OBJECTIVES: Probabilistic sensitivity analysis (PSA) is embedded as a requirement in many pharmacoeconomic guidelines and therefore widely used in practice as a tool for assessing uncertainty around cost-effectiveness outcomes. In current PSAs, many points overlap in high density areas thereby blurring individual points and making it hard to visually assess relative densities. Of course, the general shape of the distribution can be interpreted. We do however believe that the current visualization can be improved; potentially making PSAs more informative. The goal of this study is to illustrate the advantages of more informative PSA results through our novel graphical representation using two real world case studies.

METHODS: Decision analytic models for therapeutic drug monitoring (TDM) of endoxifen and an early HTA assessment of a genetic test were used as case studies. Simulations (50.000) were performed to provide sufficient data to produce the full stochastic representation of the underlying probabilities. Using R, kernel density estimation was performed. With the R package Plotly, relative density plots and cumulative probability contour plots were generated. A color gradient was used to illustrate uncertainty around the point estimate including the probability distribution. An online supplement will be provided as a manual to allow others to use our method.

RESULTS: The TDM model relative density plot showed a very small area with a very high relative density. This effect and the shape of this area was invisible in the traditional PSA representation. For the genetic case, a density plot visually provides limited information. The cumulative probability contour plot is more informative for both models in displaying what areas contributed to the overall probability distribution.

CONCLUSIONS: The relative density plot clearly provided additional information in one of the cases. The contour plots did not provide true additional information but could provide easier visual interpretation of the probability distribution.