WASHINGTON - Government scientists turned regular blood cells into tumor attackers that wiped out all signs of cancer in two men with advanced melanoma. The striking finding, unveiled Thursday, marks an important step in the quest for gene therapy for cancer.

But the genetically altered cells didn't help 15 other melanoma victims. So scientists are trying to strengthen the shots.

Still, the National Cancer Institute called its experiment the first real success in cancer gene therapy because it fought cancer's worst stage, when it has spread through the body, unlike earlier attempts that targeted single tumors.

The government hopes to soon begin testing the gene therapy in small numbers of patients dying of more common cancers, such as advanced breast or colon cancer.

The hope is that, one day, such treatment will provide long-lasting tumor suppression.

"It's not like chemotherapy or radiation where as soon as you're done, you're done," said Dr. Steven Rosenberg, lead researcher and the institute's surgery chief. "We're giving living cells, which continue to grow and function in the body."

The first two successful patients appear melanoma-free almost two years after infusions of tumor fighters made from their own blood. Doctors can't predict how the men will fare long-term. Melanoma, the most aggressive skin cancer and killer of almost 8,000 Americans a year, is notorious for returning years after patients think they've subdued it.

"I'm cured for now," is how a grateful Mark Origer, 53, of Watertown, Wis., put it after a checkup from institute doctors this week. "I know how fortunate I am to have gone through this and responded to this. Not everybody's that lucky."

Cancer specialists praised the work, published by the journal Science, but warned that years of additional research are needed.

"Clearly, this is a first step," cautioned Dr. Len Lichtenfeld of the American Cancer Society. "We have to be very cautious about not raising hopes too much."

But "it is exciting," he added. "It certainly is a proof of concept that this approach will work."

The institute's Rosenberg has long led a tantalizing research field: how to harness the body's immune system to fight cancer.

White blood cells called T-lymphocytes hunt down germs and other foreign tissue. But cancerous cells look a lot like healthy cells, making it hard for those T-cells to spot a problem.

By 2002, Rosenberg had made a breakthrough. He found small numbers of cancer-fighting T-cells inside some patients with advanced melanoma.

He literally pulled those cells out of their blood and grew billions more of them in laboratory dishes, enough to have a chance at overwhelming a tumor when they're pumped back into patients.

About half significantly improve after this so-called cell-transfer therapy.

But few melanoma patients make enough cancer-fighting T-cells naturally for scientists to spot in their bloodstream, and T-cells that attack other cancers are virtually impossible to find. So Rosenberg and colleagues set out to create those tumor fighters from scratch.

The scientists took normal lymphocytes, or ones that don't recognize cancer, out of 17 patients with advanced melanoma who had exhausted their treatment options.

They infected those cells with a virus carrying genes that create T-cell receptors, essentially homing devices for, in this case, melanoma. (Different genes create receptors for other cancers.)

"We can take a normal cell from you or me or any patient and ... convert that cell into a cell that recognizes the cancer," Rosenberg explained.

In 15 of the 17 patients who tried it, the newly armed cells took root and grew at low levels for a few months.

But only two saw their tumors gradually fade away: Origer and a 30-year-old whose T-cell levels remained superhigh for over a year.

Why did the genetically altered cells flourish in only two people?

"That's the critical question," said Dr. Patrick Hwu, melanoma chairman at the University of Texas M.D. Anderson Cancer Center, who once worked with the National Cancer Institute's team.

Picking the right lymphocyte to genetically alter isn't easy - there are many different kinds - or perhaps more precise T-cell receptors were needed, Hwu suggested.

"(But) these are all solvable issues," he stressed, calling the study "one of the first documented, effective