Empirical tuberculosis treatment in infants living with HIV hospitalised with severe pneumonia in Africa (EMPIRICAL): a multicentre, randomised, open-label, 2 × 2 factorial, controlled superiority trial.
Sara Domínguez-Rodríguez, Alfredo Tagarro, Cinta Moraleda et al.
✦ AI-curated · Sources linked
In 30 seconds
The EMPIRICAL trial was a multicentre, open-label, randomised 2 × 2 factorial controlled trial in infants aged 28–365 days hospitalised with severe HIV-associated pneumonia at 19 African hospitals. It compared empirical tuberculosis treatment added to standard care, with or without valganciclovir, against standard care with or without valganciclovir and tuberculosis treatment guided by test results or clinical suspicion. At 12 months, 132 (48%) of 276 infants in the empirical-treatment groups and 121 (43%) of 282 in the comparison groups had died; the adjusted hazard ratio was 1·12 (95% CI 0·87–1·43; p=0·3868), showing no evidence of improved survival.
Key findings
- At 12 months, mortality was 48% with empirical tuberculosis treatment and 43% with the comparison strategy.
- The adjusted hazard ratio for death with empirical treatment was 1·12 (95% CI 0·87–1·43; p=0·3868).
- In the comparison groups, 130 (46%) of 282 infants subsequently started tuberculosis treatment, at a median of 4·0 days after recruitment.
- Serious adverse events did not differ between groups (1·19, 0·82–1·73; p=0·3523); no treatment-related deaths occurred.
Why it matters
Severe HIV-associated pneumonia in infants carries substantial mortality, and tuberculosis may be difficult to diagnose promptly. In this trial, adding empirical tuberculosis treatment did not demonstrate a survival advantage over a strategy that initiated treatment based on microbiological findings or clinical suspicion. The results inform assessment of routine empirical treatment in this population while underscoring the importance of timely diagnostic and clinical evaluation.
What to keep in mind
The trial was open-label, and 17 (6%) and 20 (7%) participants in the respective groups were lost to follow-up. The confidence interval around the mortality estimate includes both lower and higher hazards, so the result does not establish equivalence.