For decades our biological and clinical understanding of breast cancer has been based on three therapeutically predictive biomarkers: estrogen (ER), progesterone (PR) receptors and the human epidermal growth factor receptor-2 (HER2). Today, we recognize that breast cancer biology is more complex;as well, clinical oncologists routinely use additional biomarkers and gene expression signatures (e.g. Ki-67/IHC4, MammaPrint or Oncotype-Dx) to recommend breast cancer treatments. Despite this deeper understanding of breast cancer biology and increasing clinical use of biology-driven breast cancer therapeutics, we lack population-based estimates of the extent that these ever more costly breast cancer subtype-targeted diagnostics and therapeutics actually reduce breast cancer mortality (BCM), improve quality of life (QOL), or otherwise prove cost-effective. To address this need and overcome the challenging constraints imposed by cross-sectional US population modeling efforts now being advanced by the Cancer Intervention and Surveillance Modeling Network (CISNET), we will model an expanded repertoire of prognostic and predictive biomarkers linked to biology. We will employ a unique longitudinal population dataset not available in the US: the 40+ year old Stockholm Breast Cancer Registry, which currently tracks ~40,000 individual breast cancer patients over time and is annotated for screening, tumor biomarkers, treatments and outcomes, and which can be linked to the Stockholm Mammography Registry through unique identifiers, providing an unparalleled longitudinal population dataset for modeling. To model the population benefits of more modern predictive biomarkers and tailored adjuvant therapies, we will utilize our access to two other unique breast cancer randomized trials: the Stockholm-1 and I-SPY clinical trial datasets. Stockholm-1 consists of 729 women randomized to tamoxifen vs. no systemic therapy with 30-year follow-up;and the I-SPY trials are fully characterized biomarker-driven trials of pathway targeted agents that include response to therapy and event-free survival outcomes. Finally, we will update the CISNET model to estimate the population level benefits (BCM and cost effectiveness) of a more biologically targeted approach to treatment and screening.
The specific aims for this study include:
Aim 1. Develop and program a bridging model using longitudinal Swedish population data to determine the impact of assigning treatments on the basis of biological subtypes. This model will then be tailored to the US population using biased sampling to reflect SEER characteristics.
Aim 2. Use the model in Aim 1 to evaluate the population effects on breast cancer mortality of tailored therapy employing highly characterized data sets with survival benefits and/or response rates from biomarker-driven outcomes and or/treatment.
Aim 3. Estimate the population level cost effectiveness of biologically targeted therapy.
Breast cancer is the most common cancer among women. We now understand that breast cancer represents a collection of diseases that are best approached by using biomarkers to better characterize the disease and treat it accordingly. We intend to model the impact of more targeted screening and treatment of breast cancer for the population level impact to drive changes in our approach to this disease.