BACKGROUND: Established methods exist for evaluating the effects of uncertainty around the model structure and parameters on the results generated by traditional cost-effectiveness analyses (CEAs) and include one-way and probabilistic sensitivity analyses (SAs). In contrast to the primary outcome of a traditional CEA—the ICER— the primary outcomes of a threshold CEA conducted for a product early in development include (1) the value-based price opportunity given a hypothetical or target product profile and (2) the magnitude of effect required to justify a target price. Because the outputs of a threshold model pertain to a new drug or indication where little or no data have been collected, and because the outputs are multiple, representing the set of product attributes, including price, that will define drug value, there is a need to explore the sensitivity of the results to factors that go beyond uncertainty. In analyses that generate potential value-based price or product attribute levels, new methods and applications of SA are required.
METHODS: We present example one-way and probabilistic SAs, highlighting problems in interpretation that arise when traditional sensitivity analyses are applied to threshold models. We propose alternative SA methods and analyses and present interpretations of results. a Pricing Contribution Diagram is presented as a means of characterizing the extent to which each product attribute (efficacy, safety, tolerability, quality of life, position in care pathway) influences the value-based price opportunity. Probabilistic SAs are presented to examine the relationship between price (value-based and target) and individual product attributes, and the influence of uncertainty in other model inputs.
CONCLUSIONS: Traditional methods of conducting SA are insufficient when applied to the threshold application of CEA. Instead, SAs specific to threshold models supporting decisions regarding early stage development should be employed.
