Pharmacokinetic and pharmacodynamic assessment of co-amoxiclav in the treatment of melioidosis

Abstract

Objectives: We conducted a prospective pharmacokinetic study of oral co-amoxiclav in patients with melioidosis to determine the optimal dosage and dosing interval in this potentially fatal infection.

Patients and methods: Serial plasma concentrations were measured after administration of two 1 g tablets of Augmentin^® (1750 mg of amoxicillin and 250 mg of clavulanate) to 14 adult patients with melioidosis. Monte Carlo simulation was used to predict the concentration of each drug following multiple doses of co-amoxiclav at different dosages and dose intervals. The proportion of the dose-interval above MIC (T > MIC) was calculated from 10 000 simulated subject plasma concentration profiles together with chequerboard MIC data from 46 clinical isolates and four reference strains of Burkholderia pseudomallei.

Results: The median (range) observed maximum plasma concentrations of amoxicillin and clavulanate were 11.5 (3.3–40.2) mg/L and 5.1 (0.8–12.1) mg/L, respectively. The median (range) elimination half-lives were 94 (73–215) and 89 (57–140) min, respectively. Simulation indicated that co-amoxiclav 1750/250 mg given at 4, 6, 8 or 12 hourly dosing intervals would be associated with a T > MIC of ≤50% in 0.7%, 2.8%, 8.6% and 33.2% of patients, respectively. Corresponding proportions for T > MIC of ≥90% were 95.8%, 78.6%, 50.2% and 10.8%, respectively.

Conclusions: The dosing interval for co-amoxiclav (750/250 mg) in melioidosis should not be greater than 6 h.

Monte Carlo simulation, Burkholderia pseudomallei, amoxicillin, clavulanate, PK/PD

Introduction

Melioidosis is a serious human infection caused by Burkholderia pseudomallei. Antibiotic treatment is protracted and involves 10–14 days of parenteral therapy followed by oral therapy to complete 20 weeks of treatment.1 Around 6% of patients develop recurrent infection in the first year after the primary episode, the majority representing treatment failure and relapse with the original strain. The single most important risk factor for relapse is choice and duration of oral antimicrobial treatment. Recommended oral treatment is a combination of trimethoprim/sulfamethoxazole and doxycycline.2 However, many patients have contraindications to, or intolerance of, one or more of these drugs.2 Oral co-amoxiclav is the recommended alternative. A dose of 1000 mg amoxicillin with 250 mg clavulanate given three times daily for 20 weeks is well tolerated and has a better adverse effects profile than the conventional regimen but it is associated with a higher relapse rate.3

The current dose of co-amoxiclav used in the oral treatment of melioidosis was introduced over 15 years ago based on in vitro studies of B. pseudomallei susceptibility,4,5 and represented the most intensive dosing regimen used at the time. In an evaluation of intravenous co-amoxiclav in severe melioidosis it was concluded that 1000/200 mg of amoxicillin/clavulanate should be given every 4 h in the absence of significant renal impairment.6,B. pseudomallei is intrinsically highly resistant to β-lactam antibiotics including amoxicillin. The presence of clavulanate is crucial for bacterial inhibition. As β-lactams exert no post-antibiotic effect against B. pseudomallei,7–9 the concentration of clavulanate needs to be maintained at therapeutic concentrations throughout treatment.

Although the theoretical basis for 4 hourly dosing of intravenous co-amoxiclav in the treatment of melioidosis is well founded, it goes against the current trend to prescribe the oral drug twice daily for infections such as upper and lower respiratory tract infections. A twice daily dosing schedule is based on clinical efficacy studies in regions of the world where B. pseudomallei infection does not occur,10 and pharmacokinetic data indicating that serum levels adequate to kill common pathogens such as Streptococcus pneumoniae are achieved throughout most of the 12 h dosing interval.11 Given this shift in practice, combined with concern that twice daily dosing is inadequate in our population of melioidosis patients, we have conducted a pharmacokinetic study of oral co-amoxiclav in melioidosis in order to determine optimal dose spacing. As the pharmacokinetic properties of amoxicillin and clavulanic acid are unlinked, and these in turn are unlinked to the antimicrobial susceptibility of the infecting organisms, we have extended this evaluation to the use of Monte Carlo simulation (MCS) to associate MIC data with pharmacokinetic profiles and thereby to predict the probabilities of attaining pharmacodynamic targets with different dosing regimens.12,13

Patients and methods

Adults (age ≥ 15 years) with culture proven melioidosis admitted to Sappasithiprasong Hospital between October 2002 to January 2003 and who were able to take oral medication were eligible for enrolment in the trial following informed consent. Exclusion criteria included: known or suspected hypersensitivity to penicillin, amoxicillin or clavulanate; treatment with co-amoxiclav within the preceding 48 h; pregnancy; haemodialysis or peritoneal dialysis; and the presence of any factors likely to affect the absorption of drug after oral dosing such as vomiting or treatment with histamine-2-receptor antagonists.

A single dose of two co-amoxiclav tablets (Augmentin^® 1 g; GlaxoSmithKline Ltd, UK) containing 1750 mg amoxicillin and 250 mg clavulanate was given after overnight fasting. Blood was collected into lithium heparin tubes at 0, 0.25, 0.5, 1, 1.5, 2, 3, 4, 6, 8 and 12 h after oral co-amoxiclav administration, for assay of amoxicillin and clavulanate plasma concentrations. The samples were centrifuged and separated immediately and the plasma stored at −80°C until shipment to the Department of Pharmaceutical Chemistry, Khon Kaen University for clavulanate analysis, and to the Faculty of Tropical Medicine, Mahidol University, Bangkok for amoxicillin analysis. The study was approved by the Ethics Subcommittee of the Ministry of Public Health, Royal Government of Thailand.

Measurement of plasma clavulanate concentration

Plasma clavulanate was extracted using a column switching liquid chromatography (LC)-system with an extraction column [Guard-Pak™ Inserts μBondapak™ C₁₈ (Waters, Milford, USA)] and an analytical column [Hypersil C18 (5 μm, 125 × 4 mm) (Agilent, Palo Alto, USA)]. In brief, 25 μL of internal standard (0.145 mg/mL salicylamide) and 50 μL of imidazole (33% w/v in water, adjusted to pH 6.8 with concentrated hydrochloric acid) were added to 250 μL of plasma. The sample was vortexed, immersed in a water bath at 35°C for 20 min and injected immediately into the extraction column. The extraction column was washed for 2 min before switching the valve. The analytical mobile phase consisted of 7% acetonitrile in 10 mM NaH₂PO₄, pH 2.5, containing 5 mM octanesulphonate sodium and 10 mM tetraethylammonium bromide pumped at a flow rate of 1.2 mL/min. The switching valve was operated for a period of 1 min to allow the sample to be injected onto the analytical column. The extraction column was conditioned before the next injection by washing with 3 × 50 μL of methanol. The injection volume used in the assay was 10 μL and the UV detector was set to 311 nm. The limit of quantification for the assay was 0.07 mg/L with 2.27% precision.

Measurement of amoxicillin plasma concentration

Amoxicillin plasma levels were extracted and measured as described previously.14 Briefly, 100 μL of plasma was precipitated with acetonitrile before centrifugation. The supernatants were then loaded manually onto acetonitrile preconditioned solid phase extraction (SPE) columns. The SPE columns were washed with acetonitrile and dried under vacuum. The elution was performed by slowly pushing HPLC grade water through the columns. The SPE eluates were gently mixed before injection into the LC-system. The detector was set at 230 nm. The samples were analysed on an Aquasil (150 × 4.6 mm) column (Thermo Electron Corporation, Cheshire, UK) using a mobile phase containing acetonitrile-phosphate buffer (pH 2.5; 0.1 M) (7:93, v/v) at a flow rate of 1 mL/min. A short wash out gradient from 7% to 30% acetonitrile was used between each sample to remove late eluting interfering peaks. The limit of quantification for the assay was 0.05 mg/L. The total assay precision during the analysis of the study samples was <6% at all quality control levels.

Pharmacokinetic analysis

Pharmacokinetic parameters of amoxicillin and clavulanate were determined using non-compartmental and compartmental analysis (WinNonlin version 4.1, Pharsight Corporation, Cary, USA). The actual sampling time was used in the analysis. To determine which compartmental model best fitted the data, the log likelihood function (Akaike information criterion) was used. The parameters characterized were: absorption rate constant (k_a), elimination rate constant (k_el), apparent clearance (CL/F), apparent volume of distribution (V/F, where F is the fraction of drug absorbed), time prior to the time corresponding to the first measurable (non-zero) concentration (T_lag), and area under the concentration–time curve (AUC).

Microbiology data

MIC values of amoxicillin and clavulanate for B. pseudomallei were derived from a previous study.4 Forty-six clinical isolates of B. pseudomallei from Northeast Thailand and four reference strains from the National Collection of Type Cultures were selected for study. Standard agar dilution MICs were determined as described previously, using a chequerboard design for all possible combinations of 2-fold dilutions of amoxicillin and clavulanate. Amoxicillin concentrations ranged from 0 to 16 mg/L and clavulanate concentrations ranged from 0.03 to 16 mg/L. The majority of strains required at least 0.5 mg/L of clavulanate for inhibition of growth, irrespective of the amoxicillin concentration.

The effect of amoxicillin concentration on growth inhibition of B. pseudomallei depends on the clavulanate concentration. The maximum effect does not occur when the clavulanate concentration is <1 mg/L (Figure 1). Figure 2 shows the plot and fitted curve between amoxicillin concentration and clavulanate concentration for inhibition of each B. pseudomallei strain derived directly from the result of chequerboard MIC. Using TableCurve 2D Version 5.01.05 (SYSTAT Software Inc., 2002, CA, USA), the observed clavulanate concentrations versus amoxicillin concentrations to inhibit growth for 50 strains of B. pseudomallei were best fitted using the exponential formula [median residual standard error of fit (range) = 0.34 (0.05–3.96), median adjusted R² (range) = 1.00 (0.61–1.00)]:

$$C_{\text{clav}}(\text{mg}/\text{L})=a+b\text{exp}\left[\frac{-C_{\text{amox}}(\text{mg}/\text{L})}{c}\right]$$

(1)

where C_clav is the clavulanate concentration needed for a given amoxicillin concentration (C_amox); a, b and c are constants. The distributions of a and c were normal after square root [mean √a ± SD = 0.68 ± 0.23; P = 0.09, Kolmogorov–Smirnov (KS) test] and reciprocal root (mean 1/√c ± SD = 1.64 ± 0.28; P = 0.41, KS test) transformations, respectively. b was divided into two normally distributed ranges; 22/50 (44%) were more than 18 (mean ± SD = 31.45 ± 0.28; P = 0.51, KS test) and 28/50 (56%) were less than 18 (mean ± SD = 15.78 ± 0.29; P = 0.72, KS test).

Figure 1

Amoxicillin concentration effect curves using a sigmoid maximal attainable effect (E_max) model plotted at different clavulanate concentrations. The effect is cumulative inhibition of 50 tested B. pseudomallei isolates. The label on each line represents clavulanate concentration doubling from 0.03125 to 16 mg/L.

Figure 2

Exponential-fitted curve of inhibitory amoxicillin and clavulanate concentration for 50 strains of B. pseudomallei: each fitted line represents a single isolate.

Monte Carlo simulation (MCS)

To characterize drug concentration profiles in a population of patients infected with organisms of different susceptibilities we performed a MCS. A sample size of 10 000 was used for the MCS. The simulations were performed using S-PLUS 6.0 Professional Release 2 (1988–2001 Insightful Corp., Seattle, WA, USA). The study patients were simulated using the distribution of ‘real melioidosis patient’ pharmacokinetic parameters (as derived from the above). Each patient was infected with a B. pseudomallei strain which was simulated from the distribution of constants a, b and c described above. The simulations were generated for different dosages of co-amoxiclav at 4, 6, 8 and 12 h drug intervals repeatedly. The amoxicillin concentrations needed in the presence of the corresponding clavulanate concentrations to inhibit assigned simulated B. pseudomallei growth were calculated at each time point using Equation 1.

A one-compartment model with lag time, first-order absorption and first-order elimination was selected as the most suitable pharmacokinetic model for both amoxicillin and clavulanate. The basic assumptions, based on previous studies, are that the drug is eliminated by first-order kinetics, there is no dose-dependency and that the pharmacokinetics of the drug after a single dose (first dose) are not altered after taking multiple doses.

The predicted curves of the concentrations of amoxicillin and clavulanate over time curves after multiple oral administrations of co-amoxiclav were plotted using the formula for a one-compartment model with lag time:

where C(t_n) is the concentration at any given time (t_n) after the n_th dose, D is the fixed amount of drug taken at each time interval (τ) (4, 6, 8 and 12 hourly) in minutes (125 and 250 mg for clavulanate, and 500, 750, 1000 and 1750 mg for amoxicillin), and k_a, k_el, T_lag and V/F of clavulanate and amoxicillin were derived from the observed distribution of the pharmacokinetic profiles of the 14 melioidosis patients studied. The percentage of time above MIC (%T > MIC) at each dose and dosing interval for each simulated patient was calculated, and histograms were plotted for each group of patients.

Statistical methods

Groups were compared using ANOVA. The ‘normality’ of distributions was assessed with the KS test. Correlations were performed by Pearson's method on normally distributed data. Correlations between the pharmacokinetic variables k_el, T_lag and V/F of amoxicillin and clavularate were incorporated in the MCS using simple linear regression models.

Results

Fourteen patients with culture proven melioidosis were recruited into the pharmacokinetic study. Six of 14 patients were septicaemic including three patients with disseminated disease (septicaemia and two or more organs involved). All but one patient had underlying diseases; 12 had diabetes mellitus, and one chronic liver disease from alcohol abuse. One patient died 3 weeks after the study, and one had a treatment failure 2 months after primary infection. Median (range) time to fever defervescence was 17.5 (10–40) days.

There was considerable inter-individual variability in pharmacokinetics. Peak amoxicillin concentrations ranged between 3.3 and 40.2 mg/L and peak clavulanate concentrations ranged between 0.8 and 12.1 mg/L. The median ± SD terminal elimination half-lives (t_1/2) of amoxicillin and clavulanate were similar; 94 ± 40 (range 73–215) and 89 ± 24 (range 57–140) min, respectively. The pharmacokinetic parameters derived from the one-compartmental analysis with lag time are shown in Table 1. Logarithmic transformations of parameters used in the simulations were normally distributed. The following values were used: mean (SD) of log(clavulanate k_a in h⁻¹) = −0.02 (0.52); mean (SD) of log(amoxicillin k_a in h⁻¹) = −0.54 (0.36); mean (SD) of log(clavulanate k_el in h⁻¹) = −0.50 (0.30); mean (SD) of log(clavulanate V/F in L) = 4.39 (0.30); and mean (SD) of log(clavulanate T_lag in min) = 1.49 (0.23), P ≥ 0.15 for all by KS test. Using a simple regression model, the estimates of k_el, T_lag and V/F of amoxicillin and clavulanate were all correlated significantly for k_el (R² = 0.61, P = 0.02), for T_lag (R² = 0.70, P = 0.01) and for V/F (R² = 0.84, P < 0.001), so k_el, T_lag and V/F of amoxicillin were calculated after k_el, T_lag and V/F for clavulanate for each patient had been simulated.

Table 1

Pharmacokinetic parameters of clavulanate and amoxicillin in patients with melioidosis following oral administration of 1750/250 mg co-amoxiclav, estimated using a one-compartment model with lag time (n = 14)

Parameters	Amoxicillin median (range)	Clavulanate median (range)
ka (/h)	0.55 (0.32–1.0)	0.84 (0.48–2.7)
kel (/h)	0.53 (0.26–0.99)	0.58 (0.41–1.0)
Tlag (min)	23.3 (9.5–58.0)	29.0 (10.8–70.0)
Tmax (min)	140.2 (83.6–243.3)	102.3 (73.6–203.2)
Cmax (mg/L)	11.5 (3.3–40.2)	5.1 (0.8–12.1)
AUC (mg·h/L)	62.1 (13.3–339.5)	15.3 (2.9–39.2)
V/F (L/kg)	1.1 (0.3–6.3)	0.4 (0.2–2.5)
CL/F (mL/min/kg)	9.8 (1.9–49.6)	4.6 (1.9–32.5)

k_a, first-order absorption rate constant; k_el, first-order elimination rate constant; T_lag, time prior to the time corresponding to the first measurable (non-zero) concentration; T_max, time to maximum concentration; C_max, maximum concentration; AUC, area under the concentration–time curve; V/F, apparent volume of distribution; CL/F, apparent clearance; F, fraction of drug absorbed.

Examples of the concentrations of each drug over time after administration of 1000/250 mg of amoxicillin/clavulanate every 8 h, and the amoxicillin concentrations needed in the presence of the corresponding clavulanate concentrations in patient blood to inhibit B. pseudomallei growth in simulated patients infected with simulated organisms are shown in Figure 3.

Figure 3

Example of simulated concentration curves of amoxicillin (dashed line) and clavulanate (thin line) over time in a simulated patient receiving 1000/250 mg of co-amoxiclav 8 hourly. Thick lines represent the amoxicillin concentration needed to inhibit B. pseudomallei growth in the presence of the corresponding clavulanate plasma concentration.

The mean %T > MIC for the 4 hourly dosing interval was 96–99% depending on the dosage of amoxicillin used; for the 6 hourly dosing interval it was more than 90% only if the dosage was 1000/250 or higher (Figure 4). The %T > MIC values were significantly different between the four dosing intervals irrespective of the dosage of amoxicillin taken (P < 0.0001 for each dosage of amoxicillin, by ANOVA).

Figure 4

Mean (95% CI) of %T > MIC after administration of co-amoxiclav 4, 6, 8 and 12 hourly from 10 000 simulated melioidosis patients, each line representing one dosage of co-amoxiclav (1750/250, 1000/250, 750/250 and 500/250 mg).

Depending on the dose administered between 0.2% and 2.1% of simulated patients would not have had sufficient clavulanate to make amoxicillin effective against B. pseudomallei (%T > MIC = 0), despite a very high dose of amoxicillin (Table 2). Comparing all four dosing intervals, 4 hourly treatment gave the best coverage with more than 90% of patients having a T > MIC of greater than 90%.

Table 2

Percentage of 10 000 simulated melioidosis patients having 0%, ≤50%, ≥90% and 100% T > MIC after the oral administration of different dosages of co-amoxiclav (1750/250, 1000/250 and 750/250 mg) at different dosing intervals (4, 6, 8 and 12 hourly)

	Proportion of patients achieving target values (%)
	1750/250 mg co-amoxiclav				1000/250 mg co-amoxiclav				750/250 mg co-amoxiclav
	Dosing interval (hourly)
T > MIC	4	6	8	12	4	6	8	12	4	6	8	12
0%	0.2	0.3	0.4	0.6	0.2	0.6	1.1	0.8	0.5	0.9	1.6	2.1
≤50%	0.7	2.8	8.6	33.2	0.9	3.9	12.0	41.8	1.3	5.3	14.9	48.1
≥90%	95.8	78.6	50.2	10.8	94.8	73.8	42.1	6.7	93.2	68.8	36.5	4.3
100%	93.5	70.1	39.4	6.1	92.4	65.7	31.5	3.2	90.9	59.7	26.2	1.9

Uncertainty and sensitivity analysis was performed. The coefficient of variation for the %T > MIC of 10 000 simulated patients given 1000/250 mg of amoxicillin/clavulanate 8 hourly was 28.4%. Variation in five parameters; the three constants for the organism MIC (a, b and c), and both amoxicillin and clavulanate k_a values, influence the model prediction (partial rank correlation coefficients for each P < 0.05).

Discussion

In this study, a single oral dose of 1750/250 mg co-amoxiclav produced peak concentrations of amoxicillin and clavulanate at median times of 140 and 102 min, respectively, after administration, consistent with previous reports that oral amoxicillin T_max varies between 1–6 h.15 The median C_max values were 11.5 and 5.1 mg/L for amoxicillin and clavulanate, respectively, also consistent with a previous report,15 but there was considerable inter-individual variation over a more than 10-fold range for both drugs.

This study shows that the %T > MIC following 12 hourly dosing intervals of co-amoxiclav were insufficient in patients with melioidosis. A new ‘pharmacokinetically enhanced’ adult tablet formulation of 2000/125 co-amoxiclav has become available that contains even less clavulanate; the mean T > MIC [95% confidence interval (CI)] for this formulation was predicted to be only 50.2% (range 49.7–50.6) for a twice daily dosing interval (data not shown). Patients with melioidosis are currently prescribed an 8 hourly, 1000/250 mg regimen. We predict that 12% of these patients would have less than 50% T > MIC and only 42% of patients would achieve more than 90% T > MIC with this regimen. This may well explain the relatively high rate of relapse in patients with melioidosis who are treated with co-amoxiclav.

We conclude that the use of a 12 h interval of co-amoxiclav for the oral eradication treatment of melioidosis is not appropriate, and an 8 hourly regimen is sub-optimal. A 4 hourly dosing interval of co-amoxiclav gives the highest %T > MIC and is predicted to be the best regimen, although this dosing interval for an oral drug is not practical and intolerance may occur. We predict that approximately 95% of patients should achieve >50% T > MIC, and nearly 70% should have ≥90% T > MIC with 6 hourly administration of 750/250 mg. Further evaluation is required to determine the clinical efficacy of this regimen.

Transparency declarations

Conflicting interests: none to declare.

We are grateful to the directors of Sappasithiprasong and Srinagarind Hospitals and the medical and nursing staff of the Department of Medicine, Nongluk Getchalarat, Premjit Amornchai, Gumphol Wongsuvan, Sayan Langla and Jintana Suwannapruk for their support. We also thank Dr Kasia Stepniewska and Mr Noprath Singtoroj for their kind help with the mathematical modelling. This study was part of the Wellcome Trust-Mahidol University-Oxford Tropical Medicine Research Programme, funded by the Wellcome Trust of Great Britain.