Health Risks of Ionizing Radiation: - Clark University

0.1-3.5 WLM; this is roughly equivalent to a
range of 1-20 mSv lung dose according to the
conversions suggested by UNSCEAR (2000).
Uncertainty and judgement. At low doses, with
low associated risks, it is difficult for epidemiology
to detect an excess incidence of disease. Background
variations in the rate of a disease, caused by variations
in demographic characteristics, unknown risk
factors, and the stochastic nature of cancer, create
situations where it is often impossible to say with
any statistical certainty that an observed outcome is
attributable to a particular exposure. This problem is
exacerbated by small study populations. Land (1980)
presents a good discussion of statistical power.
Studies with low expected risks tend to have low
power because the size of a cohort needed to detect
the risk is unrealistic. In these cases we introduce a
bias when we only consider risk estimates that are
significantly greater than zero (Land 1980).
As an example of this problem we might
consider atomic bomb survivors who were exposed
to radiation in utero. In this group there were two
cases of childhood cancer; based on the background
cancer rate less than one case was expected. This is
not a statistically significant excess and one might
say something like “atomic bomb survivors exposed
in utero did not demonstrate an increase in childhood
cancer”. On the other hand, the excess relative risk
estimate based on these two cases could be as high as
44 per Gy 13 . Another example is multiple myeloma
incidence among veterans of the 1957 nuclear test
“Smoky”; although one case of the disease was
expected, none were observed. This outcome might
have occurred by chance even if the true relative
risk was as high 2.8 14 . The most truthful assessment
of data such as these is that they are insufficient to
tell us anything specific and they are consistent with
a wide range of possibilities.
Synthesis of information. Although many
individual studies of low doses are inconclusive by
themselves they become more meaningful when
they are considered together with other information.
With a set of uncertain information in hand we can
Discussion 171
attempt to describe our understanding in terms that
are meaningful if not precise.
As an illustration we can consider the leukemia
risk of adult exposures to low doses of radiation.
We have seen convincing evidence that there is a
leukemia risk in children exposed to low doses of
radiation in utero, from Chernobyl, or from the
Nevada Test Site (above and in the leukemia section).
There are physiological reasons why adult risks
might be different, including different rates of blood
production; we can see evidence of this difference
in the fact that childhood leukemia is often the acute
lymphocytic type and adult leukemia is often the
chronic myeloid type. It is therefore worthwhile to
examine adult risks independently, keeping in mind
what we know about childhood risks.
The best sets of data on adult exposures
come from nuclear workers and the atomic bomb
survivors, and we might also consider veterans who
participated in nuclear weapons testing. As noted
above, Cardis et al. (1995) found a significant dose
response for non-CLL leukemia mortality among
workers in three countries. This dose-response
estimate included doses over 0.4 Sv and so it is
not, by itself, evidence of a risk at low doses. The
authors note, however, that although the slope is not
significant at lower doses it is compatible with the
estimate for the full cohort (unfortunately this data
is not shown in the report). In a study of Canadian
workers, over 98% of whom received doses less
than 0.1 Sv, a similar estimate of the leukemia risk
(for incidence) was derived (Sont et al. 2001), and a
compatible mortality risk estimate was also derived
(Ashmore et al. 1998). Gilbert (2001) compares
these estimates to the male atomic bomb survivors
who were exposed as adults and an estimate from
the National Registry of Radiation Workers (UK) 15 .
Table 13-2 is based on a table presented by Gilbert
(2001). Although three of the estimates presented
in Table 13-2 are not significant there is notable
consistency; the estimate of Cardis et al. (1995) is
stronger in this context. The vast majority of these
workers were exposed to low doses.
In addition to these dose-response estimates we
13 The ERR estimate for childhood cancer incidence following in utero exposure to the atomic bombs was reported to
be 11/Gy with a 95% confidence interval of –1 to 44 (Wakeford and Little 2003).
14 The RR for multiple myeloma incidence was 0.0 (0.0-2.8; Caldwell et al. 1983).
15 There is some overlap between the UK study and the three-country study.