[ RadSafe ] Can anyone help with this mystery?

[farbersa at optonline.net]

Dr. Cohen,
I've not seen the paper. However, a basic question. Is there any difference in the distribution, mean,  or median number of years worked by workers in each interval?  For example, if the lower exposure intervals involved persons irradiated over a shorter number of work years, accumulating lower exposure on average over that time, the number who would die of cancer at their  younger ages would be expected to be lower, given the lower fraction of all deaths due to cancer at lower average age.

Thus the findings could in this case be an artifact of the age distribution and exposure per year of these workers.  Does the paper have any information on these questions about age distribution by interval?

Stewart Farber
=======================

----- Original Message -----
From: Bernard Cohen <blc+ at pitt.edu>
Date: Monday, January 23, 2006 1:45 pm
Subject: [ RadSafe ] Can anyone help with this mystery?

> A paper by W.N. Sont et al in Am. J. Epidemiol. 154:309-318:2001 
> on 
> radiation monitored Canadian workers in industrial, medical, 
> dental, and 
> nuclear power jobs gives the following percentages of people dying 
> from 
> cancer vs lifetime accumulated radiation dose (as I crudely 
> calculate 
> from the data they present):
> 
>          Dose        % who died      95% confid.             
> 
>          in rem      from cancer      interval     
> 
>            0.25             1.8%             1.8 - 1.8       
> 
>             0.75            2.3%             1.9 - 2.7      
> 
>             1.5              2.8%             2.4 - 3.2        
> 
>             3.5              2.5%             2.1 - 2.9        
> 
>              7.5             3.9%             3.2 - 4.6        
> 
>              15              3.8%             2.9 - 4.7        
> 
>              30              4.8%             3.3 - 6.3        
> 
>            >40              6.8%             3.4 - 4.2        
> 
>          On the face of it, these data give very strong evidence 
> in 
> favor of a linear-no threshold dose response relationship 
> extending well 
> below 1.0 rem.
> 
>          However, by the time a person dies, he receives a dose 
> averaging about 20 rem from non-occupational exposure, and these 
> exposures vary widely, typically between about 10 rem and 30 rem, 
> in a 
> manner not correlated with occupational exposures. Roughly 
> speaking, 
> that means that the numbers for dose in the first column above 
> should be 
> increased by about 20 rem and assigned an uncertaincy of about 10 rem.
> 
>          Crudely, this converts the above table to:
> 
>         
> 
>          Dose        % who died      95% confid.             
> 
>          in rem      from cancer      interval     
> 
> 
> 
>     10.25-30.25      1.8%             1.8 - 1.8     
>                                                                   
> 10.75     -30.75      2.3%             1.9 - 2.7      
> 
>      11.5-31.5         2.8%             2.4 - 3.2        
> 
>      13.5-33.5         2.5%             2.1 - 2.9        
> 
>      17.5-37.5         3.9%             3.2 - 4.6        
> 
>         25-45             3.8%             2.9 - 4.7        
> 
>          40-60            4.8%             3.3 - 6.3        
> 
>            >50              6.8%             3.4 - 4.2            
> 
> 
> 
>          No one could claim that this table gives any info on low 
> level 
> radiation in the dose range below 20 rem. But this raises another 
> question: why was the first table so deceiving?
> 
>            Can anyone offer an explanation for this?
> _______________________________________________