A PROBABILISTIC DECISION MODEL TO GUIDE OPTIMAL HEALTH POLICY DECISIONS FOR LUNG CANCER SCREENING

Purpose: It has yet to be determined whether earlier detection of lung cancer with screening translates into reduced mortality. Trials have been initiated, but in the interim, debate continues over the implications given uncertain efficacy and risks of screening in practice.  We address this lack of information with a decision-analytic model.

Method: We developed a probabilistic Markov model of cost-effectiveness for lung cancer screening with helical computed tomography (hCT) compared with chest x-ray (CXR) and no screening (NS).  The base case was 55-year old US current/former smokers screened annually until age 74. Data were from published literature, Surveillance Epidemiology and End Results database, and Lung Screening Study (LSS), a nationally representative trial.  We assumed that stage shifts observed with screening translate into survival benefits.  Analyses explored the impact of alternative cohort ages for starting and stopping screening, smoking status, overdiagnosis, compliance, and false positive (FP) rates. We estimated incremental cost-effectiveness ratios (ICERs) for life years (LYs) and quality-adjusted LYs (QALYs), and cost-effectiveness acceptability curves.  We performed value of information analysis to estimate the potential value of further research to reduce uncertainty.

Result: no In the base case: (a) CXR cost $362,291/LY vs. NS; (b) hCT cost $634,982/LY vs. CXR; and (c) both CXR and hCT were dominated by NS in the QALY analyses.  The probability that NS is cost-effective was 91% at a maximum acceptable ratio of $100,000/QALY (7% and 2% for CXR and hCT, respectively).  When only current smokers were screened, screening was performed once, screening was initiated at either age 45 or 65, or there was no overdiagnosis, CXR and hCT remained dominated by NS when QALYs were the outcome.  If hCT were associated with FPs, false negatives, or overdiagnosis, hCT cost $144,671/QALY vs. NS, and CXR was dominated by NS.  Replacing the base case LSS FP rates for hCT with those from the Early Lung Cancer Action Program resulted in an ICER of ~$1.3M/QALY for hCT vs. NS.  When the FP rate for hCT was set at 0, this ICER decreased to $297,759/QALY.

Conclusion: As expected, both CXR and hCT were associated with increased LYs and costs. However, the associated negative quality-of-life effects of FPs and overdiagnosis resulted in both screening options being dominated by NS when QALYs were the outcome.