Animal tests of substances for toxic properties have always had their critics. Even the most sincere may not realize that before the testing of new substances became routine, thousands of people died after taking remedies that turned out to be poisonous.

We are now less concerned with substances that might kill people and more concerned with those that have long-term cumulative effects. Of course, the long-term cumulative effect that is of greatest concern is cancer, which usually appears after many years -- even decades -- of exposure to, for example, cigarettes. During this time, exposure to many other substances occurs, some of which might augment or inhibit the effects of the first substance.

Not everyone who is exposed to carcinogenic agents develops cancer, but the chance of doing so increases with the number of exposures. More than a million Americans will develop cancer, and 500,000 will die of cancer-related illnesses; the number increases each year despite advances in treatment.

Early in this century, cancer was a relatively rare disease. Little notice was taken of the observations of a few physicians that some malignant tumors were associated with exposure to particular kinds of chemicals -- for example, bladder cancer among workers in dye factories and rubber manufacturing plants and skin cancer among those who worked in shale oil and coal tar processing factories.

As the population aged, cancer became more common and, in the early 1950s, the most rapidly increasing malignancy -- lung cancer -- was definitively linked to smoking. By this time, many scientists had become interested in studying the development of several types of cancer in animals by giving them particular chemicals. The larger the dose, the more animals developed tumors and the sooner the tumors appeared.

Because cancer-causing chemicals often caused the same type of cancer in experimental animals as they caused in humans, it seemed that human cancer was probably a consequence of exposure to carcinogens. This was not an entirely new idea, but when it took hold it led to enactment of the Delaney Amendment to the Food, Drug and Cosmetic Act, which forbade the use as a food additive of any substance shown by appropriate tests to cause cancer in people or animals.

This required that with some exceptions -- substances "generally recognized as safe," labeled GRAS -- those that might become food additives either intentionally or accidentally had to be tested for their carcinogenic potential. The only way of doing this was through animal tests. In these tests, small groups of rats or mice were given the substance for most of their lifetime to see if they developed tumors.

Because the tests could not use thousands of animals and because the animals lived only two years, they had to be treated with larger amounts of the substances than people would be exposed to at one time in order to see an effect. This is the reason that saccharin was tested by feeding apparently

ridiculous amounts to rats. The high dose partly compensates for a possible 50 or more years of exposure for the human population.

If the number of treated animals that develop tumors is not statistically significant, the test substance is accepted as not carcinogenic.

A finding of a particular tumor in fewer than four animals of any group would normally indicate that the substance was unlikely to pose a carcinogenic risk to humans. However, tumor frequencies higher than this do suggest a carcinogenic risk, which would be more certain the more animal groups were positive in this way.

Even in the case of a definite positive finding in such a test, it cannot be concluded that the substance causes cancer in humans, only that it is more likely to contribute to an increased risk of cancer than another substance that is not carcinogenic.

In spite of the high cost of a well-conducted test for cancer risk in animals and the criticisms regarding the outcome, it remains the only reliable method of screening synthetic or natural substances that might contribute to humans' risk of cancer. The risk, however, is substantial: 15 to 25 percent of the population will develop some form of cancer during their lifetimes, and we do not know the cause of most of it.

Numerous short-term tests that are more economical have been developed during the past 20 years to investigate the cancer-causing potential of substances, especially tests based on measurement of mutations in bacteria, yeast or mammalian cells in culture. However, they are not reliable individually or as a group because many carcinogens, including some known to cause cancer in humans, produce no response in the tests.

No one knows the mechanism by which any chemical produces cancer, but it seems that many steps are involved, most of which cannot take place in a short-term test. Animal tests, with their complexity and inherent problems, remain the only reliable means of demonstrating the safety of a substance for human use and, as a corollary, the only means of uncovering the cancer-causing properties of a substance.

This is likely to be the case for some time, and a careful and unbiased assessment of the results of such tests can provide reasonable assurance of safety without relinquishing improvements in the quality of life.