Research in context
Evidence before this study
Drug-resistant tuberculosis is one of the major challenges impeding global tuberculosis control. Isoniazid-resistant, rifampicin-susceptible (INH-R) tuberculosis is the most common form of drug-resistant tuberculosis. In settings where drug susceptibility testing is not accessible or there is access only to Xpert MTB/RIF assay, INH-R tuberculosis will be missed and treated with standard regimens. Despite the frequent occurrence of INH-R tuberculosis and its major effect on outcomes, remarkably little research on therapy for INH-R tuberculosis has been done. As such, the optimum regimen for INH-R tuberculosis, including use of fluoroquinolones and duration of treatment, remains controversial. We built this individual patient data meta-analysis upon a 2017 systematic review and aggregate data meta-analysis, in which four electronic databases (Cochrane Database of Systematic Reviews and randomised trials, PubMed, Embase, and HealthStar) were searched with the terms “tuberculosis” AND “treatment” OR “therapy” AND “INH” OR “isoniazid resistance” for prospective or retrospective cohorts or randomised clinical trials published in English, French, or Spanish between Jan 1, 1948, and March 31, 2015. The systematic review found that treatment of patients with unrecognised INH-R tuberculosis with the standard regimen recommended for newly diagnosed patients would result in combined failure and relapse rates of 16%, and 8% of patients would acquire rifampicin resistance. All studies included in this review were deemed eligible for the individual patient data meta-analysis. We added previously excluded studies that might have been suitable for individual data analysis, plus studies included in a review on INH-R tuberculosis in children, and seven additional studies published after May 31, 2015, identified from an updated search finalised on Feb 10, 2016, using the same search terms and databases and eligibility criteria as the original review. Additionally, five of the authors contacted for papers identified during the systematic review provided other unpublished datasets (three now published), and three regional or national surveillance datasets were provided by authors responding to an invitation to all participants at a WHO European regional resistant tuberculosis surveillance meeting.
Added value of this study
To our knowledge, this is the first individual patient data meta-analysis on treatment outcome for INH-R tuberculosis. Individual-level data were compiled from 33 studies, and among 23 of these studies (21 cohorts and two randomised clinical trials), a total of 3923 patients received one of the regimens of interest. We assessed bias using an eight-item scale: two critical criteria (sampling method and outcome definition) and six important criteria (participation rate, attrition rate, completeness of information for age, HIV status, cavity at chest-x ray, and smear). On the basis of these criteria, the quality was judged low in one study, moderate in four, and high in the remainder. Compared with 6 months of REZ (rifampicin, ethambutol, and pyrazinamide), longer duration of REZ did not result in significantly improved treatment success or less acquired drug resistance. Addition of a fluoroquinolone to a core regimen of at least 6 months of REZ was associated with improved success and less acquired drug resistance, but no difference in mortality. Addition of a fluoroquinolone to a regimen with 2–3 months of pyrazinamide plus 6 or more months of rifampicin and ethambutol had no significant effect on success. The retreatment regimen (streptomycin added to 6 months of rifampicin and ethambutol plus 1–3 months of pyrazinamide) was associated with significantly worse success compared with at least 6 months of REZ.
Implications of all the available evidence
The findings of this study emphasise the importance of detecting INH-R tuberculosis and support the use of a fluoroquinolone in addition to a core regimen of 6 months of REZ. The addition of isoniazid, and prolongation of REZ beyond 6 months, seem to provide no benefits for this condition. Our results support a move away from the use of the streptomycin-containing retreatment regimens previously recommended. Because of the observational nature of the data, these results are graded very low quality evidence; as such, these results are insufficient to support strong treatment recommendations. However, the results do support the conduct of randomised trials to define the optimum treatment of this common and overlooked condition, particularly to assess the optimum type and duration of fluoroquinolone treatment and the optimum duration of pyrazinamide treatment.