(FORTUNE Magazine) – More than a century ago doctors noticed a curious coincidence: Cancer patients' tumors sometimes shriveled after the patients contracted bacterial infections. That inspired the idea of cancer vaccines, which would stir the body's immune system to attack tumor cells as if they were germs, replicating what seemed to have happened in the lucky patients. Unlike the familiar preventive shots we get to ward off the measles, flu, and other infectious ills, cancer vaccines would be therapeutic: Doctors would administer them only after disease struck, to focus the immune system's xenophobia on the growths within.

The good news about this idea is that it has seemed to work since the 19th century--bulging neck tumors withered in one of the first patients on whom it was tried. The bad news is that it has never worked consistently. Like a diabolical lottery, experiments with cancer vaccines typically succeed just enough to lure ambitious doctors and desperate patients to try again and again.

Laura Abrams, a 48-year-old San Diego resident, is one of the winners whose story has helped enthusiasts keep the faith. In 1985 she was diagnosed with metastatic melanoma--skin cancer that had spread to her lungs and other tissues. "I was getting ready to not be here," she says. "They give you six months."

Then she heard of an experimental melanoma vaccine concocted by a Los Angeles doctor named Malcolm Mitchell. Just as polio vaccines are made from the virus that causes polio, Mitchell's compound contained a molecular likeness of its target: human skin-cancer cells. He had produced the vaccine by mincing the cells in a stainless-steel blender.

Seven weeks after Abrams' first injection, her lung tumor disappeared. Then dozens of lumps under her skin melted away. She has fought off two recurrences with the help of other therapies. Now she's the longest-known survivor from the group of gravely ill patients who first tried the vaccine. "I'm a melanoma poster child," she says, reeling off some of the articles and TV shows that have featured her case. "But I don't do a lot of interviews anymore. It makes people think, 'Oh, yeah, it's just a one-time thing.'"

Finally, though, cancer vaccines have begun to shed their aura of tabloid miracle, thanks to a gradual accretion of solid data. In a telling sign of the change, big drug companies have recently made major bets on the vaccines. Last year part of France's Rhone-Poulenc Group, Pasteur Merieux Connaught, launched a ten-year effort budgeted at up to $350 million to develop cancer vaccines. In October, SmithKline Beecham agreed to invest up to $200 million on cancer vaccines and related drugs at Corixa, a Seattle biotech company. Schering-Plough, Bristol-Myers Squibb, Chiron, Genzyme, and other biotech companies are also investing.

It has been a remarkably quiet change. Cancer vaccine researchers, mindful of their field's history of premature hopes, rarely flaunt their promising data. After 30 years of work, Dr. Donald L. Morton recently reached a major milestone: The National Cancer Institute deemed preliminary human tests on his CancerVax vaccine so promising that it awarded $34 million for clinical trials that may lead to Food and Drug Administration approval. But Morton, medical director of the John Wayne Cancer Institute in Santa Monica, Calif., isn't popping corks: "We're at the end of the beginning," he says.

He and other doctors are acutely aware that the war on cancer is incremental, a tough hedgerow-to-hedgerow fight. Most new cancer drugs are tested initially on people with advanced disease. For them long-term survival usually means a year or two, and "cure" is a loaded word. But unlike the hype last spring about anticancer drugs that have shown promise in mouse studies, the rising hopes for cancer vaccines spring mainly from human data.

Since the mid-1980s more than 2,000 patients have received a variety of the vaccines in clinical trials. More than half a dozen of the medicines are in final "phase III" tests needed for FDA approval. Most of this first set aim at melanoma, which seems more susceptible than most cancers to immune attack. Two companies, Intracel of Rockville, Md., and Ribi ImmunoChem Research of Hamilton, Mont., say they plan to seek approval for vaccines within months.

Oncologists don't expect miracles from these first vaccines, developed more than a decade ago. Clinical trials with Ribi's Melacine showed it shrank tumors in only about 20% of patients and was comparable with conventional drugs in terms of overall survival. But reports in the past two years suggest greater efficacy is coming:

--When Dr. Mitchell injected 18 patients who had failed on Melacine with an immune stimulant called alpha interferon, tumors regressed in eight, suggesting that combining the drugs could triple Melacine's response rate to 60%. Schering-Plough, which sells alpha interferon, licensed rights to Melacine in March and is funding clinical tests of the duo.

--Researchers at Thomas Jefferson University in Philadelphia last year reported that about 60% of metastatic-melanoma patients given a cancer vaccine survived at least five years, vs. a third or fewer in studies on other therapies. AVAX Technologies in Kansas City recently began a phase III trial of the vaccine, which is produced by extracting a patient's tumor cells, chemically treating them so they'll spark immune attacks, then reinjecting them.

--A 1996 report showed that a group of advanced-melanoma patients who took Dr. Morton's CancerVax had a five-year survival rate of 25%, four times that of similar patients on other therapies.

These heartening success rates may decline in the larger, more demanding trials needed for FDA approval. Cancer vaccines may have additional serious limitations. They don't seem to help much against bulky tumors--their best use in such cases may be mopping up tumor cells after surgery. Moreover, vaccines like AVAX's may never see wide use because they involve extracting and reinjecting patients' own tumor cells. Getting enough cells for such "autologous" medicines may be infeasible when tumors are caught early and small, and drugmakers are reluctant to back autologous vaccines, which they view more as a service than as a product.

But vaccines offer a plus that could offset such drawbacks: Their side effects tend to be mild, often no more than injection-site soreness. That's no surprise: Scientists have long known the immune system can single out diseased cells with little collateral damage--otherwise we'd have to regrow our noses after colds. If cancer vaccines win approval, that hateful principle, "no pain, no gain," could see a rare reversal.

Most of the drugs and radiation used to treat cancer bludgeon all fast-multiplying cells and suppress the immune system. Besides making patients vulnerable to deadly infections, this erodes the body's natural defenses against cancer. Scientists believe the immune system culls cells whose genetic machinery has gone haywire; errant cells that escape immune surveillance can become self-replicating monsters.

Unlike germs, cancer cells may differ only subtly from normal ones, appearing as what scientists call Self--jargon for cells that display on their surface just the right idiosyncratic molecules, the cellular equivalent of our own fingerprints, to our immune police. The immune system is strongly constrained from attacking such cells, a phenomenon called tolerance. The Grail in cancer vaccine research is overcoming tolerance enough to engender some well-placed Self destructiveness. "If you can break tolerance, you can cure cancer," says Drew Pardoll, a vaccine researcher at Johns Hopkins University.

For decades researchers tried overwhelming tolerance by injecting general immune stimulants like bacterial extracts. These throw the immune system into a state of high agitation--a strategy akin to loosing frantic hounds after an escaped convict without first making them sniff his clothes. The agitation might step up attacks on cells that appeared defective, but the injections didn't help immune cells solve the basic problem: distinguishing healthy Self from cancer.

Later researchers switched with high hopes to bioengineered immune boosters like interleukin-2. Again they were frustrated. Occasionally tumors regressed, but more commonly the drugs caused debilitating fever with little benefit.

The field almost died, but a few stalwarts kept at it. One was California's Dr. Morton, who found antibodies in the blood of melanoma patients that attacked their tumor cells in the test tube--a clue that the immune system could indeed spot the demented Self of cancer, even if it couldn't fully break tolerance. After decades of failure, he finally began realizing some success with CancerVax in the late 1980s. The vaccine contains three types of live melanoma cells that are rendered harmless with radiation; when injected, they sometimes spark a strong immune-system attack.

Laura Abrams' physician, Dr. Mitchell, now at Detroit's Karmanos Cancer Institute, was another diehard. He had long dreamed of using biotechnology's growing tool kit to dissect immune attacks on tumors at the molecular level, in order to learn how to amplify the reactions. Injecting his pulped cancer cells alone didn't incite the level of immune agitation he needed. That led him in 1985 to Edgar Ribi, a researcher who had developed one of the most potent immune stimulants known.

It was based on endotoxins, bacterial molecules that even in tiny doses can spark violent immune reactions leading to fatal shock--infected patients receiving antibiotics sometimes die as the bacteria in their bodies break up, releasing bursts of the toxins. Ribi had found a way to soften such reactions by chemically altering the compounds. A Swiss expatriate, he had spent most of his career at the Rocky Mountain Laboratory, a federal installation in Hamilton, Mont. In 1981 he formed Ribi ImmunoChem, and when Mitchell called, he gladly supplied some of his booster for the vaccine studies. Then serendipity and tragedy struck.

"The first patient I injected had tumor shrinkage," says Mitchell, who had expected at best a mild response useful for research. When Laura Abrams' and other patients' tumors also responded, the surprised scientist switched to his doctor hat: "I knew I had something of therapeutic value," he says. Soon after, he met Ribi at a scientific meeting in Hungary. As Mitchell talked, Ribi excitedly jotted down the oncologist's findings and ideas for more ambitious clinical tests. But returning from the meeting, Ribi died when he crashed his private plane into a mountain not far from home.

"I was one of the last people to see him alive," says Mitchell. "The plane didn't burn, and his notes about our meeting were recovered. Right after that, the company had me come up, and I explained what they meant." Chasing its founder's last dream, Ribi's company set off on the long trek to market with Mitchell's mix.

Seeking the quickest way to show efficacy, the Ribi brain trust made a risky decision: to test the vaccine first in patients with advanced Stage 4 melanoma. Clinical trials involving such end-stage patients don't last very long--a harsh truth, but a major plus for a strapped startup. The disadvantage, of course, is that it takes a real breakthrough to show much benefit in such advanced cases. The lackluster results Ribi finally reported in 1994 showed that Melacine was no breakthrough.

Still, the data do demonstrate that the vaccine offers advantages over conventional chemotherapy, contends CEO Robert E. Ivy. Melacine's mild side effects can make a patient's final months less miserable. Patients whose tumors respond to the vaccine appear to survive longer than those whose cancer responds to chemo. Ribi may have a blockbuster yet if the rate of such responses is shown to rise dramatically in the ongoing trial of its vaccine plus alpha interferon.

Even as Ribi struggled, discoveries about the immune system were opening the possibility of far-better vaccines. Importantly, scientists elucidated how key immune-system players called T cells singled out their targets. All our cells, it seems, present samples of themselves for T-cell inspection by dismantling inner proteins and holding out the pieces on MHC molecules--picture little arms jutting from the cell surface, waving ID cards. If a virus has snuck into the cell, the invader's telltale pieces, called antigens, get displayed too. That tips off "killer" T cells to glom onto the cell and destroy it.

This suggested a way to help the immune system target cancerous Self. First, find a cancer patient whose tumors have responded to a vaccine. Next, isolate T cells from the patient's blood. Then, in the test tube, observe whether any of the T cells attack molecules scissored out of the patient's tumors with gene-splicing techniques. If they do, you've found the long-sought antigens that mark cancerous cells.

The approach works: In 1991 a European team electrified cancer researchers by identifying a melanoma-cell protein that provoked T-cell attack. Other tumor antigens were soon found, opening a new chapter in the vaccine quest--if immune cells are like bloodhounds, the novel antigens can be seen as really smelly pieces of an escaped convict's clothing.

In a cramped, one-story brick building in Cambridge, Mass., bioengineers at Therion Biologics have stitched genes for such antigens into benign pox viruses, including the one Edward Jenner used for his historic smallpox vaccine. When injected into cancer patients, the viruses cause the cells they infect to display both viral and tumor antigens. The viral antigens rile T cells, which, it is hoped, will spill over into a bodywide intolerance for cells displaying the closely associated tumor antigens--so cancerous Self would be doomed. Teamed with Pasteur Merieux and the NCI, Therion has six such vaccines in preliminary clinical trials, including ones to fight skin, prostate, lung, and colon cancers. It's too early to judge efficacy, says CEO Dennis Panicali, but preliminary data show "we can break tolerance and induce a T-cell response to tumors."

The advent of such bioengineered vaccines is a major reason big drug companies are now interested. Such vaccines can be mass-produced like traditional drugs. They also pose no risk of contamination with infectious agents, as do vaccines made from, say, human cancer cells. Indeed, researchers working on the new compounds refer to vaccines like Dr. Mitchell's as "crude" vaccines. Exasperated by that, Mitchell counters, "Maybe I should call theirs 'ineffective vaccines.'"

That epithet has real bite--bioengineered vaccines based on single antigens haven't worked as hoped in early clinical tests. But researchers aren't discouraged about the new tack and recently have achieved exciting results with bioengineered vaccines designed to block "tumor escape mechanisms."

Cancer cells are notoriously changeable--not surprising, given that they're genetically running amok. Thus, they may easily evolve into variants that fool immune cells primed by vaccines to look for just one or two antigens. Such evasion would be harder if attacking T cells were keyed to many of their targets' antigens. Some of the new bioengineered cancer vaccines mount a host of tumor antigens in the immune equivalent of a post office's bulletin board of most-wanted posters.

Here's how: Before T cells attack, they're briefed on their targets by "antigen-presenting" cells. Some antigen presenters, called dendritic cells, lie in wait for enemies, which they engulf, dismantle, and display to T cells as sets of antigens on MHC molecules; others, called macrophages, move around the body doing the same thing. All antigen presenters have an added power: When showing antigens to T cells, they can also extend so-called B7 proteins to touch counterpart receptors on the T cells. This dual signal hits T cells' hot buttons as nothing else does, prompting mass attacks on cells exhibiting the antigens.

Now that researchers know about the hot buttons, they're racing to develop vaccines that press them. One way may be to inject tumor cells that have been genetically reengineered to secrete GM-CSF, a substance the immune system uses to help mobilize antigen presenters when going on the attack. Such altered tumor cells would effectively scream to the immune sentinels, "Come here--I'm a vicious killer." The altered cells would also provide the antigens that the excited immune authorities need to hunt down and destroy other tumor cells.

In October a team at Boston's Dana-Farber Cancer Institute reported that in a pilot study for melanoma patients, injecting such a vaccine caused widespread tumors to crumble under immune attack in 11 of 16 subjects. A dramatic series of microscope pictures from tumor biopsies show fat, happy tumor cells undergoing lethal onslaught by killer T cells, which resemble a swarm of angry hornets.

It's too early to assess efficacy, however. What's more, the Boston team's vaccine recipe is too complex for wide use, says senior author Glenn Dranoff--it involves bioengineering a patient's own cells before reinjecting them. Dranoff's team has developed a much faster, simpler way to make such vaccines, though, and Cell Genesys, in Foster City, Calif., recently announced plans to test it in the clinic.

A recent advance promises to mesh with the Boston team's approach. It involves a drug that may provoke immune attacks of unprecedented ferocity by prolonging the time that T cells remain on the assault. Scientists have long known that something quells T cells once their targets are destroyed. In the past few years they've discovered that this signal is probably conveyed by a protein called CTLA-4, which dots T cells like the emergency-stop buttons on a train. Researchers at the University of California at Berkeley recently showed that mouse tumors resistant to a vaccine like the one tested in Boston did respond when a drug that blocks CTLA-4 was added to the treatment.

James P. Allison, who led the Berkeley research, offers a simple analogy to explain this: The T-cell receptor that registers tumor antigens is like a car's ignition switch. After it's turned on, stimulating the T cell with B7 proteins is like stepping on the accelerator. But CTLA-4 can get in the way, says Allison: "It's like the hand brake." Unless it's deactivated, which is what the blocking drug does, a nascent immune attack may not build up much speed.

In one experiment, Allison's team gave mice afflicted with melanoma a vaccine in combination with the CTLA-4 blocker. Afterward, the rodents turned from black to splotchy white as both cancerous and normal skin-pigment cells died--a vivid sign that tolerance had been broken. "They began looking like Holstein cows," he says. After decades of trying to break tolerance, researchers may soon face the opposite problem: keeping the powerful immune reactions they incite from snowballing into autoimmune disease.

Perhaps the best sign of progress is the growing number of patients who have received cancer vaccines and continued to thrive for years. Canadian jockey Sandy Hawley, one of Dr. Mitchell's early patients, just retired at age 49 with a prodigious 6,449 wins, more than a decade after being diagnosed with metastatic melanoma. "The vaccine saved my life," he says. "It should be an option for everyone with skin cancer."

Bias of a lucky winner? Maybe not.