OncoSim-Lung (CPAC-StatCan) Canadian Partnership against Cancer, Statistics Canada

Summary

The OncoSim-Lung model is led and supported by the Canadian Partnership Against Cancer, with model development by Statistics Canada, and is made possible through funding from Health Canada. The OncoSim-Lung model is used to project the population trends in lung cancer in Canada. It has been used for evaluation of various interventions, particularly screening strategies and smoking cessation interventions for the Canadian population (Figure 1).

Figure 1: Overview of OncoSim-Lung model

OncoSim-Lung models the Canadian population consisting of individual life histories in which lung cancer may develop. The life histories consist of:

• Exposure to risk factors, particularly cigarette smoking and radon
• Diagnosis of lung cancer
• Stage-specific survival from lung cancer
• Influence of screening and smoking cessation on time of diagnosis and death

Individual life histories are created with events drawn from probability distributions. A life history starts with a birth date (earliest year of birth is 1872), assigning sex, and the jurisdiction of birth. Risk and age of death from causes other than lung cancer is determined from observed mortality according to the Canadian Vital Statistics Death database.

OncoSim-Lung simulates the smoking history of the Canadian population using smoking initiation rates, quit rates, and smoking intensity derived from 3 large Canadian health surveys (i.e., 1979 Canada Health Survey, 1994–1995 National Population Health Survey, and 2007-2008 Canadian Community Health Survey).^1-3 OncoSim-Lung models the changes in smoking rates and intensity across time periods for each age group, by sex and province. Smoking trajectories were externally validated against other survey years and tobacco manufacturers’ data.^4,5

The model calculates the annual risk of developing lung cancer for each person based on their smoking history and radon exposure in two steps. First, the model calculates the risk of lung cancer using the age and smoking-related coefficients in the PLCOall2014 risk equation, which was developed from the control group (never and ever-smokers) in the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO) and validated using data from the PLCO intervention arm.⁶ Then, OncoSim-Lung includes the impact of radon exposure on lung cancer risk by including a relative risk to the PLCOall2014 equation, where the relative risk was derived from a cohort study of uranium miners.⁷ The resulting overall risk equation has a residual term that takes into account the effects of other factors (i.e., factors that are not related to age, smoking and radon exposure) on lung cancer risk. The residual term was calibrated to match the lung cancer incidence cases from the Canadian Cancer Registry (diagnosed between 2015 and 2020).

For each incident lung cancer case, a histologic subtype—i.e., non-small cell lung cancer (NSCLC) or small cell lung cancer (SCLC)—is assigned at diagnosis. The stage distribution of invasive NSCLC (i.e., I, II, III, or IV) and SCLC (i.e., limited, extensive) varies by sex, age group, and jurisdiction, and is based on Canadian Cancer Registry data between 2010 and 2014. OncoSim-Lung uses a two-piece Weibull survival model that has been fit to Kaplan-Meier survival curves based on survival data from a cohort of individuals diagnosed with lung cancer in Ontario, Canada, with up to 3 years of follow-up data. Survival parameters were estimated by stage and first assigned treatment (i.e., surgery, chemotherapy, radiation therapy, none) and subsequently adjusted to align with age, sex, and provincial differences from the death-cleared Canadian Cancer Registry data.

The OncoSim-Lung screening module includes eligibility criteria, follow-up diagnostic procedures for positive screens, and screening-related factors such as sensitivity and specificity of LDCT, and a screening-related stage shift in diagnosis. OncoSim-Lung estimates the impact of screening using a preclinical sojourn time. Sojourn time, which is assigned to the individual at birth, is modelled as an exponential distribution and is not stage-specific. As the individual ages, OncoSim-Lung compares the assigned detectable preclinical cancer phase (i.e., sojourn time) with the wait time to clinical detection to determine the possibility of early detection by screening. OncoSim-Lung allows users of the model to choose between two follow-up protocols for positive screen results. The first, referred to as the “NLST protocol”, can be useful for validating the screening module in OncoSim-Lung against the NLST results and mainly means using the sensitivity and specificity and estimated detectable preclinical cancer phase (years). The other, referred to as the “custom protocol”, consists of several parameters that allow for screen results to fall into one of six risk categories, which are generically identified as 1 through 6. The follow-up actions and timing can be specified by risk category.

References

1. Statistics Canada. Canada Health Survey 1978/1979. https://www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSurvey&SDDS=3217
2. Statistics Canada. National Population Health Survey (NPHS):1994-2010. https://www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSurvey&SDDS=3225
3. Statistics Canada. Canadian Community Health Survey. https://www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSurvey&SDDS=3226
4. Statistics Canada. Table 16-10-0044-01 Tobacco, sales and inventories, monthly production (x 1,000). https://doi.org/10.25318/1610004401-eng
5. Forey B, Hamling J, Lee P, Wald N. International Smoking Statistics.. 2009.
6. Tammemägi MC, Church TR, Hocking WG, Silvestri GA, Kvale PA, Riley TL, Commins J, Berg CD. Evaluation of the Lung Cancer Risks at Which to Screen Ever- and Never-Smokers: Screening Rules Applied to the PLCO and NLST Cohorts. PLoS Medicine. 2014;11(12).
7. Whittemore AS, McMillan A. Lung cancer mortality among U.S. uranium miners: A reappraisal. Journal of the National Cancer Institute. 1983;71(3).