Sensitivity analysis in economic evaluation: a review of published studies

A structured methodological review of journal articles published in 1992 was undertaken to determine whether recently published economic evaluation studies deal systematically and comprehensively with uncertainty. Ninety three journal articles were identified from a range of searches including a computerised search of the MEDLINE CD-Rom database. Articles were reviewed to determine how they had handled uncertainty in: a) data sources; b) generalisability; c) extrapolation; and d) analytic method. Articles were subsequently assessed to determine how they had represented this uncertainty in terms of the overall results of their analysis. Finally, studies were rated on the basis of their overall performance with respect to dealing systematically and comprehensively with uncertainty. Despite the numerous books and articles devoted to the appropriate methods to be employed by analysts conducting economic evaluation, 22 (24%) studies failed to consider uncertainty at all and 35 (38%) studies employed sensitivity analysis in a manner judged as inadequate. In all, 36 (39%) studies were judged to have given at least an adequate account of uncertainty with 13 (14%) of those judged to have provided a good account of uncertainty. Such disappointing results may reflect a general lack of detail in much of the methods literature concerning how sensitivity analysis should be applied and how results should be presented. Journal editors and readers of economic evaluation articles should acquaint themselves with the methods for handling uncertainty in order that they can critically evaluate the extent to which authors have allowed for uncertainties inherent in their analysis.