Radon: is a little good for you?

Radon: Is a little good for you?

Risks now associated with low-dose exposures to ionizing radiation hae been extrapolated from effects seen in people exposed to high doses -- generally atomic-bomb survivors, recipients of early X-rays, or workers in uranium mines. Because no one has established that there is a threshold to radiation effects -- a level below which no hazard exixts -- policymakers have conservatively assumed that even tiny exposures present some risk. However, controversial new radon studies in humans now challenge the no-threshold view -- and even go a step further. They hint, as a few animal studies have, that it's possible some radiation may actually be beneficial.

The studies, by Bernard Cohen at the University of Pittsburgh, compared U.S. data on average indoor-radon levels with average lung-cancer rates for the county in which each measurement was taken. According to the no-threshold theory, Cohen says, one would expect to find a trend toward higher lung-cancer rates for those counties with the highest indoor-radon averages. But to the contrary, he says, "we found there's a strong tendency for counties that have high radon levels to have low lung-cancer rates."

One study, representing data from 415 counties, was based on 39,000 measurements taken in the main living rooms (not basements, where readings are typically highest) of homes in which the residents had purchased their first radon test kit. Based on the radon average, a no-threshold estimate would have predicted female lung-cancer rates 25 percent higher than the national average. Instead, Cohen says, "the data show a 30 percent decrease." Comparisons for men and women in the 10 states for which there are data on 10 or more counties give similar "negative correlations in 80 percent of the cases. And in the states where there is a positive correlation," he adds, "it is very slight and not statistically significant." But this study was clearly nonrandom, since it involved only homes where the residents were worried enough to pay for radon measurements.

In a separate study, Cohen made similar comparisons for about 1,200 homes -- this time selected at random -- in 40 counties having the highest and lowest U.S. lung-cancer rates. Again, Cohen reports, in every case the radon level for low-lung-cancer counties was much lower than had been predicted, and the radon level in high-lung-cancer counties was much higher than predicted.

He reports similarly perplexing data from Scandinavia. For example, though Finland's average indoor-radon level is about 2.5 picocuries per liter (pCi/1) in air--about 2.5 times the world average--its female lung-cancer rate is oly about 70 percent of the average for industrialized countries, he says. Cohen also cites five state-sponsored studies completed within the past year -- in Florida, south Carolina, New Jersey and two in New York -- that "showed the same trends."

These data do not suggest that people exposed to high radon levels have a low cancer risk, Cohen says, because a large body of data compellingly links high-radon exposures to lung cancer in underground miners. Rather, he says, it calls into question the no-threshold theory -- because if there is not threshold, average county measurements should correlate directly with observed lung-cancer incidence. However, should further studies support the negative association found in these studies, Cohen says, scientists may soon be forced to ask the even more revolutionary question: Do small radiation exposures actually confer some sort of protection against lung cancer?

These data "certainly look counter to what you'd expect," says C. Richard Cothern, a radon-risk analyst and executive secretary of the Environmental Protection Agency's scientific advisory board committee on environmental health. But even if Cohen's interpretation is right, Cothern says, the Pittsburgh scientist can't prove it with these studies because "none of his data are truly random--they all have some kind of bias." Rather than prompting criticism of the study design, Cothern says, these biases should be recognized as inherent limitations in the available data.