OBJECTIVES: Multi-state natural history models are often used to represent the progression of a disease across a patient’s lifetime. These models can be used in a health technology assessment to represent the current standard of care against which a novel treatment can be compared. The construction of these models for rare diseases is problematic, since evidence sources are typically much sparser and more heterogeneous. Ignoring this heterogeneity can lead to biased estimates of disease progression. The objective of this study was to assess the impact of failing to account for study heterogeneity on health economic measures of interest that health technology assessments/decision makers rely on.

METHODS: Four different methods were applied to estimate disease progression within a model for a rare disease called Duchenne Muscular Dystrophy (DMD); one that did not account for study source and three that did. The model structure, costs and utilities were obtained from a published model. A hypothetical treatment cohort was simulated consistent with an example of an existing treatment for DMD. Life years, quality-adjusted life years (QALYs) and incremental cost-effectiveness ratios (ICERs) were calculated.

RESULTS: Disease progression data were available for 3,166 patients with DMD across 25 international studies. ICERs from the different model applications varied significantly; the ICER from the model not accounting for study source was consistently lower (~£2.2 million per QALY) than from the models that did (~£2.4-2.5 million per QALY).

CONCLUSIONS: Adjusting for study source is a key component of natural history modelling for rare diseases. It is likely that there will be heterogeneity between data sources, and models that do not account for this can give inaccurate predictions of times spent in different stages of a disease, which impact the economic modelling and conclusions. This could be the difference between a decision maker funding/reimbursing a novel treatment or not.