Young children from three months to five years of age, attending the clinics of Shoklo Malaria Research Unit in Maela Refugee camp on the western border of Thailand with Burma, were considered for enrolment if they presented with symptomatic, microscopically confirmed Plasmodium falciparum malaria. They were excluded if they had signs of severe or complicated malaria30, a concomitant severe disease which required hospitalization, or any underlying neuro-developmental condition, including epilepsy. Controls were not randomly selected. They were the relatives of children in the study who were selected because their parents volunteered them after invitation, they had a negative malaria smear, and were free from medications and fever for at least 48 hours, and had a normal body temperature prior to neurological testing.

If consent was forthcoming a full medical history and clinical examination was carried out. This included assessing the duration of symptoms before presentation and any drugs taken prior to arrival at the clinic. Daily malaria smears were made until the patient was malaria smear negative. Temperature was measured daily until the patient was afebrile. Patients under treatment had daily clinical examination, drug administration, and adverse events documented on a case record form.

Participants with acute uncomplicated falciparum malaria received either artesunate alone (AS7), or artesunate in combination with mefloquine (MAS7). The total dose of artesunate was 12 mg/kg and the total dose of mefloquine was 25 mg/kg. Treatment regimes were as follows:

AS7: Artesunate 2 mg/kg for days 0 to 4, then 1 mg/kg for days 5 and 6.

MAS7: Artesunate 2 mg/kg for days 0 to 4, then 1 mg/kg on days 5 and 6; Mefloquine 15 mg/kg day 7 and 10 mg/kg day 8.

The protocol for drug administration was to calculate the target dose per body weight, crush the tablets in water and give with sugar, biscuits or milk. This was done by injecting the suspension into the mouth with a 5-ml syringe followed by milk and food if tolerated. If vomiting occurred within the first 30 minutes, the entire dose was repeated. If vomiting occurred between 30 minutes and 1 hour later, half the initial dose was repeated. Repeat dosing was given via nasogastric tube.

Reappearance of P. falciparum was treated with 7 days of artesunate, and mefloquine if it had not been prescribed in the previous 2 months. Treatment of Plasmodium vivax infections was with a standard chloroquine regimen (25 mg base/kg over 3 days).

Clinical assessment was performed daily from day 0 to day 5, then on days 7, 9, 10, then weekly until week 9. Only the results up to day 28 are reported here. Follow-up included basic clinical examination and weekly malaria smear and haematocrit from day 0. In addition, on each follow-up day, the parents or guardians of the participant completed a questionnaire regarding occurrence of symptoms and side-effects.

Children with malaria were tested on days 0, 7, 9, 10, 14 and 28. Children completed the neurological assessment before treatment on day 0, and on all other days were treated then assessed. Control children were invited for neurological examination on days 0, 7, 14 and 28. Testing until day 28 was considered necessary due to the pharmacokinetic properties of mefloquine31. Mefloquine was administered on days 7 and 8 so both treatment groups had repeat neurological testing on days 9 and 10.

Neurological items predominantly assessing hand coordination (box, cube, ring, coin, and a timed item to place 6 and 12 coins into a slot in a coin box as fast as possible) were adapted from the Griffiths Developmental Scales 32 and the Movement ABC 33; those assessing tone and behaviour from the Hammersmith Infant Neurological Examination 34and the Bayley Scales of Infant Development35. The complete testing procedure, applicability and scoring in this population have been described previously 36. These tests were chosen because they focused on coordination and concentration which were the CNS adverse affects of mefloquine and artesunate that were of greatest concern at the time. While an adult can be asked about dizziness young children cannot but it may be observed in items testing coordination by for example ataxic movements.

Children were randomized to one of two treatment groups using a list of random numbers that were allocated in a 1:1 ratio in blocks of 10. The randomization list was generated by an independent statistician who was not involved in conducting the study. The treatment allocation was concealed in envelopes labeled with the study code by another person not involved in conducting the study. The study envelopes were sorted by code and kept at the field sites. Inclusion was sequential. Patients who met the inclusion criteria were assigned the next available code. The envelope was then opened and the patient treated according to the allocation.

The testing room was separate to the medical outpatient department. Testers were kept blind to the treatment group of the patients with malaria. Testers knew only when patients belonged to the control group.

Blood smears (thin and thick films) were prepared using Giemsa staining and were read for 200 fields before being declared negative. Uncomplicated falciparum malaria was defined as slide-confirmed P. falciparum, with an asexual parasitaemia (between 6/500 white blood cells and 40/1,000 red blood cells, equivalent to 96 - 150,000 parasites/uL), in the absence of signs of severe malaria. Hyperparasitaemia was defined by a parasitaemia of ≥ 40/1000 RBC.

Haematocrit capillary tube samples were taken by finger prick and centrifuged at 1,500 rev/min for three minutes using a standard Hawksley™ haemotocrit reader. Anaemia was defined as a haematocrit of less than 30%, and severe anaemia by a haematocrit <20%.

The combination of mefloquine and artesunate was expected to cause a 3 fold increase in worse outcomes based on the increase in severe dizziness found in children treated with high dose mefloquine in this area 37. A sample size of 42 children in each group allowed a 3 fold higher incidence of worse outcomes (15 to 45%), to be detected with 95% confidence and 80% power. A 15% drop out was expected and 6 was added i.e 49 in each group.

Only children who completed a full course of treatment according to the protocol and who had at least two days of neurological testing, which included a test on day 0 (baseline), were included. Rather than use the overall scores for the items of hand co-ordination, behaviour and tone, some of which were age dependent, results of testing on a specified day were compared to a prior day e.g. day 7 compared to day 0. Children could refuse to participate in tests although they were all of an age where they could potentially score on the test item. Hence a child might refuse one week and get a top score the next week or vice versa. Therefore for each item examined we chose to assign the score as performed worse (score = 1), performed the same (score = 2) or improved performance (score = 3) in comparison to the previous score. Behaviour was scored in the same way and tone was scored as tone decreased (score = 1), tone stayed the same (score = 2) or tone increased (score = 3). Neurological test results could no longer be used once children were diagnosed with recurrent malaria of any species. Differences in proportions were compared using the Chi-squared test or Fisher's exact test when appropriate. Normally distributed data was compared using the mean and Student's T test. Data that was not normally distributed was summarized by the median and compared using the Mann Whitney U test when independent or with Wilcoxon Signed Ranks Test when paired at specified days. Statistical programmes used were SPSS for Windows, version 11.0 (SPSS), and Epi Info, version 3.4.0 (Centres for Disease Control and Prevention).

Ninety-eight patients between the ages of three months and five years with uncomplicated falciparum malaria were randomized between 5^th December 1994 and 29^th July 1997. Six patients did not have any paired neurological data and one patient was treated before testing on day 0 and all these children were excluded from analysis. Of the remaining 91 children, 45 were treated with artesunate alone and 46 were treated with mefloquine and artesunate. There were 36 control children, with a mean age of 2.2 ± 1.2 [0.3-4.7] years, of whom 61.1% (22/36) were male which was not significantly different from the children with malaria (P = 0.871, P = 0.574, respectively).

In children with malaria, there were no significant differences between treatment groups on admission characteristics (Table 1) or symptoms on admission (Table 2). After starting treatment fever and other symptoms decreased, and there was no significant difference in the proportion of children who developed new symptoms (data not shown). There was no difference in fever clearance time between the groups. Participants who became smear-positive for either P. falciparum or P. vivax during follow-up were retreated. Significantly more participants in the AS7 group become positive for malaria by day 28 than in the MAS7 group: 21.7% (10/46) vs 4.4% (2/45), P = 0.033. In the AS7 group one child had a mixed infection (P. falciparum and P. vivax) identified by day 14; the remaining nine all had P. vivax, one of whom was diagnosed by day 14, and the rest were diagnosed by day 28. In the MAS7 group two children had P. falciparum identified at day 28.

Many children refused testing on one or more items. The proportion of valid test scores was consistently lower in the malarial children than control children from day 0 to day 28 (Figure 1). The differences between malaria and control children were significant for box, cube, ring and coin items on day 0 (P < 0.001 for all) and on day 7 (P = 0.037, P = 0.019, P = 0.032, P < 0.001, respectively), day 14 cube (P = 0.037) and coin (P = 0.001) and on day 28 only coin (P = 0.012). As the days post-treatment increased there was a tendency in the malaria group towards increasing proportions of valid test scores however it never reached the same high proportion as the control children. By day 28, only the coin task showed significantly fewer valid test scores in the malaria group compared to the control group (Figure 1). There were no significant differences between the AS and MAS7 treatment groups in the proportion of valid test scores (data not shown) at each day of follow-up, on any neurological item. Tone and behaviour test scores were nearly always valid (Figure 1) regardless of the day of testing or group.

On day 0 children with fever were less likely to have valid test scores than non-febrile children for most items: box 62.7% (32/51) vs 82.9 (63/76), P = 0.013; cube 62.7% (32/51) vs 86.8% (66/76), P = 0.002; and ring 64.7% (33/51) vs 85.5% (65/76), P = 0.009. The proportion of febrile children who were timed while putting 6 or12 coins through a slot in a box (coin 6 and 12) was less than for non-febrile children but the difference was not significant. The median tone score in children with fever compared to those without fever was significantly higher 11 89101112131415 vs 11 6789101112131415, P = 0.020. The median behaviour score in children with fever compared to those without fever was significantly lower: 11 6789101112131415 vs 14 6789101112131415, P < 0.001. Tone was negatively and significantly correlated with behaviour (P = 0.007). Vomiting was also associated with a significant increase in tone and decrease in behaviour scores compared to those who did not vomit: 12 89101112131415 vs 11 6789101112131415, P = 0.013; 10 6789101112131415 vs 13 6789101112131415, P = 0.042, respectively.

Up to day 7, both malaria treatment groups received identical treatment with artesunate alone. There were no significant differences in the proportion of hand co-ordination, behaviour or tone scores that were worse, compared to the same or improved, in the AS7 vs MAS7 groups for any item (Table 3). There were also no differences between malaria and controls on items of hand co-ordination on the day 7 - day 0 comparison or on the day 28 - day 0 comparison, except for the ring task, which was significantly worse in the controls. Behaviour improved significantly more in the malaria treated children compared to controls at both day 7 to day 0 and day 28 to day 0, with most of the improvement apparent by day 7. Concomitant to the fever clearance, tone decreased significantly in malaria treated children at both the day 7 to day 0 and day 28 to day 0 comparisons (Table 3).

The proportion of children with worse scores for items of hand co-ordination on days 9, 10 and 14 compared to day 7 items, was not significantly different between mefloquine-and artesunate-treated children and those treated with artesunate alone (Table 4). Likewise behaviour scores were not significantly worse and tone scores did not decrease significantly once mefloquine treatment was given to the children (Table 4).

For this item children were required to put 6 or 12 coins into the coin box as fast as they could. Approximately half the children were unable to score on this item (Figure 1). Improvement on this item was reflected by a reduced time (seconds) to complete the task. The median time for each day of testing was summarized and paired data was compared for the specified days (Table 5). The mean age of children who managed to score on this item was significantly higher than those who did not complete this task: 6 coins; 3.4 ± 0.9 vs 1.7 ± 1.0 yrs and 12 coins; 3.4 ± 0.9 vs 1.8 ± 1.1 yrs, P < 0.001 for both. Nevertheless there was no trend for significantly worse times at day 7 compared to day 0 when an effect of artesunate may be noticed, nor at day 9, 10 and 14 compared to day 7, when an effect of mefloquine may be noticed. Although there are some significantly faster times in the artesunate-mefloquine group compared to artesunate alone group on day 7 and 9 and day 7 and 10, a practice effect cannot be ruled out.