When Edwina Schreiber showed up at the National Cancer Institute three years ago, her illness was already far beyond the reach of conventional therapy. Malignant melanoma, a cancer justly famous for its virulence, had spread to more than 30 sites in her body, from the inside of her arm to the roof of her mouth. Tumors were devouring her lungs and tonsils, and one had burst through her palate, hindering her ability to swallow. Instead of making futile gestures with radiation and chemotherapy, Dr. Steven Rosenberg, the NCI's chief of surgery, enrolled the young Atlanta woman in an audacious experiment. With her consent, he cultivated vast numbers of her own white blood cells in laboratory dishes and gave them back to her as medicine, hoping they would mount an all-out attack on her tumors. They did.

In his new book, "The Transformed Cell" (353 pages. Putnam. $24.95), Rosenberg describes how Schreiber looked when she returned this year for a follow-up exam. The cancer had "left a few marks," he writes, "but they were only marks--small bluish spots on the inside of her arm, on the roof of her mouth, and at a few other sites .... The tumors in her tonsils had disappeared. The tumors in her chest, gone. In her breast, gone. In her lungs, gone. The ulcerating tumor in her palate, gone. The thirty other subcutaneous tumors carefully counted and measured, gone."

Stories like Schreiber's are still rare, and the treatment she received is years from general use. But her experience reflects the emergence of a new strategy in the war against cancer. Disheartened by the archaic rituals of radiation and chemotherapy, researchers are turning increasingly to the immune system as the best hope against the disease. While some teams, including Rosenberg's, work to produce tumor-fighting immune cells in the laboratory, others are using vaccines to arouse the body's natural defenses. Aided by biotechnology and a growing knowledge of the immune system, both endeavors are now making advances that could lead ultimately to a new era in treatment. In the coming century, says Rosenberg, immunotherapy could "transform the practice of medicine."

The immune system's basic job is to keep foreign material out of the body. Its agents, the antibodies and white blood cells, course through our systems interacting with the distinctive protein molecules (antigens) found on the substances they encounter. A person's own cells display antigens identifying them as such. Foreign materials exhibit foreign antigens. By following a simple rule-tolerate substances marked "self," attack those marked "nonself"--the immune system does a good job of shielding the body from viruses and bacteria. But cancer poses a special problem, for the hostile forces are none other than the body's own mutinous cells. Even when a tumor displays unusual antigens, as many tumors do, the immune system tends to respond weakly, if at all. There are exceptions, however. Once in a great while, a patient kicks a malignancy like a cold. The challenge is to figure out why-and to make it happen more often.

Rosenberg's long quest began with just such an incident. In 1968, as a surgical resident at Boston's West Roxbury Veterans Hospital, he met a patient named James DeAngelo. Twelve years earlier DeAngelo had shown up at the same hospital with tumors devouring his stomach and liver and spreading through his lymph nodes. A surgeon had cut the largest mass from the man's stomach, then sent him home to die peacefully. DeAngelo got no further treatment, yet as Rosenberg confirmed in examining him, the cancer had utterly vanished. Dazzled by the thought of duplicating DeAngelo's feat, Rosenberg made a crude attempt at what's now called adoptive immunotherapy: he transfused a small amount of DeAngelo's blood into a man with advanced stomach cancer. The treatment accomplished nothing (the recipient died within two months). But medicine has advanced since then, and Rosenberg has polished his technique.

The first advance came in the late 1970s, when scientists started identifying cytokines, the hormones that regulate the immune system. In the early 1980s Rosenberg began treating advanced cancer patients with a cytokine called interleukin-2 (IL-2). The immune system uses IL-2 to make white cells proliferate. Rosenberg reasoned that if he administered the stuff artificially, the resulting onslaught of white cells might help combat the cancer. The treatment didn't work at first, nor did Rosenberg's early attempts to treat melanoma patients with cultured immune cells. But the record improved in late 1984, when he started using the cells and hormone together. And it improved further after he learned to cultivate tumor-infiltrating lymphocytes (TILs), immune cells directed specifically against tumors.

Rosenberg has now tried these various treatments on 1,200 patients, most of them suffering from advanced melanoma or kidney cancer. Only a quarter have responded, and just one in 10 has ended up cancer free. But that's an accomplishment in itself, and the record is still improving. TIL infusions have helped nearly half of the melanoma patients receiving them, and Rosenberg is now busy redesigning the cells to bolster their effects. In an experiment launched two years ago, he and his collaborators are outfitting TILs with the gene that produces a toxic enzyme called tumor necrosis factor (TNF). Eight melanoma patients have now received the supercharged TILs, and at least one who had failed treatment with regular TILs is responding.

As the modified-TIL experiment continues, Rosenberg is trying other innovations as well. The latest involves splicing the genes for IL-2 and TNF into patients' tumor cells, then reinfusing them as suicide bombers. "We're only starting," he says of the gene-transfer experiments. "We have the whole gene pool of the planet to work with. We just have to figure out how to use it."

Compared with cutting, burning and poisoning, the idea of treating cancer with tumor-fighting lymphocytes is elegance itself But the ideal immune therapy would require even less meddling. Instead of delivering potentially toxic substances through tubes and needles, doctors would simply trick the body into fighting its own battle. There is ample precedent for such an approach. Vaccines have rid the world of smallpox and helped control various other diseases simply by presenting antigens to the immune system in harmless but recognizable forms. If researchers could just package the right tumor antigens in the right way, the resulting concoctions might serve both to combat existing malignancies and to prevent future ones. Scientists are now testing a number of possible cancer vaccines, and they're not feeling discouraged. "People in this field are going into high gear," says Dr. Phillip Livingston of New York's Memorial Sloan-Kettering Cancer Center. "This is a time of enormous promise."

Dr. Donald Morton of the John Wayne Cancer Institute in Santa Monica, Calif., has spent 25 years trying to get melanoma patients to reject their own tumors. His technique is almost quaint in an age of designer lymphocytes, but his recent results have been eye-popping. In the early 1960s, Morton started extracting tumor cells from cancerous mice and reinjecting them along with BCG, the weakened bacteria used in tuberculosis vaccines. He found that the treatment often triggered a response not just against the bacteria but also against the cancer. He saw a similar effect when he injected BCG directly into tumors in people; the presence of a foreign substance somehow inspired the immune system to attack adjacent cancer cells. The challenge was to extend that phenomenon throughout the body.

Humans didn't respond as well as the mice had when Morton infused them with BCG and their own immobilized tumor cells. But he eventually devised a better strategy. Today, instead of simply giving people their own cells, he collects and cultivates cells that exhibit the known melanoma antigens. By disabling the cells (to keep them from causing new tumors) and administering them with BCG, Morton supplies the recipient's immune system with a range of possible targets. The effect can be dramatic. Once melanoma becomes disseminated in the body, patients on conventional therapy live an average of seven months. Only 6 percent survive five years. But in a newly published study of 136 patients treated with his vaccine, Morton reports an average survival time of nearly two years. Fully 26 percent of his patients-more than four times the usual proportion-have passed the five-year mark.

The survivors are naturally elated. Peggy Maddox, a teacher in Redondo Beach, Calif., had tumors in 16 lymph nodes eight years ago, when she became the first patient to receive Morton's vaccine. Today she's cancer-free, and feeling cocky enough to take out two-year magazine subscriptions. No one denies that vaccines like Morton's can work; scientists at Ribi ImmunoChem Research of Hamilton, Mont., have achieved similar results with their own melanoma blend. Yet many researchers are lukewarm. Clinical successes are hard to improve on if you don't know exactly how they came about, says Dr. Drew Pardoll of Johns Hopkins University. In order to design still better vaccine therapies in the future, he says, scientists need to determine how individual tumor antigens are affecting the immune system.

That effort is well underway. At Sloan-Kettering and other centers, researchers are now experimenting with "conjugate" melanoma vaccines. They consist of individual tumor antigens grafted onto large, foreign molecules that the immune system is sure to notice (chart). Taking a slightly different tack, Seattle researchers have synthesized the gene that produces a melanoma antigen and spliced it into the DNA of a harmless live virus (the one used in the smallpox vaccine). The virus provokes a vigorous immune response, and the researchers hope the tumor antigen will now be a target. They have yet to report results from human trials, but the trick has worked perfectly in animals.

Melanoma is an ideal candidate for immunotherapy; it's often impervious to conventional treatment, and its cells display a number of distinctive antigens. But it's by no means the only target. Scientists are testing vaccines against a range of other cancers as well. Dr. Georg Springer of Chicago's H.M. Bligh Cancer Biology Research Laboratory is particularly interested in breast carcinoma, the disease that claimed both his mother and his wife. Springer has vaccinated a handful of patients with carcinoma-associated antigens known as T and Tn, and the results have been tantalizing. Nationwide, only one in 10 patients with advanced (stage IV) breast cancer survives five years; the rate is 41 percent among less ravaged "stage III" patients. But of the 11 patients treated with Springer's vaccine at those advanced stages of illness, every one has passed the five-year mark in good health. No one knows whether that pattern will hold, but researchers at Sloan-Kettering and at the Biomira Co. in Alberta, Canada, are now testing similar vaccines in larger groups of patients to find out.

Most experts assume that different cancers will require different vaccines-that the war will be won in small steps. Not Hernan Acevedo, a researcher at Pittsburgh's Allegheny General Hospital. Acevedo suspects that cancer cells share a common denominator, an antigen that not only sets them apart from healthy cells but helps them make their mischief His candidate is a molecule called hCG, or human chorionic gonadotropin. Adult cells don't normally produce the hormone, but it helps shield embryonic cells from immune attack during pregnancy. Acevedo suspects it gives tumor cells the same kind of protection.

Whatever its role, the molecule does show up in suspicious places. Last spring Acevedo reported isolating pieces of hCG in 74 different cancer-cell lines. He has shown previously that metastatic cells-the ones that break free of established tumors to take root elsewhere-are especially rich in hCG. He has also used a conjugate vaccine containing a small fragment of hCG to shrink solid tumors and prevent metastasis in animals. Researchers at Ohio State University are now testing an anti-hCG vaccine in people, both as a contraceptive and as therapy for various cancers. No one yet knows whether the cancer patients will benefit, but Acevedo has high hopes. If his hypothesis holds up, he says, an anti-hCG vaccine might someday serve not only to treat cancer but to prevent it. Few others expect the quest to be wrapped up so neatly. But mainstream scientists now discuss such ideas without blushing. "A decade ago the perception that little could be done with immunotherapy was realistic," says Dr. Philip Greenberg of Seattle's Fred Hutchinson Cancer Research Center. "Today it's not."

1 One approach is to identify a tumor antigen, a molecule found on cancerous cells but not on healthy ones.

2 The tumor antigen is attached to a molecule that reliably provokes an immune response.

3 The resulting conjugate vaccine mobilizes the to immune system against the antigen and, with luck, against the tumor cells that exhibit it.

1 Doctors remove a small piece of tumor and extract immune cells called tumor-infiltrating lymphocytes (TILs).

2 TILs are then outfitted with foreign genes that make them more toxic to cancer cells.

3 The genetically altered TILs are reinjected along with interleuken-2, a hormone that helps them multiply.