Digital Breast Tomosynthesis for the Dutch National Breast Cancer Screening Program

In the last decade, several large-scale European studies have shown that digital breast tomosynthesis (DBT) increases the detection rate from \~6 to \~9 cancers per 1000 women screened, while the recall rate approaches 3.5% to 4%, independent of the initial recall rate. In the Dutch screening setting with digital mammography (DM), the recall rate is only 2.4%. Therefore, introducing DBT in the Netherlands will likely generate an increase in false positives, but the magnitude of this increase is unknown. Furthermore, the \~50x increase in the number of images resulting from a DBT exam compared to a DM exam doubles the reading time. For DBT screening in the Netherlands to be feasible from the economic and human resources point of view, the interpretation of these cases must be accelerated, to result in an average reading time comparable to that of reviewing DM exams. The aim of this project is to make the replacement of DM with DBT in the Dutch National Breast Cancer Screening Program feasible and sustainable, by ensuring its acceptability by the various stakeholders, optimizing the acquisition and reading strategy, determining its expected impact on the national recall rate, and determining its cost-effectiveness. In addition, completion of this project will result in the creation of a DBT-based screening database available for future research. Work Package (WP) 1 will involve the introduction of DBT in 6 of the 71 screening units in the country to screen 18,200 women for two screening rounds. This study is powered to detect a combined one-third reduction in the interval and advanced cancer rates with DBT screening as the primary endpoint. In addition, as secondary endpoints, the Dutch DBT-based screening recall rates will be determined for the first and second rounds of DBT screening. A contemporary control group of about 86,400 women screened with DM will be selected from the screening database (ScreenIT) of the national screening program. This control group will be selected to reflect the distribution of urban and rural areas as well as the mixture of mobile and fixed units of the screening units selected for the intervention group. In addition, the attendees, and the screening radiographers and radiologists will be surveyed to identify any barriers that limit acceptability of this new modality. Finally, WP1 will also involve the creation of a DBT-based screening database for further research. In WP2, the optimal DBT reading strategy to be used in the Dutch screening program to minimize the increase in reading time with DBT, will be identified. For this, retrospective observer studies will be performed, using the images from WP1, to test eight different acquisition and reading strategies, some involving the use of cutting-edge artificial intelligence computer systems. These strategies have been shown in initial studies, many performed by the sponsor, to result in a significant reduction in reading time, with no or minimal loss in performance. Once the optimal reading strategies is identified, it will be used to interpret the entire second round of DBT screening acquired in WP1 at the screening reading units retrospectively. In this way, its actual performance and impact on reading time in the real screening environment will be determined. Finally, in WP3, the results from the other two WPs will be used to model the entire DBT screening program and therefore predict the long-term outcomes and cost-effectiveness of this new screening program. The only method possible to estimate the impact of screening on long-term effects, such as breast cancer mortality and morbidity, as well as cost-effectiveness, is the use of modelling. Therefore, the well-validated MISCAN model will be used, with the data and results gathered, to estimate the long-term impact of DBT screening. As a result of this project, if successful, breast cancer screening with DBT will be ready for nationwide implementation, having addressed all the country-specific questions and having optimized the DBT workflow specifically for a high-volume population-based screening program.