EPA ASSESSMENT OF RISKS FROM RADON IN HOMES June 2003 UPDATED RADON RISK INCREASED

EXECUTIVE SUMMARY Radon-222 is a noble gas produced by radioactive decay of radium-226, which is widely distributed in soils and rocks. Radon-222 decays into a series of short-lived radioisotopes. These decay products are often referred to as radon progeny or daughters. Because it is chemically inert, most inhaled radon is rapidly exhaled, but the inhaled decay products readily deposit in the lung, where they irradiate sensitive cells in the airways, thereby enhancing the risk of lung cancer. In 1999, the National Research Council of the National Academy of Sciences published the BEIR VI report, Health Effects of Exposure to Radon (NAS 1999), which assessed the risks to the U.S. population from radon in homes. The authors of this study, sponsored by the EPA, had the benefit of extensive new information not available to the authors of the Academy’s previous BEIR IV report on the risks from radon and other alpha emitters (NAS 1988). On the basis of epidemiologic evidence from miners and an understanding of the biologic effects of alpha radiation, the committee concluded that residential exposure to radon is expected to be a cause of lung cancer in the population. Based on a statistical analysis of epidemiologic data on 11 cohorts of occupationally exposed underground miners, the committee developed two preferred risk models from which they projected, respectively, 15,400 or 21,800 excess lung cancer cases in the U.S. each year. An analysis of the uncertainties suggested a range of 3,000 to 33,000 cases per year. The committee concluded that “this indicates a public health problem and makes indoor radon the second leading cause of lung cancer after cigarette smoking.” Both of the preferred BEIR VI models are framed in terms of excess relative risk (ERR), which represents the fractional increase in lung cancer risk due to a specified exposure.1 To estimate the risk at any given age from a past exposure, one multiplies the ERR times the baseline lung cancer rate for an individual of that age (and, if appropriate, sex or smoking category). The lifetime risk from an arbitrary exposure can be calculated using a specified risk model in conjunction with life-table methods that incorporate competing causes of death. In both of these BEIR VI models the ERR falls off with time-since-exposure and with age at risk; nevertheless, because of the increasing baseline rate of lung cancer with age, the calculated risk from a given exposure often increases with increasing age. An important finding in BEIR VI, based on updated and expanded miner data, is 1 Exposures are measured in units of working level months (WLM), a measure of potential alpha particle energy that will be released by short-lived radon decay products per liter of air. 1 that risk from a given exposure tends to increase when that exposure is more spread out over time. For the relatively low exposure rates or long time durations of most concern for EPA, the risk per unit (WLM) exposure is maximal and increases linearly with radon exposure. Another new finding is that the estimated ERR is about twice as high for never smokers (NS) as for ever smokers (ES). Estimates indicate that radon exposure accounts for about 1 in 8 ES lung cancer deaths and 1 in 4 NS lung cancer deaths. However, since ES have a much higher baseline lung cancer rate than NS, the risk of a radon-induced lung cancer, on an absolute scale, is still much higher than for NS. Although there is a growing body of data from epidemiological (case-control) studies showing a correlation between lung cancer and radon exposures in homes, these results do not conclusively demonstrate an excess risk in homes with elevated radon and are inadequate as a basis for quantitative risk estimation. Thus, estimates of risk for indoor exposures must still be extrapolated using models derived from the miner data. There are a number of important differences between mine and indoor exposures that must be considered in making this extrapolation. First, due to physical and physiological factors, the alpha particle dose to target cells in the lung per WLM could be higher or lower in the case of residential exposures than for mine exposures. Since the risk is presumed to be proportional to dose, a model derived from the miner data might need to be adjusted to account for these differences. The BEIR VI risk estimates were based on the premise that the effects of these differences approximately counterbalanced each other in such a way that no adjustment was warranted. Doubts about this premise were expressed by Cavallo (2000). Cavallo correctly noted inconsistencies in portions of BEIR VI relating to how doses from exposures in mines and homes compare, and suggested that as a result the BEIR VI report may have overstated risks from residential exposures. More recently, James et al. (2003) submitted a report which carefully reexamined issues raised by Cavallo. James et al. reaffirmed that the effects on doses of differences between homes and mines do approximately counterbalance each other so that no adjustment would be needed for in-home risk calculations. It follows that the inconsistencies in BEIR VI noted by Cavallo did not lead to an overestimate of the risks from radon. Second, other agents in the atmosphere of underground mines, such as arsenic, silica, and diesel fumes, could modify the lung cancer risk associated with exposure to radon progeny. BEIR VI cited evidence that the latter two types of exposures were probably not strong modifiers of risk but that arsenic might be a source of positive bias in the risk estimates. Third, the exposure rates in homes are generally lower than the lowest levels for which we have clear evidence of excess risk in mines. Consequently, assessment of indoor radon risks requires an extrapolation to lower exposure rates. Although the 2 miner data and radiobiological data are both suggestive of a constant risk per unit exposure as one extrapolates downward from the lowest miner exposures, this assumption has been questioned. An ecological study has indicated that lung cancer rates are negatively correlated with average radon concentrations across U.S. counties (Cohen 1995), suggesting that the risks from very low levels of radon have been overestimated, or that such exposure levels might even protective against lung cancer. Biologically based models have also been proposed that could project substantially reduced carcinogenicity at low doses (for example, Moolgavkar and Luebeck 1990, Elkind 1994). Numerous critics, including the BEIR VI committee, have discounted the ecological study results because of methodological limitations, and the biologically based models remain highly speculative. The BEIR VI committee adopted the linear nothreshold assumption based on our current understanding of the mechanisms of radoninduced lung cancer, but recognized that this understanding is incomplete and that therefore the evidence for this assumption is not conclusive. In this document EPA updates its assessment of the health risks from indoor radon, based primarily on the BEIR VI report, with some technical adjustments and extensions. First, EPA constructs a single model that yields numerical results midway between what would be obtained using the two BEIR VI preferred models. Second, noting that the BEIR VI definition of excess risk effectively omits premature deaths caused by radon in people who would otherwise have eventually died of lung cancer, EPA modifies the BEIR VI calculations so as to include all radon-induced lung cancer deaths. Third, whereas the BEIR VI committee assumed that a fixed percentage of adult males or females were ES, EPA uses age-specific smoking prevalence data. Fourth, whereas BEIR VI estimated the fractional increase in lung cancers due to radon, EPA also provides numerical estimates of the risk per unit exposure [lung cancer deaths per working level month (WLM)] and the number of years of life lost per cancer death. Based on its analysis, EPA estimates that out of a total of 157,400 lung cancer deaths nationally in 1995, 21,100 (13.4%) were radon related. Although it is not feasible to totally eliminate radon from the air, it is estimated that about one-fourth of the radon-related lung cancers could be averted by reducing radon concentrations in homes that exceed EPA’s recommended 4 picocurie per liter (pCi/L) action level (NAS 1999). It is estimated that 86% of the radon-related lung cancer deaths were in ES, compared to 93% for all lung cancer deaths. The projected average years of life lost are higher for the radon-related cases (17 y) than for lung cancer deaths generally (12 y). Estimates of risk per unit exposure are as follows: 5.38×10-4/WLM (all); 9.68×10-4/WLM (ES); and 1.67×10-4/WLM (NS). Based on an assumed average equilibrium fraction of 40% between radon and its decay products and an indoor occupancy of 70%, the estimated risks from lifetime exposure at the 4 pCi/L action level are: 2.3% (all), 4.1% (ES), and 0.73% (NS). Although estimated absolute risks are much higher for ES than NS, estimated relative risks are higher for NS. It is estimated that among NS about one-quarter (26%) of lung cancers are due to radon compared to 3 about one-eighth (12%) among ES. It was more difficult to estimate risks for current smokers. Because of limitations of the data from the miner cohorts, the BEIR VI models did not specify excess relative risks for current smokers. Estimates of risk for current smokers (calculated by presuming that they start smoking at age 18 y and do not quit) are 1.5×10-3 per WLM, or over 6% for a lifetime exposure at 4 pCi/L. EPA also reexamines the issue of uncertainty in the risk estimates. Emphasizing the uncertainty in extrapolating risk estimates from observations on miners exposed to higher levels of radon than are ordinarily found in homes, BEIR VI derived its preferred uncertainty bounds (95% confidence limits 3,300 to 32,600) using a constant relative risk model obtained by a statistical fit to a restricted set of data on miners exposed to less than 50 WLM — levels that are comparable to lifetime residential exposures. The sampling errors are large with this limited data base; as a consequence the resulting confidence range may be overly broad. EPA adopts an alternative approach, deriving its estimates of uncertainty using the BEIR VI preferred models, with some explicit consideration of model uncertainties. However, like BEIR VI, EPA was unable to quantify all the relevant sources of uncertainty. These uncertainties are discussed qualitatively (or semi-quantitatively) and, for perspective, results of sensitivity analyses for some of these variables are included. From a Monte Carlo analysis of those uncertainties that could be quantified, EPA estimates a 90% subjective confidence interval of 2 to 12 ×10-4 lung cancer deaths per WLM, for the general population. The corresponding 90% interval for radon-induced lung cancer cases in 1995 is 8,000 to 45,000. Since the interval would be wider if additional sources of uncertainty had been accounted for in the analysis, it is plausible that the number of radon-induced deaths is smaller than 8,000 (but unlikely that it would be as small as 3,300). However, given the predominant role smoking is known to play in the causation of lung cancer, it is unlikely that radon accounts for as many as 45,000 deaths or 12 ×10-4 lung cancer deaths per WLM. Risk estimates for exposures to specific subgroups, especially children, NS and former smokers, have a higher degree of uncertainty than estimates for the general population. The effects of radon and cigarette smoking are synergistic, so that smokers are at higher risk from radon. Consequently, if projected reductions in U.S. smoking rates hold up, some decrease in radon-induced lung cancers is expected, concomitant with decreases in lung cancer, generally; nevertheless, it is anticipated that indoor radon will remain an important public health problem, contributing to thousands of lung cancer deaths annually. 4