Pharmacokinetic Properties of Conventional and Double-Dose Sulfadoxine-Pyrimethamine Given as Intermittent Preventive Treatment in Infancy

Study site, sample, and approvals.The present study was conducted at Alexishafen Health Centre, Madang Province, on the north coast of Papua New Guinea (PNG). Infants between the ages of 2 and 13 months from the surrounding area were eligible for recruitment provided that they (i) did not have features of severe malaria or significant nonmalarial illness, (ii) had not been treated with SDX or PYR in the previous 4 weeks, (iii) did not have a known allergy to either SDX or PYR, and (iv) were available for assessment for the duration of follow-up. Written informed consent was obtained from the parents/guardians of all recruited infants. The study was approved by the Medical Research Advisory Committee of PNG and the Human Ethics Research Committee at the University of Western Australia.

Clinical procedures.At enrollment, a clinical assessment was performed that included a standard baseline symptom questionnaire completed by parents/guardians. A 500-μl finger prick capillary blood sample was taken for preparation of blood smears for microscopy, baseline drug assay, biochemical tests, and hemoglobin concentration (HemoCue, Angelholm, Sweden). Subjects were randomized to receive either the recommended dose of SDX/PYR (25/1.25 mg/kg Fansidar; Roche, Basel, Switzerland) or a double dose (50/2.5 mg/kg). Table 1 shows the dose administered based on body weight. All dosing was directly observed, with subsequent monitoring and readministration of the dose if the infant vomited within 30 min. Infants with a positive blood film were also given a 3-day course of amodiaquine according to PNG national treatment guidelines (33). All drugs were crushed and mixed with either water or breast milk before administration by mouth using a syringe.

All infants were reassessed on days 1, 2, 3, 5, 7, 14, 21, and 28. A hemoglobin concentration was determined on each occasion and a repeat symptom questionnaire administered at each visit up to day 7. Blood films were repeated on day 28 and/or when fever or a recent history of a fever was reported. For pharmacokinetic analysis, four additional 500-μl capillary blood samples were taken from each infant. The times for the first three of these were randomly selected for each infant from either 4 to 8 h or 2, 5, 7, 14, or 21 days postdose. A final sample was taken in all cases on day 28. The exact timing of each blood sample was recorded. All samples were centrifuged promptly, with red cells and separated plasma stored frozen at −80°C until assay.

Laboratory methods.Giemsa-stained thick blood smears were examined independently by at least two skilled microscopists who were blinded to dose group. Each microscopist viewed >100 fields at ×1,000 magnification before a slide was considered negative. Any slide discrepant for positivity/negativity or identification to the species level was referred to a third microscopist.

Cystatin C (CysC) concentrations were measured by particle-enhanced immunoturbidimetry (PETIA) using the Tina-quant cystatin C kit run on an Elecsys 2010 analyzer (Roche, Indianapolis, IN). Sodium, urea, creatinine, albumin, γ-glutamyl transferase, and bilirubin were measured using an Integra 800 analyzer (Roche) when sufficient plasma was available.

Sulfadoxine, sulfamethazine, and pyrimethamine were obtained from Sigma-Aldrich (Castle Hill, Australia), and midazolam hydrochloride was obtained from Pfizer (West Ryde, Australia). N₄-acetylsulfadoxine (NSX) was synthesized according to the method of Whelpton et al. (39) and found to have a melting point of 230°C and >99.9% purity by high-performance liquid chromatography (HPLC). Acetonitrile was obtained from Merck (Darmstadt, Germany). All other chemicals were of analytical or HPLC grade.

For PYR, SDX, and NSX, extraction and separation were performed based on previously published HPLC-UV methods (26, 37). The internal standards were midazolam HCl for PYR and sulfamethazine for SDX and NSX. Analytes were assayed using UV detection at 270 nm. Chemstation software (version 9; Agilent Technology, Waldbronn, Germany) was used for analysis of chromatograms. Standard curves were linear from 5 to 1,000 μg/liter, 0.1 to 200 mg/liter, and 0.02 to 10 mg/liter for PYR, SDX, and NSX, respectively. Intra- and interday relative standard deviations (RSDs) were <15% for all analytes at all concentrations. The limits of quantification (LOQ) were 2.5 μg/liter, 0.1 mg/liter, and 0.02 mg/liter, and the limits of detection (LOD; determined as a signal-to-noise ratio of 5) were 1 μg/liter, 0.05 mg/liter, and 0.01 mg/liter for PYR, SDX, and NSX, respectively.

Population pharmacokinetic analysis.Log_e concentration-versus-time data sets for PYR, SDX, and NSX were analyzed by nonlinear mixed effect modeling using NONMEM (version 6.2.0; ICON Development Solutions, Ellicott City, MD) with an Intel Visual FORTRAN 10.0 compiler. Linear mammillary model subroutines within NONMEM (ADVAN2/TRANS2 and ADVAN4/TRANS4), first order conditional estimation (FOCE) with η-ε interaction, and the objective function value (OFV) were used to construct and compare plausible models. Unless otherwise specified, a difference in OFV of ≥6.63 (χ² distribution with 1 df, P < 0.01) was considered significant. Due to the small number of samples with low concentrations, those below the LOD were not included in the analysis, while levels between the LOD and LOQ were kept at their measured concentrations.

As the subjects were infants with a range of ages, it was important to incorporate maturation of clearance into the model. Therefore, total clearance (CL_T) was defined as the sum of hepatic clearance (CL_H) and renal clearance (CL_R), i.e., CL_T = CL_H + CL_R. The age-adjusted hepatic clearance, CL_H, was determined using a sigmoid maximum effect (E_max) model (7) as TVCL_H × [PMA^HillCL/(PMA^HillCL + MATCL₅₀^HillCL)], where TVCL_H is the population average value for hepatic clearance, PMA is the postmenstrual age (the age of the infant recorded from the last menstrual cycle of the mother during pregnancy rather than birth), HillCL is the Hill coefficient for hepatic clearance, and MATCL₅₀ is the PMA at which CL_H is 50% of the mature value. When an accurate PMA could not be obtained, it was estimated from the postnatal age (PNA) and average gestation in PNG (3, 15, 21). CL_R was adjusted to a standardized value for an estimated glomerular filtration rate (eGFR) of 120 ml/min/1.76 m², i.e., TVCL_R × (eGFR/120), where CL_R is the adjusted renal clearance, TVCL_R is the population average value for renal clearance, and the eGFR was determined from the cystatin C concentration (CysC) as 91.62 × (1/CysC^1.123) (20).

Allometric scaling using weight (WT) was also used on all volume and clearance terms, which were multiplied by (WT/70) and (WT/70)^0.75, respectively. One- and two-compartment models with first order absorption without lag time were assessed for both SDX and PYR. As few data exist to describe the absorption phase of both drugs, the absorption rate constant (k_a) was fixed to the previously published value for infants (18). Between-subject variability (BSV) was added to parameters for which it could be estimated reasonably from available data. As log_e concentration data were used, an additive model (representing proportional error) was used for residual unexplained variability (RUV).

In the development of the final models, we investigated the influence of the covariates dosing group, relative dose (milligram/kilogram), PMA, malaria status, concomitant treatment with amodiaquine, and initial hemoglobin concentration using the generalized additive modeling procedure within Xpose (http://xpose.sourceforge.net) and by inspection of correlation plots. Covariate relationships identified by this procedure were evaluated within the NONMEM model, and inclusion of the covariate required a significant decrease in OFV accompanied by a decrease in the BSV of that parameter. Correlations among BSV terms and weighted residuals (WRES) plots were also used in model evaluation.

Once a final model for SDX was obtained, the parameter estimates were fixed and an additional compartment was added in order to model NSX concentrations. In order to allow identifiability in the model, the percentage conversion of SDX to NSX was fixed to 60% based on the product information (35). The elimination of NSX was assumed to be entirely renal (25).The influence of the covariates was assessed on new model parameters using the method described above.

A bootstrap procedure using Perl-speaks-NONMEM (PSN) (http://psn.sourceforge.net) and the resulting parameters were then summarized as median and 2.5th and 97.5th percentiles (95% empirical confidence interval [CI]) to facilitate validation of the final model parameter estimates. In addition, stratified visual predictive checks (VPCs) and numerical predictive checks (NPCs) were also performed using PSN with 1,000 replicate data sets simulated from the original data set. NPCs stratified according to PMA were assessed by comparing the actual with the expected number of data points within the 20, 40, 60, 80, 90, and 95% prediction intervals (PI). The resulting 80% PI for drug concentrations were plotted with the observed data to assess the predictive performance of the model.

Statistical analysis.As previously reported in a study of SDX/PYR pharmacokinetics in pregnant versus nonpregnant women (26), and using estimates of centrality and variance for pharmacokinetic parameters from previous pediatric studies (11, 19, 32, 37, 40) and an assumed 20% attrition rate, a sample size of 35 in each group in the present study would be expected to show a >30% increase in the magnitude of any pharmacokinetic parameter in the double-dose group at α = 0.05 and β = 0.1. SPSS 17.0 (SPSS inc. Chicago, IL) was used for all statistical analysis unless otherwise specified. Data are summarized as mean ± standard deviation (SD) or median and interquartile range (IQR) as appropriate. Student's t test or the Mann-Whitney U test was used for two-sample comparisons. Categorical data were compared using either Pearson chi-squared or Fisher's exact test and multiple means by repeated measures analysis of variance (ANOVA). A two-tailed level of significance of 0.05 was used.

Patient characteristics.Seventy infants were enrolled between April 2008 and December 2008, with equal numbers in each dose group. Baseline subject characteristics are summarized in Table 2. The double-dose group received a significantly higher milligram/kilogram dose than the conventional dose group (P < 0.001) and was taller by a mean of 4.3 cm (P = 0.015). The double-dose group was also older (by a mean of 47 days) and heavier (by 0.4 kg) than the conventional dose group, but these differences were not statistically significant (P > 0.05).

Tolerability, safety, and efficacy.Both doses were well tolerated. There were no changes in symptoms in either group compared to predose profiles, including an absence of dermatological conditions. There were no significant changes in hemoglobin, or in plasma urea, creatinine, or CysC, over time. In the conventional dose group, there was a significant but transient mean fall in plasma albumin of 2 g/liter at day 2 (from 38 to 36 g/liter; P < 0.01), but there were no concomitant increases in plasma bilirubin or hepatic enzymes in either group.

Five infants with vivax malaria and one infant with a mixed Plasmodium vivax/P. falciparum infection at enrollment responded to treatment. Three other infants in the conventional dose group and two in the double-dose group were administered antimalarial drugs during follow-up at an external health care facility, and no blood smears were available for review. No other subjects became symptomatic during the study. Only two infants in the conventional dosing group and one in the double-dosing group who were aparasitemic at entry had a positive blood slide on day 28 (all for P. vivax). All were asymptomatic, and each was treated according to PNG national treatment guidelines.

Pharmacokinetic modeling.There were 248, 255, and 247 drug concentration measurements available for pharmacokinetic modeling for PYR, SDX, and NSX, respectively. There were four samples with PYR concentrations between the LOD and LOQ and a further four with concentrations below the LOD for PYR. In addition, seven samples were of insufficient volume for measurement of PYR after the SDX/NSX assay. There were no SDX or NSX concentrations below the LOQ, but NSX concentrations could not be determined in eight samples due to an unidentified interfering peak. For PYR, a two-compartment model was superior to a one-compartment model with a lower OFV (−87.081 versus −30.030) and a less-biased weighted residuals versus time (WRES) plot. The model parameters were k_a, CL_H/F, CL_R/F, central compartment volume of distribution (V₂/F), peripheral compartment volume of distribution (V₃/F), intercompartmental clearance (Q/F), HillCL, and MATCL₅₀. BSV was estimable on CL_T/F, V₂/F, and Q/F. As the correlation between the variability of V₂/F and Q/F was very close to 1, it was subsequently fixed to unity to assist with successful determination of the covariance matrix. None of the covariates tested improved the model significantly; therefore, the final model contained only the effects of PMA and WT anticipated from maturation and allometric scaling, respectively.

The final parameter estimates and the results of the bootstrap procedure for PYR are shown in Table 3. All model parameters had a bias of <11%. Goodness-of-fit plots for PYR are shown in Fig. 1. NPCs of the data showed good predictive performance, as did VPC plots of the observed drug concentrations and their 80% PI (the 10th and 90th percentile boundaries) stratified by dosing group (Fig. 2A and B). Post hoc parameter estimates are shown in Table 4. There was no difference between the two groups for any of these parameters except for AUC from 0 h to infinity (AUC_0-∞), which was significantly higher in the double-dose group (4,915 versus 2,844 μg/day/liter). Median steady-state volume (V_SS) for the combined study sample was 27.8 liters, and the half-lives at α and β phases (t_1/2α and t_1/2β, respectively) were 72.7 and 374 h, respectively.

• Open in new tab
• Download powerpoint

FIG. 1.

Goodness-of-fit plots for PYR showing observed versus model predicted concentrations (A) and individual predicted concentrations (B) (both log scale) and conditional weighted residuals versus time (C). For panels A and B, the solid gray line represents the line of identity, while the dashed black line represents the linear regression line of best fit; in panel C, the solid gray line represents the locally weighted scatterplot smoothing (LOESS) smoothed fit.

• Open in new tab
• Download powerpoint

FIG. 2.

Visual predicted check plots for PYR showing simulated 10th (short dashed line), 50th (dotted line), and 90th (solid line) percentile concentrations and observed concentration (log scale) data (gray open circles) versus time (log scale) for conventional dose (A) and double-dose (B) participants.

Initial modeling of SDX revealed that a one-compartment model was appropriate, as there was minimal bias in the WRES plot that was not improved when a two-compartment model was fitted. The model parameters were k_a, CL_H/F, CL_R/F, volume of distribution (V/F), HillCL, and MATCL₅₀. BSV was able to be estimated on CL_T/F and V/F. There was a significant relationship between relative dose (in milligrams/kilogram) and relative bioavailability which conformed to a power function; specifically, individual relative bioavailability = 1 × ([individual relative dose]/[average relative dose]^{effect parameter}). The value of the power effect parameter was −0.56, indicating that, when the dose is doubled, the bioavailability falls by 32.2%. The final parameter estimates and the results of the bootstrap procedure are shown in Table 5. With the exception of CL_R, all parameter estimates had biases of <13%. The median bootstrap value for CL_R was almost double the initial estimate (195%), demonstrating the difficulty in delineating the difference between and estimating the hepatic and renal clearance using this methodology. Goodness-of-fit plots for SDX are shown in Fig. 3. NPCs of the data showed good predictive performance, as did VPC plots of the observed drug concentrations and their 80% PI stratified by dose group in Fig. 4.

• Open in new tab
• Download powerpoint

FIG. 3.

Goodness-of-fit plots for SDX showing observed versus model predicted concentrations (A) and individual predicted concentrations (B) (both log scale) and conditional weighted residuals versus time (C). For panels A and B, the solid gray line represents the line of identity, while the dashed black line represents the linear regression line of best fit; in panel C, the solid gray line represents the LOESS smoothed fit.

• Open in new tab
• Download powerpoint

FIG. 4.

Visual predicted check plots for SDX showing simulated 10th (short dashed line), 50th (dotted line), and 90th (solid line) percentile concentrations and observed concentration (log scale) data (gray open circles) versus time (log scale) for conventional dose (A) and double-dose (B) participants.

An additional compartment was added to the final SDX PK model to incorporate the data for NSX. This resulted in three additional model parameters: volume of distribution of NSX (V_NSX/F), clearance of NSX (CL_NSX/F), and percentage of total SDX elimination representing conversion of SDX to NSX (%NSX). As these three parameters cannot be estimated simultaneously, %NSX was fixed to 60% based on published data (35). The estimates of V_NSX/F and CL_NSX/F are directly related to %NSX; therefore, the value of these parameters should be interpreted with caution. However, AUC and t_1/2 for NSX remain unchanged for different values of %NSX. V_NSX/F and CL_NSX/F were not influenced by any of the available covariates. Final parameter estimates and results of the bootstrap procedure are shown in Table 5. Bias was <5% for all NSX parameters, and NPCs and VPCs were performed on the NSX data set and indicated good predictive performance of the model (data not shown).

There were some significant differences between conventional and double-dose groups in the post hoc parameter estimates for both SDX and NSX (Table 6). These included expected differences in the AUC_0-∞ for both SDX and NSX, but also differences in the half-life and clearance for both drugs which were not revealed by the model covariate building stage. A higher clearance and lower half-life (t_1/2) in the double-dose group can be attributed to organ maturation, as these infants were older than those in the conventional dose group. The median t_1/2 of NSX for the combined study sample was shorter than that of SDX (8.9 versus 218 h). The percentage of the AUC_0-∞ of NSX compared to that for SDX was the same for both dose groups (approximately 5%).

Sigmoid E_max curves of hepatic maturity for SDX and PYR by PMA are shown in Fig. 5. They are closely related to MATCL₅₀ values of 318 days and 271 days for PYR and SDX, respectively. Of the 70 infants, 48 (69%) and 38 (54%) had an estimated hepatic clearance that was >90% of adult values for PYR and SDX, respectively.

• Open in new tab
• Download powerpoint

FIG. 5.

Maturation as a fraction of adult clearance for PYR (dashed line) and SDX (solid line) predicted from the PK model plotted against PMA. A box plot of the PMA in the recruited subjects is included to show its distribution in relation to maturation of clearance. d, days.