Government clinical trial investigators changed the primary metric for measuring the success of Gilead’s experimental drug remdesivir as a coronavirus treatment two weeks before Anthony S. Fauci’s announcement that the drug would be the new “standard of care.”

Instead of counting how many people taking the drug were kept alive on ventilators or died, among other measures, the National Institute of Allergy and Infectious Diseases said it would judge the drug primarily on a different outcome: how long it took surviving patients to recover.

Death and other negative outcomes were moved to secondary measure status: They would still be tracked, but they would no longer be the key measure of remdesivir’s performance. The switch — which specialists said is unusual in major clinical trials but not unheard of — was publicly disclosed on the government’s clinicaltrials.gov website on April 16 but did not receive much attention at the time.

The change reflects evolving scientific understanding of the fast-moving nature of the virus and uncertainties around how the lethal effects reveal themselves in patients, said NIAID, Gilead, and outside specialists. But the change also adds weight to the assessment of government and medical researchers that remdesivir is not a knockout drug that will change the trajectory of the coronavirus pandemic.

On Friday, as expected, the Food and Drug Administration approved an emergency use authorization for the drug that will allow it to be prescribed for hospitalized patients infected with the coronavirus.

The newly adopted criteria were a central feature of this week’s declaration by Fauci, NIAID’s director, that remdesivir reduced the time to recovery for surviving patients from 15 days to 11 days, a 31 percent improvement.

“The data shows that remdesivir has a clear-cut, significant, positive effect in diminishing the time to recovery,'' Fauci said as he sat in an Oval Office meeting with President Trump and other members of the president’s coronavirus task force. “It’s highly significant.''

The difference in death rate, one of the original primary measures, was not statistically significant, Fauci said, showing only a marginal reduction from 11 percent in patients given a placebo to 8 percent in patients given remdesivir. Full release of the trial results would be made soon, Fauci said.

Some medical research specialists questioned the change in the primary outcome measure of the trial, which had 1,063 patients.

“I think that they thought they weren’t going to win, and they wanted to change it to something they could win on,'' said Steven Nissen, a Cleveland Clinic cardiologist and expert clinical investigator who has led numerous drug trials. “I prefer the original outcome. It’s harder. It’s a more meaningful endpoint.

“Getting out of the hospital early is useful,'' he said, “but it’s not a game-changer.''

Nissen expressed dismay that the placebo phase of the trial was declared over so quickly, when a few more weeks might have provided a pool of patients large enough to show a statistically beneficial difference in death.

But Fauci said Wednesday that ethical considerations drove the announcement: As soon as a clear evidence of shorter hospitalizations was available, trial investigators owed it to patients on placebo to stop that phase of the trial so they could access the drug.

When the trial began on Feb. 21, it was designed to focus on collecting eight patient outcomes, to be measured on the 15th day after treatment. The list of outcomes was similar to guidance the World Health Organization issued in February for coronavirus clinical trials.

The original NIAID trial list started with death, followed by five categories of hospitalized patients: on ventilator or ECMO machine (which oxygenates blood outside the body); on high-flow oxygen therapy; on basic supplemental oxygen; not requiring oxygen but requiring ongoing care; not requiring care. The final two categories covered patients released from the hospital.

NIAID said in response to questions Thursday that it made the switch eight weeks later to the more limited measure of “time to recovery'' based on modeling that took into account new information about the course of the disease. The initial measurement period of two weeks, it said, was deemed to be too short as scientists learned more about the lengthy time patients could be seriously ill with covid-19, NIAID said.

“Little was known regarding the natural course of covid-19 when the trial was initially designed, and the initial endpoint chosen specified a single timepoint for evaluation, namely day 14,” the institute said. “However, with the growing knowledge during the epidemic, we learned that covid-19 had a more protracted course than previously known.

“NIAID statisticians performed modeling of what happens if the right day is not picked for assessment, which revealed that meaningful treatment effects could be missed with that primary endpoint,'' NIAID said. “Time to recovery avoids this issue, and the change in primary endpoint seemed appropriate given the evolving clinical data.''

Government researchers who decided to make the switch in outcome measure did not have access to clinical data, NIAID added.

Gilead did not respond to a question about whether it had input on the decision to change the endpoint.

“NIAID changed the primary endpoint while the study was blinded,'' Gilead spokesman Ryan McKeel said in an email, a decision he said was “based on continuing discussions and evolving understanding of the disease.'' He added that it is important to note that death, ventilation and other measures are included in the list of the trial’s secondary outcome measures “and will be reported when the data are published."

An exact timetable for publication of the results has not been disclosed. In the absence of any other treatment for coronavirus, Fauci declared that remdesivir would become the standard of care for certain hospitalized patients. Clinicians are eagerly awaiting more data from the trial so they will know which patients stand to benefit the most.

Fauci’s announcement coincided with more negative results for remdesivir published in the medical journal The Lancet; that trial, a Chinese study that was stopped early because it did not recruit enough patients, showed no benefit of remdesivir over placebo, and also no benefit in survival.

Some experts in clinical trials raised questions about the change in outcome measures.

“It raises a lot of flags, and it requires a lot of answers,” Walid F. Gellad, a professor of health policy and management at the University of Pittsburgh’s Department of Medicine, said in an interview, “especially when people start saying it’s become the standard of care, and all we saw was a news release in a trial with an outcome that was changed two weeks ago. It really is striking.''

An expert in transparency in clinical trials at the University of Oxford, Henry Drysdale, said the Oval Office setting of Fauci’s announcement, combined with the limited data disclosed, raised important questions about the outcomes change that will need to be answered when the full results are published in a peer-reviewed journal. Drysdale is part of a team of researchers who, in a landmark study published last year, found discrepancies in trial designs and reported outcomes in major academic journals.

“It’s extremely worrying that those very important outcomes were dropped from the primary outcome,'' Drysdale said in an interview. Asked to assess NIAID’s statement issued to reporters on Thursday, Drysdale said, “Whenever I see an explanation like this, when an outcome-switching has happened, that’s fine, but you were not open about this when you reported your quote-unquote exciting results."

But a strong degree of transparency about the change, combined with NIAID’s assurances that trial leaders did not know trial results when they switched the outcome, should settle any lingering questions, experts agreed.

Milton Packer, a cardiologist and clinical trial investigator at Baylor University Medical Center, said that, based on Fauci’s disclosures, roughly 95 people died in both arms of the trial. That combined number may have led investigators to believe they would not have a large enough sample of deaths to be statistically significant, he said.

“If you knew that the number of observations was inadequate to answer the question, and you did not know the breakdown, then shifting the endpoint is not a problem,” Packer said.