Mechanistic Model Predicts a U-Shaped Relation of Radon Exposure to Lung Cancer
Risk Reflected in Combined Occupational and U.S.
Residential Data
Kenneth T. Bogen, Dr. P.H.
Health and Ecological Assessment Div. (L-396),
Lawrence Livermore National Laboratory,
Livermore, CA 94550, USA
Tel: (925) 422-0902
Fax: (925) 424-3255
Email: bogen@LLNL.gov.
Abbreviations: BEIR = NRC Committee on the Biological Effects of Ionizing Radiations, CD2 = cytodynamic 2-stage, CL = confidence limit, LCM = lung cancer mortality, NRC = National Research Council, PY = person-year, RR = relative risk, WF = U.S. white female, WLM = Working Level Month.
A mechanistically based cytodynamic 2-stage (CD2) cancer model was shown recently to predict both ecologic U.S. county data and underground-miner data on lung-cancer mortality (LCM) vs. radon concentration, indicating biological plausibility of the apparent negative dose-response relation exhibited by the ecologic data (Bogen, Hum. Ecol. Risk Assess. 3, 157-186; 1997). To further investigate this hypothesis, the CD2 model was fit to combined age-specific LCM data vs. estimated radon-exposure in white females of age 40+ y in 2,821 U.S. counties during 1950-54 using new estimates of county-specific mean residential radon exposure, and in five cohorts of underground nonsmoking miners. The negative association of radon levels and corresponding county-level LCM rates apparent in women dying in 1950-54 (11% of whom ever smoked) was also apparent in women of age 60+ y (5% of whom ever smoked). The CD2 fit obtained to the combined residential and occupational data was found to predict the combined data using biologically plausible parameter values, and also to predict inverse dose-rate effects exhibited in nonsmoking miner data to which the CD2 model was not fit. These results are consistent with the hypothesis that residential radon exposure has a nonlinear U-shaped relation to LCM risk, and that current linear extrapolation models substantially overestimate such risk.
Radon decay yields alpha radiation that is genotoxic and cytotoxic, even at low doses. Increased lung cancer mortality (LCM) is associated with radon exposure in experimental animals and in underground miners (NRC, 1988,1998). LCM risk from residential radon exposurenow estimated by linear-no-threshold extrapolation from data on LCM in minersis thought to pose the greatest threat from indoor air pollution in the U.S., causing ~10% of all lung cancer and ~20-30% of all lung cancer in nonsmokers (NRC, 1988,1998; Puskin, 1992). But age-adjusted U.S. county-level LCM rates exhibit a substantial apparent negative correlation with residential radon levels, inconsistent with linear-no-threshold predictions (Cohen, 1995,1997). Both this negative correlation and elevated LCM in miners were shown to be predicted by a biologically based "cytodynamic 2-stage" (CD2) model of lung cancer (Bogen, 1997). Notably, although this model was fit only to data on residential/occupational LCM, the resulting fit also predicted the general pattern of "inverse dose-rate" effects seen in miners (for whom greater risks have been posed by a given cumulative exposure when incurred over a longer duration; see Lubin et al., 1994,1995), and did so with biologically plausible parameter estimates.
As applied to radon, the CD2 model (Fig. 1) realistically posits linear-no-threshold dose-response relations for alpha-induced cell killing and critical mutations. However, unlike previous applications of the related 2-stage stochastic "MVK" model (e.g., to radon; see Moolgavkar et al., 1993; Luebeck et al., 1996) in which premalignant cell proliferation is presumed to increase monotonically with cytotoxic dose, a key feature of the CD2 model is that it may reflect net cytotoxic loss induced in exposed premalignant as well as exposed normal stem cells. The CD2 model thus predicts reduced cancer risk whenever (i) induced cytotoxicity is sufficient to negate a slight net proliferative advantage presumed for spontaneous premalignant clones, but (ii) induced mutations yield insufficiently many new premalignant clones to offset the latter effect on tumor likelihood.
Analyses involving ecologic data are always subject
to possibly irresolvable confounding, and thus
typically serve as a heuristic source and screen of hypotheses
that require testing by more rigorous methods (Piantadosi
et al., 1988; Piantadosi, 1994). With this caveat in mind,
the preliminary study (Bogen, 1997) had several
additional limitations insofar as it involved modeling relative
risk (RR) estimates based on county-specific
age-adjusted LCM rates for U.S. males in the 1980s and
corresponding residential radon estimates published by
Cohen (1995), and on BEIR IV summary data on LCM
in Colorado Plateau uranium miners in specified ranges
of cumulative occupational radon exposure (NRC,
1988). Consequently, biases may have arisen from
spurious smoking/radon/age-distribution correlations within
the residential data used (Piantadosi, 1994; NRC,
1998), by nonrandom errors in the residential-exposure
estimates used (which were derived by ad hoc methods
from disparate sources), and by aggregation errors
introduced by use of summary occupational data. Typical of
other MVK-type model applications, the preliminary study
also involved fitting lifetime rather than age-specific risks
of cancer mortality, without examining the consistency
of age-specific risks implied by the fit obtained to
corresponding empirical (or plausible) age-specific
risk patterns. Fits to age-specific risk (or "hazard")
patterns are preferable, since many age-specific risk patterns
can yield the same lifetime risk (i.e., integrated
hazard) pattern. Another drawback of the preliminary study
was that the parameter reflecting alpha cytotoxic potency
was estimated (to be ~3-fold greater than indicated by
Figure 1.
Cytodynamic 2-stage (CD2) model of bronchial
carcinogenesis (Bogen, 1997), incorporating a mechanistic
2-stage ("MVK") framework (dashed box), whereby normal
epithelial stem cells (S) may each with probability
m1 per cell division give rise to a premalignant cell
(P), which may proliferate clonally and with probability
m2 give rise to a malignant cell
(M). The CD2 model adds a reservoir of unexposed cells
(R) may play a alpha-enhanced role in replacing
S-cells lost, e.g., at rate
kr to a pool of reproductively dead cells
(D). R-cells may progress to premalignant
(Q) and malignant (M) cells via
independent, parallel processes involving S and
P cells. Rates of birth (b) and death/differentiation
(d) are specified for each cell type,
fR is the ratio
R/S under normal conditions, and bold
arrows indicate potential cytotoxic/mitotic induction.
These limitations were addressed, to some degree, by refitting the CD2 model, conditional on likely alpha cytotoxicity, to combined age-specific data on LCM in white females of age 40+ y in 2,821 U.S. counties during 1950-54 using new estimates of county-specific mean residential radon levels, and in 5 cohorts of nonsmoking miners. What follows is intended as only a brief outline of methods used for this new analysis, and a summary of key results and implications.
Residential Mortality and Related Smoking Data. Age- and county-specific 1950-54 LCM rates were obtained for U.S. white females (WF) in 5-y age intervals ending in 85+ y, excluding data for Virginia due to unreliability of VA-county-level data at that time (Marsh et al., 1996). Analyses excluded data on women <40 y, for whom LCM was quite rare. Only ~11% (vs. ~5%) of WF who died at 40+ (vs. 60+) y in 1950-54 ever smoked, based on survey data covering this period (Mills and Porter, 1953; Haenzsel and Shimkin, 1956). Based on the latter studies' account of how those smokers tended to smoke less and start at a later age relative to subsequent smoking women, it can be estimated that LCM in all WF who died at 40+ (vs. 60+) y in 1950-54 was increased by an RR of 1.74 (vs. 1.16) over the baseline rate(s) expected for WF never-smokers who died during that period. This ~5-fold difference in expected excess RR for 1950-54 LCM in WF of age 40+ vs. 60+ y indicates that any ecologic association between radon exposure and LCM in these women due solely or primarily to confounding by smoking should be reduced in WF aged 60+ vs. 40+ y. Rn-related trends for 1950-54 LCM adjusted for age and other factors were therefore examined for WF of age 40+ and 60+ y as described below. WF data were modeled for age <80 y only, because the general pattern of LCM increase (a very nearly cubic function of age) did not hold for the 80-84 and 85+ age groups. The mortality-rate decline in the oldest age groupswell known to pertain to nearly all types of cancer (Armitage and Doll, 1957)may be due, e.g., to data unreliability (Doll and Peto, 1981) and/or population heterogeneity in cancer susceptibility, neither of which are addressed by the CD2 model. For modeling, county-level LCM data adjusted by family income (see below) were pooled within 6 ranges of mean household radon exposure.
Rn-Exposure, Socioeconomic and Climatic Data for U.S. Counties. This study used new estimates of annual geometric mean household radon concentrations in U.S. counties based on analyses of long-term vs. short-term U.S. EPA monitoring data adjusted for climatic and other variates (Price, 1997; Price et al., 1998). Geometric mean levels were scaled uniformly to corresponding arithmetic means assuming lognormal intra-county distributions with a common geometric standard deviation of 2.0 (Cohen, 1992a; Price, 1997). In addition to VA data as noted, data for major retirement states (AZ, CA, FL) were dropped in view of survey data indicating the high % of lifetime spent near residence at time of death in non-retirement states (Cohen, 1992b)expected even more so for WF dying in 1950-54. For the remaining 2,821 counties, 12 types of 1950 demographic/socioeconomic data (USBC, 1953) and 5 typical 1953-1975 climatic measures (Apte et al., 1997) were binned into county quintiles, and were, together with a 3-level dietary Se index (Clark et al., 1991) and U.S. region (among 9), included separately and in pairs as factors used together with age in adjusted trend analyses.
Occupational Data. Information from 5 of 6 cohorts for which data on LCM in nonsmoking underground miners are available (Lubin et al., 1994,1995; NRC, 1998) was kindly provided by Dr. J. Lubin and coworkers. These person-year (PY) data (n = 2,488, 44,600.7 PY, 53 cases) were summarized by total LCM, PY and corresponding PY-mean attained age, exposure duration (DUR), and cumulative exposure (in Working Level Months, WLM), for the 5 WLM and 3 attained-age bins used by Lubin et al. (1994, pp. 84-5), and for the DUR bins: 0-8, 8-16, and >16 y.
Cancer Risk Model. The CD2 model (Fig. 1) was
used with the changes noted. This model adapts the "MVK"
2-stage mechanistic framework, in which transition
of normal (S) to premalignant (P) cells and of P-cells
to malignant cells (M) is modeled as a doubly
stochastic filtered Poisson process (Moolgavkar et al.,
1993). As applied to radon, the CD2 model additionally
assumes: (i) alpha-induced transition from S to a pool of
reproductively dead cells (D), (ii) replacement of
S-cells partly by virtually unexposed cells (R) via a Verhulst
feedback-inhibition process, and (iii) similarly unexposed
premalignant (Q) cells derived from R-cells and subject
to malignant transformation via a process similar to
and independent from the S
PM process. In terms
of notation for CD2 parameters and relations
previously described (Bogen, 1997), new assumptions used for
the present study were that: (1) dose rate (E) in cGy
y-1 to surface cells in lobar/segmental bronchi is given by
3.3 vs. 4.4 cGy WLM-1 for residents vs. miners (from
NRC, 1991); (2) excess relative risk = sxE (unitless) for
SP and PM transitions; (3) kr =
E/D0, with D0 = 35 cGy, a weighted mean of human lung-cell values
(Raju et al., 1993; Simmons et al., 1996); (4)
dQ = bQ g[1 +
c(bRb-11)]; (5) RQ and
QM (vs. SP and PM) transitions occur at a background rate (per cell division) of
wxm (vs. m); and (6) b = 4y-1 (a plausible target-cell turnover
ratesee Bogen, 1997). As before, 0.1935 WLM L
pCi-1 y-1 was assumed for residents
(Puskin, 1992), the fraction fR (= R/S under normal conditions) was estimated, and
other CD2 parameters were assigned biologically
plausible values (given in Bogen, 1997).
Data Analysis. Under these assumptions, the CD2 model with 6 estimated parameters (m , w , fR, g, c, and s) was fit to 8x6 = 48 income-adjusted age- and exposure-specific residential LCM rates, plus 3x5 = 15 age- and WLM-bin-specific occupational LCM rates, assuming the corresponding errors estimated by standard methods (Chiang, 1984) are Poisson (and hence, particularly those of adjusted WF LCM rates, ~normally) distributed. Otherwise, the CD2 model was evaluated analytically and fit by weighted chi-square minimization as previously described, except that CD2 hazards were calculated rather than lifetime probabilities (see Bogen, 1997). Outlying data were assessed by a corresponding F-test.
The resulting fit was compared graphically with age-adjusted RRs and associated confidence limit (CL) values corresponding to residential and miner LCM data. (The residential comparison also shows predictions made using "preferred" BEIR VI risk-extrapolation models; NRC, 1998). Similarly, age-adjusted RRs estimated for nonsmoking miners exposed over different durations to 800-1600 WLM (who clearly exhibit an inverse dose-rate effect) were compared to those implied by the CD2 fit obtained to the former LCM (not the latter exposure-duration) data. For these graphical comparisons, (internally) standardized RR implied by the CD2 fit (or predicted by BEIR VI models) were defined as the corresponding weighted mean of predicted age-specific RRs, using inverse variances of age-specific 1950-54 rates of WF LCM as common weights (and, for the CD2 model, an occupational dose-duration relation reflecting PY data for the nonsmoking miners). Numerical maximum likelihood methods (Breslow and Day, 1987) were used to obtain RR and CL values for the latter analyses, and adjusted chi-square values for trend analyses involving the residential LCM data. RR trend was also summarized as the standardized RR slope, B = badj/a, where badj (the adjusted LCM slope) and a (the unadjusted LCM intercept) were estimated by standard methods (Fleiss, 1981).
Results of trend analyses involving the WF LCM data, summarized in Table 1, indicate an apparently robust negative LCM-vs.-radon association adjusted for age and one additional variate among those considered. Notably, the fact that results using data on women of age 40+ y do not differ substantially from those for women of age 60+ y, indicates that confounding by smoking is unlikely to account for the apparent negative trend (see Methods). Similar results (not shown) were obtained using age and all combinations of two additional variates as adjustment factors.
The adequate CD2 fit obtained to combined set
of age-specific residential/miner LCM data
(
2 = 73.8, df=57, p = 0.066) was improved significantly
(F2,55 = 9.72, p=0.00024) by dropping one outlying data point
from each data subset, to yield an excellent fit
(2 = 54.6, df=55, p = 0.49) with parameter estimates
(±100% x SE/estimate) of: mb = 0.76 x
10-8 y-1 (± 12%), w = 3.7
(± 27%), fR = 0.063 (± 50%),
g = 0.0893 y-1 (± 4.9%), c = 0.35
(± 65%), and s = 0.11 y cGy-1
(± 160%). RR data summarizing LCM in U.S. WF in 1950-54 as a function of
county-mean household radon concentration are compared
in Fig. 2a to those predicted by the latter CD2 fit
under residential-exposure assumptions. This figure also
shows how these RRs are underestimated substantially
by "preferred" BEIR VI models (NRC, 1998).
Figure 2b shows how, under mining-exposure assumptions
reflecting the actual experience of nonsmoking miners,
the CD2-model predicts RRs consistent with those
summarizing the age-specific miner data used. Figure
3 shows how, under similar assumptions concerning
nonsmoking miners exposed to 800-1,600 WLM, the
CD2-model predicts RRs consistent with the significant "inverse
dose-rate" effect indicated for those miners, even though
the model was not fit to these data.
Table 1.
Trend in the relative risk (RR) of lung cancer mortality (LMC) among women in U.S. countries from 1950-54 as a function of country mean residential radon level, adjusted for various factors.
aLCM was compared among 6 groups of
counties classified by mean residential radon level,
after adjusting for age and the following factors classified, unless otherwise specified, into
U.S.-county quintiles: Agwork = % employed in agriculture, Density = persons km-2, Fem. work
= % females in total labor force, Heating IDD = heating infiltration degree-days, High school
= % completed high school or more, Income = median family income, Migration = #
persons living in a different county or abroad in 1945
vs. 1950, Poor = % with income <$2000, PrecipHr
= mean daily hours of precipitation, Region = location within 9 U.S. divisions, Rich = %
with income > $5000, Rural/Urban = % rural-farm/urban population, School = median years
of schooling completed, SeBin = index (0, 1, or 2) of relative exposure to dietary selenium
based on foliage Se content, TempJan/TempJul = mean daily temperature for Jan/Jul, Uneduc
= % who completed grade <5, Wind = mean daily wind speed.
b = standardized RR slope (see Methods); CV
= 100%±(standard deviation of b)/b.
Figure 2.
Relative risk (RR) of increased lung cancer
mortality (LCM) experienced by (a) U.S. white females (WF)
during 1950-54 as a function of county-mean residential
radon concentration (within six concentration ranges), and
(b) a total of 2,488 nonsmoking underground miners as a
function of cumulative underground mining exposure adjusted for
age and cohort. Both sets of RR estimates are based on
internal comparisons to data (solid points) corresponding to the
lowest exposure group (RR = 1, dashed line). The RR estimates
are compared to values predicted by the 6-parameter CD2
model fit to >60 age-specific LCM rates for the WF and
miners corresponding to the RR estimates shown. Plot (a) also
shows RRs predicted for female nonsmokers by the "preferred"
(12- and 13-parameter) BEIR VI linear-extrapolation models:
BEIR VIc = age-exposure-concentration model, BEIR VId =
age-exposure-duration model (NRC, 1998).
Figure 3.
Relative risk (RR) of increased lung cancer
mortality (LCM) experienced by nonsmoking underground
miners exposed to 800-1,600 WLM as a function of duration of
mining exposure, adjusted for age and year of observation, based
on internal comparison to LCM in miners exposed for only 0-8
y (solid point on dashed line indicating RR = 1). The
RR estimates are compared to values predicted by the
6-parameter CD2 model fit to >60 age-specific LCM rates for 1950-54
WF and nonsmoking miners that correspond to the summary
RR data points shown.
Results obtained from this analysis are consistent with those obtained previously using the CD2 model (Bogen, 1997). For the several reasons mentioned in the introduction, but particularly insofar as the potential role of smoking as a confounding factor was reduced in the present study by using data pertaining largely or exclusively to nonsmokers, the new results may be considered more convincing than the previous study. As both studies relied on ecologic data, they can only indicate a mechanistically based, biologically plausible basis for challenging linear-no-threshold models (such as the 12- or 13-parameter BEIR VI models) applied to radon, and for supporting the hypothesis long argued by Cohen (1995,1997) that LCM may typically be negatively associated with residential radon exposure in the U.S. Specifically, results from the present study provide a mechanistic basis for the plausibility of a U-shaped (or "hormetic") dose-response for radon's effect on LCM risk. This plausibility certainly highlights the issue of how best to address fundamental model uncertainty in the context of quantitatively based risk management (Bogen and Layton, 1998). But CD2 modeling results also pose specific, mechanistic hypotheses concerning the effect of subchronic or chronic exposure to relatively cytotoxic genotoxins, such as alpha radiation, on the growth kinetics of premalignant foci. Because these hypotheses can now be tested experimentally, the current challenge is to initiate and organize support for this effort.
This work was performed under the auspices of the U.S. Department of Energy at Lawrence Livermore National Laboratory under contract W-7405-ENG-48. Many thanks to Dr. J. Lubin and coworkers for access to miner data, and to J. Cullen (U. Calif. Berkeley), L. Wilder (LLNL) and Dr. G. Keating (LLNL) for data preparation/management that supported this analysis.
Apte, M. G., Nero, A. V., and Rezvan, K. L. 1997. Meteorological database for the United States. Indoor Air 8, 1-7.
Armitage, P., and Doll, R. 1957. A two-stage theory of carcinogenesis in relation to the age distribution of human cancer. Br. J. Cancer 11, 161-169.
Bogen, K. T. 1997. Do U.S. county data disprove linear no-threshold predictions of lung cancer risk for residential radon? a preliminary assessment of biological plausibility. Hum. Ecol. Risk Assess. 3, 157-186.
Bogen, K. T., and Layton, D. W. 1998. Risk management for plausibly hormetic environmental carcinogens: The case of radon. BELLE Newsletter (in press).
Breslow, N. E., and Day, N. E. 1987. Statistical Methods in Cancer Research, Volume IIThe Design and Analysis of Cohort Studies. IARC Scientific Publ. No. 82. International Agency for Cancer Research (IARC), Lyon, pp. 106-118.
Chiang, C. L. 1984. The Life Table and its Applications. R.E. Krieger Publ. Co., Malabar, FL, pp. 91-111.
Clark, L. C., Cantor, K. P., and Allaway, W. H. 1991. Selenium in forage crops and cancer mortality in U.S. counties. Arch. Environ. Health 46, 37-42.
Cohen, B. L. 1992a. Compilation and integration of studies of radon levels in U.S. homes by states and counties. Crit. Rev. Environ. Control 22, 243-364
Cohen, B. L. 1992b. Percentage of lifetime spent in area of residence at time of death. Environ. Res. 57, 208-211.
Cohen, B. L. 1995. Test of the linear-no threshold theory of radiation carcinogenesis for inhaled radon decay products. Health Phys. 68, 157-174.
Cohen, B. L. 1997. Lung cancer rate vs. mean radon level in U.S. counties of various characteristics. Health Phys. 72, 114-119.
Doll, R., and Peto, R. 1981. The causes of cancer: Quantitative estimates of avoidable risks of cancer in the United States. J. Natl. Cancer Inst. 66, 1191-1308.
Fleiss, J. L. 1981. Statistical Methods for Rates and Proportions, 2nd ed. John Wiley & Sons, New York, NY, pp. 143-146.
Haenszel, W., and Shimkin, M. B. 1956. Smoking patterns and epidemiology of lung cancer in the United States: Are they compatible? J. Natl. Cancer Inst. 13, 1417-1441.
Lubin, J. H., Boice Jr., J. D., Edling, C., Hornung, R. W., Howe, G., Kunz, E., Kusiak, R. A., Morrison, H. I., Radford, E. P., Samet, J. M., Tirmarche, M., Woodward, A., Yao, S. X., and Pierce, D. A. 1994. Lung cancer and radon: A Joint Analysis of 11 Underground Miners Studies. NIH Publ. No. 94-3644. U.S. National Institutes of Health, Bethesda, MD.
Lubin, J. H., Boice, J. D., Edling, C., Hornung, R. W., Howe, G., Kunz, E., Kusiak, R. A., Morrison, H. I., Radford, E. P., Samet, J. M., Tirmarche, M., Woodward, A., and Yao, S. X. 1995. Radon-exposed underground miners and inverse dose-rate (protraction enhancement) effects. Health Phys. 69, 494-500.
Luebeck, E. G., Cutis, S. B., Cross, F. T., and Moolgavkar, S. H. 1996. Two-stage model of radon-induced malignant lung tumors in rats: Effects of cell killing. Radiat. Res. 145, 163-173.
Marsh, G. M., Ehland, J., Sefcik, S., and Alcorn, C. 1996. Mortality and Population Data System (MPDS). MPDS Technical Report. University of Pittsburgh, Department of Biostatistics, Pittsburgh, PA.
Mills, C. A., and Porter, M. M. 1953. Tobacco-smoking habits in an American city. J. Natl. Cancer Inst. 13, 1283-1297.
Moolgavkar, S. H., Luebeck, E. G., Krewski, D., and Zielinski, J. M. 1993. Radon, cigarette smoke, and lung cancer: A re-analysis of the Colorado Plateau uranium miners' data. Epidemiol. 4, 204-217.
National Research Council (NRC). 1988. Health Risks of Radon and Other Internally Deposited Alpha-Emitters - BEIR IV. NRC Committee on the Biological Effects of Ionizing Radiations (BEIR), National Academy Press, Washington, DC.
National Research Council (NRC) 1991. Comparative Dosimetry of Radon in Mines and Homes. National Academy Press, Washington, DC.
National Research Council (NRC). 1998. Health Risks of Exposure to Radon, BEIR VI (prepublication copy). NRC Committee on the Biological Effects of Ionizing Radiations (BEIR), National Academy Press, Washington, DC.
Piantadosi, S., Byar, D. P., and Green, S. B. 1988. The ecological fallacy. Am. J. Epidemiol. 127, 893-904.
Piantadosi, S. 1994. Invited commentary: Ecologic biases. Am. J. Epidemiol. 139, 761-764.
Price, P. N. 1997. Predictions and maps of county mean indoor radon concentrations in the mid-Atlantic states. Health Phys. 72, 893-906.
Price, P. N., Revzan, K., and Nero, A. V. 1998. Estimated County Mean Radon Concentrations in the United States. LBNL-41115. Lawrence Berkeley National Laboratory, Berkeley, CA.
Puskin, J. S. 1992. An analysis of the uncertainties in estimates of radon-induced lung cancer. Risk Anal. 12, 277-285.
Raju, M.R., Eisen, Y., Carpenter, S., Jarrett, K., and Harvey, W.F. 1993. Radiobiology of a particles. IV. Cell inactivation by a particles of energies 0.43.5 MeV. Radiat. Res. 133, 289-296.
Simmons, J.A., Cohn, P., and Min, T. 1996. Survival and yields of chromosome aberrations in hamster and human lung cells irradiated by alpha particles. Radiat. Res. 145, 174-180.
U.S. Bureau of the Census (USBC). 1953. County and City Data
Book 1952. U.S. Department of Commerce. U.S.
Government Printing Office, Washington, DC.