Health Effects of Exposure to Radon: Time for Reassessment? (1994)

Underground miners, especially those working in uranium or tin mines, are exposed to a wide range of, and in some cases to high levels of, radon and its progeny. This exposure of underground miners to radon progeny has been causally linked to lung cancer (National Research Council, 1988). It is therefore not surprising that those populations of miners have served as a valuable resource for epidemiology studies, and the resulting database has been used to establish risk estimates for lung cancer that have then been extrapolated to the lower exposure levels found in homes. However, a majority of uranium miners are cigarette-smokers, and the miners are also exposed to silica, arsenic, blasting fumes, and, in some mines, diesel exhaust. The contributions of the other agents to the excess risk of lung cancer in miners have not been adequately characterized (National Research Council, 1991), and nonmalignant respiratory disorders might also be a risk factor for lung cancer (Tockman, 1994).

The BEIR IV report, published in 1988, reviewed epidemiologic studies of underground-miner cohorts that had been published through 1987:

• Uranium miners in the Colorado Plateau; Ontario, Port Radium, and Beaverlodge, Canada; Czechoslovakia; and France.

• Tin miners in Cornwall, U.K.

• Fluorspar miners in Canada.

• Iron miners in Sweden.

• Other miners in Sweden.

• Niobium miners in Norway.

The BEIR IV committee analyzed data on four of the cohorts Malmberget (Sweden) iron miners, Colorado Plateau uranium miners, Beaverlodge uranium miners, and Ontario uranium miners-to develop a risk model. Those four yielded the only data on radon-progeny exposures of individual participants to which the committee could gain access; at that time, data from the Czechoslovakian miners could not be obtained, and several other investigations were still in progress.

Since publication of the BEIR IV report, the findings on additional cohorts have been reported; all confirm the excess lung cancer incidence found in previous studies, and all demonstrate increasing risk with increasing exposure to radon progeny. The more recent cohorts include Chinese tin miners (Xuan et al., 1993), notable for the large population of exposed miners, the large number first exposed as children, and the complications caused by arsenic exposure; New Mexico uranium miners in the

Grants area (Samet et al., 1991a); Australian uranium miners in the Radium Hill mine (Woodward et al., 1991); and French uranium miners (Tirmarche et al., 1993).

Although carcinogenicity of radon progeny had been well established in humans by 1988, the year of publication of the BEIR IV report, the new studies have contributed new estimates of the relationship between radon-progeny exposure and lung-cancer risk and new evidence on the validity of assumptions used for risk modeling. The Chinese tin miners included a unique group that had started working underground as children; age at first exposure did not modify the eventual risk of lung cancer (Xuan et al., 1993). In a pilot case-control study conducted within the cohort, skeletal lead-210 concentrations were estimated by counting cranial lead-210 activity (Laurer et al., 1993). In this study of 19 lung-cancer cases and 141 age-matched controls, there was a smooth gradient of lung-cancer risk with category of lead 210 concentration; this finding indicates the potential for using skeletal lead-210 to improve the retrospective exposure estimates constructed for the epidemiologic studies.

The study of New Mexico uranium miners included information on smoking for most miners (Samet et al., 1991a). The data indicated a multiplicative interaction between smoking and radon progeny exposure, although the combined effect of the two agents could not be described with precision, because of the small number (68) of lung-cancer deaths. A case-control study conducted within the New Mexico cohort found that the presence of chest-radiograph abnormalities indicative of silicosis was not associated with lung-cancer risk (Samet et al., 1994). The studies of Port Radium, Radium Hill, and French uranium miners provided new data on exposures lower than in some of the earlier cohorts. 

The published reports on uranium miners in Czechoslovakia were reviewed in the BEIR IV report. Followup of these miners has been extended, and patterns found for lung cancer (Tomásek et al., 1994) and cancers other than lung cancer (Tomásek et al., 1993) have been reported. Lung-cancer risk varied with exposure rate, stratified by the authors at intervals of 10 WL. Among men whose exposure rates never exceeded 10 WL, the risk increased linearly with time-weighted cumulative exposure, was higher in younger men, and declined with lengthening interval since exposure. These patterns are similar to those found in the BEIR IV analysis. Higher risks in men who had worked in one particular mine suggested an effect of arsenic in the dust on lung-cancer risk. When the analysis included all miners, the patterns of risk were more complex.

The pooled analysis of four cohorts performed by the BEIR IV committee demonstrated the informativeness of combining data sets to derive risk models. The sets used by that committee included data on 360 lung-cancer deaths among 22,190 miners. With the publication of additional studies and the new opportunity to work with the team investigating the Czechoslovakian uranium miners, the U.S. National Cancer Institute (NCI) took the lead in a pooled analysis of data from 11 studies of underground miners, each of which was large (at least 40 lung-cancer deaths) and included estimates of individual exposures to radon progeny. The pooled dataset included over 2,700 lung-cancer deaths among 68,000 miners followed for nearly 1.2 million person-years. A

full description of the analysis has recently been published as an NCI monograph (Lubin et al., 1994a). We summarize here the findings most relevant to radon risk assessment.

The cohorts in the joint analysis are listed in Table 2; the NCI monograph provides detailed descriptions of the cohorts and the methods for exposure assessment. The cohorts included uranium, tin, iron, and fluorspar miners. Various methods were used to estimate exposure to radon progeny. Six studies had some data on cigarette-smoking, and a few had information on exposures in addition to radon progeny, including arsenic and silica. The birth years and followup intervals varied widely among the cohorts; the Malmberget miners were born in 1880-1919, whereas the other cohorts were much younger.

The data were analyzed with Poisson regression methods similar to those used by the BEIR IV committee, whose approach was generally followed. The effect of cumulative exposure to radon progeny was estimated, and the relation between risk and various potential modifiers was considered: attained age, duration of exposure, rate of exposure, age at first exposure, time since last exposure, and time since exposure during specified temporally defined exposure windows. Most analyses were based on a linear excess-relative risk (ERR) model:

where RR is relative risk, w is cumulative exposure to radon progeny in WLM, and is a parameter representing the unit increase in ERR per unit increase in w. As in the BEIR IV analysis, cumulative exposure was divided into the exposures received during windows defined by time since exposure. Effect modification was examined with categorical approaches and parametric functions. Curvilinearity in exposure-response trends was assessed, and attributable-risk models were fitted to the data.

As anticipated, ERR was linearly related to cumulative exposure to radon progeny. The ERR/WLM ratio decreased significantly with attained age, time since exposure, and time after cessation of exposure, but it was not affected significantly by age at first exposure. Over a wide range of total cumulative exposures, lung-cancer risk increased as exposure rate declined; that effect was not included in the BEIR IV committee's model. Although an effect of exposure rate was found in the BEIR IV analysis of data on the Colorado cohort, such an effect was not evident in the three other cohorts, and exposure rate was not considered further by the BEIR IV committee. On the basis of comparison of the risk of lung

cancer in Beaverlodge and Port Radium uranium miners, two groups considered to have substantially different exposure rates, Darby and Doll (1990) had also hypothesized that lung-cancer risk would increase with declining exposure rate.

The finding in the pooled analysis of an exposure-rate effect confirms the pattern reported from the Colorado Plateau study and supports the previously mentioned hypothesis. The inverse exposure rate has potentially important implications for the use of the miner studies to estimate risk associated with typical indoor exposure levels. However, the pooled analysis does not directly address the exposure-rate range typical of indoor environments, which lies within the lowest categories used in the models discussed below.

Information on tobacco use was available for six cohorts, and the combined effect of smoking and radon progeny was estimated with a mixed model. The combined data were consistent with a relationship between additive and multiplicative. Over 50,000 person-years and 64 lung-cancer deaths were accrued by miners who were identified as having never smoked. In this group, there was a linear exposure-response trend that was about 3 times greater than that observed in smokers. That difference might lead to modification of the multiplicative relationship of smoking and radon used in the BEIR IV model.

Other occupational agents that might confound or modify the relationship between radon progeny exposure and lung-cancer risk were also considered. Such information was available on five of the 11 cohorts. Information was available on previous hard-rock mining experience of the Colorado, New Mexico, and French cohorts of uranium miners. The effect of radon-progeny exposure did not vary significantly across strata of prior hard-rock mining. Similarly, in the Ontario uranium miners, prior gold-mining had no significant impact on risk associated with exposure to radon

progeny. Information on arsenic exposure, a causal risk factor for lung cancer, was available for the Chinese tin miners and the Ontario uranium miners. In the Chinese cohort, adjustment for arsenic exposure substantially reduced the risk associated with exposure to radon progeny, but was not statistically significant.

The NCI monograph (Lubin et al., 1994a) offers preferred models in categorical or contiguous forms. The models incorporate radon-progeny exposure during time-since-exposure windows and variables for modification of the effect of radon-progeny exposure by attained age and exposure rate or by attained age and duration of exposure. These models are similar in form to the BEIR IV model but have terms for either rate of exposure or duration of exposure (which is also an indicator of exposure rate). In addition, the larger data set available for the pooled analysis made it possible to estimate the effects of age and exposure in four windows of time since radon-progeny exposure, rather than three as in the BEIR IV model. As in the BEIR IV model, the effect of exposure declined with increasing time since exposure and attained age. A major departure from the BEIR IV model was the inclusion of terms expressing an inverse dose-rate effect. However, the range of exposure rates encompassed by the miner studies does not include the usual indoor exposure rates.

To make the best decision regarding the handling of risks, the uncertainties in risk estimates must be understood. Uncertainties arise from sampling variation in estimated parameters, potential biases in data (such as exposure-measurement error), and, especially, lack of knowledge regarding the correct model. The latter

is particularly important when it is necessary to extrapolate to age sex groups, time-since-exposure periods, and doses and dose rates that are outside the range of the data analyzed.

The report of the BEIR IV committee discussed in detail these and other uncertainty sources, but its quantitative assessment of uncertainty included only the sampling variation. Standard statistical theory can be used to assess uncertainty from this source, possibly including computer simulations as well as analytic solutions. Methods are also available to assess uncertainty resulting from exposure-measurement errors and other errors in the data, but these are generally difficult to apply and require a thorough understanding of the magnitude and nature of the errors. Less rigorous methods are available to evaluate the third source of uncertainties, and they generally require subjective judgments about the nature and magnitude of the uncertainties from various sources. Examples of such subjective assessments are found in the BEIR V report (National Research Council, 1990) and in the National Institutes of Health radioepidemiologic tables (National Institutes of Health, 1985). Such problems can be studied in part by ''sensitivity analysis" in which parts of the model are varied to assess effects on conclusions.

An overall assessment of uncertainty that combines uncertainties from all or most relevant sources has not been attempted in past radiation risk assessments. Such an assessment would be difficult and would require many assumptions, but it is a highly desirable objective for a Phase II (BEIR VI) committee.

After the publication of the BEIR IV report, ecologic analyses

showed associations between exposure indexes for radon and the incidence of several cancers, including myeloid leukemias and lymphomas (Henshaw et al., 1990; Eatough and Henshaw, 1991). The BEIR IV report summarized mortality data from several sources for cancer sites other than the lung. Because most of the individual cohort studies have inadequate statistical power for addressing malignancies other than lung cancer, an additional pooled analysis of the 11 cohorts directed at cancers other than lung cancer has been organized by Sarah Darby at the Imperial Cancer Research Fund Cancer Epidemiology Unit, University of Oxford. Publication of the findings of this study is expected by the end of 1994.

A recent report addressed cancers other than lung cancer among 4,320 male uranium miners in West Bohemia (Tomásek et al., 1993). These men had been followed for an average of 25 years. There was a small but not statistically significant excess mortality from cancers other than lung cancer compared with that expected from national rates: the observed/expected mortality (O/E) was 1.11, and the 95% percent confidence interval was 0.98-1.24. The investigators examined 28 sites and types of cancer and found significantly increased risks of cancers of the liver (O/E= 1.67) and of the gallbladder and extrahepatic bile ducts (O/E=2.26). For liver cancer, mortality did not increase significantly with exposure to radon progeny; the trend for cancers of the gallbladder and extrahepatic bile ducts was statistically significant. Those findings need to be evaluated in the light of findings from other cohorts; the analyses being conducted by Dr. Darby should be very useful in this regard.  

An appendix to the BEIR IV report addressed nonmalignant respiratory diseases in miners exposed to radon. The report noted that silicosis had been documented among underground uranium miners in the United States and that adverse effects of underground uranium mining on lung function had been shown in miners in the Colorado Plateau and New Mexico. The report concluded that the investigations had not separated the effects of radon-progeny exposure from those of exposure to other potentially harmful contaminants of the air in mines, including silica, diesel-engine exhaust, and blasting fumes. Since the release of the BEIR IV report, detailed literature reviews have reached similar conclusions (Samet and Morgan, 1991; Samet and Simpson, 1991). Further evaluation of all these studies is needed.

Risk assessments of lung cancer associated with residential radon exposure have been based on extrapolation of estimates obtained from studies of underground miners. Concentrations in the mines were generally much higher than typical concentrations in residences, and factors influencing exposure-dose relations, such as breathing and activity patterns, are also expected to be different between the residential and mining environments. In addition, the exposed underground miners were predominantly male smokers of

a narrow age range. Those differences necessarily lead to uncertainties in applying miner-based estimates of risk to the general population, which includes males and females of all ages, both smokers and nonsmokers, who are exposed to the lower levels found in homes.

To reduce the uncertainties, epidemiologic investigations designed to estimate the risk associated with indoor radon directly have been undertaken. Since the publication of the BEIR IV report, numerous case-control and ecologic studies on indoor radon and lung cancer have been initiated. The findings of some of the studies have already been reported and the results of the principal case-control studies now in progress should become available over the next few years. Both case-control and ecologic studies are discussed below. For each type of study, the current status of the research is reviewed first, and then the problems and limitations of the studies are discussed.

In case-control studies of the effects of residential radon, persons with lung cancer (cases) and persons without lung cancer (controls) are characterized and compared with regard to radon exposure and other factors. In most recent studies, cases and controls have been matched on age, sex, and possibly other factors; and interviews have been conducted to obtain information on residential histories, time spent in the home, occupation, smoking, and other potential risk factors. Most recent studies have also included long-term (several months) residential radon measurements, and some have been limited to females or to nonsmoking females. Because these studies include information on individuals,

they are regarded as much less subject to bias than ecologic studies. The case-control design was endorsed by the 1989 International Workshop on Residential Radon Epidemiology as the method of choice for the direct assessment of risks associated with residential radon exposure (Samet et al., 1991b).

When the BEIR IV report was prepared, only a few small studies in Sweden, including about 100 lung-cancer cases, had been completed. The findings of these studies suggested an association of lung cancer with residential radon exposure, but radon had been measured in only one study. Since the BEIR IV committee completed its review, results of additional studies have been reported, and many new studies have been initiated. Neuberger (1992), in a recent review article on studies in progress, listed 20 studies that included a total of more than 12,000 lung-cancer cases and more than 19,000 controls. Most of these studies incorporate long-term radon measurements in the homes of cases and controls and the collection of data on smoking and other risk factors.

Results of three such studies that included measurements in homes of individual subjects had been published at the time of this review. These studies--conducted in New Jersey (Schoenberg et al., 1990), Sweden (Pershagen et al., 1992), and China (Blot et al., 1990)-include 480 cases and 442 controls, 210 cases and 400 controls, and 308 cases and 356 controls, respectively, and were considered in a pooled analysis recently by Lubin et al. (1994b). The analyses indicate that the findings from the studies are not inconsistent with one another and that pooling of all three studies

makes the estimate of risk sufficiently imprecise to be consistent both with no effect and with predictions of the BEIR IV model. This initial pooled analysis provides warning of the difficulty of precisely estimating the risk associated with indoor radon directly from case-control data.

Results of a second study in Sweden (with 1,360 cases and 2,847 controls) (Pershagen, personal communication) and of a study in Winnipeg, Manitoba, Canada (with 738 cases and 738 controls) (Létourneau, personal communication) are in press and would be available for evaluation by a Phase II (BEIR VI) committee. For studies presented at the Second International Workshop on Residential Radon, held in July 1991, but not yet reported in print, investigators were queried; Table 3 shows when investigators expect results to be available. The study in Florida has been completed, and results of the Missouri study are expected to be published within a year. Results of a few additional studies might be available in time to be included in a BEIR VI report, but results of other studies, such as the large study in Germany, are unlikely to be available.

The participants in the second international workshop strongly recommended that pooled analyses of data from case-control studies be conducted (U.S. Department of Energy, 1991). Such analyses are needed to obtain the most precise direct quantitative assessment of risk, to evaluate the consistency of findings from different studies, and to achieve sufficient power to address questions related to effect modification by smoking, sex, age at exposure, and other factors. As indicated above, a pooled analysis has already been conducted, and pooled analyses involving additional studies might be available for evaluation by a BEIR VI committee. Planning for pooling of the data from European studies has also been initiated. However, it is clear that a pooled analysis includ-

ing all the major studies now in progress will not be complete in time for BEIR VI evaluation.

The direct study of the effects of residential radon exposure is subject to several methodologic limitations. The most obvious is that large samples are needed if predicted risks based on the studies of underground miners are correct. Lubin and colleagues (Lubin et al., 1990) evaluated the sample sizes required in a case control study to detect an association between lung cancer and residential radon exposure under the assumption predictions based on underground miners are correct, to detect that the quantitative effect is half that predicted from underground miners, and to detect a departure from an additive interaction between smoking and radon exposure if the true interaction is multiplicative rather than additive. Their sample-size estimates for a case-control study assume that the distribution of exposure in the population being studied is similar to that described by Nero and colleagues (Nero et al., 1986) for the U.S. population, that cases and controls are 60-65 years old, and that the study includes twice as many controls as cases. Under the assumptions that subjects lived in a single residence for 60 years and that the underground-miner risk is that from the BEIR IV analysis, the numbers of cases required to achieve 90% power for testing the three hypotheses were 251, 1,610, and 764, respectively. However, both lowering the underground miner-risk and increasing the number of assumed residences increased the sample sizes substantially. For example, the assumption of three residences for each case and control and lowering the assumed underground-miner risk from 1.5 %/WLM to

1.0%/WLM increased the sample size requirements for testing the three hypotheses to 952, 5,163, and 3,344, respectively.

A second limitation of residential radon studies is that exposure estimates are subject to many sources of error. Even in the most straightforward scenario in which subjects have lived their whole life in a single residence for which long-term exposure measurements are available, there are uncertainties in exposure estimates. For example, the measurements might not adequately reflect conditions in the home over the full period of residence, the amount of time the subject spent in different rooms in the home cannot be determined precisely, and exposures received in places other than the home cannot be estimated. When persons have resided in more than one residence, additional error can be introduced, particularly if measurements cannot be made in all residences.

It is well known that even unbiased measurement error can lead to underestimation of the risk coefficient and can reduce the power to detect effects. The sample-size determinations described above were based on the assumption that exposures of all study subjects are measured perfectly. Modification of this assumption to include the possibility of 50% misspecification of exposure (that is, the error is assumed to constitute 50% of the true exposure) increased required sample sizes for testing the hypothesis of no effect (hypothesis 1) by a factor ranging from 1.5 to 2.8, depending on assumptions made about the number of residences occupied by each subject. Statistical analyses of case-control studies conducted thus far have not accounted for exposure measurement error; thus, both risks and uncertainties were probably underestimated.

A third limitation is posed by the observational nature of epidemiologic studies and the resulting potential for confounding bias; confounding is of particular concern when small changes in risk

are studied (Monson, 1980; Boice and Land, 1982; Rothman, 1986), as expected in most studies of residential radon exposure. In the studies of residential radon, information on key variables of interest is obtained through interviews, and cases and controls might differ in the adequacy and validity of their responses, especially when interviews must be conducted with persons other than the subject. Information bias is probably not of great concern in residential histories, but it could be important with regard to other factors, such as smoking. Even if the information on potential confounding factors is equally valid for cases and controls, data on smoking and other risk factors are not likely to be sufficiently detailed and accurate to permit complete adjustment in analyses. Like exposure measurement error, potential confounding introduces additional uncertainty that is not reflected in usual statistical confidence limits. Because of the enormous confounding effect of cigarette use, some studies limited to life-long nonsmokers may be needed despite the costs and sample sizes.

In summary, because of the limitations described here, case-control studies of exposure to indoor radon probably cannot provide quantitative risk estimates that are sufficiently precise to allow these studies to supplant extrapolation from underground miners for risk-assessment purposes, particularly if the consequences of measurement error and confounding factors are fully acknowledged. Reported results indicate that individual studies are likely to yield a range of results, extending from no apparent effect to effects greater than anticipated from the studies of miners. A definitive assessment of risk should not be anticipated from any individual study, and combined analyses are clearly needed. In interpreting the findings of the case-control studies, it will also be important to consider uncertainties from sources other than statistical variation. An important task for a BEIR VI committee will be

to evaluate the appropriate role of case-control studies in radon risk assessment.

In ecologic studies, groups, rather than individuals, are used as the unit of analysis. Because the groups are usually defined in terms of geographic areas in which mortality or incidence data are available, ecologic studies can often make use of existing data and so can be conducted quickly and inexpensively. However, methodologic problems greatly limit the usefulness of ecologic studies, and they are generally considered to be useful primarily for developing new hypotheses regarding exposure-disease relationships.

Stidley and Samet (1993) provide a detailed review of the results and methods of 15 ecologic studies of lung cancer and indoor radon exposure. In eight of the studies, lung-cancer rates were compared for two or more groups defined by exposure status, and seven were regression studies in which rates were modeled as a function (usually linear) of exposure. Exposure estimates used in these studies included both surrogate measures, such as the geologic characteristics of an area, and estimates based on current measurements of indoor radon from samples of homes in an area. Most of the studies were published recently, and only three were evaluated by the BEIR IV committee. Seven of the 15 studies reported positive associations, six no

association, and two negative associations. Of the seven positive studies, only one included adjustment for smoking, and six of these were based on surrogate measures of indoor radon exposure. Of the six studies showing no association, three included adjustment for smoking, and three of these were based on surrogate measures of exposure. All were judged by Stidley and Samet to have low power for detecting effects under reasonable alternative hypotheses. The two studies showing negative associations were both regression studies that included adjustment for smoking and were based on samples of exposure measurements (Haynes, 1988; Cohen, 1990; Cohen and Colditz, 1992); an additional comparison study recently reported a negative association but did not include adjustment for smoking (Neuberger et al., 1992). Several methodologic shortcomings could have biased all studies.

Limitations of the ecologic-study design and the special biases to which ecologic studies are subject have been discussed extensively (Piantadosi et al., 1988; Greenland and Morgenstern 1989; Greenland, 1992; Stidley and Samet, 1994). There are often problems with the exposure measurements used in these studies. The need to use groups on which vital statistics are available can obscure much of the variability in indoor radon exposure and lead to low statistical power for detecting effects, even if the average exposures are assessed correctly. Estimation of exposure from surrogate measures or from sampled homes is likely to reduce power still further and might also introduce bias.

An additional difficulty is that grouped data can be especially subject to confounding if the geographic regions chosen reflect

differences in other risk factors. It is often not possible to control for potential confounders adequately because suitable data are not available or because, with grouped data, adjustment for confounders cannot be correctly carried out unless the risk is a linear function of both indoor radon concentration and other variables for which adjustment is needed. In evaluating lung-cancer risks, smoking is the factor of greatest concern, and data on smoking are generally not available for finely defined geographic regions. Stidley and Samet (1994) show that even a small negative correlation of radon exposure and smoking could induce a negative correlation of radon exposure and lung-cancer risk. As noted before, ecologic studies are regarded as useful primarily for the generation of hypotheses that must be further examined by means of with other study designs. There is already strong evidence from studies of underground miners that exposure to indoor radon can cause lung cancer. The major uncertainty is the magnitude of risk at residential concentrations.

Ecologic studies by Henshaw and colleagues (1990) examined average radon exposures and disease rates for leukemia and other cancers in 14 countries and in regions of the United Kingdom and Canada. They identified significant positive correlations for childhood and adult leukemia, kidney cancer, melanoma, and prostatic cancer. Henshaw and colleagues (1992) presented calculations for radon-derived a-radiation dose to bone marrow and skin and proposed that these calculations supported a causal explanation for the identified correlations.

Butland and co-workers (1990) noted the poor quality of radon exposure data from some countries, indicating that these data did not always cover the same areas as the cancer data. They also noted that countries with the highest-quality cancer data (e.g., Scandinavian countries) tended to be those with the highest radon exposures, and they pointed to problems in the statistical methods used by Henshaw and colleagues. Muirhead and co-workers (1991) conducted an ecologic analysis based on small areas of the United Kingdom and found no significant associations with radon exposure, even though analyses by aggregated areas (counties) did show a correlation.

Results of several epidemiology studies have been published since the BEIR IV report. The results from miner studies have increased our knowledge of the association between radon-progeny exposure and lung-cancer risk, including an increase in the population database, the introduction of new information in humans first exposed as children, and information regarding the potential role of other agents and nonmalignant respiratory disease in causing lung cancer. Results of one of these studies (Laurer et al., 1993) suggest that skeletal lead-210 measurements might be useful for dose reconstructions to improve the epidemiologic studies. Data have also been used to investigate the interaction between smoking and radon-progeny exposure. A pooled analysis of data from 11 studies of underground miners has recently been published (Lubin et al., 1994), and it contains several observations relevant to risk reassessment, including information regarding the confounding

effects of smoking. Ecologic analyses have shown associations between exposure indexes for radon and the incidence of nonrespiratory cancers.

Numerous case-control and ecologic studies on indoor radon and lung cancer have been initiated. Results of some studies have been published, and results of the case-control studies now in progress should become available in the next few years. The most recent case-control studies include long-term residential radon measurements and matching on age, sex, and other factors. At least 20 studies in progress include more than 12,000 lung-cancer cases and more than 19,000 controls (Neuberger, 1992). Results of some of these studies might be available in time to be included in a BEIR VI reassessment, but the committee recognizes that the results of several studies will not be completed in time to be available to the Phase II committee. Ecologic studies of other cancers and indoor radon exposure have been also published (Stidley and Samet, 1993), but limitations in these studies have made it difficult to ascertain whether there is a correlation between radon exposures and cancers other than lung cancer.

• Because the pooled analyses of 11 cohorts are likely to serve as the basis for developing a new risk model for radon and lung cancer, a Phase II committee will have to critique this work, identify additional analyses needed, and possibly gain access to the data on the 11 cohorts.

• The Phase II committee should consider biologically driven modeling of the pooled data.

• The Phase II committee should formally evaluate sources of exposure error in the miner cohorts and the consequence of these errors for risk estimation and risk assessment.

• The Phase II committee should consider new evidence on arsenic, silica, and other factors that can modify or confound the risks estimated from miner studies.

• Formal analysis of uncertainties in extending risk models from miners to the general population should be undertaken.

• The Phase II committee should evaluate possible risks of cancer at sites other than lung, particularly with the data from the pooled analysis of the miner studies in progress.

• The Phase II committee must evaluate and interpret the results of case-control and ecologic studies that have been reported and of the forthcoming studies, including evaluation of the limitations and uncertainties in these studies.

• Results of case-control studies reported thus far have generally been consistent with estimates based on extrapolation from data on underground miners but have not provided enough precision to rule out the possibility of no risk or of risks that are substantially larger than those obtained through extrapolation. The Phase II committee should determine the potential role of current and future case-control studies for validating risk models based on miners and more generally for developing risk models of residential exposure to radon.

• The Phase II committee should make recommendations regarding whether it is desirable to initiate new case-control or ecologic studies of residential radon.