J Antimicrob Chemother 2015; 70: 1798 – 1803

doi:10.1093/jac/dkv039 Advance Access publication 10 March 2015

Pharmacokinetics of anti-TB drugs in Malawian children: reconsidering

the role of ethambutol
R. Mlotha1, D. Waterhouse2, F. Dzinjalamala3, A. Ardrey2, E. Molyneux1, G. R. Davies4* and S. Ward3
1
Department of Paediatrics, Queen Elizabeth Central Hospital, Blantyre, Malawi; 2Department of Molecular Parasitology, Liverpool School
of Tropical Medicine, Liverpool, UK; 3Faculty of Pharmacy, College of Medicine, University of Malawi, Blantyre, Malawi; 4Institutes of
Infection and Global Health and Translational Medicine, University of Liverpool, Liverpool, UK

*Corresponding author. Tel: +44-(0)151-7063036; E-mail: [email protected]

Received 14 October 2014; returned 27 November 2014; revised 27 January 2015; accepted 30 January 2015

Background: Current guidelines for dosing of anti-TB drugs in children advocate higher doses for rifampicin and
isoniazid despite limited availability of paediatric data on the pharmacokinetics of these drugs, especially from
Africa, where the burden of childhood disease remains high.
Methods: Thirty children aged 6 months to 15 years underwent intensive pharmacokinetic sampling for first-line
anti-TB drugs at Queen Elizabeth Central Hospital, Blantyre, Malawi. Rifampicin, isoniazid, pyrazinamide and eth-
ambutol were dosed at 10, 5, 25 and 20 mg/kg, respectively. Plasma drug concentrations were determined using
sensitive, validated bioanalytical methods and summary pharmacokinetic parameters were estimated using
non-compartmental analysis.
Results: The median (IQR) Cmax was 2.90 (2.08–3.43), 3.37 (2.55–4.59), 34.60 (32.30–40.90) and 1.20 (0.85–
1.68) mg/L while the median (IQR) AUC0 – 1 was 16.92 (11.10–22.74), 11.48 (7.35–18.93), 333.50 (279.50–487.2)
and 8.65 (5.96–11.47) mg.h/L for rifampicin, isoniazid, pyrazinamide and ethambutol, respectively. For all drugs,
pharmacokinetic parameters relating to drug absorption and exposure were lower than those published for adults,
though similar to existing paediatric data from sub-Saharan Africa. Weight and/or dose predicted at least one meas-
ure of exposure for all drugs. Age-related decreases in CL/F for rifampicin and pyrazinamide and a biphasic elimination
pattern of isoniazid were observed. Predicted AUC0 – 1 for rifampicin dosed at 15 mg/kg was comparable to that of
adults while the dose required to achieve ethambutol exposure similar to that in adults was 55 mg/kg or higher.
Conclusions: These data support recently revised WHO recommendations for dosing of anti-TB drugs in children,
but dosing of ethambutol in children also appears inadequate by comparison with adult pharmacokinetic data.

Keywords: tuberculosis, paediatrics, PK, Africa

comprises data from only 133 South African and 45 Malawian

Introduction children.7 – 12 Furthermore, most of these studies employed rela-
Clinical diagnosis and treatment of childhood TB is challenging, tively sparse sampling schemes, did not include data for all of
particularly in resource-poor settings where the burden of disease the first-line drugs deployed in the regimens studied and pre-
is highest.1 Consequently much of the evidence supporting treat- sented only descriptive analyses. Only a single small study has
ment of paediatric TB is derived from clinical trials of interventions subsequently reported data on PK associated with the revised
in adults. Appropriate paediatric investigation plans were not car- WHO recommendations.13 We report the results of a PK study
ried out for existing first-line anti-TB drugs2 and dosing recom- employing intensive sampling at steady-state in children repre-
mendations have been derived from schedules employed for sentative of the spectrum of paediatric TB in Blantyre, Malawi,
adults, with a relative lack of supportive data on efficacy or including those with HIV coinfection and malnutrition.
safety.3 – 5 In 2010, on the basis of limited existing pharmacoki-
netic (PK) studies, the WHO recommended that weight-based Methods
doses for rifampicin and isoniazid in children should be increased.6
However, there remains a paucity of reliable data on the PK of Clinical protocol
these drugs in children. The current literature relating to first-line Children were recruited from the Department of Paediatrics at Queen
drugs in the treatment of active disease from sub-Saharan Africa Elizabeth Central Hospital, Blantyre from January 2007 to February

# The Author 2015. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/
4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

1798
Anti-TB drugs in Malawian children JAC
2008. All parents or guardians gave written informed consent for study timepoints. Data summaries, graphics and analysis of variance of the
procedures. The study protocol was approved by the Ethics Committee summary PK parameters were performed in R 2.14.1 (R Foundation for
of the College of Medicine of the University of Malawi. Children were eligible Statistical Computing, Vienna, Austria). Analysis of variance was per-
if they were aged between 6 months and 15 years and had been com- formed with log-transformed PK parameters where appropriate and
menced on treatment for TB as inpatients or outpatients. Baseline evalu- model assumptions checked using routine graphical diagnostics.
ation included TB contact history and Mantoux testing using 0.1 mL of Generalized additive models were used to evaluate continuous covariate
tuberculin PPD RT23 (1: 1000), read between 48 and 72 h. Induration of relationships using the package mgcV and clustering analysis for subpopu-
10 mm or more (or 6 mm or more in an HIV-infected child) was regarded lation detection in the parameter distributions was performed using the
as positive. Chest radiographs were considered suggestive of pulmonary package mclust.
TB in the presence of mediastinal adenopathy, perihilar adenopathy or per-
sistent lobar consolidation. The diagnosis of tuberculous meningitis was
accepted with appropriate CSF changes in addition to clinical signs. Results
Following counselling, HIV status was determined by enzyme-linked
immunoassay using HIV DETERMINE (Invernos Med, Japan Co Ltd) and Table 1 shows the characteristics of children recruited into the
confirmed by a second assay with UNIGOLD (Trinity Biotech PK Ireland). study. Thirty children aged between 6 months and 15 years
CD4 counts were determined for HIV-infected children. Body mass (kg) were enrolled with a mean age and weight of 7 years and
and height (m) were measured by standard anthropometric methods. 18 kg, respectively. Nine children were aged ,2 years and 50%
Classification of nutritional state was based on the Waterlow14 classifica- were female. The most common form of TB was pulmonary (21
tion of weight for height: .90% was taken as normal; 81%– 90% as mild; patients, 70% of children), and the rest of the cases were extrapul-
70%–80% as moderate; and ,70% as severe wasting. Haematocrits were monary, including lymph node and meningeal TB. Twenty (67%)
assessed and children with values ,25% were not enrolled. Other labora-
were HIV infected. Eighteen of the HIV-infected children (90%)
tory tests performed included tests for serum alanine transferase and
were receiving co-trimoxazole prophylaxis and nine (45%) were
creatinine.
All the children were recruited at least 2 weeks after initiation of the on ART at the time of PK sampling.
intensive phase of treatment and received fixed-dose combinations The range of actual weight-adjusted dose for each drug con-
(FDCs) of anti-TB drugs approved by the National TB Programme. Each tained in the FDC product (rifampicin, isoniazid and pyrazinamide)
FDC tablet or sachet contained 60 mg of rifampicin, 30 mg of isoniazid and the loose ethambutol tablets is presented in Figure 1. The
and 150 mg of pyrazinamide, supplemented by loose 100 mg tablets of
ethambutol. Doses were administered orally once daily within the follow-
ing weight bands: ,7 kg, 1 tablet; 8 –9 kg, 1.5 tablets; 10– 14 kg, 2 tablets; Table 1. Study population (N¼30)
15 – 19 kg, 3 tablets; 20 – 24 kg, 4 tablets; and 25 – 29 kg, 5 tablets. This
schedule corresponds to the weight-based doses recommended by the Age (months), median (range) 90 (7–187)
national programme at the time of the study: isoniazid, 5 mg/kg; rifampi- Male, n (%) 15 (50)
cin, 10 mg/kg; pyrazinamide, 25 mg/kg; and ethambutol, 20 mg/kg. HIV positive, n (%) 20 (67)
Children with tuberculous meningitis received intramuscular streptomycin
Weight (kg), median (range) 18 (4.8–45)
instead of ethambutol.
Height (m), median (range) 102 (60– 150)
PK sampling was performed at the hospital at least 2 weeks after treat-
ment initiation. Dosing was administered under observation by ward nurs- Pulmonary TB, n (%) 21 (70)
ing staff with no restrictions on access to food and water. An intravenous Co-trimoxazole prophylaxis, n (%) 18 (60)
cannula was inserted and maintained using heparin– saline flushes. Blood ART, n (%) 9 (30)
samples were collected at 0, 0.5, 1, 2, 3, 4, 6, 8 and 24 h into lithium hep-
arin tubes, centrifuged immediately and the separated plasma frozen at
2708C until bioanalysis was performed.
1.8
Paediatric formulation Adult formulation
Bioanalytical methods 1.6
Rifampicin and pyrazinamide plasma concentrations were determined
using HPLC on a Shimadzu LC 2010 HT system (Shimadzu, Manchester, 1.4
Log10 mg per kg

UK). Isoniazid and ethambutol concentrations were determined simultan- PZA

eously using LC-MS/MS on a triple-quadrupole TSQ Quantum Access mass 1.2
spectrometer (Thermo Scientific, Hemel Hempstead, UK). All methods ETH
incorporated appropriate internal standards and were validated to inter-
1.0
nationally recognized acceptance criteria. The lower limits of quantifica-
tion for the assays were 0.5, 2.5, 0.020 and 0.010 mg/L for rifampicin, RIF
pyrazinamide, isoniazid and ethambutol, respectively. Full details of the 0.8
bioanalytical methods are contained in the supplementary methods
(available as Supplementary data at JAC Online). 0.6 INH

10 20 30 40
Statistical analysis
Bodyweight (kg)
Non-compartmental PK analysis of plasma concentration –time data was
performed using Kinetica 4.1.1 (Adept Scientific Ltd, Armor Way, Figure 1. Weight-adjusted dose by weight band for the four drugs in
Letchworth Garden City, UK) using the trapezoidal rule with the log up – two different formulations. PZA, pyrazinamide; ETH, ethambutol; RIF,
linear down option and manual adjustment of the range of included rifampicin; INH, isoniazid.

1799
Mlotha et al.

Rifampicin Isoniazid

10.0

5.00
5.0
Rifampicin (mg/L)

Isoniazid (mg/L)
0.50
2.0

1.0
0.05

0.5
0.01
0 5 10 15 20 0 5 10 15 20
Time (h) Time (h)

Pyrazinamide Ethambutol
100
2.00
50
1.00
Pyrazinamide (mg/L)

Ethambutol (mg/L)

0.50
20
0.20
10
0.10

0.05
5

0.02
2
0 5 10 15 20 0 5 10 15 20
Time (h) Time (h)

Figure 2. Summary semi-logarithmic scatterplots of plasma concentrations of anti-TB drugs. Continuous lines represent median concentrations and
broken lines represent upper and lower quartiles.

recommended weight-banding controls the weight-adjusted the LOQ and were omitted from the analysis. Estimates of the
dose within a narrow proportional range for each drug, although apparent terminal elimination half-life were based on at least
the absolute range of mg/kg dose increased with dose. Hence, the three observations with means of 3.8, 3.6, 3.7 and 3.6 for the
absolute range was largest for pyrazinamide (almost 10 mg/kg four drugs, respectively.
across each weight band) and smallest for isoniazid (,2 mg/kg The mean weight-adjusted dose of rifampicin received was
across each weight band). Three children received a similar 9.92 mg/kg with a median observed Cmax of 2.90 mg/L and an
weight-adjusted dose using the adult formulations of the drugs. AUC0 – 1 of 16.92 mg.h/L (adult reference IQR: 4.2 – 9.4 and
Plasma concentrations for each of the drugs over time are 16.6 – 36, respectively14). Some 87% and 70% of children were
summarized in Figure 2. Summary PK parameters derived from below the lower quartile for Cmax and AUC0 – 1, respectively, in
non-compartmental analysis are summarized in Table 2. Owing adults. Only 4 of 30 children exceeded the threshold of 4 mg/L
to data below the limit of quantification (LOQ) of the PK assay cited in adults as representing ‘very low’ exposure to rifampicin.
or a non-credible PK profile, parameters of rifampicin, isoniazid, Of note, in eight children absorption appeared to be delayed
pyrazinamide and ethambutol could only be estimated for 28, with a Tmax of 4 h or greater. Mean Cmax in children with delayed
30, 29 and 28 of the children with mean percentage extrapolation absorption was lower than in those with early absorption (2.38
of the AUC0 – 1 of 28.6%, 3.2%, 30.8% and 10.9%, respectively; versus 3.58 mg/L), although this was not a statistically significant
54.4%, 11.1%, 21.1% and 10.3% of datapoints for rifampicin, iso- finding (t-test P ¼ 0.07). In multivariate analysis, AUC0 – 1was
niazid, pyrazinamide and ethambutol, respectively, were below related to weight-adjusted dose, increasing by 0.12 mg.h/L for

1800
Anti-TB drugs in Malawian children JAC
Table 2. Summary PK parameters derived from non-compartmental analysis

Rifampicin (n¼28) Isoniazid (n¼30) Pyrazinamide (n ¼29) Ethambutol (n¼28)

Cmax (mg/L) 2.90 (2.08– 3.43) 3.37 (2.55– 4.59) 34.60 (32.30–40.90) 1.20 (0.85–1.68)
Tmax (h) 2.00 (1.00– 4.00) 2.00 (1.00– 2.75) 2.00 (1.00–3.00) 3.00 (2.00–4.00)
AUC0 – last (mg.h/L) 7.50 (5.59– 13.06) 11.21 (7.03– 18.40) 194.70 (163.40– 382.40) 8.00 (4.92–10.07)
AUC0 – 1 (mg.h/L) 16.92 (11.10–22.74)a 11.48 (7.35– 18.93)b 333.50 (279.50– 487.2) 8.65 (5.96–11.47)
t1/2 (h) 2.01 (1.64– 3.27) 3.54 (2.95– 4.50) 5.64 (4.47–6.81) 6.49 (5.69–8.04)
CL/F (L/h) 5.41 (3.38– 10.64) 5.67 (4.34– 8.27) 0.75 (0.48–1.16) 34.0 (19 –43)
Vz/F (L) 20.840 (14.850 –43.060) 28.93 (21.6– 42.9) 6.193 (3.691–9.334) 289 (207– 410)
Vss/F (L) 33.040 (20.10–49.790) 28.3 (19.2– 34.8) 7.642 (4.205–10.280) 333 (229– 413)

Cmax, maximum observed plasma concentration; Tmax, time of maximum observed plasma concentration; AUC0-last, area under the curve to last
observed plasma concentration; AUC0-1, area under the curve extrapolated to infinity; t1/2, apparent elimination half-life; CL/F, apparent clearance;
Vz/F, volume of distribution; Vss/F, volume of distribution at steady-state.
Data are presented as median (IQR).
a
Based on 17 subjects.
b
Based on 28 subjects.

each additional mg/kg (P ¼ 0.028). This was not true for Cmax, increased with absolute dose (P¼0.049). The CL/F of pyrazinamide
possibly because of the delayed absorption observed in some appeared to decrease significantly with age (P ¼ 0.036) and this
children. CL/F of rifampicin decreased with weight, weight-for- trend appeared linear in a generalized additive model.
height, age and co-administration of co-trimoxazole in univariate The mean weight-adjusted dose of ethambutol was 17.05 mg/kg.
analysis, although only weight remained significant in multivari- The median observed Cmax at 2 h was 1.2 mg/L and the AUC0 – 1 was
ate analysis. In a generalized additive model CL/F decreased 8.65 mg.h/L (adult reference IQR: 4.1–6.3 and 19.2–30.8, respect-
steadily with age after a threshold of 4 years, a trend also ively14). Some 100% and 93% of children were below the lower quar-
observed for the highly correlated weight and weight-for-height tile for Cmax and AUC0 – 1, respectively, in adults. Cmax, AUClast and
variables (Figure 3a). Median CL/F for children under 5 years of AUC0 – 1 were all strongly predicted by absolute, but not weight-
age was 9.76 L/h, but 4.78 L/h for children over 5 years of age, adjusted, dose with AUC0 – 1 increasing by 0.13 mg.h/L for each
although this was not a statistically significant difference additional milligram. Age also appeared to affect measures of
(Wilcoxon test P¼0.41). exposure with AUC0 – 1 increasing by 0.013 mg.h/L for each add-
The mean weight-adjusted dose of isoniazid was 5.18 mg/kg. itional month of age. Serum creatinine was also a significant
The median observed Cmax at 2 h was 3.37 mg/L and the AUC0 – 1 predictor for AUC0 – 1 with exposure increasing by 8.649 mg.h/L
was 11.48 mg.h/L (adult reference IQR: 4.9 – 8.7 and 22.5 – 42.4, for each additional g/dL (P ¼ 0.048). Ethambutol exhibited a very
respectively14 ). Some 80% and 90% of children were below high apparent volume of distribution (333 L or 18.5 L/kg).
the lower quartile for Cmax and AUC0 – 1, respectively, in adults. HIV coinfection was not associated with any significant change
AUC0 – 1 did not increase with absolute or weight-adjusted dose in PK parameters.
(P ¼ 0.862 and 0.147). The rate of absorption as represented by Co-administration of co-trimoxazole prophylaxis appeared to
Tmax increased with both absolute and weight-adjusted dose increase the CL/F of rifampicin alone (P ¼ 0.028), while concomi-
(P ¼ 0.032 and 0.041). The overall median half-life of isoniazid tant ART did not impact any PK parameter. Weight-for-height
was 3.54 h, but, as expected, clustering analysis separated the was not identified as an important covariate for any of the
children into two main groups with mean half-lives of 1.54 h drugs or parameters.
(19.9%, likely fast acetylators) and 3.85 h (80.1%, likely inter- Given the complex relationships observed with dose and weight
mediate or slow acetylators) (Figure 3b). Of note, 18 of 30 children and lack of information on NAT2 genotype, it was only possible to
had detectable concentrations of isoniazid at 24 h post-dose, a cautiously evaluate alternative dosing recommendations for rifam-
finding highly suggestive of biphasic elimination. Re-analysis picin and ethambutol. For rifampicin, assuming a linear relationship
omitting this timepoint was performed to examine the influence with weight-adjusted dose, the expected AUC0 – 1 at the newly
of this information and showed that the overall median half-life recommended target of 15 mg/kg would be 65.1 mg.h/L, which
would have been estimated at 1.65 h, a figure more similar to exceeds the upper quartile of the observed adult range. For etham-
many previous reports, with a similar predicted proportion of butol, however, linear extrapolation of average weight-adjusted dose
fast acetylators. suggested that an AUC0 – 1 comparable to adults (.19.2 mg.h/L)
The mean weight-adjusted dose of pyrazinamide was could only be achieved at a dose of 55 mg/kg or higher.
24.80 mg/kg. The median observed Cmax at 2 h was 34.6 mg/L
and the AUC0 – 1 was 333.50 mg.h/L (adult reference IQR: 46.0–
Discussion
61.4 and 406.2 – 632.3, respectively14). Some 83% and 67% of
children were below the lower quartile for Cmax and AUC0 – 1, respect- This study is the most intensive PK study of all four first-line
ively, in adults. Cmax was significantly related to absolute dose anti-TB drugs yet undertaken in children with TB and adds sub-
(P¼0.02), but AUC0 – 1 was not (P¼0.11). The Tmax of pyrazinamide stantially to the evidence available, especially in sub-Saharan

1801
Mlotha et al.

(a) observed in African adults using the same analytical approach.15

0.4 For most drugs, some relationship of exposure with absolute or
less commonly weight-adjusted dose was observed for at least
0.2 one measure of exposure. In addition, for rifampicin and pyrazina-
mide, there appeared to be an independent relationship of CL/F
with age. Isoniazid elimination appeared biphasic, with a longer
0.0 apparent half-life than previously reported, but a frequency of
Spline (age)

fast acetylator phenotype similar to reports from South Africa.

This finding probably resulted from the availability of a sample
–0.2 at 24 h and the highly sensitive bioanalytical method used for iso-
niazid. Although observed in some previous studies, the possible
–0.4
metabolic basis of this finding is currently unknown. Exposure of
ethambutol appeared to be significantly related to creatinine.
Neither HIV nor nutritional status significantly affected the PK
–0.6 parameters of any of the drugs. Co-administration of ART or
co-trimoxazole prophylaxis also did not appear to have any
impact with the possible exception of a small increase in rifampi-
50 100 150
cin CL/F associated with co-trimoxazole co-administration.
Age (months)
These findings are generally in agreement with limited data
available from elsewhere in Africa and support recent recommen-
(b) 1.5
dations by the WHO to increase the dose of isoniazid and rifampi-
cin in children with TB to 10 and 15 mg/kg, respectively. Although
PK targets for treatment success in TB are not well-defined, ensur-
ing exposure in children at least equivalent to adults is logical
given the excellent overall performance of first-line therapy in
Frequency density

1.0
adults. The data presented here might also suggest that dose
modification could be considered for pyrazinamide and etham-
butol. However, it was difficult to accurately predict the equivalent
doses required due to sometimes complex relationships observed
0.5 between dose, weight and the PK parameters. This may be
attributable to the relatively small sample size and in the case
of isoniazid to tight weight-banding in a small dose range and
confounding effects of acetylator status. Absolute dose size did
appear to independently predict exposure of pyrazinamide and
0.0 ethambutol, however. In the case of ethambutol, extrapolation
of this relationship with AUC0 – 1 appeared to confirm previous
0.5 1.0 1.5 suggestions that current dosing of ethambutol in children may
Log half-life INH be inadequate and that a substantial dose increase would be
Figure 3. (a) Spline function of rifampicin CL/F with age derived from a
needed to achieve plasma concentrations comparable to those
generalized additive model (mean estimate shown by the continuous in adults. This finding is consistent with the results of a previous
line and 95% Bayesian credible intervals shown by the broken lines). (b) study in Blantyre, which evaluated thrice weekly dosing at
Estimated frequency density of subpopulations of isoniazid CL/F within 35 mg/kg, achieving a Cmax of only 1.8 mg/L. These extrapolations
the dataset derived from empirical clustering analysis. INH, isoniazid. assume linearity of PK at higher doses and better predictions
might be obtained using a more sophisticated modelling
approach and dose-ranging data. A single small study from
Africa, which bears the highest burden of childhood disease. All South Africa has also suggested that the new dosing recommen-
studies of African children have to date been performed in dations can achieve the targeted exposures desired.13
South Africa,7 – 11 with the exception of a single study from In addition to relationships observed with dose and weight,
Malawi, which evaluated intermittent dosing of pyrazinamide other features relating to the metabolism of certain drugs were
and ethambutol alone.12 The size of the cohort, nine-point sam- observed. Using empirical clustering, an unequivocally fast acety-
pling strategy and sensitive bioanalytical methods adopted in the lator phenotype could be attributed to 20% of the children, con-
current study allowed for more accurate and precise estimates of sistent with previous studies in South Africa, which reported a
PK parameters using non-compartmental techniques than was homozygous slow acetylator NAT2 genotype in 36% – 40% of chil-
previously possible, particularly Cmax, which may be underesti- dren.8,10,13 However, the African region exhibits the greatest diver-
mated by sparse sampling. More importantly, improved estimates sity at this locus worldwide16,17 and further characterization of the
of AUC0 – 1, believed to be the key determinant of clinical efficacy distribution of NAT2 alleles in Malawi would be useful in view of
for most first-line drugs, could also be obtained. the prolonged half-life of isoniazid that was observed. Both rifam-
The most important measures of absorption and exposure picin and pyrazinamide appeared to exhibit age-related changes
(Cmax and AUC0 – 1) for all of the drugs assayed in this representa- in CL/F independent of body size that were judged statistically sig-
tive group of African children with TB were lower than typically nificant. For pyrazinamide this trend appeared linear over the age

1802
Anti-TB drugs in Malawian children JAC
range studied, whereas for rifampicin CL/F appeared to decrease
around 4 years, ultimately halving over the age range. These References
changes may reflect the greater relative liver size and blood 1 Sandgren A, Cuevas LE, Dara M et al. Childhood tuberculosis: progress
flow per kilogram in younger children.18 Why the shape of requires an advocacy strategy now. Eur Respir J 2012; 40: 294– 7.
these trends should differ is unclear, although it might reflect 2 Donald PR, Ahmed A, Burman WJ et al. Requirements for the clinical
maturation processes specific to the dominant metabolic path- evaluation of new anti-tuberculosis agents in children. Int J Tuberc Lung
ways for the two drugs (hepatic carboxyesterases for rifampicin Dis 2013; 17: 794– 9.
and amidase for pyrazinamide). The apparent CL/F of pyrazina- 3 Donald PR, Maritz JS, Diacon AH. Pyrazinamide pharmacokinetics and
mide reported here is lower than that observed in adults, but efficacy in adults and children. Tuberculosis (Edinb) 2012; 92: 1– 8.
this could reflect changes in bioavailability relative to adults rather 4 Donald PR, Maritz JS, Diacon AH. The pharmacokinetics and pharmaco-
than truly reduced clearance, given the sachet formulation used dynamics of rifampicin in adults and children in relation to the dosage
in this study. Ethambutol exposure was related to serum creatin- recommended for children. Tuberculosis (Edinb) 2011; 91: 196– 207.
ine, which is consistent with the renal route of elimination of the 5 Donald PR, Maher D, Maritz JS et al. Ethambutol dosage for the treat-
drug. CL/F was not clearly related to serum creatinine, but this ment of children: literature review and recommendations. Int J Tuberc
may be due to the restricted range of the covariate in this dataset. Lung Dis 2006; 10: 1318–30.
From a clinical perspective the study provides insight into dos-
6 WHO. Rapid Advice: Treatment of Tuberculosis in Children. Geneva: WHO,
ing in potentially important subgroups and with other commonly 2010. WHO/HTM/TB/2010.13.
co-administered medications. We found no evidence that HIV or
7 Seifart HI, Donald PR, de Villiers JN et al. Isoniazid elimination kinetics in
nutritional status significantly affected PK parameters or that
children with protein-energy malnutrition treated for tuberculous menin-
there are any important PK interactions with co-trimoxazole
gitis with a four-component antimicrobial regimen. Ann Trop Paediatr
prophylaxis or nevirapine-based ART. Although the analysis 1995; 15: 249–54.
suggested a possible increase in CL/F of rifampicin with
8 Schaaf HS, Parkin DP, Seifart HI et al. Isoniazid pharmacokinetics in chil-
co-administration of co-trimoxazole prophylaxis, this has not
dren treated for respiratory tuberculosis. Arch Dis Child 2005; 90: 614–8.
been reported elsewhere and could be a chance finding requiring
independent confirmation. Overall, these data provide reassur- 9 Schaaf HS, Willemse M, Cilliers K et al. Rifampin pharmacokinetics in chil-
dren, with and without human immunodeficiency virus infection, hospita-
ance that dosage adjustment of anti-TB drugs is not needed for
lized for the management of severe forms of tuberculosis. BMC Med
any of these specific reasons. However, it should be noted that
2009; 7: 19.
the study was not designed to evaluate different phenotypes of
malnutrition and further work would be necessary to substantiate 10 McIlleron H, Willemse M, Werely CJ et al. Isoniazid plasma concentra-
tions in a cohort of South African children with tuberculosis: implications
this conclusion.
for international pediatric dosing guidelines. Clin Infect Dis 2009; 48:
In conclusion, this intensive PK study demonstrates that
1547– 53.
plasma exposures of first-line anti-TB drugs in Malawian children
are lower than those observed in African adults, confirming the 11 McIlleron H, Willemse M, Schaaf HS et al. Pyrazinamide plasma con-
centrations in young children with tuberculosis. Pediatr Infect Dis J
limited data available to date from paediatric PK studies in the
2011; 30: 262–5.
region. These findings are supportive of recent WHO recommen-
dations to increase the doses of rifampicin and isoniazid in routine 12 Graham SM, Bell DJ, Nyirongo S et al. Low levels of pyrazinamide and
paediatric practice, but the low exposures of ethambutol ethambutol in children with tuberculosis and impact of age, nutritional
status, and human immunodeficiency virus infection. Antimicrob Agents
observed also suggest that the dose and role of this drug in child-
Chemother 2006; 50: 407–13.
hood TB should be revisited.
13 Thee S, Seddon JA, Donald PR et al. Pharmacokinetics of isoniazid,
rifampin, and pyrazinamide in children younger than two years of age
with tuberculosis: evidence for implementation of revised World Health
Organization recommendations. Antimicrob Agents Chemother 2011; 55:
Funding 5560– 7.
This work was supported by the Wellcome Trust UK. 14 Waterlow JC. Classification and definition of protein-calorie malnutri-
tion. Br Med J 1972; 3: 566– 9.
15 McIlleron H, Wash P, Burger A et al. Determinants of rifampin, isoniazid,
pyrazinamide, and ethambutol pharmacokinetics in a cohort of tubercu-
losis patients. Antimicrob Agents Chemother 2006; 50: 1170 –7.
Transparency declarations
16 Sabbagh A, Langaney A, Darlu P et al. Worldwide distribution of
None to declare.
NAT2 diversity: implications for NAT2 evolutionary history. BMC Genet
2008; 9: 21.
17 Mortensen HM, Froment A, Lema G et al. Characterization of genetic
variation and natural selection at the arylamine N-acetyltransferase genes
Supplementary data in global human populations. Pharmacogenomics 2011; 12: 1545–58.
Supplementary methods are available as Supplementary data at JAC 18 Anderson BJ, Holford NH. Understanding dosing: children are small
Online (http://jac.oxfordjournals.org/). adults, neonates are immature children. Arch Dis Child 2013; 98: 737–44.

1803