[ RadSafe ] New Chernobyl Study Predicts up to 60, 000 Excess Cancer Deaths

[Ivor Surveyor]

I would be interested to know what precision should be placed on the 
numerous and contradictory estimated projected future deaths from the 
Chernobyl incident.

It is clear that the basic calculation is a triple product of mean 
exposure dose to a population, the size of the population, and 
presumably nominal risk coefficients based on ICRP.

However, current estimates make claims which usually vary from 4,000 
to over 60,000.  However, on to day's radio national program, OCKAMS 
RAZOR an even greater number of deaths has been claimed.

<http://www.abc.net.au/rn/science/ockham/stories/s1613742.htm>

Dr John Coulter, ex Australian Senator and President of Sustainable 
Population Australia quoted the figure of 950,000.  His stated 
authority is Gofman J.

I wonder how many would agree with me that these figures are 
meaningless, unless authors make explicit their methodology and 
address the following points.

    * What actual numbers are used for the basic triple 
product?  Together with errors in their assumed "risk 
coefficient."   Is the distribution of exposure in a specific 
population assumed to be symmetrically distributed or skew 
distributed around the mean dose?
    * Where the additional exposure to a population from the 
Chernobyl incident is low, then theoretically should not other 
sources of exposure such as background variation, medical etc be 
factored in?   Or is this really too difficult?
    *  What allowance, if any has been made to transfer "risk 
coefficient" assumed based on the atomic bomb experience to another 
population separated both in time and in geography. Populations with 
different baseline health, socio-economic factors and so on.
    * Am I correct in making the following statement:   The brake 
down of health services, infectious illness and nutrition of a 
decimated post war population must have been orders of magnitude 
worse that Russia in 1986?
    * What allowance if any is made for yearly variations in cancer incidence.
    * In 1945 it is very likely that for the majority of cancers, 
incidence rates were very close to mortality rates.   Because of 
advances in medical knowledge this is no longer true.  I suppose it 
would be very difficult if not impossible to factor this into estimates?
    * In making estimates of projected deaths, authors ought to make 
clear the model that they are using.  Models are important to 
standardise for age, sex, of different populations.  I do not write 
as either statistician or epidemiologist. However it seems to me that 
one approach would be modify the probability of death column in a 
life table by a radiation risk factor. Another approach may be to use 
a multivariable approach (? Cox's model).   I am just interested to 
know how these estimates are fashioned.
    * Is the "control population" counterfactual?  In the sense that 
the estimated cancer incidence after the Chernobyl exposure is being 
compared with what it would have been if there had been no Chernobyl exposure.
    * The estimates of radiation deaths are almost invariably quoted 
with out stating confidence limits.  So one can have little idea of 
either the accuracy or precision of the estimates.
    * I will make a final a pragmatic point.  If what ever estimate 
is made it is stated that the number of deaths is too low to be 
detected or empirically verified by epidemiological surveys. Do these 
estimates have any real practical purpose?


Ivor Surveyor, MD (Brist), FRACP, FRCP
Emeritus Consultant Physician, Nuclear Medicine,
[isurveyor at vianet.net.au]