Demographic, clinical, parasitological, and laboratory data for all patients by study site are presented in and . All patients fulfilled the criteria for moderately severe malaria with parasitaemias >0.1%, moderately elevated plasma lactate levels (except in Bangkok and Malawi, where three and seven patients, respectively, had lactate concentrations >5 mmol/l), and moderate anaemia. The total number of suppositories given to each patient was usually between one and three (89%), although up to six suppositories were given to South African adults. Few (<10%) suppositories were expelled, the exception being Malawi, where 19% of children expelled suppositories.
Baseline Laboratory Data of All Patients
Safety and Clinical Efficacy
One patient from Mae-Sot died. The most likely cause of death was considered to be over-transfusion of 5% dextrose leading to iatrogenic water intoxication. Nineteen patients required rescue treatment ().
Efficacy Data of All Patients
The integrated adverse-event count yielded 66 adverse events from the 179 patients. The leading adverse events reported were gastrointestinal (mostly nausea, vomiting, and abdominal pain). For five patients, central nervous system or neurological adverse events were observed. These were convulsions, dizziness, impairment of consciousness, abnormal reflexes, and vertigo. However, none of the reported events led to treatment discontinuation or were classified as causally related to rectal ARS therapy.
A total of 307 levels from 136 patients were available for the statistical modelling (one outlier with an ARS concentration of 9,224 ng/ml 13 min after drug administration was excluded because of the likelihood of sample contamination). Of these, 32 (10.4%) levels were below the limit of quantification and one (0.3%) level was above the range.
shows a scatter plot of the observed ARS concentrations versus time, by study site. There are no major differences in the distribution of data across the five different study sites. A clear pharmacokinetic profile is difficult to discern, precluding formal pharmacokinetic modelling. The median (range) of the observed individual peak concentrations was 269 ng/ml (11–4,720 ng/ml).
Observed Artesunate Concentrations (Semilogarithmic Scale; ng/ml) versus Time post Dose (h)
A total of 424 levels were available for statistical modelling from 164 patients (three outliers were excluded with levels >6,000 ng/ml). Of these, 33 (7.8%) levels were below the limit of detection and three (0.7%) were above the range. shows a scatter plot of the observed DHA concentrations versus time, by study site. A one-compartment model with first-order appearance and elimination kinetics including lag time best fits the data (appearance rate was fixed at 0.2/h and lag time at 0.14 h).
Observed and Population Predicted Concentrations (Solid Black Line) of DHA (Semilogarithmic Scale; ng/ml) versus Time post Dose (h)
The data did not support a full variance–covariance matrix for the random effects of CL/F and V/F. The mean population estimate of CL/F was 2.64 (l/kg/h) with 66% inter-patient variability, and V/F was 2.75 (l/kg) with 96% inter-patient variability. These estimates correspond to an elimination half-life of 43 min. The residual error model that gave the minimum value for the objective function was a lognormal residual error model for all the data. Including a covariate for “study site” or “levels outside the limits of the assay” in the residual error modelling did not improve the objective function. shows the observed and population-predicted concentrations versus time. The residual plot indicated no bias in estimation (see ), and the patient-specific profiles were characterised adequately (unpublished data).
Residual (loge Units) versus Population Predicted Concentrations of DHA (loge Units)
Relationship between DHA Pharmacokinetic Variables and Covariates
A total of 164 of the 179 patients had posterior individual estimates of CL/F and V/F derived from the population pharmacokinetic modelling of DHA.
Age, Body Weight, Gender, and Geographic Group (Southeast Asia versus Africa)
When age was treated as a dichotomous variable, children (age ≤15 y) versus adults (age >15 y), there was no significant difference between adults and children for the distribution of the patient-specific deviations from the population estimate of CL/F (ηCL/F), whereas the patient-specific deviations from the population estimate of V/F (ηV/F) were higher for adults compared with children (p < 0.001). No correlation was observed between ηCL/F and body weight, whereas there was a positive correlation between ηV/F and body weight (r = 0.42). There was no significant difference between males and females for the distribution of ηV/F, whereas significantly higher values of ηCL/F were observed for males compared with females (p = 0.003). There were no significant differences between patients from Southeast Asia and Africa, for both the distributions of ηCL/F and ηV/F.
Baseline Parasitaemia, PCV, Lactate and Glucose, and Number of Suppositories
No correlations were observed between ηCL/F and ηV/F and baseline parasitaemia, lactate and glucose, and baseline PCV and ηCL/F. There was a positive correlation between baseline PCV and ηV/F (r = 0.22, p = 0.005). There were no significant differences between both the distributions of ηCL/F and ηV/F for the number of suppositories given (either one, two, or three or more).
presents the final population parameter estimates and the inter-individual and intra-individual variability. Including the covariates “gender” and “body weight” reduced the inter-individual variability of CL/F from 66% to 62% and V/F from 96% to 75%, respectively. The addition of baseline PCV and age did not significantly improve the objective function.
Population Pharmacokinetic Parameters for DHA
Relationship between DHA Pharmacokinetic Parameters and Pharmacodynamic Variables
The pharmacodynamic variables defined prospectively as important were treatment outcome, early rescue treatment, failure of baseline parasitaemia to fall 40% by 12 h (PC60), and time to clear 50% (PCT50) and 90% (PCT90) of baseline parasitaemia.
The DHA pharmacokinetic parameters (available for 164 of the 179 study participants) investigated were posterior individual estimates of CL/F, V/F, and AUC0–6h.
For the statistical modelling, treatment outcome was a dichotomous variable with death, positive smear from day 7 to day 28, or undetermined response grouped together (termed “treatment failure”, n = 91) versus “treatment success” (negative smear from day 7 to day 28; n = 73). This definition of treatment outcome does not include the “early rescues”, which are considered under the “Early rescue treatment” section below.
No significant difference in the distribution of the posterior individual estimates of CL/F and AUC0–6h was observed between the two groups of patients, treatment failure, and treatment success. Patients who had failed their treatment (n = 91) had significantly lower values of V/F compared with those with treatment success (n = 73) (geometric mean [95% CI]: 2.46 [2.25, 2.66] versus 3.16 [2.80, 3.60] l/kg; p = 0.001), but after multiple logistic regression analysis, with adjustment for baseline parasitaemia, baseline PCV, age, and gender (i.e., pre-defined confounders), no significant association between V/F and “treatment outcome” was observed (p = 0.12). The observed univariate association between “treatment outcome” and V/F was confounded by age.
Early Rescue Treatment
For the statistical modelling, rescue treatment was a dichotomous variable with absconded, death, or rescued by day 7 grouped together (n = 16) versus not rescued (n = 148). No significant association between rescue treatment and both CL/F and AUC0–6h was observed. Those rescued had slightly higher values of V/F compared with those patients not requiring rescue treatment (geometric mean [95% CI]: 3.42 [2.58, 4.53] versus 2.69 [2.49, 2.90] l/kg, p = 0.056). From the multiple logistic regression analysis with adjustment for confounders, a significant association between V/F and “rescue treatment” was observed (p = 0.01). As the V/F increased, so did the risk of requiring rescue treatment (see ). Whether this reflected a truly larger apparent volume of distribution or a reduced fraction of drug absorbed (F) cannot be distinguished.
Relationship of Estimated Probability of Requiring Rescue Treatment and V/F of DHA (l/kg)
Failure of Baseline Parasitaemia to Fall 40% by 12 h
There were 20 patients whose parasitaemia failed to drop by 40% of baseline values by 12 h. No significant association between failure and both CL/F and AUC0–6h was observed, but patients who failed to drop parasitaemia by 40% by 12 h had higher V/F values compared with the remainder (geometric mean [95% CI]: 3.36 [2.17, 3.63] versus 2.67 [2.44, 2.80] l/kg; p = 0.045). Multiple logistic regression analysis with adjustment for confounders showed no significant association between V/F and “failure to drop to 60% parasitaemia at 12 h” (p = 0.09).
Time to Clear 50% (PCT50) and 90% (PCT90) of Baseline Parasitaemia
Two adult patients from South Africa with baseline parasitaemia <10,000/μl were excluded from the analyses of the efficacy variables, PCT50 and PCT90. No significant correlations between both PCT50 and PCT90 and the pharmacokinetic parameters CL/F, V/F, and AUC0–6h were observed (Spearman rank correlation coefficients ranged from −0.07 to 0.17). These results were unchanged when adjusting for baseline parasitaemia, PCV, gender, and age. A and B illustrate the lack of association between PCT50, PCT90, and AUC0–6h, respectively.
Observed PCT50 (A) and PCT90 (B) versus Posterior Estimate of DHA AUC0–6h (ng/ml × h)
As ARS is rapidly hydrolysed in vivo to DHA and both have approximately equal antimalarial activity, the molar sums of ARS plus DHA plasma concentrations were also modelled. A total of 442 ARS-plus-DHA levels from 168 patients were available. The associations between the posterior individual PK parameters of ARS plus DHA (i.e., CL/F, V/F, and AUC0–6) and the efficacy variables (defined above) were the same as observed for the posterior individual PK parameters of DHA alone (unpublished data).