Design: Pragmatic, individually randomised superiority trial, 11 hospitals
Population: 1,172 adults with HIV admitted to hospital in Tanzania and Mozambique (median CD4 232)
Primary outcome: Microbiologically confirmed TB starting treatment within 72 h
Effect: 16.0% vs 15.3% (difference 0.7%, 95% CI −3.4 to 4.8); 8-week mortality HR 0.86 (0.69–1.07)

EXULTANT, published in May 2026, randomised 1,172 adults with HIV admitted to 11 hospitals in Tanzania and Mozambique. The intervention was universal testing within 24 hours regardless of symptoms: Xpert MTB/RIF Ultra on sputum, stool and urine, plus urine lateral flow lipoarabinomannan (LF-LAM). The control was WHO-recommended, symptom-guided sputum Xpert Ultra and LF-LAM. Median CD4 was 232 cells/µL and 75% were on antiretroviral therapy.

Microbiologically confirmed TB treated within 72 hours occurred in 16.0% vs 15.3% (difference 0.7%, 95% CI −3.4 to 4.8). Eight-week mortality was 25.8% vs 28.8% (HR 0.86, 0.69 to 1.07). Time to treatment was under a day in both groups.

The control arm already worked well: 86% had TB-compatible symptoms and 91% qualified for LAM. In settings where symptom-guided testing is done properly, adding more samples added little. Mortality of one in four at eight weeks shows the problem lies beyond diagnosis.

In adults with HIV admitted to hospital, apply WHO-recommended testing reliably on day one: sputum Xpert Ultra for TB symptoms, and urine LF-LAM for symptoms, advanced HIV disease, serious illness or CD4 below 200.
Urine LAM needs no sputum and is especially useful when the patient cannot produce a sample.
Adding stool and urine Xpert for everyone did not significantly increase confirmed diagnoses or reduce deaths.
Eight-week mortality was about 1 in 4; look for other causes of death such as cryptococcal disease and sepsis.

Why it matters

It suggests effort in HIV inpatient care is better spent on getting standard testing done than on testing more samples from everyone.

The statistics, in plain English

The difference in the primary outcome (0.7%) has a 95% CI from −3.4 to 4.8, so any real effect is small. The mortality hazard ratio of 0.86 crosses 1.0 (0.69 to 1.07), so the lower death rate may be chance.