Drug concentration at the site of disease in children with pulmonary tuberculosis

Abstract

Background

Current TB treatment for children is not optimized to provide adequate drug levels in TB lesions. Dose optimization of first-line antituberculosis drugs to increase exposure at the site of disease could facilitate more optimal treatment and future treatment-shortening strategies across the disease spectrum in children with pulmonary TB.

Objectives

To determine the concentrations of first-line antituberculosis drugs at the site of disease in children with intrathoracic TB.

Methods

We quantified drug concentrations in tissue samples from 13 children, median age 8.6 months, with complicated forms of pulmonary TB requiring bronchoscopy or transthoracic surgical lymph node decompression in a tertiary hospital in Cape Town, South Africa. Pharmacokinetic models were used to describe drug penetration characteristics and to simulate concentration profiles for bronchoalveolar lavage, homogenized lymph nodes, and cellular and necrotic lymph node lesions.

Results

Isoniazid, rifampicin and pyrazinamide showed lower penetration in most lymph node areas compared with plasma, while ethambutol accumulated in tissue. None of the drugs studied was able to reach target concentration in necrotic lesions.

Conclusions

Despite similar penetration characteristics compared with adults, low plasma exposures in children led to low site of disease exposures for all drugs except for isoniazid.

Introduction

Adequate exposure of antituberculosis drugs at the site of disease is required for sterilization of tuberculous lesions. Only half of the one million children estimated to have TB are treated each year.¹ Those who do receive treatment are often underdosed using current weight-based treatment guidelines.² Children typically have lower plasma concentrations for the same milligrams per kilogram dose compared with adults for most first-line antituberculosis drugs, particularly for rifampicin, an important sterilizing drug.³ Additionally, young children, those with the highest risk of TB, are also at risk of developing more severe forms of TB, with increased mortality.

The treatment of paediatric intrathoracic TB (the most common form of TB in children, occurring in >75% of cases), including drug regimens, dosing and treatment duration, has historically been extrapolated from efficacy trials in adults, with limited consideration of differences in the clinical spectrum of disease or the pharmacokinetic (PK) variability observed in children. While many children with TB have paucibacillary disease limited to the mediastinal lymph nodes (LN), many (up to 40%) have more extensive forms of disease, including miliary TB, parenchymal pathology or complicated LN involvement with breakthrough to endobronchial disease.⁴ Although a shorter treatment of 4 months at standard doses was recently shown to be non-inferior to the currently recommended first-line 6 months treatment in children with non-severe intrathoracic and extrathoracic LN disease in the SHINE trial,⁵ children with TB currently still receive the same regimen, regardless of disease spectrum, severity and bacillary burden.

Antituberculosis drug dosing is based on target drug concentrations measured in plasma, despite evidence showing a low correlation between drug concentration measured in plasma versus at sites of disease, where penetration may be both lesion- and drug-specific.^6,7 To date, few studies in humans have evaluated the penetration of drugs into intrathoracic TB lesions. The limited studies to date have been in adults and have focused on lung tissue,^6,8–10 despite the fact that TB can be thought of as a lymphatic disease.¹¹

Understanding drug exposure at the site of disease could facilitate the design of more-effective and potentially shorter regimens in children across the disease spectrum, and may have broader therapeutic implications, including for adults. We aimed to characterize the concentrations of first-line antituberculosis drugs at the site of disease in children with complicated intrathoracic TB requiring bronchoscopy or surgical LN decompression (SD), and to compare drug exposures in children with adult levels in plasma and lesions, against current targets correlated with efficacy.^6,7,12–20

Patients and methods

Study design and setting

From November 2018 through March 2019, we prospectively enrolled children with complicated intrathoracic TB routinely referred to Tygerberg Hospital, Cape Town, South Africa. Children underwent bronchoscopy or transthoracic surgical LN decompression (SD) for diagnosis or to establish airway patency. Children were eligible if they were on a rifampicin-containing regimen for at least 10 days.

Ethics statement

The Health Research Ethics Committee at Stellenbosch University (N18/05/059) approved this study. Written informed consent was obtained from parents/legal guardians.

Clinical procedures

On the day of the procedure, antituberculosis treatment was administered by the study team after an overnight fast. The study was designed to ensure that different participants contributed samples at different timepoints, from peak plasma to the end of the distribution phase, to reconstruct the PK profile at the site of disease, with dosing times planned at 2 h before SD, or randomized to 2, 4 or 6 h prior to bronchoscopy. Randomization in the SD group was not feasible owing to the complexity and duration of the procedure. Treatment consisted of weight-banded once-daily oral doses of isoniazid (10–15 mg/kg), rifampicin (10–20 mg/kg) and pyrazinamide (30–40 mg/kg) using fixed-dose dispersible paediatric combinations (Table S1, available as Supplementary data at JAC Online). Ethambutol (15–25 mg/kg) was added in children who were HIV-coinfected and at clinical discretion.^21,22 Children with significant airway obstruction received oral prednisone (2 mg/kg daily).

Plasma and tissue sampling

Venous or arterial blood samples were collected pre-dose, and at approximately 2, 4 and 6 h post-dose. Blood sampling was targeted to be coupled to the site of disease sample and was recorded in relation to the time of bronchoscopy/SD. Bronchoalveolar lavage (BAL) samples were taken from children undergoing bronchoscopy and, when available, endobronchial LN biopsy specimens were collected for PK analysis and microbiological investigation. Lymph node tissue samples were collected during SD for PK, histology and microbiology. Depending on the size and consistency of lymph node fragments, samples were collected in either homogenizing tubes or in Cryomold for PK analysis. Further details are described in the Supplementary data.

Analytical methods

Assays to quantify rifampicin, isoniazid, pyrazinamide and ethambutol concentrations in plasma, BAL and LN tissue used validated methods.⁶ Plasma and BAL samples were analysed using a validated high-performance liquid chromatography-tandem mass spectrometry (LC-MS/MS) assay.^6,8 LN material obtained through endobronchial biopsy and liquefied LNs from SD were analysed using previously described validated LC-MS/MS methods.^6,8 Solid SD specimens larger than 5 mm were further classified by histology on frozen section into cellular, necrotic, and mixed lesions (cellular with necrosis in the background). For these specimens, drug concentrations were measured from pathologically distinct regions using laser capture microdissection (LCM) combined with LC-MS/MS as previously described.⁸

PK modelling and simulations

Published rifampicin, isoniazid and pyrazinamide population PK models in children, and an adult ethambutol model, were optimized, and the final models were fit to the new data.^23,12 All drug PK models included weight-based allometric scaling on clearance and volume, and for rifampicin and isoniazid, included postmenstrual age as a covariate on clearance.²³ To model ethambutol in children, South African adult PK parameters from Jönsson et al.¹² were scaled allometrically to the children’s weights. The parameters from these models were fixed and the model fit evaluated and subsequently re-estimated where necessary.

Concentration–time profiles and derived area under the concentration–time curve from 0 to 24 h (AUC_0–24) at steady-state were simulated for the median age and weight of children in the cohort (8.6 months, 8.2 kg) using our cohort plasma and tissue estimates. To compare these outputs with adult exposure from the same population, we used published population PK models from South African adults for rifampicin, isoniazid, pyrazinamide and ethambutol.^12–15 Concentration profiles for the four drugs in the different compartments were compared with appropriate inhibitory and bactericidal target concentrations. We used published distributions of wild-type MIC for plasma, BAL, homogenized LN and mixed tissue, intracellular macrophage IC₅₀ for cellular lesions, and caseum MBC for necrotic tissue or caseum (Table S2). Further details are described in the Supplementary data.

Data sharing

Data underlying these findings may be requested from the corresponding author.

Results

Participant characteristics

We enrolled 13 children who collectively underwent 15 procedures including eight bronchoscopies and seven SDs to resolve severe airway obstruction due to complicated TB. Baseline clinical and radiological characteristics are shown in Table 1, and individual participant characteristics in Table S3. All children had evidence of severe extensive intrathoracic disease on chest imaging (Figure S2). All were bacteriologically confirmed; one child had isoniazid monoresistant TB and a second had multidrug-resistant TB which was only diagnosed following the procedure.

Plasma and tissue samples

LN samples were obtained in 4/8 (50%) of the bronchoscopy cases through endobronchial biopsy, and in all seven (100%) SD cases (Table S3). Histological examination of all SD tissue fragments showed that most specimens were characterized by necrotizing granulomatous inflammation. None to minor residual normal tissue was identified in LNs, and only one had preserved architecture displaying reactive follicular hyperplasia. Ziehl-Neelsen staining revealed acid-fast bacilli in four of the seven samples, of which two had numerous bacilli, which were more abundant in necrotic foci. While none of the BAL samples was culture-positive, 4/7 SD cases were culture-positive. MICs were obtained for two of the Mtb isolates and values were within normal MIC range (Table S3).

The concentrations of rifampicin, isoniazid, pyrazinamide and ethambutol were measured in 44 plasma, and in 65 BAL and LN samples collected at 2 to 8 h post-dose. Drug concentrations in BAL could only be determined in 4/8 (50%) children due to urea concentrations below the limit of quantification in BAL, a requirement for dilution factor calculation. LCM was performed in nine frozen LN specimens from four participants, allowing for the spatial quantification of drugs within the spectrum of different tissue compartments identified (necrotic, cellular, or mixed lymphoid) (Figure 1). The longitudinal PK profiles showed that there was large variability in the concentrations of all drugs amongst tissue compartments, between participants, and within similar lesions in the same participant (Figure 2).

Modelling of plasma PK and tissue distribution

Model building used prior published paediatric population PK models and estimates for the plasma PK model were similar to published parameters (Table S4). The final PK estimates are shown in Table 2. The rate and penetration coefficient estimates for each drug are shown in Table 3. Simulated PK profiles with their relevant concentration targets and AUC_0–24 values for tissue compartments are shown in Figure S3 and Table S5, respectively.

Rifampicin

Rifampicin had high exposure in cellular tissue with higher concentration and AUC_0–24 levels compared with plasma. Modelling results showed a penetration coefficient of 1.37 (95% CI: 0.87–1.87) for cellular tissue. Necrotic regions had the lowest rifampicin concentrations and the lowest penetration coefficient (0.55, 95% CI: 0.477–0.641). Rifampicin concentrations were below the upper end of the MIC range for >50% of the dosing interval in all tissue compartments except necrotic regions (Figure S3).

Isoniazid

Isoniazid had good penetration into necrotic tissue with an estimated penetration coefficient of 0.84 (95% CI: 0.80–0.88), but did not reach the high caseum MBC (casMBC) target of 128 mg/kg. The penetration coefficient was lower in the cellular region of granulomas (0.56, 95% CI: 0.32–0.79), but concentration levels were above the MIC for 75% of the 24 h dosing profile. Overall, isoniazid exposure was above the MIC for approximately 50% of the dosing interval for all tissue compartments except necrotic lesions.

Pyrazinamide

Drug penetration coefficient estimates were relatively low and similar for cellular (0.42, 95% CI: 0.36–0.48) and necrotic lesions (0.40, 95% CI: 0.30–0.50). Despite low plasma pyrazinamide concentrations, our simulations showed pyrazinamide exposure was above the ‘acidic’ macrophage MIC for approximately 54% of the dosing interval in cellular compartments and below target casMBC for necrotic lesions.

Ethambutol

High penetration into all compartments was observed for ethambutol, and penetration coefficients ranged from 1.11 (95% CI: 0.43–1.95) in necrotic lesions, to 6.17 (95% CI: 0.91–14.4) in cellular tissue. In all compartments except for necrotic lesions, ethambutol appeared to accumulate in tissue. However, similar to other first-line drugs, ethambutol concentrations were never higher than its casMBC target and in other tissue types were above the lower range of their respective targets for only 29%–50% of the dosing interval.

Paediatric exposures compared with adult predictions

Penetration coefficient values for necrotic lesions were similar between children and adults (Table 3). Additional simulations to compare first-line drug concentration–time profiles of lesions in adults and children are shown in Figure 3. Mean values of the simulated steady-state AUC_0–24 in children compared with adults for plasma, cellular and necrotic lesions are shown in Table 4.

Plasma rifampicin exposures were similar to adult reference values (AUC_0–24 of 41.7 versus 38.3 mg·h/L).¹⁵ However, only 15% of the 1000 simulated rifampicin AUC_0–24 in plasma were above the proposed adults target of approximately 42 mg·h/L.^24–26 Children did appear to have more favourable drug penetration in LNs compared with penetration into lung lesions observed in adults, particularly in cellular regions, where the penetration coefficient was 1.37 (95% CI: 0.874–1.87) resulting in an AUC_0–24 of 55.2 mg·h/L.

Isoniazid plasma concentrations were similar in children compared with the simulated adult values (AUC_0–24 of 25.3 versus 27.7 mg·h/L). Particularly low plasma pyrazinamide exposures were observed compared with adult data (AUC_0–24 of 248 versus 466 mg·h/L) and resulted in children having approximately only half the exposure levels of adults in cellular and necrotic lesions. For ethambutol, the AUC_0–24 in children was six-fold lower than the simulated adult value of 78.5 mg·h/L. There were no prior adult data on ethambutol concentrations at pulmonary site of disease to allow comparison with paediatric data.

Discussion

We report for the first time, data on drug concentrations and exposure into intrathoracic LNs and BAL in children with severe intrathoracic TB. Penetration into the site of disease was similar in children compared with adults. However, the standard WHO-recommended antituberculosis drug dosing resulted in low plasma and consequently, low site of disease exposures for all drugs except for isoniazid.

In our study, histopathological examination of SD tissue fragments showed predominantly necrotizing granulomatous inflammation, with little residual recognizable LN tissue. This was probably a reflection of the nature of the procedure (decompression rather than complete LN excision) and disease severity, as these LN were affected to the extent of causing airway compression. Acid-fast bacilli were seen on Ziehl-Neelsen staining in four of seven SD samples, including two that were acid-fast bacilli negative, after at least 4 weeks of TB therapy. Although Ziehl-Neelsen staining does not differentiate viable from non-viable organisms, it is possible that some viable organisms remained in LNs after more than a month of treatment. Bacilli were more numerous in necrotic foci. This is similar to data previously reported in animal models²⁷ and in human adults.²⁸

The histological spectrum of the tissue samples between adults and children was similar, despite different excision locations. The penetration coefficients of necrotizing LNs in children were remarkably similar to those of adult lung granulomas, suggesting that penetration properties translate well between children and adults for these lesions. The penetration was highest for the two drugs with limited activity against quiescent mycobacteria: isoniazid and ethambutol. Conversely, the two drugs with sterilizing activity, rifampicin and pyrazinamide, had caseum exposures below their target concentrations (casMBC). A recent ex vivo model has shown that only rifampicin fully sterilizes bacilli in caseum,²⁹ with a casMBC of 6.5 mg/L. In the in vivo rabbit model of active TB, pyrazinamide was reported to exert bactericidal activity on phenotypically drug-tolerant bacilli in hard-to-reach caseous lesions.³⁰

We present, to our knowledge, the first human data on penetration of ethambutol into sites of disease compartments. Our results support the hypothesis that, despite its poor plasma-based PK/pharmacodynamic (PD) profile, the efficacy of ethambutol might be explained by the favourable penetration of this drug into TB lesions.⁸ Despite being a polar drug,³¹ ethambutol accumulated in all compartments relative to plasma and particularly in the cellular areas where ethambutol targets intracellular bacilli.¹⁷ Overall, ethambutol exposures in different compartments were limited by the low plasma concentrations, which were significantly lower than in adults (C_max 2.3 versus 6.4 mg/L) and below the MIC target. This is consistent with previous paediatric studies, which showed erratic absorption and low concentrations of ethambutol in children dosed at 10–20 mg/kg per day.³²

Given our results, antituberculosis treatment in children could benefit from higher ethambutol and rifampicin doses. The favourable penetration of ethambutol into LNs suggests that ethambutol should be dosed at 25 mg/kg (that is, at the higher end of the recommended 15–25 mg/kg) to achieve higher site of disease exposure. This dose is considered safe but doses beyond 25 mg/kg may be too high given ocular toxicity.³³ For rifampicin, only 15% of simulated plasma AUC_0–24 in children reached the proposed adult target.^24–26 These data again emphasize the need for higher doses of rifampicin.³⁴ Low exposure to rifampicin is associated with worse treatment outcomes in children.^35–39 Several ongoing trials are evaluating higher doses of rifampicin of up to 50 mg/kg in adults, and short-term dosing of up to 70 mg/kg in children.⁴⁰ In accordance with the national guidelines at the time of the study, pyrazinamide was administered at 25 mg/kg rather than the current WHO recommended weight bands of 30–40 mg/kg. Higher dosing will likely improve time above the acidic MIC in cellular compartments but will still remain below the casMBC for necrotic lesions. We recommend that where possible pyrazinamide should be dosed at the higher end of its dosing range to maximize coverage in cellular lesions.

Future studies should include sample times that will better characterize the absorption phase of the drugs studied and later trough levels. For this initial study, the sample times were optimized around the planned medical procedure time and C_max to ensure measurable drug concentration. Peripheral volume and intercompartmental clearance estimates of isoniazid were not consistent with previously reported parameters. Data were collected only up to 6 h making this compartment difficult to parameterize. Therefore, isoniazid simulations used previously published PK parameters with lesion parameters from our model to ensure accurate prediction of terminal concentrations. The macrophage IC₅₀ and casMBC potency assays used to measure drug potency at the site of disease are performed in matrices that reproduce in vivo drug binding and correct for protein binding. Therefore, our results are presented as total concentrations.

Limitations of our study include the fact that we studied a small sample of young children with severe forms of pulmonary TB undergoing surgery at a single site, and thus external generalizability should be evaluated further. All children were anaesthetized at the time of the procedure, and particularly those from the SD group received a significant number of concomitant drugs and intravenous fluids, all of which could have affected drug disposition. The classification of the LN histology on the fresh frozen tissue selected for PK into cellular, mixed and necrotic was challenging in some of our study samples due to freezing artefacts, absence of clear area demarcation and the fragmented nature of the tissue samples. However, routine histology samples (formalin-fixed paraffin-embedded) were usually available and provided guidance on interpretation in areas that were difficult to classify on frozen section. Finally, our PK/PD simulations in BAL should be interpreted with caution given the limited number of children in whom BAL urea concentration was successfully measured and the technical caveats such as dwelling time and volume of instilled lavage fluid.

In conclusion, we have shown that measuring the penetration of antituberculosis drugs in intrathoracic LN compartments and BAL in young children was feasible and that the penetration coefficients of first-line antituberculosis drugs into these compartments were similar to adults. The overall plasma exposures of all the drugs were low, particularly for ethambutol. Similarly, all the first-line drugs had exposure in necrotic tissue at levels lower than current adult target concentrations. Our results indicate that current paediatric dosing guidelines do not reach target exposures at sites of disease. An improved understanding of lesion penetration, and the use of a data-driven modelling approach will allow for more-appropriate dosing in children and improved efficacy of current regimens. Our data also informs the design of novel regimens for TB treatment-shortening in children across the spectrum of pulmonary disease.

Acknowledgements

We are grateful to the children and families who participated in the study. We thank Professor Peter Donald for his comments on the manuscript and Corne Bosch and Marianna de Kock for their support in obtaining the MIC data. We thank the Sub-Saharan Africa Regional Biospecimen Repository at Tygerberg Hospital for providing liquid nitrogen and dry ice. We thank the data and clinical team from DTTC and the clinical team from Tygerberg Hospital.

Funding

This study was supported by the Spanish Ministry of Science and Innovation and State Research Agency through the ‘Centro de Excelencia Severo Ochoa 2019-2023’ Program (CEX2018-000806-S), and support from the Generalitat de Catalunya through the CERCA Program. E.L.V. is supported by a Spanish Paediatrics Association (AEP) fellowship and a Ramon Areces Foundation fellowship. A.A.A. and E.M.S. are supported by the European and Developing Countries Clinical Trials Partnership (EDCTP), grant number TRIA.2015.1102. J.A.S. is supported by a Clinician Scientist Fellowship jointly funded by the UK Medical Research Council (MRC) and the UK Department for International Development (DFID) under the MRC/DFID Concordat agreement (MR/R007942/1). A.M.D. is supported by National Institute of Allergy and Infectious Diseases (NIAID) of the National Institutes of Health (NIH) under Award Number UM1AI106716. A.C.H. is supported by the South African National Research Foundation (SARchi Chair), SA NRF. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Transparency declarations

None to declare.

Supplementary data

Additional Methods, Figures S1 to S4 and Tables S1 to S5 are available as Supplementary data at JAC Online.

References

Author notes