WASHINGTON (AP) _ University of Texas scientists are destroying cancerous tumors in mice by engineering blood clots that starve the tumors to death, an advance that could be tested in people within two years.

The therapy, much like killing a plant by cutting its roots, caused rapid cancer-cell death within 24 hours, Dr. Philip Thorpe of UT's Southwestern Medical Center reports Friday in the journal Science.

Two weeks later, tumors had disappeared in 38 percent of the mice and had shrunk by more than half in another 24 percent.

Much work is needed to prove the treatment could work in people. But it could one day offer doctors a less-toxic alternative to chemotherapy for breast, lung, ovarian and other cancers.

``It would be wonderful,'' said Dr. James Pluda, a National Cancer Institute senior drug investigator. ``What this paper demonstrates is proof of the concept that ... this kind of therapy can be effective.''

Said Harvard University professor Dr. Judah Folkman, whose research into blood vessels that feed tumors formed a foundation for the discovery: ``This is very promising and very elegant work.''

Solid tumors, which represent most major cancers, depend on blood for oxygen and nutrients. Blood vessels grow rampantly through the cancer mass, often making surgery difficult because of heavy bleeding. The vessels eventually snake into other organs and spread the malignancy.

Thorpe theorized that by clogging vessels deep inside a tumor would make it die from the inside out. The question was how to avoid life-threatening blood clots in arteries throughout the body.

A human protein called tissue factor, or TF, is vital in helping people's blood clot. Coagulation won't start until TF latches onto cells, so Thorpe removed the molecule that allows it to connect.

Then Thorpe attached an antibody that recognizes a substance found only inside the tumor's blood vessels. And once that substance hooks TF to these vessels, the TF starts creating blood clots inside the tumor.

Clogged vessels appeared throughout mice tumors in 30 minutes and caused rapid cancer-cell death within 24 hours. Two weeks later, tumors large enough to be the equivalent of 2-5 pounds in a person had died in 38 percent of the mice.

Key to making the process work in people is finding the right homing device to direct TF to a tumor's blood vessels. Thorpe already has engineered drugs that would target one substance, called vascular endothelial cell growth factor.

A biotechnology firm, Techniclone Corp. of Tustine, Calif., is licensing the therapy and plans to begin testing it in people within two years.

``We give the tumor a stroke,'' Thorpe said. ``We can envision making a single drug for treating all types of solid tumors, whereas previously we had to tailor to each disease.''

Before Thorpe's work, scientists were trying to stop the runaway growth of new blood vessels that allows cancer to spread, a field pioneered by Folkman called ``anti-angiogenesis'' that recently yielded potential drugs to fight metastasis.

Thorpe's approach severs the cancer's original blood vessels, using the body's own clotting mechanisms instead of a chemical.

The two methods are complementary, the cancer institute's Pluda said, and it is unclear which would prove best for certain tumors.

But unlike chemotherapy, the risk of serious side effects from Thorpe's method is low, as is the chance of tumors mutating to resist the treatment, Pluda said.

The therapy is not a cure-all. Not all the mice responded. And those whose tumors shrank, even by huge magnitudes, eventually relapsed because cells on the cancer's outer edge got sufficient nourishment from neighboring blood vessels to cling to life.

Thus, if the therapy works in people, it likely would be used to shrink a tumor to make it easier to remove surgically or to require a lower dose of chemotherapy, Harvard's Folkman said.

``These are speculations,'' he cautioned, ``but that's why I think it is an exciting advance.''