The analysis of censored treatment cost data in economic evaluation

This article investigates the way in which the presence of censored cost data in clinical trials should dictate the inferential tests adopted when comparing treatment and nontreatment groups for the purpose of economic evaluation. The authors argue that the techniques of survival analysis are appropriate where censoring is present, and that bias will be imparted if cruder methods are used to analyze cost data, even if that data is drawn from a relevant population. The first section of the article discusses the problem of censoring and survival analysis, while the second examines three methods of dealing with censored cost data and possible biases resulting from them. The third section presents results from actual trial data using the three methods described in the preceding section. Conclusions are presented in section four, where it is argued that these methodological issues are likely to become more important as economists are called upon to evaluate the treatment of chronic conditions using data from clinical trials with finite end points.