Abstract
Background: Most low-dose computed tomography (LDCT) lung cancer screening guidelines recommend shared-decision making (SDM) before initiating screening. Indeed, Medicare requires evidence of SDM for reimbursement. The clinical benefit of screening, however, varies dramatically across eligible patients. Also, the harms of LDCT, such as fear and unnecessary procedures incurred by false positive results, can be quite substantive. Clinicians and health systems, therefore, need individually tailored screening guidance, depending on the extent of benefit. To this end, we developed the Personalized Lung Cancer Screening Model, a microsimulation model that estimates individual-specific health gain from LDCT screening. This model evaluates the potential effects of patient preferences on health gains across low- and high-benefit groups.
Methods: We estimated the effects of LDCT screening on lung cancer outcomes and quality-adjusted life years (QALYs). Our natural history model was built based on previously validated lung cancer models, constructed by utilizing different data sources: two large randomized lung cancer screening trials (NLST and PLCO) and the Surveillance, Epidemiology and End-Results cancer registry. For this study, we simulated a nationally representative sample of 1 million patients eligible for LDCT screening, whose risk profiles mimic adult smokers participated in the National Health Interview Study (NHIS) from 2010 to 2014. We quantified patient preferences using literature-derived utilities (e.g., the burden of testing, false-positive diagnoses, treatment, and complications that result from the screening and treatment process). Besides inherent uncertainty in some utility measures, our primary aim was to understand the effect of varying patient preferences on the net benefit of screening. Therefore, we performed a further analysis by varying utilities over a plausible range.
Results: Our model predictions of lung cancer incidence and mortality rates in the NLST and PLCO participants matched well to the observed rates. Similarly, average incremental QALY gains were consistent with that found in a previous NLST-based cost-effectiveness analysis. Among the simulated NHIS population, incremental QALY gains varied significantly across differing baseline risk of developing lung cancer (range in base-case analysis: 2 QALYs lost per 100 people screened to 6 QALYs gained per 100 screened). Our analysis for patient preferences showed that the magnitude of net benefit from LDCT screening is not very sensitive to patient's views of the burdens and harms of testing and treatment if the patient's baseline lung cancer risk was above the third decile. That is, even assuming unfavorable preferences, those above 3rd decile of risk generally experienced net benefit, while the less than 3rd decile of baseline risk was a more preferences sensitive zone.
Conclusion: Results from our Personalized Lung Cancer Screening Model demonstrate the importance of an individual's estimated baseline lung cancer risk in determining net benefit from LDCT screening. In addition, we found that patient preferences play an important role to determine the extent of net benefit. These findings support the use of a decision-support tool through shared decision making, rather than recommending screening uniformly.
This abstract is also being presented as Poster B19.
Citation Format: Pianpian Cao, Tanner Caverly, Rodney Hayward, Rafael Meza. Effect of cancer risk and patient preferences on net benefit of lung cancer screening: A personalized lung cancer screening model. [abstract]. In: Proceedings of the AACR Special Conference: Improving Cancer Risk Prediction for Prevention and Early Detection; Nov 16-19, 2016; Orlando, FL. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2017;26(5 Suppl):Abstract nr PR16.