Interventions:
The rationale for the selection of the comparators was clear as the objective was to assess the net value of aripiprazole as an addition to standard care, which was the background comparator. The dose and duration of treatment were reported.
Effectiveness/benefits:
The treatment effect and patients' characteristics were from two head-to-head randomised controlled trials, the results of which were pooled. This should guarantee high internal validity, but little information on the two trials was given. The time horizon of the additional treatment part of these two trials was only six weeks, and a longer follow-up would have been interesting. A disease-specific benefit measure and a generic outcome measure (QALYs) were appropriately used. The utility data were from a published study that used a valid approach to elicit the preferences.
Costs:
The cost categories and the data sources were consistent with the stated perspective. Extensive details of the unit costs and resource quantities were provided, enhancing the transparency of the analysis. The author stated that some cost items, such as adverse events and psychiatric consultations, were excluded, but no justification for their exclusion was provided. Reflation exercises will be possible as the price year was clearly reported. The economic data were treated deterministically and, except for the price of aripiprazole, the economic inputs were not varied in the sensitivity analyses.
Analysis and results:
Average cost-effectiveness and cost-utility ratios were calculated to synthesise the costs and benefits of the two strategies, but the projected costs and QALYs were not clearly reported. The method used to calculate the incremental cost-effectiveness ratios appears to have been incorrect. It should be the difference in costs divided by the difference in effectiveness, not the difference in average cost-effectiveness ratios. The author did not provide a criterion (cost-effectiveness threshold) to identify the best treatment strategy. The uncertainty was partly investigated focusing on two model inputs, which were varied separately. A more comprehensive assessment of uncertainty would have improved the validity of the results. The findings appear to be specific to the author’s setting and their transferability to other settings was not discussed.
Concluding remarks:
The study was well presented, but an apparent error in the calculation of the results might affect the validity of the author’s conclusions.
