Abstract Background: Ductal carcinoma in situ (DCIS) is a non-obligate precursor to invasive ductal carcinoma. A minority of DCIS cases will ever progress to ipsilateral invasive breast cancer (iIBC), but almost all are treated with breast-conserving surgery and radiotherapy. Reliable biomarkers of progression risk are needed to prevent overtreatment of low-risk lesions. We present a DCIS gene expression classifier to predict the risk of iIBC occurring within the first 5 years of diagnosis. Methods: The model was trained on a dataset of pure primary DCIS RNA-seq samples from a Dutch population-based cohort collected from 1989 to 2005. Samples from 188 patients treated with breast-conserving surgery only were retained for the training data to remove radiotherapy as a confounding factor in iIBC risk. The classifier is a logistic regression model with elastic net penalization, trained in a nested 5x5 cross-validation scheme. The final model was validated on an external, independent dataset of 91 British DCIS samples from the NHS Sloane project. Results: The classifier achieved an overall AUC of 0.642 on the outer loop test sets in the Dutch training set, and 0.694 in the Sloane independent validation set. The classifier risk score is shown in a logistic model to be associated with an increased risk of iIBC within 5 years in the Sloane validation dataset [OR: 1.16; CI: 1.05-1.27; p = 0.0496] where HER2 status, ER status, histopathological grade and age at diagnosis were not. A threshold to categorize risk scores into low- and high-risk categories was chosen on the training set by selecting the cut-point that maximized balanced accuracy. Gene set enrichment analysis showed enrichment of cell cycle and proliferation gene sets in the high-risk category (HALLMARK_E2F_TARGETS, HALLMARK_G2M_CHECKPOINT, GNF2_MKI67). Discussion: The performance of this classifier on an external, independent validation dataset shows that signals of risk of developing iIBC within 5 years of diagnosis can be detected within the gene expression profile of pure primary DCIS lesions. That the classifier was trained and validated on samples from patients who received the least aggressive treatment available without the confounding factors of radiotherapy or mastectomy brings us closer to an understanding of the biology underlying the risk of progression to iIBC in DCIS untreated at diagnosis. Such an understanding could be valuable information for including DCIS patients in active surveillance trials, and a step forward in preventing women having to undergo unnecessary surgery and radiotherapy. Citation Format: William Joseph Harley, Maria Roman-Escorza, Jelle Wesseling, Elinor Sawyer, Renee X. de Menezes, Esther Lips. A gene expression classifier to predict progression risk of ductal carcinoma in situ to invasive breast cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 7626.

Cancer research emphasises early detection, yet quantitative methods for normal tissue analysis remain limited. Digitised haematoxylin and eosin (H&E)-stained slides enable computational histopathology, but artificial intelligence (AI)-based analysis of normal breast tissue (NBT) in whole slide images (WSIs) remains scarce. We curated 70 WSIs of NBTs from multiple sources and cohorts with pathologist-guided manual annotations of epithelium, stroma, and adipocytes (https://github.com/cancerbioinformatics/OASIS). We developed robust convolutional neural network (CNN)-based, patch-level classification models, named NBT-Classifiers, to tessellate and classify NBTs at different scales. Across three external cohorts, NBT-Classifiers trained on 128 × 128 µm and 256 × 256 µm patches achieved AUCs of 0.98–1.00. The model learned independent normal features different from those of precancerous and cancerous epithelium, which were further visualised using two explainable AI techniques. When integrated into an end-to-end preprocessing pipeline, NBT-Classifiers facilitate efficient downstream analysis within peri-lobular regions. NBT-Classifiers provide robust compartment-specific analytical tools and enhance our understanding of NBT appearances, which serve as valuable reference points for identifying premalignant changes and guiding early breast cancer prevention strategies.

Ductal Carcinoma In Situ (DCIS) is a non-obligate precursor of invasive breast cancer. Due to a lack of reliable prognostic markers, nearly all women with DCIS undergo intensive treatment-often unnecessarily. The LORD trial addresses this by offering active surveillance to women with screen-detected, ER-positive, HER2-negative, grade 1 or 2 DCIS, aiming to reduce overtreatment. To support this, we developed a deep learning pipeline based on foundation models to predict grade, ER, and HER2 status directly from H&E-stained digitized pathology slides. Models were trained and tested on a Dutch multicenter dataset (n = 887) and externally validated on a UK dataset (n = 259). On the Dutch data, the models achieved mean AUROCs of 0.90 (ER), 0.84 (HER2), and 0.86 (grade); external validation yielded 0.80, 0.74, and 0.75, respectively. Using these outputs, we stratified patients according to active surveillance criteria, reaching balanced accuracies of 0.81 (Dutch) and 0.64 (UK), with corresponding NPVs of 0.86 and 0.76. Our models generalize across cohorts and reliably predict key biomarkers, supporting the identification of DCIS patients eligible for less aggressive management.

Pathologic complete response after neoadjuvant treatment is considered a surrogate of cure in triple-negative breast cancer, yet around 10% of patients still relapse. Whether baseline stromal tumor-infiltrating lymphocytes can stratify this residual risk is unknown. Here, we report on GAMBIT, a multicentric real-world retrospective study of 2457 patients with triple-negative breast cancer or estrogen receptor-low disease, HER2-negative, of whom 1192 obtained a pathological complete response and 690 have evaluable tumor infiltrating lymphocytes. Among patients with pathological complete response, clinical nodal status and tumor infiltrating lymphocytes are independently prognostic and patients with clinical node-positive/low-tumor infiltrating lymphocytes tumors experience substantially worse outcomes, with five-year distant relapse-free survival of 83.4% and overall survival of 85.8%. In this high-risk subgroup, five-year cumulative incidence of central nervous system reaches 7.5%, including 6.9% presenting as isolated central nervous system relapse. In this work, we identify a high-risk subgroup despite pathologic complete response and provide a framework supporting risk-adapted trial design incorporating central nervous system-directed strategies.

Early prediction of the response to neoadjuvant chemotherapy (NAC) enables tailoring treatment strategies to the specific needs of individual breast cancer patients. Circulating tumor DNA (ctDNA) has shown to be a prognostic factor for response on NAC during treatment. However, at this point in time mostly tumor-informed ctDNA detection methods are used which are costly, have relatively long turnaround times and are subsequently potentially less feasible for widespread clinical application. In this study, we investigated four tumor-agnostic methods to determine their ability to accurately detect circulating tumor DNA (ctDNA) at baseline. These methods were the Oncomine Breast cell free DNA (cfDNA) NGS panel, the LINE-1 sequencing assay mFAST-SeqS, shallow whole genome sequencing and the genome-wide methylation profiling assay MeD-Seq. In total 40 patients with triple negative or luminal B breast cancer were included and cell free DNA (cfDNA) from plasma before the start of NAC was analyzed with the four assays. We detected ctDNA in 3/24 (12.5

The lower all-cause mortality in women with Ductal carcinoma in situ (DCIS) compared with the general population has been hypothesized to be due to a "healthy-user effect," but this has not been studied in large cohorts. In a population-based, retrospective cohort study comprising 18,942 women with primary DCIS between 1999 and 2015 in the Netherlands, the cumulative incidence of breast cancer death (BCD) was estimated using death by other cause as a competing risk. The cause-specific mortality risk of women with DCIS was compared with that of the Dutch female population. Multivariable competing risk regression was used to quantify the effects of the method of detection and socio-economic status (SES). With 289 BCDs, the 10-year cumulative incidence of BCD was 1.3% (95% CI, 1.1-1.5). Compared to the Dutch female population, women with DCIS had a 2.1-times higher risk of BCD, but a 7% lower risk of all-cause mortality. Women with screen-detected DCIS had lower risks of BCD compared to women with non-screen-detected DCIS (subdistribution hazard ratio [sHR]:0.60, 95% CI 0.47-0.77), as did women with high SES versus low SES (sHR 0.54, 95% CI 0.30-0.97) in the first 4 years of follow-up, adjusted for age and year at diagnosis, and DCIS characteristics. In conclusion, overall mortality in women with DCIS is not higher compared to the Dutch female population, though death due to invasive breast cancer is increased. Within all women with DCIS, those with screen-detected DCIS or high SES had lower BCD and all-cause mortality, suggesting a healthy-user effect.

Ductal carcinoma in situ (DCIS), often misclassified as early-stage breast cancer, rarely progresses to invasive disease. Yet, the inability to distinguish between indolent and aggressive DCIS leads to widespread overtreatment, burdening thousands of women globally. Addressing this, the PRECISION team, under the Cancer Grand Challenges initiative, has achieved remarkable advancements, driving a paradigm shift in DCIS management. PRECISION exemplifies next-level team science, merging expertise from 10 institutions across the USA, UK, and the Netherlands. Spanning disciplines such as epidemiology, molecular biology, AI, pathology, clinical expertise, health outcomes, and patient advocacy, the team’s interdisciplinary approach enabled transformative breakthroughs. Key achievements include: 1. Understanding DCIS Biology: Using innovative patient-derived DCIS-MIND models and single-cell genomic techniques, PRECISION revealed that genetic changes alone do not dictate progression. These findings challenge long-held assumptions, emphasizing the need for comprehensive approaches to assess risk. 2. Risk Stratification Methodologies: Novel biomarkers, AI-driven morphometric analyses, and an RNA-seq classifier reliably differentiate high-risk from low-risk DCIS. For example, the RNA-seq classifier identified low-risk DCIS with a 98% negative predictive value, paving the way for treatment de-escalation. 3. Testing Active Surveillance: Clinical trials embedded in PRECISION’ (COMET, LORD, LORIS) explored the safety of active surveillance, potentially replacing invasive treatments for low-risk DCIS with cost-effective outcomes. Patient engagement and shared decision-making tools further enhance personalized care strategies. PRECISION’s success stems from dynamic international collaborations and integration of advanced technologies. Its findings not only redefine DCIS management but also establish a blueprint for addressing other low-risk malignancies. By focusing on biology-based, patient-centered solutions, PRECISION ensures that treatment aligns with the true nature of the disease, paving the path to transform DCIS care worldwide. To do so, the team aims to develop dynamic risk forecasting techniques, powered by evolutionary AI, drawing inspiration from geoscience methodologies. The team also plans to establish a Global Virtual DCIS Monitoring Clinic, enabling worldwide participation. This initiative will support personalized care, foster shared decision-making, and minimize clinic visits through standardized data collection and seamless online follow-ups. Jelle Wesseling, Proteeti Bhattacharjee, Esther H. Lips, Elinor J. Sawyer, E. Shelley Hwang, Alastair M. Thompson, Grand Challenge PRECISION Consortium. Conquering overtreatment of ductal carcinoma in situ through integrative team science [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 7477.

BACKGROUND:Ductal Carcinoma In Situ (DCIS) can progress to ipsilateral invasive breast cancer (IBC) but over 75% of DCIS lesions do not progress if untreated. Currently, DCIS that might progress to IBC cannot reliably be identified. Therefore, most patients with DCIS undergo treatment resembling IBC. To facilitate identification of low-risk DCIS, we developed deep learning models using histology whole-slide images (WSIs) and clinico-pathological data. METHODS:We predicted invasive recurrence in patients with primary, pure DCIS treated with breast-conserving surgery using clinical Cox proportional hazards models and deep learning. Deep learning models were trained end-to-end with only WSIs or in combination with clinical data (integrative). We employed nested k-fold cross-validation (k = 5) on a Dutch multicentre dataset (n = 558). Models were also tested on the UK-based Sloane dataset (n = 94). FINDINGS:Evaluated over 20 years on the Dutch dataset, deep learning models using only WSIs effectively stratified patients into low-risk (no recurrence) and high-risk (invasive recurrence) groups (negative predictive value (NPV) = 0.79 (95% CI: 0.74-0.83); hazard ratio (HR) = 4.48 (95% CI: 3.41-5.88, p < 0.0001); area under the receiver operating characteristic curve (AUC) = 0.75 (95% CI: 0.70-0.79)). Integrative models achieved similar results with slightly enhanced hazard ratios compared to the image-only models (NPV = 0.77 (95% CI 0.73-0.82); HR = 4.85 (95% CI 3.65-6.45, p < 0.0001); AUC = 0.75 (95% CI 0.7-0.79)). In contrast, clinical models were borderline significant (NPV = 0.64 (95% CI 0.59-0.69); HR = 1.37 (95% CI 1.03-1.81, p = 0.041); AUC = 0.57 (95% CI 0.52-0.62)). Furthermore, external validation of the models was unsuccessful, limited by the small size and low number of cases (22/94) in our external dataset, WSI quality, as well as the lack of well-annotated datasets that allow robust validation. INTERPRETATION:Deep learning models using routinely processed WSIs hold promise for DCIS risk stratification, while the benefits of integrating clinical data merit further investigation. Obtaining a larger, high-quality external multicentre dataset would be highly valuable, as successful generalisation of these models could demonstrate their potential to reduce overtreatment in DCIS by enabling active surveillance for women at low risk. FUNDING:Cancer Research UK, the Dutch Cancer Society (KWF), and the Dutch Ministry of Health, Welfare and Sport.

Ductal carcinoma in situ (DCIS) may progress to ipsilateral invasive breast cancer (iIBC), but often never will. Because DCIS is treated as early breast cancer, many women with harmless DCIS face overtreatment. To identify features associated with progression, we developed an artificial intelligence-based DCIS morphometric analysis pipeline (AIDmap) on hematoxylin-eosin-stained (H&E) tissue sections. We analyzed 689 digitized H&Es of pure primary DCIS of which 226 were diagnosed with subsequent iIBC and 463 were not. The distribution of 15 duct morphological measurements was summarized in 55 morphometric variables. A ridge regression classifier with cross validation predicted 5-years-free of iIBC with an area-under the curve of 0.67 (95% CI 0.57–0.77). A combined clinical-morphometric signature, characterized by small-sized ducts, a low number of cells and a low DCIS/stroma ratio, was associated with outcome (HR = 0.56; 95% CI 0.28–0.78). AIDmap has potential to identify harmless DCIS that may not need treatment.

The current clinical paradigm around Ductal Carcinoma in Situ (DCIS) is that it consists of malignant cells confined to the breast ducts, and therefore cannot metastasize. Nonetheless, several studies have reported DCIS with metastasis in the sentinel lymph node (SN+). For accurate risk communication and management, we aimed to assess to what extent registered “metastatic spread” in DCIS could be explained by limitations in registration or missed invasive breast cancer at time of diagnosis. Data from the nationwide cancer registry and national pathology database on women diagnosed with DCIS SN+ in the Netherlands, spanning from 2005 to 2020, was curated and reviewed, taking into account their histories of prior DCIS, invasive breast cancer, or other malignancies. Cases were excluded from further analysis if pathology data indicated registration errors, DCIS mixed with other lesion types, or diagnostic uncertainties. Next, hematoxylin and eosin-stained tissue slides of eligible DCIS SN+ cases were independently reviewed by two pathologists to assess the presence of invasive breast cancer and SN status. Additional immunohistochemical staining (CK 5/6 or CK 8/18) was performed when findings were unclear. Inter-observer agreement was evaluated using the linearly weighted Kappa statistic. A total of 30, 863 patients were identified with a DCIS diagnosis between 2005 and 2020, of which 16, 070 (52%) underwent SN biopsy according to cancer registry data. SN+ was registered in 454 (3 %) patients: 47 (10%) had macrometastases (>2 mm), 78 (17%) had micrometastases (>0.2 to <= 2 mm), and 329 (73%) were positive for isolated tumor cells (<= 0.2 mm). Out of the 454 registered DCIS SN+ cases, 273 (60%) were excluded from further investigation based on pathology data, including registration errors (n=44), DCIS mixed with other lesions (n=147), and diagnostic uncertainties (n=82). Tissue material of 46 (37%) out of 125 registered cases with macro- and micrometastases was reviewed. Observer variability in assessing the presence of invasive breast cancer was high (k= 0.14; 95% CI 0.08 - 0.34; p = 0.12) and additional CK 5/6 staining was requested for 38 cases. Two cases were classified as primary invasive breast cancer, 25 as pure DCIS, while 19 remained inconclusive, due to variation in tissue sections or suboptimal tissue quality. In six cases the SN was scored as negative by pathology revision, likely due to tissue section variability. Our study indicates that DCIS in itself has minimal to no metastatic potential. Ongoing clonality analysis of 9 cases with macrometastases aims to determine whether the SN metastases are clonally related to the DCIS lesions. Merle van Leeuwen, Sandra van den Belt- Dusebout, Petra Kristel, Lennart Mulder, Joyce Sanders, Carmen Vlahu, Esther H. Lips, Jelle Wesseling. Ductal carcinoma in situ: Potential to metastasize? A nationwide cancer registry-based study [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2332.

Ductal carcinoma in situ (DCIS) is a non-obligate precursor to invasive breast cancer, but distinguishing patients with harmless from potentially hazardous DCIS remains a challenge. Consequently, a cornerstone of DCIS research is finding prognostic biomarkers. One such recent biomarker is adipocyte hypertrophy, which has been shown to be prognostic of ipsilateral invasive breast cancer (iIBC) in post-menopausal women with primary DCIS. However, little is known about the correlation between adipocyte size, clinical factors and mammographic density. Archival hematoxylin and eosin- stained breast biopsy and excision specimens were retrieved from 669 women diagnosed with primary DCIS between 2000 and 2020 treated at the Netherlands Cancer Institute. These slides were digitized whereafter a machine-learning algorithm using HALO®, was applied to retrieve adipocyte size. Radiology reports were obtained to extract mammographic BI-RADS density. Age at diagnosis, body mass index (BMI), menopausal status and information on comorbidities were extracted from electronic patient records. Associations between adipocyte size and clinical factors and mammographic density were investigated using univariable and multivariable linear regression models. Using a clinically relevant cutoff point for adipocyte size, multivariable logistic regression was performed. The median age at primary DCIS diagnosis was 55 years (interquartile range (IQR): 49.0 -63.0) and most DCIS lesions were grade 3 (40.9%). The median BMI was 24.1 (IQR: 21.9 - 27.1). The majority of women were post-menopausal (47.8%) and had dense breasts (53.8%). Significant positive correlations were found between adipocyte size and age, BMI and all metabolic comorbidities with the exception of smoking. Strong negative correlations were found between adipocyte size and mammographic density categories C and D. In univariable linear models with age and BMI, metabolic risk was able to further differentiate between patients with and without adipocyte hypertrophy beyond age and BMI. This was confirmed using multivariable logistic regression comparing models adjusted and unadjusted for metabolic risk. BMI is the strongest predictor of adipocyte hypertrophy in DCIS patients, however metabolic risk further differentiates between patients with and without adipocyte hypertrophy. Future work should be done on collecting long term follow-up and ascertaining subsequent breast cancer in these patients. If successful, BMI and metabolic risk can potentially be incorporated into risk prediction models for subsequent invasive breast cancer. Charlotta V. Mulder, Mathilde Almekinders, Renaud Tissier, Lennart Mulder, Petra Kristel, Esther Lips, Marjanka Schmidt, Jelle Wesseling. Metabolic risk is an important determinant of adipocyte hypertrophy beyond age, BMI and breast density in patients with ductal carcinoma in situ [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 6215.

Background Cancer research emphasises early detection, yet quantitative methods for analysing normal tissue remain limited. Hematoxylin and eosin (H&E)-stained tissues in digitised whole slide images (WSIs) enable computational histopathology; however, artificial intelligence (AI)-based analyses of normal breast tissue (NBT) remain scarce. Methods We curated 70 WSIs of NBTs from multiple sources with pathologist-guided manual annotations of epithelium, stroma, and adipocytes, and developed robust convolutional neural network (CNN)-based, patch-level classification models, named NBT-Classifiers , to tessellate and classify NBTs at different scales. Data and code are available at and . Findings Across three external cohorts, NBT-Classifiers trained on 128 x 128 µm and 256 x 256 µm patches achieved AUCs of 0·98–1·00. Two explainable artificial-intelligence (AI)- visualisation techniques confirmed the biological relevance of tissue class predictions. An end-to-end WSI pre-processing framework was then integrated, capable of localising lobules and peri-lobular stroma. The outputs are compatible with the QuPath v0.3.0 platform and enable downstream digital image analysis, such as texture and nuclei morphology assessment, at the patch level. Interpretation NBT-Classifiers represent a robust, generalisable, end-to-end deep learning framework, which will enable broader application in studies of normal tissues, in the context of breast and breast cancer. Funding The Breast Cancer Research Trust, Breast Cancer Now (and their legacy charity Breakthrough Breast Cancer), the Medical Research Council (MRC) [MR/X012476/1], Cancer Research UK [CRUK/07/012, KCL-BCN-Q3], and CRUK City of London Centre Award [CTRQQR-2021/100004]. Evidence before this study We searched PubMed from Jan 1, 2013, to Mar 20, 2025, using the query (("normal breast" OR "terminal duct lobular unit" OR "TDLU" OR "acini") AND ("artificial intelligence" OR "deep learning" OR "convolutional neural network" OR "CNN" OR "machine learning" OR "automation" OR "classification" OR "segmentation" OR "detection")) AND (journal article[pt] NOT review[pt]) NOT ("computed tomography" OR "CT" OR "ultrasound" OR "sonography" OR "US" OR "mammography" OR "mammogram" OR "Raman" OR "single-cell" OR "MRI"). This yielded 264 initial hits and after a thorough manual review, only eight studies were found to be relevant for using deep learning or machine learning to automate tissue recognition or quantification of normal breast tissues (NBTs) on digitised histological images. All these methods are highly specialised for detecting and quantifying lobules or quantifying tissue composition, and none of them are specifically tailored for providing patch-level tissue classifications on WSIs of NBTs. Moreover, NBTs are heavily underrepresented in large annotated WSI databases. A recent literature review summarised publicly available breast hematoxylin and eosin (H&E) WSI datasets from 2015 to 2023, identifying 17 datasets comprising a total of 10,385 breast H&E WSIs. Of these, two datasets contain female normal breast WSIs with pathology-guided manual annotations. Added value of this study Motivated by the research gap, we collected a comprehensive WSI dataset that captures real-world variability present across NBTs and provided by expert ground-truth manual annotations. Based on extensive cross-validation and external testing, we developed robust patch-level classification models to classify three major tissue compartments within NBTs, at different scales. Implications of all the available evidence With high AUCs of 0·98–1·00 across three external cohorts, our artificial-intelligence (AI) tool can be used to automatically classify tissue compartments and localise lobular regions for downstream normal breast research. These approaches have the potential to enhance our understanding of how various NBT components contribute to both benign and malignant breast pathology and lay the groundwork for the development of more advanced deep learning models and spatial-defined molecular large-scale analyses in the future. ### Competing Interest Statement Anita Grigoriadis, Louise J. Jones and Greg Verghese are Co-Founders of PharosAI. Salim Arslan and Pahini Pandya and employed by Panakeia Technology, UK. All other authors declare no conflict of interest. WSIs involved in this study are stored at the OASIS repository: , which currently can be accessed upon request. The Breast Cancer Research TrustThe Breast Cancer Research Trust, , Breast Cancer NowBreast Cancer Now, , the Medical Research Council (MRC)the Medical Research Council (MRC), , MR/X012476/1 Cancer Research UKCancer Research UK, , CRUK/07/012, KCL-BCN-Q3 CRUK City of London Centre AwardCRUK City of London Centre Award, , CTRQQR-2021/100004

Summary: As we cannot reliably distinguish indolent, low-risk ductal carcinoma in situ (DCIS) from potentially progressive, high-risk DCIS, all women with DCIS diagnosis undergo intensive treatment without any benefit. The PREvent ductal Carcinoma In Situ Invasive Overtreatment Now team was established to unravel DCIS biology and develop new multidisciplinary approaches for accurate risk stratification to tackle the global problem of DCIS overdiagnosis and overtreatment. See related article by Bressan et al., p. 16 See related article by Stratton et al., p. 22 See related article by Goodwin et al., p. 34

Upon the inception of population-based screening programs, the incidence of ductal carcinoma in situ (DCIS) increased 6-fold. DCIS is a non-obligate precursor lesion of invasive breast cancer (IBC), of which the majority does not progress. This implies that women with non-progressive DCIS are overtreated. Within the Cancer Grand Challenge PRECISION project, we identified prognostic markers, including immunohistochemical, morphological markers, and an RNA-seq classifier holding promise in distinguishing progressive from non-progressive DCIS. We now aim to validate these externally and build a clinical prediction model. We conducted a case-cohort study nested in a Dutch population-based cohort of 8987 patients with DCIS treated with breast-conserving surgery (BCS) between 2005 and 2015. Women who subsequently developed ipsilateral IBC were considered cases and controls were those who did not. Our study population consisted of a random sample of 10.7% of the full cohort as our subcohort, and all other additional cases, totaling 1237 women (cases n=308; controls n=929). Tissue blocks were requested for all women, of which 940 were received and eligible for immunohistochemical analysis of ER, HER2, COX-2, Ki67 and P16 1. A new H&E slide was cut for the measurement of adipocyte size 2, level of periductal fibrosis 3 and a ductal morphometric analysis 4. RNA-sequencing was carried out on a selection of BCS-only patients (cases n = 100; controls n = 100). To assess the association of each marker and iIBC risk, prentice-weighted Cox proportional hazards models, with age as the underlying time variable, were implemented. Multivariable models were also implemented and adjusted for treatment, margins status, DCIS grade and size. To account for multiple comparisons, a false discovery rate (FDR) < 0.05 was used to define statistical significance. The median follow-up time for the case-cohort was 7.2 years (interquartile range (IQR): 5.2-10.1). The age at primary DCIS diagnosis was similar for breast cancer cases and controls (58.0, IQR: 51.0-64.0 vs 58.0, IQR: 51.0-66.0), as was the size of the tumor (p = 0.06). After final surgery, 10.4% of cases and 6.7% of controls had involved margins of <2mm (p = 0.06). Cases were treated more frequently with radiotherapy than controls (76.0% vs 68.4%, p = 0.01) and were more often high grade DCIS (46.4% vs 38.0%, p = 0.04). In December, we will present the validation of immunohistochemical, morphological markers, and an RNA-seq classifier for predicting subsequent ipsilateral IBC after DCIS and the benefit of using these markers in a clinical prediction model. Ultimately, this will aid individual risk stratification of women with primary DCIS, and can be used to diminish the current overtreatment of harmless, low-risk DCIS. References 1. Visser, L. L. et al. Clinicopathological risk factors for an invasive breast cancer recurrence after ductal carcinoma in situ-a nested case-control study. Clinical Cancer Research 24, 3593–3601 (2018). 2. Almekinders, M. M. M. et al. Breast adipocyte size associates with ipsilateral invasive breast cancer risk after ductal carcinoma in situ. NPJ Breast Cancer 7, (2021). 3. Visser, L. L. et al. Predictors of an invasive breast cancer recurrence after DCIS: A Systematic Review and Meta-analyses. Cancer Epidemiology Biomarkers and Prevention vol. 28 835–845. 4. Sobral-Leite, M. et al. Articial intelligence-based morphometric signature to identify ductal carcinoma in situ with low risk of progression to invasive breast cancer. doi:10.21203/rs.3.rs-3639521/v1. Citation Format: Charlotta Mulder, Will Harley, Petra Kristel, Lennart Mulder, Sten Cornelissen, Renee Menezes, Michael Schaapveld, Marjanka K. Schmidt, Jelle Wesseling, Esther H. Lips. A clinical risk prediction model for subsequent invasive breast cancer after ductal carcinoma in situ [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P3-05-28.

Ductal carcinoma in situ (DCIS) is a potential precursor to invasive breast cancer (IBC). The trajectory of an individual’s DCIS, if it will progress to IBC or remain as DCIS is difficult to predict. Currently >80% of DCIS is detected through mammographic screening of breast calcifications. Despite the close association of calcifications with DCIS, their role in the development of DCIS and/or its progression to IBC remains largely unexplored. In this study, we present a spectroscopy based analytical approach to probe chemical compositional changes of both DCIS associated breast calcifications and surrounding soft tissue aiming to identify a cohort at increased risk for invasive progression. Tissue samples from 316 DCIS patients without invasive cancer were obtained from multiple centres as part of the PRECISION consortium (The Netherlands, UK and USA). Three consecutive tissue sections were obtained each of the tissue biopsies. Mid-infrared (mid-IR) and Raman hyperspectral imaging was performed independently on two sections that were left unstained. The third H&E-stained section was used to annotate calcifications and histopathological features. All DCIS samples had known outcome (i) ‘pure DCIS as controls’ (DCIS without progression to invasion) (n=193), (ii) ‘DCIS with progression to invasion as cases’ (DCIS from patients who subsequently developed invasive disease after initial treatment) (n=123). Spectral features of DCIS calcifications and surrounding soft tissue were used as inputs for analysis. Data was divided into a discovery and a validation set. Cluster analysis followed by Principal component analysis fed linear discriminant analysis was carried out on the discovery set to develop DCIS prediction models. For the Raman data, a mean area under the receiver operating characteristic curve (AUROC) value of 0.85 was obtained using calcification spectral features, and 0.75 using soft tissue spectral features in distinguishing controls from cases (N=118 vs 52). Similar analysis on the IR data showed a mean AUROC value of 0.68 for calcification, 0.78 for epithelial and 0.80 for stromal components (N=97 vs 61). Preliminary analysis shows changes in phosphate to carbonate ratio and variations in magnesium whitlockite content in calcifications, and protein secondary structural changes in soft tissue, between the two groups. The models will be tested independently on the validation set and the outcomes will be presented at the AACR conference. Spectroscopic chemical analysis of breast calcifications and soft tissue show promise in predicting the likely progression of DCIS to IBC. Pending further independent validation, these techniques appear to be novel image-based risk assessment tools that can potentially be utilised to inform DCIS prognosis and treatment options. Jayakrupakar Nallala, Doriana Calabrese, Sarah Gosling, Esther Lips, Ihssane Bouybayoune, Rachel Factor, Sarah Pinder, Lorraine King, Jeffrey Marks, Thomas Lynch, Donna Pinto, Alastair Thompson, Elinor Sawyer, Jelle Wesseling, Shelley Hwang, Keith Rogers, Nick Stone, Grand Challenge PRECISION consortium. Predicting the prognosis of ductal carcinoma in situ through chemical analysis of breast microcalcifications and soft tissue using infrared and Raman spectroscopy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3353.

Background: Ductal carcinoma in situ (DCIS) is a non-obligate precursor to invasive breast cancer, but distinguishing patients with harmless from potentially hazardous ductal carcinoma in situ (DCIS) remains a challenge. Consequently, a cornerstone of DCIS research is finding prognostic biomarkers. One such recent biomarker is adipocyte hypertrophy, which has been shown to be predictive of iIBC in post-menopausal women with primary DCIS 1. Little is known about the correlation between adipocyte size and clinical factors, mammographic breast density and other tissue composition metrics conferring risk of breast cancer. Methods: Using the digital pathology software program HALO®, we automated the adipocyte detection pipeline by building a machine learning algorithm for tissue segmentation, whereafter individual adipocytes were characterized using the vacuole module version 3.2.2. The correlation between the automated adipocyte measurement and the pathologist’s generated measurement was assessed for 249 patients. Thereafter, archival hematoxylin and eosin- stained breast biopsy and excision specimens were retrieved from 700 women diagnosed with primary DCIS between 2000 and 2020 treated at the Netherlands Cancer Institute. These slides were digitized whereafter the machine-learning algorithms were applied to retrieve adipocyte size, proportion of epithelial cells, stromal cells and fibroglandular tissue. Radiology reports were obtained to extract mammographic BI-RADS density. Age at diagnosis, weight, length and menopausal status were extracted from electronic patient records. Associations between adipocyte size and clinical factors, BI-RADS density and tissue metrics were investigated using multivariable linear regression models. Results: The median age at primary DCIS diagnosis was 55 years (interquartile range (IQR): 49.0 -63.0) and most DCIS lesions were grade 3 (40.9%). The median BMI was 24.1 (IQR: 21.9 - 27.1). The majority of women were post-menopausal (47.8%) and had dense breasts, whereby 40.6% had heterogeneously dense breasts and 13.7% had extremely dense breasts 2. Furthermore, there was good concordance between the automated adipocyte measurements in comparison to the pathologist’s measurement with an intraclass correlation coefficient of 0.96 (95% CI = 0.95-0.97). In December, we will present the associations between adipocyte size and the clinical factors, mammographic breast density and tissue composition metrics. Conclusion: Adipocyte hypertrophy has been shown to be predictive of iIBC in post-menopausal women with primary DCIS. This biomarker can be confidently measured with machine learning algorithms. This study will elucidate the associations between adipocyte size and clinical factors, mammographic breast density and other tissue composition metrics. References 1. Almekinders, M. M. M. et al. Breast adipocyte size associates with ipsilateral invasive breast cancer risk after ductal carcinoma in situ. NPJ Breast Cancer 7, (2021). 2. Breast Imaging Reporting and Data System (BI-RADS). (American College of Radiology, Reston, VA, 2013). Citation Format: Charlotta Mulder, Mathilde Almekinders, Renaud Tissier, Petra Kristel, Esther Lips, Marjanka Schmidt, Jelle Wesseling. Adipocyte size in relation to clinical factors, mammographic density and quantitative histologic metrics in ductal carcinoma in situ (DCIS) [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P1-07-13.

Automated semantic segmentation of whole-slide images (WSIs) stained with hematoxylin and eosin (H&E) is essential for large-scale artificial intelligence-based biomarker analysis in breast cancer. However, existing public datasets for breast cancer segmentation lack the morphological diversity needed to support model generalizability and robust biomarker validation across heterogeneous patient cohorts. We introduce BrEast cancEr hisTopathoLogy sEgmentation (BEETLE), a dataset for multiclass semantic segmentation of H&E-stained breast cancer WSIs. It consists of 587 biopsies and resections from three collaborating clinical centers and two public datasets, digitized using seven scanners, and covers all molecular subtypes and histological grades. Using diverse annotation strategies, we collected annotations across four classes - invasive epithelium, non-invasive epithelium, necrosis, and other - with particular focus on morphologies underrepresented in existing datasets, such as ductal carcinoma in situ and dispersed lobular tumor cells. The dataset's diversity and relevance to the rapidly growing field of automated biomarker quantification in breast cancer ensure its high potential for reuse. Finally, we provide a well-curated, multicentric external evaluation set to enable standardized benchmarking of breast cancer segmentation models.

Background: Benign breast disease (BBD) is commonly detected in women participating in breast cancer screening programs and comprises a diverse group of lesions. The clinical significance of BBD lies in its association with an increased risk of developing breast cancer, which depends on the histological subtype. Calcifications, frequently observed in mammographic screenings, are critical in the detection and diagnosis of both benign and malignant breast conditions. While many calcifications are benign, some patterns are indicative of ductal carcinoma in situ (DCIS) or invasive breast cancer (IBC). Better characterization of breast cancer risk among women with BBD, considering both clinical and molecular findings, can improve surveillance, early diagnosis, and survival. This study aims to identify mammographic and chemical characteristics of calcifications that are associated with subsequent development of DCIS or IBC. Methods: A matched case-control study was conducted of women diagnosed with BBD at the Netherlands Cancer Institute and Albert Schweitzer Hospital between 2004 and 2023. Cases (n=65) were women with BBD who developed ipsilateral DCIS or IBC ≥ 6 months after a first BBD diagnosis, whereas controls (n=244) were BBD patients who did not develop subsequent ipsilateral DCIS or IBC during the follow-up (FU) duration of their matching cases. Additionally, controls were matched based on the year and age at the time of BBD diagnosis. Patient characteristics (e.g. age at diagnosis, vital status) and characteristics of both the benign lesions and subsequent malignant lesions were extracted from pathology reports using text searches and Palga codes. Mammographic lesion types (e.g. calcifications, masses, architectural distortion, asymmetries) were extracted from radiology reports. Qualitative mammographic features including calcification morphology and distribution were extracted from mammograms by two researchers and a trained radiologist. Quantitative mammographic features including breast density score and calcification cluster size and number will be extracted using TRANSPARA 2.0, an radiology artificial intelligence decision support system. In a subset of cases (n = 29) and controls (n=59) chemical characteristics were measured using infrared and Raman spectroscopy. Results: The baseline comparison of mammographic qualitative features comprised 65 cases and 244 matched controls.Median age at BBD diagnosis was 51 years (range 35-80). Median FU from BBD diagnosis to DCIS or IBC was 6 years (range 1-17). Most cases and controls had non-proliferative BBD (89% and 94%) rather than proliferative BBD. Among cases, 19 developed DCIS while 46 developed IBC. While cases and controls showed comparable proportions of mammographic lesions, calcifications were more prevalent among cases (48.0% vs. 34.0%), approaching statistical significance (p = 0.06). Significant differences in calcification morphology were observed (p = 0.009), with cases more likely to display fine pleomorphic calcifications (23 % vs. 8.5%). The distribution of calcifications was similar between cases and controls (p = 0.49). Multivariate-adjusted conditional regression models showed an odds ratio (OR) of 1.7 (95% CI: 0.9-3.0) for the association between calcification presence and DCIS/IBC development, albeit with considerable uncertainty. Presence of suspicious calcification morphologies (amorphous, fine pleomorphic, linear) suggested an OR of 3.0 (95% CI: 0.9-10.0) compared to benign morphology. Conclusions: The trends observed in this study suggest potential prognostic value of calcification morphology in women with BBD. Additional results on quantitative mammographic features and chemical characteristics of cases and controls will be presented at the conference. Citation Format: Merle van Leeuwen, Sandra van den Belt-Dusebout, Jia Ning Zhuchen, Shannon Doyle, Petra Kristel, Lennart Mulder, Jayakrupakar Nallala, Pieter Westenend, Nick Stone, Esther Lips, Ritse Mann, Jelle Wesseling. Calcification characteristics in women with benign breast disease and the risk of subsequent breast cancer: a case-control study [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P3-03-29.

Background: The current clinical paradigm around Ductal Carcinoma in Situ (DCIS) is that it consists of malignant cells confined to the breast ducts, and therefore cannot metastasize. Nonetheless, several studies have reported DCIS with metastasis in the sentinel lymph node. For accurate risk communication and management, we aimed to assess to what extent registered “metastatic spread” in DCIS could be explained by limitations in registration or missed invasive breast cancer at time of diagnosis. Methods: Nationwide data on women diagnosed with DCIS and a positive sentinel node (DCIS SN+) between 2005 and 2021 were obtained from the Netherlands Cancer Registry (NCR) and the Dutch Nationwide Pathology Databank (Palga). The incidence of registered DCIS SN+ and the size of the metastasis was determined with data from NCR. Pathology data of the primary DCIS diagnosis was thoroughly reviewed, with all known corresponding history pathology reports of in-situ or invasive breast cancer or unknown primary. Cases were excluded from further analysis if PALGA data indicated registration errors, DCIS mixed with other types of lesions, or diagnostic uncertainties such as suspicion of microinvasion, uncertain sentinel node status, poor tissue quality or positive surgical margins. Next, hematoxylin and eosin stained tissue slides of eligible DCIS SN+ cases were independently reviewed by two expert breast pathologists to assess the presence of (micro)invasion and sentinel node status. Presence of invasion and sentinel node positivity was scored as yes, uncertain, no or not applicable. For cases scored as uncertain, additional immunohistochemical (IHC) stainings with cytokeratin 5/6 or 8/18 were used. Agreement between the two pathologists was assessed using the linearly weighted Kappa statistic. Results: A total of 30,863 patients were identified with a DCIS diagnosis between 2005 and 2020, of which 16,070 (52.1%) underwent sentinel lymph node biopsy according to NCR data. SN+ was identified in 454 (2.8 %) patients: 47 (10 %) had macrometastases (>2 mm), 78 (17%) had micrometastases (>0.2 - <= 2 mm), and 329 (73%) were positive for isolated tumor cells (ITCs) (<= 0.2 mm). Out of the 454 registered DCIS SN+ cases, 273 (60%) were excluded from further investigation based on pathology data from Palga, based on registration errors (n=44), due to the presence of other lesions (n=147), or diagnostic uncertainties (n=82). There was no significant difference in reasons for exclusion between micrometastases, macrometastases or ITCs (p = 0.36). Tissue material of 47 out of 181 DCIS cases with macro- and micrometastases were reviewed by the pathologists. Initial agreement on sentinel node status was weak with a kappa statistic of 0.33 (95% CI 0.15 – 0.52; p = 0.003. Agreement on the presence of an invasive component was minimal with a kappa statistic of 0.14 (95% CI 0.08 – 0.34 ; p = 0.12). The most frequent discrepancies were between ‘uncertain’ and ‘no’ scorings. After reaching consensus, three cases were scored as having an invasive component, and additional IHC staining was requested for 40 cases due to suspicion of the presence of (micro)invasion. Review results of these 40 cases will be presented at the conference. Conclusions: Our study offers a nuanced understanding of DCIS SN+ cases, suggesting that while these cases pose diagnostic challenges, the metastatic potential of pure DCIS remains low. Citation Format: Merle van Leeuwen, Sandra van den Belt-Dusebout, Petra Kristel, Lennart Mulder, Joyce Sanders, Carmen Vlahu, Esther Lips, Jelle Wesseling. Does Ductal Carcinoma in Situ have metastatic potential? A nationwide cancer registry-based study of Ductal Carcinoma in Situ with sentinel lymph node positivity [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P5-12-13.