Twice the NRC Was Pushed to Drop Its Radiation Model. So Far, It Hasn’t.

In February 2015, three petitions reached the U.S. Nuclear Regulatory Commission within about two weeks of one another. The first, from Dr. Carol S. Marcus, was submitted on February 9. A petition from health physicist Mark L. Miller followed on February 13, and one from Dr. Mohan Doss and colleagues on February 24. All three asked the agency to abandon the linear no-threshold model, or LNT. That model assumes cancer risk from ionizing radiation rises in direct proportion to dose, all the way down to zero.

The petitioners advanced the concept of radiation hormesis, the idea that low doses of ionizing radiation protect against the harmful effects of high ones. They asked the NRC to keep worker doses “at present levels, with allowance of up to 100 mSv (10 rem) effective dose per year if the doses are chronic” and to remove the principle of keeping exposures “as low as reasonably achievable,” or ALARA, entirely. The current occupational limit is 5 rem per year, so the allowance they proposed was twice that.

The request rested on a gap in the evidence. At low doses, cancer effects are weak and hard to detect, and epidemiology struggles to resolve the shape of the dose-response curve near zero. Some studies do report low-dose associations, but no study has established a threshold below which risk falls to zero. The NRC took six years to answer the petitions. In 2026, under an executive order directing it to reconsider LNT, it issued a proposed rule that would remove ALARA. Both times the model survived, and the reason lies in what the data can and cannot show.

A threshold that was set aside

For much of the first half of the twentieth century, it was commonly assumed that radiation-induced health effects occur only when the dose exceeds a certain threshold. Below that line, the body was presumed to cope.

The idea was discarded for stochastic effects as the mechanisms of cancer became better understood. In a stochastic effect, dose changes the probability of the outcome, not its severity. The no-threshold assumption grew out of that biology. At low doses it has not been confirmed by direct measurement, and regulators extrapolate from high-dose data toward zero.

Where the signal falls below the noise

The figures below mix absorbed dose, in mGy, with effective dose, in mSv and rem, so they are not interchangeable.

The United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR) defines low doses as those below 0.1 Gy of absorbed dose. In that range, as a 2024 article by Andrzej Wojcik and Friedo Zölzer puts it, biological effects “are very weak so they are easily influenced by random environmental factors making results difficult to reproduce.” Their piece appeared as a “Debate” contribution in Radiation and Environmental Biophysics and defends LNT on scientific and precautionary grounds, so it is one side’s argument.

Cancer is common and has many causes. A study looking for a small excess against that high baseline needs a very large population, and small low-dose studies lack the statistical power to do so. Physicist Nicole Martinez has noted that 100 mSv of effective dose is widely cited as the point where uncertainty becomes high, adding “but there are studies that show that there is attributable risk lower than that.” The 100 mSv figure marks where confidence thins out, not where the evidence ends.

Critics of LNT push in opposite directions. Some argue that bystander effects mean the linear model underestimates risk. Others argue that adaptation creates a threshold or even a protective effect. In its discussion of Laurier and colleagues (2023), who argue for LNT, the Wojcik and Zölzer article reports that any threshold “could not be greater than a few tens of mGy.”

Large occupational cohorts are still being followed. So far, results from the International Nuclear Workers Study (INWORKS) and the Million Person Study support a linear association between some types of cancer risk and cumulative dose, even at low dose levels.

That is why the model has lasted. No established non-zero threshold exists. That absence does not prove harm at doses comparable to natural background. It means regulators must choose an assumption under uncertainty, and the evidence that does exist leans linear.

Six years on the docket

Miller’s petition stated that animal studies have not shown increased cancer at doses below 100 mSv (10 rem), that many show thresholds, and that about 40% show hormesis. That was his own characterization of the literature, not a verified finding.

Several federal agencies disagreed. The National Cancer Institute, the National Institute for Occupational Safety and Health, and the Environmental Protection Agency all supported keeping LNT. In comments quoted by the Nuclear Information and Resource Service, the EPA described the threshold and protective-effect proposals as “poorly supported and highly speculative.”

On August 17, 2021, the NRC denied the petitions and closed the dockets. Its stated reason was that they fail to present an adequate basis supporting the request to discontinue use of the LNT model.

The 2026 proposal and the line that stayed

In May 2025, Executive Order 14300 ordered NRC reforms, including reconsideration of ALARA and LNT. The agency announced its proposed rule on July 1, 2026. It was published in the Federal Register on July 15 as “Reforming and Modernizing the NRC’s Radiation Protection Framework,” at 91 FR 43456.

The proposal would remove ALARA, a principle more than 50 years old. LNT stays. According to NPR, the proposal discusses the model’s limits but finds that “no consensus-supported, regulation-ready alternative model exists at this time.” An analysis by the law firm Greenberg Traurig describes the rule as retaining LNT “though narrows its applicability.”

According to the American Nuclear Society, the proposal quotes the NRC’s 2021 position that there is no compelling evidence for hormesis. Its section on effluents, however, states without citing a source that there is evidence for an adaptive cellular response. Neither hormesis nor adaptive response is established by that statement.

The agency also used the linear model to estimate the effect of one of its own changes. It calculates that raising the effluent limit from 10 to 25 mrem per year would change the excess fatal cancer risk over 70 years from 0.04 cases per 100 people to 0.09 cases per 100 people. That figure is an output of the NRC’s own model, not an observation.

NRC Chairman Ho K. Nieh said, “Our radiation dose limits remain unchanged.” The worker limit stays at 5 rem per year, though the proposal would allow up to twice that in a single year as long as the five-year average stays below the annual limit.

Public comments closed on August 31, 2026, with 2,307 received. As of mid-October 2026, the rule was still a proposal. The bottom of the dose-response curve, the stretch that started the argument in 2015, was still drawn the same way: a straight line extended downward from where the data are strongest.