The search for drugs to fight cancer has been a story of hit-and-miss experimentation with highly toxic chemicals since at least 1865, when doctors attempted to combat leukemia by giving patients doses of arsenic. From then to now, the problem has been the same: finding an effective drug, which does not have horrendous side effects, that kills only cancer cells, not healthy ones.

At the center of the problem is the simple fact that, after decades of research, cancer essentially remains a mystery. Scientists know very little about the differences between a cancerous cell and a normal one. They know that a cancer cell keeps growing and dividing while a normal cell does not, but they do not know why or how a normal cell becomes cancerous.

How, then, can they develop drugs to fight a disease that is so little understood? "We've been working in the dark," acknowledges Gordon Zubrod, one of the pioneers of the National Cancer Institute's drug program. "It really is a black box."

Scientists in universities and research laboratories across the country recently have begun to learn how certain parts of a gene within a cell may transmit chemicals that may change a normal cell into a cancer cell. The hope of most researchers is that their work will lead to an understanding of cancer cell biology and genetics -- an understanding that might eventually be used to control and contain cancer.

Up to this point, such scientific research has played a very small role in the nation's anticancer drug development program.

The first successful step in the use of drugs to fight cancer came during World War II, when a team of researchers examined chemical warfare compounds. They found that mustard gas killed cancer cells. From that work, came the first modern anticancer drug, Nitrogen Mustard, which was put on the market in 1948.

In the 1950s and early 1960s, researchers developed other anticancer drugs, many of them derived from poisonous compounds. While none of the common cancers could be cured by these drugs, doctors were excited that at least in a few cancers -- especially in childhood leukemias -- the new drugs held out some hope where in the past there had been none.

But from the start there was a serious problem. Although their main function was to kill cancer cells, these drugs also killed normal cells. They were so potent that they often caused severe side effects, such as nausea and vomiting, organ damage and, in some cases, death. That problem remains largely unsolved today.

Based on the early drug discoveries, Congress appropriated $5 million in 1955 for the National Cancer Institute to begin a drug development program. In 1971, when the federal War on Cancer was declared, the drug program grew enormously. Clinical drug studies were conducted on a large scale across the country. The drug development budget grew accordingly, to the point where today the yearly expense exceeds $50 million.

Each year, the National Cancer Institute searches for chemicals from throughout the world to be used as possible anticancer drugs. The chemicals generally come from industry, including such firms as Clairol, Dow Chemical, the 3M Company, Gulf Oil, Uniroyal, U.S. Naval Weapons, Procter and Gamble, Eastman Kodak. Other drugs are derived from plants and animals, including poison ivy, foxglove, monarch butterfly wings, wasps, extract from the bark of an Ethiopian tree and guinea pig serum.

"We would literally pull leaves out of the jungle," recalls Stephen Carter, a former NCI division director. "Our hypothesis was that we never knew where our next anticancer drug was coming from."

In all, the National Cancer Institute examines about 15,000 chemical compounds each year. About 50 of those come from scientists who specifically designed compounds as possible anticancer drugs.

All of the compounds are tested against mice with leukemia to see if they show signs of killing cancer cells.

Many NCI officials consider these tests to be unreliable and they are currently designing new test systems. "It's not scientific, but it's all we've had to work with," says Jack Driscoll, an NCI chemist.

Many cancer officials also say that the vast majority of chemicals that show activity against leukemia in mice show little activity against human cancers. Vincent DeVita, the head of the National Cancer Institute, said in a recent interview that NCI officials used to joke that Smith Brothers cough drops would show promise against cancer in the earlier mouse system.

Actually, about 100 of the 15,000 compounds tested in the mouse system each year show some level of anticancer activity. Of those, the NCI usually selects five to 10 for possible tests in humans.

Before the chemicals are given to people, they are injected in mice and dogs to determine how large a dose would be lethal. The starting dose for humans is determined by taking one-tenth of the dose that kills only 10 percent of the mice. The mice and dogs are also studied to determine toxic side effects.

Once dose levels are determined, the drugs are tried on humans -- beginning with what is known as Phase I. Doctors in such experiments continually raise the doses to see how much of the drug humans can tolerate and to observe what adverse side effects they suffer. It is out of these experiments that dosage is established for the next round of tests.

The primary purpose of a Phase I study is to determine toxicities and dose levels, not to treat the patient. In the next two phases of tests, humans are given experimental drugs to see what effect, if any, the chemical has against cancer.

The Food and Drug Administration has responsibility for protecting the patient from dangerous drugs and for warning the patient of risks. But some at FDA say the agency has shied away from close monitoring of what they see as its larger -- and politically more powerful -- subject, NCI.

"Cancer chemotherapy is off-limits for FDA investigators . . . the rule book is thrown out when it comes to cancer chemotherapy," said Michael Hensley, a doctor and FDA investigator. "I don't know if it's written in as many words, but there's absolutely no question about it. Ask anyone around here. I have been told many times that NCI was off-limits."