INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Although quinine is one of the oldest drugs in the pharmacopoeia, the optimum usage of quinine in children with severe malaria is still debated (12, 29). The choice of route and dose of quinine for children with severe malaria will vary depending on circumstances and particularly on the capability of administering intravenous infusions reliably. Quinine is the drug of choice for the management of severe malaria in most areas of the world, and it is frequently deployed in conditions where intravenous infusions cannot be rapidly established or reliably monitored.

Recent pharmacokinetic studies using African children have revived the intramuscular route as an alternative, cheaper, practicable, and potentially safer route for quinine administration (15, 25, 27). However, classical pharmacokinetic studies are not always applicable to populations at highest risk of death from Plasmodium falciparum infection. One of the most important risk factors that identify these children is the complication of lactic acidosis (plasma or whole blood lactate concentration of 5 mmol/liter) (11). Dichloroacetate (DCA) is a potential treatment for malaria-associated lactic acidosis (7).

We recently conducted a randomized, double-blind, placebo-controlled investigation to test the hypothesis that treatment with the lactate-lowering drug DCA, given with quinine, significantly improved morbidity and mortality in Ghanaian children with lactic acidosis due to severe P. falciparum malaria infection. This report describes the population pharmacokinetics of a loading dose of intramuscular quinine dihydrochloride (20 mg/kg of body weight) in 120 patients. The size of this study also allows assessment of covariables that may be important in influencing quinine kinetics. The major clinical results of the study will be presented elsewhere.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Patients. The study was carried out at the Komfo-Anokye Teaching Hospital in Kumasi, Ghana, and was approved by the Committee of Research, Publication and Ethics of the School of Medical Sciences, University of Science and Technology, Kumasi, Ghana, and the Institutional Review Board at the University of Florida. Between June 1997 and February 1999, 1,654 children with a suspected case of malaria were referred to the study team. Patients were examined by a member of the team, and samples were taken to measure glucose or lactate (100 µl of whole blood or plasma), hematocrit, and parasitemia. After written, informed consent from parents or guardians was obtained, children were entered into this study if they fulfilled the following inclusion criteria: age of 12 months to 10 years inclusive, positive blood film for asexual stages of P. falciparum, and plasma venous or capillary blood lactate concentration of 5 mmol/liter. Exclusion criteria were pregnancy, assessed by a urine -human chorionic gonadotropin test in patients aged 10 years, and brain death, determined by clinical examination. Cerebral malaria was defined using the Blantyre Coma Score (BCS) (2 on a 5-point scale) (16). A lumbar puncture was performed on all patients with coma (BCS  2) to exclude the possibility of meningitis or encephalitis. The cerebrospinal fluid was analyzed immediately by microscopy and cultured.

Quinine and DCA treatments. Quinine (20 mg/kg as a loading dose; diluted 1:1 [vol/vol] in water for injection, giving a final concentration for quinine dihydrochloride of 150 mg/ml) (Rotomed [Rotex, Trittau, Germany] or quinine dihydrochloride [Martindale, Romford, United Kingdom]) was given to patients who had not received quinine treatment in the past 24 h. Half the loading dose was injected intramuscularly into the anterior aspect of each thigh. Maintenance doses of quinine (10 mg/kg, intramuscularly) were given into alternate thighs every 12 h (after dilution as before) until the patient was able to tolerate oral quinine. Patients were randomized to receive either DCA, 50 mg/kg formulated in 100 mg/ml as described previously (22), or normal saline (as placebo) after the first dose of quinine. DCA or placebo was given as a single infusion over 10 min by manual injection timed with a stopwatch.

Supportive treatment. Patients were managed according to published guidelines (17), and supportive therapy included the following:

(i) Correction of possible thiamine deficiency. To correct for a possible deficiency (10), all patients received a single intramuscular dose of thiamine (100 mg; Martindale) before DCA was administered.

(ii) Prevention of hypoglycemia. A constant infusion of glucose (3 mg/kg/min, 5% dextrose) (Intravenous Infusions Ltd., Koforidua, Ghana) was administered to minimize the risk of hypoglycemia. Hypoglycemia (capillary or venous blood glucose concentration of 2.2 mmol/liter) was treated with a slow infusion of 50 or 25% glucose (1).

(iii) Prevention and treatment of seizures. All patients with cerebral malaria received phenobarbitone (7 mg/kg, intramuscularly) (28), and seizures after admission were treated with diazepam (0.3 mg/kg, intravenously).

(iv) Correction of anemia. A blood transfusion (20 ml/kg) was given to children with a hematocrit of 15%, and careful attention was given to maintenance of intravascular volume.

Monitoring and sampling. Vital signs (respiratory rate, pulse, systolic and diastolic blood pressure, and temperature), BCS, hematocrit, glucose, and lactate were monitored every 4 h for 24 h after admission or more frequently, if clinically indicated. Venous blood samples (2 ml) were taken at baseline and at two time points between 0 and 12 h after the first DCA dose. Randomization of sampling times was done at 0, 10, 20, and 40 min and at 1, 2, 4, 8, and 12 h after DCA administration for collection of plasma to determine blood gas, metabolite, DCA, and quinine levels. As DCA was given a few minutes after quinine, this individual time difference for each patient was accounted for in analyses. Samples were collected in heparinized tubes (15 IU; Leo Pharmaceuticals, Risborough, United Kingdom). Plasma was separated within 15 min and stored and transported frozen (<70°C) for analysis of quinine. Thick and thin blood films were also prepared at each pharmacokinetic time point and at every 4 h after admission, up to 24 h. Thereafter, sampling was performed every 6 h until parasitemia cleared.

Analysis of quinine. Quinine was quantified in plasma samples by a previously reported high-performance liquid chromatography (HPLC) method (8). Briefly, plasma (100 µl) was mixed with quinidine hydrochloride (10 µl of a 200-µg/ml solution), ammonia (60 µl), and water (230 µl). To this solution, methylene chloride (500 µl) was added, and the mixture was vortexed (45 s). The two layers were separated by centrifugation (3,000 × g, 10 min). The organic layer was vortex mixed with HCl (200 µl, 0.15 M) and was centrifuged again. The upper aqueous layer was separated, and sample (5 µl) was injected and analyzed by HPLC. HPLC was performed with a Hewlett-Packard 1100 Series system (Palo Alto, Calif.) made up of a G1322A Vacuum Degasser, a G1311A quaternary pump, a G1313A autosampler, and a G1315 diode array UV-visible detector. This instrument was coupled with a Hewlett-Packard Vectra Windows NT Workstation controlling pumps and detectors. A Hewlett-Packard Hypersil BDS-C-18 column (125 by 4 mm; inside diameter, 5 µm) was used for isocratic chromatographic separation with a mobile phase consisting of acetonitrile and 1% triethylamine in water adjusted to pH 3 with o-phosphoric acid (15%:85% [vol/vol]). The flow rate was kept at 1 ml/min. Quinine and quinidine were detected at 254 nm, and retention times for quinine and quinidine were 5.2 and 4.5 min, respectively. The assay method was linear in the range of 2.5 to 40 µg of quinine/ml. The inter- and intrabatch bias and relative standard deviation were <5% at all concentrations.

Data analysis. (i) Pharmacostatistical model. Population modeling was developed by an expectation-and-maximization algorithm (described below) using P-Pharm software (InnaPhase, Champs sur Marne, France). Expectation-and-maximization analysis is an iterative two-step process suitable for computing maximum-likelihood (ML) population estimates of primary pharmacokinetic parameters. The algorithm computes the ML estimates by (i) an expectation step (E step), in which the individual parameters in the subjects are estimated by a Bayesian approach, given the present population parameters and the individual concentration-time data, and (ii) a maximization step (M step), in which the population parameters are estimated by ML methodology given the present individual-parameter estimates. These two steps are iterated up to convergence or until the fractional changes of fixed, random, and residual error variance parameters between two consecutive iterations become less than 0.001.

(ii) Model validation. The model was validated for standardized concentration prediction error (SCPE) and standardized parameter prediction error (SPPE) for clearance (CL), intercompartmental clearance (Q), volume of distribution in the central compartment (V₁), and volume of distribution in the peripheral compartment (V_p). SCPE for each concentration was calculated using the relationship:
C_obs represents the observed concentrations, and SD (C_exp) represents the estimated standard deviation on the expected values computed, using all sources of random variability including residual error. For each pharmacokinetic parameter (CL and V₁), the normalized SPPEs were computed as
P_pop is the population pharmacokinetic parameter, and SD (P_pop) is the corresponding standard deviation. To assess the posterior distribution properties of the residuals and the individual parameters, a t test was used to compare the mean of SCPE and SPPE to zero and the Kolmogorov-Smirnov test was used to compare the sampled distribution to the expected one [n (0,1)].

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Patients. Patients with malaria-associated lactic acidosis (n = 124) were enrolled. Sixteen patients (12.9%) died, the majority of them (11 of 16; 69%) within 24 h of admission. Demographic, clinical admission, and laboratory features of patients are summarized in Table 1. All patients had one or more defining features of severe malaria, including hyperlactatemia, cerebral malaria (n = 38; 31%), acidemia (39 of 81; 48%), or respiratory distress (n = 67; 56%). Fifteen patients who did not have cerebral malaria on admission became comatose during the first 24 h of quinine treatment. Half the patients (n = 62) were randomized to receive DCA. More detailed descriptions of these patients' clinical courses will be published elsewhere. Ninety-two patients (74%) had a history of antimalarial treatment (1 with amodiaquine, 1 with artesunate, 10 with an unspecified antimalarial, and the remainder with chloroquine).

Population modeling of quinine pharmacokinetics. Each patient contributed 1 to 3 plasma samples, depending on the actual time of blood sampling. Five patients with a history of prior quinine pretreatment had measurable baseline quinine levels. Three patients with baseline quinine levels above 2 µg/ml (range, 6.6 to 13.4 µg/ml) and one patient without measurable quinine levels for up to 1 h after admission were excluded from the population analysis. The patient without measurable quinine survived. Another surviving patient's quinine concentration was >60 µg/ml (0.5-h sample), and this time point was also excluded from analysis. Plasma samples (n = 282) from 120 patients were available for population pharmacokinetic analysis of quinine. Based on earlier analysis and published reports (12), a two-compartment model with first-order absorption and elimination was used to develop a population pharmacokinetic model for quinine. Total CL, Q, V₁, V_p, and absorption rate constant (k_a) were used as the model parameters. The parameter k_a was fixed at 3.5/h (absorption half-life of 12 min) (8) to arrive at reliable estimates of CL and V₁. Intersubject variability in CL had a log-normal distribution, while the values for intersubject variability for Q, V₁, and V_p showed normal distributions. The distribution of residual errors was best explained by a multiplicative (proportional to the observed concentration) error model.

View larger version (16K): [in this window] [in a new window]   FIG. 1.   (a) Population pharmacokinetic profile with 95% confidence intervals for quinine after a single intramuscular dose (20 mg/kg). (b) Correlation between the observed quinine plasma levels and the Bayesian estimates predicted by the model.

View larger version (21K): [in this window] [in a new window]   FIG. 2.   Histograms and cumulative distribution graphs for normalized residuals (A), CL (B), V1 (C), Q (D), and Vp (E).

Analysis of covariates. Based on a preselected critical percentage (5%) of the F distribution to assess the contribution of a covariate in multiple, stepwise, linear regression, none of the following covariates influenced weight-normalized CL and V₁ : age, sex (0 = male, 1 = female), admission arterial pH, pO₂, pCO₂, body temperature, parasite count, cerebral malaria (0 = no, 1 = yes), phenobarbital treatment, diazepam treatment (0 = no, 1 = yes), and previous antimalarial treatment (0 = no, 1 = yes). When all of these factors were forcibly included as covariates and the data set was reanalyzed, weight, sex, age, hematocrit, and DCA (0 = placebo, 1 = DCA) treatment were selected as covariates for CL. Covariates for other parameters were weight for V₁; hematocrit and cerebral malaria for Q; and cerebral malaria, DCA, and hematocrit for V_p. Intersubject variability estimates of all the parameters were lower with the covariate model (Table 3). However, there were no significant changes in the ML ratio and AIC. Analysis of the diagnostic graphs also did not indicate any improvements in the predictions. Hence, it was inferred that none of these patient-specific factors could be used reliably as covariables to explain interpatient variability in CL and V₁.

Efficacy and tolerability of quinine. Figure 3a represents parasite CL data following quinine as a survival curve. The time taken for half the patients to clear parasites completely is 43 h. Figure 3b shows normalized parasite CL kinetics (median PC₅₀ = 16 h and PC₉₀ = 27 h, where PC₅₀ is the time taken for parasite numbers to fall by 50% of baseline values and PC₉₀ is the time required for parasite numbers to fall by 90% of baseline values). The median (interquartile range) parasite clearance time was 48 (36 to 54) h.

View larger version (13K): [in this window] [in a new window]   FIG. 3.   (a) Proportion of cases remaining positive for P. falciparum following admission. (b) Median change in parasitemia after admission. Numbers of patients contributing to each value are also given beside the data points.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Debate about the best way to administer quinine has continued since it was first used to treat severe malaria (12, 13, 29). Ross and others, working in India (14, 19), cautioned against using "strong" quinine solutions, because they might cause necrosis at the intramuscular injection site and might not be adequately absorbed. Diluted quinine was advocated as it was "practically painless and rapidly effective" (3). Fletcher provided a lucid review of clinical and laboratory experience on intramuscular quinine and suggested that it should be reserved for "patients who are most dangerously ill with malaria" (5). More recent studies have examined absorption of undiluted (300-mg/ml) (27) and diluted (150-mg/ml) (8) quinine and 60-mg/ml (18) quinine following a loading dose (20 mg of salt/kg). Dilution of quinine decreases the absorption half-life (mean ± standard deviation) from 38 ± 25 min to 10.4 ± 9 min (for 150-mg/ml quinine) and 8.7 ± 7.8 min (60-mg/ml quinine). There was no evidence of major local or systemic toxicity in these studies. A larger prospective evaluation of intramuscular (n = 57) versus intravenous (n = 47) quinine confirmed similar efficacy, safety, and blood levels of quinine for both routes (21) but did not include a pharmacokinetic analysis. Parasite CL estimates after quinine in our population are similar to those reported for children with cerebral malaria in The Gambia: median (with the interquartile range in parentheses) parasite CL time was 48 (36 to 54) h in our study, compared with 60 (48 to 72) h in the study from The Gambia (26).

This study on the pharmacokinetics of intramuscular quinine in children with severe malaria exceeds other reports in size and detail. Population estimates for pharmacokinetic parameters presented here are consistent with previous, smaller, classical pharmacokinetic studies. For example, quinine CL is estimated here as equal to 0.05 liter/h/kg. In previous studies, estimates ranged from 0.027 to 0.0816 liter/h/kg (reviewed in reference 12). In this study k_a was fixed in order to obtain reliable estimates for other parameters. The suitability of distribution models is reflected in Fig. 1b, which shows the correlation between observed plasma concentrations of quinine and those predicted by Bayesian analysis.

Quinine (20 mg/kg) is absorbed rapidly and reliably after dilution and intramuscular injection into both anterior thighs. Mean peak plasma concentration versus time profiles between 15 and 20 µg/ml (Fig. 1a) are within the notional "therapeutic range" for quinine (12). The relatively broad confidence intervals around this value may reflect heterogeneity in quinine disposition rather than variability due to disease, as children were selected for study by strict and objective entry criteria and represent patients with the highest mortality (9, 11). The absence of any identifiable relationship between clinical and laboratory variables and population pharmacokinetic indices supports this suggestion. In particular, it is reassuring that intramuscular quinine is reliably absorbed, even in patients who may have severe acidemia, cerebral malaria, or anemia. Concomitant use of other medications (DCA, phenobarbital, and diazepam) also did not affect first-dose pharmacokinetics of quinine. Quinine is metabolized to 3-hydroxyquinine predominantly by the hepatic CYP4503A4 system (12). Since the expression of this enzyme system exhibits considerable interindividual variation (ranging from between 10 and >60% total hepatic cytochrome P450 activity) (6), genetic factors may contribute to quinine's variable disposition.

There was no major local toxicity associated with our regimen of intramuscular quinine. Local side effects were self-limiting, and none required surgical intervention, in contrast to findings from The Gambia, where some children (5 of 288) (1.7%) given intramuscular quinine (diluted 1:5) needed drainage of abscesses (26).

Patients who had hypoglycemia prior to admission were at highest risk of postadmission hypoglycemia, despite receiving a constant infusion of glucose (3 mg/kg/min). Taken together with observations on glucose kinetics in children with severe malaria receiving quinine (1), children who are hypoglycemic on admission to the hospital may require larger amounts of glucose (up to 6 mg/kg/min) than what we routinely used to prevent hypoglycemia. In any case, this high-risk group should be monitored particularly carefully. These observations are consistent with studies in Kenyan children, where admission hypoglycemia (n = 27 of 171; 16%) also identified children at risk of postadmission hypoglycemia (n = 9; relative risk = 5.33 [95% confidence interval = 2.33 to 12.2; P < 0.0001]). As in this study, some children who were euglycemic on admission subsequently developed hypoglycemia despite receiving dextrose (4). Postadmission rates of hypoglycemia were ~15% in a large Gambian study (26). These observations contrast with those from Malawi indicating that glucose infusions prevented postadmission hypoglycemia after quinine treatment (23) and confirm that blood glucose should be monitored regularly, whenever practicable, in children with severe malaria.

Figure 4 displays predicted quinine pharmacokinetic profiles of three different doses of quinine, based on our population analysis. Both 10- and 15-mg/kg doses are likely to undertreat a significant proportion of children in areas where parasites are not fully quinine sensitive. The safety of a 20-mg/kg loading dose of quinine and the potential for undertreatment suggest that this dose should be preferred in the management of severe malaria in African children. The risks of undertreatment with quinine were recently highlighted by a retrospective analysis that showed a significantly higher mortality rate in patients who received a 10-mg/kg dose than in those who received a 20-mg/kg dose of quinine (24). A few patients (5%) in our study had quinine levels of >30 µg/ml, but none suffered toxicity. Only one patient had levels below 5 µg/ml 12 h after the first dose of quinine. Many factors must be considered in choosing between intravenous and intramuscular routes for quinine use. Ease of administration, the lack of requirement for immediate intravenous access, more expensive fluid administration sets, predictable pharmacokinetics, usefulness in severe malaria, and safety favor the intramuscular route. However, the intramuscular route has some potential disadvantages, in particular the risk of infection (rarely tetanus or poliomyelitis) that may "seed" to areas of muscle necrosis (2, 5, 30). Both parenteral routes for quinine administration may be associated with hyperinsulinemic hypoglycemia and the risk of transmitting blood-borne infections. The intravenous route incurs a risk of major quinine toxicity if infusion rates cannot be reliably managed. Thus, policies governing selection of one route over another must take into account these considerations. Our findings should increase confidence in the efficacy and safety of intramuscular quinine as the first choice for management of severe malaria in children.

View larger version (13K): [in this window] [in a new window]   FIG. 4.   Predicted pharmacokinetic profiles of four different doses of quinine based on population estimates: 20 mg/kg (), 10 mg/kg (), 15 mg/kg (--), 20 mg/kg + 10 mg/kg at 12 h (-).