The man nominated to run the government's main controls over toxic substances yesterday said he plans to change the ways that the Environmental Protection Agency evaluates health risks.

Although Dr. John A. Todhunter didn't say so, the changes will tend to find substances less dangerous than current procedures do.

The question of risk is thorny and central to all federal regulations, which in effect choose a level of risk and label it acceptable or unacceptable.

Todhunter, 32, told the Senate Environment and Public Works Committee that as EPA's assistant administrator for pesticides and toxic substances, he will make those choices using computer models that are different from the ones that guide those choices now.

Asked by Sen. Robert T. Stafford (R-Vt.), chairman of the committee, about his views on using data from tests on laboratory animals, Todhunter said he would continue to rely on animal tests.

"What may change is that the agency has been in the habit of using only one of several available curve-generating computer models" to evaluate the tests, he said. "I will add the use of other models.

"Some of them agree with each other , and where they do, you have greater confidence," he added.

The action could affect everything EPA does in the toxics regulation field. Laboratory animals are commonly exposed to the chemicals in tests in an effort to determine whether the substances are dangerous to people. But they are tested at such high doses and have such short normal life spans that relating the tests to human risk is a complex business.

Scientists use computer models to process the raw test information in order to be able to say, in effect, that if so many rats develop cancer when dosed with a substance in a certain manner, at a certain level and over a certain period of time, then the risk to humans is "X" at such-and-such a dose.

Todhunter's science adviser, Dr. Richard N. Hill, said the results would be different for each model, especially at the low levels of exposure that are usually the case for people.

"It opens the option of finding lower risk than we could have found before," he said. "Also it could find that the previous model was on target, or that the risk was even somewhat worse."

The models now most widely used, he said, tend to assume that there is no threshold level of exposure, that is, that even "one hit" of a toxic substance may cause damage.

He stressed that risk is expressed in terms of the probability of developing a health problem, like cancer, when a person is exposed to a certain amount of the substance.

Stafford asked Todhunter to provide the committee with all documents relating to EPA's decision against emergency regulation of two alleged cancer-causing substances, formaldehyde and a chemical called DEHP.