The rationale for the selection of the comparators was clear as the proposed programme was compared with usual care in the study setting.
Limited information was given on the methodological aspects of the clinical side of the study. The approach used to identify data sources was not reported and few details were given. Some data were derived from Chinese databases, but other inputs were based on estimates from other countries, which raised issues on the differences in epidemiological patterns of disease transmission across countries (as acknowledged by the authors). However, no local data were available; the authors tried to overcome this by performing extensive sensitivity analyses. Extensive details of the methods used to calculate benefit measures were reported. Both benefit measures appeared appropriate for capturing the burden of the disease. In particular, DALYs would allow cross-diseases comparisons.
A broad perspective was adopted, so long-term costs of HIV management were taken into account. No formal justification was given for the exclusion of costs associated with HIV-related productivity costs. Costs and quantities associated with the voluntary counselling and testing programme were broken down, which enhanced the transparency of the analysis. All cost items included were presented in detail. Costs were treated deterministically and varied in the sensitivity analyses. Other details, such as the discount rate and price year, were reported, with appropriate conversions to international dollars. The authors highlighted the fact that these costs were representative of a specific area of China (Shandong province) which might be different from other Chinese areas.
Analysis and results:
Projected costs and benefits were synthesised using an incremental approach, which allowed the identification of the optimal intervention using the conventional benchmark of the pro capita GDP. Uncertainty was investigated using a deterministic approach, which considered variations of model inputs singly rather than simultaneously. The results were presented in detail for the two populations examined. The authors acknowledged some limitations of their analysis in the discussion, mainly the lack of local data for some clinical inputs and the poor quality of other estimates. They also stated that behavioural parameters were based on self-reported values that might be biased. The analysis was based on a one-year programme, so the effect of a longer counselling and testing programme was unclear. The results are specific to the local context of China and did not appear to be transferable to other countries.
The study used a conventional cost-effectiveness framework. The authors’ conclusions appear robust.