The findings in this paper focus on two topics: (i) the cost-effectiveness of IPTi at averting anaemia and (ii) the influence of using ATP as the measure of effect as opposed to the more commonly used ITT.
Independently of the type of analysis used, this study assessed that the average cost of a first or only episode of moderate anaemia averted during infancy/early childhood is lower than US$13. Considering the high burden associated with anaemia in the country, this can be considered as a very valuable intervention. Specifically, the cost per case of anaemia averted in the ITT analysis was, on average, US$12.88. It is difficult to compare this finding to others which show an effect on anaemia as so few studies have looked at the cost-effectiveness of averting anaemia cases; and those which have, tend to use DALYs, not episodes, as their unit of effectiveness. The only other study the authors have identified using a similar approach to this study was an economic evaluation of IPT in schoolchildren undertaken in Kenya, in which the cost per case of anaemia averted was calculated at US$29.84 (2006 US$) .
Using both types of results, ITT and ATP, can give important information to policy makers. The main finding of this study is that although IPTp-SP is inexpensive and efficacious in reducing moderate anaemia, its cost-effectiveness worsens when drop-out is considered. This result may be intuitive but it is really important in a setting like Lambaréné where drop-out is primarily caused by migration and where any type of health intervention is likely to face the same challenge. The difference between ITT and ATP allows a comparison of the cost-effectiveness of IPTi-SP in two scenarios: one in which several of the initial recipients of the intervention move elsewhere and do not complete the programme (ITT); another, in which a smaller number of recipients complete the prevention programme (ATP). This difference has both cost and health system/organizational repercussion. The more cost-effective between the two was the ATP, meaning that total coverage and total adherence constitute the best situation. However, in a context of migration, reaching complete coverage and adherence is likely to imply high costs.
In this study, it was considered more appropriate to report ICERs in terms of the number of episodes of anaemia averted rather than in terms of DALYs averted. The magnitude of DALYs averted is particularly sensitive to long-term and severe disabilities and to the case fatality rate of a disease . IPTi results in Gabon pointed to a reduction in non-severe cases of anaemia, which are likely to lead neither to long-term, severe disabilities nor to death. In addition, even if IPTi would have shown reductions of severe anaemia, which is associated to death, there is no evidence in Gabon that severe anaemia in children leads to death . Without a sufficiently robust estimate of the anaemia case fatality rate, any DALY calculation is debatable.
IPTi was more cost-effective at preventing moderate anaemia when ATP results were used to measure both efficacy (protective efficacy of IPTi as outcome of the trial) and effectiveness (episodes of anaemia averted due to IPTi). ICERs in terms of intervention costs divided by protective efficacy of IPTi are not reported. However, when analysing efficacy, ATP intervention costs were lower than ITT intervention costs because the ATP population was a sub-group of the ITT population and, therefore, fewer infants were administered with the prevention. In addition, ATP efficacy was higher than ITT efficacy. Moving from efficacy to effectiveness, effectiveness was, on average, higher in the ITT than in the ATP population (ITT cases averted were 16.45; ATP cases averted were 13.73). This was due to the higher number of infants included in the ITT compared to the ATP population. ITT however, still remained less cost-effective than ATP due to the higher total intervention costs.
The threshold analysis was conducted to explore the extent that two influential ICER parameters could change in order for the cost-effectiveness of the ATP population to equal the more commonly used ITT population. Findings suggested that intervention costs should rise from 118.38 to 134.00 US$ for the ICER of the ATP analysis to equal the ITT one. Since both fixed unit costs and target population (594 infants) were assumed to be identical across both types of analysis, the variable that made the difference in determining total intervention costs was the number of infants actually administered with the three doses. For the ATP ICER to reach the ITT ICER, 435 infants instead of 315 would need to take all three doses in order for the intervention costs to rise up to 134.00 US$. However, the rise in the number of infants taking three doses would also have a positive effect on the protective efficacy of the intervention, thus affecting the cost-effectiveness ratio at the denominator.
The use of ATP efficacy reduced the uncertainty existing in the cost-effectiveness analysis, since ATP efficacy was statistically significant. This reduction in uncertainty was visible with reference to the minimum and the maximum value of the cost-effectiveness ratios (Table 2). Both in this analysis and in the one presented in Conteh et al , ICERs are based on borderline statistically significant and statistically non-significant primary outcomes from the clinical trial in Gabon . There is some debate about the value of undertaking economic analysis based on non-significant clinical results. In accordance with Johnston et al , the authors believe it is worth performing economic evaluation with such clinical results. Cost minimization, where cost differences are reported without the calculation of an ICER, was traditionally used as response to non-significant efficacy differences of a trial. However, cost minimization is often inappropriate as it assumes the trial was powered to show equivalence of effects, which is rarely the case and it was not in IPTi trials . Johnston et al  advocate that cost-effectiveness ratios should be calculated even in absence of significance of efficacy, but uncertainty surrounding ICERs needs to be quantified and highlighted.
While randomized controlled trials (RCTs) provide crucial information on the efficacy of an intervention (represented by the protective efficacy of IPTi in this study), they do not always provide clear information on the effectiveness (represented here by the actual number of moderate anaemia cases averted) as they are conducted under atypical conditions. However, RCTs can give some signs of what coverage might be in a real world situation. As to be expected, the health benefits that infants get from the three doses is higher than the benefits from fewer doses, however, it is evident that in Gabon adherence to three doses of IPTi spread out over 15 months will prove a challenge.
In Gabon, vaccination coverage, represented by DTP3 coverage, scheduled at 14 weeks was 14% in 1983, reached a maximum level of 70% in 1995 and dropped to 45% in 2009, ranking Gabon as the country with the fourth lowest DTP3 coverage in the African region (after Chad, Equatorial Guinea and Somalia) . While migration became a relevant phenomenon from the early 1980s across Gabon and could have had an impact, the little information available on this topic does not explain the changes in coverage entirely . In addition to migration, the fluctuating EPI coverage and the lack of the demand for vaccinations has been associated with side effects, attitudes towards vaccinations, limited access to providers and bad previous experiences at mother and child clinics. Gysels et al reported that in Gabon 80% (184/231) of respondents interviewed about their use of EPI services and perceptions of IPTi claimed that shame related to having sick or malnourished infants, having too many children, or being a young mother, were common reasons for avoiding routine clinic visits . All these factors that inhibited demand for EPI may well have played a role in the fall in uptake of IPTi.
Although Conteh et al , Hutton et al  and this study are economic evaluations based on the same study trial, the aim of this study was different and subsequently the methodology followed differs. In Conteh et al and in Hutton et al the dropout from the intervention was calculated starting from the intake of the first dose of SP. In this study, in order to focus on the differences between ITT and ATP definitions of efficacy, the drop-out was calculated starting from one step prior to this. If IPTi was to be introduced, mothers would be sensitized about the intervention while in the health facility having just given birth, and would then be expected to start IPTi on a later occasion in line with their infant's vaccination schedule. To reflect this in our analysis, the drop-out rate was considered to start from when infants were initially recruited for the trial in the health facility having just given birth, three months before the administration of the first IPTi dose.
The IPTi Gabon trial was the only trial in which coverage dropped to such an extent between the enrolment and the first dose. Considering coverage loss from the target population to the first dose is important from a policy perspective. If preventive health interventions to infants, at local level, are planned to target newborns, such plans should take into account the potential migration of the target population. More research is needed to explore the migration patterns in Gabon and how these can shape an individual's treatment seeking behaviour.
Cost-effectiveness is an important tool when it comes to assessing an intervention for possible implementation. IPTi-SP has shown to be an additional malaria control tool of value  and has, consequently, recently been added to the WHO's malaria control instrument portfolio . However, IPTi-SP is recognized to be of use in particular settings, i.e. in areas of high endemicity, stable transmission and low SP resistance. For a variety of reasons, such as the efficacy of IPTi bordering on significance [21, 46] and high levels of SP resistance , IPTi-SP will most likely not be considered as an intervention of high priority in many countries. Before such country-level decisions are made, it should be noted that IPTi-SP can have positive effects in addition to averting malaria. Seasonal IPTi in areas of unstable transmission is under evaluation  and recent modelling has predicted IPTi can be cost effective even in settings of low transmission, seasonality, using short-acting and more expensive drugs and in conditions of increasing drug resistance . In addition to averting malaria, the cost-effectiveness of IPTi at averting anaemia has been explored in this paper, and other evidence exists to suggest IPTi can have an impact on all cause admissions  and increase EPI vaccine uptake . It is important therefore to assess any intervention, not only on the impact it has on the primary endpoint of a randomized controlled trial, but to also consider the often difficult to measure, wider, health benefits to the health system and the individual.