Levofloxacin versus placebo for the prevention of tuberculosis disease in child contacts of multidrug-resistant tuberculosis: study protocol for a phase III cluster randomised controlled trial (TB-CHAMP)

This trial specifically targets child household contacts of MDR-TB cases aged < 5 years for two reasons. First, children aged < five years are at the highest risk of progressing to TB disease following infection [2]. Second, concordance of drug susceptibility test (DST) results is high between adults with MDR-TB and young child household contacts [27‐29]. Young child household contacts are therefore most likely to benefit from MDR-TB preventive therapy. In addition, global policy and most national guidelines in high-burden settings recommend TB preventive therapy only for child contacts aged < 5 years or for children living with HIV, following exposure to a person with drug-susceptible TB [30]. Although children aged > 5 years, living with HIV, would likely benefit from preventive therapy, the potential for reduced concordance with the identified source case meant that the decision was made to restrict the trial to children aged < 5 years. Children aged < 5 years will be enrolled, regardless of tuberculin skin test (TST) or interferon-gamma release assay (IGRA) status. This is a deliberate decision to ensure programmatic relevance, since the WHO and most National TB Programmes in high-burden TB settings do not require a positive test of TB infection before initiating preventive therapy in children who have been exposed to a person living with infectious drug-susceptible TB [31].

We considered a wide range of anti-TB drugs and drug combinations for the intervention arm of the trial, including first-line TB drugs, third generation fluoroquinolones, ethionamide, para-aminosalicylic acid (PAS), cycloserine, linezolid and clofazimine, and the novel drugs, bedaquiline and delamanid. We discounted drugs if: (1) resistance was likely to be present to that drug in MDR strains (the first-line drugs) [32, 33]; (2) the drug is likely to be poorly tolerated (ethionamide and PAS); (3) the drug is associated with frequent or serious AEs (cycloserine, linezolid, clofazimine); (4) the drug is only weakly effective in killing M. tuberculosis (cycloserine, clofazimine, PAS); or (5) appropriate formulations are unavailable or pharmacokinetic parameters are not well understood (delamanid and bedaquiline).

We therefore decided to use a third-generation fluoroquinolone. These drugs have good efficacy against M. tuberculosis in vitro and are a core component of MDR-TB disease treatment regimens in both adults and children. Historically, clinicians had been hesitant to use fluoroquinolones in children following early animal research demonstrating cartilage damage in juvenile beagles [34]. However, a significant body of evidence has now demonstrated that drugs of this class are safe in children, even for long-term use [9, 35‐38]. Although moxifloxacin has good efficacy against metabolically active mycobacteria, levofloxacin may have better activity against metabolically inactive mycobacteria [39]. A mouse model which evaluated a number of novel TB infection treatment regimens in a latent infection model found that levofloxacin, moxifloxacin and isoniazid had similar efficacy against drug-susceptible strains of M. tuberculosis. Moxifloxacin is poorly tolerated in its current formulation due to its bitter taste and there are challenges in taste masking the active ingredient [40]. In addition, the large milligram size of the tablet (400 mg) makes dosing challenging in children. Levofloxacin, in contrast, has a low tablet milligram dose, is licensed for children, has been widely used to treat paediatric MDR-TB disease and other bacterial infections, has known pharmacokinetic parameters when used at once-daily dosing [41], does not interact with antiretroviral therapy drugs and appears to have low toxicity at currently used dosages [42].

Some limited rationale exists for the additional use of isoniazid in a preventive therapy regimen, as it would be efficacious in preventing disease if the child had been exposed to a person with drug-susceptible TB in addition to the identified MDR-TB case. As young children are likely to be highly concordant with the identified source case, we anticipate few children acquiring drug-susceptible organisms. Isoniazid may also provide some efficacy against MDR strains with inhA promoter region mutations (typically conferring low-level resistance) [43]. Globally, however, isoniazid resistance is more commonly caused by katG gene mutations [44], in which isoniazid is less likely to be effective. For these reasons we elected not to include isoniazid in the intervention regimen. Paediatric pharmacokinetic and safety data for delamanid are becoming increasingly available and the WHO now recommends delamanid for the treatment of MDR-TB disease in children aged as young as six years, but not yet to the youngest children, where treatment of infection is most needed [45]. We plan to adapt the trial design to include delamanid to treat child contacts of fluoroquinolone-resistant MDR-TB, in an observational cohort, once pharmacokinetic data and formulations become available for the youngest children, dependent on additional funding.

In designing our intervention, we also considered that a simple regimen with one drug would likely be safer and would facilitate easier adherence for children. Although concerns have been raised that the use of a single drug could lead to the acquisition of secondary resistance in individuals who develop TB disease, this has not been shown to occur with the use of isoniazid as a single agent for TB preventive therapy in adult and child contacts of drug-susceptible TB [21, 46]. We therefore decided to use levofloxacin alone.

As outlined above, isoniazid may be active against M. tuberculosis with inhA promoter region mutations, as well as susceptible strains from exposures other than the identified index case. For these reasons, the PHOENIx randomised controlled trial (A5300/IMPAACT2003), evaluating MDR-TB infection treatment with delamanid, chose isoniazid for their control arm. However, as also stated above, we anticipate high concordance between index case and child contact (with few children infected with drug-susceptible strains), as well as globally limited utility of isoniazid to treat MDR strains (due to the preponderance of katG gene mutations worldwide). In addition, the inclusion of isoniazid would increase the (albeit low) risk of harm that could arise through the use of a potentially hepatotoxic drug. Given the lack of evidence for the efficacy for isoniazid to treat MDR-TB infection and the lack of guidance on this topic, the protocol team, after broad scientific and ethical consideration, determined that it would be most appropriate to use no active drug in the control arm. This represents current WHO guidance [31].

We considered using an open-label design. However, the primary trial endpoint, incident TB, although standardised to as great a degree as possible, remains, to some extent, subjective. Many children who develop incident TB during the trial will be diagnosed clinically (i.e. not bacteriologically confirmed). The study team felt that with an open-label design there may be bias in the ascertainment of the endpoint. Children known not to be receiving any medication may be followed more closely and investigated more intensively than children known to be on treatment. In addition, there would be potential for bias in the ascertainment and reporting of AEs, with families/caregivers potentially more likely to report AEs if they knew their child was taking an active treatment. We therefore decided blinding and use of placebo would support unbiased assessment of endpoints and toxicity.

As we planned to identify and treat children in their households, we felt that all children in the same household should be randomised to the same trial arm. Randomisation by household allows for ease of drug administration and protection from bias through accidental or deliberate switching of tablets (levofloxacin/placebo) between child household contacts. The statistical implications of the clustering are accounted for in the sample size estimation and analysis plan.

Adult MDR-TB index cases are identified via routine National TB Programme TB clinics/hospitals or via other (e.g. laboratory-based) surveillance methods and assessed for eligibility as index cases (Table 2). Adults are approached at their local clinic/hospital or at home to provide consent for collection of their TB episode data and a sputum sample, and for permission to conduct a home visit to enumerate all household members and eligible child contacts (Fig. 1).

Following the evaluation of an adult MDR-TB index case, home visits are undertaken and the household contacts enumerated. Comprehensive infection control procedures are followed to protect staff. Screening is conducted in all household contacts to identify those with prevalent TB disease, to rule out additional individuals who may have drug-susceptible-TB and to identify child household contacts who do not have prevalent TB disease and who are eligible for enrolment. Household characteristics and children’s TB exposure are captured using a standard approach [48]. All household members aged ≥ 5 years with suggestive symptoms are referred for TB investigation and TB/HIV care as appropriate.

Households are defined using a standard inclusive definition [49]. A household contact is defined as a person who currently lives or lived in the same dwelling unit or plot of land and shares or shared the same housekeeping arrangements as the adult MDR-TB index case, and where there is reported exposure within six months before the index case starting MDR-TB treatment.

Written informed consent from caregivers is obtained to screen for study entry and eligibility (Table 2). Children aged < 5 years who are identified as having been living in the same household as an enrolled adult MDR-TB index case at any point during the preceding six months are eligible, if the exposure has lasted > 2 weeks. Anthropometric measurements are completed, TB symptoms and signs captured, and chest radiographs (CXR), HIV testing and IGRA (QuantiFERON-TB Gold Plus; Qiagen) done. Any children with suggestive symptoms or signs of intrathoracic or extrathoracic TB disease, or with concerning CXR features, are investigated for TB disease (including sampling for bacteriological testing) and are referred for TB care if TB is diagnosed. If any adult is identified with isoniazid- or rifampicin-susceptible TB in the household, then all children in the household become ineligible. Any child found to have been in contact with an index case who has isoniazid- or rifampicin-susceptible TB will be referred to the local clinic to access preventive therapy, as per WHO guidelines. Prevalent TB in child contacts is defined using standard international case definitions for paediatric TB in contact investigation studies (Table 1) which have been harmonised between three international MDR-TB preventive therapy protocols (TB-CHAMP, V-QUIN and the ACTG/IMPAACT PHOENIx trials). CXR findings suggestive of intrathoracic (pulmonary) TB are defined in Table 3.

Carers/legal guardians of child household contacts are approached; consent is obtained for screening and, if eligibility is confirmed, subsequent enrolment.

All eligible contacts in a household consenting to participate are allocated to the same study arm, but participation of all eligible child household contacts is not required for the household to participate. Enrolment of children and randomisation of households into the study is via a centralised web-based system to maintain allocation concealment. Randomisation is stratified by site and households are randomised 1:1 to either levofloxacin or placebo. The randomisation lists are prepared by an individual who is not involved in the day-to-day running of the trial.

Medications are dosed by pragmatic weight-bands (Table 4). Children randomised to the intervention arm receive levofloxacin at a target dose of 15–20 mg/kg per day once daily for 24 weeks. Although this is the recommended target dose [50], we acknowledge that higher dosages may be needed to achieve the serum concentrations seen in adults when dosed with 750 mg once daily [51]. In addition, as can be seen in Table 4, the use of weight-bands leads to some over-dosing and some under-dosing for children at the boundaries of each weight-band. Children randomised to the control arm receive placebo formulated to look and taste identical to the 250 mg (or 100 mg) levofloxacin dispersible tablet. Drug (and placebo) is being supplied initially as 250 mg tablets. During the course of the trial, a 100 mg dispersible tablet (and placebo) is expected to become available for use in the lower weight-bands.

Children are seen at 4, 8, 12, 16, 24, 36, 48, 72 and 96 weeks for study visits and at any time that caregivers feel that the child may have developed TB disease or they, or the study team, have other concerns.

Written informed consent is obtained from the adult MDR-TB index case to approach the household and the parent/legal guardian of the eligible child for their screening and possible subsequent enrolment into the trial. The consent process is conducted in the home language of the person from whom consent is requested and that individual is given a written information sheet, also in their home language, explaining the study. Consent is obtained before any trial-related procedures are performed.

The trial has been registered, and details can be viewed at: http://​www.​isrctn.​com/​ISRCTN92634082. The trial has been approved by the Stellenbosch University Human Research Ethics Committee (M16/02/009), the University of the Witwatersrand Ethics Committee (160409) and the Medicines Control Council, South Africa.

The trial is run primarily by the Trial Management Group and is overseen by the Trial Steering Committee. An Independent Data Monitoring Committee sees confidential, unblinded data for the trial and advises the Trial Steering Committee on whether the trial needs to be prematurely closed. An Endpoint Review Committee, blinded to treatment arm, evaluates clinical trial endpoints and causes of death.

The Social Science team will describe: (1) the perceptions that patients, families and health workers have about TB care and prevention, as well as the TB-CHAMP trial; (2) how the study intervention is implemented in the context of local health systems; and (3) the families’ experiences of the trial. This involves key informant interviews with health service personnel, mixed-methods research around implementation and a nested qualitative cohort.

Costing and cost-effectiveness of the trial interventions for both families and health systems at the three collaborating sites will be evaluated. This involves collecting data on costs associated with the interventions as well as costs associated with children developing MDR-TB disease. The primary outcome is the incremental cost-effectiveness ratio of levofloxacin against placebo.

Palatability and acceptability of the new child-friendly dispersible 100 mg levofloxacin formulation has been evaluated in 24 children before trial start-up as part of a lead-in pharmacokinetic sub-study. Bioavailability is also being evaluated in healthy adult volunteers. Ongoing evaluation of the 250 mg formulation (and 100 mg formulation when it becomes available) will take place throughout the trial using questionnaires and qualitative research.