Optimal antimalarial prophylaxis with maximum compliance would be provided by a regimen that is long acting so that administration is not required more frequently than monthly. Sulphadoxine, mefloquine and DP have sufficiently long half-lives to be considered [28]. However, there is high-level resistance to sulphadoxine in many parts of east and southern Africa, precluding its use for this purpose in these malaria endemic areas [29]. Both amodiaquine [30] and mefloquine are poorly tolerated, which is an important consideration when providing drugs for malaria prevention to recipients with few or no symptoms [31, 32]. DP is very effective, well tolerated and provides 4 to 5 weeks of post-treatment prophylaxis. It is therefore currently the drug of choice for use for evaluation as part of IPT and malaria chemoprevention in areas with high-grade parasite resistance to sulphadoxine [33–38]. Furthermore, recent studies show that artemether-lumefantrine and DP exert inverse selective pressure on Plasmodium falciparum drug sensitivity [39], suggesting that DP may be a good choice for chemoprevention in areas where artemether-lumefantrine is the first-line drug of choice for case management.

The primary study population involves children with severe anaemia, rather than only children with severe malarial anaemia, which was the study population in the previous trial in Malawi [17]. This is based on observational studies in Malawi, Uganda and western Kenya showing that children admitted with severe anaemia appear to be at increased risk of readmission and death regardless of whether they had evidence of malaria infection at the time of admission or not (Desai et al., unpublished observations; Richard Opoko, unpublished observations) [17]. Second, reliable diagnosis of the presence of malaria is difficult, and the differentiation between severe anaemia and severe malarial anaemia is not always feasible, as it is common practice in many hospitals in sub-Saharan Africa to start parenteral treatment with antimalarials before the laboratory diagnosis of malaria is available. Furthermore, the interpretation of malaria diagnostic tests on admission may be complicated in children who received antimalarial treatment just prior to admission [40].

This current study is an efficacy trial, and each treatment course will be provided by study staff directly. The first dose of each course will be observed, and where feasible, doses on days 2 and 3 will also be given under supervision, or compliance verified by home visits or contacting caretakers by mobile phone. A separate trial, focusing on the effectiveness of different delivery mechanisms, is being conducted by our consortium members in Malawi (NCT02721420).

Use of clinical malaria as primary outcome would require a smaller study; however, the composite outcome is used because it is more likely to drive policy. We use a composite outcome rather than a single severe outcome, such as death, to keep sample size requirements manageable.

The period 2–26 weeks instead of 0–26 weeks is used for the primary efficacy analysis because children will not be randomised until 2 weeks after enrolment. Prior to 2 weeks, all children, including those in the placebo arm, will receive a 3-day course of artemether-lumefantrine as part of standard in-hospital care, which will be started before discharge and completed at home after discharge. The duration of post-treatment prophylaxis in our previous trial with artemether-lumefantrine is about 2 to 3 weeks [17], and we therefore do not anticipate any differential effect between the arms until children receive their first study-specific intervention upon randomisation. The protective drug levels have waned in many children by 14 weeks (i.e. about 4 weeks after the last PMC course of DP), but we follow the children for a total of 26 weeks to capture any potential prolonged benefits or rebound effects.

Children will be randomised to one of the two treatment groups: DP or placebo. Children in both arms will receive standard in-hospital care and, at discharge (enrolment) a 3-day course of artemether-lumefantrine regardless of whether they were admitted with severe malarial anaemia or severe anaemia without evidence of malaria.

The primary efficacy outcome is the number of all-cause deaths or all-cause readmissions between 2 and 26 weeks after enrolment (composite outcome).

Readmission due to severe malaria (defined as any treatment with parenteral quinine or artesunate, or presence of severe anaemia and treatment with oral antimalarials) by 26 weeks from randomisation

Readmission due to severe malaria-specific anaemia (severe anaemia plus parenteral or oral antimalarial treatment and parasite density > 5000/μL) by 26 weeks from randomisation

Serious adverse events, excluding primary and secondary efficacy outcomes, by 26 weeks from randomisation

The study timelines consist of an in-patient pre-study screening period while the patient is acutely ill (visit 1), followed by a screening, consent and enrolment visit (visit 2). During the convalescence phase in the hospital, patients receive artemether-lumefantrine (visit 3) prior to discharge. The patient returns to the study clinic 14 days later (visit 4) for randomisation. Home treatment visits are made at 6 (visit 5) and 10 (visit 6) weeks. The PMC period starts at 2 weeks and ends at 14 weeks, but participants receive passive follow-up for an additional 12 weeks and are then seen at 26 weeks (visit 7) for an end-of-study assessment (see Additional file 2, Fig. 2 and Additional file 1, section 8.7, pages 37 to 40)Fig. 2

Study Design and Schedule of Assessment (Spirit figure)

A passive surveillance system is in place to monitor intercurrent illnesses during the observation period. Parents are instructed to bring their child to the study clinic for any suspected illness. Blood samples for Hb, malaria diagnosis (RDT and smear) and filter paper dry blood spot (DBS) for parasite genetics are obtained. Verbal autopsy is conducted for children who die at home during the follow-up period. Adverse events and vital status are assessed during all scheduled or unscheduled visits.

The initial estimate of the required sample size was 2212 children (1106 per arm) across both countries pooled. This estimate was designed to detect a 30% reduction in the incidence rate of the composite primary outcome (death or all-cause readmission) from 469 per 1000 child years in the control arm to 328 per 1000 child years in the intervention arm (power 90%, α = 0.05), which allowed for one interim analysis and 15.7% loss to follow-up. For these estimations, we assumed an average pooled event rate of 399 per 1000 child years across the two arms, with 328 and 469 events per 1000 child years in the intervention and control arms respectively (Rate Ratio (RR) = 0.70). We based this assumption on observations in western Kenya (Desai et al., unpublished) and Malawi [17]. However, the observed event rate, pooled across both arms, during the first year of the study was 1120/1000 child years, which is almost three times higher than the assumed event rates. The higher rate is consistent with the recently published observations in Uganda [11]. Furthermore, the observed rate of loss to follow-up in the first 533 participants recruited and followed up for 6 months was 7% rather than the assumed 15%.

Following recommendations from the Data Monitoring and Ethics Committee (DMEC) and the Trial Steering Committee (TSC), a blinded interim sample size re-estimation was conducted to take into account the lower than expected rate of loss to follow-up and the higher than expected pooled incidence rate of the composite primary endpoint (death or all-cause readmission). This was favoured over an interim analysis, because the available funding did not allow an extension of the recruitment period, even if the results of any interim analysis had suggested that this would be required.

The revised sample size calculations show that a total sample size of 1040 children (520 per arm) is required to detect a 25% reduction in the incidence of the composite primary outcome from 1152 per 1000 child years (530 events per 1000 children) in the control arm to 864 per 1000 child years (398 per 1000 children) in the intervention arm (power 80%, α = 0.05), allowing for 10% loss to follow-up. The same sample size also provides 90% power to detect a 28.7% reduction in the primary endpoint from 1152 to 822 events per 1000 child years.

Eligible children are randomly assigned (1:1) to either PMC-DP or placebo by a computer-generated randomisation schedule stratified by weight (per DP dosing schedule) and study site using permuted blocks of random sizes (see Additional file 1: page 30, Table S4) [12]. Recruitment is ‘competitive’ between the sites in the trial.

The study is double-blinded to both participants/caretakers and study staff. Allocation concealment is achieved by the use of sealed opaque envelopes, with each envelope containing three other small envelopes (one for each PMC course). The envelopes containing active DP or placebo look identical, and the appearance and consistency of the tablets are also identical.

We will use stratified analysis to assess to what extent the effect of the intervention on the primary outcome is influenced by country, demographic parameters (e.g. age, ethnicity and socio-economic status), clinical parameters, malaria transmission variables (malaria transmission intensity, residence (urban/rural), season, insecticide-treated nets use, site), time of assessment and potential intervention modifiers. Because we did not power the study for sub-group analyses, we will interpret the results of the sub-group analysis cautiously. No adjustment will be made for multiple comparisons.

A number of sensitivity analyses will be conducted to assess the robustness of the primary endpoint analysis. These include analysis of the PP subject population and a covariate adjusted analysis. Other regression models will also be explored. Additional post hoc analyses may also be conducted if deemed appropriate. In addition, we will compare the results of the covariate-adjusted analyses with and without imputation for missing values for covariate values at baseline.

Secondary efficacy outcomes include all-cause and malaria-specific clinic visits. They will be assessed through passive case detection as well as questionnaires administered during visits 4–7 at 2, 6, 10 and 26 weeks and during unscheduled sick visits. Details of clinic visits are recorded including malaria diagnosis results to allow for differentiation between malaria and non-malaria clinic visits.