OBJECTIVES: Semi-Markov models (SMM) are criticised in National Institute for Health and Care Excellence guidance (Technical Support Document 19) for their modelling of post-progression survival (PPS), which is often biased by “early progressors”. To avoid this modelling of PPS, a ‘novel’ estimation approach is defined in which parameters are estimated by minimising the area between the SMM-predicted overall survival (OS) and progression-free survival (PFS) curves and the corresponding OS and PFS Kaplan-Meier curves, rather than fitting over the individual state transitions.

METHODS: SMMs were fitted to the SQUIRE trial data in non-small-cell lung cancer (obtained from Project Data Sphere). Two methods were used: a ‘standard’ approach whereby the maximum likelihood was utilised for the individual transitions and the best-fitting parametric model selected based on AIC/BIC, and a ‘novel’ approach which directly optimises the area between the estimated PFS and OS distributions and the PFS and OS Kaplan-Meier curves. Probabilistic sensitivity analyses were run to assess parameter uncertainty.

RESULTS: The ‘novel’ approach resulted in better estimates to the restricted mean survival time in 10 of the 12 parametric distributions compared to the ‘standard’ approach. The uncertainty associated with the ‘novel’ approach was slightly larger.

CONCLUSIONS: An approach is defined which successfully overcomes the drawback of SMMs described in Technical Support Document 19 related to PPS modelling. This enhances the value of SMMs for oncology models and can easily be extended to SMMs with more health states.