The scientists implanted the human glioblastomas inside the brains of mice, then injected the experimental virus directly into the tumors.

Untreated mice died in 19 days, but 60 percent of the treated mice were alive and thriving for four months. Healthy tissue nearby remained unharmed.

Scientists euthanized the survivors to see what was happening inside their brains — and found only empty cavities and scar tissue where the tumors once were.

"Everyone here is excited about it because we've never seen anything happen with the mice like that," says lead researcher Dr. Frederick Lang, a neurosurgeon at Houston's M.D. Anderson Cancer Center.

He cautioned that the dramatic results don't assure the virus will work in people: Scientists have cured lots of mice of cancer only to see the therapies fail in patients.

"This is an interesting study," said Dr. Len Lichtenfeld of the American Cancer Society (news - web sites). But he echoed Lang's caution, adding that mutant viruses could prove too toxic to use.

"We need to be very, very careful" in studying the experimental treatment, Lichtenfeld said. "There are too many situations where doctors and patients and families have gotten very, very excited about drugs and it's turned out ... that the drugs weren't effective. We need to avoid that."

Still, the National Cancer Institute (news - web sites) is intrigued enough that it is providing $1 million to produce enough of the mutant virus to begin human testing, said Lang, who hopes to start enrolling brain-tumor patients in a study of the treatment by winter 2004.

The virus should target other solid tumors, too, he said.

But "if there's any disease that needs a novel approach, it's really brain tumors," said Lang, who reported the experiment in this week's Journal of the National Cancer Institute.

Glioblastomas are the most common primary brain tumor in adults, striking about 7,000 Americans a year, and the most lethal. Survival is only about a year, a dismal rate that hasn't changed in decades despite improvements in surgery and radiation and chemotherapy treatments.

A question is whether people's immune systems will attack the mutant virus before it can penetrate and spread through a brain tumor. The virus would have to be injected through the skull directly into the brain tumor, which could delay the immune reaction.