Lancet. 2026 Nov 14;408(10568):2109-2124. doi: 10.1016/S0140-6736(26)01418-2.
ABSTRACT
BACKGROUND: Mortality among infants with severe HIV-associated pneumonia remains high globally. The EMPIRICAL trial evaluated whether empirical tuberculosis treatment (eTB-T) improves survival compared with the currently recommended microbiologically and clinically guided strategy.
METHODS: This multicentre, open-label, group-sequential designed, 2×2 factorial, randomised, controlled, superiority trial was conducted in 19 hospitals across Côte d’Ivoire, Malawi, Mozambique, Uganda, Zambia, and Zimbabwe. Infants aged 28-365 days admitted with severe HIV-associated pneumonia, without a current tuberculosis diagnosis or contact, were centrally and individually randomly assigned (1:1:1:1) using a secure web-based system stratified by site and severity. A computer-generated block randomisation schedule was used, with fixed blocks of eight allocations (two allocations to each factorial treatment group) within each stratum. Participants received standard of care (SoC) comprising treatment for bacterial (WHO-recommended antibiotics) and Pneumocystis jirovecii (cotrimoxazole and steroids) pneumonia, SoC plus 15 days of oral valganciclovir (16 mg/kg per 12 h), or SoC plus 6 months of tuberculosis treatment (isoniazid, rifampicin, pyrazinamide, and ethambutol for 2 months, plus isoniazid and rifampicin for 4 subsequent months), or both interventions combined. Here, we report the comparison between the eTB-T (ie, SoC plus tuberculosis treatment and SoC plus oral valganciclovir plus tuberculosis treatment) and non-eTB-T (ie, SoC or SoC plus 15 days of oral valganciclovir) groups. Nasopharyngeal aspirates, stools for Xpert MTB/RIF Ultra testing, and urine for tuberculosis lipoarabinomannan assay were collected at recruitment. In the non-eTB-T groups, tuberculosis treatment was initiated if microbiological results collected at recruitment yielded positive results or if, during follow-up, the treating clinician suspected tuberculosis on clinical grounds, including radiological findings, in accordance with routine clinical practice and guidelines. The primary endpoint was all-cause mortality (at 15 days and 360 days). Survival was estimated using flexible parametric models with covariate-adjusted conditional estimates; missing data were imputed by chained equations. The trial is registered with ClinicalTrials.gov (NCT03915366) and is now complete.
FINDINGS: Between March 15, 2020, and Jan 31, 2024, 563 participants were enrolled; 558 were analysed after excluding five patients who did not meet eligibility criteria. 276 participants were allocated to the eTB-T groups (n=140 eTB-T and n=136 valganciclovir plus eTB-T) and 282 to the non-eTB-T groups (n=142 SoC and n=140 valganciclovir). 17 (6%) of 276 in the eTB-T groups and 20 (7%) of 282 in the non-eTB-T groups were lost to follow-up. Median age was 4·4 months (IQR 3·2-7·4) and 274 (49%) of participants were female and 284 (51%) were male. In the non-eTB-T groups, 130 (46%) of 282 participants eventually initiated tuberculosis treatment at a median of 4·0 days (IQR 3·0-15·0) after recruitment due to positive microbiological findings in 55 (20%) cases (53 confirmed by tuberculosis lipoarabinomannan assay and two by nasopharyngeal aspirates Xpert MTB/RIF) or clinical decision in 75 (27%) cases. There was no evidence of interaction between valganciclovir and eTB-T (adjusted hazard ratio 1·19, 95% CI 0·73-1·95; pheterogeneity=0·4836). At 12 months, 132 (48%) of 276 participants in eTB-T groups and 121 (43%) of 282 participants in non-eTB-T groups died. In the eTB-T groups, the adjusted hazard ratio for death was 1·12 (95% CI 0·87-1·43; p=0·3868). At 15 days, the odds ratio was 1·24 (95% CI 0·83-1·84; p=0·2950). Serious adverse events did not differ between groups (1·19, 0·82-1·73; p=0·3523). There were no treatment-related deaths. The most common adverse events were elevation of transaminases (72 [26%] cases in the eTB-T groups and 74 [26%] in the non-eTB-T group), anaemia (163 [59%] cases in the eTB-T groups and 169 [60%] in the non-eTB-T groups), and infectious disease requiring antimicrobial treatment (161 [58%] cases in the eTB-T groups and 176 [62%] in the non-eTB-T groups).
INTERPRETATION: In contrast to empirical treatment with valganciclovir, there was no evidence that eTB-T was superior to a clinically and microbiologically guided strategy in reducing all-cause mortality among infants with HIV-associated pneumonia. However, the high proportion of infants who eventually initiated tuberculosis treatment during follow-up in the non-eTB-T groups might have reduced the contrast between strategies and attenuated the ability to detect differences in outcomes.
FUNDING: European and Developing Countries Clinical Trials Partnership.
PMID:43020097 | DOI:10.1016/S0140-6736(26)01418-2
