RSH Data & Documents

"Low Level
Radiation Health
Effects: Compiling
the Data"

Revision 2
March 30, 1999
by Radiation, Science, and
Health, Inc.,
Edited by J. Muckerheide

1.2.6.3.1
Ecological Studies

"Counties in the U.S. with high lung cancer rates should have higher average 222Rn levels than counties with low lung cancer rates, assuming the average 222Rn level in a county is not correlated with other factors that cause lung cancer. The magnitude of this effect was calculated, using the absolute risk model, the relative risk model, and an intermediate model, for females who died in 1950-1969. The results were similar for all three models. We concluded that, ignoring migration, the average Rn level in the highest lung cancer counties should be about three times higher than in the lowest lung cancer counties according to the theory. Preliminary data indicate that the situation is quite the opposite: The average Rn level in the highest lung cancer counties was only about one-half that in the lowest lung cancer counties."

"A straightforward approach to testing the linear, non-threshold theory for radiation-induced lung cancer would be to study lung cancer rates in areas where Rn levels are very high and very low and compare them with the predictions of the theory. This was done in a few cases (Cohen 1987), but the approach was limited because we did not know the Rn levels in most areas. An alternative approach is to turn the problem around and study the Rn levels in areas with very high and very low lung cancer; the above difficulty is then eliminated since lung cancer rates are readily available for all U.S. counties. However, with this approach, the predictions of the theory are not as obvious. It is the purpose of this paper to derive those predictions and make the alternative approach useful for testing the theory."

"As the cohort to be studied, we chose white females who died during the period 1950-1969. The advantage of this choice was that this was the most recent cohort, which lung cancer rates were not heavily influenced by cigarette smoking; this made effects of Rn relatively more important."

"…the 253 counties with white female population between 50,000 and 300,000 constitute a single distribution… The reason for limiting consideration to counties with populations over 50,000 was because statistical uncertainties were large for lower populations."

"The average Rn levels, r, for various counties constituted a log-normal distribution centered about the geometric mean, r0. ..the width (i.e., standard deviation) of this log-normal distribution is ln q; that is, 68% of all counties were assumed to have an average Rn level between qr0 and r0q-1."

"…we noted that 6.5% of 253 counties (16 counties) had lung cancer mortality rates above 8 x 10-5 y-1, and 25 counties had rates below 5 x 10-5 y-1. By measuring the Rn level in 100 randomly selected houses in a county, the average Rn level in that county can be estimated with an uncertainty of about 10% (Cohen 1986). When this is done for 16 counties, the grand average should be uncertain only by about … 2.5%. For 25 counties, this is 2.0%. Thus, the ratio of average Rn levels in the 16 high lung cancer counties to those in the 25 low lung cancer counties should be determined with an uncertainty of … 3.2%. It was, therefore, evident that the ratio predicted by the linear, non-threshold theory, estimated … as in the range 1.66-4.7, can be tested with very high sensitivity if the value of q is accurately known."

Table 2

"The data from Table 4 were seasonally adjusted and averaged to derive a single point for the average annual Rn level in each county; these points were plotted in Fig. 4 along with the least-squares fit of the data to a straight line. (The line appears curved because the ordinate scale is logarithmic.) The expected results from Table 2 for the absolute and relative risk models with q = 1.76 are shown by dashed lines. The discrepancy between measured and expected Rn levels is evident. Each of the 18 points for low lung cancer rates is above the expected value, while each of the 26 points for high lung cancer rates is below the expected value. The lines indicating the expected trend of Rn levels vs. lung cancer rate are essentially perpendicular to the line showing the least-squares fit to the data. If these results are confirmed, it would seem that the theory is unsupportable."

Table 4, p.1     Table 4, p.2

Figure 4