Scale-up of TB/HIV care in a high HIV prevalence setting changed the patterns of the TB epidemic. TB notification declined and gender disparity shifted from more women affected in earlier periods to more male TB cases in recent years. An explanation may be that women better accessed ART during TB/HIV care decentralization resulting in better immune reconstitution at population level and thus being less susceptible to TB. Decentralization of TB/HIV care should be prioritized for improved access to health care and ART to control the TB epidemic in high HIV prevalence settings. Special attention need to be paid to mortality in HIV co-infected TB cases. New programmatic approaches and diagnostic tools are deemed necessary for more rapid TB case detection and earlier treatment initiation in severely immunosuppressed patients. Methods Study eligibility was defined as all ≥16 year old patients initiating first line TB drug treatment at 3 secondary and 22 primary care facilities in Shiselweni region, from 01/2009 to 12/2013. i) We described temporal trends in TB epidemiology of new and recurrent first line drug treatment cases using frequency statistics (01/200912/2013). ii) For HIV/TB co-infected patients, logistic regression was used to assess associations between baseline factors (calendar year, primary/secondary care, sex, age, new-/re-treatment case, TB-site, CTX and ART status, CD4 count) and the composite endpoint LTFU/death/treatment failure (01/2009-12/2012). Background Tuberculosis (TB) is the main cause of mortality in HIV co-infected individuals. Swaziland has one of the highest HIV prevalence (31% [1]) and TB incidence (1,380/100,000) in the world. Since 2007, TB/HIV care was decentralized in the rural Shiselweni region (Swaziland) and coinciding with increasing antiretroviral-therapy (ART) coverage from an estimated 40% in 2009/10 to 80% in 2012. Concurrently, TB case finding was accelerated through changes in national TB guidelines, intensified TB symptomatic screening and role out of new diagnostic tools (GeneXpert, TLA (thin layer agar method)).