The interventions were chosen as the most commonly prescribed treatments in the authors’ setting. They were appropriately described.
No systematic review of the literature was reported and the sources searched, the inclusion and exclusion criteria, and the details of the selected studies, such as their design and population, were not reported. It is not clear if the best available evidence was used. The use of the clinical evidence was inconsistent: the three drugs were not compared in one analysis because no three-way head-to-head trials were found, but the effectiveness of fluvoxamine was assumed to be the same as that of citalopram because there were no escitalopram versus fluvoxamine trials. The remission rate was a disease-specific measure that did not reflect the impact of the interventions on a patients’ quality of life. It also does not allow cross-disease comparisons and did not take account of adverse events. It was not clear that the treatment period of six months was adequate to capture the differential health effects of the drugs. However, including adverse events in the measure of benefit and extending the time horizon could make escitalopram more cost-effective, based on the data in the model.
The costs included in the analysis reflected the perspective stated. The unit costs and resource quantities were reported separately, which enhances the transparency of the analysis. The effect of increasing the time horizon on the cost difference between the drugs is not clear. The resource use was based on experts’ opinion, but extensive sensitivity analysis was conducted on these estimates. The sources for the unit costs were reported and appear to have been appropriate for the study setting. The currency and the price year were reported, which will aid future reflation exercises.
Analysis and results:
A decision analytic model was used to synthesise all the available evidence. The methods used and the modelling assumptions were reported, with a diagram of the model. The parameter uncertainty was investigated thoroughly using a deterministic and a probabilistic approach. The results of the base case, as well as the sensitivity analyses were reported adequately. The authors compared their results with those of previous studies and highlighted the possible reasons for differences. They reported the limitations of their study and the main ones were the use of effectiveness data from studies conducted in European settings and the lack of head-to-head comparisons of escitalopram with fluvoxamine.
The methods had some limitations, but it appears that these should not affect the conclusions.