A Review of the Radiological Assessments Corporation's Fernald Dose Reconstruction Report (1997)

Untoward health effects in members of the public residing in the vicinity of FMPC could arise through exposure to ionizing radiation, through the radiations emitted by radionuclides released from the facility, through the chemical toxicity of uranium released from the facility, or from other causes. The possible effects of the releases from the facility are considered in this report.

The primary result of the Fernald Dose Reconstruction Project is an estimate of the probability that the radiation dose received by members of the public during the operation of FMPC might result in cancer, particularly lung cancer. Before the dose-reconstruction project, most scientists believed that the radiation dose to members of the public at Fernald was smaller than that from other sources of radiation exposure (such as radon in homes) and that the radiation dose from inhalation of uranium released into the environment would be the dominant source of the public's radiation exposure. Results of the dose-reconstruction project indicate 2 major points: that the radiation doses resulting from the facility were small relative to background radiation (see tables 22 and 23 in the RAC report) and that radon from the K-65 silos was a more important source of radiation exposure than was inhalation of uranium. The uncertainty of the radon exposures is larger than the uncertainty of the uranium exposures simply because fewer release or environmental data were available on which to base the dose-reconstruction models or to corroborate their predictions.

Most estimates of risk associated with exposure to radon are based on exposure to radon progeny in terms of working-level months (WLM) or in SI units (Jm^-3s). The methods used by RAC are based on estimating the concentrations of radon progeny where people resided. It would have been straightforward to compute WLM at this point. However, RAC computed the absorbed dose to the bronchial epithelium. Then (see page S-6) they used an inverse conversion from dose back to WLM for comparing risk. This is very unconventional.

There is ample evidence that very high doses of radon cause an increase in lung cancer in humans. Numerous studies in the United States and in other countries have confirmed that

uranium miners have a measurable increase in lung cancer that is clearly related to radon exposure. For the purposes of protecting the public from the health effects of radiation exposure, many scientists and public-health officials have assumed that even small doses of radiation (from radon or other sources) cause cancer and that the number of cancers that radiation causes is linearly related to the radiation dose. That is called the linear no-threshold hypothesis. There are large uncertainties in the epidemiologic studies from which the hypothesis and the associated risk coefficients are derived. Obtaining direct evidence as to whether radon or other sources of radiation cause cancer at low doses is difficult for a number of reasons. First, the number of radiation-caused cancers that are predicted by the linear no-threshold hypothesis at low doses is small, compared with the natural incidence of cancer; when the difference is small, it is nearly impossible to detect it unless the exposed population is very large. Second, radiation-caused cancers are indistinguishable from non-radiation-related cancers. Third, most cancers occur long after the exposure to radiation; the determination of a possible cause-effect relationship necessitates, as in the Fernald experience, retrospective estimation of dose over long periods, which leads to large uncertainty in the estimated doses. The more uncertainty there is in the estimates of dose, the more difficult it is to identify potential increases in risk.

The harmful chemical effects of uranium might be larger than the harmful effects of the radiation dose. The RAC report's discussion of the chemical toxicity of uranium is confusing. Unlike cancer, the major end point for radiation effects, the end points for the chemical toxicity of uranium vary from study to study, and the threshold dose for toxic effects depends on the end point that is used. To some extent, the “controversy” about the threshold for the toxic effects of uranium can be explained by the use of different end points in different studies.

Many toxicants (such as alcohol) cause damage that can be detected and can be reversed. The kidney has excess functional capacity, so even permanent damage might not result in clinically important decreases in renal function. On the basis of studies of occupationally exposed uranium workers, it is unlikely that even the most highly exposed members of the public living near Fernald suffered important renal damage from their exposure to chronic or acute releases of uranium.

In defining various dose and risk-estimate parameters, the RAC report has leaned in the conservative direction, that is, toward overestimating dose and risk. Although the degree of overestimation varies among different factors, the cumulative effect might be substantial. On the dose-conversion and risk-estimation side, the following factors have led to an overestimate of risk (see also the additional specific comments in the appendix of this report):

• The report assumes that environmental radon produces risk that is larger than that associated with radon among miners, whereas a Research Council report (NRC 1991) indicates a factor of 0.7-0.8 on the basis of differences in breathing rates, fraction unattached to particles, and other conditions.

• RAC incorporated a radon age 0-19 risk factor of 2, whereas a Research Council BEIR IV report (NRC 1988) indicates that a factor of 1 appears appropriate.

• RAC multiplied the all-age risk coefficient (rather than the adult risk coefficient) by 2 for irradiation at ages 0-19.

• RAC incorporated a sex factor larger than 1 (numeric details not given, but only shown on a graph in volume II, page S-11), whereas there is little evidence of a gender difference for the organs that were under consideration.

• Of the several estimates of radon risk available from authoritative bodies, RAC chose the highest (and oldest) estimate available as a starting point.

• The radon risk estimates are unduly high for nonsmokers (whose lifetime radon risk is one-fourth to one-third that for smokers).

• The report overemphasizes the importance of radon-dose estimates. In practice, lung-cancer risk is estimated directly from exposure to radon-decay products (progeny). Such risk estimates are typically about one-third to one-half those based on dose.

A radon dose-conversion factor for general-population exposure relative to miners' exposure (called the K factor) has been given by the Research Council (NRC 1991) as about 0.8 for ages under 10 and 0.7 otherwise. RAC did not apply this factor but claims that “part

of this difference is already accounted for in the conversion factor from WLM to rads” (page S-6, line 1); but the relative conversion factor that it refers to (0.65/0.5 = 1.3) appears to be in the opposite direction to the K factor. The discussion of a lung-cancer risk coefficient (see pages S-5 and S-6 in volume II) is incomplete, although the final result appears reasonable. Notably lacking was any discussion of uncertainties in radon risk estimates associated with smoking and with other lung-cancer cofactors (such as silica, arsenic, and diesel exhaust) in the mining studies. As mentioned above, the RAC assessment of lung-cancer risk did not take into account the apparent interaction between radon exposure and smoking. The joint analyses of the miner data have resulted in a larger than additive relation between smoking and radon exposure with respect to lung cancer (Lubin and others 1995). By taking the average risk across both smokers and nonsmokers, RAC has thereby overestimated the lifetime risk for nonsmokers and underestimated it for smokers. In particular, the absolute risks associated with radon for smokers are about 3-4 times higher than those for nonsmokers. Lubin and others (1995) have found relative risks 3 times higher for nonsmokers than smokers, but when the relative risks are applied to the baseline rate among nonsmokers, which is one-tenth that of smokers, the absolute risk is one-fourth to one-third in nonsmokers. Ignoring a difference that large is questionable. It is suggested that RAC present separate estimates for at least the larger risk scenarios contingent on whether the person is a smoker or nonsmoker.

The RAC examined the estimates of lung-cancer risk from uranium-worker studies to compare them with the risk estimate from the highest doses received by atomic-bomb survivors. In comparing the major study of multiple uranium facilities by Dupree and others (1995) (page S-8 in volume II) with other results, RAC did not use the overall risk estimate of Dupree and others, but rather concentrated on an ad hoc subgroup. For combined internal and external radiation dose, Dupree and others (1995) found a relative risk (RR) for total lung dose of 1.000 (95% CI, 0.859-1.094), whereas for internal irradiation alone the RR was 1.009 (CI, 0.983-1.036); neither indicates an increased risk. That was based on generally larger doses than the residents around Fernald received (and, in fact, the study of Dupree and others included the Fernald-worker population), so it should provide a measure of assurance to the residents regarding their exposure to uranium dust. Similarly, the other major multifacility uranium-worker study (Waxweiler and others 1983) could not detect an association between uranium exposure and lung-cancer risk at the doses the workers received.

Referring to the worker studies in volume II page S-8, the report says, “whenever an estimate of risk is possible, the result is not very different from what would be expected on the basis of the A-bomb survivor data.” But the authors have selected only subsets of data and findings that support an increased risk, rather than looking at the data as a whole, which suggest that the risk is very small.

The committee is concerned about the presentation of risks on page S-11 and in table S-3 and figure S-1 in volume II. In the Japanese atomic-bomb study (on which the risk estimates are based), much of the excess in lifetime radiation risks for females over that for males comes about by virtue of risks to the female breast and genital organs. There is little evidence for assuming higher female risks for the organs of interest in the present RAC study. In particular, the Research Council's BEIR IV report (NRC 1988) indicates that the lifetime risk of lung cancer associated with radon exposure is equal for males and females.

With respect to age adjustment for radon exposure (as on page S-12), a recent study indicated that there is no evident age differential in risk associated with radon exposure (ICRP 1994b). However, RAC chose to use an age-adjustment factor of 2 for exposures occurring before age 20 and 1 after age 20. It thereby increased the risk estimates appreciably for the various scenarios. The age bias played a major role in that most of the 9 hypothetical scenarios assumed that exposure was primarily in childhood. If RAC wished to maintain the same overall risk coefficient as that used by ICRP or BEIR IV, while maintaining an age factor, it should have made the juvenile factor somewhat larger than 1 but the adult factor somewhat less than 1. Instead, by virtue of the age adjustment, it has arbitrarily projected larger risks than have the studies that used the major data sets. In a similar vein, this section projects a higher risk for females than for males. In contrast, the Research Council's BEIR IV report (page 48) states that “the committee could identify no biologic rationale for considering that sex influences the development of radon-related carcinoma of the lung. . . . The relative risks appear to be similar in males and females when quantitative aspects of smoking habits are adjusted for.”

With respect to the uncertainties in risk estimation, although the discussion covers the principal categories of uncertainty, a closer inspection of the categories shows that not all sources of uncertainty are covered within the categories. For example, RAC mentions that

they do not include uncertainties in some parameter values, such as dose conversion factors (page S-24).

It is difficult even for scientists to understand the nuances of a 90% uncertainty range so as to interpret it appropriately. Hence, it would be helpful in the summary booklet to have a paragraph describing it. One common misconception about uncertainty ranges is that all possible values in the range are equally likely. It would therefore be informative to state the concept that values near the median or mean are appreciably more likely than ones out on the tails and to illustrate it with a sample probability-distribution curve from one of the Monte Carlo simulations superimposed on error bars. The last paragraph on page S-31 would be better balanced (and less alarming to the public) if RAC had also described how small the risks might be (at the 5th percentile), rather than emphasizing only the 95th-percentile risk.

Several figures in the RAC report are confusing or misleading. For example, the legend to figure 48 on page 88 of volume I indicates that cumulative doses are being presented and presumably all contributions are being added together. It is not clear why the 5% quantile always corresponds exactly to background—does this mean that the central 90% interval of the excess dose from Fernald is simply stacked on top of the background? It is not clear how the median in scenario 5 can be below the contribution of background if they are added together. Figure 51 is another example where the graph is misleading. Why is there no uncertainty associated with background? Box plots might be a more appropriate method for comparing central tendencies, spreads, and tails in a simple and consistent manner. Side-by-side bar graphs similar to figure 46 would also illustrate the comparison that is intended and would eliminate much of the confusion.

The public would probably like a means to compare the risk from the silo radon emanations with the levels of exposure that are common to the general public. To provide a context for the radon doses from the silos, RAC compares them with the background radon dose in homes for the same number of years (see table 23 on page 86 and the discussion on page 87 in volume I). A more useful comparison would probably be with the background radon dose for a lifetime, inasmuch as the 38-year dose from the silos is likewise the total that will be received over a lifetime. If a house has the median background radon level of 1 pCi/L, a person would receive 10-20 WLM in a lifetime (Lubin and others 1995). Using the dose-

conversion factor of (20 × 0.65) rems/WLM yields lung doses in the range of 1.3-2.6 Sv. This context allows people to see that the cumulative doses being talked about in the 9 scenarios (0.4-3.4 Sv) are less than what many people naturally receive in their homes. A comparison with indoor radon levels on page S-31 would be useful as a way to provide a context for these risks. In addition, the lower bound of the EPA action level of indoor radon is 4 pCi/I, which yields a lifetime lung dose of 5.2-10.4 Sv. It can be seen that the best estimates of cumulative doses from all 9 scenarios are below the ones that EPA has set as the action level.