On Monday, they posted their data online, in advance of full peer-review.
They compared 2,104 coronavirus patients who received dexamethasone with 4,321 patients who did not. Overall, 21.6% of patients who got dexamethasone and 24.6% of those who got standard care died within 28 days. But the effect on mortality rates varied depending on how sick patients were when they entered the study.
Dexamethasone reduced the death rate by one-third in patients who needed a ventilator to help them breath (from 40.7% to 29.0%) and by one-fifth in patients receiving supportive oxygen without an invasive ventilator (from 25.0% to 21.5%).
The steroid did not reduce deaths in patients not receiving respiratory support.
The World Health Organization has said dexamethasone should be reserved for serious COVID-19 cases in which it has been shown to provide benefits.