Introduction The burden of malaria has declined substantially in several areas of sub-Saharan Africa, particularly in the past 3–5 y . Such a change has been attributed to a combination of factors , including large scale indoor residual spraying campaigns ,, massive distribution of insecticide-treated bed nets , and the introduction of artemisinin-based combination treatments (ACTs) ,. The scale-up of the interventions has been possible thanks to the availability of more funding, especially from the Global Fund to Fight AIDS, Tuberculosis and Malaria , that has allowed an increasing number of countries to include ACTs in their national treatment guidelines as first and, in some cases, second-line treatments, and to the massive scale-up implementation of treatment programs . In addition, increased funding for research, often through effective public–private partnerships , has resulted in the availability of several ACTs . However, data to guide individual countries in choosing the most appropriate ACTs are limited. The World Health Organization (WHO), which recently produced revised guidelines for the treatment of malaria, states that the choice of ACT in a country or region should be based on the level of resistance to the medicine partnered to the artemisinin derivative in the combination . However, up-to-date treatment efficacy data for the partner medicine to the artemisinin derivative are scarce. The WHO recommends five ACTs, namely artemether-lumefantrine (AL), amodiaquine-artesunate (ASAQ), mefloquine-artesunate, sulfadoxine-pyrimethamine-artesunate, and, most recently included, dihydroartemisinin-piperaquine (DHAPQ) . Each of these combinations may have different advantages and disadvantages that vary according to a number of factors, including malaria endemicity, safety, tolerability, dosing, post-treatment prophylactic effect, resistance to the partner drug of the prevailing parasites in the area, and price. Accordingly, we carried out a head-to-head comparison of the safety and efficacy of several ACTs, with the aim of providing the information necessary to make an informed choice for the formulation of relevant national antimalarial treatment policies. The ACTs tested included three of those recommended by the WHO, namely AL, ASAQ, DHAPQ, and one that was under development, chlorproguanil-dapsone-artesunate (CD+A). AL was the first co-formulated ACT to become available and, together with ASAQ, is the most common ACT used in sub-Saharan Africa. DHAPQ has been recently submitted for registration to the European Medicines Agency under the orphan drug legislation  following two phase III trials that were carried out in Africa  and Asia . DHAPQ is currently used as a recommended treatment only in Asia , and a formulation approved by a stringent drug regulatory authority, such as the European Medicine Agency, or pre-qualified by the WHO is not yet available. At the time our trial started, combining chlorproguanil-dapsone (Lapdap, GlaxoSmithKline) with artesunate (CD+A) was considered to be a promising combination. Lapdap was on the market until 2008, when the producer withdrew it following the results of several studies showing that it caused significant reductions in hemoglobin (Hb) levels in patients with glucose-6-phosphate dehydrogenase deficiency . Therefore, the CD+A arm was stopped (because the drug was no longer in development owing to concerns over safety), and our trial continued with the other three ACTs under evaluation. Methods Study Design, Sites, and Concealment of Patient Allocation Between 9 July 2007 and 19 June 2009, a randomized, open-label, multicenter, non-inferiority clinical trial was carried out at 12 sites located in seven African countries (Nanoro, Burkina Faso; Fougamou and Lambaréné, Gabon; Afokang and Pamol, Nigeria; Mashesha and Rukara, Rwanda; Jinja, Tororo, and Mbarara, Uganda; Ndola, Zambia; and Manhiça, Mozambique). See protocol (Text S1) and amendments (Texts S3–S5), and CONSORT checklist (Text S2). Each site compared three of the four ACTs under investigation, ASAQ, DHAPQ, AL, or CD+A. The decision of which treatments to test at a given site was made by considering the current first-line treatments, the known antimalarial resistance profile, and local malaria endemicity (Table 1). Patients were individually randomized according to a 1∶1∶1 scheme, with six sites testing ASAQ versus DHAPQ versus AL, four testing DHAPQ versus CD+A versus AL, and two testing ASAQ versus CD+A versus DHAPQ (Table 1). A randomization list was produced for each recruiting site by the National Institute for Health Research Medicines for Children Research Network Clinical Trials Unit, University of Liverpool, UK, with each treatment allocation concealed in opaque sealed envelopes that were opened only after the patient's recruitment. 10.1371/journal.pmed.1001119.t001 Table 1 Study treatment to be tested by country. Country Sites Transmission (Entomological Inoculation Rate) Percent with Chloroquine Resistance Percent with Sulfadoxine-Pyrimethamine Resistance Study Treatments Burkina Faso Nanoro Seasonal, high (50–60) 24  4  ASAQ DHAPQ AL Gabon Fougamou, Lambaréné Perennial, high (50) 100  23  ASAQ DHAPQ AL Nigeria Afokang, Pamol Perennial, high 45  30  ASAQ DHAPQ AL Zambia Ndola Seasonal, mesoendemic High 19 (in adults)  ASAQ DHAPQ AL Rwanda Rukara Seasonal, high 40  36  DHAPQ CD+A AL Rwanda Mashesha Seasonal, high 50  12  DHAPQ CD+A AL Uganda Jinja Perennial, low (6)  28 , 49 , DHAPQ CD+A AL Uganda Tororo Perennial (>563)  45 , 9–15 , DHAPQ CD+A AL Mozambique Manhiça Perennial, mesoendemic  78  22  ASAQ CD+A DHAPQ Uganda Mbarara Mesoendemic 81  25  ASAQ CD+A DHAPQ Entomological inoculation rate is infective bites/person/year. Children 6–59 mo old (12–59 mo old at sites where CD+A was used) attending the health facilities with suspected uncomplicated malaria were included in the study if they fulfilled all the following inclusion criteria: body weight >5 kg, microscopically confirmed Plasmodium falciparum mono-infection with asexual parasite densities between 2,000 and 200,000/µl, fever (axillary temperature ≥37.5°C) or history of fever in the preceding 24 h, and Hb ≥7.0 g/dl. Patients were not recruited if they met at least one of the following exclusion criteria: participation in any other investigational drug study during the previous 30 d; known hypersensitivity to the study drugs; severe malaria  or other danger signs, e.g., not able to drink or breast-feed, vomiting (more than twice in 24 h), recent history of convulsions (more than once in 24 h), unconscious state, or unable to sit or stand; severe malnutrition (weight for height 200,000/µl at day 0, no fever or history of fever and parasite density 90% and the trial aimed at showing non-inferiority at a 10% difference threshold. Non-inferiority was demonstrated for the three pair-wise comparisons involving DHAPQ, ASAQ, and AL at most sites, with a few exceptions where the individual sites' 95% CI cross the non-inferiority limit. In west Africa, ASAQ continues to have excellent efficacy, comparable to that of AL ,, while in east Africa doubts about its use have been expressed . Indeed, in a study carried out in Kampala, Uganda, ASAQ had a lower efficacy than AL, a result confirmed in subsequent years . The superior efficacy of AL compared to ASAQ was explained by the presence of amodiaquine resistance in east Africa that may render this ACT increasingly less efficacious, similar to what has been observed for sulfadoxine-pyrimetamine-artesunate in east Africa . In Tanzania, AL was significantly better than ASAQ, but this was an effectiveness study, i.e., the treatment administration was not supervised, and ASAQ was not co-formulated, two important factors that may have influenced the treatment outcome . In our study, although ASAQ was tested mainly in west Africa, it also had excellent efficacy at sites located in eastern and southern Africa, namely Mbarara (Uganda), Ndola (Zambia), and Manhiça (Mozambique). However, the choice of the ACTs to test at the individual sites was influenced by their known drug resistance profile, i.e., ASAQ was not tested in sites with known high amodiaquine resistance. Therefore, although ASAQ is a possible option for some countries in east Africa, it should be not be deployed where amodiaquine resistance is known to be high. For the PCR-unadjusted efficacy at days 28 and 63, none of the pair-wise comparisons could show non-inferiority. Instead, DHAPQ was the best treatment, followed by ASAQ, AL, and then CD+A, which was consistently inferior to the three other ACTs. Considering that most of the recurrent infections were due to new infections and that the risk of re-infection depends on the activity of the non-artemisinin component, this result is largely expected. Indeed, piperaquine has the longer elimination half-life (about 23–28 d), followed by amodiaquine (3 wk), lumefantrine (3.2 d) , chlorproguanil (35 h), and dapsone (27 h) . The length and efficacy of the post-treatment prophylaxis may also be influenced by the transmission intensity. In Tororo (Uganda), a site with very high entomological inoculation rates (>500 infective bites/person/year) , the difference in the cumulative risk of recurrent malaria between AL and DHAPQ decreased when the follow-up period was extended to 63 d, suggesting that piperaquine's long elimination half-life could do little against the overwhelming risk of recurrent malaria . Nevertheless, in our study and at the same site, the risk of recurrent infections at day 63, though still extremely high, was lower in the DHAPQ than in the AL arm. When considering the forest plot comparing AL and DHAPQ, the OR tended to be lower at sites with the highest transmission intensity, suggesting that piperaquine's longer post-treatment prophylaxis still had an effect despite the high transmission. It should also be noted that places with an intensity of transmission as high as or higher than that of Tororo are not common, particularly in the current context of decreasing malaria burden in sub-Saharan Africa . Though delaying the occurrence of a second clinical attack, via a long post-treatment prophylactic effect, represents an advantage at the individual level, the increased risk of selecting resistant parasites among the new infections should be considered. Such risk occurs during a specific period, the “window of selection,” whose opening and duration is proportional to the drug terminal elimination half-life . Therefore, according to this model, such window would be shorter for lumefantrine (3–5 wk) than for piperaquine. Though this should not be a deterrent for the large-scale deployment of DHAPQ, setting up a reliable early warning system for the detection of resistance, possibly by both in vivo and in vitro tests, would be essential. In children treated with AL, gametocyte prevalence during follow-up and gametocyte carriage time were significantly lower than in children treated with either DHAPQ or ASAQ. The difference remained significant for gametocyte carriage time even when excluding patients with gametocytes at enrollment. Higher gametocyte carriage after treatment with DHAPQ, when compared to either AL or mefloquine-artesunate, has already been reported in some ,,– but not all trials . Similarly, gametocyte carriage was higher after treatment with ASAQ than with AL in some , but not all studies . The meaning of such differences in terms of transmission potential is unclear. Compared to molecular methods, microscopy detects only a small fraction of gametocyte carriers, both in individuals with asymptomatic infections  and in patients treated with an antimalarial . Children with microscopically detectable gametocytes are more likely to be infectious but those with sub-microscopic gametocytes can also transmit, albeit less efficiently . The gametocyte prevalence as determined by molecular methods has been observed to be higher in patients treated with DHAPQ than with AL . However, about 60% of children treated with AL and without microscopically detectable gametocytes were infectious to mosquitoes, with little difference between treatments, though the probability of a mosquito becoming infected was significantly lower for the ACT (AL and sulfadoxine-pyrimetamine-artesunate) than for monotherapy or non-ACT combinations . This indicates the difficulty of determining the transmission potential on the basis of the gametocyte carriage time as determined by microscopy, so that the differences in gametocyte carriage observed in our trial may not necessarily relate to a significantly different transmission potential. Considering that ACTs reduce the production of gametocytes by both decreasing the asexual reservoir and destroying a substantial proportion of immature, developing gametocytes, still sequestered in the microvasculature , and that parasite clearance was similar in the four study arms, the difference observed between the three ACTs may relate to their ability to clear almost mature forms not released in the blood stream yet. Hematological recovery up to day 28 post-treatment was similar for all ACTs tested except for CD+A, for which this was significantly slower, with a more marked Hb decrease up to day 7, confirming previous results . In addition, in the CD+A group, anemia was diagnosed more frequently as AE, providing additional evidence for the higher risk of anemia for this treatment. Aside from anemia risk related to CD+A, all regimens were well tolerated. In conclusion, this is, to our knowledge, the largest head-to-head comparison of most of the currently available ACTs for falciparum malaria in sub-Saharan Africa. CD+A was suspended partway through the trial, leaving AL, ASAQ, and DHAPQ under investigation. These three ACTs showed excellent efficacy, up to day 63 post-treatment, but the risk of recurrent infections was significantly lower, even in areas of high transmission, for DHAPQ, followed by ASAQ, and then AL. Although the gametocyte carriage rate differed between regimens, with those treated with AL having the lowest carriage rate and those treated with ASAQ having the highest carriage rate, the meaning of these different carriage rates with relation to transmission potential is unclear. The possibility of adding a single dose of primaquine to any of these three ACTs, with the objective of further reducing gametocyte carriage, should be explored ,. AL and/or ASAQ are already included in the antimalarial drug policies of many sub-Saharan African countries. This study confirms that DHAPQ is a valid third option for the treatment of uncomplicated P. falciparum malaria, as its efficacy is excellent and comparable to the other ACTs, while its long post-treatment prophylaxis could be an additional advantage. Supporting Information Text S1 Study protocol. (PDF) Click here for additional data file. Text S2 CONSORT checklist. (PDF) Click here for additional data file. Text S3 Protocol amendment 1. (PDF) Click here for additional data file. Text S4 Protocol amendment 2. (PDF) Click here for additional data file. Text S5 Protocol amendment 3. (PDF) Click here for additional data file.