People living with HIV and hospitalised with sepsis in Africa are at risk of death due to tuberculosis but diagnostics for tuberculosis might be delayed or inaccessible for people presenting for critical care. We aimed to compare the effects of immediate empirical and high-dose antituberculosis therapy on 28-day mortality in adult people living with HIV with sepsis in east Africa.

ATLAS was a phase 3, open-label, randomised, controlled, 2 × 2 factorial, superiority trial conducted at four hospitals in Tanzania and Uganda. Participants were aged 18 years or older, living with HIV, and had been admitted to hospital with sepsis with two or more modified quick Sequential Organ Failure Assessment score criteria. Exclusion criteria were active tuberculosis or receipt of antituberculosis therapy within 6 months of hospitalisation, pregnancy or lactation, allergies to antituberculosis therapy, another investigational drug within the past month, chronic liver disease, heavy alcohol use, positive serum cryptococcal antigen, or anticipated significant drug–drug interaction with rifampicin. A computer-generated permuted-block algorithm with allocation concealment and random block sizes of four and eight randomly assigned participants (1:1) to receive either immediate or diagnosis-dependent antituberculosis therapy, and (1:1) to receive either high-dose or conventional WHO-recommended weight-based dose antituberculosis therapy. Allocation was stratified by country and the presence of altered mental status at the time of randomisation. Participants randomly assigned to conventional-dose antituberculosis therapy received fixed-dose combination tablets of rifampicin approximately 10 mg/kg, isoniazid approximately 5 mg/kg, pyrazinamide, and ethambutol, plus pyridoxine 50 mg orally. Participants randomly assigned to receive high-dose antituberculosis therapy received rifampicin approximately 30 mg/kg and isoniazid approximately 7·5 mg/kg as a combination of single formulation tablets and fixed-dose combination tablets that included WHO-recommended weight-based doses of pyrazinamide and ethambutol, plus pyridoxine. Participants continued immediate antituberculosis therapy for 28 days. All medications were administered daily. Participants in the diagnosis-dependent antituberculosis therapy groups received treatment based on a clinical or microbiological diagnosis of tuberculosis. All participants received 2 g intravenous ceftriaxone daily for 7 days. The primary endpoint was 28-day mortality analysed in the modified intention-to-treat population and in the subgroup with later confirmed tuberculosis. We defined the survival time for each participant as the time from randomisation until death, discharged (alive) by day 28, or censored (alive) at day 28. We graded adverse events according to recommendations from the National Institutes of Health Division of AIDS. This study was registered with ClinicalTrials.gov , NCT04618198 , and is completed.

Between Jan 5, 2022, and Dec 9, 2024, 707 people were screened for eligibility and 437 were randomly assigned (110 to immediate conventional-dose, 112 to immediate high-dose, 107 to diagnosis-dependent conventional-dose, and 108 to diagnosis-dependent high-dose antituberculosis therapy). 395 patients (226 [57%] of whom were female and 169 [43%] were male; all participants were Black) received study intervention and were analysed for the primary outcome of 28-day mortality. We confirmed tuberculosis in 204 (52%) patients. There was no evidence of differences in 28-day mortality in immediate antituberculosis therapy groups (50 deaths [25%] in 198 patients) compared with diagnosis-dependent groups (50 deaths [25%] in 197 patients; adjusted hazard ratio [aHR] 0·99 [95% CI 0·67–1·46]; p=0·95), or in high-dose antituberculosis therapy groups (51 deaths [26%] in 199 patients) compared with conventional-dose groups (49 deaths [25%] in 196 patients; aHR 1·07 [0·72–1·59]; p=0·73). In patients with microbiologically confirmed tuberculosis, the 28-day mortality relative to the diagnosis-dependent conventional-dose group (18 deaths [34%] in 53 patients) was lower for the immediate conventional-dose group (six deaths [12%] in 51 participants; aHR 0·32 [95% CI 0·13–0·82]; p=0·015); for the diagnosis-dependent high-dose group (11 deaths [20%] in 56 patients) was 0·51 (0·24–1·08; p=0·079); and for the immediate high-dose group (11 deaths [26%] in 43 patients) was 0·63 (0·29–1·36; p=0·24). No significant differences in adverse events occurred between treatment groups but numerically more events of drug-induced liver injury occurred in the immediate high-dose group compared with any other group.

Among all participants with HIV-related sepsis, 28-day mortality was not significantly reduced with immediate or high-dose antituberculosis therapy. In the subgroup with later confirmed tuberculosis, immediate conventional-dose antituberculosis therapy significantly reduced 28-day mortality, suggesting, as in other forms of bacterial sepsis, that hours to active treatment might determine survival.

US National Institutes of Health.