OBJECTIVES: Selection of an appropriate modeling structure is a key consideration in economic evaluations.  Factors that influence the choice of modeling structure include data availability and consensus within the clinical community regarding clinically meaningful definitions of treatment response and defined and measureable health states. In the absence of an agreed consensus, selecting a model structure is challenging. We explored two model structures to assess the cost-effectiveness of onabotulinumtoxinA for the treatment of urinary incontinence (UI) due to neurogenic detrusor overactivity (NDO). METHODS: The merits and limitations of a model based on treatment response versus an absolute model structure were considered. In the response model, health states were defined by precise trial outcomes (percent reduction in UI episodes from baseline). In the absolute model, health states were defined by categories of UI episodes/week. In the absence of clinically meaningful cutoffs, we plotted health-related quality of life (HRQoL) scores versus UI episodes to derive meaningful cutoffs for health states based on HRQoL. RESULTS: In the response model, response was defined as a ≥50% reduction in UI episodes. The primary limitation with this approach is the heterogeneity within ‘non-responder’ patients (e.g. patients with either a 0% or a 49% reduction are both ‘non-responders’).  No clear health state cutoffs were observed with respect to HRQoL and UI episodes for the absolute model. Health states were defined on percentile distribution within the trial population: dry, 0-25% [1-14 UI episodes/week], >25%-75% [15-32 UI episodes/week], and >75% [>32 UI episodes/week]. The primary limitations to this approach are the inability to capture health improvement within health states and the lack of clinical relevance for health states based on percentiles. CONCLUSIONS: In the absence of clinical consensus, model structure selection should be a key consideration to capture the true economic value of a therapy.