P28 Novel Bayesian network analysis allows systematic comparison of the safety and efficacy of different latent TB infection treatments. (14th November 2013)

• Record Type: Journal Article
• Authors: Stagg, HR; Zenner, D; Harris, RJ; Munoz, L; Lipman, M; Abubakar, I
• Is Part Of: Thorax. Volume 68(2013)Supplement 3
• Page Start: A86
• Page End: A87
• Publication Date: 2013-11-14
• Subjects: Chest -- Diseases -- Periodicals; Thorax; Chest -- Diseases; Periodicals; Periodicals; 617.54
• DOI: 10.1136/thoraxjnl-2013-204457.178
• Languages: English
• ISSNs: 0040-6376

Abstract

Abstract : Background: Although many randomised controlled trials (RCTs) and systematic reviews of treatment for latent tuberculosis (TB) infection (LTBI) have been conducted, previous analyses have not been able directly compare all utilised regimens. To address this we systematically searched for RCTs of LTBI treatment, then used a Bayesian network approach, which allows indirect head-to-head comparisons, to determine the most efficacious regimens at preventing active TB and those that caused the fewest adverse events. Methods: PubMed, EMBASE and Web of Science were systematically mined using a search strategy developed to find RCTs of LTBI treatment. Animal studies, non-RCTs, and RCTs without at least one of our two endpoints were excluded. No language restrictions were made. Extracted data were inputted into a full random effects mixed treatment compartment model, based on code by Ades, Welton and Lu, and implemented in WinBUGS. Odds ratios for all possible comparisons in the network and hierarchical rankings for the different treatments were obtained from the model with point estimates taken as the median of the posterior distribution and 95% credibility intervals (CrI) from the appropriate percentiles. Study quality was individually assessed. Results: 1, 344 publications were generated by our search strategy, of which 52 fitted our criteria. 31 studies contained extractable data on adverse events and 44 on the development of active TB. 14 regimens were compared; an extract of the full results is presented (Table 1 ). Conclusion: Our Bayesian approach allows a novel, integrated, overview of the comparative efficacy and safety of different LTBI regimens, as well as a clear identification of the knowledge gaps where inference is difficult due to sparse data. The results of our study can therefore be used to inform guidelines and plan vital future LTBI treatment RCTs.