Patients and Methods

Area and population. This trial was performed from May through November 2007 at 2 health care centers in Allada and Sekou, southern Benin. P. falciparum transmission is intense and perennial, with recrudescence during the rainy season. P. falciparum is the predominant malaria‐causing species [1].

Study design. Until 2004, SP was the first‐line treatment of malaria in Benin. Since then, the Ministry of Health has adopted ACTs as first‐line drugs, and it made AL available in public heath care centers in 2008. The more recent marketing of ASAQ prompted us to compare the effectiveness of these 3 drug regimens. The study was a postmarketing open‐label, noninferiority comparative trial. Ethical clearance was obtained from the Ethics Committee, Faculty of Medicine, Benin National University, Cotonou, and the Ethics and Deontology Committee, Institut de Recherche pour le Développement, Paris, France. The study was open to children aged 6–60 months who presented to the outpatient clinic. Inclusion criteria were (1) fever (tympanic temperature, ⩾38°C) or fever during the previous 24 h, (2) uncomplicated P. falciparum infection with asexual parasite density >1000/μL, (3) weight >5 kg, and (4) written informed consent from the child’s guardian. Exclusion criteria were (1) receipt of an adequate antimalarial drug (SP, quinine, or ACT) within the previous 3 days, (2) clinical signs of disease severity, (3) screening hemoglobin level <5 g/dL, and (4) hypersensitivity to SP, AL, amodiaquine, or artesunate.

The study was performed in 3 phases: (1) screening; (2) supervised administration of the first drug intake, with doses provided to the mother or guardian and the drug regimen carefully explained; and (3) clinical and parasitological follow‐up for 42 days. Scheduled follow‐up visits were on days 3, 7, 14, 21, 28, 35, and 42. Children with early or late treatment failures were treated with mefloquine (25 mg/kg) or with intravenous quinine when parenteral treatment (24 mg/kg per day) was required. On day 3, information on adherence to the drug regimen was obtained at the patient’s home. When available, study drug blisters were collected to determine the number of missing pills.

Sample size calculations were based on results from a study in the same area, assuming day 28 cure rates of 60% in the SP group [1] and 95% each in the AL and ASAQ groups. We intended to establish noninferiority between ASAQ and AL and superiority of either ACT over SP with respect to treatment effectiveness. With a noninferiority margin of 10%, the sample size for 80% power (α=.05) was 76 children in each ACT group. To allow for loss to follow‐up (20%), 96 children were enrolled in each ACT group, and 48 were enrolled in the SP group.

Efficacy end points. The primary end point was the day 28 polymerase chain reaction (PCR)–adjusted cure rate. The secondary end point was the day 42 PCR‐adjusted cure rate. Clinical and parasitological outcomes were graded according to World Health Organization 2002 guidelines [11].

Study drugs and randomization. Treatment in the SP group was given once as 25 mg/500 mg tablets (Fansidar; Roche). Pills were given according to the child’s weight: 5–6 kg, 0.25 tablet; >6 to 10 kg, 0.5 tablet; >10 to 15 kg, 0.75 tablet; or >15 to 20 kg, 1 tablet. Treatment in the AL group was provided as 20 mg/120 mg tablets (Coartem; Novartis), given in a 3‐day, 6‐dose regimen. Children who weighed 5–15 kg received 1 tablet per dose, and children who weighed >15 to 20 kg were given 2 tablets per dose. Treatment with ASAQ (1 tablet once daily for 3 days) was provided as 25 mg/67.5 mg tablets (Coarsucam; Sanofi) for children who weighed 5–9 kg, 50 mg/135 mg tablets for children who weighed >9 to 18 kg, and 100 mg/270 mg tablets for children who weighed >18 kg. Treatment allocation was randomized using computer‐generated blocks of 5 children; in each block, 1 child received SP, 2 received AL, and 2 received ASAQ.

Laboratory analysis. A Giemsa‐stained thick blood film was prepared at inclusion and at each follow‐up visit. Slide quality control was achieved by masked rereading of 10% of slides, selected randomly. At enrollment and on day 28 after treatment, laboratory assessments were performed (hemogram, platelet count, alanine aminotransferase [ALT] level, creatinine level, total bilirubin level).

On day 3, 100‐μL blood samples were collected on Whatman 3MM filter paper from all patients assigned to receive an ACT. Lumefantrine (AL arm) or monodesethylamodiaquine (ASAQ arm) levels were determined using reversed‐phase liquid chromatography with ultraviolet (AL, 335 nm) or colorimetric (ASAQ, +0.8 V) detection after liquid‐liquid or solid‐phase extractions [12, 13]. Day 3 blood concentrations were also obtained for 60 children of similar ages and from the same area who were recruited solely for comparing drug levels and were given supervised treatments; 30 were treated with AL, and 30 were treated with ASAQ. These 60 children were not included in the comparison of treatment effectiveness. Half of the patients who were given unsupervised treatments were randomly selected for to have blood samples obtained on day 7 for additional drug level assessment.

DNA was prepared from blood collected at enrollment and on the day of treatment failure, when parasites reappeared in blood. Blood collected on filter paper was dried and conserved at room temperature until extraction. DNA was prepared by Chelex extraction [14]. Msp1 and Msp2 polymorphic genes were amplified as described elsewhere, with slight modifications [15]. DNA patterns of blood collected at baseline (day 0) and on the day of treatment failure were compared according to band size and number, considering the 3 families of msp1 and the 2 families of msp2. This allowed therapeutic failures (⩾1 band in the pattern on the day of treatment failure also present in the pattern at baseline, although additional bands may have been present at baseline) to be distinguished from reinfections (⩾1 new band on the day of treatment failure for either of the 2 markers and no common bands between day 0 and the day of treatment failure).

Data management. Data were entered into individual source documents. Data were double entered, validated with EpiData software (version 3.1; Centers for Disease Control and Prevention), and analyzed with Stata software (version 9.1; Stata). Drug concentration data were analyzed with JMP software (version 6.0; SAS). The intention‐to‐treat (ITT) analysis excluded randomized patients who failed to take ⩾1 dose of trial medication. Patients were not excluded because of violations of entry criteria, such as wrong age, wrong trial medication dosage, P. falciparum infection with <1000 trophozoites/μL, or coinfection with Plasmodium malariae. Patients unavailable for follow‐up and those withdrawn from the study because of an adverse event or the use of another drug with antimalarial activity were considered as having experienced treatment failure.

The day 28 and day 42 per‐protocol analysis excluded all violations of entry criteria, all wrong drug dosages, and all study withdrawals. Comparability between treatment groups was evaluated according to the following variables: age, sex, body weight, body temperature, baseline parasitemia, baseline hemoglobin level, diarrhea and/or vomiting before enrollment in the study, and baseline leukocyte count and ALT, total bilirubin, and creatinine levels. Patients excluded from the per‐protocol analysis were compared with evaluable patients according to the same variables.

Day 28 and day 42 effectiveness rates, both unadjusted and PCR adjusted, were compared. The superiority of both AL and ASAQ over SP has been tested by logistic regression analysis. ASAQ was considered to be noninferior to AL if the 1‐sided 95% confidence interval (CI) for the difference in treatment effectiveness was entirely within the interval of −∞ to 10% in both ITT and per‐protocol analyses. When noninferiority was not shown, the difference in effectiveness between AL and ASAQ was tested by logistic regression analysis. Changes in laboratory values between admission and the last study day (if the last day was after day 14) were compared across treatment groups with use of the paired Wilcoxon test.

Results

Two hundred forty patients were randomized (48 to receive SP, 96 to receive AL, and 96 to receive ASAQ). At enrollment, the 3 groups had similar baseline characteristics (table 1), as did patients excluded and those not excluded from per‐protocol analysis (data not shown). Eleven children had a blood parasite density ⩾200,000 parasites/μL at enrollment (median, 260,000 parasites/μL [range, 208,000–788,800 parasites/μL]; 6 received AL, 3 received SP, and 2 received ASAQ).

Figure 1 shows the flow of patients through the trial. All children except 1 received ⩾1 study drug dose and were included in the ITT population (48 in the SP arm, 96 in the AL arm, and 95 in the ASAQ arm). The excluded child received the first ASAQ dose but did not ingest it because of repeated vomiting, reported by investigators to the study drug.

Comparisons of effectiveness rates are shown in tables 2⇓–4. Noninferiority of ASAQ to AL was established both in the day 28 PCR‐adjusted ITT and per‐protocol analyses and in the day 42 PCR‐adjusted ITT analysis. In the per‐protocol analysis, AL was slightly more effective than ASAQ before PCR adjustment only (odds ratio [OR] for day 28, 2.2 [P=.05]; OR for day 42, 2.3 [P=.02]) (table 3). ACTs were much more effective than SP in both ITT and per‐protocol analyses (OR, 9.1–41.1; P<.001). The 89 treatment failures recorded on day 42 (PCR‐unadjusted per‐protocol analysis) included 17 early treatment failures (all in the SP arm), 16 late clinical failures (3 in the SP arm, 6 in the AL arm, and 7 in the ASAQ arm), and 56 late parasitological failures (19 in the SP arm, 10 in the AL arm, and 27 in the ASAQ arm). The 53 treatment failures in the day 42 PCR‐adjusted per‐protocol analysis included 17 early treatment failures (all in the SP arm), 7 late clinical failures (3 in the SP arm, 2 in the AL arm, and 2 in the ASAQ arm), and 29 late parasitological failures (16 in the SP arm, 4 in the AL arm, and 9 in the ASAQ arm). Nearly all recurrent episodes of parasitemia in the ACT groups occurred after day 14. The only recurrent episode of parasitemia observed on day 14 in an ACT group was proved to be pure reinfection.

Adverse events likely to be related to study drugs, as judged by investigators, are listed in table 5; all of these events were described as mild or moderate. There was no difference in the incidence of adverse events between treatment arms (10.4% in the SP arm, 6.25% in the AL arm, and 13.7% in the ASAQ arm; P=.23, by χ² test). No deaths were noted. Three children had to be hospitalized because of serious adverse events during follow‐up (all 3 within the first week after enrollment); 2 were hospitalized for severe anemia concomitant to persistent parasitemia (enrollment parasite densities, 220,610 parasites/μL and 280,000 parasites/μL) The third child received ASAQ, and pneumonia requiring parenteral antibiotics was diagnosed on day 1.

Enrollment and last study day safety laboratory parameters are shown in table 6. Two children (1 in the AL arm and 1 in the ASAQ arm) had a neutrophil count <800 neutrophils/μL on day 28. Both abnormalities were spontaneously reversible without clinical consequences. An asymptomatic and spontaneously reversible ALT level of 385 IU/L occurred on day 28 in a child who was given AL.

Genotyping was successful in 72 of 73 samples with recurrent parasitemia. Pure recurrent parasites, pure new infections, and a mixture of recurrent parasites and new infections were found in 15, 36, and 21 samples, respectively. The mean numbers of alleles were 3.25 for msp1 and 2.6 for msp2 at baseline and 2.9 and 2.8, respectively, on the day of treatment failure. New infections during the follow‐up period were more frequent in the ASAQ group (34.9%) than in the AL group (18.3%; P=.02). The median time from baseline to new infection was 35 days in the AL arm and 28 days in the ASAQ arm (P=.11, by Kruskal‐Wallis test restricted to newly infected patients).

Drug blisters were recovered at home from 81 children (84.4%) who received AL and 89 (93.7%) who received ASAQ. Rates of full adherence were similar in the 2 groups (83% in the AL arm and 91% in the ASAQ arm; P=.16). Overall, of the 171 children who received an ACT and were included in the per‐protocol analysis, 155 (75 in the AL arm and 80 in the ASAQ arm) were reported as having taken all prescribed doses at home. When the per‐protocol analysis was restricted to these 155 children, the PCR‐adjusted success rates were 94.7% (95% confidence interval [CI], 91.2%–98.2%) in the AL arm and 92.5% (95% CI, 88.4%–96.7%) in the ASAQ arm.

On day 3 after initiation of treatment, the mean levels (± standard deviation [SD]) of lumefantrine were 1.31±0.87 μg/mL in the supervised group (n=28) and 0.97±0.7 μg/mL in the unsupervised group (n=82; P=.03). The mean levels (±SD) of monodesethylamodiaquine were 0.31±0.08 μg/mL in the supervised group (n=30) and 0.28±0.16 μg/mL in the unsupervised group (n=91; P=.05). On day 7, the mean levels (±SD) of lumefantrine (n=44) and monodesethylamodiaquine (n=46) were 0.3±0.23 and 0.01±0.02 μg/mL, respectively. The mean day 3 and day 7 lumefantrine and monodesethylamodiaquine values did not differ according to outcome (success vs. failure) or according to the presence or absence of reinfection during the follow‐up period. The day 7 levels of lumefantrine and monodesethylamodiaquine were below the threshold of detection in 10 (23%) of 44 patients and in 26 (57%) of 46 patients, respectively (P<.001).

Discussion

Our trial highlighted the unacceptably low efficacy of SP in the management of uncomplicated malaria attacks in children aged <5 years in Benin. Two recent studies evaluated the efficacy of SP in uncomplicated malaria in Benin and found day 28 efficacy rates of 50% or slightly higher [1, 2]. Hitherto, the very low efficacy rate of SP has been a major concern, given the current situation of malaria treatment in Benin and elsewhere in Africa. Data are lacking on amodiaquine efficacy in Benin. Mefloquine has had high success rates [1], but most experts worldwide have given priority to antimalarial combinations in an attempt to prevent spreading of drug resistance. The SP‐artesunate combination efficacy is controversial in areas where resistance to SP is high [2, 16]. Since 2004, AL and ASAQ have been recommended in Benin as first‐line therapies. ACTs were not yet available when our trial ended but are now fully implemented in public health care facilities.

SP is now widely recommended in Africa as the basis of intermittent preventive therapy for malaria in pregnant women and infants [17, 18]. Because the current trend is a worsening of SP efficacy in the management of uncomplicated malaria, the efficacy of SP in intermittent preventive therapy is likely to wane and should be carefully monitored, and research studies on alternatives to SP in intermittent preventive therapy are essential.

Comparisons of day 28 PCR‐adjusted efficacies between AL and ASAQ combinations in Africa have produced inconsistent results [8, 19–23]. We used the ASAQ fixed‐dose formulation in this trial for the first time; previous studies have used amodiaquine and artesunate in separate formulations. Two studies in East Africa showed that AL was superior to ASAQ [8, 19], but other trials, like ours, did not find any difference between the combinations in day 28 per‐protocol effectiveness. One of the 2 East African studies was performed where the PCR‐adjusted failure rate for amodiaquine monotherapy was 48% [8]. In countries where ASAQ is implemented, efforts should be made to obtain in vitro and in vivo data on amodiaquine efficacy.

The posttreatment prophylaxis effect is an issue in terms of public health policy, because national authorities may favor treatment options that reduce the rate of any type of recurrent infection. Three trials showed that reinfection was less frequent after treatment with AL than after treatment with ASAQ [8, 20, 21]; 1 trial did not but found a similar trend [19]. Although some degree of misclassification with PCR adjustment is likely in hyperendemic areas, this finding seems reproducible.

In the current study, the posttreatment prophylaxis effect was superior with AL, compared with ASAQ, in the per‐protocol analysis only. Because of logistic and economic constraints, we had to choose a large noninferiority margin (10%), which reduces the power and may account for the nonsignificant results of superiority tests at day 28 (ITT analysis). Larger trials are needed to validate these results and to determine the posttreatment prophylaxis effect in both ITT and per‐protocol comparisons. However, even if post hoc analyses cannot be applied to the present study, noninferiority would probably be shown with a lower threshold in both PCR‐adjusted per‐protocol and ITT comparisons, reflecting the actual efficacy of treatments.

Checchi et al. [24] reported that reinfections occurred only when day 7 lumefantrine levels were low. Hietala et al. [25] found a high rate of undetectable monodesethylamodiaquine levels on day 7 but did not assess adherence; their estimation of the risk of parasitemia during follow‐up was 40% among patients with an undetectable monodesethylamodiaquine level on day 7. In our trial, the proportion of patients with an undetectable drug level on day 7 was significantly higher for ASAQ than for AL, but reported adherence and day 3 drug levels did not indicate low adherence in the ASAQ group. The elimination half‐life of amodiaquine metabolites was formerly reported to be 1–3 weeks [26]. It was recently shown in healthy human volunteers that adding artesunate to amodiaquine modifies several pharmacokinetic parameters, including reducing the elimination half‐life of monodesethylamodiaquine [27]. We believe that our day 7 pharmacology findings reflect rather fast elimination of amodiaquine metabolites, which might affect the difference in posttreatment prophylaxis effect between ASAQ and AL, but additional studies are needed to test this hypothesis. In the selection of drug‐resistant parasites, the longer half‐life of lumefantrine is theoretically not beneficial.

We observed no severe adverse events related to treatment. In small children, we were unable to accurately assess adverse events such as nausea and abdominal pain, which may be a concern in the clinical management of older patients. Because our trial was open label, investigators may have been influenced by their knowledge of study drugs when identifying adverse events. Few adverse biological changes occurred during the follow‐up period, and all were subclinical and transient.

Our study design attempted to give the best‐possible true picture of the expected improvement related to drug policy change in the management of malaria. First, we ignored an upper limit for parasitemia at inclusion. The Roll Back Malaria program promoted the implementation of rapid diagnostic tests for malaria, and this policy will not provide health care providers with quantitative parasite density at diagnosis. Among patients whose baseline parasite load was >200,000 parasites/μL, 2 of 3 children treated with SP had to be hospitalized for severe anemia within the first week. We suggest that, when the efficacy of a study drug is doubtful, the upper limit of parasitemia should be strictly taken into account at inclusion.

Second, although it is difficult to assess effectiveness in a clinical trial (the mothers’ understanding of the study drug regimens and their motivation to give a full course of therapy may have been boosted in several ways), we did not fully supervise administration of ACTs; however, they still achieved high success rates. Suboptimal adherence was not rare with AL and ASAQ regimens but did not affect effectiveness. Compared with supervised treatments, unsupervised AL treatment achieved lower drug levels, and a similar trend was observed for ASAQ. Another study found decreased plasma lumefantrine levels after unsupervised AL treatment and no change in effectiveness [28]. We advised the mothers to administer AL with fat (mainly palm oil), but this is not common practice, and the lack of concomitant ingestion of fat may partly explain the trend to a more pronounced decrease in lumefantrine levels after unsupervised AL. Drug level data are a complementary way of estimating adherence during chronic diseases and may also provide critical information for interpreting bad outcomes in the treatment of acute infectious diseases, especially with unsupervised treatment.

In conclusion, this trial compared unsupervised AL with unsupervised ASAQ fixed‐dose formulation, used for the first time in an effectiveness trial. Both treatments provided similar PCR‐adjusted day 28 effectiveness rates of >90%. The impact of replacing SP by ACTs for the treatment of malaria in children aged 6 months to 5 years in the study area should be highly significant.