When did the coronavirus arrive in the United States?

The first infection was confirmed on Jan. 21, 2020, in a resident of Washington State who had recently returned from Wuhan, China. Soon after, experts concluded that the virus had been in the country for weeks.

A study published on Tuesday offers new evidence: Based on an analysis of blood tests, scientists identified seven people in five states who may have been infected well before the first confirmed cases in those states. The results suggest that the virus may have been circulating in Illinois, for example, as early as Dec. 24, 2019, although the first case in that state was confirmed a month later.

But the new study is flawed, some experts said: It did not adequately address the possibility that the antibodies were to coronaviruses that cause common colds, and the results could be a quirk of the tests used. In addition, the researchers also did not have travel information for any of the patients, which might have helped explain the test results.

“This is an interesting paper because it raises the idea that everyone thinks is true, that there were infections that were going undiagnosed,” said Scott Hensley, an immunologist at the University of Pennsylvania.

But the small number of samples that tested positive made it difficult to be sure that they were true cases of infection and not just a methodological error. “It’s hard to know what is a real signal and what isn’t,” he said.

If the findings are accurate, however, then they underscore the notion that poor testing in the United States missed most cases during the early weeks of the pandemic.

“Without testing, you can’t see what’s going on,” said Keri Althoff, an epidemiologist at the Johns Hopkins Bloomberg School of Public Health and lead author of the study. “In these earlier months in some of these states where we were not suspecting, there was a lot of infection going on there.”

It is not a surprise that there may have been undocumented cases early in the pandemic, said Sarah Cobey, an evolutionary biologist at the University of Chicago. Experts “already knew that was the case from studying trends in excess mortality and hospitalizations,” she said.

Dr. Cobey’s most recent model estimated that there were roughly 10,000 infections in Illinois on March 1, 2020. “Given the horrible state of testing, there was never any doubt we were missing most early transmission,” she added.

In the study, published in the journal Clinical Infectious Diseases, Dr. Althoff and her colleagues analyzed blood samples from more than 24,000 people. They found nine people who had donated blood between Jan. 2 and March 18 of last year who appeared to have antibodies to the coronavirus.

Seven of the samples came from blood donated before the date of the first diagnosis in their states — Illinois, Wisconsin, Pennsylvania, Mississippi and Massachusetts. The results are consistent with those of another study that identified coronavirus antibodies in blood donated as early as mid-December 2019.

The participants were enrolled in a long-term National Institutes of Health project, called All of Us, that aims to include a million people in the United States in order to increase the representation of minorities in research. Only about half of the people in the study were white.

Early in the pandemic, the virus would have infected very few people. A low prevalence increases the odds that an antibody test mistakenly identifies a sample as having antibodies when it does not, Dr. Hensley said — a false positive.

The researchers tried to minimize that possibility by using two antibody tests in sequence. The first test flagged 147 samples as possibly having antibodies to the coronavirus; the second slashed that number down to nine.

The team also analyzed 1,000 samples of blood from the 2018-19 cold and flu season, and found none that tested positive for antibodies to the coronavirus.

“It’s still very possible that some of them might be false positives,” said Dr. Josh Denny, chief executive of All of Us. But “the fact that all of them would be false positives seems pretty unlikely with what we’ve done.”

The researchers said they planned to contact the participants to ask about travel history and would continue to analyze additional samples to estimate when the coronavirus reached American shores.

“The exact month at which it probably came into the U.S. is still unknown,” Dr. Althoff said. “It’s essentially a puzzle right now, and our study is just one piece of that puzzle.”