From S. Jay Smith, Andrew Chen, Samuel Caudill, John Reidy, (Lowell E. Sever), Centers for Disease control and Prevention Atlanta, Georgia, US (Centers for Public Health Research and Evaluation, Seattle, Washington)
We read with interest your article on radiation safety limits (This Week,
2 December 1995, p 10), which quotes Maurice Tubiana of the French Academy of
Sciences as saying: “Studies [to determine what levels are safe] can be carried
out in a few years.”
The Centers for Disease Control and Prevention is evaluating workers and
communities exposed to elevated levels of manmade radiation. Another way to
study radiation health effects is to investigate large cohorts of individuals
exposed to differing levels of natural background radiation (which increases
with altitude).
Pochin (Health Physics, vol 31, p 148) has estimated that 10 000
people exposed to an increase in annual radiation dose of 5 millisieverts would
need to be observed for 56 years to detect even a change in mortality risk as
large as 7 per cent. But Pochin’s example has limited statistical power. To
increase this power, it would require a sample size of 100 000 and 35 years of
observation to detect such a change in risk. To detect a change in risk as small
as 1 per cent, six million persons would need to be observed for 30 years.
Thus, studying small relative risk changes associated with low levels of
radiation exposure will require a very large cohort exposed to background
radiation as great as 5 millisieverts per year. Finding such cohorts will be
difficult because most humans are not naturally exposed to these levels.
Worldwide, only about two million people live in six cities above 10 000 feet
with background exposures of 3 to 4 millisieverts per year.
Advertisement
Large sample size
increases may not be sufficient for evaluating small changes in risk, however,
since confounding factors and misclassification become more problematic.
With current study designs, Tubiana’s above statement may be valid for
detecting large changes in risk, but not smaller changes, such as 5 per cent or
less. In future, alternative study designs or better methods for measuring
exposures and outcomes could possibly reduce the resources and time required to
detect such small differences.
