Sensitivity analysis in health economic and pharmacoeconomic studies - An appraisal of the literature

The objective of this study was to analyse the extent of reporting of sensitivity analyses in the health economics, medical and pharmacy literature between journal types and over time. 90 articles were chosen from each of the bodies of literature on health economics, medicine and pharmacy. MEDLINE, EMBASE and International Pharmaceutical Abstracts were searched for English-language economic studies published between 1989 and 1993. The studies chosen for inclusion had to be original articles published in one of the selected journals between January 1989 and December 1993, involving a comparison between drugs, treatments or services, and evaluating both costs and outcomes. 123 articles initially met these criteria; however, 16 were inappropriate, 17 were randomised out, leaving 90 studies (73%) that were used (30 from each literature group). Data were extracted independently by 5 raters using a validated checklist. Inter-rater reliability was assessed by calculating kappa. 53 of the 90 articles (59%) conducted sensitivity analyses. 39 (74%) stated explicitly that a sensitivity analysis was being performed; this was noted in the Methods section of 35 papers (67%). 80% of health economics journals, 70% of medical journals and 20% of pharmacy journals conducted sensitivity analyses. Despite the fact that all published pharmacoeconomic guidelines suggest the use of sensitivity analysis, only 5% of studies between 1989 and 1993 did so. Improvement is required, especially in the pharmacy literature. No time trends in the conduct of sensitivity analyses were detected. However, the sample may not have been sufficient to detect such trends. Pharmacoeconomic guidelines should provide more details on preferred methods of sensitivity analysis and on desired parameters.