In studies where IPTi was shown to have a statistically significant impact on reducing malaria, it was cost effective in all sites with all drugs and highly cost effective in all but one site that used MQ. As mentioned previously, the thresholds that have been used in the literature to determine highly cost effective interventions and cost-effectiveness interventions vary. In this analysis using the most conservative cut-off points, IPTi is highly cost effective in the majority of studies. Had we chosen the WHO threshold of under 1× Gross Domestic Product per capita, all of the studies that had a statistically significant impact on malaria would have been considered well within the highly cost effective range 
Although not part of this analysis, if we were to add the benefits of the additional health gains and subsequent cost savings from averting anaemia and those associated with averting the non-malaria admissions included in ‘all-cause’ hospitalisations (here we included only hospital admissions with parasitaemia), the ICERS would be even more cost effective. For example, IPTi with SP was not seen to have a statistically significant impact on clinical malaria in the trial in Gabon. This was due to a number of reasons, including a steady decline in the malaria incidence in Lambaréné area over the past decade (unpublished data), the high mobility of the local population and a study design with a close-knit passive and active follow-up system that led to the creation of an outstandingly healthy study cohort 
. However, Lambaréné did show a 26% (0%, 45%) PE against moderate anaemia in the first year of life, the benefits of which are not measured in this analysis.
Cost effectiveness analysis aims to inform policy makers on the cost-effectiveness associated with different interventions when decisions have to be made about where to allocate limited funds. However, caution should be exercised when comparing the cost-effectiveness of different malaria control strategies 
as there needs to be an understanding of site specific epidemiological and health system characteristics, the costing perspective and how different malaria control strategies complement and/or substitute one another. With this in mind, delivery cost of IPTi was between USD 0.13 (per dose of SP in southern Tanzanian and Ghana) to USD 1.92 (per dose of CD at 3 days each dose in northern Tanzania). Other malaria prevention strategies have reported annual costs (also adjusted to USD 2007) of providing insecticide treated nets (ITNs) of USD 1.40–USD 3.85 
, USD 3.42–USD 5.83 for indoor residual household spraying 
, USD 1.94 to deliver IPT to school children (3 doses, SP & AQ3) 
, and USD 2.60 when delivering a full course of IPT to pregnant women (2 doses, IPTp-SP) via community care and USD2.30 via health centres 
The incremental benefit of IPTi in addition to ITN use needs to be explored further 
. In the sites included in this analysis ITN ownership and use varied. For example in western Kenya, ITNs were provided alongside the timing of IPTi, thus the PE of IPTi was in the context of high ITN use 
, whereas in Manhica ITN use was zero at the time of the study 
The potential impact of IPTi on the EPI needs careful consideration: will it overburden EPI activities and lead to inequities 
or conversely will the additional benefits of IPTi provide extra resources and momentum that will strengthen the EPI and increase vaccination uptake? The level of EPI coverage will also impact the ICERs as there are certain fixed costs that remain constant regardless of EPI coverage and the subsequent number of IPTi doses given (such as communication and sensitisation materials and a minimum number of training workshops) and certain variable costs that are related to coverage (such as IPTi drugs dispensed).
For the multi-dose IPTi drug regimens used in western Kenya, Korogwe and Same, additional costs were incurred delivering day two and three doses to achieve maximal efficacy. In a bid to reflect effectiveness rather than trial efficacy the costs of research staff used as adherence monitors were excluded from this analysis. It is important to recognise that there is likely to be a gap between trial efficacy and programmatic effectiveness for multi-day regimens. The threshold analysis shows the scope for additional IPTi delivery costs associated with monitoring adherence or a potential fall in protective efficacy if day two and three IPTi doses are not taken. For example, in western Kenya the threshold analysis presented in Table S5
shows that the PE of IPTi with SP-AS3 could fall from the trial level of 22% to 9% and from 25% to 10% with AQ3-AS3 and still remain highly cost effective. Alternatively, the IPTi cost per dose would need to increase from USD 0.60 to 1.33 with SP-AS3 and USD 0.44 to 1.64 with AQ3-AS3 before it was no longer highly cost effective. The costs and effects of using community health workers to prompt caretakers to administer IPTi doses in days 2 and 3 of multi dose IPTi regimens still need to be evaluated.
Every effort was made to conduct a rigorous analysis, but some limitations remain. Costing the intervention was a challenge as we had to extrapolate data from Mtwara, Tanzania to other settings and countries. The use of PPP adjustments is a recognised approach 
, but it would have been advantageous to look at cost variation across sites using primary data. However, the other sites in our analysis were randomized control trials and therefore it would not have been possible to measure real system delivery costs. Cost variation, within and across countries, has important implications for planning health services and budgets, however there is surprisingly little data published on this topic 
. A within-country cost and cost-effectiveness analysis of nationwide school-based helminth control in Uganda showed substantial variation between six districts in the cost per individual treated (USD0.41–USD0.91)
. Hutton and others investigated variation of maternity costs in Thailand and Cuba in the context of multicountry, multicentre randomised controlled trials. Unit costs per antenatal visit and per pregnancy showed considerable variation, due largely to staffing patterns and productivity 
. Across 5 Sub-Saharan African countries, the annualized economic costs per ITN distributed varied from USD2.75 in Togo to USD8.05 in Senegal 
, explained mainly by differences in the composition of each programme, levels of existing resources and spare capacity.
The implications of cost variation for decision making depend critically on the cost-effectiveness thresholds applied 
. By undertaking a threshold analysis, using a particularly conservative threshold of US$36, we were able to show that the cost per dose of IPTi, especially IPTi-SP, could vary, more specifically increase considerably, and still remain highly cost effective across most of the settings.
To be consistent with Hutton et al (2009) DALYs were calculated with no age weighting, however we recognize that the debate on the use of age weighting continues 
. Supporters of age weighting suggest all societies have age-based biases when deciding resource allocation. Detractors suggest DALYs can be criticized on equity grounds as every year of life is of equal value a priori, and on empirical grounds as the standard age weights may not accurately reflect social values.
All the ICERs presented in this analysis reflect the PE during the intervention period and using the pooled IPTi-SP PE. The analysis does not present the potential cost implications of an increase in drug resistance which is likely to lead to other health, health system and household costs 
. The threshold analysis presented here shows that there is room for the PE to fall and still remain highly cost effective in all sites where IPTi had a statistically significant effect on clinical malaria.
While IPTi is shown to be low cost and highly cost effective in this analysis, this does not guarantee that it will be adopted as a strategy. The funding of the intervention is vital and given the scarce resources and competing interventions (malaria and non-malaria related) countries may recognise the advantages of introducing IPTi but struggle to secure the funds. One of the great advantages of delivering IPTi is that it relies on an existing, well established delivery strategy such as the EPI scheme that already reaches a high proportion of the target IPTi recipients across all malarious countries.
Given the limited public health expenditure in many low income countries, decision makers need cost-effectiveness data to prioritise potential interventions for scale-up. This analysis shows that in many settings IPTi is a highly cost effective intervention, and that IPTi-SP would remain highly cost effective even if the level of PE of the intervention or the malaria incidence or the CFR were to decline. IPTi benefits from an already existing delivery system, EPI, which is a routine point of contact for many infants, making IPTi potentially one of the most cost effective malaria interventions available in areas where malaria transmission is moderate to high.