OBJECTIVES: The NICE technology appraisals programme commissioned the development of a pathways model to explore cost-effectiveness of interventions in non-small cell lung cancer. Pathways models incorporate multiple decision nodes in a treatment sequence. To inform the model structure, we conducted a review to identify methodologies used to evaluate cost-effectiveness of interventions in pathways disease modelling within oncology.

METHODS: A search was designed to identify papers on methodological approaches to oncology pathway modelling. Systematic reviews were included if they provided a critique of methodological approaches. The search terms included pathways models, core/core disease and whole disease models. We also ran a supplementary update review of Lewis et al. to identify studies on methods for evidence synthesis to inform pathways models.

RESULTS: 8 full texts were included on modelling methodology, with best practice recommendations extracted and summarised. Key themes identified included: a preference for patient-level model structures; preference for data coming from a single source; choice of software; terminology; potential for further research into methods for developing pathways models when optimal data is not available. 5were included in the supplementary review of evidence synthesis methods. Methods included: analysis of combined individual patient datasets, separate parametric models at each line of therapy, and multi-state network meta-analysis with flexible survival models.

CONCLUSIONS: Existing methodological papers exploring pathways model development outlines a process for developing pathway models, emphasise the need for individual patient data and data from a single source. However, there is value in further methods development allowing for the inclusion of multiple data sets limited to published data whilst adhering to the processes outlined in the literature. Multi-state network meta-analysis with flexible survival models was most appropriate for evidence synthesis for pathways models. However it may be necessary to conduct separate syntheses at each line of therapy due to data limitations.