Contrast Enhancement Mammography vs MRI for the Surveillance of Women at High Risk of Breast Cancer: Con-trust Randomized Controlled Trial

Description and distribution of activities of each operating unit The project comprises six work packages (WPs) to ensure efficient project coordination, quality control and data collection for conducting the trial, breast cancer risk prediction using machine learning models, cost analysis, and evidence synthesis for guidelines development. In the Project Coordination WP, regular communication channels will be established, project management tools developed,and progress monitored to ensure compliance with ethical and regulatory requirements. The deliverables include ethics, safety, and administrative approval, electronic data collection and transfer tools, training materials for data management, progress reports, meeting agendas and minutes, budget and expenditure reports, and dissemination materials. The leading unit is AUSL-RE, and the WP is instrumental to all three aims. Administrative coordination of the trial accrual in the recruiting centers will be led by AUSL-RE and the centers in Southern Italy. The following centers have agreed to contribute: AOUI Verona, Policlinico S. Matteo Pavia, Irst-Meldola-Ausl Romagna, Policlinico S. Donato, Taormina, AOU Modena. The QualityControls WP ensures the quality control of CEM and MRI equipment used in the project and develops a tool to monitor the stability of CEM over time. The deliverables include quality control protocols, staff training materials, a consensus meeting of the radiologists and physicists, inspection and maintenance reports, a stability monitoring tool, and a database for collected data. The leading units are IOV and Messina; the WP contributes to Aims 1 and 2. The Machine Learning Models for BC Risk WP aims to develop machine learning models that can predict the risk of breast cancer using clinical and histopathological information and radiomic features. Deliverables include the machine learning models developed, documentation of their algorithms and validation results, and reports on their performance in predicting short-term breast cancer risk. The WP is led by Bari and all units will participate; it supports Aim 2. The Breast Density WP evaluates breast density using various automated software. The assessment of breast density will allow us to stratify the trial results. Deliverables include reports on the performance of different software tools, a database of breast density measurements, and an analysis of the impact of breast density on screening outcomes. The leading unit is IOV; the WP contributes to Aims 1 and 3 defining different scenarios for the introduction of CEM and MRI in different groups of women defined by breast density. The Radiation Dose WP aims to ensure that the radiation dose associated with CEM is within acceptable limits. Deliverables include a report on the radiation dose associated with CEM and recommendations for optimizing imaging protocols to minimize radiation exposure. The leading unit is AUSL-RE; the WP supports Aim 1 and will inform Aim 3 about the undesirable effects of the test. Finally, the HTA (health technology assessment) WP will assess the financial and organizational impact and explore the ethical, legal, and social issues of different scenarios of introducing CEM or MRI in the surveillance of high-risk non-mutated women. Deliverables are systematic reviews of the evidence integrated with Con-TRUST results and Evidence to Decision tables to be used by the Italian guideline initiative on breast cancer screening. The WP supports Aim 3 and is led by AUSLRE, with the contribution of other units.

Specific aim 1 The Con-Trust (CONtrast enhanced mammography Tailored Risk-Using Screening Trial) will include about 2200 high-risk women in the following centers: IOV of Padua, AUSL-IRCCS of Reggio Emilia, AOU of Messina, AUO of Bari, AOUI of Verona, Policlinico S. Matteo of Pavia, Policlinico San Donato Milanese, Irst-Meldola-Ausl Romagna, AOU of Modena. Women aged 30 to 65 years with a 5-year breast cancer (BC) risk \>5% will be contacted and informed about the study. The number of women enrolled in the dedicated surveillance programs at the participating centers, receiving a visit every 12 months, is expected to be adequate to reach the target sample size in less than 12 months, assuming a participation rate of less than 50%. Participants will be assessed using validated tools such as Tyrer-Cuzick IBIS, BOADICEA, BCSC or MyPeBS to determine their level of risk. All these tools rely on family and screening history, hormonal and reproductive history, breast density, and genetic information (BRCA1/2, TP53). In addition, MyPeBS includes a genomic risk score based on 313 single nucleotide polymorphisms and is available to women who previously participated in the MyPeBS study in Reggio Emilia. Women with known contraindication for MRI will undergo contrast-enhanced mammography (CEM) only, while those with known genetic mutation will receive the center-specific surveillance, usually MRI+DM. Finally, the remaining women (expected 1,400) will be randomized (700 per arm) to two rounds of CEM surveillance or to two rounds of MRI+DM surveillance. After 2 years, all women who were randomly assigned to receive either MRI or CEM, and who were eligible for both, will undergo the same exit test. This exit test is the standard test used for the surveillance of high-risk nonmutated women at each participating center. The co-primary endpoints are cumulative incidence of BC, including invasive BCs and ductal carcinoma in situ, interval, and screen-detected BCs, and cumulative recall-rate. The main analysis will only cover the subpopulation that was randomized. If the two techniques are equivalent, the predicted sample size will have 80% power to exclude a non-inferiority threshold of 2.5% higher incidence of BC, with 95% confidence. This would be equivalent to a reduction of less than one-third of cancers diagnosed early by the screening test. The cumulative incidence, based on an average risk of 7% at 5-years and a lead time of 4 years from the exit test, has been estimated at 8.4%, with 5.3% detected by screening at baseline and 3.1% occurring during follow-up based on results from the DENSE trial). Secondary endpoints include biopsy and false positive rate, positive predictive value, interval cancer rate, and cumulative incidence of advanced (T2+) cancers. The project involves two rounds of screening, with an interim analysis at 12 months (plus 3 months to include assessment of women positive at 12 months). The final analysis will be conducted at 24 months (plus 3 months) following the end of the project. The safety outcomes are mean glandular dose, adverse events, and reactions to contrast agents. Because eligible women have a high lifetime risk of BC, overdiagnosis is not considered a safety outcome. However, the entire study population will be considered for false-positive rates, glandular dose, and reactions to contrast agents. Subgroup analyses will be conducted based on key risk factors (age group, breast density, and risk level). Retrospective evaluation of images will allow estimation of the cancer detection rate of restricted imaging protocols such as unenhanced MRI (specifically, diffusion-weighted imaging) and MRI plus single medio-lateral oblique mammographic projection. The final study protocol will be defined with the active involvement of women and public health decision-makers.

Specific aim 2 Mammographic features, particularly breast density, are related to both BC risk and interval cancer incidence. Researchers are exploring ways to combine mammography and tomosynthesis data with standard variables to improve BC risk prediction. Limited studies on breast MRI or CEM suggest promising prognostic value and ability to predict BC. The Con-TRUST trial will have access to both mammographic and contrast-enhanced images for women participating in at least two rounds of screening. Accurate stratification of women according to their risk is critical for effective personalized screening. The potential benefits of screening are greatest for high-risk women, as they are more likely to develop breast cancer. However, the potential harms of screening are largely unrelated to the women\'s risk and are instead associated with false positives results or direct effects of the test. Therefore, stratifying the population according to individual risk can help optimize the balance between benefits and harms by recommending intensive and potentially risky screening strategies only to women with sufficiently high risk that the expected benefits outweigh the harms. The study will analyze images to extract numerical data on contrast, luminosity, texture indexes, and machine learning-derived features, as well as collect radiologist-defined characteristics such as the presence of calcification, visually assessed density, background enhancement, and other patterns. Several computational methods will be used to explore the ability to predict cancer characteristics along with clinical information: artificial intelligence methods as well traditional logistic models. The endpoint will be incidence of breast cancer, including invasive BCs and ductal carcinoma in situ. Sensitivity analyses will be conducted by limiting the clinical endpoint to invasive cancer and incident cancers, i.e. excluding prevalent cancer detected through imaging examinations made at the time of recruitment and excluding BRCA1/2 and P53 mutated women. The area under the curve at receiver operating characteristics analysis and its 95% confidence intervals (according to the exact binomial distribution) will be calculated, classifying women according to the model\'s predicted risk of BC and observing how the sensitivity and specificity of model classification vary at increasing thresholds of risk. All women included in the study who are randomized, assigned to CEM for contraindications to MRI, or mutated and following the standard practice will contribute to this aim.

Specific aim 3 To make evidence generation more usable for the health system and accelerate decision-making on the best allocation of CEM or MRI in the surveillance of high-risk women, the trial will include substudies on costs, feasibility, and organizationalimpact, and acceptability. The principles of HTA will guide this evaluation. Costs will be collected through an activity-based cost analysis, assessing the consumption of human and technological resources for activities that may differ between the two strategies. If non-inferiority is demonstrated, costs related to the two strategies will be projected in a budget impact analysis with a 5-year horizon. A cost-minimization analysis will also be conducted assuming equal effectiveness, considering both the public and individual perspectives, including costs and time spent by users. Organizational feasibility and impact will be assessed through interviews with key people, such as decision-makers, screening program coordinators, directors of imaging department, and professionals involved in the surveillance program. Women\'s preferences and values given to the considered outcomes will be collected through interviews and focus groups involving participants to investigate the acceptability of the two technologies and possible surveillance protocols. The process will be conducted with the advice of a user and stakeholder engagement board, which will include high-risk women, representatives of patient associations, public health system decision makers, and health professional representatives. Systematic reviews will integrate study results into the framework of existing knowledge on the use of CEM in women at high and intermediate risk of BC. The Steering committee of the Italian breast cancer screening guidelines adolpment project, coordinated by the Osservatorio Nazionale Screening, will prioritize relevant clinical questions and frame them as PICO (Population Intervention Comparator Outcome). Evidence from systematic reviews on efficacy and safety, costs, and women\'s preferences will be integrated with context-specific information from trial sub-studies on costs, feasibility, organizational impact, and acceptability. The evidence will be summarized in Summary of finding and Evidence to Decision tables, using the GRADE (Grading of Recommendations Assessment, Development and Evaluation) methodology, and proposed for developing recommendations to the Scientific Committee of the breast cancer screening guideline adolopment of the European Guidelines.

Experimental design aim 1 High-risk women who are scheduled for a surveillance test will be informed about the study through mail or phone. If willing, women can schedule an appointment for detailed information and provide consent. Recruitment can also occur during their first episode on-site. Validated risk assessment tools like Tyrer-Cuzick IBIS, BOADICEA, Breast Cancer Screening Consortium, and Mammorisk will be used to assess eligibility based on a 5% estimated risk of BC in the next five years. Algorithms will be provided to calculate the risk. Exclusion criteria are pregnancy and non-comprehension of information. After assessing eligibility for randomization, the investigator will assign women directly to Contrast-Enhanced Mammography (CEM) if they have claustrophobia, MRI-unsafe devices, intolerance to gadolinium, or are unable to enter the MRI gantry. Women with a known intolerance to iodine contrast medium or known mutations in DNA repair genes (BRCA1, BRCA2, TP53) will be assigned directly to MRI. MRI will be performed during the indicated phase of the menstrual cycle (day 5-14) according to each center\'s protocol. It will include T2W suppressed, post-contrast dynamic imaging with MIP reconstruction, and diffusion-weighted imaging (DWI). MRI and Digital Mammography (DM) will preferably be performed on the same day and interpreted together. CEM, involving two views per breast, will also be conducted during the indicated menstrual cycle phase (day 5-14) following the injection of contrast media, with the first image acquired 2 minutes post-injection. The order of projection (MLO and CC for each breast) will not affect the results. Experienced breast radiologists will interpret the images locally. Ultrasound (US) second look will be considered based on local protocols but only for specific cases. Women with positive or suspicious findings will be recalled for further assessment, which may include ultrasound-guided core needle biopsy (CNB) or stereotactic vacuum-assisted breast biopsy (VABB) depending on the easiest imaging to localize the finding. Women with inconclusive CNB results will undergo VABB. Those with malignancy or B3/B4 results on VABB will be discussed by a multidisciplinary team. Short-term follow-up may be recommended in exceptional cases, usually after six months. The study will include these screen-detected findings in the baseline outcomes. Women with negative assessments will be referred to regular screening and actively re-invited after 12 months. In the third screening round, all women in the randomized sub-cohort will be rescreened using the standard test used for high-risk non-mutated women at each center. Retrospective imaging reading, blinded to the outcome, will be conducted to determine the accuracy of different imaging protocols, such as unenhanced MRI (DWI-only with ADC maps; DWI plus T1-weighted plus T2-weighted sequences), MRI abbreviated protocol, MRI plus single mediolateral oblique Mx projection, and reversed hanging protocol for CEM. Breast density will be objectively assessed to conduct sub-group analyses based on breast density categories. In addition, Aim 1 of the study will include site visits for quality control of MRI and CEM equipment. This will ensure that the imaging devices are functioning optimally and producing accurate results. Furthermore, as part of the study, the radiation dose delivered by CEM and DM will be accurately measured and recorded for the CEM and MRI+DM arms of the randomized sub-cohort, respectively. This information will also be obtained for women who undergo CEM alone or to MRI+DM in non-randomized cohorts. By collecting this data, the study aims to assess the radiation exposure associated with CEM and evaluate its safety and effectiveness as a screening method for high-risk women.

Experimental design aim 2 The objective of this experimental design is to develop a model for predicting breast cancer (BC) risk during the study period. The design considers multiple outcomes: 1) all cancers (including invasive and ductal carcinoma in situ [DCIS], both prevalent and incident, and those detected by screening or as interval cancers); 2) only incident cancers (excluding screendetected cancers that were positive at the entry test); and 3) interval cancers. Sensitivity analyses excluding DCIS will be conducted. The experimental design involves processing low-energy CEM images from the randomized CEM arm and the observational cohort, as well as digital mammography (DM) images from the randomized MRI+DM arm and the observational cohort. An artificial intelligence (AI) software tool will be used to assess breast density and calculate a breast cancer risk score. The goal is to refine or potentially replace existing breast cancer risk models with this AI-based approach. Radiomics features will be extracted from mammographic images using AI tools for risk assessment. Radiomics refers to high-dimensional quantitative features extracted from medical images that capture information about tumor heterogeneity, texture, shape, and other characteristics that are not easily discernible to the naked eye. These features may include statistical measures such as mean, standard deviation, skewness, and kurtosis of pixel intensities within the breast region. Texture features, such as entropy, contrast, and homogeneity, will also be extracted to quantify patterns and spatial relationships within the breast tissue. In addition, shape-based features, such as compactness, circularity, and sphericity can be computed to capture the geometric properties of lesions or regions of interest. These features provide a rich set of data that can be used to improve risk prediction models. To incorporate these radiomics features into risk prediction models, various machine learning algorithms will be employed. These algorithms are able to learn patterns and relationships within the data and make predictions based on the learned models. Various machine learning models can be used, such as logistic regression, support vector machines (SVM), random forests, or gradient boosting machines (GBM). Training and validation datasets will be carefully constructed to ensure representativeness of the study population and to capture the full spectrum of breast cancer risk. The datasets will include cases from both the CEM and DM arms, as well as the observational cohort. Variability among different imaging systems and technologists will be considered to account potential sources of bias. During the model development process, the performance of machine learning models will be evaluated using appropriate metrics, such as accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curve (AUC). The AUC represents the ability of the model to discriminate between individuals with and without breast cancer. It provides a comprehensive measure of the predictive performance of the model across different risk thresholds. In addition, the performance of AI-based risk assessment models will be compared with existing state-of-the-art breast cancer risk models. This comparison will help determine the added value of radiomics features and machine learning algorithms in improving risk prediction. Overall, the combination of radiomics features extracted from mammography images and machine learning models will enable the development of a more accurate and robust breast cancer risk prediction model, potentially improving personalized screening strategies and optimizing the balance between benefits and harms of screening.

Experimental design aim 3 Prior to the start of the project and before the protocol is submitted to the ethics committee, a stakeholder committee will be established, composed of representatives of high-risk women, patients\' associations, public health system decision-makers, and health professionals. This committee will play a crucial role in assessing the relevance of the study outcomes to women\'s health and to the health system in general. In addition, it will provide input into the review of informational materials and women¿s outreach strategies. The stakeholder committee will serve as an external advisory board for the study and will prioritize the clinical questions framed as PICOs (Population Intervention Comparator Outcomes) that will guide the systematic review of the use of contrast-enhanced mammography (CEM) and breast MRI in the surveillance of high-risk women. The systematic review protocol will be registered in the PROSPERO database, ensuring transparency and adherence to predefined methods. The review will adopt the standard methods of the Cochrane Collaboration to ensure rigor and minimize bias. The GRADE system will be used to assess the importance of outcomes and certainty of evidence. This approach allows the quality and strength of the evidence to be assessed, helping to formulate reliable conclusions. To examine the economic evidence, the NICE methodological checklist for economic studies will be used to assess applicability and methodological limitations. If the study results support equal effectiveness between the two surveillance strategies, a cost minimization analysis will be conducted. This analysis will consider both the direct costs incurred by the healthcare system and the time spent by women participating in the surveillance. To explore the organizational impact of the two surveillance strategies, interviews will be conducted with decision makers at the hospital and local health authority levels, screening program coordinators, directors of diagnostic imaging department, and professionals involved in the surveillance program. These interviews will provide insights into the practical implications and potential challenges associated with implementing the different strategies within the existing healthcare infrastructure. The acceptability of the two technologies and potential surveillance strategies will be investigated through interviews with participating women and focus groups. Face-to-face discussions with women will gather their perspectives and experiences, helping them to understand their preferences and concerns. Data extracted from the systematic review and information gathered through the interviews and focus groups will be summarized in summary tables of findings using the electronic online tool GRADEpro. These tables will present a concise and structured overview of the evidence, facilitating interpretation and synthesis of the results. In addition, Evidence to Decision Tables will be developed for submission to the Scientific Technical Committee of the Italian Breast Cancer Screening Guidelines Development Project, which contributes to the European recommendations on breast cancer screening. These tables will help consider and potentially include PICOs in the guidelines, ensuring the formulation of evidence-based recommendations. By involving stakeholders, conducting systematic reviews, evaluating economic aspects, assessing organizational impact, and investigating acceptability, the study aims to generate comprehensive evidence that can inform decision making and guide the best allocation of resources between CEM and MRI in the surveillance of high-risk women.

Hypothesis and significance International guidelines recommend annual MRI, usually in combination with digital mammography (DM), for women with a known BRCA1/2 mutation and those with a similarly high risk of BC. This is due to its higher sensitivity, as demonstrated by several multicentre studies, including one conducted in Italy. It is estimated that 1 to 3% of females aged 40 to 70 years, depending on the model used, can be classified with a mutated-like risk of BC and may benefit from surveillance by MRI. Breast MRI has been shown to reduce interval cancer in women with extremely dense breasts, and has been suggested as potential supplemental screening for women with dense breasts. However, MRI is a limited resource, expensive, and burdened by specific contraindications, such as claustrophobia and unsafe medical devices for MRI. As a result, many women at high risk for BC do not receive adequate surveillance. A very recent meta-analysis, which included more than 10,000 patients, showed that contrast-enhanced mammography (CEM) has a high accuracy in detecting BC, similar to that of MRI. CEM is already preferred over MRI for preoperative assessment by European Commission guidelines, and smaller studies on screening with CEM of high-risk women are underway (e.g., SCEMAM in the USA, and CESM in high and intermediate-risk women, which just concluded recruitment at IOV, Padua, one of the centers participating in this proposal). However, no study has compared MRI and CEM for downstream consequences of screening. The techniques use different contrast agents (CAs) that exhibit the same two-compartment intravascular-interstitial biodistribution. The gadolinium-based CAs used in MRI, apart from rare allergic reactions, are of concern for long-term accumulation in tissues when considering the high cumulative doses reached in annual screening programs, including the brain. In contrast, iodine-based CAs can cause nephropathy. CEM provides a radiation dose comparable to that of DM but offers both morphological evaluation (similar to standard DM) and functional evaluation (similar to MRI), simultaneously. Morphological evaluation allows for the detection of calcification-associated lesions, such as ductal carcinoma in situ, with less neoangiogenesis and the possibility of false negatives with MRI, especially in women who underwent thoracic radiotherapy. Moreover, CEM is estimated to cost less than one-third of MRI. Given its comparable efficacy and safety, as well as much lower costs and resource consumption, a non-inferiority study design aimed to investigate whether CEM can be used as an alternative to MRI+DM for early detection of BC in high-risk women with \>5% BC risk in the next 5 years.