Comparison of different treatments for isoniazid-resistant tuberculosis: an individual patient data meta-analysis

Background: Isoniazid-resistant, rifampicin-susceptible (INH-R) tuberculosis is the most common form of drug resistance, and is associated with failure, relapse, and acquired rifampicin resistance if treated with first-line anti-tuberculosis drugs. The aim of the study was to compare success, mortality, and acquired rifampicin resistance in patients with INH-R pulmonary tuberculosis given different durations of rifampicin, ethambutol, and pyrazinamide (REZ); a fluoroquinolone plus 6 months or more of REZ; and streptomycin plus a core regimen of REZ. Methods: Studies with regimens and outcomes known for individual patients with INH-R tuberculosis were eligible, irrespective of the number of patients if randomised trials, or with at least 20 participants if a cohort study. Studies were identified from two relevant systematic reviews, an updated search of one of the systematic reviews (for papers published between April 1, 2015, and Feb 10, 2016), and personal communications. Individual patient data were obtained from authors of eligible studies. The individual patient data meta-analysis was performed with propensity score matched logistic regression to estimate adjusted odds ratios (aOR) and risk differences of treatment success (cure or treatment completion), death during treatment, and acquired rifampicin resistance. Outcomes were measured across different treatment regimens to assess the effects of: different durations of REZ (≤6 months vs >6 months); addition of a fluoroquinolone to REZ (fluoroquinolone plus 6 months or more of REZ vs 6 months or more of REZ); and addition of streptomycin to REZ (streptomycin plus 6 months of rifampicin and ethambutol and 1–3 months of pyrazinamide vs 6 months or more of REZ). The overall quality of the evidence was assessed using GRADE methodology. Findings: Individual patient data were requested for 57 cohort studies and 17 randomised trials including 8089 patients with INH-R tuberculosis. We received 33 datasets with 6424 patients, of which 3923 patients in 23 studies received regimens related to the study objectives. Compared with a daily regimen of 6 months of (H)REZ (REZ with or without isoniazid), extending the duration to 8–9 months had similar outcomes; as such, 6 months or more of (H)REZ was used for subsequent comparisons. Addition of a fluoroquinolone to 6 months or more of (H)REZ was associated with significantly greater treatment success (aOR 2·8, 95% CI 1·1–7·3), but no significant effect on mortality (aOR 0·7, 0·4–1·1) or acquired rifampicin resistance (aOR 0·1, 0·0–1·2). Compared with 6 months or more of (H)REZ, the standardised retreatment regimen (2 months of streptomycin, 3 months of pyrazinamide, and 8 months of isoniazid, rifampicin, and ethambutol) was associated with significantly worse treatment success (aOR 0·4, 0·2–0·7). The quality of the evidence was very low for all outcomes and treatment regimens assessed, owing to the observational nature of most of the data, the diverse settings, and the imprecision of estimates. Interpretation: In patients with INH-R tuberculosis, compared with treatment with at least 6 months of daily REZ, addition of a fluoroquinolone was associated with better treatment success, whereas addition of streptomycin was associated with less treatment success; however, the quality of the evidence was very low. These results support the conduct of randomised trials to identify the optimum regimen for this important and common form of drug-resistant tuberculosis. Funding: World Health Organization and Canadian Institutes of Health Research.