Background: People with human immunodeficiency virus (HIV) and advanced immunosuppression initiating antiretroviral therapy (ART) remain vulnerable to tuberculosis (TB) and early mortality. To improve early survival, isoniazid preventive therapy (IPT) or empiric TB treatment have been evaluated; however, their benefit on longer-term outcomes warrants investigation. Methods: We present a 96-week preplanned secondary analysis among 850 ART-naive outpatients (≥13 years) enrolled in a multicountry, randomized trial of efavirenz-containing ART plus either 6-month IPT (n = 426) or empiric 4-drug TB treatment (n = 424). Inclusion criteria were CD4 count <50 cells/mm3 and no confirmed or probable TB. Death and incident TB were compared by strategy arm using the Kaplan-Meier method. The impact of self-reported adherence (calculated as the proportion of 100% adherence) was assessed using Cox-proportional hazards models. Results: By 96 weeks, 85 deaths and 63 TB events occurred. Kaplan-Meier estimated mortality (10.1% vs 10.5%; P = .86) and time-to-death (P = .77) did not differ by arm. Empiric had higher TB risk (6.1% vs 2.7%; risk difference, -3.4% [95% confidence interval, -6.2% to -0.6%]; P = .02) and shorter time to TB (P = .02) than IPT. Tuberculosis medication adherence lowered the hazards of death by ≥23% (P < .0001) in empiric and ≥20% (P < .035) in IPT and incident TB by ≥17% (P ≤ .0324) only in IPT. Conclusions: Empiric TB treatment offered no longer-term advantage over IPT in our population with advanced immunosuppression initiating ART. High IPT adherence significantly lowered death and TB incidence through 96 weeks, emphasizing the benefit of ART plus IPT initiation and completion, in persons with advanced HIV living in high TB-burden, resource-limited settings.