In the present analysis, a population pharmacokinetic model was developed for AS and its active metabolite DHA using extensive sampling data from 26 pregnant and 25 non-pregnant women in the DRC. The model consists of a one-compartment model for AS and a one-compartment model for DHA, with AS absorption occurring through a mixed zero-order, lagged first-order absorption process. Upon absorption, AS is rapidly converted to DHA, with an approximate AS elimination half-life of 9.1 minutes. The model indicates that DHA apparent clearance is approximately 42% higher in pregnant than non-pregnant subjects, with resultant DHA elimination half-life estimates of 45 minutes and 59 minutes for pregnant and non-pregnant subjects, respectively.
The rapid elimination of AS found in this analysis is consistent with findings of pharmacokinetic analyses with IV AS, with an AS half-life estimate of 13.2 minutes obtained by Newton et al  and estimates of less than ten minutes found by Binh et al  and Batty et al . Given this rapid conversion of AS to DHA, the rate of DHA formation may be limited by the rate of AS absorption. The multiple samples collected in this study during the early period following AS administration allowed for AS absorption to be characterized using a mixed zero-order, lagged first-order absorption process that offered marked improvement in model fit over simpler absorption models. Given that AS is a weak acid with a pKa of 4.6 , absorption though this mixed-order process may reflect AS solubility and permeability changes occurring in the differing pH environments encountered in gastrointestinal transit. Gastric absorption of AS may be limited by the solubility of the free acid form of AS (168.2 μg/mL) ; such solubility-limited absorption would plausibly correspond to a zero-order process  such as the process characterizing the initial phase of AS absorption in the mixed-order absorption model utilized in the present analysis.
Erratic AS absorption in the postpartum women appears to have contributed substantially to the difficulty in identifying a satisfactorily predictive structural model for describing data from the postpartum subjects. Given the rapid conversion of AS to DHA upon AS absorption, unpredictable AS absorption would be expected to produce a pattern of DHA appearance inconsistent with standard compartmental modelling. The source of this atypical absorption may relate to breastfeeding; the women in the study were encouraged to bring their infants to the study site and to feed the children prior to AS administration. The effects of lactation on maternal kinetics have not been extensively studied. However, some studies have been performed evaluating the effects of lactation on ethanol kinetics. These studies report changes in ethanol pharmacokinetics, which may represent altered patterns of ethanol absorption, associated with the lactational state in general, as well as more acute effects induced by breast pumping or, presumably, infant suckling [21–23]. Suckling appears to trigger the release of various hormones responsible for regulation of digestion; these hormones can alter rates of processes such as gastric emptying . Therefore, it is plausible that the erratic AS absorption patterns observed for the postpartum subjects in this study may have resulted from the effects of lactation and recent infant suckling on AS absorption.
The only significant covariate identified in the present analysis was the effect of pregnancy on the clearance of DHA; this effect was estimated to produce a 42.3% (95% CI: 19.7% - 72.3%) proportional increase in DHA clearance in pregnant women as compared to non-pregnant controls. Clearance did not appear to differ substantially between the two windows of pregnancy, but more subjects in each trimester would likely be required for a difference between trimesters to be reliably detected. Apparent DHA clearance values for pregnancy and control subjects in the present analysis are similar to those obtained by non-compartmental methods, although apparent DHA volume of distribution was somewhat lower in the population, as compared to the non-compartmental, analysis . In the present analysis, the apparent volume of distribution of DHA trended higher for pregnant subjects, but the association between pregnancy status and increased volume of distribution did not meet the statistical significance criteria (p < 0.001) for the described covariate analysis methods. However, given that this association was statistically significant in the initial step of covariate modelling, the association would likely attain significance if assessed in a larger number of subjects.
The source of the pregnancy-related accelerated DHA clearance identified in this analysis is difficult to determine, as pharmacokinetic changes resulting from the physiological changes of pregnancy are not presently well understood. As DHA is metabolized through hepatic glucuronidation by UGT1A9 and UGT2B7 , induction of these enzymes could result in accelerated DHA clearance. Induction of hepatic glucuronidation, potentially by elevated sex hormone levels in pregnancy, may be responsible for the substantial pregnancy-related increases in glucuronidation observed for various drugs, including lamotrigine [24, 25], oxcarbazepine , and lorazepam . Alterations in hepatic blood flow could also produce changes in DHA clearance. Although such alterations in blood flow during pregnancy have been investigated, the results of these investigations are not in agreement . Additionally, blood flow changes may not be consistent across trimesters . Therefore, the manner in which hepatic blood flow alterations would be expected to contribute to pregnancy-associated DHA pharmacokinetic changes is difficult to predict.
The results of the present analysis are comparable to those of McGready et al. They assessed DHA pharmacokinetics following oral AS administration to 24 pregnant women (2nd or 3rd trimester) of the Karen ethnic group in Thailand with acute uncomplicated falciparum malaria . In their study, patients received a three-day regimen of orally administered AS (4 mg/kg/day) with 250 mg atovaquone and 100 mg proguanil; medications were given once daily with high fat milk. Samples used for pharmacokinetic analysis were obtained prior to the third daily dose and at seven time points between 0.5 and 12 hours following that dose. Since AS was detectable in only 21 of 323 samples, the investigators limited their analysis to DHA. Using population pharmacokinetic analysis, they modelled DHA data using a one-compartment model with a first-order rate of formation modelled as a fixed effect. The parameter estimates they obtained, adjusted for the median weight of their subjects (50 kg), were 88.5 L/hr [95% CI 60 - 117 L/h] for oral DHA clearance and 231.5 L [95% CI 57 - 406 L] for DHA volume of distribution. Their estimate for DHA clearance in pregnant patients is similar to the estimate in the present analysis (91 L/h). The volume of distribution estimate from the present analysis (91.4 L) is lower than found by these investigators. However, both the 91.4 L estimate and the 95% bootstrap confidence interval for that estimate (78.5-109 L), fall within their 95% confidence interval. Additionally, the 91.4 L estimate is similar to that obtained by other analysts, albeit obtained from the study of exclusively non-pregnant patients. Specifically, Teja-Isavadharm et al conducted non-compartmental analysis of DHA kinetics following oral AS administration to patients with falciparum malaria; their estimate of 1.33 L/kg [range: 0.70 - 2.70 L/kg]  is similar to the estimate obtained in the current analysis, adjusted for the median weight of the pregnancy and control groups, of 1.58 L/kg. A similar estimate of 1.33 L/kg [95% CI 1.02 - 1.64] was obtained by Newton et al when examining the kinetics of DHA following oral AS administration to acute falciparum malaria patients .
The pregnant women included in the present study were asymptomatic, displayed low-grade parasitaemia, and were otherwise generally healthy. Therefore, the results from this study can be generalized to populations for which intermittent preventative treatment regimens are indicated. Given that the model was not constructed using data from pregnant women with acute symptomatic malaria, it is not known if the model would optimally describe AS and DHA pharmacokinetics in such patients. However, given the findings of McGready et al, it seems probable that the significant pregnancy-associated increase in DHA oral clearance identified in the present analysis would be observed in pregnant women with acute malaria. In these patients, lower DHA blood levels resulting from accelerated DHA clearance could translate into reduced efficacy of AS and related compounds. Lower levels could also select for survival of parasites more tolerant to these compounds, increasing the risk of resistance development.