OBJECTIVES: Treatment switching commonly occurs in the pivotal HTA evidence for advanced or metastatic cancer treatments submitted to reimbursement agencies. Simple approaches, such as Intention-to-treat (ITT) analysis, have typically been applied to data with treatment switching, despite simulation studies showing these to drastically underestimate the underlying treatment effect. Therefore, before these studies are included in secondary analyses, the data must be re-analysed appropriately. When only summary data are available, individual patient level data can be reconstructed using a simulation approach. Given patients switch on disease progression, their progression time is assumed equivalent to their switch time. Simulating this effectively requires an illness-death modelling framework; the process of which is the aim of this research. METHODS: An example was used, where Kaplan-Meier curves for all three transition rates were available. The coordinates were extracted digitally from these scanned survival curves, and used to model the times for each transition. Many datasets were created, where the times for the transitions were simulated from the respective models. ITT summary statistics were calculated for each dataset; then averaged over. Examples with increasingly less information on which to estimate the transition rates were also systematically investigated. RESULTS: When information on transition rates is available, the process is easy to implement; giving data that are, on average, broadly representative of the original dataset with median survival times and overall survival hazard ratio differing by less than 10% and 0.05 respectively. As the information becomes more limited, the process requires additional assumptions, and ultimately may not be feasible. CONCLUSIONS: Using an illness-death modelling framework to simulate individual patient level data is affected by the information available to the analyst. However, this approach is important, when addressing treatment switching where only summary data are available, as the relationship between time to progression and overall survival is modelled correctly.