The finding by the National Academy of Sciences panel is viewed as critical because it is likely to significantly influence what radiation levels government agencies will allow at abandoned nuclear power plants, nuclear weapons production facilities and elsewhere.

The nuclear industry,, as well as some independent scientists, have argued that there is a threshold of very low level radiation where exposure is not harmful, or possibly even beneficial. They said current risk modeling may exaggerate the health impact.

The panel, after five years of study, rejected that claim.

"The scientific research base shows that there is no threshold of exposure below which low levels of ionized radiation can be demonstrated to be harmless or beneficial," said Richard R. Monson, the panel chairman and a professor of epidemiology at Harvard's School of Public Health.

The committee gave support to the so-called "linear, no threshold" model that is currently the generally acceptable approach to radiation risk assessment. This approach assumes that the health risks from radiation exposure declines as the dose levels decline, but that each unit of radiation — no matter how small — still is assumed to cause cancer.

The panel, formally known as the Committee on Biological Effects of Ionizing Radiaton, or BEIR, generally supported previous cancer risk estimates — the last one by an earlier BEIR group in 1990.

Contrary to assertions that risks from exposure from low-level radiation may have been overstated, the panel said "the availability of new and more extensive data have strengthened confidence in these (earlier) estimates."

The committee examined doses of radiation of up to 100 millisievert, a measurement of accumulated radiation to an individual over a year. By comparison, a single chest X-ray accounts for 0.1 millisievert and average background radiation 3 millisievert.

The committee estmated that 1 out of 100 people would likely develop solid cancer or leukemia from an exposure of 100 millisievert of radiation over a lifetime.