BACKGROUND: Isoniazid preventive therapy prevents active tuberculosis in people with HIV, but previous studies have found no evidence of benefit in people with HIV who had a negative tuberculin skin test, and a non-significant effect on mortality. We aimed to estimate the effect of isoniazid preventive therapy given with antiretroviral therapy (ART) for the prevention of tuberculosis and death among people with HIV across population subgroups. METHODS: We searched PubMed, Embase, the Cochrane database, and conference abstracts from database inception to Jan 15, 2019, to identify potentially eligible randomised trials. Eligible studies were trials that enrolled HIV-positive adults (age ≥15 years) taking ART who were randomly assigned to either daily isoniazid preventive therapy plus ART or ART alone and followed up longitudinally for outcomes of incident tuberculosis and mortality. We approached all authors of included trials and requested individual participant data: coprimary outcomes were relative risk of incident tuberculosis and all-cause mortality. We did a single-stage meta-analysis of individual participant data using stratified Cox-proportional hazards models. We did prespecified subgroup analyses by sex, CD4 cell count, and evidence of immune sensitisation to tuberculosis (indicated by tuberculin skin test or interferon-γ release assays [IGRAs]). We also assessed the relative risk of liver injury in an additional prespecified analysis. This study is registered with PROSPERO, CRD42019121400. FINDINGS: Of 838 records, we included three trials with data for 2611 participants and 8584·8 person-years of follow-up for the outcome of incident tuberculosis, and a subset of 2362 participants with 8631·6 person-years of follow-up for the coprimary outcome of all-cause mortality. Risk for tuberculosis was lower in participants given isoniazid preventive therapy and ART than participants given ART alone (hazard ratio [HR] 0·68, 95% CI 0·49-0·95, p=0·02). Risk of all-cause mortality was lower in participants given isoniazid preventive therapy and ART than participants given ART alone, but this difference was non-significant (HR 0·69, 95% CI 0·43-1·10, p=0·12). Participants with baseline CD4 counts of less than 500 cells per μL had increased risk of tuberculosis, but there was no significant difference in the benefit of isoniazid preventive therapy with ART by sex, baseline CD4 count, or results of tuberculin skin test or IGRAs. 65 (2·5%) of 2611 participants had raised alanine aminotransferase, but data were insufficient to calculate an HR. INTERPRETATION: Isoniazid preventive therapy with ART prevents tuberculosis across demographic and HIV-specific and tuberculosis-specific subgroups, which supports efforts to further increase use of isoniazid preventive therapy with ART broadly among people living with HIV. FUNDING: National Institutes of Health and National Institute of Allergy and Infectious Diseases.