Interventions:
The authors justified their selection of the strategies, which were appropriately chosen.
Effectiveness/benefits:
The bulk of the evidence came from a published study. The methods and characteristics of this study were not presented, except for the sample size, but its randomised design should have ensured a high internal validity. The benefit measure was derived from the RCT and was disease-specific, which means it cannot be directly compared with the benefits of other health care interventions.
Costs:
The analysis of costs was consistent with the perspective; only those costs borne by the hospital were included. The analysis did not consider the costs of medication preparation, distribution, and administration, which could have been relevant. Also, the costs of severe adverse events were excluded, but a threshold analysis showed that these would have to have been unrealistically high to alter the cost-effectiveness results. The sources of costs were reported, but no details were given of the case-control study that was used to derive the costs of invasive fungal infections, which might limit the transparency of the economic analysis. Appropriate alternative assumptions were made in the sensitivity analysis. The price year was reported, which will facilitate reflation exercises for other time periods.
Analysis and results:
The costs and benefits were analysed in an incremental analysis, which was appropriate to show the dominance of one treatment over the other. The issue of uncertainty was extensively investigated using valid approaches, which considered both individual inputs and the overall uncertainty in the model. The authors stated that a conservative approach was taken to bias the study against posaconazole, especially in the sensitivity analysis. They acknowledged a number of limitations, including the fact that cost savings depended on the rates of invasive fungal infections, which might vary across institutions.
Concluding remarks:
This study was based on valid methodology, but some sources of data were not presented in detail. The authors’ conclusions appear to be robust due to the appropriate investigation of uncertainty.