Interventions:
The intervention was well explained and the comparator was clear. The interventions compared were justified because the study retrospectively analysed the cost-effectiveness of the implementation of a health programme.
Effectiveness/benefits:
The total number of infants born to HIV-positive women was only 15. HIV was transmitted to one infant. The small sample size suggests the existence of uncertainty around the transmission rate estimate. The methods used to identify the literature used to estimate the clinical effectiveness estimates were not reported. The main clinical parameters of transmission rates were not well reported. Termination of pregnancy was listed as an intervention to reduce the risk of vertical transmission. This was not included in the calculation of the measure of benefit, without justification.
Life-years is an adequate measure of benefit when considering the mortality associated with HIV/AIDS. However, it does not fully capture the health benefits.
Costs:
All of the relevant costs consistent with the study perspective appear to have been included in the analysis. The sources of the resource use and the unit costs, and the data themselves, were well reported. The cost estimates were all relevant to the study population and setting. Discounting was appropriately conducted.
Analysis and results:
The cost results were well reported but the benefit results could have been clearer. A sensitivity analysis on antiretroviral effectiveness rates was conducted, but it was not explained how the therapy effectiveness rates fitted into the model. The uncertainty surrounding the HIV mother-to-child transmission rate was evaluated through varying the estimate in a sensitivity analysis. However, the small number of HIV-positive mothers and infants in the programme suggests that a greater range of the estimate could have been analysed. There was no indication that any analysis of uncertainty was conducted on the cost estimates in the model.
Concluding remarks:
The clinical estimates used in the model were not well reported and there is uncertainty in the results. Nevertheless, the authors' conclusions are broadly appropriate.