Something's Fishy in the Shipyard

The nuclear worker groups had a lower death rate from all causes, leukemia, and LHC than the non-nuclear workers. - Professor Emeritus Myron Pollycove, M.D., University of California at San Francisco, Medicine and Radiology

Suppose you're not convinced about the concept of radiation hormesis and want to do a statistical analysis to settle the matter in your own mind. What is your concept of a "convincing" study? How would you design an experiment so that you would have no doubt about the trustworthiness of the results? Some safeguards I'd like to see would be:

1. The study would have to be supported by deep pockets, because there would be a lot of expense in collecting and analyzing the mountains of data involved.

2. I would want those in charge of the actual research (as opposed to those who are paying for it) to be scientists from a reputable institution.

3. There must be a very large number of exposed persons and the same order of unexposed controls, with both chosen randomly from the same employment pool in order to make the study statistically meaningful and to avoid any possible "healthy worker effect."

4. The doses to the individuals would have to be as accurately known as possible, at least up to industrial or military standards.

5. The study would have to look at not only cancer but also total mortality, in order to test the hypothesis that radiation hormesis not only might reduce cancer, but also might lessen the effects of infectious diseases and other immune system breakdowns.

6. Finally, I would want the researchers to believe that they were attempting to measure a positive correlation between radiation and disease, without even suspecting that any hormesis effect was of interest.

The following investigation at Johns Hopkins meets all my criteria.

United States

Sources of significant background radiation in the United States are (1) terrestrial sources, such as granite and certain other types of rocks; (2) radon and its progeny from the decay of thorium, uranium and other radionuclides; and (3) cosmic radiation - which doubles each 6,000 feet in altitude in the temperate latitudes. Several of the Rocky Mountain states, particularly Idaho, Colorado and New Mexico, have higher than normal levels of each of these categories, combined to make a significant difference between these states and others, especially the Gulf Coast states of Louisiana, Mississippi and Alabama.

We'll take a look at the cancer rates in these areas (American Cancer Society 1998 data) and compare them with background sources. The Linear No-Threshold (LNT) theory would predict an increase in cancer; the hormesis model forecasts a decrease in cancer - and other diseases or conditions affected by immune competence - as the radiation levels increase in the hormetic range. You be the judge.

Jagger investigated the average background exposures and cancer death rates among the 5.84 million people living in Idaho, Colorado and New Mexico, compared with the same factors for the 10.83 million residents of Louisiana, Mississippi and Alabama. His results are shown in Figure 25. While the study does not examine the large number of confounding factors that could possibly influence the data, it does illustrate a trend diametrically opposite to the LNT and is strongly indicative of hormesis. [Jagger, H. Natural background radiation and cancer death in Rocky Mountain states and Gulf Coast states. Health Physics, 75(4), 1998.]

Source for Figure 25: Background Radiation vs. Cancer Rate: Jagger, H. Natural background radiation and cancer death in Rocky Mountain states and Gulf Coast states. Health Physics, 75(4), 1998. Cancer data from the American Cancer Society, 1998.

If you are unaccustomed to reading graphical data, please note that Figures 25 and 26 show two different parameters - radiation dose and cancer rate - for two different geographical areas. The scale on the left side of the graph relates to the bar graphs, while the right-hand values pertain to the cancer deaths per 100,000 persons, as shown by the data points and connecting trend line. What is intended to be shown is the increase in cancer rate (as evidenced by the upward sloping line) compared with the decrease in background radiation indicated by the magnitude of the bar graphs.

A 1994 study by Cohen compares the average radon level and lung cancer rate in the Rocky Mountain states with that in the Gulf Coast states. Radon data come from state agencies, the EPA, and University of Pittsburgh researchers; cancer data are from the American Cancer Society. [Cohen, B. Dose-response relationship for radiation carcinogenesis in the low-dose region. Int. Arch. Occupational Environmental Health, 66, 1994.]

Source for Figure 26: Residential Radon vs. Lung Cancer Rate: Cohen, B. Dose-response relationship for radiation carcinogenesis in the low-dose region. Int. Arch. Occupational Environmental Health, 66, 1994.

Were the data, plotted in Figure 26, to show that lung cancer increased with increasing radon levels, one would have to concede as very likely that the higher residential radon levels were a cause of cancer. Since the evidence shows the exact opposite, one might expect our regulatory agencies to take note and consider revising their policies accordingly. Unfortunately, they apparently don't think they should be bothered with such trivial matters as evidence. "It is the radiation protector's task to protect people from radiation, regardless of whether the radiation has bionegative or biopositive effects."

Craig and Seideman studied the rate of leukemia and lymphocytic lymphoma versus altitude in the United States. [Craig, L. and Seidman, H. Leukemia and lymphoma mortality in relation to cosmic radiation. Blood, 17, 1961.] This, of course, should be a "no brainer" - everyone knows that leukemia is caused by radiation. Since there is about a 4,000 foot difference between the low data points and the high point - and thus a near doubling in cosmic radiation - we will no doubt find, in Figure 27, a doubling of radiation-sensitive cancers like leukemia, right?

Source for Figure 27: Leukemia and Lymphocytic Lymphoma vs. Altitude (U.S.): Craig, L, and Seidman, H. Leukemia and lymphoma mortality in relation to cosmic radiation. Blood, 17, 1961.

Oops. Something is obviously wrong here. I guess it's back to the old drawing board again for the LNTers. Really, this does go on and on. Allow me to mention a few of the more interesting cases - without the plots, since I suspect you're starting to tire of graphs and charts.

Leukemia Mortality Among Survivors

Leukemia is a family of cancer involving the white blood cells. With the exception of lymphocytic leukemia - which is often erroneously included - the disease can be induced by ionizing radiation, and hence is the model of a radiation-engendered disorder. One would therefore expect a sizable increase in leukemia as the exposure level increases from background level of 0.1 cGy as shown in Figure 15. The data - taken from M. Delpha's "Fear of nuclear power could be met with data from Hiroshima" [Delpha, M. Nuclear Europe, 42, 3 1989] - indicate that the leukemia mortality rate shows a minimum at 3.5 cGy, or about ten times the average annual U.S. background level. Only a single data point gives and indication of hormesis; however, a threshold is positively demonstrated, and a clear difference in the effect of low- and high-level radiation is evident - both in conflict with expectations of the LNT.

Source of Figure 15: Leukemia Mortality Among A-Bomb Survivors: Delpha, M. Fear of nuclear power could be met statistically with data from Hiroshima. Nuclear Europe, 42, 3, 1989.

Effects of Radiation on Cancer - Leukemia Mortality

Of the myriad varieties of cancer, leukemia is most often considered to be associated with exposure to ionizing radiation, so we'll look at it, first, in an experiment involving 1,000 young adult mice per group (about 12,000 mice in all), which were exposed to a single dose of gamma radiation from 20 to 600 cGy at the rate of 300 cGy (300 rad) per minute. (Ouch.) This experiment was directed by J.R. Maisin and reported in Radiation Research, 113, 300, 1988 (see Figure 7). To realize just how far apart the Linear No-Threshold Theory and the hormesis model are from one another, the LNT predicts a 60% increase in leukemia at an exposure of 200 cGy, while the actual data show a 35% decrease. One can argue all day the beauty of the LNT and how it is a terrific standard for regulatory control; but these data show that it just isn't true when compared to experiment.

Caption for Figure 7: Leukemia Mortality in Mice: Source: Maisin, J.R., Wambersie, A., Gerber, G.B., Mattelin, G., Lambert-Collier, M., and Guelette, J., Life shortening and disease incidence in C57BL mice after single and fractionated gamma and high energy neutron exposure. Radiation Research, 113, 300, 1988.

A Slippery Slope

When we read the statistics on deaths involving automobile accidents, we are given the actual count of deaths as compiled by various law enforcement agencies. But when the anti-nuclear zealots tell us about the number of people who died as a result of radiation from exposure to, say, radon, they don't have a single victim they can point to with any degree of certainty. Their "statistical deaths" come from an extrapolation based on the Linear No-Threshold (LNT) theory. Just as with our falling analogy, they correctly note that very high exposures, like falling from very tall buildings, increase the likelihood of death (by cancer, in the case of radiation). Their argument falls apart when they try to extend, or extrapolate, the high-dose exposure to much lower exposures.

For a moment let's jump to an example detailed in a later chapter. Studies of the Japanese indicated that exposure to the equivalent of 100 SXR units (20 rem, if you're ahead of me) in a short time would double the number of leukemia deaths in a population of one million people, from the expected fifty deaths to one-hundred. The LNT extrapolation would predict one-tenth the increase in deaths (in this case, five) if the population were exposed to one-tenth that additional exposure (in this case, 10 SXR units).

Could they point to any bodies? No, they only have their theoretical corpses based on the LNT extrapolation. But in this case, there is actual data that completely contradict the LNT theory's prediction. Not only did the death rate not increase; it actually decreased - by an astounding 40%! To summarize:
Fifty deaths expected in unexposed population
Fifty-five deaths predicted by LNT extrapolation
Only thirty deaths occurred, according to actual data

* * *

Therein lies the crux of the hormesis/LNT controversy: Those who advocate the Linear No-Threshold theory base their belief on the extrapolation of high-level exposure responses down to low levels. But when low-level data are available, they almost always show a bio-positive - or stimulatory - response. It is this response, called hormesis, that we will be discussing in the next chapter.