The authors justified their selection of the comparators. Fluoxetine was the most commonly prescribed selective serotonin re-uptake inhibitor in the UK, venlafaxine was a typical selective serotonin and noradrenaline re-uptake inhibitor, and amitriptyline was a typical tricyclic antidepressant. The selection of the comparators appears to have been appropriate.
The use of a literature review to identify the sources of data was appropriate and aimed to include all relevant studies. Some key details were given on the inclusion and exclusion criteria used to select these studies. The selection of head-to-head RCTs improves the internal validity of the clinical estimates. Where data were not available, experts’ opinions were used. Extensive information on the derivation of the QALYs was provided, especially for the assessment of utility values and a disease-specific benefit measure was also considered in the sensitivity analysis.
The perspective of the NHS appears to have been adopted although this was not explicitly stated. This viewpoint was indicated by the categories of costs and their sources. Details on the unit costs and quantities of resources used were reported, for most items, and the price year was reported. In general, the economic analysis was presented transparently. As in the clinical analysis, some assumptions were required and made by a Delphi panel of experts for the data on resource consumption.
Analysis and results:
The costs and benefits were appropriately combined using an incremental approach, and the results were clearly presented and discussed. The issue of uncertainty was extensively investigated, although the methodology was not clearly described. The current study was an update of a previous model and some aspects of the analysis were described elsewhere. The authors acknowledged some limitations of their analysis, such as the need for assumptions and the fact that only one RCT comparing venlafaxine with amitriptyline was found.
The analysis appears to have been based on valid methodology, but some assumptions were needed. In general, the authors’ conclusions appear to be valid.