The World Health Organization recently released a risk assessment for the Fukushima radiation releases that began in March 2011. Their estimates are based on population dose estimates provided in an earlier report and risks of cancer at those radiation levels derived from the Life Span Study (LSS) of survivors of the atomic bombs dropped on Hiroshima and Nagasaki. The dose assessment is incomplete as there are a number of components that the report ignores. For example, the committee chose not to assess the doses within twenty kilometers of the nuclear plant, occupational radiation exposure, doses from beta radiation emitted by gaseous releases, and in utero doses.
Radiation risk assessments are largely based on the Life Span Study (LSS) of survivors of the U.S. nuclear attacks on Hiroshima and Nagasaki. Follow-up of LSS participants did not begin until five or more years after the bombing, and many people did not survive long enough to be included. This introduced several sources of potential bias into radiation risk estimates. If mortality from the immediate effects of the bombings was related to longer-term risks, then the most radiosensitive people died before the study began, meaning that the LSS selected for healthier people with lower radiation risks than the people who were excluded. Furthermore, the LSS’s monitoring of cancer incidence (new diagnoses of cancer rather than death) did not begin until 1958. Therefore, all cancers that occurred within thirteen years of exposure are omitted from LSS cancer incidence risk estimates. These caveats are routinely omitted when risk estimates from the LSS are applied to other populations.
The LSS has focused on the penetrating gamma and neutron radiation from the nuclear detonations, prompt radiation that was gone in seconds. However, fallout led people to be exposed to radioactive particles. Because the primary sites of fallout were not at ground zero, people most affected by fallout, also known as “black rain,” were exposed to lower doses from prompt radiation. This would tend to raise disease rates among the LSS survivors with the lowest prompt radiation doses and create a downward bias in LSS radiation risk estimates. Despite the signing of a treaty banning atmospheric nuclear weapons testing in 1963, in large part due to concerns about health effects of fallout at great distances from aboveground tests, the impacts of fallout in the LSS have not been factored into radiation risk estimates.
A Radiation Effects Research Foundation (RERF) report on black rain was published in December 2012. The report summarizes analyses of survey responses of survivors who were asked about their exposures to fallout. Out of the 86,671 survivors in the primary analyses, approximately 12,000 said they had been exposed to black rain. For over 21,000, however, there is no information on exposure to black rain. This missing data is a major gap in the Life Span Study that has been ignored for half a century.
RERF reported on the mortality rates in the LSS between 1950 and 2003, and between 1962 and 2003, comparing people who said they had been exposed to black rain, those who said they had not, and those for whom exposure was not determined. The groups reporting exposure to black rain and no exposure to black rain had similar mortality in both time periods, but Hiroshima survivors for whom black rain exposure was unknown experienced 27 percent higher mortality, and Nagasaki survivors for whom black rain exposure was unknown experienced 46 percent higher mortality between 1950 and 2003. Excess mortality of people whose exposure to black rain was unknown occurred primarily during the period between 1950 and 1962.
In addition to fallout, another type of residual radiation that exposed the A-bomb survivors came from neutron activation near ground zero. This caused elevated levels of gamma radiation during the early hours and days after the detonations. People closest to ground zero, especially those with little shielding, did not survive to be included in the LSS, and residual radiation exposures from neutron activation was not important for them. However, many people who were farther from the hypocenter at the time of the bombings and who were not badly injured by heat, blast, and radiation moved through areas near ground zero shortly after the detonations, some of them looking for their relatives. As with black rain, to the extent that survivors with lower prompt gamma and neutron doses had more exposure to residual radiation than those with higher doses, there is a downward bias in LSS risk estimates.
All survivors were entered into follow-up on October 1, 1950, even though not all survivors completed sufficient interviews to be assigned a dose until 1965. This creates a phenomenon that epidemiologists call “immortal person time,” which inflates the denominator of the disease rates for the proximal survivors, resulting in an underestimate of their disease rates and an underestimate of radiation risks. More important, survivors within three kilometers of ground zero who had insufficient information about location and shielding for calculating their radiation doses were excluded from analyses that produce radiation risk estimates, whereas none of the survivors who were more than three kilometers from ground zero were excluded because of the lack of such information. The excluded survivors had higher mortality from cancer and leukemia, especially during early years of follow-up, compared to survivors who were included. Because exclusions were made only for proximal survivors and not for distal survivors whose doses were lower, there is a downward bias in LSS radiation risk estimates. Existing statistical techniques for reducing bias from dose-related exclusions have not been used to correct LSS risk estimates.
The LSS is restricted to survivors exposed postnatally. A separate study has been conducted of A-bomb survivors exposed in utero, however due to their small numbers, this study has not been of much use for developing dose–response relationships for risk assessment. Perhaps in part for this reason, the Fukushima risk estimates of the World Health Organization did not include disease that would result from in utero exposure. However, ever since the 1950s, when Alice Stewart first showed that obstetric X-rays cause childhood cancer, it has been well known that the embryo and fetus are especially sensitive to low-dose radiation. Although the prenatal population exposed to Fukushima Daiichi emissions during the early time period following the meltdowns was small compared to those exposed postnatally, they are especially sensitive and should be included in risk assessments.
As I noted at the outset, risk assessment uses figures drawn from studies such as the LSS and dose estimates for populations to estimate radiation casualties in populations whose disease incidence has not been quantified. Biases in radiation effects such as those in the LSS affect risk estimation. In contrast, epidemiologic studies such as the LSS directly estimate relationships between radiation and disease; however, they can only evaluate effects after they have occurred.
I started working on radiation epidemiology in 1988 when I was assigned to lead a study of the mortality of workers employed at one of the first nuclear weapons plants, the Oak Ridge National Laboratory in eastern Tennessee. The workers’ radiation doses had been monitored from very early on with individual dosimeters. I was told that we would not find any radiation effects in this population because the doses were too low, so my first clash with the dominant wisdom in this field was when we did see dose–response relationships: the higher the readings on the badges, the higher the cancer death rates of the workers. I had been told this was impossible, and yet there it was.
The disaster at Chornobyl was followed by similarly dismissive rhetoric regarding risk estimates. In a 1991 document, five years after the explosion of the Chornobyl reactor, the International Atomic Energy Agency stated: “On the basis of the doses estimated by the project teams and currently accepted radiation risk estimates, future increases over the natural incidence of cancers or hereditary effects would be difficult to discern even with large and well-designed long term epidemiological studies.” Since that time a large number of epidemiologic studies have documented increases in cancer from Chornobyl’s radiation releases. As in the Oak Ridge study, this prediction was made based on a risk assessment that used assumptions from the LSS.