Naïve-pooled pharmacokinetic analysis of pyrazinamide, isoniazid and rifampicin in plasma and cerebrospinal fluid of Vietnamese children with tuberculous meningitis

The study was a prospective observational study of 100 patients under 15 years of age presenting with suspected TBM to the Pham Ngoc Thach Hospital for Tuberculosis and Lung Diseases (PNT) in Ho Chi Minh City. This hospital is the tertiary referral centre for patients with TB from the whole of southern Vietnam (population around 45 million). All patients aged ≤15 years presenting to PNT with meningitis symptoms and signs (e.g. fever, headache, neck stiffness, vomiting, confusion, coma, convulsions, cranial nerve palsies, hemiplegia or paraplegia) and who were suspected to have TBM by their attending physician were eligible to enter the study. Patients were graded for severity at the point of study entry. For children more than 5 years of age severity was graded using the modified UK Medical Research Council (MRC) criteria, based on Glasgow Coma Score (GCS). For children less than 5 years of age, severity was graded according to the Blantyre Coma Score (BCS).

Ethical approval for the study was granted from the ethical review board of Pham Ngoc Thach Hospital, Viet Nam, Health Services of Ho Chi Minh City and the Oxford Tropical Research Ethics Committee (OxTREC), UK. Written informed consent was obtained from the parents or guardians of all participants.

Children diagnosed with TBM were treated with the standard Vietnamese National TB Programme (NTP) 8-month regimen. At the time the study was initiated, Vietnam had not yet adopted the new doses recommended by WHO [11] in 2011. Therefore, children received once daily treatment with pyrazinamide (25 mg/kg), isoniazid (5 mg/kg), rifampicin (10 mg/kg), ethambutol (15 mg/kg) and streptomycin (15 mg/kg) for the first 2 months; then isoniazid, rifampicin, ethambutol and pyrazinamide for the third month; and then isoniazid, rifampicin and ethambutol for the continuation phase of 5 months. Dexamethasone was used as adjunctive therapy for the first 6 (MRC/BCS grade I disease) or 8 (MRC/BCS grades II or III disease) weeks of treatment in all patients according to standard practice [12].

Locally produced adult fixed dose combination (FDC) tablets were used for treatment (Turbezid® Namha Pharmaceutical Joint-Stock Company). These consisted of film-coated and scored tablets containing 150 mg of rifampicin, 75 mg of isoniazid and 400 mg of pyrazinamide. Each tablet was designed for ≥ 15 kg of body weight and used a drug ratio compatible with the adult TB regimen. Tablets were split in thirds for the lightest patients (<5 kg), halves (5–7.5 kg), 2 thirds (7.5–10 kg) or used as whole plus extra split for over 15 kg. Ethambutol was given orally as a single drug tablet and streptomycin was injected intravenously.

Each patient enrolled in the study was asked to provide plasma on day 1 (first day of treatment), day 14 (steady-state), and plasma and CSF at months 1 and 3. Enrolled patients were stratified into two groups (below and above 4 years of age) and randomisation of plasma sampling times was stratified by group. Previous admission patterns led us to believe that this age grouping would deliver to groups of approximately equal size. For each patient, two plasma samples were collected at two of ten randomly allocated pre-specified time points (i.e. at 1, 2, 3, 4, 5, 6, 8, 12, 18 or 24 h post-dose) on each of days 1 and 14. One additional plasma sample was collected for each patient at month 1 and 3, again at a randomly allocated time points between 3 to 5 h post-dose. Each child provided a maximum of 6 plasma samples over the trial.

CSF sampling by lumbar puncture (LP) was conformed to standard clinical care with CSF examination executed to assess treatment response. LPs were performed at months 1 and 3 post study entry at randomly allocated time points either 3, 4 or 5 h post-dose. Each LP was followed by a blood draw performed within 15 min. Freshly collected clinical samples were immediately processed and stored at −80 °C pending bioanalysis.

Acetylator genotype and predicted phenotype for each patient was determined by sequencing the second exon of the arylamine N-acetyltransferase 2 (NAT2) gene which contains the functional polymorphisms [13, 14]. Genomic DNA was extracted from blood using the Nucleon Genomic DNA extraction kit (GE Healthcare, Amersham, United Kingdom). Two primers NAT2F 5′-TGGGCTTAGAGGCTATTT and NAT2R 5′-GAGTTGGGTGATACATACAC were designed using Primer Express version 2.0 software (Applied Biosystems Inc, Foster City, CA, USA) to amplify a 768 bp sequence of the second exon of the NAT2 gene which contains the relevant polymorphisms (G191A, C282T, T341C, C481T, G590A, A803G, G857A). Following purification (QIAgen PCR purification kits, QIAGEN, United Kingdom) amplicons were sequenced (3130xl Genetic Analyzer Applied Biosystem, Hitachi, Singapore) and phenotypes predicted from genotypes previously defined in a cohort of healthy Vietnamese volunteers in a study correlating urinary caffeine metabolites ratios with NAT2 genotypes (data not shown) [15, 16].

The concentrations of pyrazinamide, isoniazid and rifampicin in plasma and CSF were measured by a fully validated HPLC method with UV detection following the Food and Drug Administration (FDA) and International Conference on Harmonization (ICH) recommendations [17]. Ethambutol is only detectable with mass spectrometry and was not analysed in this study. Thawed samples (plasma or CSF) were mixed for at least 30 s and left undisturbed on the bench for further 10 min. Then, 300 μL of IS solution (Guanosine and Phenacetine in saturated NaCl) were added to 300 μL of sample (plasma or CSF) in a 1.5 mL microcentrifuge tube. The tube was mixed for 15 s and left at room temperature for 10 min followed by a centrifugation at 3000 g for 5 min. The samples were loaded into a conditioned ISOLUTE-C18 50 mg/mL 96 well-plate (Biotage, Uppsala, Sweden) and washed with 0.5 mL of saturated NaCl until dryness. Isoniazid, pyrazinamide and guanosine were eluted with 2 × 250 μL of methanol–water (10:90, v/v). A second elution containing rifampicin and phenacetine was carried out by using 250 μL acetonitrile followed by 250 μL methanol. Ten μL of each eluate were injected into 2 independent equilibrated LC-UV systems (Lachrom Elite – Hitachi – Merck system). Separation was performed on a 5 μm LichroCart 125 × 4.6 mm Purosphere RP-18 end-capped column, equipped with 5 μm guard column LichroCart 4 × 4 mm, RP-18e (Merck, Darmstadt, Germany), and an isocratic flow rate of 1 mL/min of phosphate buffer (pH 4.2, 50 mM) and acetonitrile 0.25 % (v/v) for isoniazid, pyrazinamide and guanosine, and phosphate buffer (pH 4.2, 50 mM) and acetonitrile 36 % (v/v) for rifampicin and phenacetine. The chromatography time for all assays was set to 7 min with detection at 263 nm for isoniazid, pyrazinamide and guanosine and 247 nm for rifampicin and phenacetine, respectively. In plasma, the calibration ranged [lower limit of quantification (LOQ) to upper limit of quantification (ULOQ)] from 0.2 to 80 μg/mL, 0.2 to 32 μg/mL and 0.1 to 20 μg/mL for pyrazinamide, isoniazid and rifampicin, respectively. In CSF, the ranges were 0.2 to 40 μg/mL, 0.2 to 32 μg/mL and 0.1 to 16 μg/mL for pyrazinamide, isoniazid and rifampicin, respectively. Triplicates of quality control samples (low, middle and high) were analyzed within each batch to ensure accuracy and precision throughout the analysis. The precision of all quality control samples were below 15 % CV during routine drug measurements.

The naïve-pooled data analysis was based on plasma and CSF data collected at the first steady-state occasion (i.e. at day 14 for plasma samples and at month 1 for CSF samples). Drugs reach their respective steady-states after a minimum of 5 half-lives. With anti-TB drugs, the steady-state is assumed to be reached after 2 weeks even for RIF and its well described autoinduction metabolism [18, 19]. We also anticipated no significant pharmacokinetic changes between day 14 and month 1. The concentration-time data were pooled dependent on sampling matrix and split into 2 independent age-groups: ≤ 4 years and > 4 years. All concentration-time data were pooled for the following time windows: 0–0.5, 0.5–1, 1–2, 2–3, 3–4, 4–5, 5–6, 6–8, 8–12, 12–18, and 18–24 h post-dose. The 25^th percentile (Q1), 50^th percentile (median), and 75^th percentile (Q3) of drug concentrations were calculated within each time window. Drug measurements below the LOQ were set to LOQ/2 to avoid bias on account of data censoring. Median concentrations and mid-time intervals were used in the non-compartmental analysis, and Q1 and Q3 were used for graphical representation only.

Naïve-pooled concentration-time data were evaluated using a non-compartmental analysis approach implemented in WinNonlin version 6.3 (Pharsight Corporation, California, USA). Median drug exposure up to the last time point (AUC_0-t) was calculated using the linear trapezoidal method for ascending concentrations and the logarithmic trapezoidal method for descending concentrations. The median terminal elimination rate constant (λ_Z) was determined as the slope of the terminal elimination phase by best-fit ordinary linear regression. Drug exposure was extrapolated from the last observed median concentration to time infinity by C_LAST/λ_Z for each individual group to compute total drug exposure (AUC_0-∞). The terminal elimination half-life (t_1/2) was calculated as ln(2)/λ_Z. Maximum median concentration (C_MAX) and time to C_MAX (T_MAX) were taken directly from the observed data. Apparent median volume of distribution (V_Z/F) and median oral clearance (CL/F) were computed according to standard procedures. The wild-type minimum inhibitory concentrations (MIC) were arbitrary chosen from reference values from the literature [20, 21]. They were set at 12.5 mg/L, 0.2 mg/L and 1.0 mg/L for pyrazinamide, isoniazid and rifampicin, respectively. These MIC values were implemented in WinNonlin for the calculation of time above MIC (T > MIC) and exposure above MIC (AUC > MIC) as therapeutic indices. Sub-therapeutic thresholds were applied using maximum plasma concentrations (C_MAX) associated with poor outcome in adults with pulmonary TB [18, 22‐25]. The use of these cut-off concentrations is difficult in the context of a multiple drugs therapy and they are still controversial and debated in the literature. However, although there are no consensus values for sub-therapeutic thresholds associated with TBM, or in children, they can be considered as “biomarkers” of clinical outcome. More importantly even if only used graphically in this study, they are able to visualize under-exposure and compare peak concentrations between observed values in children and an adult reference. The sub-therapeutic thresholds were set at 35 μg/mL, 3 μg/mL and 8 μg/mL for pyrazinamide, isoniazid and rifampicin, respectively. Naïve-pooled pharmacokinetic parameter estimates are summarized in Tables 2 to 5, and the relative differences (%) in individual median parameter estimates were calculated between the two age strata for each drug, using the older age group as a reference.

One hundred children were enrolled in this study. The overall median age of the studied children was 36 months with the largest proportion of children (64 %) below 4 years of age. Among the 33 children ≥ 5 years of age, 16 (48 %) patients were admitted with stage I TBM according to the MRC criteria, 11 (33 %) with stage II and 6 (18 %) with stage III. Among the remaining 67 children < 5 years of age, 43 (64 %) were admitted with a BCS of 4–5 (BCS I), 12 (18 %) with a BCS of 2–3 (BCS II) and 12 (18 %) with a BCS of 0–1 (BCS III). An important proportion (39 %) of the Vietnamese children in this study were classified as malnourished according to the WHO growth criteria with a body mass index (BMI) Z-score below −2 (negative two) [26]. However, a significant weight gain occurred during the first 3 months of treatment (Table 1). The treatment doses in mg/kg, based on the body weight measured at enrollment, did not show significant differences between the two age-groups (Table 2, 3 and 5 for pyrazinamide, isoniazid and rifampicin, respectively).

Fifteen patients (15 %) died. Ten of 64 children under 4 (16 %) vs 5 of 36 children ≥ 4 years (14 %). The majority of deaths occurred in patients with advanced grades of disease at presentation (MRC grade II or III, or BCS from 0–3), but a single death occurred in a patient presenting with MRC grade-I TBM with concomitant HIV infection. It must be noticed that a higher proportion of older children had definite or probable TBM (30/36: 83 %) compared to younger children (42/64: 66 %) which might influence the respective mortality and morbidity in each group. Eight deaths occurred within the first 6 days, six between 6 and 45 days and the last death occurred at day 92. There were no significant differences in mortality or morbidity between the two age groups. At the end of the treatment, 6 out of 81 children (7 %) had severe symptoms and/or required help day and night, 21 of 81 children (26 %) had intermediate sequelae but independent life (modified Rankin scale from 2–3 [12]), and 54 out of 81 children (67 %) recovered fully (4 children did not complete the final study visit). In this study, AST and ALT levels were measured repeatedly for all patients for 90 days (pre-dose, 30, 60 and 90 days) or at any time the children developed symptoms. Two out of 99 children (2 %) met the criteria for drug-induced liver toxicity (DILI) but no child died due to DILI in this study.

All pharmacokinetic blood draws were close to the pre-specified protocol sampling times (<10 % CV for both age-groups). The absolute difference between the actual sampling time and mean time in the assigned window ranged between −8.2 min to +6.4 min for children below 4 years of age, and between −6.5 min to +0.3 min for children above 4 years of age. The density of individual data points used in each bin was also comparable between the two age groups (after adjusting for the different sample size).

The median and inter-quartile range of naïve-pooled plasma concentrations represented the central trends and the variability of the measured concentration-time profiles well in both age-groups for all drugs (Figs. 1, 2, 3 and 4). As seen in Tables 2, 3 4 and 5, the dosages were in agreement with the 2006 WHO recommendations at the time of the study. However, significant differences were found between young and old children for all drugs but INH when the dosing of fast and slow acetylators was compared. This is likely an inevitable consequence of larger relative body weight differences for younger children enhanced by the need to fit the weight bands and the lack of uniformity when tablets have to be split [27].