Concordant and Discordant Breast Density Patterns by Different approaches for Assessing Breast Density and Breast Cancer Risk

Concordant and Discordant Breast Density Patterns by Different approaches for Assessing Breast Density and Breast Cancer Risk | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Concordant and Discordant Breast Density Patterns by Different approaches for Assessing Breast Density and Breast Cancer Risk Yoosun Cho, Eun Kyung Park, Yoosoo Chang, Mi-ri Kwon, Eun Young Kim, and 12 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4471074/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 01 Nov, 2024 Read the published version in Breast Cancer Research and Treatment → Version 1 posted 9 You are reading this latest preprint version Abstract Purpose To examine the discrepancy in breast density assessments by radiologists, LIBRA software, and AI algorithm and their association with breast cancer risk. Methods Among 74,610 Korean women aged ≥ 34 years, who underwent screening mammography, density estimates obtained from both LIBRA and the AI algorithm were compared to radiologists using BI-RADS density categories (A–D, designating C and D as dense breasts). The breast cancer risks were compared according to concordant or discordant dense breasts identified by radiologists, LIBRA, and AI. Cox-proportional hazards models were used to determine adjusted hazard ratios (aHRs) [95% confidence intervals (CIs)]. Results During a median follow-up of 9.9 years, 479 breast cancer cases developed. Compared to the reference non-dense breast group, the aHRs (95% CIs) for breast cancer were 2.37 (1.68–3.36) for radiologist-classified dense breasts, 1.30 (1.05–1.62) for LIBRA, and 2.55 (1.84–3.56) for AI. For different combinations of breast density assessment, aHRs (95% CI) for breast cancer were 2.40 (1.69–3.41) for radiologist-dense/LIBRA-non-dense, 11.99 (1.64–87.62) for radiologist-non-dense/LIBRA-dense, and 2.99 (1.99–4.50) for both dense breasts, compared to concordant non-dense breasts. Similar trends were observed with radiologists/AI classification: the aHRs (95% CI) were 1.79 (1.02–3.12) for radiologist-dense/AI-non-dense, 2.43 (1.24–4.78) for radiologist-non-dense/AI-dense, and 3.23 (2.15–4.86) for both dense breasts. Conclusion The risk of breast cancer was highest in concordant dense breasts. Discordant dense breast cases also had a significantly higher risk of breast cancer, especially when identified as dense by either AI or LIBRA, but not radiologists, compared to concordant non-dense breast cases. Mammography Screening Breast density Artificial intelligence Laboratory Individualized Breast Radiodensity Assessment Cohort study Figures Figure 1 Introduction Dense breast tissue, as detected through mammography, is considered one of strongest risk factor for breast cancer [ 1 – 3 ]. This finding is consistently observed across diverse populations, including Western and Asian pre- and postmenopausal women, highlighting the crucial role of breast density in predicting future breast cancer risk [ 3 – 7 ]. Additionally, more than 50% of Asian women have dense breast tissue, which is not only independent risk factor that used in various risk prediction models but also for the identification of eligible women for supplemental screening methods [ 8 , 9 ]. Thus, improving breast density assessment could potentially enhance the effectiveness and management of breast cancer screening. It has been reported that breast cancer risk increases with higher mammographic density, whether measured quantitatively or qualitatively [ 1 ]. Qualitative assessments by radiologists, using the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS), are the most common standard for evaluating mammographic breast density [ 10 ]. However, the reliability of BI-RADS categories among radiologists can vary from moderate to substantial, potentially leading to misclassification and, consequently, the under- or overestimation of breast cancer risk [ 11 , 12 ]. To address these issues, automated density assessment methods such as AI or quantitative tools like the open-source LIBRA (Laboratory for Individualized Breast Radiodensity Assessment) software, which calculates the fibroglandular tissue volume relative to total breast volume, provide more objective and reproducible measurements [ 13 ]. AI-driven qualitative assessments, trained on extensive data sets with ground truths provided by radiologists, have shown fair agreement with radiologist assessments [ 14 , 15 ]. Despite this, the comparative risks of breast cancer development based on breast density assessments between radiologists and AI-driven assessments, particularly in cohort studies involving Asian women with dense breast tissue, remain underexplored. Moreover, the use of AI and quantitative methods can lead to either concordant or discordant density assessments, and the implications of these outcomes require further investigation. Therefore, our study aimed to achieve two objectives: 1) assess the relationship between mammographic dense breasts, as defined by radiologists, LIBRA, and AI, and incident breast cancer, and 2) examine how concordance or discordance in breast density classifications between radiologists and either LIBRA or AI affects the risk of breast cancer. Material and Methods Study population The Kangbuk Samsung Health Study is a cohort study of Korean men and women aged ≥ 18 years who underwent comprehensive annual or biennial health examinations at the Kangbuk Samsung Hospital Total Healthcare Centers located in Seoul and Suwon, South Korea, as previously described [ 7 , 16 ]. This study received approval from the Institutional Review Board (2023-02-004). This retrospective study primarily focused on Korean women aged ≥ 34 years who underwent their initial digital screening mammography at our institution as part of a health examination between January 2009 and 2014, and followed up until December 2020 (Fig. 1 ). Participants provided informed consent for linkage to the national cancer registry data were included in the study. Notably, although national breast cancer screening recommendations in Korea start at age 40, there is a common practice of private screening organizations offering screenings from age 35 [ 17 , 18 ]. Moreover, acknowledging the differentiation between Korean age and Western age, resulting in a potential variance of 1–2 years in documented data, we have incorporated instances where the verified age distribution is 34 years, backed by authentic data. Participants were excluded if they had insufficient follow-up duration of < 12 months, had a history of breast surgery, history of mammoplasty, had a history of breast cancer or a prior registered breast cancer before mammography. After excluding missing or inappropriate values related to breast density or BI-RADS classification, the final analysis included 74,610 women (Fig. 1 ). Data collection Demographic information, first-degree family history of breast cancer, behavioral factors, reproductive factors, and medical history were collected through standardized, self-administered questionnaires. Body height and weight were measured by trained nurses, with participants wearing a hospital gown and no shoes. Body mass index (BMI) was then classified according to Asian-specific criteria [ 19 ]: underweight, < 18.5 kg/m 2 ; normal weight, 18.5 to 23 kg/m 2 ; overweight, 23 to 25 kg/m 2 ; and obese, ≥ 25 kg/m 2 . The Gail scores, predicting breast cancer over 5 year, were calculated [ 20 ], and were categorized into three risk groups; low (less than 0.8%), intermediate (0.8–1.67%), and high (greater than or equal to 1.67%). Mammographic Breast Density Measures Mammographic imaging data, including BI-RADS final assessment categories and mammographic density, were retrieved from the original radiological reports. Participants in the study underwent a standard four-view digital mammography examination, which involved bilateral craniocaudal (CC) and mediolateral oblique (MLO) views, using a full-field digital mammography (FFDM) system (Sonograph 2000D/DMR/DS, GE Healthcare, Chicago, IL, USA, or Selenia, Hologic, Marlborough, MA, USA) at the Suwon and Seoul Total Healthcare Centers. The assessment of mammography was conducted by experienced breast imaging radiologists at the two centers, employing the BI-RADS classification system. Breast density was assessed by radiologists and categorized based on the BI-RADS assessment, including category A (almost entirely fatty), category B (scattered fibroglandular densities), category C (heterogeneously dense), or category D (extremely dense). While the 5th edition of BI-RADS was released in 2013, most of our baseline data on breast density, collected between 2009 and 2014, was obtained before its implementation in our centers. Therefore, the breast density categories may reflect assessments made with the earlier 4th edition of BI-RADS, as radiologists reviewed the FFDM data from this study using that edition [ 21 ]. LIBRA, completely automated, freely available open-source software, can be utilized on both raw and processed FFDM images to produce area-based measurements of mammographic breast density [ 22 ]. LIBRA provides an estimate of the total absolute dense area (DA), whereas normalizing DA by the total breast area results in breast percent density (PD). For this study, we used the DA and PD estimates obtained with LIBRA from images processed and stored in DICOM format for each woman, averaging density measures from all four mammographic views, bilateral CC and MLO views [ 13 ]. Density assessments by LIBRA were strongly correlated with widely-used Volpara® methods in our study (r = 0.89, P < 0.001). Quantitative breast density assessed by LIBRA were categorized based on the BI-RADS assessment (type A, almost entirely fatty, 75% fibroglandular tissue) [ 21 ]. For the study, the AI algorithm (Lunit INSIGHT MMG, version 1.1.7.2, Seoul, Korea) was retrospectively applied to stored mammographic images. The AI density task module provided the average density results for four (bilateral CC, MLO) views of each mammography. The average mammographic density was presented on a 1–10 scale, with density categories defined as follows: Density A (Scores 1 to 2), Density B (Scores 3 to 5), Density C (Scores 6 to 8), and Density D (Scores 9 to 10), following the recommendations of provider [ 15 ]. Definition of breast cancer The breast cancer diagnosis after a screening mammography was established by linking the study data to the Korean Central Cancer Registry [ 23 ]. In this context, breast cancer was defined as either invasive cancer (International Classification of Diseases-10 code C50) or ductal carcinoma in situ (International Classification of Diseases-10 code D05.1). Statistical analyses The primary endpoint was the development of breast cancer. The occurrence of breast cancer was measured in terms of the number of cases per 1000 person-years, and the follow-up period extended from the baseline visit until the date of the primary endpoint or the last health screening examination (December 31, 2020), whichever came first. The participants' breast density was classified into the following categories: A-B (non-dense), and C-D (dense). We evaluated incident breast cancer cases based on concordant or discordant breast density patterns between radiologists and LIBRA or AI. Cases where radiologists classified breasts as non-dense, in agreement with LIBRA or AI assessments, were designated as "concordant non-dense breast." Likewise, cases where radiologists and LIBRA or AI both categorized breasts as dense were designated as "concordant dense breast." Discordant breast density patterns were defined as cases where radiologists reported breasts as non-dense while LIBRA or AI indicated dense breast tissue, or vice versa. To assess the relationship of 1) breast density assessment and incident breast cancer, and 2) concordant and discordant breast density patterns and incident breast cancer, according to different measurement, radiologists, LIBRA, and AI-driven, cox proportional hazard models were used to calculate adjusted hazard ratios (aHRs) with 95% confidence intervals (CIs) for the primary endpoint. The multivariable-adjusted model was gradually adjusted for covariates, including age, BMI, family history of breast cancer, reproductive histories (age at menarche, parity, menopausal status, and female hormone use), education level (below college graduate, college graduate or higher, or unknown), smoking status (never, former, current smoker, or unknown), alcohol consumption (< 10 or ≥ 10 g/day), and physical activity level (inactive, minimally active, HEPA, or unknown). Harrell's C-index (the area under the receiver operating characteristic curve [AUROC]), a measure of the concordance probability adapted for survival analysis, was used to assess whether the addition of radiologists, LIBRA, and AI individually or concurrently to the base model, including conventional risk factors, improved prediction of breast cancer. C statistics are routinely applied for global assessments of discrimination in a survival model [ 24 ]. All analyses were carried out using Stata software version 18.0 (StataCorp LLC, College Station, TX, USA). Statistical significance was defined as a two-tailed P-value less than 0.05. Results Baseline characteristics Table 1 presents the baseline characteristics of the participants, categorized into non-dense and dense groups based on breast density measurements from radiologists, LIBRA, and AI algorithm. At baseline, dense breasts were classified by radiologists (87.7%), AI-program (84.5%), and LIBRA (27.6%). Among the 74,610 Asian women selected (mean age: 42.1 ± 8.2 years), individuals with dense breasts determined by those measures were consistently younger, had an earlier menarche, were more likely to be nulliparous, had lower BMI, and attained higher education levels. Table 1 Characteristics of the women in the study according to breast density assessment measures (n = 74,610) Total Radiologist AI-driven LIBRA Non-dense Dense Non-dense Dense Non-dense Dense N 74,610 (100.0) 9,185 (12.3) 65,425 (87.7) 11,576 (15.5) 63,034 (84.5) 53,987 (72.4) 20,623 (27.6) Age (SD), year a 42.10 (8.17) 51.08 (11.09) 40.84(6.77) 49.80 (11.40) 40.69(6.49) 43.47(8.75) 38.52 (4.79) Age at menarche (SD), year a 14.08 (1.63) 14.79 (1.95) 13.99(1.54) 14.650 (1.93) 13.98(1.55) 14.14(1.68) 13.94(1.49) BMI category b ≤ 22.9 kg/m2 50,758 (68.0) 3,351 (36.5) 47,407 (72.5) 4,187 (36.2) 46,571 (73.9) 31,399 (58.2) 19,359 (93.9) 23–24.9 kg/m2 12,450 (16.7) 2,399 (26.1) 10,051 (15.4) 3,012 (26.0) 9,438 (15.0) 11,493 (21.3) 957 (4.6) ≥ 25 kg/m2 11,274 (15.1) 3,414 (37.2) 7,860 (12.0) 4,358 (37.6) 6,916 (11.0) 10,999 (20.4) 275 (1.3) missing 128 (0.2) 21 (0.2) 107 (0.2) 19 (0.2) 109 (0.2) 96 (0.2) 32 (0.2) Family history of breast cancer No 72,196 (96.8) 8,841 (96.3) 63,355 (96.8) 11,194 (96.7) 61,002 (96.8) 52,234 (96.8) 19,962 (96.8) Yes 2,190 (2.9) 313 (3.4) 1,877 (2.9) 346 (3.0) 1,844 (2.9) 1,585 (2.9) 605 (2.9) missing 224 (0.3) 31 (0.3) 193 (0.3) 36 (0.3) 188 (0.3) 168 (0.3) 56 (0.3) Parity nulliparous 5,617 (7.5) 275 (3.0) 5,342 (8.2) 288 (2.5) 5,329 (8.5) 3,235 (6.0) 2,382 (11.6) parous 63,419 (85.0) 7,889 (85.9) 55,530 (84.9) 10,082 (87.1) 53,337 (84.6) 46,518 (86.2) 16,901 (82.0) missing 5,574 (7.5) 1,021 (11.1) 4,553 (7.0) 1,206 (10.4) 4,368 (6.9) 4,234 (7.8) 1,340 (6.5) Menopausal status premenopausal 52,184 (69.9) 3,026 (32.9) 49,158 (75.1) 4,889 (42.2) 47,295 (75.0) 35,351 (65.5) 16,833 (81.6) postmenopausal 11,440 (15.3) 4,940 (53.8) 6,500 (9.9) 5,702 (49.3) 5,738 (9.1) 10,891 (20.2) 549 (2.7) missing 10,986 (14.7) 1,219 (13.3) 9,767 (14.9) 985 (8.5) 10,001 (15.9) 7,745 (14.3) 3,241 (15.7) Female hormone use No 73,047 (97.9) 71,797 (97.9) 1,250 (98.0) 11,228 (97.0) 61,819 (98.1) 52,725 (97.7) 20,322 (98.5) Yes 1,371 (1.8) 1,350 (1.8) 21 (1.6) 319 (2.8) 1,052 (1.7) 1,116 (2.1) 255 (1.2) missing 192 (0.3) 188 (0.3) 4 (0.3) 29 (0.3) 163 (0.3) 146 (0.3) 46 (0.2) Education < College graduate 20,730 (27.8) 4,469 (48.7) 16,261 (24.9) 5,562 (48.0) 15,168 (24.1) 17,122 (31.7) 3,608 (17.5) ≥ College graduate 48,771 (65.4) 3,945 (43.0) 44,826 (68.5) 5,076 (43.8) 43,695 (69.3) 33,113 (61.3) 15,658 (75.9) missing 5,109 (6.8) 771 (8.4) 4,338 (6.6) 938 (8.1) 4,171 (6.6) 3,752 (6.9) 1,357 (6.6) Physical activity Low 31,886 (42.7) 31,405 (42.8) 481 (37.7) 4,440 (38.4) 27,446 (43.5) 22,341 (41.4) 9,545 (46.3) Moderate 18,236 (24.4) 17,935 (24.5) 301 (23.6) 3,043 (26.3) 15,193 (24.1) 13,287 (24.6) 4,949 (24.0) High 9,834 (13.2) 9,659 (13.2) 175 (13.7) 1,924 (16.6) 7,910 (12.5) 7,486 (13.9) 2,348 (11.4) missing 14,654 (19.6) 14,336 (19.5) 318 (24.9) 2,169 (18.7) 12,485 (19.8) 10,873 (20.1) 3,781 (18.3) Alcohol consumption (in grams/day) < 10 52,435 (70.3) 5,726 (62.3) 46,709 (71.4) 7,421 (64.1) 45,014 (71.4) 37,083 (68.7) 15,352 (74.4) ≥ 10 8,530 (11.4) 881 (9.6) 7,649 (11.7) 1,184 (10.2) 7,346 (11.7) 6,148 (11.4) 2,382 (11.6) missing 13,645 (18.3) 2,578 (28.1) 11,067 (16.9) 2,971 (25.7) 10,674 (16.9) 10,756 (19.9) 2,889 (14.0) Smoking status never smoker 53,923 (72.3) 6,555 (71.4) 47,368 (72.4) 8,068 (69.7) 45,855 (72.7) 38,678 (71.6) 15,245 (73.9) former smoker 5,766 (7.7) 513 (5.6) 5,253 (8.0) 690 (6.0) 5,076 (8.1) 3,980 (7.4) 1,786 (8.7) current smoker 1,576 (2.1) 186 (2.0) 1,390 (2.1) 222 (1.9) 1,354 (2.1) 1,090 (2.0) 486 (2.4) missing 13,345 (17.9) 1,931 (21.0) 11,414 (17.4) 2,596 (22.4) 10,749 (17.1) 10,239 (19.0) 3,106 (15.1) Five-year risk based on Gail model < 0.83% 74,184 (99.4) 9,065 (98.7) 65,119 (99.5) 11,443 (98.9) 62,741 (99.5) 53,630 (99.3) 20,554 (99.7) 0.83–1.66% 191 (0.3) 86 (0.9) 105 (0.2) 97 (0.8) 94 (0.1) 184 (0.3) 7 (0.0) ≥ 1.67% 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) missing 235 (0.3) 34 (0.4) 201 (0.3) 36 (0.3) 199 (0.3) 173 (0.3) 62 (0.3) Breast density dichotomized into dense and non-dense categories based on the BI-RADS; Dense breast: ≥50%, Heterogeneously dense (C) and Extremely dense (D). Non-dense breast: 0 ~ 50%, Almost entirely fat (A) and Scattered fibroglandular densities (B) Data are presented as a means (standard deviation); b BMI: normal to underweight (≤ 22.9 kg/m2), overweight (23–24.9 kg/m2), and obese (≥ 25 kg/m2) Abbreviation: BMI, body mass index; LIBRA, Laboratory for Individualized Breast Radiodensity Assessment; AI, Artificial Intelligence Breast cancer risk associated with dense breasts determined by LIBRA, radiologist, and AI-driven methods During a median follow-up of 9.9 years, 479 breast cancer cases were newly identified. Table 2 presents the risk of incident breast cancer by dense breasts, as measured by three methods, including radiologist, LIBRA and AI-driven breast density. Comparing dense breast groups estimated by LIBRA, radiologist, and AI to their respective non-dense counterparts (used as the reference), the age-adjusted HRs (95%CIs) for overall breast cancer development were 1.32 (1.06–1.63) for LIBRA, 2.71 (1.93–3.80) for radiologists, and 2.99 (2.18–4.11) for AI methods. These association remained significant with corresponding aHR (95% CI) of 1.30 (1.05–1.62), 2.37 (1.68–3.36), and 2.55 (1.84–3.56), after adjusting confounders, including age, BMI, age at menarche, family history of breast cancer, parity, menopausal status, female hormone use, education, physical activity, smoking, and alcohol intake. The findings were consistent across separate analyses of invasive cancer and DCIS ( Table S1 ), although dense breasts identified by LIBRA were no longer associated with an increased risk of DCIS. Table 2 Risk of incident breast cancer by ordinal LIBRA, radiologist, and AI-driven breast density (N = 74,610) Breast cancer (total) Person-years (PY) Incidence Cases Incidence rate (/10^3 PY) Age-adjusted HR (95% CI) Multivariable-adjusted HR a (95% CI) LIBRA breast density Non-dense (N = 53,987) 192.54 144 1.74 1.00 (reference) 1.00 (reference) Dense (N = 20,623) 73.17 335 1.97 1.32 (1.06–1.63) 1.30 (1.05–1.62) Radiologist breast density Non-dense (N = 9,185) 41.35 44 1.06 1.00 (reference) 1.00 (reference) Dense (N = 65,425) 224.36 435 1.94 2.71 (1.93–3.80 2.37 (1.68–3.36) AI-driven breast density Non-dense (N = 11,576) 51.07 53 1.04 1.00 (reference) 1.00 (reference) Dense (N = 63,034) 214.64 426 1.98 2.99 (2.18–4.11) 2.55 (1.84–3.56) Breast density dichotomized into dense and non-dense categories based on the BI-RADS; Dense breast: ≥50%, Heterogeneously dense (C) and Extremely dense (D). Non-dense breast: 0 ~ 50%, Almost entirely fat (A) and Scattered fibroglandular densities (B) a. Hazard ratios adjusted for age, BMI, age at menarche, family history of breast cancer, parity, menopausal status, female hormone use, education, physical activity, smoking, and alcohol intake Breast cancer risk according to concordant and discordant dense breast assessment by Radiologists and LIBRA or AI-driven methods When comparing discordant and concordant dense breast patterns to the concordant non-dense breast pattern, identified by both radiologists and LIBRA, the aHRs (95% CI) for incident breast cancer were as follows; 2.40 (1.69–3.41) for radiologist-identified dense but LIBRA-identified non-dense breasts, 11.99 (1.64–87.62) for radiologist-identified non-dense but LIBRA-identified dense breasts (notably, only one case of breast cancer occurred in this category), and 2.99 (1.99–4.50) for both radiologist- and LIBRA-identified dense breasts (Table 3 ). Additionally, when considering breast density assessments provided by radiologists and AI-driven methods, the aHRs for incident breast cancer were as follows; 1.79 (1.02–3.12) for radiologist-identified dense breasts but AI-identified non-dense breasts, 2.43 (1.24–4.78) for radiologist-identified non-dense but AI-identified dense breasts, and 3.23 (2.15–4.86) for both radiologist- and AI-identified dense breasts (Table 3 ). When analyses were conducted separately for breast cancer subtypes, including invasive cancer and DCIS, the findings remained similar for invasive cancer ( Tables S2 and S3 ). However, the number of incident DCIS cases were too small to estimate the risk in discordant breast density patterns assessed by radiologist and LIBRA, or radiologist and Al algorithm. Table 3 Risk of breast cancer according to the discrepancy between Radiologist's, LIBRA and AI-method breast density (n = 74,610) No Person-years (PY) Incidence Cases Incidence rate (/10 3 PY) Multivariable-adjusted HR a (95% CI) Radiologist LIBRA Nondense Nondense 9,138 41,240 43 1.04 1.00 (reference) Dense Nondense 44,849 151,300 292 1.93 2.4 (1.69–3.41) Nondense Dense 47 112 1 8.89 11.99 (1.64–87.62) Dense Dense 20,576 73,060 143 1.96 2.99 (1.99–4.50) Radiologist AI-methods Nondense Nondense 7,095 35,080 32 0.91 1.00 (reference) Dense Nondense 4,481 15,990 21 1.31 1.79 (1.02–3.12) Nondense Dense 2,090 6,268 12 1.92 2.43 (1.24–4.78) Dense Dense 60,944 208,400 414 1.99 3.23 (2.15–4.86) Breast density dichotomized into dense and non-dense categories based on the BI-RADS; Dense breast: ≥50%, Heterogeneously dense (C) and Extremely dense (D). Non-dense breast: 0 ~ 50%, Almost entirely fat (A) and Scattered fibroglandular densities (B) a. Hazard ratios adjusted for age, BMI, age at menarche, family history of breast cancer, parity, female hormone use, menopausal status, education, physical activity, smoking, and alcohol intake Predictive ability for incident breast cancer determined after adding radiologists, LIBRA, and AI –identified dense breasts to the Gail model The incremental predictive ability for incident breast cancer was determined after adding breast density assessments from radiologists, LIBRA, or AI individually, in pairs or altogether to a base model composed of conventional breast cancer risk factors, represented as the Gail score ( Table S4 ). Adding dense breast assessed by either radiologists or AI alone to the base model (Gail model) significantly improved the AUROCs for predicting incident breast cancer. The AUROC for additions by radiologists was 0.533 (95% CI: 0.519–0.548), and for AI, it was 0.540 (95% CI: 0.524–0.556). Although the inclusion of dense breast assessments by LIBRA also improved cancer prediction, it did not achieve statistical significance, with an AUROC of 0.518 (95% CI: 0.496–0.539) compared to the Gail model. The addition of dense breast assessment by both radiologist and AI-method to the base model further improved the predictive ability of developing breast cancer, resulting in an AUROC, 95% CI 0.542, 0.526–0.559) ( Table S4 ). Discussion Our study compared qualitative, quantitative, and AI-driven assessments of breast density and found that each method, when evaluated individually by radiologists, LIBRA, or AI, demonstrated a significant association with increased risk of breast cancer. Furthermore, discordant breast density patterns between radiologists and LIBRA or AI were also associated with an increased risk of incident breast cancer. Notably, breasts identified as dense by LIBRA or AI methods—even when deemed non-dense by radiologists—showed a consistently higher risk of breast cancer, with a relatively greater hazard ratio than cases identified solely by radiologists. The dense breast assessed by radiologist and/or AI improved risk prediction for incident breast cancer (based on Harrell’s C−index) compared to conventional breast cancer risk factors, estimated as Gail model. Our results suggest that dense breasts identified by AI or LIBRA may provide complementary information into the risk of developing breast cancer, compared to those identified solely by radiologists. Our study is the first cohort study to assess the risk of developing subsequent breast cancer based on individual measures of breast density (radiologist, LIBRA, and AI) and the implications of concordant and discordant assessments of dense breast by radiologist and LIBRA or AI. A cross-sectional study for 488 Korean women evaluated and compared the inter-rater agreements between radiologists, AI and another commercial automated density assessment program (Volpara®) [ 15 ], and found that density assessments by AI showed similar agreement with those of radiologists compared to the Volpara® (κ = 0.52 and 0.50, respectively). The study, however, did not extend to investigate the longitudinal association between either dense breast assessment by individual breast density measures or their discrepant interpretations in dense breast and the risk of incident breast cancer [ 15 ]. In our study, breast density assessed by different approaches was all associated with increased risk of incident breast cancer. This study showed slightly weaker breast cancer risk for LIBRA-defined dense breast than radiologists or AI. However, few studies investigated the LIBRA assessments for images before a diagnosis of breast cancer. A case-control study by Gastounioti et al. demonstrated that LIBRA PD showed comparable breast cancer association (OR, 1.3; processed images and OR, 1.2; raw images) with semiautomated area-based assessment tool Cumulus (OR, 1.5; processed images) and Volpara (OR, 1.4; raw images) [ 13 ]. A recent cohort study, comprising 21,000 non-Hispanic white women aged 40–74, evaluated the long-term breast cancer risk (up to 10 years) using LIBRA [ 25 ]. The study found that LIBRA measures of density were associated with breast cancer risk, similar to the results obtained with Cumulus measures, a semi-automated validated tool [ 25 ]. After adjusting for BMI and reproductive variables, the aHR for breast cancer associated with each standard deviation increase in percent density, measured by LIBRA, was 1.44 (95%CI: 1.26–1.66) [ 25 ]. Our study showed a comparable age-adjusted HR of 1.32 (95% CI: 1.06–1.63) for LIBRA, and AI achieved the strongest capacity in breast cancer risk prediction with age-adjusted HR of 2.99 (95% CI: 2.18–4.11). Further study in diverse population is necessary to examine which approach has the greatest robustness in accurate risk assessment of future breast cancer. There have been several studies on the use of various AI models for mammographic density assessments. A previous study compared two automated methods and visual assessment in contralateral breasts of women with breast cancer showed similar associations with breast cancer [ 26 ]. However, a study of five methods of measuring breast density on prior mammograms demonstrated that visual assessment predicted subsequent breast cancer risk significantly than all other density methods [ 27 ]. In our study, AI-driven breast density had the strongest association with subsequent breast cancer. According to the AI models, the algorithms are diverse, in terms of algorithm operation and development dataset. In this study, we used commercially available AI model based on BI-RADS density classification by experienced radiologists. Although it remains undiscovered which factor has a greater impact on breast cancer risk, our study suggests the feasibility of AI-driven mammographic density as a reliable marker. Our findings revealed that reader-dependent, subjective measures of density and categorical density data (fatty, scattered, heterogeneously dense, and extremely dense in accordance with BI-RADS) could be concordant or discordant to LIBRA or AI-defined density, potentially contributing to breast cancer risk. The assessment of breast density by LIBRA or AI could complement the evaluation made by radiologists, especially in cases where there's a discrepancy, such as AI identifying density while the radiologist does not. In this study, discordance resulted from the radiologists’ tendency to assign density grades higher than those obtained by LIBRA, and radiologist-defined non-dense but LIBRA-defined dense breasts yielded a sample size that was too small to estimate the risk of breast cancer. Our results favored the AI-derived dense breast model demonstrating the improvement of predictive performance when adding to the Gail model (0.540 vs. 0.503, p < 0.01), although the C-statistic for the Gail model, the most frequently used breast cancer risk model, was low as much as 0.50, corresponding to previous studies found that the risk models have demonstrated only limited prediction performance [ 28 ]. The modified Gail model has allowed for risk estimates for Asian-American women [ 29 ], albeit its C-statistics is low and has been estimated to be equal to 0.54 in an external validation study for Korean 40 229 women [ 30 ]. According to an external validation study in a U.S. screening cohort, the predictive performance of a mammography-derived AI risk model was significantly higher than that of the Gail model (0.68 vs. 0.55, p < 0.01) [ 31 ]. The limited discriminatory accuracy might be attributed to recent increase in the incidence of breast cancer in Korea; previous validation study population dataset was outdated, without updating recent increase in the incidence of breast cancer in Korea. Moreover, our study is not a prospective cohort of the general population but we used retrospective data from breast cancer screening patients, predominantly comprising young and relatively healthy women with mostly low Gail risk scores. Future research should consider using an Asian-specific breast cancer risk prediction model to provide a more profound understanding of the relationship between breast density and breast cancer, supported by further external validation studies. Our study had several limitations that should be considered. First, we evaluated a single automated density assessment program and AI program. The performance of different programs varies and our results cannot be generalized to other programs immediately. Secondly, we did not consider changes in breast density over time, the potential impact on the development of breast cancer, or the discriminatory accuracy in predicting breast cancer. Further studies using longitudinal digital mammograms accounting dynamic changes in density over time may be useful to refine more accurate breast cancer risk. Third, the study population comprised well-educated women with potentially high accessibility to medical services, limiting generalizability. Lastly, our analysis was based on retrospective data collected during routine health screening examinations and previous radiologic reports of screening mammography. Therefore, we did not assess the utility of the AI algorithm or LIBRA for radiologists in a real screening setting, nor did we evaluate its potential impact on screening performance when used by radiologists. To comprehensively understand the effectiveness of the AI algorithm or LIBRA in a real-world screening environment, further prospective studies are necessary. Currently, there is a lack of study regarding the assessment of breast cancer risks based on concordant and discordant breast density patterns, utilizing various approaches for assessing breast density. Accurate identification of women with dense breast tissue could facilitate the development of strategies for additional screening and primary breast cancer prevention. Future prospective studies with a more diverse population are required whether incorporating LIBRA or AI measurements to radiologists' interpretations for dense breasts could enhance breast cancer risk assessments in real clinical settings. Abbreviations BI-RADS, Breast Imaging Reporting & Data System CI, confidence interval HR, hazard ratio LIBRA, Laboratory for Individualized Breast Radiodensity Assessment AI, Artificial Intelligence FFDM, full-field digital mammography PD, percent density VPD, volumetric percent density DA, dense area SD, standard deviation Declarations Funding This work was supported by the SKKU Excellence in Research Award Research Fund, Sungkyunkwan University (2021) and Lunit Inc. Competing Interests The authors declare that they have no conflict of interest. Author Contributions YCho, EK, YChang, and SR planned, designed and implemented the study, including quality assurance and control. YChang, MK and SR analyzed the data and designed the study’s analytic strategy. YChang, and SR supervised field activities. Material preparation, and data collection were performed by all authors. The first draft of the manuscript was written by YCho, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Data Availability Statement The datasets generated during and/or analyzed during the current study are not publicly available due outside of Institutional Review Board restrictions (the data were not collected in a way that could be distributed widely) but are available from the corresponding author on reasonable request. Ethics approval This study was approved by the Institutional Review Board of Kangbuk Samsung Hospital (2023-02-004), and the need for informed consent was waived due to the use of de-identified retrospective data collected during the health screening process. Consent to participate Not applicable Consent for publication Not applicable Acknowledgments This work was supported by the Medical Research Funds from Kangbuk Samsung Hospital and the National Research Foundation of Korea (NRF) grant funded by the Korea government (Ministry of Science and ICT, MSIT) (RS-2023-00253017). References McCormack VA, dos Santos Silva I (2006) Breast density and parenchymal patterns as markers of breast cancer risk: a meta-analysis. Cancer Epidemiol Biomarkers Prev 15(6):1159–1169 Boyd NF, Guo H, Martin LJ, Sun L, Stone J, Fishell E, Jong RA, Hislop G, Chiarelli A, Minkin S, Yaffe MJ (2007) Mammographic density and the risk and detection of breast cancer. N Engl J Med 356(3):227–236 Pettersson A, Graff RE, Ursin G, Santos Silva ID, McCormack V, Baglietto L, Vachon C, Bakker MF, Giles GG, Chia KS et al (2014) Mammographic density phenotypes and risk of breast cancer: a meta-analysis. J Natl Cancer Inst 106(5) Boyd NF, Guo H, Martin LJ, Sun L, Stone J, Fishell E, Jong RA, Hislop G, Chiarelli A, Minkin S, Yaffe MJ (2007) Mammographic Density and the Risk and Detection of Breast Cancer. N Engl J Med 356(3):227–236 Mariapun S, Li J, Yip CH, Taib NA, Teo SH (2015) Ethnic differences in mammographic densities: an Asian cross-sectional study. PLoS ONE 10(2):e0117568 Kim S, Tran TXM, Song H, Ryu S, Chang Y, Park B (2022) Mammographic Breast Density, Benign Breast Disease, and Subsequent Breast Cancer Risk in 3.9 Million Korean Women. Radiology 304(3):534–541 Kim EY, Chang Y, Ahn J, Yun JS, Park YL, Park CH, Shin H, Ryu S (2020) Mammographic breast density, its changes, and breast cancer risk in premenopausal and postmenopausal women. Cancer 126(21):4687–4696 Lau S, Abdul Aziz YF, Ng KH (2017) Mammographic compression in Asian women. PLoS ONE 12(4):e0175781 Chalfant JS, Hoyt AC (2022) Breast Density: Current Knowledge, Assessment Methods, and Clinical Implications. J Breast Imaging 4(4):357–370 Radiology ACo (2003) Breast imaging reporting and data system. BI-RADS Berg WA, Campassi C, Langenberg P, Sexton MJ (2000) Breast Imaging Reporting and Data System: inter- and intraobserver variability in feature analysis and final assessment. AJR Am J Roentgenol 174(6):1769–1777 Boyd NF, Wolfson C, Moskowitz M, Carlile T, Petitclerc C, Ferri HA, Fishell E, Gregoire A, Kiernan M, Longley JD et al (1986) Observer variation in the classification of mammographic parenchymal patterns. J Chronic Dis 39(6):465–472 Gastounioti A, Kasi CD, Scott CG, Brandt KR, Jensen MR, Hruska CB, Wu FF, Norman AD, Conant EF, Winham SJ et al (2020) Evaluation of LIBRA Software for Fully Automated Mammographic Density Assessment in Breast Cancer Risk Prediction. Radiology 296(1):24–31 Schaffter T, Buist DSM, Lee CI, Nikulin Y, Ribli D, Guan Y, Lotter W, Jie Z, Du H, Wang S et al (2020) Evaluation of Combined Artificial Intelligence and Radiologist Assessment to Interpret Screening Mammograms. JAMA Netw Open 3(3):e200265 Lee SE, Son NH, Kim MH, Kim EK (2022) Mammographic Density Assessment by Artificial Intelligence-Based Computer-Assisted Diagnosis: A Comparison with Automated Volumetric Assessment. J Digit Imaging 35(2):173–179 Chang Y, Ryu S, Choi Y, Zhang Y, Cho J, Kwon MJ, Hyun YY, Lee KB, Kim H, Jung HS et al (2016) Metabolically Healthy Obesity and Development of Chronic Kidney Disease: A Cohort Study. Ann Intern Med 164(5):305–312 Lee SY, Jeong SH, Kim YN, Kim J, Kang DR, Kim HC, Nam CM (2009) Cost-effective mammography screening in Korea: high incidence of breast cancer in young women. Cancer Sci 100(6):1105–1111 Lee EH, Park B, Kim NS, Seo HJ, Ko KL, Min JW, Shin MH, Lee K, Lee S, Choi N et al (2015) The Korean guideline for breast cancer screening. J Korean Med Assoc 58(5):408–419 World Health Organization, Regional Office for the Western Pacific (2000) The Asia-Pacific perspective: redefining obesity and its treatment. Health Communications Australia, Sydney Gail MH, Brinton LA, Byar DP, Corle DK, Green SB, Schairer C, Mulvihill JJ (1989) Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst 81(24):1879–1886 D’Orsi CJME, Ikeda DM et al (2003) Breast Imaging Reporting and Data System: ACR BI-RADS—Breast Imaging Atlas. American College of Radiology Keller BM, Chen J, Daye D, Conant EF, Kontos D (2015) Preliminary evaluation of the publicly available Laboratory for Breast Radiodensity Assessment (LIBRA) software tool: comparison of fully automated area and volumetric density measures in a case-control study with digital mammography. Breast Cancer Res 17:117 Kwon MR, Chang Y, Park B, Ryu S, Kook SH (2023) Performance analysis of screening mammography in Asian women under 40 years. Breast Cancer 30(2):241–248 Uno H, Cai T, Pencina MJ, D'Agostino RB, Wei LJ (2011) On the C-statistics for evaluating overall adequacy of risk prediction procedures with censored survival data. Stat Med 30(10):1105–1117 Habel LA, Alexeeff SE, Achacoso N, Arasu VA, Gastounioti A, Gerstley L, Klein RJ, Liang RY, Lipson JA, Mankowski W et al (2023) Examination of fully automated mammographic density measures using LIBRA and breast cancer risk in a cohort of 21,000 non-Hispanic white women. Breast Cancer Res 25(1):92 Brandt KR, Scott CG, Ma L, Mahmoudzadeh AP, Jensen MR, Whaley DH, Wu FF, Malkov S, Hruska CB, Norman AD et al (2016) Comparison of Clinical and Automated Breast Density Measurements: Implications for Risk Prediction and Supplemental Screening. Radiology 279(3):710–719 Astley SM, Harkness EF, Sergeant JC, Warwick J, Stavrinos P, Warren R, Wilson M, Beetles U, Gadde S, Lim Y et al (2018) A comparison of five methods of measuring mammographic density: a case-control study. Breast Cancer Res 20(1):10 Vilmun BM, Vejborg I, Lynge E, Lillholm M, Nielsen M, Nielsen MB, Carlsen JF (2020) Impact of adding breast density to breast cancer risk models: A systematic review. Eur J Radiol 127:109019 Costantino JP, Gail MH, Pee D, Anderson S, Redmond CK, Benichou J, Wieand HS (1999) Validation studies for models projecting the risk of invasive and total breast cancer incidence. J Natl Cancer Inst 91(18):1541–1548 Min JW, Chang MC, Lee HK, Hur MH, Noh DY, Yoon JH, Jung Y, Yang JH (2014) Korean Breast Cancer S: Validation of risk assessment models for predicting the incidence of breast cancer in korean women. J Breast Cancer 17(3):226–235 Gastounioti A, Eriksson M, Cohen EA, Mankowski W, Pantalone L, Ehsan S, McCarthy AM, Kontos D, Hall P, Conant EF (2022) External Validation of a Mammography-Derived AI-Based Risk Model in a U.S. Breast Cancer Screening Cohort of White and Black Women. Cancers (Basel) 14(19) Supplementary file 1 Supplementary Table 1 Risk of incident breast cancer by ordinal LIBRA, radiologist, and AI-driven breast density (N = 74,610) Supplementary Table 2 Risk of invasive breast cancer according to the discrepancy between Radiologist's and LIBRA and AI-method breast density (n = 74,610) Supplementary Table 3 Risk of DCIS according to the discrepancy between Radiologist's, LIBRA and AI-method breast density (n = 74,610) Supplementary Table 4 Comparison of the discriminatory power of LIBRA, Radiologist, and AI-driven breast density in prediction of breast cancer Additional Declarations No competing interests reported. Supplementary Files BRCTLIBRAAISupplementaryfilev1.1.docx Supplementary Table 1. Risk of incident breast cancer by ordinal LIBRA, radiologist, and AI-driven breast density (N=74,610) Supplementary Table 2. Risk of invasive breast cancer according to the discrepancy between Radiologist's and LIBRA and AI-method breast density (n= 74,610) Supplementary Table 3. Risk of DCIS according to the discrepancy between Radiologist's, LIBRA and AI-method breast density (n= 74,610) Supplementary Table 4. Comparison of the discriminatory power of LIBRA, Radiologist, and AI-driven breast density in prediction of breast cancer Cite Share Download PDF Status: Published Journal Publication published 01 Nov, 2024 Read the published version in Breast Cancer Research and Treatment → Version 1 posted Editorial decision: Revision requested 20 Sep, 2024 Reviews received at journal 14 Sep, 2024 Reviewers agreed at journal 07 Sep, 2024 Reviewers agreed at journal 14 Jul, 2024 Reviewers agreed at journal 11 Jun, 2024 Reviewers invited by journal 05 Jun, 2024 Submission checks completed at journal 25 May, 2024 Editor assigned by journal 25 May, 2024 First submitted to journal 24 May, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4471074","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":310942123,"identity":"6d5f7542-0948-42d2-865c-35f12632a8ba","order_by":0,"name":"Yoosun Cho","email":"","orcid":"","institution":"Total Healthcare Center, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yoosun","middleName":"","lastName":"Cho","suffix":""},{"id":310942124,"identity":"27a3f3bc-20bd-4cb1-8a5e-c8e3b7613999","order_by":1,"name":"Eun Kyung Park","email":"","orcid":"","institution":"Lunit","correspondingAuthor":false,"prefix":"","firstName":"Eun","middleName":"Kyung","lastName":"Park","suffix":""},{"id":310942126,"identity":"4eb0698e-e87f-4082-ad96-1d986a9b64d4","order_by":2,"name":"Yoosoo Chang","email":"","orcid":"","institution":"Total Healthcare Center, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yoosoo","middleName":"","lastName":"Chang","suffix":""},{"id":310942127,"identity":"8040fd63-75db-4d9f-bcb0-3d089d91c0e5","order_by":3,"name":"Mi-ri Kwon","email":"","orcid":"","institution":"Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Mi-ri","middleName":"","lastName":"Kwon","suffix":""},{"id":310942130,"identity":"17bd75d9-4dc6-4570-a1ab-95953bdba1cf","order_by":4,"name":"Eun Young Kim","email":"","orcid":"","institution":"Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Eun","middleName":"Young","lastName":"Kim","suffix":""},{"id":310942132,"identity":"4f2bc5e5-abfb-49d4-b2f6-fb50b8798e98","order_by":5,"name":"Minjeong Kim","email":"","orcid":"","institution":"Lunit","correspondingAuthor":false,"prefix":"","firstName":"Minjeong","middleName":"","lastName":"Kim","suffix":""},{"id":310942134,"identity":"7da7ab0e-ba58-4346-986d-306f6f0e2b65","order_by":6,"name":"Boyoung Park","email":"","orcid":"","institution":"Hanyang University College of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Boyoung","middleName":"","lastName":"Park","suffix":""},{"id":310942135,"identity":"7bd7d661-1826-4e97-90b1-dea25e150fe1","order_by":7,"name":"Sanghyup Lee","email":"","orcid":"","institution":"Lunit","correspondingAuthor":false,"prefix":"","firstName":"Sanghyup","middleName":"","lastName":"Lee","suffix":""},{"id":310942140,"identity":"390861ab-176e-485e-9f44-0b8a5a710305","order_by":8,"name":"Han Eol Jeong","email":"","orcid":"","institution":"Lunit","correspondingAuthor":false,"prefix":"","firstName":"Han","middleName":"Eol","lastName":"Jeong","suffix":""},{"id":310942145,"identity":"150ca632-1e7c-48ba-8c0b-3d83c3fdfb63","order_by":9,"name":"Ki Hwan Kim","email":"","orcid":"","institution":"Lunit","correspondingAuthor":false,"prefix":"","firstName":"Ki","middleName":"Hwan","lastName":"Kim","suffix":""},{"id":310942147,"identity":"76fa5d7c-7ec8-46e3-ad42-3604ae835cbc","order_by":10,"name":"Tae Soo Kim","email":"","orcid":"","institution":"Lunit","correspondingAuthor":false,"prefix":"","firstName":"Tae","middleName":"Soo","lastName":"Kim","suffix":""},{"id":310942149,"identity":"152fbfc4-119b-4494-8abe-905060f152e3","order_by":11,"name":"Hyeonsoo Lee","email":"","orcid":"","institution":"Lunit","correspondingAuthor":false,"prefix":"","firstName":"Hyeonsoo","middleName":"","lastName":"Lee","suffix":""},{"id":310942154,"identity":"9f1b6186-e79a-4d08-a573-6ffcd34b6a77","order_by":12,"name":"Ria Kwon","email":"","orcid":"","institution":"Total Healthcare Center, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Ria","middleName":"","lastName":"Kwon","suffix":""},{"id":310942156,"identity":"fc6a92a9-edee-482a-9b63-b59fd90c2774","order_by":13,"name":"Ga-Young Lim","email":"","orcid":"","institution":"Total Healthcare Center, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Ga-Young","middleName":"","lastName":"Lim","suffix":""},{"id":310942159,"identity":"6d3c6200-e21a-4aba-877f-c7608ee23e14","order_by":14,"name":"JunHyeok Choi","email":"","orcid":"","institution":"Sungkyunkwan University","correspondingAuthor":false,"prefix":"","firstName":"JunHyeok","middleName":"","lastName":"Choi","suffix":""},{"id":310942160,"identity":"b444708c-c94f-4f0d-b306-55ef9702704f","order_by":15,"name":"Shin Ho Kook","email":"","orcid":"","institution":"Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Shin","middleName":"Ho","lastName":"Kook","suffix":""},{"id":310942165,"identity":"b5840c40-be6a-4222-a7a9-01fd27fbfc89","order_by":16,"name":"Seungho Ryu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA40lEQVRIiWNgGAWjYDACCf4PBz+AGOwNQMLAghgtDIaPJUAMngMgLRJEaTE24AEzEiBcgkB+dkOahATDYTl+yedXN/wokGDgb+9OwKvF4M6BYxIFDIeNJWfnlN3sATpM4szZDfi1SCS2gWxJ3HA7J+0GD1CLgUQufi3yM5LZJHhAWm6eSbv5hxgtDDfSmA3AWm6wH7tNlC0GN3IYH0sYpBtL9uSw3ZYxkOAh6Bf5GTkMBz9UWMvxsx9/dvPNHxs5/vZeAg6D2NUMJHgMQEweIpSDQR0Qsz8gVvUoGAWjYBSMMAAAlCdFdPDNkngAAAAASUVORK5CYII=","orcid":"","institution":"Total Healthcare Center, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine","correspondingAuthor":true,"prefix":"","firstName":"Seungho","middleName":"","lastName":"Ryu","suffix":""}],"badges":[],"createdAt":"2024-05-24 08:26:34","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4471074/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4471074/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s10549-024-07541-1","type":"published","date":"2024-11-01T16:20:02+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":58077599,"identity":"5c6810e6-76d2-4ec0-b5fd-f745517ab24e","added_by":"auto","created_at":"2024-06-10 22:35:21","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":36804,"visible":true,"origin":"","legend":"\u003cp\u003eA flow chart of study population; A total of 74,610 women were finally participated in this study.\u003c/p\u003e","description":"","filename":"BRCTLIBRAAIFigure1.1.png","url":"https://assets-eu.researchsquare.com/files/rs-4471074/v1/4cd7ba8d27513eb33c9f55ff.png"},{"id":68207124,"identity":"9e73f7ce-9232-44e0-b2ab-6edf2f69553c","added_by":"auto","created_at":"2024-11-04 16:35:07","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1087078,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4471074/v1/eedd1985-203e-4117-b864-be71d791c847.pdf"},{"id":58077600,"identity":"66f88620-674d-43e0-9f3f-2abce9f11acf","added_by":"auto","created_at":"2024-06-10 22:35:21","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":38923,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupplementary Table 1\u003c/strong\u003e. Risk of incident breast cancer by ordinal LIBRA, radiologist, and AI-driven breast density (N=74,610)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSupplementary Table 2\u003c/strong\u003e. Risk of invasive breast cancer according to the discrepancy between Radiologist's and LIBRA and AI-method breast density (n= 74,610)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSupplementary \u0026nbsp;Table 3\u003c/strong\u003e. Risk of DCIS according to the discrepancy between Radiologist's, LIBRA and AI-method breast density (n= 74,610)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSupplementary \u0026nbsp;Table 4\u003c/strong\u003e. Comparison of the discriminatory power of LIBRA, Radiologist, and AI-driven breast density in prediction of breast cancer\u003c/p\u003e","description":"","filename":"BRCTLIBRAAISupplementaryfilev1.1.docx","url":"https://assets-eu.researchsquare.com/files/rs-4471074/v1/d44d671a340911c199779cf5.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Concordant and Discordant Breast Density Patterns by Different approaches for Assessing Breast Density and Breast Cancer Risk","fulltext":[{"header":"Introduction","content":"\u003cp\u003eDense breast tissue, as detected through mammography, is considered one of strongest risk factor for breast cancer [\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. This finding is consistently observed across diverse populations, including Western and Asian pre- and postmenopausal women, highlighting the crucial role of breast density in predicting future breast cancer risk [\u003cspan additionalcitationids=\"CR4 CR5 CR6\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Additionally, more than 50% of Asian women have dense breast tissue, which is not only independent risk factor that used in various risk prediction models but also for the identification of eligible women for supplemental screening methods [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Thus, improving breast density assessment could potentially enhance the effectiveness and management of breast cancer screening.\u003c/p\u003e \u003cp\u003eIt has been reported that breast cancer risk increases with higher mammographic density, whether measured quantitatively or qualitatively [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Qualitative assessments by radiologists, using the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS), are the most common standard for evaluating mammographic breast density [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. However, the reliability of BI-RADS categories among radiologists can vary from moderate to substantial, potentially leading to misclassification and, consequently, the under- or overestimation of breast cancer risk [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. To address these issues, automated density assessment methods such as AI or quantitative tools like the open-source LIBRA (Laboratory for Individualized Breast Radiodensity Assessment) software, which calculates the fibroglandular tissue volume relative to total breast volume, provide more objective and reproducible measurements [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. AI-driven qualitative assessments, trained on extensive data sets with ground truths provided by radiologists, have shown fair agreement with radiologist assessments [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Despite this, the comparative risks of breast cancer development based on breast density assessments between radiologists and AI-driven assessments, particularly in cohort studies involving Asian women with dense breast tissue, remain underexplored. Moreover, the use of AI and quantitative methods can lead to either concordant or discordant density assessments, and the implications of these outcomes require further investigation.\u003c/p\u003e \u003cp\u003eTherefore, our study aimed to achieve two objectives: 1) assess the relationship between mammographic dense breasts, as defined by radiologists, LIBRA, and AI, and incident breast cancer, and 2) examine how concordance or discordance in breast density classifications between radiologists and either LIBRA or AI affects the risk of breast cancer.\u003c/p\u003e"},{"header":"Material and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy population\u003c/h2\u003e \u003cp\u003eThe Kangbuk Samsung Health Study is a cohort study of Korean men and women aged\u0026thinsp;\u0026ge;\u0026thinsp;18 years who underwent comprehensive annual or biennial health examinations at the Kangbuk Samsung Hospital Total Healthcare Centers located in Seoul and Suwon, South Korea, as previously described [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. This study received approval from the Institutional Review Board (2023-02-004).\u003c/p\u003e \u003cp\u003eThis retrospective study primarily focused on Korean women aged\u0026thinsp;\u0026ge;\u0026thinsp;34 years who underwent their initial digital screening mammography at our institution as part of a health examination between January 2009 and 2014, and followed up until December 2020 (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Participants provided informed consent for linkage to the national cancer registry data were included in the study. Notably, although national breast cancer screening recommendations in Korea start at age 40, there is a common practice of private screening organizations offering screenings from age 35 [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Moreover, acknowledging the differentiation between Korean age and Western age, resulting in a potential variance of 1\u0026ndash;2 years in documented data, we have incorporated instances where the verified age distribution is 34 years, backed by authentic data. Participants were excluded if they had insufficient follow-up duration of \u0026lt;\u0026thinsp;12 months, had a history of breast surgery, history of mammoplasty, had a history of breast cancer or a prior registered breast cancer before mammography. After excluding missing or inappropriate values related to breast density or BI-RADS classification, the final analysis included 74,610 women (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eData collection\u003c/h2\u003e \u003cp\u003eDemographic information, first-degree family history of breast cancer, behavioral factors, reproductive factors, and medical history were collected through standardized, self-administered questionnaires. Body height and weight were measured by trained nurses, with participants wearing a hospital gown and no shoes. Body mass index (BMI) was then classified according to Asian-specific criteria [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]: underweight, \u0026lt;\u0026thinsp;18.5 kg/m\u003csup\u003e2\u003c/sup\u003e; normal weight, 18.5 to 23 kg/m\u003csup\u003e2\u003c/sup\u003e; overweight, 23 to 25 kg/m\u003csup\u003e2\u003c/sup\u003e; and obese, \u0026ge;\u0026thinsp;25 kg/m\u003csup\u003e2\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe Gail scores, predicting breast cancer over 5 year, were calculated [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], and were categorized into three risk groups; low (less than 0.8%), intermediate (0.8\u0026ndash;1.67%), and high (greater than or equal to 1.67%).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eMammographic Breast Density Measures\u003c/h2\u003e \u003cp\u003eMammographic imaging data, including BI-RADS final assessment categories and mammographic density, were retrieved from the original radiological reports. Participants in the study underwent a standard four-view digital mammography examination, which involved bilateral craniocaudal (CC) and mediolateral oblique (MLO) views, using a full-field digital mammography (FFDM) system (Sonograph 2000D/DMR/DS, GE Healthcare, Chicago, IL, USA, or Selenia, Hologic, Marlborough, MA, USA) at the Suwon and Seoul Total Healthcare Centers. The assessment of mammography was conducted by experienced breast imaging radiologists at the two centers, employing the BI-RADS classification system. Breast density was assessed by radiologists and categorized based on the BI-RADS assessment, including category A (almost entirely fatty), category B (scattered fibroglandular densities), category C (heterogeneously dense), or category D (extremely dense). While the 5th edition of BI-RADS was released in 2013, most of our baseline data on breast density, collected between 2009 and 2014, was obtained before its implementation in our centers. Therefore, the breast density categories may reflect assessments made with the earlier 4th edition of BI-RADS, as radiologists reviewed the FFDM data from this study using that edition [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eLIBRA, completely automated, freely available open-source software, can be utilized on both raw and processed FFDM images to produce area-based measurements of mammographic breast density [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. LIBRA provides an estimate of the total absolute dense area (DA), whereas normalizing DA by the total breast area results in breast percent density (PD). For this study, we used the DA and PD estimates obtained with LIBRA from images processed and stored in DICOM format for each woman, averaging density measures from all four mammographic views, bilateral CC and MLO views [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Density assessments by LIBRA were strongly correlated with widely-used Volpara\u0026reg; methods in our study (r\u0026thinsp;=\u0026thinsp;0.89, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Quantitative breast density assessed by LIBRA were categorized based on the BI-RADS assessment (type A, almost entirely fatty, \u0026lt; 25% fibroglandular tissue; type B, scattered fibroglandular tissue, 25\u0026ndash;50% fibroglandular tissue; type C, heterogeneously dense, 51\u0026ndash;75% fibroglandular tissue; type D, extremely dense, \u0026gt; 75% fibroglandular tissue) [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFor the study, the AI algorithm (Lunit INSIGHT MMG, version 1.1.7.2, Seoul, Korea) was retrospectively applied to stored mammographic images. The AI density task module provided the average density results for four (bilateral CC, MLO) views of each mammography. The average mammographic density was presented on a 1\u0026ndash;10 scale, with density categories defined as follows: Density A (Scores 1 to 2), Density B (Scores 3 to 5), Density C (Scores 6 to 8), and Density D (Scores 9 to 10), following the recommendations of provider [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e].\u003c/p\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003eDefinition of breast cancer\u003c/h2\u003e \u003cp\u003eThe breast cancer diagnosis after a screening mammography was established by linking the study data to the Korean Central Cancer Registry [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. In this context, breast cancer was defined as either invasive cancer (International Classification of Diseases-10 code C50) or ductal carcinoma in situ (International Classification of Diseases-10 code D05.1).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analyses\u003c/h2\u003e \u003cp\u003eThe primary endpoint was the development of breast cancer. The occurrence of breast cancer was measured in terms of the number of cases per 1000 person-years, and the follow-up period extended from the baseline visit until the date of the primary endpoint or the last health screening examination (December 31, 2020), whichever came first. The participants' breast density was classified into the following categories: A-B (non-dense), and C-D (dense).\u003c/p\u003e \u003cp\u003eWe evaluated incident breast cancer cases based on concordant or discordant breast density patterns between radiologists and LIBRA or AI. Cases where radiologists classified breasts as non-dense, in agreement with LIBRA or AI assessments, were designated as \"concordant non-dense breast.\" Likewise, cases where radiologists and LIBRA or AI both categorized breasts as dense were designated as \"concordant dense breast.\" Discordant breast density patterns were defined as cases where radiologists reported breasts as non-dense while LIBRA or AI indicated dense breast tissue, or vice versa.\u003c/p\u003e \u003cp\u003eTo assess the relationship of 1) breast density assessment and incident breast cancer, and 2) concordant and discordant breast density patterns and incident breast cancer, according to different measurement, radiologists, LIBRA, and AI-driven, cox proportional hazard models were used to calculate adjusted hazard ratios (aHRs) with 95% confidence intervals (CIs) for the primary endpoint. The multivariable-adjusted model was gradually adjusted for covariates, including age, BMI, family history of breast cancer, reproductive histories (age at menarche, parity, menopausal status, and female hormone use), education level (below college graduate, college graduate or higher, or unknown), smoking status (never, former, current smoker, or unknown), alcohol consumption (\u0026lt;\u0026thinsp;10 or \u0026ge;\u0026thinsp;10 g/day), and physical activity level (inactive, minimally active, HEPA, or unknown).\u003c/p\u003e \u003cp\u003eHarrell's C-index (the area under the receiver operating characteristic curve [AUROC]), a measure of the concordance probability adapted for survival analysis, was used to assess whether the addition of radiologists, LIBRA, and AI individually or concurrently to the base model, including conventional risk factors, improved prediction of breast cancer. C statistics are routinely applied for global assessments of discrimination in a survival model [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. All analyses were carried out using Stata software version 18.0 (StataCorp LLC, College Station, TX, USA). Statistical significance was defined as a two-tailed P-value less than 0.05.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eBaseline characteristics\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents the baseline characteristics of the participants, categorized into non-dense and dense groups based on breast density measurements from radiologists, LIBRA, and AI algorithm. At baseline, dense breasts were classified by radiologists (87.7%), AI-program (84.5%), and LIBRA (27.6%). Among the 74,610 Asian women selected (mean age: 42.1\u0026thinsp;\u0026plusmn;\u0026thinsp;8.2 years), individuals with dense breasts determined by those measures were consistently younger, had an earlier menarche, were more likely to be nulliparous, had lower BMI, and attained higher education levels.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCharacteristics of the women in the study according to breast density assessment measures (n\u0026thinsp;=\u0026thinsp;74,610)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003eRadiologist\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003eAI-driven\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c8\" namest=\"c7\"\u003e \u003cp\u003eLIBRA\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNon-dense\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDense\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNon-dense\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eDense\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eNon-dense\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eDense\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eN\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e74,610 (100.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9,185 (12.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e65,425 (87.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e11,576 (15.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e63,034 (84.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e53,987 (72.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e20,623 (27.6)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge (SD), year\u003c/b\u003e\u003csup\u003e\u003cb\u003ea\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e42.10 (8.17)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e51.08 (11.09)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e40.84(6.77)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e49.80 (11.40)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e40.69(6.49)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e43.47(8.75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e38.52 (4.79)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge at menarche (SD), year\u003c/b\u003e\u003csup\u003e\u003cb\u003ea\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e14.08 (1.63)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14.79 (1.95)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e13.99(1.54)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e14.650 (1.93)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e13.98(1.55)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e14.14(1.68)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e13.94(1.49)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eBMI category\u003c/b\u003e\u003csup\u003e\u003cb\u003eb\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c8\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026le;\u0026thinsp;22.9 kg/m2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e50,758 (68.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3,351 (36.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e47,407 (72.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4,187 (36.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e46,571 (73.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e31,399 (58.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e19,359 (93.9)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e23\u0026ndash;24.9 kg/m2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12,450 (16.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2,399 (26.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10,051 (15.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3,012 (26.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e9,438 (15.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e11,493 (21.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e957 (4.6)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026ge;\u0026thinsp;25 kg/m2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11,274 (15.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3,414 (37.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7,860 (12.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4,358 (37.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e6,916 (11.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e10,999 (20.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e275 (1.3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emissing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e128 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e107 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e19 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e109 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e96 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e32 (0.2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eFamily history of breast cancer\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c8\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e72,196 (96.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e8,841 (96.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e63,355 (96.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e11,194 (96.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e61,002 (96.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e52,234 (96.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e19,962 (96.8)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2,190 (2.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e313 (3.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1,877 (2.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e346 (3.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1,844 (2.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1,585 (2.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e605 (2.9)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emissing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e224 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e31 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e193 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e36 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e188 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e168 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e56 (0.3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eParity\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c8\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003enulliparous\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5,617 (7.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e275 (3.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5,342 (8.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e288 (2.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5,329 (8.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e3,235 (6.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2,382 (11.6)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eparous\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e63,419 (85.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7,889 (85.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e55,530 (84.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e10,082 (87.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e53,337 (84.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e46,518 (86.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e16,901 (82.0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emissing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5,574 (7.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1,021 (11.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4,553 (7.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1,206 (10.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e4,368 (6.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e4,234 (7.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1,340 (6.5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMenopausal status\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c8\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epremenopausal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e52,184 (69.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3,026 (32.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e49,158 (75.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4,889 (42.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e47,295 (75.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e35,351 (65.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e16,833 (81.6)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epostmenopausal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11,440 (15.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4,940 (53.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6,500 (9.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5,702 (49.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5,738 (9.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e10,891 (20.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e549 (2.7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emissing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10,986 (14.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1,219 (13.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9,767 (14.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e985 (8.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e10,001 (15.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e7,745 (14.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3,241 (15.7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eFemale hormone use\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c8\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e73,047 (97.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e71,797 (97.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1,250 (98.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e11,228 (97.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e61,819 (98.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e52,725 (97.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e20,322 (98.5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1,371 (1.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1,350 (1.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e21 (1.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e319 (2.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1,052 (1.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1,116 (2.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e255 (1.2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emissing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e192 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e188 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e29 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e163 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e146 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e46 (0.2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEducation\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c8\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026lt; College graduate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e20,730 (27.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4,469 (48.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16,261 (24.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5,562 (48.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e15,168 (24.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e17,122 (31.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3,608 (17.5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026ge; College graduate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e48,771 (65.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3,945 (43.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e44,826 (68.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5,076 (43.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e43,695 (69.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e33,113 (61.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e15,658 (75.9)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emissing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5,109 (6.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e771 (8.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4,338 (6.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e938 (8.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e4,171 (6.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e3,752 (6.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1,357 (6.6)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePhysical activity\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c8\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLow\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e31,886 (42.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e31,405 (42.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e481 (37.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4,440 (38.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e27,446 (43.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e22,341 (41.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e9,545 (46.3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModerate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18,236 (24.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e17,935 (24.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e301 (23.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3,043 (26.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e15,193 (24.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e13,287 (24.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e4,949 (24.0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHigh\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9,834 (13.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9,659 (13.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e175 (13.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1,924 (16.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7,910 (12.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e7,486 (13.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2,348 (11.4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emissing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e14,654 (19.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14,336 (19.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e318 (24.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2,169 (18.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e12,485 (19.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e10,873 (20.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3,781 (18.3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAlcohol consumption\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u003cb\u003e(in grams/day)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c8\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e52,435 (70.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5,726 (62.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e46,709 (71.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7,421 (64.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e45,014 (71.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e37,083 (68.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e15,352 (74.4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026ge;\u0026thinsp;10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8,530 (11.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e881 (9.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7,649 (11.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1,184 (10.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7,346 (11.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e6,148 (11.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2,382 (11.6)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emissing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e13,645 (18.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2,578 (28.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e11,067 (16.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2,971 (25.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e10,674 (16.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e10,756 (19.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2,889 (14.0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSmoking status\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c8\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003enever smoker\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e53,923 (72.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6,555 (71.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e47,368 (72.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e8,068 (69.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e45,855 (72.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e38,678 (71.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e15,245 (73.9)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eformer smoker\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5,766 (7.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e513 (5.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5,253 (8.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e690 (6.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5,076 (8.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e3,980 (7.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1,786 (8.7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ecurrent smoker\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1,576 (2.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e186 (2.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1,390 (2.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e222 (1.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1,354 (2.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1,090 (2.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e486 (2.4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emissing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e13,345 (17.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1,931 (21.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e11,414 (17.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2,596 (22.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e10,749 (17.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e10,239 (19.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e3,106 (15.1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eFive-year risk based on Gail model\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c8\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026lt; 0.83%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e74,184 (99.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9,065 (98.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e65,119 (99.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e11,443 (98.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e62,741 (99.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e53,630 (99.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e20,554 (99.7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0.83\u0026ndash;1.66%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e191 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e86 (0.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e105 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e97 (0.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e94 (0.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e184 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e7 (0.0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026ge; 1.67%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0 (0.0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emissing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e235 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e34 (0.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e201 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e36 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e199 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e173 (0.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e62 (0.3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003eBreast density dichotomized into dense and non-dense categories based on the BI-RADS; Dense breast: \u0026ge;50%, Heterogeneously dense (C) and Extremely dense (D). Non-dense breast: 0\u0026thinsp;~\u0026thinsp;50%, Almost entirely fat (A) and Scattered fibroglandular densities (B)\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003eData are presented as \u003csup\u003ea\u003c/sup\u003emeans (standard deviation); \u003csup\u003eb\u003c/sup\u003eBMI: normal to underweight (\u0026le;\u0026thinsp;22.9 kg/m2), overweight (23\u0026ndash;24.9 kg/m2), and obese (\u0026ge;\u0026thinsp;25 kg/m2)\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003eAbbreviation: BMI, body mass index; LIBRA, Laboratory for Individualized Breast Radiodensity Assessment; AI, Artificial Intelligence\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eBreast cancer risk associated with dense breasts determined by LIBRA, radiologist, and AI-driven methods\u003c/h2\u003e \u003cp\u003eDuring a median follow-up of 9.9 years, 479 breast cancer cases were newly identified. Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents the risk of incident breast cancer by dense breasts, as measured by three methods, including radiologist, LIBRA and AI-driven breast density. Comparing dense breast groups estimated by LIBRA, radiologist, and AI to their respective non-dense counterparts (used as the reference), the age-adjusted HRs (95%CIs) for overall breast cancer development were 1.32 (1.06\u0026ndash;1.63) for LIBRA, 2.71 (1.93\u0026ndash;3.80) for radiologists, and 2.99 (2.18\u0026ndash;4.11) for AI methods. These association remained significant with corresponding aHR (95% CI) of 1.30 (1.05\u0026ndash;1.62), 2.37 (1.68\u0026ndash;3.36), and 2.55 (1.84\u0026ndash;3.56), after adjusting confounders, including age, BMI, age at menarche, family history of breast cancer, parity, menopausal status, female hormone use, education, physical activity, smoking, and alcohol intake. The findings were consistent across separate analyses of invasive cancer and DCIS (\u003cb\u003eTable \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e), although dense breasts identified by LIBRA were no longer associated with an increased risk of DCIS.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eRisk of incident breast cancer by ordinal LIBRA, radiologist, and AI-driven breast density (N\u0026thinsp;=\u0026thinsp;74,610)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBreast cancer (total)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePerson-years\u003c/p\u003e \u003cp\u003e(PY)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eIncidence\u003c/p\u003e \u003cp\u003eCases\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eIncidence rate\u003c/p\u003e \u003cp\u003e(/10^3 PY)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAge-adjusted\u003c/p\u003e \u003cp\u003eHR (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eMultivariable-adjusted\u003c/p\u003e \u003cp\u003eHR\u003csup\u003ea\u003c/sup\u003e (95% CI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLIBRA breast density\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNon-dense (N\u0026thinsp;=\u0026thinsp;53,987)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e192.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e144\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.00 (reference)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.00 (reference)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDense (N\u0026thinsp;=\u0026thinsp;20,623)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e73.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e335\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.32 (1.06\u0026ndash;1.63)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.30 (1.05\u0026ndash;1.62)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRadiologist breast density\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNon-dense (N\u0026thinsp;=\u0026thinsp;9,185)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e41.35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.06\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.00 (reference)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.00 (reference)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDense (N\u0026thinsp;=\u0026thinsp;65,425)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e224.36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e435\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.71 (1.93\u0026ndash;3.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2.37 (1.68\u0026ndash;3.36)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAI-driven breast density\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNon-dense (N\u0026thinsp;=\u0026thinsp;11,576)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e51.07\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e53\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.04\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.00 (reference)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.00 (reference)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDense (N\u0026thinsp;=\u0026thinsp;63,034)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e214.64\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e426\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.99 (2.18\u0026ndash;4.11)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2.55 (1.84\u0026ndash;3.56)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"6\"\u003eBreast density dichotomized into dense and non-dense categories based on the BI-RADS; Dense breast: \u0026ge;50%, Heterogeneously dense (C) and Extremely dense (D). Non-dense breast: 0\u0026thinsp;~\u0026thinsp;50%, Almost entirely fat (A) and Scattered fibroglandular densities (B)\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"6\"\u003ea. Hazard ratios adjusted for age, BMI, age at menarche, family history of breast cancer, parity, menopausal status, female hormone use, education, physical activity, smoking, and alcohol intake\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eBreast cancer risk according to concordant and discordant dense breast assessment by Radiologists and LIBRA or AI-driven methods\u003c/b\u003e \u003c/p\u003e \u003cp\u003eWhen comparing discordant and concordant dense breast patterns to the concordant non-dense breast pattern, identified by both radiologists and LIBRA, the aHRs (95% CI) for incident breast cancer were as follows; 2.40 (1.69\u0026ndash;3.41) for radiologist-identified dense but LIBRA-identified non-dense breasts, 11.99 (1.64\u0026ndash;87.62) for radiologist-identified non-dense but LIBRA-identified dense breasts (notably, only one case of breast cancer occurred in this category), and 2.99 (1.99\u0026ndash;4.50) for both radiologist- and LIBRA-identified dense breasts (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Additionally, when considering breast density assessments provided by radiologists and AI-driven methods, the aHRs for incident breast cancer were as follows; 1.79 (1.02\u0026ndash;3.12) for radiologist-identified dense breasts but AI-identified non-dense breasts, 2.43 (1.24\u0026ndash;4.78) for radiologist-identified non-dense but AI-identified dense breasts, and 3.23 (2.15\u0026ndash;4.86) for both radiologist- and AI-identified dense breasts (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). When analyses were conducted separately for breast cancer subtypes, including invasive cancer and DCIS, the findings remained similar for invasive cancer (\u003cb\u003eTables S2 and S3\u003c/b\u003e). However, the number of incident DCIS cases were too small to estimate the risk in discordant breast density patterns assessed by radiologist and LIBRA, or radiologist and Al algorithm.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eRisk of breast cancer according to the discrepancy between Radiologist's, LIBRA and AI-method breast density (n\u0026thinsp;=\u0026thinsp;74,610)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePerson-years\u003c/p\u003e \u003cp\u003e(PY)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eIncidence\u003c/p\u003e \u003cp\u003eCases\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eIncidence rate\u003c/p\u003e \u003cp\u003e(/10\u003csup\u003e3\u003c/sup\u003e PY)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eMultivariable-adjusted\u003c/p\u003e \u003cp\u003eHR\u003csup\u003ea\u003c/sup\u003e (95% CI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRadiologist\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eLIBRA\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNondense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNondense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e9,138\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e41,240\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.04\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.00 (reference)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNondense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e44,849\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e151,300\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e292\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2.4 (1.69\u0026ndash;3.41)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNondense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e112\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e8.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e11.99 (1.64\u0026ndash;87.62)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e20,576\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e73,060\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e143\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2.99 (1.99\u0026ndash;4.50)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRadiologist\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eAI-methods\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNondense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNondense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e7,095\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e35,080\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.00 (reference)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNondense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4,481\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e15,990\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.31\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.79 (1.02\u0026ndash;3.12)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNondense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2,090\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6,268\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e2.43 (1.24\u0026ndash;4.78)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e60,944\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e208,400\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e414\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e3.23 (2.15\u0026ndash;4.86)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"7\"\u003eBreast density dichotomized into dense and non-dense categories based on the BI-RADS; Dense breast: \u0026ge;50%, Heterogeneously dense (C) and Extremely dense (D). Non-dense breast: 0\u0026thinsp;~\u0026thinsp;50%, Almost entirely fat (A) and Scattered fibroglandular densities (B)\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"7\"\u003ea. Hazard ratios adjusted for age, BMI, age at menarche, family history of breast cancer, parity, female hormone use, menopausal status, education, physical activity, smoking, and alcohol intake\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003ePredictive ability for incident breast cancer determined after adding radiologists, LIBRA, and AI \u0026ndash;identified dense breasts to the Gail model\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe incremental predictive ability for incident breast cancer was determined after adding breast density assessments from radiologists, LIBRA, or AI individually, in pairs or altogether to a base model composed of conventional breast cancer risk factors, represented as the Gail score (\u003cb\u003eTable S4\u003c/b\u003e). Adding dense breast assessed by either radiologists or AI alone to the base model (Gail model) significantly improved the AUROCs for predicting incident breast cancer. The AUROC for additions by radiologists was 0.533 (95% CI: 0.519\u0026ndash;0.548), and for AI, it was 0.540 (95% CI: 0.524\u0026ndash;0.556). Although the inclusion of dense breast assessments by LIBRA also improved cancer prediction, it did not achieve statistical significance, with an AUROC of 0.518 (95% CI: 0.496\u0026ndash;0.539) compared to the Gail model. The addition of dense breast assessment by both radiologist and AI-method to the base model further improved the predictive ability of developing breast cancer, resulting in an AUROC, 95% CI 0.542, 0.526\u0026ndash;0.559) (\u003cb\u003eTable S4\u003c/b\u003e).\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur study compared qualitative, quantitative, and AI-driven assessments of breast density and found that each method, when evaluated individually by radiologists, LIBRA, or AI, demonstrated a significant association with increased risk of breast cancer. Furthermore, discordant breast density patterns between radiologists and LIBRA or AI were also associated with an increased risk of incident breast cancer. Notably, breasts identified as dense by LIBRA or AI methods\u0026mdash;even when deemed non-dense by radiologists\u0026mdash;showed a consistently higher risk of breast cancer, with a relatively greater hazard ratio than cases identified solely by radiologists. The dense breast assessed by radiologist and/or AI improved risk prediction for incident breast cancer (based on Harrell\u0026rsquo;s C\u0026minus;index) compared to conventional breast cancer risk factors, estimated as Gail model. Our results suggest that dense breasts identified by AI or LIBRA may provide complementary information into the risk of developing breast cancer, compared to those identified solely by radiologists.\u003c/p\u003e \u003cp\u003eOur study is the first cohort study to assess the risk of developing subsequent breast cancer based on individual measures of breast density (radiologist, LIBRA, and AI) and the implications of concordant and discordant assessments of dense breast by radiologist and LIBRA or AI. A cross-sectional study for 488 Korean women evaluated and compared the inter-rater agreements between radiologists, AI and another commercial automated density assessment program (Volpara\u0026reg;) [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e], and found that density assessments by AI showed similar agreement with those of radiologists compared to the Volpara\u0026reg; (κ\u0026thinsp;=\u0026thinsp;0.52 and 0.50, respectively). The study, however, did not extend to investigate the longitudinal association between either dense breast assessment by individual breast density measures or their discrepant interpretations in dense breast and the risk of incident breast cancer [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn our study, breast density assessed by different approaches was all associated with increased risk of incident breast cancer. This study showed slightly weaker breast cancer risk for LIBRA-defined dense breast than radiologists or AI. However, few studies investigated the LIBRA assessments for images before a diagnosis of breast cancer. A case-control study by Gastounioti et al. demonstrated that LIBRA PD showed comparable breast cancer association (OR, 1.3; processed images and OR, 1.2; raw images) with semiautomated area-based assessment tool Cumulus (OR, 1.5; processed images) and Volpara (OR, 1.4; raw images) [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. A recent cohort study, comprising 21,000 non-Hispanic white women aged 40\u0026ndash;74, evaluated the long-term breast cancer risk (up to 10 years) using LIBRA [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. The study found that LIBRA measures of density were associated with breast cancer risk, similar to the results obtained with Cumulus measures, a semi-automated validated tool [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. After adjusting for BMI and reproductive variables, the aHR for breast cancer associated with each standard deviation increase in percent density, measured by LIBRA, was 1.44 (95%CI: 1.26\u0026ndash;1.66) [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Our study showed a comparable age-adjusted HR of 1.32 (95% CI: 1.06\u0026ndash;1.63) for LIBRA, and AI achieved the strongest capacity in breast cancer risk prediction with age-adjusted HR of 2.99 (95% CI: 2.18\u0026ndash;4.11). Further study in diverse population is necessary to examine which approach has the greatest robustness in accurate risk assessment of future breast cancer.\u003c/p\u003e \u003cp\u003eThere have been several studies on the use of various AI models for mammographic density assessments. A previous study compared two automated methods and visual assessment in contralateral breasts of women with breast cancer showed similar associations with breast cancer [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. However, a study of five methods of measuring breast density on prior mammograms demonstrated that visual assessment predicted subsequent breast cancer risk significantly than all other density methods [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. In our study, AI-driven breast density had the strongest association with subsequent breast cancer. According to the AI models, the algorithms are diverse, in terms of algorithm operation and development dataset. In this study, we used commercially available AI model based on BI-RADS density classification by experienced radiologists. Although it remains undiscovered which factor has a greater impact on breast cancer risk, our study suggests the feasibility of AI-driven mammographic density as a reliable marker.\u003c/p\u003e \u003cp\u003eOur findings revealed that reader-dependent, subjective measures of density and categorical density data (fatty, scattered, heterogeneously dense, and extremely dense in accordance with BI-RADS) could be concordant or discordant to LIBRA or AI-defined density, potentially contributing to breast cancer risk. The assessment of breast density by LIBRA or AI could complement the evaluation made by radiologists, especially in cases where there's a discrepancy, such as AI identifying density while the radiologist does not. In this study, discordance resulted from the radiologists\u0026rsquo; tendency to assign density grades higher than those obtained by LIBRA, and radiologist-defined non-dense but LIBRA-defined dense breasts yielded a sample size that was too small to estimate the risk of breast cancer.\u003c/p\u003e \u003cp\u003eOur results favored the AI-derived dense breast model demonstrating the improvement of predictive performance when adding to the Gail model (0.540 vs. 0.503, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01), although the C-statistic for the Gail model, the most frequently used breast cancer risk model, was low as much as 0.50, corresponding to previous studies found that the risk models have demonstrated only limited prediction performance [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. The modified Gail model has allowed for risk estimates for Asian-American women [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e], albeit its C-statistics is low and has been estimated to be equal to 0.54 in an external validation study for Korean 40 229 women [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. According to an external validation study in a U.S. screening cohort, the predictive performance of a mammography-derived AI risk model was significantly higher than that of the Gail model (0.68 vs. 0.55, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01) [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. The limited discriminatory accuracy might be attributed to recent increase in the incidence of breast cancer in Korea; previous validation study population dataset was outdated, without updating recent increase in the incidence of breast cancer in Korea. Moreover, our study is not a prospective cohort of the general population but we used retrospective data from breast cancer screening patients, predominantly comprising young and relatively healthy women with mostly low Gail risk scores. Future research should consider using an Asian-specific breast cancer risk prediction model to provide a more profound understanding of the relationship between breast density and breast cancer, supported by further external validation studies.\u003c/p\u003e \u003cp\u003eOur study had several limitations that should be considered. First, we evaluated a single automated density assessment program and AI program. The performance of different programs varies and our results cannot be generalized to other programs immediately. Secondly, we did not consider changes in breast density over time, the potential impact on the development of breast cancer, or the discriminatory accuracy in predicting breast cancer. Further studies using longitudinal digital mammograms accounting dynamic changes in density over time may be useful to refine more accurate breast cancer risk. Third, the study population comprised well-educated women with potentially high accessibility to medical services, limiting generalizability. Lastly, our analysis was based on retrospective data collected during routine health screening examinations and previous radiologic reports of screening mammography. Therefore, we did not assess the utility of the AI algorithm or LIBRA for radiologists in a real screening setting, nor did we evaluate its potential impact on screening performance when used by radiologists. To comprehensively understand the effectiveness of the AI algorithm or LIBRA in a real-world screening environment, further prospective studies are necessary.\u003c/p\u003e \u003cp\u003eCurrently, there is a lack of study regarding the assessment of breast cancer risks based on concordant and discordant breast density patterns, utilizing various approaches for assessing breast density. Accurate identification of women with dense breast tissue could facilitate the development of strategies for additional screening and primary breast cancer prevention. Future prospective studies with a more diverse population are required whether incorporating LIBRA or AI measurements to radiologists' interpretations for dense breasts could enhance breast cancer risk assessments in real clinical settings.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eBI-RADS,\u0026nbsp;Breast Imaging Reporting \u0026amp; Data System\u003c/p\u003e\n\u003cp\u003eCI, confidence interval\u003c/p\u003e\n\u003cp\u003eHR, hazard ratio\u003c/p\u003e\n\u003cp\u003eLIBRA, Laboratory for Individualized Breast Radiodensity Assessment\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAI, Artificial Intelligence\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFFDM, full-field digital mammography\u003c/p\u003e\n\u003cp\u003ePD, percent density\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eVPD, volumetric percent density\u003c/p\u003e\n\u003cp\u003eDA, dense area\u003c/p\u003e\n\u003cp\u003eSD, standard deviation\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the SKKU Excellence in Research Award Research Fund, Sungkyunkwan University (2021) and Lunit Inc.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYCho, EK, YChang, and SR planned, designed and implemented the study, including quality assurance and control. YChang,\u0026nbsp;MK\u0026nbsp;and SR analyzed the data and designed the study\u0026rsquo;s analytic strategy. YChang, and SR supervised field activities.\u0026nbsp;Material preparation, and data collection were performed by all authors. The first draft of the manuscript was written by YCho, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated during and/or analyzed during the current study are not publicly available due outside of Institutional Review Board restrictions (the data were not collected in a way that could be distributed widely) but are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved by the Institutional Review Board of Kangbuk Samsung Hospital (2023-02-004), and the need for informed consent was waived due to the use of de-identified retrospective data collected during the health screening process.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the Medical Research Funds from Kangbuk Samsung Hospital and the National Research Foundation of Korea (NRF) grant funded by the Korea government (Ministry of Science and ICT, MSIT) (RS-2023-00253017).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eMcCormack VA, dos Santos Silva I (2006) Breast density and parenchymal patterns as markers of breast cancer risk: a meta-analysis. Cancer Epidemiol Biomarkers Prev 15(6):1159\u0026ndash;1169\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoyd NF, Guo H, Martin LJ, Sun L, Stone J, Fishell E, Jong RA, Hislop G, Chiarelli A, Minkin S, Yaffe MJ (2007) Mammographic density and the risk and detection of breast cancer. N Engl J Med 356(3):227\u0026ndash;236\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePettersson A, Graff RE, Ursin G, Santos Silva ID, McCormack V, Baglietto L, Vachon C, Bakker MF, Giles GG, Chia KS et al (2014) Mammographic density phenotypes and risk of breast cancer: a meta-analysis. J Natl Cancer Inst 106(5)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoyd NF, Guo H, Martin LJ, Sun L, Stone J, Fishell E, Jong RA, Hislop G, Chiarelli A, Minkin S, Yaffe MJ (2007) Mammographic Density and the Risk and Detection of Breast Cancer. N Engl J Med 356(3):227\u0026ndash;236\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMariapun S, Li J, Yip CH, Taib NA, Teo SH (2015) Ethnic differences in mammographic densities: an Asian cross-sectional study. PLoS ONE 10(2):e0117568\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim S, Tran TXM, Song H, Ryu S, Chang Y, Park B (2022) Mammographic Breast Density, Benign Breast Disease, and Subsequent Breast Cancer Risk in 3.9 Million Korean Women. Radiology 304(3):534\u0026ndash;541\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim EY, Chang Y, Ahn J, Yun JS, Park YL, Park CH, Shin H, Ryu S (2020) Mammographic breast density, its changes, and breast cancer risk in premenopausal and postmenopausal women. Cancer 126(21):4687\u0026ndash;4696\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLau S, Abdul Aziz YF, Ng KH (2017) Mammographic compression in Asian women. PLoS ONE 12(4):e0175781\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChalfant JS, Hoyt AC (2022) Breast Density: Current Knowledge, Assessment Methods, and Clinical Implications. J Breast Imaging 4(4):357\u0026ndash;370\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRadiology ACo (2003) Breast imaging reporting and data system. \u003cem\u003eBI-RADS\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerg WA, Campassi C, Langenberg P, Sexton MJ (2000) Breast Imaging Reporting and Data System: inter- and intraobserver variability in feature analysis and final assessment. AJR Am J Roentgenol 174(6):1769\u0026ndash;1777\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoyd NF, Wolfson C, Moskowitz M, Carlile T, Petitclerc C, Ferri HA, Fishell E, Gregoire A, Kiernan M, Longley JD et al (1986) Observer variation in the classification of mammographic parenchymal patterns. J Chronic Dis 39(6):465\u0026ndash;472\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGastounioti A, Kasi CD, Scott CG, Brandt KR, Jensen MR, Hruska CB, Wu FF, Norman AD, Conant EF, Winham SJ et al (2020) Evaluation of LIBRA Software for Fully Automated Mammographic Density Assessment in Breast Cancer Risk Prediction. Radiology 296(1):24\u0026ndash;31\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchaffter T, Buist DSM, Lee CI, Nikulin Y, Ribli D, Guan Y, Lotter W, Jie Z, Du H, Wang S et al (2020) Evaluation of Combined Artificial Intelligence and Radiologist Assessment to Interpret Screening Mammograms. JAMA Netw Open 3(3):e200265\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee SE, Son NH, Kim MH, Kim EK (2022) Mammographic Density Assessment by Artificial Intelligence-Based Computer-Assisted Diagnosis: A Comparison with Automated Volumetric Assessment. J Digit Imaging 35(2):173\u0026ndash;179\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChang Y, Ryu S, Choi Y, Zhang Y, Cho J, Kwon MJ, Hyun YY, Lee KB, Kim H, Jung HS et al (2016) Metabolically Healthy Obesity and Development of Chronic Kidney Disease: A Cohort Study. Ann Intern Med 164(5):305\u0026ndash;312\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee SY, Jeong SH, Kim YN, Kim J, Kang DR, Kim HC, Nam CM (2009) Cost-effective mammography screening in Korea: high incidence of breast cancer in young women. Cancer Sci 100(6):1105\u0026ndash;1111\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee EH, Park B, Kim NS, Seo HJ, Ko KL, Min JW, Shin MH, Lee K, Lee S, Choi N et al (2015) The Korean guideline for breast cancer screening. J Korean Med Assoc 58(5):408\u0026ndash;419\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Health Organization, Regional Office for the Western Pacific (2000) The Asia-Pacific perspective: redefining obesity and its treatment. Health Communications Australia, Sydney\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGail MH, Brinton LA, Byar DP, Corle DK, Green SB, Schairer C, Mulvihill JJ (1989) Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst 81(24):1879\u0026ndash;1886\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eD\u0026rsquo;Orsi CJME, Ikeda DM et al (2003) Breast Imaging Reporting and Data System: ACR BI-RADS\u0026mdash;Breast Imaging Atlas. American College of Radiology\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeller BM, Chen J, Daye D, Conant EF, Kontos D (2015) Preliminary evaluation of the publicly available Laboratory for Breast Radiodensity Assessment (LIBRA) software tool: comparison of fully automated area and volumetric density measures in a case-control study with digital mammography. Breast Cancer Res 17:117\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKwon MR, Chang Y, Park B, Ryu S, Kook SH (2023) Performance analysis of screening mammography in Asian women under 40 years. Breast Cancer 30(2):241\u0026ndash;248\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUno H, Cai T, Pencina MJ, D'Agostino RB, Wei LJ (2011) On the C-statistics for evaluating overall adequacy of risk prediction procedures with censored survival data. Stat Med 30(10):1105\u0026ndash;1117\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHabel LA, Alexeeff SE, Achacoso N, Arasu VA, Gastounioti A, Gerstley L, Klein RJ, Liang RY, Lipson JA, Mankowski W et al (2023) Examination of fully automated mammographic density measures using LIBRA and breast cancer risk in a cohort of 21,000 non-Hispanic white women. Breast Cancer Res 25(1):92\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrandt KR, Scott CG, Ma L, Mahmoudzadeh AP, Jensen MR, Whaley DH, Wu FF, Malkov S, Hruska CB, Norman AD et al (2016) Comparison of Clinical and Automated Breast Density Measurements: Implications for Risk Prediction and Supplemental Screening. Radiology 279(3):710\u0026ndash;719\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAstley SM, Harkness EF, Sergeant JC, Warwick J, Stavrinos P, Warren R, Wilson M, Beetles U, Gadde S, Lim Y et al (2018) A comparison of five methods of measuring mammographic density: a case-control study. Breast Cancer Res 20(1):10\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVilmun BM, Vejborg I, Lynge E, Lillholm M, Nielsen M, Nielsen MB, Carlsen JF (2020) Impact of adding breast density to breast cancer risk models: A systematic review. Eur J Radiol 127:109019\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCostantino JP, Gail MH, Pee D, Anderson S, Redmond CK, Benichou J, Wieand HS (1999) Validation studies for models projecting the risk of invasive and total breast cancer incidence. J Natl Cancer Inst 91(18):1541\u0026ndash;1548\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMin JW, Chang MC, Lee HK, Hur MH, Noh DY, Yoon JH, Jung Y, Yang JH (2014) Korean Breast Cancer S: Validation of risk assessment models for predicting the incidence of breast cancer in korean women. J Breast Cancer 17(3):226\u0026ndash;235\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGastounioti A, Eriksson M, Cohen EA, Mankowski W, Pantalone L, Ehsan S, McCarthy AM, Kontos D, Hall P, Conant EF (2022) External Validation of a Mammography-Derived AI-Based Risk Model in a U.S. Breast Cancer Screening Cohort of White and Black Women. \u003cem\u003eCancers (Basel)\u003c/em\u003e 14(19)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSupplementary file 1\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSupplementary Table 1 Risk of incident breast cancer by ordinal LIBRA, radiologist, and AI-driven breast density (N\u0026thinsp;=\u0026thinsp;74,610)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSupplementary Table 2 Risk of invasive breast cancer according to the discrepancy between Radiologist's and LIBRA and AI-method breast density (n\u0026thinsp;=\u0026thinsp;74,610)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSupplementary Table 3 Risk of DCIS according to the discrepancy between Radiologist's, LIBRA and AI-method breast density (n\u0026thinsp;=\u0026thinsp;74,610)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSupplementary Table 4 Comparison of the discriminatory power of LIBRA, Radiologist, and AI-driven breast density in prediction of breast cancer\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"breast-cancer-research-and-treatment","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"brea","sideBox":"Learn more about [Breast Cancer Research and Treatment](https://www.springer.com/journal/10549)","snPcode":"10549","submissionUrl":"https://submission.nature.com/new-submission/10549/3","title":"Breast Cancer Research and Treatment","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Mammography, Screening, Breast density, Artificial intelligence, Laboratory Individualized Breast Radiodensity Assessment, Cohort study","lastPublishedDoi":"10.21203/rs.3.rs-4471074/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4471074/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003ePurpose\u003c/h2\u003e \u003cp\u003eTo examine the discrepancy in breast density assessments by radiologists, LIBRA software, and AI algorithm and their association with breast cancer risk.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eAmong 74,610 Korean women aged\u0026thinsp;\u0026ge;\u0026thinsp;34 years, who underwent screening mammography, density estimates obtained from both LIBRA and the AI algorithm were compared to radiologists using BI-RADS density categories (A\u0026ndash;D, designating C and D as dense breasts). The breast cancer risks were compared according to concordant or discordant dense breasts identified by radiologists, LIBRA, and AI. Cox-proportional hazards models were used to determine adjusted hazard ratios (aHRs) [95% confidence intervals (CIs)].\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eDuring a median follow-up of 9.9 years, 479 breast cancer cases developed. Compared to the reference non-dense breast group, the aHRs (95% CIs) for breast cancer were 2.37 (1.68\u0026ndash;3.36) for radiologist-classified dense breasts, 1.30 (1.05\u0026ndash;1.62) for LIBRA, and 2.55 (1.84\u0026ndash;3.56) for AI. For different combinations of breast density assessment, aHRs (95% CI) for breast cancer were 2.40 (1.69\u0026ndash;3.41) for radiologist-dense/LIBRA-non-dense, 11.99 (1.64\u0026ndash;87.62) for radiologist-non-dense/LIBRA-dense, and 2.99 (1.99\u0026ndash;4.50) for both dense breasts, compared to concordant non-dense breasts. Similar trends were observed with radiologists/AI classification: the aHRs (95% CI) were 1.79 (1.02\u0026ndash;3.12) for radiologist-dense/AI-non-dense, 2.43 (1.24\u0026ndash;4.78) for radiologist-non-dense/AI-dense, and 3.23 (2.15\u0026ndash;4.86) for both dense breasts.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThe risk of breast cancer was highest in concordant dense breasts. Discordant dense breast cases also had a significantly higher risk of breast cancer, especially when identified as dense by either AI or LIBRA, but not radiologists, compared to concordant non-dense breast cases.\u003c/p\u003e","manuscriptTitle":"Concordant and Discordant Breast Density Patterns by Different approaches for Assessing Breast Density and Breast Cancer Risk","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-06-10 22:35:16","doi":"10.21203/rs.3.rs-4471074/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-09-20T13:48:38+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-09-14T11:21:26+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"127668903768715997873564509221833503473","date":"2024-09-07T14:04:33+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"217446683038064275230972688384853531845","date":"2024-07-14T15:48:04+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"213216969875580552974040581191377574537","date":"2024-06-11T21:30:56+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-06-05T13:54:40+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-05-25T05:32:35+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-05-25T05:32:35+00:00","index":"","fulltext":""},{"type":"submitted","content":"Breast Cancer Research and Treatment","date":"2024-05-24T08:25:14+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"breast-cancer-research-and-treatment","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"brea","sideBox":"Learn more about [Breast Cancer Research and Treatment](https://www.springer.com/journal/10549)","snPcode":"10549","submissionUrl":"https://submission.nature.com/new-submission/10549/3","title":"Breast Cancer Research and Treatment","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"74bc6060-7829-4f0e-bc3d-6f3567843d66","owner":[],"postedDate":"June 10th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2024-11-04T16:24:56+00:00","versionOfRecord":{"articleIdentity":"rs-4471074","link":"https://doi.org/10.1007/s10549-024-07541-1","journal":{"identity":"breast-cancer-research-and-treatment","isVorOnly":false,"title":"Breast Cancer Research and Treatment"},"publishedOn":"2024-11-01 16:20:02","publishedOnDateReadable":"November 1st, 2024"},"versionCreatedAt":"2024-06-10 22:35:16","video":"","vorDoi":"10.1007/s10549-024-07541-1","vorDoiUrl":"https://doi.org/10.1007/s10549-024-07541-1","workflowStages":[]},"version":"v1","identity":"rs-4471074","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4471074","identity":"rs-4471074","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}