Radioactive Aerosols - Aerosols: Measurement, Dosimetry, and Health Effects - page 411

Criterion (1) suggests a willingness to tolerate 15% in mortality.
From Table D-12, we can see that from the highest number of deaths (980) in the China study to
the minimum by established criteria (40), the fractional statistical error in mortality σ M will change
only from 3% to 15%. And even when the number of deaths is in the range 20-30, σ M will be in the
range of 20%.
In contrast to the criteria that mortality must be known with an accuracy of about 15%, the cri-
teria on dose (or exposure) are much more lenient.
Assessment of radon progeny concentration in mines is subject to an uncertainty of about 650%.
Furthermore, in most of the studies, exposure was reconstructed from work histories and historical
radon measurements rather than being directly measured.
Taking into account that the contribution to the dose from radon itself is negligible in comparison
with its progeny and that, according to Lubin (1994), the range of the shift of equilibrium in mines
varies from 0.2 to 0.9, we should assume that uncertainty in exposure assessment is in hundreds of
percent.
It should also be pointed out that personal dosimetry even in terms of so-called personal dosim-
eters, that is, devices that measure concentration in the breathing zone of miners, except in a French
study with a very small number of deaths, were not provided.
From this point of view, data on WLM for all studies presented in Table D-12 look completely
unrealistic (accuracy up to tenths of WLM instead of at least 50%).
Because the statistical error in mortality is already much smaller than the uncertainty in expo-
sure, it makes no sense to decrease δ M by including it in the risk assessment data from the early years
when the uncertainty in the exposure δ E was very high.
One way of improving the accuracy of dose (exposure) assessment is by using the weighted aver-
age exposure where the weights are inversely proportional to the square of the variance

1
( ) 2

W ∼

δ

If these data are excluded, the number of deaths will decrease, but the statistical error in mortality
will still be much lower than the error in dose or exposure assessment.
There is a trade-off: by choosing the largest cohort and correspondingly the greatest number of
deaths, to try to increase the statistical accuracy of mortality, we include cases with great dosimetric
(exposure) uncertainty, which results in decreasing the reliability of risk assessment.
A professor of epidemiology said at a conference, “Unfortunately, not so many people died from
this epidemic.” Paradoxically, in the case of miner studies, by improving the mortality statistic we
can make the risk assessment less reliable.
In Lubin (1994), a summary of the strengths and weaknesses of the various studies is presented.
Among the strengths, the authors mostly mentioned the large cohort and the long follow-up of the
studies. Among the weaknesses, limited exposure data, no or limited smoking data, and in six stud-
ies limited numbers of lung cancer were cited.
In the study of nonuranium miners in Tadjikistan, the number of deaths was even more limited:
34 (14 miners and 20 men from mining settlements). We can say, following the professor of epi-
demiology, “Unfortunately, only 14 miners died from lung cancer in our study.” In this case, the
fractional statistical error in mortality will be in the range of 40% for miners.
On the other hand, the assessment of dosimetric factors for all 11 year of the Tadjikistan study
was much better than in most of the studies:

• All instruments for radon and radon progeny measurements were properly calibrated with

the accuracy of PAEC measurement better than 25%-30%.

• Radon decay product concentration measurements were provided in different working sites

and for different groups of workers.