.8%) were more prevalent than prescription opioid misuse (4.3%), current cigarette smoking (3.3%), cigar smoking (2.3%), and smokeless tobacco use (1.9%). Approximately one third of students who used EVPs did so daily, and 22.4% of students who drank alcohol did so ≥6 times per month. Approximately one in three students who ever used alcohol or other drugs reported using these substances more during the pandemic. The prevalence of substance use was typically higher among non-Hispanic American Indian or Alaska Native students, older students, and gay, lesbian, or bisexual students than among students of other racial or ethnic groups, younger students, and heterosexual students. The prevalence of alcohol use also was higher among non-Hispanic White students than those of other racial or ethnic groups. Students only attending school virtually had a lower prevalence of using most of the substances examined than did students attending schools with in-person or hybrid models. These findings characterizing youth substance use during the pandemic can help inform public health interventions and messaging to address these health risks during and after the COVID-19 pandemic.

Racial and Ethnic Differences in Marijuana Use in e-Cigarettes Among US Youth in 2017, 2018, and 2020

This study reports the prevalence of ever use of marijuana in e-cigarettes among US youth in 2020 and changes in prevalence from 2017 to 2020 by racial and ethnic groups.

Breast cancer risk factors by mode of detection among screened women in the Cancer Prevention Study-II

BACKGROUND

Identifying risk factors for women at high risk of symptom-detected breast cancers that were missed by screening would enable targeting of enhanced screening regimens. To this end, we examined associations of breast cancer risk factors by mode of detection in screened women from the Cancer Prevention Study (CPS)-II Nutrition Cohort.

METHODS

Among 77,206 women followed for a median of 14.8 years, 2711 screen-detected and 1281 symptom-detected breast cancer cases were diagnosed. Multivariable-adjusted associations were estimated using joint Cox proportional hazards regression models with person-time calculated contingent on screening.

RESULTS

Factors associated with higher risks of symptom-detected and screen-detected breast cancer included current combined hormone therapy (HT) use (HR 2.07, 95% CI 1.72-2.48 and 1.45, 1.27-1.65, respectively) and history of benign breast disease (1.85, 1.64-2.08 and 1.43, 1.31-1.55, respectively). Current estrogen-only HT use was associated with symptom-detected (1.40, 1.15-1.71) but not screen-detected (0.95, 0.83-1.09) breast cancer. Higher risk of screen-detected but not symptom-detected breast cancer was observed for obese vs. normal body mass index (1.22, 1.01-1.48 and 0.76, 0.56-1.01, respectively), per 3 h/day sitting time (1.10, 1.04-1.16 and 0.97, 0.89-1.06, respectively), and ≥ 2 drinks per day vs. nondrinker (1.40, 1.16-1.69 and 1.27, 0.97-1.66, respectively).

CONCLUSIONS

Differences in risk factors for symptom-detected vs. screen-detected breast cancer were observed and most notably, use of combined and estrogen-only HT and a history of benign breast disease were associated with increased risk of symptomatic detected breast cancer.

IMPACT

If confirmed, these data suggest that such women may benefit from more intensive screening to facilitate early detection.

Epidemiologic risk factors for in situ and invasive ductal breast cancer among regularly screened postmenopausal women by grade in the Cancer Prevention Study-II Nutrition Cohort

PURPOSE

Histopathologic grade provides an integrated measure of biologic features which affects cancer prognosis. In invasive ductal breast cancer (IDBC), the grade of the ductal carcinoma in situ (DCIS) and invasive components are usually concordant, suggesting grade is established early in tumorigenesis and may be linked to etiologic factors. In this study, we used prospectively collected data from postmenopausal women in the Cancer Prevention Study-II (CPS-II) Nutrition Cohort to compare risk factor associations among low-grade and high-grade DCIS, as well as low-grade and high-grade IDBC.

METHODS

Among 73,825 cancer-free women at enrollment in the CPS-II Nutrition Cohort in 1992-1993 (mean age: 62.1 years), we verified 802 diagnosed with DCIS (C50 8500/2; n = 430 low-grade and 372 high-grade) and 3,125 with IDBC (C50 8500/3; n = 2,221 low-grade and 904 high-grade) through June 2013. Person-time contribution was conditional on screening mammograms self-reported on biennial surveys. Multivariable-adjusted joint Cox proportional hazards regression models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI).

RESULTS

A personal history of benign breast disease was more strongly associated with higher risk of low-grade DCIS (HR = 2.20, 95% CI 1.81-2.67; p for heterogeneity = 0.0004) than high-grade DCIS. Consumption of two or more alcoholic drinks/day was only associated with a higher risk of low-grade IDBC (HR = 1.58, 95% CI 1.33-1.88; p for heterogeneity = 0.005).

CONCLUSIONS

These results suggest heterogeneity by grade for breast cancer etiology. Identification of potential risk factor differences among low-grade and high-grade DCIS and IDBC may help to clarify associations, and ultimately, improve breast cancer risk prediction models.

Using Digital Pathology to Understand Epithelial Characteristics of Benign Breast Disease among Women Undergoing Diagnostic Image-Guided Breast Biopsy

Delayed terminal duct lobular unit (TDLU) involution is associated with elevated mammographic breast density (MD). Both are independent breast cancer risk factors among women with benign breast disease (BBD). Prior digital analyses of normal breast tissues revealed that epithelial nuclear density (END) and TDLU involution are inversely correlated. Accordingly, we examined associations of END, TDLU involution, and MD in BBD clinical biopsies. This study included digitized images of 262 representative image-guided hematoxylin and eosin-stained biopsies from 224 women diagnosed with BBD, enrolled within the cross-sectional BREAST-Stamp project that were visually assessed for TDLU involution (TDLU count/100 mm², median TDLU span and median acini count per TDLU). A digital algorithm estimated nuclei count per unit epithelial area, or END. Single X-ray absorptiometry of prebiopsy ipsilateral craniocaudal digital mammograms measured global and localized MD surrounding the biopsy region. Adjusted ordinal logistic regression models assessed relationships between tertiles of TDLU and END measures. Analysis of covariance examined mean differences in MD across END tertiles. TDLU measures were positively associated with increasing END tertiles [TDLU count/100 mm², OR_T3vsT1: 3.42, 95% confidence interval (CI), 1.87-6.28; acini count/TDLU_T3vsT1, OR: 2.40, 95% CI, 1.39-4.15]. END was significantly associated with localized, but not, global MD. Relationships were most apparent among patients with nonproliferative BBD. These findings suggest that quantitative END reflects different but complementary information to the histologic information captured by visual TDLU and radiologic MD measures and merits continued evaluation in assessing cellularity of breast parenchyma to understand the etiology of BBD.

Oncotype DX recurrence score implications for disparities in chemotherapy and breast cancer mortality in Georgia

Among women diagnosed with stage I-IIIa, node-negative, hormone receptor (HR)-positive breast cancer (BC), Oncotype DX recurrence scores (ODX RS) inform chemotherapy treatment decisions. Differences in recurrence scores or testing may contribute to racial disparities in BC mortality among women with HR+ tumors. We identified 12,081 non-Hispanic White (NHW) and non-Hispanic Black (NHB) BC patients in Georgia (2010-2014), eligible to receive an ODX RS. Logistic regression was used to estimate the odds of chemotherapy receipt by race and ODX RS. Cox proportional hazard regression was used to calculate the hazard ratios (HRs) comparing BC mortality rates by race and recurrence score. Receipt of Oncotype testing was consistent between NHB and NHW women. Receipt of chemotherapy was generally comparable within strata of ODX RS-although NHB women with low scores were slightly more likely to receive chemotherapy (OR = 1.16, 95% CI 0.77, 1.75), and NHB women with high scores less likely to receive chemotherapy (OR = 0.77, 95% CI 0.48, 1.24), than NHW counterparts. NHB women with a low recurrence score had the largest hazard of BC mortality (HR = 2.47 95% CI 1.22, 4.99) compared to NHW women. Our data suggest that additional tumor heterogeneity, or other downstream treatment factors, not captured by ODX, may be drivers of racial disparities in HR+ BC.

Abstract 2417: Unsupervised deep-learning to identify histopathological features among breast cancers in the Cancer Prevention Study-II Nutrition Cohort

Background: Machine learning (ML) methods are becoming more feasible for use in clinical and epidemiologic research of breast cancer, particularly when characterizing histopathology. Compared to supervised ML methods, unsupervised approaches represent an opportunity to distinguish features heretofore unknown. The purpose of this study was to use unsupervised deep learning methods to identify histopathological features in diagnostic breast cancer hematoxylin and eosin (H&E) slides that are associated with clinical characteristics and patient outcomes.

Methods: One H&E slide was scanned (Leica Biosystems Aperio Versa scanner) at 20x magnification for each of 1,716 women diagnosed with breast cancer from the Cancer Prevention Study-II Nutrition Cohort. In the pre-processing phase, the scanned images underwent color normalization, artifact detection, and tiling. We then used an un-pretrained VGG16 autoencoder with data augmentation for feature learning and extraction from tiles. These features were two-tiered clustered using the K-means algorithm. Each tile was assigned the cluster with the highest probability. The tiles were reassembled into whole slide images. For each slide, the proportion of tiles in each cluster was calculated. We will associate clusters with clinical features and 5- and 10-year breast cancer-specific survival using multivariable logistic and Cox proportional hazards regression models, respectively.

Results: Mean age at baseline enrollment (1992-1993) and breast cancer diagnosis for the cases was 60.6 years (SD=6.0) and 71.5 years (SD=7.0), respectively. The majority of cancer diagnoses occurred after 1999 (79%) and 81% of women included were diagnosed invasive breast cancer. The final pipeline for the full set of images is currently being built. Preliminary runs at the 1x magnification level with 100 cases (N=21,472 tiles) have shown clustering based on macro-level features such as adipose, stromal and epithelial content. Second-tier clustering (clustering within clusters) shows further delineation of groups within clusters of interest (i.e. epithelial-cell rich regions). The final output with all 1,716 slides will be based on analysis at the 5x magnification level.

Discussion: We expect that some histopathological features identified by ML models will be associated with conventional pathology features, clinical features, and breast cancer-specific survival. Utilization of ML methods for analyzing histology slides provides additional data that can be integrated into epidemiological studies. Future directions include analyzing images at higher magnifications (10x or 20x) and assessing the association between ML histopathological characteristics and breast cancer risk factors and incorporating these characteristics into prognostic models.

Citation Format: Samantha Puvanesarajah, James M. Hodge, Jacob L. Evans, William Seo, Michelle Yi, Michelle M. Fritz, Mary Macheski-Preston, Ted Gansler, Susan M. Gapstur, Mia M. Gaudet. Unsupervised deep-learning to identify histopathological features among breast cancers in the Cancer Prevention Study-II Nutrition Cohort [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2417.

Abstract 1635: A scalable, cloud-based, unsupervised deep learning system for identification, extraction, and summarization of potentially imperceptible patterns in whole-slide images of breast cancer tissue

Digital pathology images potentially contain novel patterns that may be perceived by modern deep learning models, but not humans. Prior unsupervised pattern recognition approaches have been used to reveal prognostically-relevant subtypes of glioblastoma (PMID: 28984190) and breast density segmentation (PMID: 26915120), and may complement supervised machine learning models trained using labeled data.

In the Cancer Prevention Study II (CPS-II) cohort (PMID: 12015775), high-resolution, digitized hemotoxylin and eosin diagnostic slides are available for approximately 1,700 breast cancer cases providing an opportunity to perform unsupervised pattern recognition image analysis for epidemiologic breast cancer studies. Given the size of the dataset and complexity of the models, we constructed an end-to-end analytical pipeline, including preprocessing, feature engineering, and clustering, using cloud-based technologies that enable analysis at scale.

Prior to training the unsupervised models, we faced issues converting raw images with open-source software. Specifically, OpenSlides could not open the Leica Versa SCN files due to their proprietary format while BioFormats inverted colors. To fix these issues, we altered the BioFormats library to successfully convert the files into a TIFF format. Since this issue likely affects other researchers, we are in discussions to provide the fix under a public license.

TIFF formatted images were then denoised through color normalization to reduce hue variance and artifact detection to remove unwanted features such as pathologist annotations. Due to the computational complexity of analyzing the full image, images were padded with white space to ensure divisibility and broken into nine tiles of a predefined size. To further reduce computation time, uninformative tiles were filtered based on a predetermined threshold of artifact and white space composition.

The remaining tiles were input to the unsupervised models. We used convolutional autoencoders, specifically a modified VGG-16 model without pretrained weights and a deep embedded clustering algorithm. These models learn representations of the images called ‘feature vectors’ and encode the images’ salient patterns. The final model was chosen based on iterative testing on a subsample of 100 images (N=21,472 tiles) and performance comparison of various VGG-inspired autoencoders. The feature vectors were clustered by K-means to summarize the information in a format suitable for statistical analyses.

Our initial results show that the system captures macro-scale tissue patterns at lower magnifications (1x and 5x) and produces clusters that can be integrated into epidemiological studies of breast cancer etiology and prognosis.

Citation Format: Jacob L. Evans, William Seo, Mary Macheski-Preston, Michelle Fritz, Samantha Puvanesarajah, James Hodge, Ted Gansler, Susan Gapstur, Mia M. Gaudet, Michelle Yi. A scalable, cloud-based, unsupervised deep learning system for identification, extraction, and summarization of potentially imperceptible patterns in whole-slide images of breast cancer tissue [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1635.

Abstract 4211: Breast cancer risk factors by mode of detection among screened women in the Cancer Prevention Study-II

Identifying risk factors for women at high risk of symptom-detected breast cancers that were missed by screening would enable targeting of alternative prevention strategies. To identify breast cancer risk factors by mode of detection, we examined these associations in heavily-screened women from the Cancer Prevention Study (CPS)-II Nutrition Cohort. Among 76,406 women followed for a median of 13.8 years, 2,469 screen-detected and 1,189 symptom-detected breast cancer cases were diagnosed. Multivariable-adjusted associations were estimated using joint Cox proportional hazards regression models with person-time calculated contingent on screening. The mean (standard deviation) age at baseline was 62.0 (6.6) years for non-cases, 61.6 (6.1) for screen-detected cases, and 61.5 (6.2) for symptom-detected cases. Women with symptom-detected tumors were diagnosed at a slightly younger age (mean=69.4) and less likely to be diagnosed with a localized tumor (65.3%), than women with a screen-detected tumor (mean age at diagnosis=70.4 and localized tumors=83.3%). Factors associated with higher risks of symptom detected vs. screen detected breast cancer, included: current combined menopausal hormone use (HR=2.04, 95% CI 1.70 - 2.46 vs. HR=1.43, 95% CI 1.25 - 1.63), current estrogen only menopausal hormone use (HR=1.41, 95% CI 1.15 - 1.72 vs. HR=0.94, 95% CI 0.82 - 1.09), and history of benign breast disease (HR=1.89, 95% CI 1.67 - 2.13 vs. HR=1.49, 95% CI 1.36 - 1.62). The opposite pattern of higher risk of screen detected vs. symptom detected breast cancer was observed for greater adult weight gain (HR=2.32, 95% CI 1.91 - 2.82 vs. HR=1.42, 95% CI 1.10 - 1.84) and alcohol intake (HR=1.39, 95% CI 1.24 - 1.64 vs. HR=1.20, 95% CI 0.92 - 1.58). Our results suggest that understanding risk factors for symptom-detected cancers may help identify women who would benefit from more intensive screening to facilitate early detection.

Citation Format: Mia M. Gaudet, Emily Deubler, W. Ryan Diver, Samantha Puvanesarajah, Alpa V. Patel, Ted Gansler, Dana Flanders, Mark Sherman, Susan Gapstur. Breast cancer risk factors by mode of detection among screened women in the Cancer Prevention Study-II [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 4211.

Mode of detection and breast cancer mortality by follow-up time and tumor characteristics among screened women in Cancer Prevention Study-II

PURPOSE

In a screened population, breast cancer-specific mortality is lower for screen-detected versus symptom-detected breast cancers; however, it is unclear whether this association varies by follow-up time and/or tumor characteristics. To further understand the prognostic utility of mode of detection, we examined its association with breast cancer-specific mortality, overall and by follow-up time, estrogen receptor status, tumor size, and grade.

METHODS

In the Cancer Prevention Study-II Nutrition Cohort, 3975 routinely screened women were diagnosed with invasive breast cancer (1992-2015). Among 2686 screen-detected and 1289 symptom-detected breast cancers, 206 and 209 breast cancer deaths, respectively, occurred up to 24 years post diagnosis. Hazard ratios (HR) and 95% confidence intervals (CI) were calculated from Cox proportional hazard regression models.

RESULTS

Controlling for prognostic factors, symptom detection was associated with higher risk of breast cancer-specific death up to 5 years after diagnosis (HR_≤5years = 1.88, 95% CI 1.21-2.91) this association was attenuated in subsequent follow-up (HR_>5years = 1.26, 95% CI 0.98-1.63). Within tumor characteristic strata, there was a 1.3-2.7-fold higher risk of breast cancer death associated with symptom-detected cancers ≤ 5 years of follow-up, although associations were only significant for women with tumors < 2 cm (HR_≤5years = 2.42, 95% CI 1.19-4.93) and for women with grade 1 or 2 tumors (HR_≤5years = 2.72, 95% CI 1.33-5.57). In subsequent follow-up, associations were closer to the null.

CONCLUSIONS

Screen detection is a powerful prognostic factor for short-term survival. Among women who survived at least 5 years after breast cancer diagnosis, other clinical factors may be more predictive of breast cancer survival.

Abstract P1-08-07: Risk factors for high-grade and low-grade DCIS in the cancer prevention study-II nutrition cohort

Background: Nuclear grade, believed to be established early in carcinogenesis, is an indicator of ductal carcinoma in situ (DCIS) prognosis. Women with high-grade DCIS have a higher risk of local recurrence compared to women with low-grade DCIS. Risk factors for DCIS overall are well-characterized but risk factors by grade are not. Given the prognostic capabilities of grade for DCIS, it is of interest to identify whether risk factors by DCIS grade differ.

Methods: Among 75,630 women enrolled in the Cancer Prevention Study-II Nutrition Cohort in 1992-1993, we identified 422 who were diagnosed with low-moderate grade DCIS (i.e. grades 1 or 2) and 355 who were diagnosed with high-grade DCIS (i.e. grade 3) during follow-up through 2013. Beginning in 1997, biennial questionnaires were administered to update exposure status, including screening mammography in the previous two years. For this analysis, these questionnaires were used to partition follow-up time into approximately two-year intervals. Because screening is strongly linked to diagnosis of DCIS, contribution of person-time within a specific interval was conditional on reporting a screening mammogram in the interval prior. Multivariate joint Cox proportional hazards regression models were used to estimate adjusted hazard ratios (HR) and 95% confidence intervals (CI) for the associations of known breast cancer risk factors with DCIS overall, and with high-grade and low-grade DCIS individually.

Results: Parity (HR=0.70; 95% CI: 0.63-0.78, parous vs. nulliparous), smoking status (HR=0.69; 95% CI: 0.58-0.82, current vs. never smoking) and menopausal status (HR=0.79; 95% CI: 0.73-0.87, natural menopause &lt;50 years vs. natural menopause ≥ 50 years) were inversely associated with risk of DCIS overall. Whereas, positive family history of breast cancer (HR=1.46; 95% CI: 1.36-1.57), personal history of benign breast disease (BBD) (HR=1.73; 95% CI: 1.62-1.85), and current use of combination estrogen and progestin hormone replacement therapy (HR=1.15; 95% CI: 1.04-1.28) were associated with higher risk of DCIS. In analyses stratified on DCIS grade, history of BBD was more strongly associated with higher risk of low-grade (HR=2.21; 95% CI: 1.81-2.70) than with high-grade DCIS (HR=1.29; 95% CI:1.04-1.60) (p for heterogeneity=0.001). Current combination estrogen and progestin hormone replacement therapy use was associated with a higher risk of high-grade DCIS (HR=1.40; 95% CI:1.03-1.90) but not low-grade DCIS (HR=1.02; 95% CI: 0.75-1.40), but the difference by grade was not statistically significant (p=0.5).

Conclusions: In this study, which is the first to comprehensively assess risk factors by DCIS grade, the association between personal history of BBD and risk of DCIS appeared to differ by grade. Due to limited power for some risk factor analyses, future studies using larger prospective cohorts or pooled data should be conducted to better identify these associations.

Citation Format: Puvanesarajah S, Gapstur SM, Gansler T, Patel AV, Gaudet MM. Risk factors for high-grade and low-grade DCIS in the cancer prevention study-II nutrition cohort [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P1-08-07.

Abstract P1-08-02: Associations of epidemiologic and clinical features with intensity of immune infiltrates in postmenopausal breast cancer cases from the cancer prevention study-II cohort

Stromal tumor-infiltrating lymphocytes (sTILs) are often prominent in HER2-enriched and triple negative breast cancers and predict a favorable prognosis for these subtypes. Clinical and epidemiological determinants of sTILs have not been defined in large cohorts. Accordingly, we analyzed these associations for 702 eligible, postmenopausal invasive breast cancer cases from the Cancer Prevention Study (CPS)-II Nutrition cohort. CPS-II included 97,786 women who completed baseline and follow-up surveys since 1992. Women with self-reported breast cancer diagnoses were consented for medical record and tumor tissue retrieval. We analyzed pre-diagnostic data from the survey prior to diagnosis on personal history of benign breast disease (BBD), menopausal hormone use, alcohol intake, cigarette smoking status, waist circumference, body mass index, adult weight gain, nonsteroidal anti-inflammatory drugs, and physical activity. One pathologist (TG) evaluated sTILs using whole slide images of H&E stained sections according to recommendations by the International TILs Working Group 2014. sTIL levels were dichotomized: none/minimal (0-10%) and moderate/high (&gt;10%). We compared the sTIL levels by clinical and epidemiologic risk factors using chi-square statistics. Odds ratio (ORs) and 95% confidence intervals (CIs) for the associations of clinical and epidemiologic risk factors with sTILs were estimated with multivariable logistic regression.

Mean age at diagnosis was 71.9 years; 44% of cancers were moderate grade, 74% were localized staged and 88% were luminal-like. Cancers with high/moderate sTIL levels (n=614), compared to none/minimal (n=88), were more likely to be high grade (54% vs. 20%; p-value &lt;0.001), node-positive (41% vs. 22%; p-value &lt;0.001), and non-luminal (36% vs. 8%; p-value &lt;0.001). In a model mutually-adjusted for age at diagnosis and clinical factors, all clinical factors were associated with moderate/high sTIL levels: node-positive (OR=2.15, 95% CI 1.28 – 3.58), high grade (OR=3.98, 95% CI 1.88 – 9.06), and non-luminal subtype (OR=3.65, 95% CI 2.00 – 6.59). Age at diagnosis was not associated with sTIL levels (per year: OR=1.01, 95% CI 0.97 – 1.05).

Women with BBD were less likely to have moderate/high sTIL levels than minimal/none sTIL levels (38% vs. 54%; age-adjusted OR=0.50, 95% CI 0.31 – 0.79). Ever smokers also were less likely to have moderate/high sTIL levels (38% vs. 49%) with an OR=0.61 (95% 0.38 – 0.97). Other risk factors were not significantly associated with sTIL levels.

In this large epidemiologic cohort, non-luminal cancers had more sTILs than luminal cancers. Women with a personal history of BBD and ever cigarette smokers had significant lower levels of sTILs. Our study is among the first to examine pre-diagnostic risk factors in relation to sTILs and provides the impetus for larger, population-based studies with phenotypic characterization of immune cells.

Citation Format: Gaudet MM, Mellow MM, Puvanesarajah S, Gapstur SM, Sherman ME, Gansler T. Associations of epidemiologic and clinical features with intensity of immune infiltrates in postmenopausal breast cancer cases from the cancer prevention study-II cohort [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P1-08-02.

Abstract 3260: Utilizing digital pathology to understand breast epithelial characteristics of benign breast disease among women undergoing diagnostic image-guided breast biopsy

Introduction: Defining the histologic correlates of mammographic breast density (MD) may provide insights into why elevated MD is related to increased breast cancer risk. Studies suggest that reduced terminal ductal lobular (TDLU) involution is associated with elevated MD, and both are independent breast cancer risk factors among women who have undergone a biopsy for benign breast disease (BBD). Prior digital histologic analyses of normal breast tissues revealed epithelial nuclear density (END) and TDLU involution are correlated. Accordingly, we examined associations of END with TDLU involution and MD in clinical biopsies. Methods: We analyzed 262 image-guided breast biopsies diagnosed as BBD from 224 women. TDLU involution was visually assessed as TDLU count/mm2 and TDLU span (inversely related to level of involution) in background normal tissue, evaluated using digitized images. The Genie Classifier (Aperio) was applied to images to estimate nuclei count per unit epithelial area, (END). Single X-ray Absorptiometry of pre-biopsy craniocaudal digital mammograms was applied to measure global MD (percent fibroglandular volume (%FGV)). Analysis of covariance, adjusted for age and body mass index, examined mean END differences across tertiles of TDLU/MD measures. Analyses were conducted at the biopsy level using SAS PROC GENMOD to account for within-woman correlations. All tests were two-tailed. Results: Overall, 67% of BBD biopsies were proliferative. Higher END was observed among proliferative than non-proliferative BBD (median END: 10,187 vs. 9,953 respectively; p=0.04). Among all women, END significantly increased with increasing tertiles of TDLU measures (p-trends: TDLU count/100mm2=0.0001, TDLU span=0.046). Whereas TDLU metrics were positively associated with %FGV, no relationship was observed between END and %FGV. In analyses stratified by BBD severity, however, END and %FGV were positively associated among women with non-proliferative disease (p-trend=0.04), findings not observed with proliferative disease. Conclusions: Automated END and visually assessed TDLU involution metrics were positively associated with each other and with MD. However, associations were diluted for proliferative lesions, suggesting that applying automated digital pathology tools to unsegmented digital images of whole sections of BBD biopsies does not demonstrate the same associations with MD as visual assessment of TDLU involution.

Citation Format: Maeve Mullooly, Samantha Puvanesarajah, Shaoqi Fan, Ruth M. Pfeiffer, Linnea Olsson, Manila Hada, Erin L. Kirk, Pamela M. Vacek, Donald L. Weaver, John A. Shepherd, Amir P. Mahmoudzadeh, Jeff Wang, Stephen M. Hewitt, Sally D. Herschorn, Mark E. Sherman, Melissa A. Troester, Gretchen L. Gierach. Utilizing digital pathology to understand breast epithelial characteristics of benign breast disease among women undergoing diagnostic image-guided breast biopsy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 3260.

Stroma modifies relationships between risk factor exposure and age-related epithelial involution in benign breast

Delayed age-related lobular involution has been previously associated with elevated breast cancer risk. However, intraindividual variability in epithelial involution status within a woman is undefined. We developed a novel measure of age-related epithelial involution, density of epithelial nuclei in epithelial areas using digital image analysis in combination with stromal characteristics (percentage of section area comprising stroma). Approximately 1800 hematoxylin and eosin stained sections of benign breast tissue were evaluated from 416 participants having breast surgery for cancer or benign conditions. Two to sixteen slides per woman from different regions of the breast were studied. Epithelial involution status varied within a woman and as a function of stromal area. Percentage stromal area varied between samples from the same woman (median difference between highest and lowest stromal area within a woman was 7.5%, but ranged from 0.01 to 86.7%). Restricting to women with at least 10% stromal area (N = 317), epithelial nuclear density decreased with age (-637.1 cells/mm² per decade of life after age 40, p < 0.0001), increased with mammographic density (457.8 cells/mm² per increasing BI-RADs density category p = 0.002), and increased non-significantly with recent parity, later age at first pregnancy, and longer and more recent oral contraceptive use. These associations were attenuated in women with mostly fat samples (<10% stroma (N = 99)). Thirty-one percent of women evaluated had both adequate stroma (≥10%) and mostly fat (<10% stroma) regions of breast tissue, with the probability of having both types increasing with the number breast tissue samplings. Several breast cancer risk factors are associated with elevated age-related epithelial content, but associations depend upon stromal context. Stromal characteristics appear to modify relationships between risk factor exposures and breast epithelial involution.

PAM50 and Risk of Recurrence Scores for Interval Breast Cancers

Breast cancers detected after a negative breast screening examination and prior to the next screening are referred to as interval cancers. These cancers generally have poor clinical characteristics compared with screen-detected cancers, but associations between interval cancer and genomic cancer characteristics are not well understood. Mammographically screened women diagnosed with primary invasive breast cancer from 1993 to 2013 (n = 370) were identified by linking the Carolina Breast Cancer Study and the Carolina Mammography Registry. Among women with a registry-identified screening mammogram 0 to 24 months before diagnosis, cancers were classified as screen-detected (N = 165) or interval-detected (N = 205). Using logistic regression, we examined the association of mode of detection with cancer characteristics (clinical, IHC, and genomic), overall, and in analyses stratified on mammographic density and race. Interval cancer was associated with large tumors [>2 cm; OR, 2.3; 95% confidence interval (CI), 1.5-3.7], positive nodal status (OR, 1.8; 95% CI, 1.1-2.8), and triple-negative subtype (OR, 2.5; 95% CI, 1.1-5.5). Interval cancers were more likely to have non-Luminal A subtype (OR, 2.9; 95% CI, 1.5-5.7), whereas screen-detected cancers tended to be more indolent (96% had low risk of recurrence genomic scores; 71% were PAM50 Luminal A). When stratifying by mammographic density and race, associations between interval detection and poor prognostic features were similar by race and density status. Strong associations between interval cancers and poor-prognosis genomic features (non-Luminal A subtype and high risk of recurrence score) suggest that aggressive tumor biology is an important contributor to interval cancer rates. Cancer Prev Res; 11(6); 327-36. ©2018 AACR.

Mammary Gland Evaluation in Juvenile Toxicity Studies: Temporal Developmental Patterns in the Male and Female Harlan Sprague-Dawley Rat

There are currently no reports describing mammary gland development in the Harlan Sprague-Dawley (HSD) rat, the current strain of choice for National Toxicology Program (NTP) testing. Our goals were to empower the NTP, contract labs, and other researchers in understanding and interpreting chemical effects in this rat strain. To delineate similarities/differences between the female and male mammary gland, data were compiled starting on embryonic day 15.5 through postnatal day 70. Mammary gland whole mounts, histology sections, and immunohistochemically stained tissues for estrogen, progesterone, and androgen receptors were evaluated in both sexes; qualitative and quantitative differences are highlighted using a comprehensive visual timeline. Research on endocrine disrupting chemicals in animal models has highlighted chemically induced mammary gland anomalies that may potentially impact human health. In order to investigate these effects within the HSD strain, 2,3,7,8-tetrachlorodibenzo-p-dioxin, diethylstilbestrol, or vehicle control was gavage dosed on gestation day 15 and 18 to demonstrate delayed, accelerated, and control mammary gland growth in offspring, respectively. We provide illustrations of normal and chemically altered mammary gland development in HSD male and female rats to help inform researchers unfamiliar with the tissue and may facilitate enhanced evaluation of both male and female mammary glands in juvenile toxicity studies.

Self-regulatory failure and the perpetration of adolescent dating violence: Examining an alcohol use by gene explanation

Studies report that alcohol use is related to partner violence, but for many, alcohol use does not culminate in violence against partners. Guided by a self-regulatory failure framework, we predicted that alcohol use would be more strongly associated with dating violence perpetration among adolescents with genotypes linked to impulsivity and emotional reactivity. The hypothesis was tested using random coefficient modeling of data from a multi-wave longitudinal study spanning grades 8-12 (ages 13-18) (n = 1,475). Analyses adjusted for multiple testing and race, and the potential for gene by environment correlation was examined. As predicted, alcohol use was more strongly associated with dating violence among adolescents who had a high rather than a low multilocus genetic profile composed of five genetic markers that influence dopamine signaling. Alcohol use was more strongly related to dating violence among boys with long rather than short 5-HTTLPR alleles, the opposite of the prediction. MAOA-uVNTR did not interact with alcohol, but it had a main effect on dating violence by boys in later grades in the expected direction: boys with more low activity alleles perpetrated more dating violence. Exploratory analyses found variation in findings by race. Our findings demonstrate the importance of incorporating genes into etiological studies of adolescent dating violence, which to date has not been done. Aggr. Behav. Aggr. Behav. 42:189-203, 2015. © 2014 Wiley Periodicals, Inc.