The Lung Cancer Outcomes Simulator (LCOS) Model (Stanford University) is based on a natural history model of lung cancer that assumes exponential growth for the primary tumor and growth of metastasis that is proportional to the tumor growth (1). The natural history model simulates the individual’s tumor growth in the absence of any intervention, based on gender and histologic subtype, providing outcomes such as tumor volume doubling time, time for onset of metastasis, tumor size at clinical detection, and survival time. The parameters of the natural history model are estimated using SEER survival data.
In the natural history model, the primary tumor grows exponentially with growth rate r , and its corresponding tumor volume doubling time is given as (log2)/r. When the tumor reaches a certain size (Vp), it prompts symptoms that lead to a clinical detection of the primary tumor. The tumor volume and the growth rate parameter are modeled as a bivariate log-normal distribution. Fatal metastases start growing at a certain time (TOM) and we assume the volume of the primary tumor at this time (Vc) is a threshold for cure (cure threshold) so that only if the primary tumor is detected and treated earlier than this point, the patient is cured.
The metastatic burden grows proportionally to the primary tumor size with fraction f until it reaches a maximum burden size (BD). The time at which BD is reached is a survival time if the primary tumor is not detected and treated before reaching the cure threshold. If this metastatic burden grows to a certain size (C1BD), which is a fraction of the maximum burden, it becomes an observable metastasis; thus if detection (either clinical or screen detection) occurs earlier than this point, the tumor is staged as early stage, otherwise as advanced. When the metastatic burden grows to a certain size (C2BD), it becomes clinically symptomatic; hence detection is due either to the primary tumor or to metastasis, depending on which becomes symptomatic first.
The LCOS superimposes a specific screening intervention to each individual and estimates individual-level outcomes, which can be aggregated for evaluating population-level outcomes. Key inputs for the LCOS include gender, individual-level smoking history (e.g. pack-years and age for starting/quitting smoking), and age of entry to the screening program. We use the Two-Stage Clonal Expansion (TSCE) model to predict annual hazards for lung cancer incidence in the absence of screening given smoking history, gender, and age at entry (2). The parameters of the TSCE model were estimated based on data from Nurses’ Health Study/ Health Professionals’ Follow-up Study. We note that this incidence of lung cancer in the absence of screening is due to detection prompted by symptoms (i.e., clinical detection) as compared to screen detection.
For each lung cancer, a histologic subtype (adenocarcinoma, squamous, large cell or small cell) is assigned by sampling from the observed proportions from Surveillance, Epidemiology, and End Results (SEER) data. The LCOS compares the performance of the screening program of interest against the no screen scenario estimating the benefit of screening. Outcome measures include the lung cancer deaths averted, all cause and lung cancer-specific mortality reduction, number of detected cases by mode of detection, number of screening exams, number of false-positive results, and number of overdiagnosed cases, among others.
References
- Lin RS, Plevritis SK. Comparing the benefits of screening for breast cancer and lung cancer using a novel natural history model. Cancer Causes Control. 2012;23(1):175-185.
- Meza R, Hazelton WD, Colditz GA, Moolgavkar SH. Analysis of lung cancer incidence in the Nurses' Health and the Health Professionals' Follow-Up Studies using a multistage carcinogenesis model. Cancer Causes Control. 2008;19(3):317-328.
