OBJECTIVES : Cost-effectiveness analyses in oncology are typically based on model structures with three health states. The aim of this study was to compare a 4-state model that explicitly simulates sequential treatment strategies with the standard 3-state approach.

METHODS : We used the open-source IVI-NSCLC model to evaluate the cost-effectiveness of a treatment strategy starting with gefintinib (comparator) to a strategy starting with erlotinib for treating epidermal growth factor receptive positive patients with metastatic non small cell lung cancer. In the 3-state model, the health states are stable disease with first line (1L) treatment (S1), progressed disease with 1L treatment (P1), and death. The 4-state model adds a fourth health state, progressed disease on 2L treatment (P2). In the 3-state model, state transitions are based on 1L treatment whereas in the 4-state model transitions from S1 are based on 1L treatment while transitions from P1 and P2 are based on 2L treatment. Costs in the 3-state model in P1 are based on 2L treatment; in the 4-state model costs in P1 are based on 2L treatment and costs in P2 are based on post 2L treatment. Transition rates were estimated using a novel multi-state network meta-analysis.

RESULTS : Incremental quality adjusted life-years and incremental costs were both higher in the 4-state model. Higher costs were driven by treatment costs after progression, which was a function of treatment duration with osimertinib before stating chemotherapy. The incremental cost effectiveness ratio increased from $75,000 in the 3-state model to $130,000 in the 4-state model.

CONCLUSIONS : Cost-effectiveness estimates with 4-state models can differ from 3-state models since 3-state models do not explicitly incorporate efficacy from 2L treatments or the duration of multiple post progression treatments. 4-state models should be considered as an alternative model structure, particularly when post progression treatments differ across the 1L evidence base used for evidence synthesis.