Treatment of latent tuberculosis infection: a network meta-analysis.

ABSTRACT

BACKGROUND: Effective treatment of latent tuberculosis infection (LTBI) is an important component of TB elimination programs. Promising new regimens that may be more effective are being introduced. Because few regimens can be directly compared, network meta-analyses, which allow indirect comparisons to be made, strengthen conclusions. PURPOSE: To determine the most efficacious regimen for preventing active TB with the lowest likelihood of adverse events to inform LTBI treatment policies. DATA SOURCES PubMed, EMBASE, and Web of Science up to 29 January 2014; clinical trial registries; and conference abstracts. STUDY SELECTION Randomized, controlled trials that evaluated LTBI treatment in humans and recorded at least 1 of 2 prespecified end points (preventing active TB or hepatotoxicity), without language or date restrictions. DATA EXTRACTION Data from eligible studies were independently extracted by 2 investigators according to a standard protocol. DATA SYNTHESIS: Of the 1516 articles identified, 53 studies met the inclusion criteria. Data on 15 regimens were available; of 105 possible comparisons, 42 (40%) were compared directly. Compared with placebo, isoniazid for 6 months (odds ratio [OR], 0.64 [95% credible interval {CrI}, 0.48 to 0.83]) or 12 months or longer (OR, 0.52 [CrI, 0.41 to 0.66]), rifampicin for 3 to 4 months (OR, 0.41 [CrI, 0.18 to 0.86]), and rifampicin-isoniazid regimens for 3 to 4 months (OR, 0.52 [CrI, 0.34 to 0.79]) were efficacious within the network. LIMITATIONS: The risk of bias was unclear for many studies across various domains. Evidence was sparse for some comparisons, particularly hepatotoxicity. CONCLUSION: Comparison of different LTBI treatment regimens showed that various therapies containing rifamycins for 3 months or more were efficacious at preventing active TB, potentially more so than isoniazid alone. Regimens containing rifamycins may be effective alternatives to isoniazid monotherapy. PRIMARY FUNDING SOURCE None.