Bird flu is believed to be a rare disease that kills more than half of the people it infects, but a US study out Thursday suggests it may be more common and less lethal than previously thought.

The research could help soothe concerns about the potential for a deadly pandemic that may kill many millions of people, sparked by the recent lab creation of a mutant bird flu that can pass between mammals.

Researchers at Mount Sinai School of Medicine in New York analyzed 20 previous international studies that tested the blood of nearly 13,000 participants worldwide, according to the study in the journal Science.

They found that between one and two percent of those tested showed evidence of a prior H5N1 avian flu infection, meaning millions of people may have been infected around the globe.

The World Health Organization's figures currently show just 573 cases in 15 countries since 2003, with 58.6 percent of those resulting in death.

The researchers said the WHO may be overlooking cases by focusing only on hospitalizations and severe illnesses, and recommended a new approach to calculating the true number of bird flu cases.

The findings could also mean that the death rate from bird flu is underestimated, largely because many of the people who get sick from it live in rural farming areas where medical care may be difficult to come by.

"We suggest that further investigation, on a large scale and by a standardized approach, is warranted to better estimate the total number of H5N1 infections that have occurred in humans," the authors wrote.

Researchers in the Netherlands and the United States have sparked international alarm with lab research that was successful in creating a mutant form of bird flu that was found to be transmissible among ferrets.

US health authorities have urged major science journals to publish only heavily edited forms of the studies in order to prevent the data from falling into terrorists' hands.

However, an international group of experts meeting at WHO headquarters in Geneva last week decided that the studies should eventually be published in full, but that a further risk assessment is needed before that can happen.