OBJECTIVES: To discuss and explore important sources of structural uncertainty in a decision model for drug treatment in Alzheimer’s disease. Progression in Alzheimer’s disease can be modelled on various domains (e.g. cognition, function, behaviour), each of which can be measured by various scales. Defining disease progression and incorporating a treatment effect can therefore be difficult, requiring many assumptions based on little or inconsistent evidence, resulting in a great deal of structural uncertainty. METHODS: We explored the impact of structural assumptions in a published three-state Markov model of treatment for Alzheimer’s disease. This ultimately led to the development of a novel model. Elements of structural uncertainty included (i) defining the model states, (ii) predicting occupancy of the states over time and (iii) allowance for disease progression within a state. RESULTS: Assessment of the step-by-step structural changes to the published model indicated a number of assumptions having a relatively large impact on the cost-effectiveness results, generally resulting in lower costs per quality-adjusted life-year (QALY) associated with the treatments. The cumulative impact of these assumptions was also considerable. However, there is little evidence to inform which assumptions are the most realistic. CONCLUSIONS: Investigation into sources of structural uncertainty has helped to identify which assumptions had the largest impact on the estimated costs per QALY. However, this does not help to reduce the uncertainty in the decision model, but does make the structural uncertainty explicit. Decision-makers are therefore forced to address this type of uncertainty as well as parameter uncertainty. Exploring structural uncertainty also helps to identify gaps in the current evidence base to help understand which assumptions may be the most credible.