EM4 Recommendations for Handling Uncertainty in Economic Evaluation: A Targeted Review of Pharmacoeconomic Guidelines


      The appropriate handling of uncertainty is an essential element of economic evaluation of healthcare interventions. Accepted methods include deterministic and probabilistic analyses to characterise parametric uncertainty, and scenario analyses to test uncertainty propagated by methodological and structural assumptions. This targeted review examined which methods for handling uncertainty are recommended in pharmacoeconomic guidelines around the world.


      Pharmacoeconomic guidelines were identified from HTA agency websites, PubMed and Google Scholar, and manual searches of references from key publications. Inclusion criteria were open access, inclusion of recommendations for handling uncertainty and publication in a language accessible to the reviewers. Two reviewers extracted data on the guidelines’ recommendation for the type of sensitivity analysis and use of technical tools (e.g. Tornado diagrams, scatter plots, CEAC).


      Forty-three national or supranational pharmacoeconomic guidelines passed the inclusion criteria. One-way deterministic sensitivity analysis (DSA) was requested in thirty-five (81%) guidelines. Notable exceptions included CADTH (Canada), which recommended against the use of DSA, and HAS (France) which considered DSA of limited use compared to probabilistic methods. Probabilistic sensitivity analysis (PSA) was compulsory in twenty-nine (67%) guidelines and a further five (12%) included it as an optional analysis. Tornado diagrams were specified in fifteen (43%) guidelines which required a DSA. The most requested tools for reporting PSA were acceptability curves (56%) and scatterplots on the cost-effectiveness plane (44%). Value of information analysis based on PSA results was recommended in ten (29%) publications. Scenario analyses to examine the impact of structural assumptions were recommended in fourteen (33%) guidelines.


      Both deterministic and probabilistic methods for characterising uncertainty were endorsed by most published guidelines, with accepted tools such as Tornado diagrams, scatter diagrams and acceptability curves commonly requested. When planning global cost-effectiveness models, manufacturers should consider additional analyses required by some authorities, such as value-of-information analyses and scenario analyses for structural assumptions.