Missing Data in Trial-Based Cost-Effectiveness Analysis: The Journey Continues

Missing data are common in trial-based cost-effectiveness analyses (CEAs) and often lead to biased estimates. Earlier reviews suggested that the use of appropriate statistical approaches to handle the missing data in trial-based CEAs has been patchy. Since then, several methodological guidelines and tutorials in this area have been published, but it remains unclear whether these have improved practice. This paper provides a contemporary picture of missing data methods used in trial-based CEAs conducted in the UK and investigates the extent to which published guidance has permeated practice. We reviewed trial-based CEAs published in the Health Technology Assessment journal between 2022 and 2024, and 63 studies were identified. Missing data remains pervasive in trial-based CEAs; the median proportion of individuals with complete cost-effectiveness data was 61%. We found an increase in the adoption of multiple imputation in the primary analysis, but the use of complete cases remains high. About 60% of studies conducted sensitivity analyses around the missing data approach, but only 10% explored departures from the missing-at-random assumption. Despite the wide availability of methodological guidance, tutorials and software tools, the adoption of appropriate missing data methods in trial-based CEAs remains a long and arduous journey. We recommend a minimum set of items to be routinely reported in trial-based CEAs, which will help improve transparency and spur better practices: (1) the proportion of complete data by treatment arm and cost-effectiveness endpoint, (2) reasons for missingness, (3) the assumption about the missing data mechanism and its justification, (4) sufficient detail on the approach to address missing data to enable replication and (5) the robustness of the findings to alternative missing data assumptions.