Treating children with tuberculosis—Using pharmacometrics to do better

Abstract

Paediatric doses of the antituberculosis drugs are intended to achieve similar exposures to those in adults on successful treatment, provided they are safe. However, the generation of robust pharmacokinetic evidence to support improved dosing has been slow, and its translation into improved dosing guidelines has been limited. Drug exposures at the site of infection and susceptibility of the mycobacteria to the drugs determine antimicrobial activity. Immunity is also important for treatment response especially in the setting of a weak regimen. A wider appreciation of pharmacokinetic evidence together with the application of pharmacometric tools would improve drug dosing and guide regimen design. Recent pharmacokinetic studies in children with tuberculosis show that current doses fail to achieve adult drug exposures, most critically in young children who are vulnerable to disseminated and severe disease.1, 2 In parallel, there is mounting evidence that adult pharmacokinetic targets should be adjusted, and that when determining optimal drug exposures, differences between patients and in disease expression should be considered.3-6 New drugs and treatment regimens are also emerging. The adoption of accurate and efficient methods is imperative to translate these advances into paediatric drug combinations and doses that optimally improve treatment outcomes and prevent drug resistance. Investment in the development and distribution of suitable formulations could then be driven with greater confidence. Roughly 1.2 million children develop tuberculosis each year. The recent SHINE study reported successful outcomes in 97% of children with non-severe drug-susceptible tuberculosis, confirming that children with paucibacillary tuberculosis respond well to treatment.7 Children under 2 years old and children with HIV are less able to contain the infection, however, and are prone to develop disseminated and severe disease with much poorer outcomes. Even amongst children with minimal disease, mortality is disproportionately high in the youngest children. In the SHINE study, the risks of unfavourable outcomes were roughly 5-fold higher in younger children compared with children more than 3 years old. Hence, it is particularly important to optimize dosing in these young children with immature immunity. That, on recommended doses, the smallest children have substantially lower drug exposures than adults is therefore worrying.1, 2 Equally concerning is that older children have been neglected by recent guidelines increasing the first-line drug doses for children under 25 kg.1, 8 Historically, doses for children have been informed by the weight-adjusted (mg/kg) dose prescribed for adults. With intent to bring drug exposures in line with those in adults on standard treatment, in 2010, the World Health Organization increased doses of the first-line antituberculosis drugs recommended for children under 25 kg. The daily dose of rifampicin was increased by 50% to 15 (10–20) mg/kg and isoniazid by 100% to 10 (7–15) mg/kg. Pyrazinamide doses of 35 (30–40) mg/kg daily are recommended. The revised doses target uniform mg/kg doses through which lower systemic exposures in smaller individuals persist, with a marked trend to reduced drug exposure with decreasing weight band, due to the nonlinear relationship between weight and clearance.1, 2 Semi-mechanistic population pharmacokinetic models have shown that reduced bioavailability of rifampicin and isoniazid exacerbates low exposures in children under 3 years and that immaturity of clearance pathways mitigates the effect of low bioavailability plus relatively increased clearance in small children, only amongst the youngest.2, 9 After repeated doses, the predicted area under the concentration–time curve (AUC) for rifampicin is up to 50% lower than the mean achieved in adults (38.7 mg·h/L).2 Given the 2-fold increase in mg/kg doses of isoniazid, median peak concentrations in the higher paediatric weight bands are well above the proposed adult normal range of 3–6 mg/L.1, 2 Reassuringly, in the SHINE study, little toxicity was reported, although this study did not evaluate toxicity in severely ill children. Simulations show that increasing the dose of rifampicin in each weight band by 75 mg will considerably improve rifampicin AUCs, but that to deliver improved doses of all three drugs in fixed dose combination (FDC), a new formulation with revised drug ratios is needed.2 Thus, pharmacokinetic models shed light on the mechanistic basis for the pharmacokinetic findings in children and can be used to predict improved dosing strategies. The WHO recommends that for children weighing more than 25 kg, adult dosing guidelines should be followed using adult formulations.8 Children in the 25- to 37.9-kg weight band receive daily 8–10, 4–6, 22–32, and 15–22 mg/kg doses for rifampicin, isoniazid, pyrazinamide, and ethambutol in two adult FDC tablets. Like adults in the same weight band, they have predictably higher drug clearance per kilogram of body weight than larger adults resulting in markedly lower exposures of all four drugs than the average exposures reported in adults.1, 10 As most drugs penetrate cavities poorly, low drug exposures are of concern in older children who tend to develop cavitary disease as adults do. Population pharmacokinetic model-based simulations show that the weight-based dosing guidelines could be simplified resulting in improved pharmacokinetic indices using the same FDCs.10 Reduced exposure to rifampicin is of particular concern. Murine models demonstrate that the AUCs achieved in patients are close to the bottom of the exposure-response curve for bactericidal activity and that higher doses achieve more rapid sterilization, which prevents the development of resistance and relapse.3 Higher rifampicin exposures result in more rapid culture conversion amongst tuberculosis patients, and studies in adults investigating the treatment shortening potential of increased doses suggest that doses up to 4-fold the current dose are safe.4 Similarly acceptable short-term toxicity was reported in the Opti-Rif study, in which doses of 60–75 mg/kg in children approximated the AUC achieved in adults receiving 35 mg/kg daily.9 In summary, there is substantial evidence to support increased rifampicin doses in children. Doses in young children and those weighing more than 25 kg should at minimum achieve the exposures in adults on standard doses. Higher rifampicin doses may improve outcomes in the most vulnerable children, might allow the development of shorter treatment regimens with less potential for the development of resistance, and are likely to be safe. Ethambutol is used to prevent the development of resistance in children with more extensive disease, in HIV co-infected children, and in settings with high levels of isoniazid resistance. While relatively few studies have characterized ethambutol pharmacokinetics in children, the exposures reported are less than half those in adults.1 Hence, this evidence casts doubt on the role of ethambutol in children as the risks of ocular toxicity would need to be evaluated at higher doses, and alternative drugs such as levofloxacin or moxifloxacin could be considered. Until recently, the evidence to support dosing of second-line antituberculosis drugs to treat multidrug-resistant tuberculosis in children has been extremely limited, with mg/kg doses following those prescribed for adults. With intensified investment in the development of antituberculosis drugs over the past two decades, new regimens for the treatment of drug-susceptible and drug-resistant tuberculosis are emerging. Recent pharmacokinetic data for new and repurposed drugs such as levofloxacin, moxifloxacin, linezolid, and delamanid have supported access to improved treatment of drug-resistant tuberculosis in children, and data for rifapentine have facilitated simplified regimens to prevent tuberculosis. However, important gaps about pharmacokinetics and safety at optimal doses remain a barrier to their use in young children, particularly for bedaquiline, pretomanid, and clofazimine. The application of modelling and simulation to emerging data is critical to provide robust estimates to support dosing guidelines and to optimize the design of further studies. One-size-fits-all approaches to tuberculosis treatment are more easily adopted under program conditions, but the evidence in favour of a more precise approach based on patient and disease factors is mounting. Cavitation, higher baseline mycobacterial load in sputum, HIV infection, and suboptimal treatment adherence are associated with worse treatment outcomes.5 Higher drug doses, longer treatment, inclusion of drugs with good lesion penetration or long half-lives, or different drug delivery systems might be more effective for specific scenarios. Prolonged treatment is recommended for osteoarticular tuberculosis and tuberculous meningitis,8 but based on pharmacometric principles, a better approach would be to increase doses to obtain more effective drug concentrations at difficult-to-reach sites of infection. Ongoing studies to confirm this approach in children are supported by the marked improvement in survival with higher systemic concentrations of rifampicin amongst adults with tuberculous meningitis.6 Recently developed computational models have integrated data describing drug penetration into tuberculosis lesions, granuloma bacterial load, host immunity, and pharmacokinetics in mechanistic frameworks describing the effects of multiple drugs on mycobacterial dynamics, with translation into clinical outcomes. Foreseeably, such pharmacometric models, when scaled to paediatric patients with specific clinical scenarios, could be leveraged to design optimal drug combinations and doses. Pharmacokinetic studies underpin drug dosing in special populations. It is therefore important to confront issues which potentially hinder translation of pharmacokinetic findings into guidelines and improved dosing practices. The variability in pharmacokinetic measures reported between studies is considerable. While sources of pharmacokinetic variability are multiple,10 product preparation and food effects are sources of pharmacokinetic variability that should be highlighted in the paediatric context. Differences in the bioavailability of products on the market is a major concern and undermines the extent to which findings can be extrapolated. Interlaboratory differences in drug concentration measurement can be large, and this is another poorly appreciated source of between study pharmacokinetic variability. Improved opportunities for laboratories to participate in proficiency testing schemes offering external quality control for antituberculosis drug assays are sorely needed. Population pharmacokinetic methods provide a platform for pooling individual data across multiple studies and thus to attribute variability due to extraneous factors such as formulation or analytical laboratory and study population differences (e.g., due to genetic factors or nutritional status). Suitable formulations are essential to implement dosing recommendations, especially within in programme settings. Multiple requirements need to be simultaneously considered in the design of FDC formulations—limiting the number of FDCs required to simplify supply and distribution, while providing satisfactory drug delivery to the range of individuals comprising the target population, in whom optimal drug ratios may differ (due to, for example, to specific effects of growth and maturation on the disposition of different drugs). Pharmacometric tools based on population pharmacokinetic models have been developed to optimize FDC design.2 Although the expense of developing and distributing new formulations is considerable, increasingly accurate pharmacometric tools should lead to more durable dosing recommendations along with more confident investment by manufacturers in the development of suitable formulations. HM is supported by the Wellcome Trust (206379/Z/17/Z). HM declares that she has no conflicts of interest.