OBJECTIVES: A cost-effectiveness model was previously built to assess the use of atezolizumab in the adjuvant treatment of stage II-IIIA non-small cell lung cancer (NSCLC) using time-invariant for the post disease-free survival (DFS) transition probabilities. The choice to build a more complex model using tunnel states to allow time-variant transitions or a simpler model using time-invariant transitions needs to be weighed between time, resource, and how robust the model would be for decision making. The potential issue with the simpler model is that the modelled overall survival (OS) may be biased. Here, we build a more complex model that allows for time-varying health state transition probabilities. The objective of this analysis is to better understand the impact of the two methodologies on the outputs.

METHODS: The cost-effectiveness model (Yip et al. (2022)) used for the NICE submission was updated to allow the use of time-variant transition probabilities using tunnel states and assessing the appropriate models using the NICE DSU 14 Technical Support Document (i.e. rather than the previous approach of assuming exponential distribution).

RESULTS: The revised model shows that using either time-invariant or time-variant post-DFS transition probabilities has little impact on the estimated incremental cost-effectiveness ratios (ICERs). Moreover, using time-invariant transition probabilities did not have a major impact on the modeled OS, which were within the 95% confidence interval of the observed Kaplan-Meier OS from IMpower010.

CONCLUSIONS: The results show that the use of a simpler approach that restricts the post-DFS health state transitions to being time-invariant is suitable for decision making. This reflects the fact that the main driver of the results is the significant improvement in DFS provided by adjuvant atezolizumab treatment.