outlines the average cost per HIV-infected mother and HIV exposed infant (mother-infant pair) under each PMTCT regimen. Dual ARV prophylaxis with breastfeeding (WHO-Option A) was the least expensive option at 738 USD per mother-infant pair, followed by Sc-HAART with 6 months breastfeeding (801 USD). The higher average cost per HIV-infected mother-infant pair is with Sc-HAART and 18 months breastfeeding (1,024 USD). Details of cost components are available from the authors upon request.
Average cost per HIV-infected pregnant woman-infant pair – base case, USD.
The findings from the analysis in Rwanda show that the fewest new HIV infections in children occur when HIV positive pregnant women receive Sc-HAART coupled with replacement feeding for infants (124 versus 3,987 in the no intervention scenario). The highest number of new HIV infections in children takes place when mothers receive Dual ARV and breastfeed for 12 months (865 new infections).
Across all Sc-HAART options throughout the breastfeeding period, the shortest breastfeeding period (6 months) averted the most HIV infections (286 new infections versus 488 in the case of 18 months breastfeeding). Dual therapy with replacement feeding is associated with 76 fewer vertical HIV infections than Sc-HAART during 18 months of breastfeeding (412 versus 488), but remains inferior to Sc-HAART during 12 months breastfeeding (387 new infections). Infections through long-term breastfeeding overcome the effects of a more effective regime of ARV prophylaxis.
Analysis of HIV-free survival, within the assumptions of the base case, changes the ranking across options. The highest numbers of HIV-uninfected children still alive at 18 months are found with Sc-HAART with 12 months breastfeeding (9,387 children) followed by Sc-HAART with 18 months breastfeeding (9,292). The other Sc-HAART scenarios (Sc-HAART with 6 months breastfeeding and with replacement feeding) still rank higher than Option A.
Net costs (Cost-Savings)
The annual cost of treatment and care per child, according to age range, is provided in . When looking at net costs (savings on future treatment and care from HIV infections averted after subtracting the cost of providing PMTCT services), all options appear cost-savings at different degrees. The highest net costs are reached with Sc-HAART with 6 months breastfeeding (−13,912,837 USD) followed by Sc-HAART and replacement feeding (−12,721,758 USD). The least net costs are reached with Sc-HAART and 18 months breastfeeding (−10,410,727 USD). Since all scenarios are cost-saving, the cost effectiveness ratios of net cost over infections averted are not reported.
Cost of treatment and care components per child per year, USD.
shows the number of HIV infections occurring in children together with the net total cost for each of the PMTCT scenarios.
Number of new HIV infections occurring in children and net total cost, USD.
shows the number of children uninfected with HIV and alive at 18 months together with the net total cost for each of the PMTCT scenarios.
HIV-free survival (at 18 months) and net total cost, USD.
Incremental cost – effectiveness analysis
When ranking scenarios based on the number of HIV infections, the analysis showed two alternatives averting more HIV infections and saving more money than all the other scenarios (): Sc-HAART with 6 months breastfeeding and Sc-HAART with replacement feeding. Sc-HAART with replacement feeding averted 163 additional HIV infections than Sc-HAART with 6 months breastfeeding for an incremental cost of 1,191,079 USD and Incremental Cost Effectiveness Ratio (ICER) of 7,319 USD.
When ordering scenarios based on the number of children alive and uninfected, ICER analysis shows that there are only two alternatives that are not dominated (): Sc-HAART with 6 months breastfeeding and Sc-HAART with 12 months breastfeeding. Sc-HAART with 12 months breastfeeding allows 170 additional children alive and HIV-uninfected at 18 months for an incremental cost of the program of 2,017,310 USD. ICER equals to 11,882 USD.
Given the uncertainty embedded in the input values of the base case scenario, we conducted a one-way sensitivity analyses for key variables of the model.
To investigate the sensitivity of findings to breastfeeding practices, ranges were used for both breastfeeding duration and for the percentage of exclusive breastfeeding. As per recommendations, the base case uses 6 months as actual duration of breastfeeding (for options other than Sc-HAART for 12 and 18 months). However, anecdotal reports from health providers in Rwanda suggest that breastfeeding duration for HIV-infected women may be higher than the recommended 6 months due to lack of financial means. Therefore, we took 18 months as the low range (or worst scenario) in terms of actual breastfeeding duration in cases when health providers recommend 6 months to the mothers. For the percentage of women practicing exclusive breastfeeding during the first 6 months, the base case uses 38% as per the rate in the general population 
. In reality, this percentage could be lower (we took a token 50% decrease from the base case, or 19%, in the absence of data) or higher. The proportion of mothers who practice exclusive breastfeeding may be higher since HIV-infected women are the more aware of the HIV transmission risk. A recent study from Uganda shows that up to 92% of women on ART exclusively breastfed their infants for a median duration of 4.0 months and stopped breastfeeding at a median age of 5.0 months 
Even in the most favorable of the breastfeeding scenarios (with 92% of mothers exclusively breastfeeding for the recommended 6 months), the highest number of HIV infections are averted by Sc-HAART with replacement feeding. The second best option in terms of number of infections averted remains Sc-HAART with 6 months breastfeeding. As expected, an extra infection averted through Sc-HAART with replacement feeding against Sc-HAART with 6 months' breastfeeding would cost more (an extra 14,229 USD versus 7,319 USD in the base case). In the worst case scenario for breastfeeding practices, when the percentage of mixed feeding and the actual breastfeeding duration increase to 18 months (independently of the duration of HAART), the number of HIV infections averted with the breastfeeding options (other than Sc-HAART for 18 months) decreases substantially. Therefore, Sc-HAART with replacement feeding becomes even more favorable and this scenario dominates all the others (the highest number of HIV infections averted for saving additional money).
When looking at HIV-free survival as an outcome of interest, we find that across the range of values for breastfeeding practices, Sc-HAART with 12 months breastfeeding still allows more HIV uninfected children alive at 18 months than Sc-HAART with 6 months breastfeeding and all other scenarios are dominated, The extra cost per child alive and uninfected is of 11,780 USD in the worst case scenario for breastfeeding practices and 10,774 USD in the most favorable breastfeeding scenario.
We conclude that our ranking across PMTCT scenarios is robust, or not sensitive, to breastfeeding practices.
Costs of ARV, replacement feeding, and laboratory tests
Acknowledging that the cost of inputs may vary over time, we explored the variability of our results for potential 50% cost increases or decreases in the inputs that account for the highest proportion of cost, namely: ARV cost, laboratory costs and the cost of replacement feeding. As expected, if prices of ARV and laboratory tests increase, the replacement feeding options become more favourable. If prices of ARV and laboratory tests decrease, Sc-HAART scenarios become more cost-effective since they become cheaper. The opposite occurs when cost of replacement feeding increases. When ARV costs increase, savings also increase.
When looking at HIV infections averted, we found that even with varying input costs, Sc-HAART with replacement feeding and Sc-HAART with 6 months breastfeeding remain highest in the ranking across PMTCT options. When all costs increase by 50%, Sc-HAART with 6 months breastfeeding still dominates all scenarios, with the exception of Sc-HAART and replacement feeding, which allows more HIV infections averted for an extra cost of 11,643 USD per extra HIV infection averted. When all costs decrease by 50%, the same is true and the extra cost per extra HIV infection averted becomes as low as 2,995 USD.
When looking at HIV-free survival as an outcome of interest, Sc-HAART with 12 months breastfeeding and Sc-HAART with 6 months breastfeeding remain the dominant scenarios across the range of cost variations (+/−50%). The ICER for Sc-HAART with 12 months breastfeeding versus Sc-HAART with 6 months breastfeeding increases to 16,590 USD when costs increase and it decreases to 7,175 USD when costs decrease.
Our findings are therefore not sensitive to potential variations (estimated at a maximum of ±50%) in input costs.
Proportion of women eligible for ART
With respect to the new eligibility criteria of a CD4 threshold of 500, data concerning the proportion of eligible women in Rwanda were not available. Thus we assumed that 73% of women in PMTCT programs may be eligible for ART, as reported in the literature 
. Given the uncertainty of this value in the Rwanda context, we explored the effect on the results if only 36% of women were eligible to ART (CD4 less than 500). This is a 50% decrease comparing to the base case value.
With 36% of women eligible to ART, we found that when looking at HIV infections averted as an outcome of interest, Sc-HAART with 6 months breastfeeding still dominates all scenarios, with the exception of Sc-HAART and replacement feeding, which allows more HIV infections averted for an extra cost of 3,415 USD per extra HIV infection averted. When looking at HIV-free survival as an outcome of interest, Sc-HAART with 12 months breastfeeding and Sc-HAART with 6 months breastfeeding remain the dominant scenarios. Sc-HAART with 12 months breastfeeding allows more HIV uninfected children alive at 18 months for an increased extra cost of 15,639 USD per uninfected child alive at 18 months. Our results are thus not sensitive to plausible changes in the proportion of eligible women in the cohort of HIV-positive pregnant women.
HIV transmission rates for each PMTCT option
Ranges for the HIV transmission rates, as per the most recent data available in the literature, are provided in . The highest and lowest values were used. Analysis for the lower bound is the more important considering that 500 CD4 was used as a cut off for ART eligibility in this study (as per Rwanda guidelines) and this may reduce the estimates of transmission as reported in the literature for a 350 CD4 cut off. As expected, there is a higher number of new HIV infections in situations with a higher HIV transmission probability. However, the overall ranking across scenarios does not change.
For HIV infections averted as an outcome of interest, when the low transmission rates are used, Sc-HAART with 6 months breastfeeding dominates all options with the exception of Sc-HAART with replacement feeding. However, the extra cost for an extra infection averted with Sc-HAART and replacement feeding versus Sc-HAART with 6 months breastfeeding increases to 38,712 USD. When the high transmission rates are used, the extra cost for an extra infection averted with Sc-HAART with replacement feeding versus Sc-HAART with 6 months breastfeeding decreases to 2,951 USD.
For HIV-free survival as an outcome of interest, Sc-HAART with 12 months breastfeeding and Sc-HAART with 6 months breastfeeding still dominate all other scenarios across the ranges of HIV transmission rates. The ICER for Sc-HAART with 12 months breastfeeding versus 6 months breastfeeding decreases to 7,542 USD when low HIV transmission rates are used and it increases to 94,544 USD when high transmission rates are used.
Mortality rates at 18 months for HIV exposed uninfected children
In the absence of national data for Rwanda, the Relative Risks (RR) of mortality rates for children breastfeeding for only 6 months or being formula fed were drawn from the literature (see ). However, several assumptions had to be made. For example, as no study was found on the RR for mortality for 6 months breastfeeding versus 12 months or more, a RR of 1.5 was taken as the middle point between the worst case of RR of 3 (when weaning at 4 months 
) and the best scenario in which there would be no difference in survival outcomes. In the case of survival rates for replacement feeding, we chose a conservative approach for the best case. Rather than using a RR of 0.5 as reported in the PIH program 
, we assumed that survival rates for such a program may be less optimal in the context of a national scale up, and therefore selected a best case, low scenario, RR equal to 1.
When redoing the analysis with this lower range (RR
1), Sc-HAART with replacement feeding allows the highest number of children alive and uninfected at 18 months, for an extra cost per child of 7,745 USD compared to Sc-HAART with 6 months breastfeeding. All other options are dominated (see ).
Figure 3 HIV-free survival (at 18 months) and net total cost, USD – sensitivity analysis – low range, RR=1.
When looking at the threshold at which the rankings of scenarios change, we found that Sc-HAART with replacement feeding becomes the scenario with highest HIV-free survival attainments if RR is 1.4 or lower. This would require a setting where replacement feeding with an available, nutritionally adequate and safe diet could be provided, such as in the model described in our study. For Sc-HAART with 6 months breastfeeding to become the favourite option in terms of HIV-free survival attainments, RR should be of 1.1 or below. This RR of mortality rate at 18 months might be the actual rate in some context, depending from the conditions of mothers/infants (and the support provided to them) during the weaning period and in the months that follow. When the analysis is re-done using the higher RR ranges (), Sc-HAART with 12 months breastfeeding allows the highest number of children alive and HIV uninfected at 18 months, for an extra cost per child of 2,089 USD compared to Sc-HAART with 6 months breastfeeding, dominating all other options.
We conclude that when using HIV-free survival as an outcome of interest, our results are sensitive to the assumptions used for the mortality rates among HIV exposed uninfected children, and therefore to the conditions (adequacy, safety) in which early weaning (after 6 months breastfeeding) or replacement feeding would take place.