Skip to main content

Breast CAncer STratification: understanding the determinants of risk and prognosis of molecular subtypes

Periodic Reporting for period 3 - B-CAST (Breast CAncer STratification: understanding the determinants of risk and prognosis of molecular subtypes)

Reporting period: 2018-09-01 to 2020-02-29

Female breast cancer is the fifth leading cause of death worldwide (627,000 deaths yearly). However, breast cancer is not one entity; different subtypes have different aetiologies and prognosis, ranging from highly fatal to fully curable cancers.
A combination of genetic and environmental factors, giving rise to distinct risk factor profiles, determines the molecular subtype of breast cancer. For example, presence of BRCA1 genetic mutations, predispose individuals to develop triple negative breast tumours. Lifestyle/environmental factor such as reproductive history influence development of breast cancer - for example nulliparous women are at higher risk than parous women. In addition, parity may be associated with the development of specific (possibly more hormone related) breast cancer subtypes.
The B-CAST project aims to pioneer development of subtype specific risk prediction models, which would more precisely identify those women who benefit from intensified or less screening and those who are likely to benefit from preventive (endocrine) therapy. For prognostication tools, the tumour expression profiles already used in clinical practice for prognosis, will be evaluated by key markers on a much larger sample size than explored before.
B-CAST project exploits existing resources, infrastructure and collaborations established through the Breast Cancer Association Consortium (BCAC). Clinical information from ~80,000 breast cancer patients with risk factor information has been collated and new data on molecular characterisation of a subset of ~20,000 tumours from a unique worldwide collection from large-scale epidemiological studies, clinical studies and biobanks is being generated. Generating new genomic information based on IHC/ISH panels on 25,873 tumours and targeted sequencing on a subset of ~10,000 tumours will inform risk and prognosis modelling of breast cancer. User-friendly online tools like CanRisk has been established that implement the risk and prognostication models like has been made that can be used to obtain personalized estimates of risk and prognosis.
Project objectives:
To define the influence of risk factors, including reproductive history, lifestyle, mammographic breast density and germline genetic variation, on breast cancer overall and by subtypes characterized by clinical and molecular markers.
To define the influence of risk factors and tumour subtypes on clinical prognosis.
To develop and validate breast cancer risk and prognostication models for breast cancer, overall and by subtypes, informed by knowledge acquired under above objectives.
To implement these models into online tools for risk prediction and prognostication; and make them available in multiple countries/languages.
To raise awareness, i.e. promote the development and integration of personalized breast cancer prevention within national public health programmes.
The B-CAST project started with a kick-off meeting together with the BRIDGES project in September 2015. For dissemination of the project results and communication a website was published:
Our database now contains over 220,000 breast cancer cases, and a similar number of healthy controls, with (varying extent of) clinic-pathological, risk factor and germline genetic data. After an inventory survey among BCAC studies, novel risk factors such as medication use, comorbidity, density, and the availability of mammograms have been added to this database. Breast density measurements from mammograms using the STRATUS software has been collected as well. A software package called PathXL was purchased, with specific features for B-CAST built in and functions as a repository for all digital TMA images and allows online access for scoring of these images. B-CAST has collected TMA data for >25,000 tumours, producing scores and scanned images for subsets of those tumours for a panel of 15 (14 immunohistochemical and one immunofluorescent) markers.
A panel for targeted DNA sequencing was developed. Significant breast cancer driver genes were identified by collecting and processing all published whole-genome and whole-exome sequencing data on breast tumours (large-scale tumour profiling studies; revision of available literature; and other previous large-scale sequencing studies). Gene selection for inclusion in the B-CAST targeted sequencing panel have been done based on prioritization using mutation frequency and relevance. The first version of the B-CAST panel included 107 genes. This panel was updated in 2018 to include 323 genes due to a change in preferred choice of panel sequencing technology. Homozygous deleted and amplified genes as well as hotspot promotor regions were added to the panel.
We allowed for inclusion of formalin-fixed paraffin-embedded (FFPE) tumour slides or cores, fresh frozen tumour cores, and already isolated DNA (for a minority of samples) from BCAC studies. DNA of tumours was centrally isolated using a standardized method. Each tumour DNA sample was quality tested and upon passing the quality criteria, a paired normal (blood derived) DNA sample was requested for sequencing as well. Over 10,000 tumour DNA samples were compiled and so far around 9,000 have been panel sequenced including a matching normal DNA sample.
B-CAST has been successful (together with BRIDGES) in improving the most comprehensive breast cancer risk prediction model BOADICEA, which among others has been updated with a new Polygenic Risk Score (PRS) that explains 20% of the familial aggregation of breast cancer. In January 2020 we released the CanRisk tool ( which implements the BOADICEA model, for health care professionals. CanRisk gained approval as a medical device (CE marking) by MHRA in November 2019.
For breast cancer prognostication, the PREDICT model (now incorporating an extended endocrine therapy option) is available with the PREDICT web tool (v2.2) (
B-CAST is expected to collate clinical data from ~80,000 breast cancer patients with risk factor information and clinical information including follow-up, and generate data on molecular characterisation of a subset of tumours from a unique worldwide collection; mostly representing women of European-descent. Molecular profiling of breast cancers on this scale, and integrating somatic genetic characteristics with germline genetics and environmental factors has not been previously attempted.
Using all this data combined, B-CAST will provide validated, easy to use online tools to identify women at risk of developing specific breast cancer subtypes of public health and clinical relevance, such as fatal cancers (~25% of all breast cancers) and interval cancers (~20%-30% of cancers detected in women undergoing screening) that tend to have more aggressive behaviour.
Additionally, B-CAST will perform a large-scale evaluation of germline genetic variation, tumour genetic variants and CNAs, IHC/ISH tumour markers, and environmental exposures in relation to prognosis. This information will lead to development of risk prediction models for breast cancer prognosis in specific tumour subgroups.
Translating these findings to the clinical practice could enable tailoring screening programmes based on individual characteristics, which could in turn improve the cost-effectiveness of screening programs within the EU aimed at reducing mortality from breast cancer.
Flow chart work packages