Low dose computed tomography-based lung cancer screening has been shown through the National Lung Cancer Screening Trial (NLST) to reduce mortality for high risk patients. We evaluate the use of stochastic simulation for cost-effectiveness analysis of CT screening, a technology which exhibits low specificity and potentially high downstream costs.
Screening effectiveness was tested using a natural history of lung cancer. Probability of metastatic disease and cure at detection are functions of tumor size. Parameters estimated were based on the Surveillance, Epidemiology and End Results cancer registry. Screening and treatment components in the simulation are based on protocols used in the Mayo CT trial. Costs of diagnostic and therapeutic procedures were obtained from the Medicare payment database. We assessed cost-effectiveness for all scenarios from ages 55-80 at 5 year starting intervals with screening ranging from 5-30 years. These scenarios were examined based on one-time, annual, biennial, and every five year intervals.
Screening approximately doubles incidence compared to the unscreened population. Clinically, the simulation model also shows a 20.5% reduction of 5-yr mortality in the screened group as opposed to the simulated, unscreened population. The most cost-effective strategies screen at every two or five year intervals from ages 60-80. Expenses in a microsimulation are tailored to show aggregate costs, with or without discounting, over the lifetime of simulated patients on a societal perspective. Varying the intensity of screening contributes to large changes in cost-effectiveness. Increased duration and frequency of screening derive marginal benefits but come at incrementally higher marginal costs. These data correlate with an even wider range of published cost-effectiveness analysis results, ranging from less than $20,000 to over $100,000 per QALY.
Simulation models can be used to predict how screening affects both clinical and economic outcomes. As extensions of clinical trials, clinical-economic model results and their conclusions should be incorporated into the development of future cancer screening and treatment programs.
Clinical trial identification
Results were obtained using a computer-based microsimulation. This research did not involve the use of personally identifiable health information and was not performed as part of an onging clinical trial.
All authors have declared no conflicts of interest.