A hypothesis to explain childhood cancers near nuclear power plants

Highlights

• Over 60 studies worldwide on increased cancers near nuclear power plants (NPPs).

• German government KiKK study provides very strong evidence.

• Hypothesis proposes cancers arise in pregnant women near NPPs.

• Nuclide spikes during refuelling could result in increased exposures.

• Explanation offered for discrepancy between small dose estimates and large risks.

Abstract

Over 60 epidemiological studies world-wide have examined cancer incidences in children near nuclear power plants (NPPs): most of them indicate leukemia increases. These include the 2008 KiKK study commissioned by the German Government which found relative risks (RR) of 1.6 in total cancers and 2.2 in leukemias among infants living within 5 km of all German NPPs. The KiKK study has retriggered the debate as to the cause(s) of these increased cancers. A suggested hypothesis is that the increased cancers arise from radiation exposures to pregnant women near NPPs. However any theory has to account for the >10,000 fold discrepancy between official dose estimates from NPP emissions and observed increased risks. An explanation may be that doses from spikes in NPP radionuclide emissions are significantly larger than those estimated by official models which are diluted through the use of annual averages. In addition, risks to embryos/fetuses are greater than those to adults and haematopoietic tissues appear more radiosensitive in embryos/fetuses than in newborn babies. The product of possible increased doses and possible increased risks per dose may provide an explanation.

Introduction

In the early 1950s, Folley et al. (1952) observed an increased risk of leukemia among Japanese bomb survivors. In the late 1950s, Stewart et al. (1958) also observed that radiation exposures can result in increased incidences of leukemia. A number of studies since then (BEIR and Committee on the Biological Effects of Ionizing Radiations, Board on Radiation Effects Research, Commission on Life Sciences, National Research Council, 1990, Preston et al., 1994, IARC, 1999) have shown that ionising radiation including medical, occupational and environmental exposures, are a risk factor for leukemia. In addition, older ecological and case–control studies (Forman et al., 1987, Gardner, 1991, Pobel and Viel, 1997) revealed an association between nuclear power plants and childhood leukemia among those living nearby.

In the late 1980s and early 1990s, increased incidences of childhood leukemias were reported near several UK nuclear facilities. Various explanations were offered for these increases, however the UK Government's Committee on the Medical Aspects of Radiation in the Environment (COMARE) concluded in a series of reports (COMARE. Committee on the Medical Aspects of Radiation in the Environment, 1986, COMARE. Committee on the Medical Aspects of Radiation in the Environment, 1988, COMARE. Committee on the Medical Aspects of Radiation in the Environment, 1989, COMARE. Committee on the Medical Aspects of Radiation in the Environment, 1996) that the cause remained unknown but was unlikely to involve radiation exposures. This was mainly because official estimates for radiation doses from these facilities were too low by orders of magnitude to explain the increased leukemias. Indeed, any theory will have to account for the >10,000 fold discrepancy between official dose estimates from NPP emissions and observed increased risks.

A pattern of epidemiological evidence world-wide now clearly indicates increased leukemia risks near nuclear power plants (NPPs). Laurier and Bard (1999) and Laurier et al. (2008) examined the literature on childhood leukemias near NPPs world-wide. These two studies identified a total of over 60 studies. An independent review of these studies (Fairlie and Körblein, 2010) indicated that the large majority of these studies revealed small increases in childhood leukemia although in many cases these were not statistically significant. Laurier and Bard and Laurier et al., mostly employees of the French Government's Institut de Radioprotection et Sûreté Nucléaire (IRSN), confirmed that clusters of childhood leukemia cases existed near most NPPs but refrained from drawing wider conclusions. Fairlie and Körblein (2010) in their review concluded that the copious evidence indicating increased leukemia rates near nuclear facilities, specifically in young children, was quite convincing.

This conclusion was supported by two meta-analyses of national multi-site studies. Baker and Hoel (2007) assessed data from 17 research studies covering 136 nuclear sites in the UK, Canada, France, the US, Germany, Japan, and Spain. In children up to nine years old, leukemia death rates were from 5 to 24% higher and leukemia incidence rates were 14–21% higher. However their analysis was criticised by Spix and Blettner (2009).

The second meta-analysis by Körblein (2009) covering NPPs in Germany, France, and the UK also found a statistically significant increased risk of child leukemias and relative risk of leukemia deaths near NPPs (RR = 1.33; one-tailed p value = 0.0246). Further studies (Guizard et al., 2001, Hoffmann et al., 2007) indicated raised leukemia incidences in France and Germany. However COMARE. Committee on the Medical Aspects of Radiation in the Environment, 2005, COMARE. Committee on the Medical Aspects of Radiation in the Environment, 2006) declined to support these conclusions.

Later, Bithell et al. (2008) and Laurier et al. (2008) found increases in child leukemias near UK and French NPPs respectively. In both cases, the numbers were low and not statistically significant – i.e. there was a greater than 5% possibility that the observations could have occurred by chance. However instead of reporting these increases, the studies incorrectly concluded that there was “no evidence” (Bithell) and “no suggestion” (Laurier) of leukemia increases near UK and French nuclear reactors, merely because their data lacked statistical significance. These conclusions were incorrect: the authors should have reported the observed leukemia increases but added there was a >5% probability they could have occurred by chance.

In more detail, p values (that is, the probabilities that observed effects may be due to chance) are affected by both the magnitude of effect and the size of study (Whitley and Ball, 2002). This means statistical tests must be used with caution as the use of an arbitrary cut-off for statistical significance (usually p = 5%) can lead to incorrectly accepting the null hypothesis (ie nil effect) merely because it is not statistically significant (Sterne and Smith, 2001): a possible type II error. This often occurs in small studies due to their small sample sizes rather than lack of effect (Everett et al., 1998). Axelson (2004) has pointed out that many epidemiology studies with negative results – statistically speaking, are of questionable validity as they may obscure existing risks.