Mammography is the primary breast cancer screening strategy. Recent methods have been developed using the mammogram image to improve breast cancer risk prediction. However, it is unclear on the extent to which the effect of risk factors on breast cancer risk is mediated through tissue features summarized in mammogram images and the extent to which it is through other pathways. While mediation analysis has been conducted using mammographic density (a summary measure within the image), the mammogram image is not necessarily well described by a single summary measure and, in addition, such a measure provides no spatial information about the relationship between the exposure risk factor and the risk of breast cancer. Thus, to better understand the role of the mammogram images that provide spatial information about the state of the breast tissue that is causally predictive of the future occurrence of breast cancer, we propose a novel method of causal mediation analysis using mammogram image mediator while accommodating the irregular shape of the breast. We apply the proposed method to data from the Joanne Knight Breast Health Cohort and leverage new insights on the decomposition of the total association between risk factor and breast cancer risk that was mediated by the texture of the underlying breast tissue summarized in the mammogram image.

Background: In the general population, sugar intake is associated with a wide range of adverse health conditions related to premature aging, including obesity, diabetes, and cardiovascular disease. Childhood cancer survivors are at increased risk of premature aging and mortality compared to their healthy peers and may be especially vulnerable to adverse consequences of excess sugar intake. Objective: To examine the association between sugar and sugar-sweetened beverage intake and premature aging in childhood cancer survivors. Method: A total of 3,322 adult survivors of childhood cancer (age range 18-65 years; mean age: 31 years) in SJLIFE self-reported their typical diet using the 110-item Block Food Frequency Questionnaire. Added sugars included all sugars added to foods during preparation or processing. Total sugar-sweetened beverages are the sum of regular and diet soda and fruit-flavored drinks. Survivors’ sociodemographics, cancer histories, and health conditions were abstracted from medical records. Premature aging was assessed using the Deficit Accumulation Index (DAI) that was a ratio of the number of age-related chronic health conditions each survivor had out of 45 conditions total. The DAI was categorized into low (<0.2), medium (0.2-0.34), and high (>0.35) aging risk groups. Multinomial logistic regressions (reference: low aging risk group) adjusting for confounders, including sociodemographics, lifestyle factors, cancer treatments, and overall diet quality, were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs). Results: Survivors’ average total sugar intake was 120 g/day and added sugar intake was 71 g/day. 41% of survivors consumed sugar-sweetened beverages ≥1 time/day, and 26% consumed soda daily; 75% of soda consumed was regular (vs. diet). Survivors with higher consumption of sugar and sugar-sweetened beverages were more likely to be non-Hispanic Black and have lower educational attainment and income. Total sugar intake was associated with a significantly increased risk of premature aging (per 25 g/1,000 kcal increment, OR=1.31 [95% CI: 1.00-1.70] in the medium-risk group; OR=1.52 [95% CI: 1.03-2.25] in the high-risk group). Added sugar intake was associated with a 19% (OR=1.19, 95% CI: 1.07-1.31, per 20 g/1,000 kcal increment) and an 18% (OR=1.18, 95% CI: 1.02-1.37) increased risk of premature aging in the medium- and high-risk group, respectively. Consuming ≥2 servings/day (vs. ≤ 1/week) of total sugar-sweetened beverage was also related to an increased risk of premature aging (OR=1.54 [95% CI: 0.83-2.83] in the medium-risk group; OR=6.71 [95% CI: 2.95-15.2] in the high-risk group). Regular soda, but not diet soda, consumption was associated with premature aging risk. Conclusion: Higher consumption of sugar and sugar-sweetened beverages was associated with an increased risk of premature aging in childhood cancer survivors. Intervention efforts to reduce sugar intake among this vulnerable population are needed. Citation Format: Tuo Lan, Mei Wang, AnnaLynn M. Williams, Matthew J. Ehrhardt, Emily R. Finch, Jennifer Q. Lanctot, Shu Jiang, Kevin R. Krull, Gregory T. Armstrong, Melissa M. Hudson, Graham A. Colditz, Leslie Robison, Kirsten K. Ness, Yikyung Park. Sugar Intake and premature aging in adult survivors of childhood cancer in the St. Jude Lifetime (SJLIFE) Cohort [abstract]. In: Proceedings of the AACR Special Conference: Aging and Cancer; 2022 Nov 17-20; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2022;83(2 Suppl_1):Abstract nr PR014.

This abstract is being presented as a short talk in the scientific program. A full abstract is available in the Short Talks from Proffered Abstracts section (PR014) of the Conference Proceedings. Citation Format: Tuo Lan, Mei Wang, AnnaLynn M. Williams, Matthew J. Ehrhardt, Emily R. Finch, Jennifer Q. Lanctot, Shu Jiang, Kevin R. Krull, Gregory T. Armstrong, Melissa M. Hudson, Graham A. Colditz, Leslie Robison, Kirsten K. Ness, Yikyung Park. Sugar Intake and premature aging in adult survivors of childhood cancer in the St. Jude Lifetime (SJLIFE) Cohort [abstract]. In: Proceedings of the AACR Special Conference: Aging and Cancer; 2022 Nov 17-20; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2022;83(2 Suppl_1):Abstract nr B004.

To efficiently capture data from mammographic breast images and classify long term risk of breast cancer, we developed methods that use the extensive existing data that are currently ignored in the context of breast cancer risk stratification. More than 20 studies support texture features add value to risk prediction beyond breast density. However, the entire mammogram imaging data has a high dimension of pixels (~13 million per image), greatly exceeding the number of women in a cohort. We apply functional principal component analysis methods to predict 5-years breast cancer incidence using baseline mammograms. We applied these methods onto women participating in the Joanne Knight Breast Health Cohort which is comprised of over 10,000 women undergoing repeated mammography screening at Siteman Cancer Center and followed since 2010. All women had baseline mammogram at entry, provided a blood sample and completed a risk factor questionnaire. Mammograms are all using the same technology (Hologic). During follow-up through October 2020, we identified 246 incident breast cancer cases (pathology confirmed) and matched them to controls from the perspective cohort based on month of mammogram and age at entry. In a baseline model we controlled for age, menopause, BMI, and mammographic breast density (BIRADs). We then added the full image (characterized by the FPC) to the base model and further compared the AUC of the new model vs the base model using the likelihood ratio test. AUC is validated with internal 10-fold cross validation. The AUC for 5-year breast cancer risk classification increased significantly from a median of 0.61 (sd 0.09 for estimated AUCs across 10-fold internal validation) for the baseline model to 0.70 (0.10) when the full image is added, p < 0.001. We conclude that using full mammogram images for breast cancer risk prediction captures additional information on breast tissue characteristics that relate to cancer risk, and improves prediction classification. This prediction algorithm can run efficiently in real time (in seconds) with processing of digital mammograms. Thus, this model can be easily implemented in mammography screening services and other clinical settings to guide real-time risk stratification to improve precision prevention of the leading cancer in women world-wide. Further analysis will quantify the value of adding other breast cancer risk factors, including polygenic risk scores. Addition of repeated mammogram images over time should further increase classification performance. This approach has the potential to improve risk classification by using data already available for the vast majority of women already having repeated screening mammograms. Citation Format: Shu Jiang, Graham A. Colditz. Whole mammogram image improves breast cancer prediction [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr LB161.

To efficiently capture data from mammographic breast images and classify long-term risk of breast cancer, we developed FLIP, a novel Cox regression-based framework that fully utilizes data in the mammograms beyond current density measures. FLIP use the extensive existing data that are currently ignored in the context of breast cancer risk stratification. More than 20 studies support texture features add value to risk prediction beyond breast density. However, the entire mammogram imaging data has a high dimension of pixels (~13 million per image), greatly exceeding the number of women in a cohort. FLIP was fitted and cross-validated within the Joanne Knight Breast Health Cohort excluding cases diagnosed in the first 6 months of entry. The Joanne Knight Breast Health Cohort is comprised of over 10,000 women undergoing repeated mammography screening at Siteman Cancer Center and followed since 2010. All women had baseline mammogram at entry, provided a blood sample and completed a risk factor questionnaire. Mammograms are all using the same technology (Hologic). During follow-up through October 2020, we identified 246 incident breast cancer cases (pathology confirmed) and matched them to controls from the perspective cohort based on month of mammogram and age at entry. We obtained an AUC of 0.68 (SE 0.03) including the whole mammogram image, age and BI-RADS (4th edition) density category; and AUC of 0.72 (SE 0.04) by adding in BMI and menopausal status to this model. These 5-year prediction performances exceed that of well-developed models based on epidemiologic risk factors (P < 0.001). FLIP offers standard statistical solutions and removes barriers to wider clinical use without prohibitive training data and extensive computational requirements, providing a transparent workflow ensuring high reproducibility. It should be accessible anywhere mammograms are used. We conclude that using full mammogram images for breast cancer risk prediction captures additional information on breast tissue characteristics that relate to cancer risk, and improves prediction classification. This prediction algorithm can run efficiently in real time (in seconds) with processing of digital mammograms. Thus, this model can be easily implemented in mammography screening services and other clinical settings to guide real-time risk stratification to improve precision prevention of the leading cancer in women world-wide. Further analysis will quantify the value of adding other breast cancer risk factors, including polygenic risk scores. Addition of repeated mammogram images over time should further increase classification performance. This approach has the potential to improve risk classification by using data already available for the vast majority of women already having repeated screening mammograms. Schema overview of FLIP The raw images are in the form of .dcm files before entering into FLIP. After automated processing and image alignment, the two CC-views (left and right) are average between the two breasts for characterization. The inputted 2D mammograms are first characterized with bivariate splines over triangulation to preserve spatial distribution of pixels and accommodate the irregular semi-circular breast boundary. The characterization is further optimized (see Supplemental Material) which provides a unique and closed-form solution. b. A simple Cox proportional hazards model is adopted using well-established risk factors (RF), including age, breast density (BI-RADS), BMI, menopausal status, parity, family history, and history of benign breast disease. The mammogram image acts as an additional risk factor in the Cox regression accompanied with a 2D coefficient surface. All inferential procedures with Cox regression are applicable to FLIP which provides a transparent workflow ensuring high reproducibility. h_i (t) denotes the hazard function at time t for individual i, and h_0 (t) denotes the nonparametric baseline hazard function. c. Women who are diagnosed with breast cancer within the first 6 month of their mammogram date have been removed from this analysis and we focus on the 5-year risk. Discriminatory performance is assessed with AUC and validated via a 10-fold cross-validation. Citation Format: Shu Jiang, Graham A. Colditz. PD14-04 Whole mammogram image-based Cox regression improves 5-year breast cancer prediction [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr PD14-04.