Modelling the Natural History of Rare Diseases Using Disparate Data Sources: pplication to Patients with Duchenne Muscular Dystrophy

A health technology assessment compares the costs and effects of two or more treatments for a disease. Generally, these will be the standard of care against a novel treatment. The disease in question will often have a number of key phases that a patient progresses through, with different treatment plans, resource requirements and quality of life at each stage. These stages can be represented by a multi-state model. To compare treatments in progressive diseases, a baseline model (termed the natural history) is estimated to represent the progression made by a typical patient under current standard of care, to which different treatments can be compared. Constructing these models for rare diseases can be problematic, since evidence sources are typically much sparser, meaning data from more than one study may need to be combined to fully model disease progression from diagnosis through to death. This thesis investigates and extends current methodology for modelling the natural history of rare progressive diseases using multiple sources of data, which are likely to be heterogeneous. Electronic health records are also evaluated as a potential source of natural history data. Data from a motivating example of Duchenne Muscular Dystrophy, a rare and progressive neuromuscular disease, is used throughout. Lastly, the identified methods are applied to an economic context, with a simulated cost-effectiveness analysis comparing these methods to a previous model framework and highlighting the importance of appropriate model selection to health technology assessments and decision making.