BCOS (Stanford) Stanford University

The Breast Cancer Outcomes Simulator (BCOS) was developed for four primary purposes. First, BCOS generates a virtual tumor registry of breast cancer patients diagnosed in the United States since 1975 and, at the individual level, specifies the patient's screening history, mode of detection, adjuvant treatment and survival; second, BCOS quantifies the impact of screening mammography and adjuvant therapy on breast cancer mortality trends from 1975 by molecular subtype; third, BCOS predicts what the incidence and mortality trends would have been had alternative age groups been targeted for screening, had there been changes to the interval between screening examinations, and/or had there been changes to the subgroups targeted for adjuvant therapy; fourth, BCOS predicts how future trends in breast cancer mortality may be affected by new screening and treatment interventions shown to be beneficial at the clinical trial level.

Contact: Sylvia Plevritis sylvia.plevritis@stanford.edu

Purpose

The Breast Cancer Outcomes Simulator (BCOS), created at Stanford University, was developed for four primary purposes. First, BCOS generates a virtual tumor registry of breast cancer patients diagnosed in the United States since 1975 and, at the individual level, specifies the patient's screening history, mode of cancer detection, adjuvant treatment, and survival. Second, BCOS quantifies the impact of screening mammography and adjuvant therapy on breast cancer mortality trends from 1975 by molecular subtype. Third, BCOS predicts what the incidence and mortality trends would have been had alternative age groups been targeted for screening, had there been changes to the interval between screening examinations, and/or had there been changes to the subgroups targeted for adjuvant therapy. Fourth, BCOS predicts how future trends in breast cancer mortality may be affected by new screening and treatment interventions shown to be beneficial at the clinical trial level.

Overview

The BCOS model aims to reproduce population-level US breast cancer incidence and mortality trends from 1975-2015 (Surveillance, Epidemiology, and End Results [SEER] data) by capturing breast cancer events that involve heterogeneity in disease progression, patient characteristics, compliance with screening, and response to adjuvant treatment. The current BCOS model incorporates an update to the background breast cancer age-period-cohort (APC) incidence and uses background breast cancer incidence derived from a novel approach developed under the APC framework; the approach explicitly considers the effects of screening and MHT.

Our model provides estimates for population–level breast cancer mortality trends by simulating the life history of individual patients then aggregating the breast cancer related outcomes at the population level. Via the Monte Carlo method, the following characteristics are generated for an individual breast cancer patient: (1) the date of her birth, (2) the age of her death of causes other than breast cancer, (3) the ages she undergoes screening examinations, (4) the age she would be detected with invasive breast cancer in the absence of screening, (5) the age she would be detected with invasive breast cancer in the presence of screening, (6) her primary tumor size, extent of nodal and distant involvement and ER/HER2 status at the time of detection in the presence and absence of screening, (7) the adjuvant treatment she received in the presence and absence of screening (it is assumed that she received primary therapy which would include surgery and possibly radiation) (8) her breast cancer survival time given her disease stage, size, age at detection and mode of detection, (9) her cause of death (i.e. breast cancer, other causes).

Natural history is modeled by subtype as a progressive disease in which tumors grow exponentially and transition from local to regional to distant stages based on a hazard function dependent on tumor size. We model the tumor size and SEER historic stage (defined as local, regional or distant) of patient’s tumor at and before the moment the tumor is detected upon clinical examination due to presentation of symptoms. The simulation model traces the tumor from the moment it clinically surfaces “backwards” in time and provides estimates of the size and stage of the tumor at any time during the preclinical phase of the disease. The current BCOS model quantifies the effects of menopausal hormonal therapy (MHT) on and on tumor onset, growth, and detectability and breast cancer incidence and mortality trends.

A screening schedule that specifies the patient’s age at the time of screening mammography is superimposed on the patient’s disease history. A patient is screen detected only if the size of her tumor is at or above the tumor size detection threshold of mammography at the time of screening. Once the patient is detected, she is assigned a breast cancer specific survival time dependent on her age, tumor size, SEER historic stage mode of detection and her use of adjuvant treatment. Treatment dissemination is estimated by treatment (adjuvant chemotherapy, adjuvant tamoxifen, both or no treatment) received for given age, stage, size and ER status at detection. Her age of death is the minimum age of breast cancer death and the age of other cause death. Individual level outcomes are aggregated and summarized as population level outcomes in terms of age–adjusted breast cancer incidence and mortality. The current version of BCOS incorporates underlying breast cancer-specific survival by ER and HER2 subtype and has been updated to incorporate non-proportional hazard to the effect of treatment depending on the ER status of each breast cancer case.