Background The implementation of antiretroviral (ARV) therapy caused a significant decrease in HIV-associated mortality worldwide. Nevertheless, mortality is still high among people living with HIV/AIDS and tuberculosis (TB). ARV-naïve HIV patients coinfected with tuberculosis (TB) have more options to treat both diseases concomitantly. Nevertheless, some TB-HIV patients undertaking ARVs (ARV-experienced) are already failing the first line efavirenz-based regimen and seem to display different response to second line ARV therapy and exhibit other predictors of mortality. Methods We performed a retrospective cohort study including 273 patients diagnosed with TB-HIV and treated at a referral center in Rio de Janeiro, Brazil, between 2008 and 2016. Multivariate analysis and Cox regression models were used to evaluate the effectiveness of ARV therapy regimens (viral load [VL] <80 copies from the 4th to 10th months after TB therapy introduction) and to identify predictors of early mortality (100 days after TB therapy initiation) considering ARV-naïve and ARV-experienced patients adjusting for sociodemographic, clinical and therapeutic covariates. Findings Survival analysis included 273 patients, out of whom 154 (56.4%) were ARV-naïve and 119 (43.6%) were ARV-experienced. Seven deaths occurred within 6 months of anti-TB treatment, 4 in ARV-naïve and 3 in ARV-experienced patients. Multivariate analysis revealed that in ARV-naïve patients, the chance of death was substantially higher in patients who developed immune reconstitution inflammatory syndrome during the study follow up (HR = 40.6, p<0.01). For ARV-experienced patients, similar analyses failed to identify factors significantly associated with mortality. Variables independently associated with treatment failure for the ARV-naïve group were previous TB (adjusted OR [aOR] = 6.1 p = 0.03) and alcohol abuse (aOR = 3.7 p = 0.01). For ARV-experienced patients, a ritonavir boosted. Protease Inhibitor-based regimen resulted in a 2.6 times higher risk of treatment failure compared to the use of efavirenz based ARV regimens (p = 0.03) and High baseline HIV VL (p = 0.03) were predictors of treatment failure. Conclusions Risk factors for mortality and ARV failure were different for ARV-naïve and ARV-experienced patients. The latter patient group should be targeted for trials with less toxic and rifampicin-compatible drugs to improve TB-HIV treatment outcomes and prevent death.